ROMANCE LINGUISTICS 2009

AMSTERDAM STUDIES IN THE THEORY AND

HISTORY OF LINGUISTIC SCIENCE
General Editor
E.F.K. KOERNER
Zentrum für Allgemeine Sprachwissenschaft, Typologie
und Universalienforschung, Berlin
efk.koerner@rz.hu-berlin.de

Series IV – CURRENT ISSUES IN LINGUISTIC THEORY

Advisory Editorial Board

Lyle Campbell (Salt Lake City)

Sheila Embleton (Toronto)
Elly van Gelderen (Tempe, Ariz.)
John E. Joseph (Edinburgh)
Manfred Krifka (Berlin)
Martin Maiden (Oxford)
Martha Ratliff (Detroit, Mich.)
E. Wyn Roberts (Vancouver, B.C.)
Joseph C. Salmons (Madison, Wis.)

Volume 315

Sonia Colina, Antxon Olarrea and Ana Maria Carvalho (eds.)

Romance Linguistics 2009. Selected papers from the 39th Linguistic Symposium
on Romance Languages (LSRL), Tucson, Arizona, March 2009
ROMANCE LINGUISTICS 2009
SELECTED PAPERS FROM
THE 39TH LINGUISTIC SYMPOSIUM
ON ROMANCE LANGUAGES (LSRL),
TUCSON, ARIZONA, MARCH 2009

Edited by

SONIA COLINA
ANTXON OLARREA
ANA MARIA CARVALHO
University of Arizona

JOHN BENJAMINS PUBLISHING COMPANY

AMSTERDAM/PHILADELPHIA
 TM
The paper used in this publication meets the minimum requirements of
8

American National Standard for Information Sciences – Permanence of

Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data

Linguistic Symposium on Romance Languages (39th : 2009 : Tucson, Arizona)

Romance linguistics 2009 : selected papers from the 39th linguistic symposium on ro-
mance languages (LSRL), Tucson, Arizona, March 2009 / Sonia Colina, Antxon
Olarrea and Ana Maria Carvalho (eds.).
p. cm. (Amsterdam studies in the theory and history of linguistic science. Series IV, Cur-
rent Issues in Linguistic Theory, issn 0304-0763 ; v. 315)
Includes bibliographical references and index.
1. Romance languages--Congresses. I. Colina, Sonia. II. Olarrea, Antxon. III. Carvalho,
Ana Maria. IV. Title.
PC11.L53â•…â•… 2010
440--dc22 2010032379
isbn 978 90 272 4833 6 (Hb ; alk. paper)
isbn 978 90 272 8764 9 (Eb)

© 2010 – John Benjamins B.V.

No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any
other means, without written permission from the publisher.
John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands
John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
 Table of contents

Foreword & acknowledgments ix

List of contributors xi
Editors’ introduction 1

part i. Phonetics/Phonology
Correcting the record on Dominican [s]-hypercorrection 15
Barbara E. Bullock & Almeida Jacqueline Toribio
V-to-V assimilation in trisyllabic words in French:
Evidence for gradience and locality 25
Christopher Carignan & Zsuzsanna Fagyal
The production and provenance of palatal nasals in Portuguese
and Spanish 43
Ryan Shosted & José Ignacio Hualde
Lenition and phonemic contrast in Majorcan Catalan 63
José Ignacio Hualde, Marianna Nadeu & Miquel Simonet
Alveolar laterals in Majorcan Spanish: Effects of contact with Catalan? 81
Miquel Simonet
Units of speech production in Italian 95
Irene Vogel & Linda Wheeldon
Pitch polarity in Palenquero: A possible locus of H tone 111
John M. Lipski
Word-minimality and sound change in Hispano-Romance 129
Fernando Martıʹnez-Gil
Multiple opacity in Eastern Regional French 153
Jean-Pierre Montreuil

part ii. Syntax

Syntactic variation in Colombian Spanish:
The case of the Focalizing Ser (FS) structure 169
Dunia Catalina Méndez Vallejo
ïš© Romance Linguistics 2009

Anaphoricity, logophoricity and intensification:

The puzzling case of son propre in French 187
Isabelle Charnavel
More on the clitic combination puzzle: Evidence from Spanish,
Catalan and Romanian 203
Anahí Alba de la Fuente
The Spanish dative alternation revisited 217
John Beavers & Chiyo Nishida
Romanian genderless pronouns and parasitic gaps 231
Ion Giurgea
To agree or not to agree: Beyond quintessentially syntactic
agreement in Spanish 249
Julio Villa-García
Variation in subject expression in Western Romance 267
Ana de Prada Pérez
A phase-based analysis of Old French genitive constructions 285
Deborah Arteaga & Julia Herschensohn
V2 loss in Old French and Old Occitan: The role of fronted clauses 301
Barbara Vance, Bryan Donaldson & B. Devan Steiner

part iii. Morphology and interfaces

The loss and survival of inflectional morphology:
Contextual vs. inherent inflection in creoles 323
Ana R. Luís
Allomorphy in pre-clitic imperatives in Formenteran Catalan:
An output-based analysis 337
Eulàlia Bonet & Francesc Torres-Tamarit
Preverbal vowels in wh-questions and declarative sentences
in Northern Italian Piacentine dialects 353
Anna Cardinaletti & Lori Repetti
Pitch accent, focus, and the interpretation of non-wh
exclamatives in French 369
Heather Burnett
Detours along the perfect path 387
Patrícia Amaral & Chad Howe
 Table of contents ïš©ïš©

Grammaticalization of commencer/cominciare
“to begin” in French and Italian 405
Jason Brazeal
Index of subjects, terms and languages 423
 Foreword & acknowledgements

The thirty-ninth annual Linguistic Symposium on Romance Languages (LSRL) was

held for the first time at the University of Arizona 27–29 March 2009. The by-now
traditional parasession was devoted to Variation and Change in Romance. Romance
linguistics has significantly evolved since the first LSRL took place in Florida in 1971,
and it is undeniable that the LSRL has played a role in forging an identity for the field
in North America, if not internationally. We trust that this LSRL has served to con-
tinue that tradition by bringing together researchers, and facilitating the sharing of
ideas. It is our hope that the present volume will serve the purpose of disseminating
the best work presented at the conference among those who were unable to attend and,
of course, among current and future generations of Romance linguists in general.
Of a total of 124 submissions, 58 were selected for presentation, along with three
plenary talks. It was an honor that Ricardo Bermúdez-Otero, Gillian Sankoff, and
Karen Zagona accepted our invitation to be the plenary speakers. This volume con-
tains a selection of 24 peer-reviewed articles presented at the conference. We express
our most sincere appreciation to the many anonymous reviewers who lent their time
and expertise to assist us with the selection of conference presentations and of the
papers included here.
Numerous people were involved in many different forms in the organization of
the conference. We are extremely grateful for the tireless efforts and dedication of the
conference graduate assistant, Miriam Díaz. We would also like to thank all the graduate
students who donated their work and volunteered to host conference participants, as
well as Mary Portillo and the staff of the Department of Spanish and Portuguese, for
their help and support, and Mark Bryant for his technical assistance with PASHA,
the online abstract submission application. Our appreciation goes to the Department
of Spanish and Portuguese, in particular, to Malcolm Compitello, Department Head,
and to the College of Humanities for their support, financial and otherwise. The fol-
lowing organizations and units provided much needed financial support in very dif-
ficult economic times: Program for Cultural Cooperation between Spain’s Ministry of
Culture and United States Universities, Associate Vice President for Research, College
of Humanities, Department of Spanish and Portuguese, Department of Linguistics,
Department of Anthropology, Department of French, Center for Latin American
Studies, Program in Cognitive Science, Program in Second Language Acquisition and
Teaching, Center for Educational Resources in Culture, Language, and Literacy. Finally,
we must thank series editor Professor E.F.K. Koerner and his advisers as well as the
 Romance Linguistics 2009

team at John Benjamins, in particular Anke de Looper, for their advice and assistance
with this project. We are grateful to both the series editor and the publisher for their
respective roles in the long-standing relation between the Current Issues in Linguistic
Theory (CILT) series and the volumes resulting from the annual LSRL meetings.

Tucson, Arizona Sonia Colina

June 2010 N.N.
 List of contributors
Ryan Shosted
Department of Linguistics
University of Illinois at
Urbana-Champaign
4080 Foreign Language Building
707 S. Mathews Avenue
Urbana, IL 61801
USA
rshosted@illinois.edu
José Ignacio Hualde
Department of Spanish,
Italian and Portuguese
University of Illinois at
Urbana-Champaign
School of Literatures,
Cultures and Linguistics
4080 Foreign Languages Building
MC-176
707 S. Mathews Avenue
Urbana, IL 61801
USA
jihualde@illinois.edu
Miquel Simonet
Department of Spanish and Portuguese
University of Arizona
Modern Languages 545
P.O. Box 210067
Tucson, AZ 85721
USA
simonet@email.arizona.edu
Barbara E. Bullock
Department of French and Italian
The University of Texas at Austin
1 University Station B7600
Austin, Texas 78712
USA
bbullock@austin.utexas.edu
toribio@austin.utexas.edu
Chris Carignan & Zsuzsanna Fagyal
Department of French
University of Illinois at
Urbana-Champaign
2090 FLB 707 S. Mathews Avenue
Urbana, IL 61801 USA
zsfagyal@illinois.edu
Irene Vogel
Department of Linguistics
University of Delaware
Newark, DE 19716
USA
ivogel@udel.edu
Linda Wheeldon
University of Birmingham
United Kingdom
l.r.wheeldon@bham.ac.uk
 Romance Linguistics 2009
John Lipski
Department of Spanish, Italian
and Portuguese
237 Burrowes Building
The Pennsylvania State University
University Park, PA 16802
USA
jlipski@psu.edu
Fernando Martínez-Gil
Department of Spanish
and Portuguese
The Ohio State University
298 Hagerty Hall
1775 College Road
Columbus, OH 43210–1340
USA
martinez-gil.1@osu.edu
Jean-Pierre Montreuil
Department of French and Italian
The University of Texas at Austin
1 University Station B7600 ↜
Austin, Texas 78712
USA
jpmontreuil@mail.utexas.edu
Dunia Catalina Méndez Vallejo
Department of Foreign Languages
West Virginia University
Chitwood Hall
P.O. Box 6298
Morgantown, WV 26506–6298
USA
catalina.mendez-vallejo@mail.wvu.edu
Isabelle Charnavel
University of California,
Los Angeles/Institut Jean
Nicod-Ecole Normale
Supérieure, Paris
Department of Linguistics
3125 Campbell Hall
UCLA
Los Angeles, CA 90095–1543
USA
icharnavel@ucla.edu
Anahí Alba de la Fuente
Department of Modern Languages
and Literatures
University of Ottawa
70 Laurier East Unit #134
Ottawa, Ontario K1N 6N5
Canada
aalba080@uottawa.ca
Chiyo Nishida
Department of Spanish
and Portuguese
The University of Texas at Austin
1 University Station B3700
Austin, Texas 78712–0192
USA
cnishida@mail.utexas.edu
John Beavers
Department of Linguistics
The University of Texas at Austin
1 University Station, B5100
Austin, TX 78712–0198
USA
jbeavers@mail.utexas.edu

Ion Giurgea
Dept. of Linguistics
FB Sprachwissenschaft
Universität Konstanz
Fach D 185
78457 Konstanz
Germany
giurgeaion@yahoo.com
Julio Villa-García
337 Mansfield Road
U-box 1145
Department of Linguistics
Arjona Building
University of Connecticut—Storrs
Storrs, CT 06269–1145, USA
julio.villa@uconn.edu
Ana de Prada Pérez
170 Dauer Hall
P.O. Box 117405
Gainesville, FL 32611–7405
USA
aprada@ufl.edu
Deborah Arteaga
Department of Foreign Languages
University of Nevada
Las Vegas€89154–5047
USA
deborah.arteaga@unlv.edu
Julia Herschensohn
Department of Linguistics
University of Washington
Box 354340
Seattle, WA 98195–4340 USA
herschen@u.washington.edu
List of contributors 
Barbara Vance
Department of French
Indiana University
Ballantine Hall 617
1020 E. Kirkwood Ave
Bloomington, IN 47405–7103
USA
bvance@indiana.edu
Heather Burnett
Linguistics Department
3125 Campbell Hall
University of California, Los Angeles
Los Angeles, California, USA, 90024
hburnett@ucla.edu
Ana R. Luís
Rua do Peneireiro 10
3040–716 Coimbra
Portugal
aluis@fl.uc.pt
Patrícia M. Amaral
SOCLAS, University of Liverpool
Cypress Building
Chatham Street
Liverpool, UK
L69 7ZR
pamaral@liverpool.ac.uk
Chad Howe
Department of Romance Languages
University of Georgia
Gilbert Hall
Athens, GA 30602–1815
USA
chowe@uga.edu
 Romance Linguistics 2009

Jason Brazeal Anna Cardinaletti

Department of French and Italian
The University of Texas at Austin
1 University Station B7600 ↜
Austin, Texas 78712
USA
brazeal@mail.utexas.edu
Dipartimento di Scienze del linguaggio
Università Ca’ Foscari di Venezia
Ca’ Bembo, Dordoduro 1075
30123 Venezia
Italy
cardin@unive.it

Eulàlia Bonet & Lori Repetti

Francesc Torres-Tamarit
Filologia Catalana, Edifici B
Universitat Autònoma de Barcelona
08193 Bellaterra
Spain
Eulalia.Bonet@uab.cat
Department of Linguistics
SUNY
Stony Brook, NY 11794–4376
USA
lrepetti@notes.cc.sunysb.edu
 Editors’ introduction

The purpose of this introduction is two-fold: first, to situate this volume within the
field of Romance linguistics and, second, to examine the contributions included within
the context of the field and as they relate to others in the volume.

1.â•… Romance linguistics

Romance linguistics has evolved significantly from its 19th-century origins in philo�
logical studies and its focus on historical phonology and morphology to where it
currently stands. It has been observed that Romance linguistics today follows the
developments of linguistic theory (Posner 1998; Dworkin 2000; Klausenburger 2006),
attracting not only scholars who specialize in Romance languages, but also linguists
who may do occasional work in Romance. Posner notes “the emphasis is on ‘linguis-
tics’ rather than ‘Romance’” and that “there is comparatively little that can serve to
demarcate Romance linguistics as a separate branch of study” (1998:€ 352). A brief
examination of recent LSRL volumes reveals that certainly the emphasis tends to be
on linguistic theory. Yet, we would like to argue that, despite the changes, there exists
a common identity to Romance linguistics, even if this identity is not quite the same
as that of the 19th century.
First, one must realize that the changes, significant as they may appear, are not
exclusive to Romance linguistics, as linguistic theory has also seen an evolution from
its origins in philology to the generative theories of the present. Klausenburger makes
a similar point in his 2006 introductory retrospective of the field “just as the ‘giants of
the field in the 19th century successfully applied the dominant paradigm of (historical)
linguistics of their time […] today’s practitioners […] avail themselves of current lin-
guistic advances to achieve equally significant results” (2006:€14).
Second, despite the less prominent position occupied by philological and histori-
cal studies, Romance linguistics today still integrates up-to-date linguistic research
with a focus on the Romance languages. Most Romance linguists see themselves as
scholars who are cognizant of linguistic theory, but who also believe that the in-depth
study of a language or language family has much to offer to the theory. This compara-
tive tradition is a crucial and defining feature of the field in an era that finally seems to
realize the significance of cross-language research. Recent developments in linguistic
theory have arguably played a role in this change of outlook: for instance, within the
 Romance Linguistics 2009

framework of Optimality Theory (OT), synchronic and diachronic divisions among

languages consist only of differences in the language-specific rankings of universal
constraints; linguistic separation is therefore only a matter of factorial typology (e.g.
the number of possible rankings). In this context, demonstrating that the theory can
account for cross-linguistic variation, be it synchronic or diachronic, through constraint
re-ranking becomes crucial, encouraging cross-dialectal and historical comparisons.
The value of cross-linguistic research exemplified by work in the Romance languages
is cited as the main reason for this focus in a current volume on laboratory phonology:

This focus on Romance languages is motivated by our belief that there is a need for
multilanguage data to test current theoretical claims and models, which often lack
precisely this sort of broad-cross-linguistic basis… The virtue of cross-linguistic
research is that it constitutes a valuable tool to explore similarities and differences
between languages and thus allows us to construct general linguistic theories,
while at the same time ensuring that the peculiarities of individual languages can
be characterized within the theory. (Prieto, Mascaró & Solé 2007: vii)

As we show below, the papers included here demonstrate the vitality and the defining
nature of the cross-linguistic aspect of Romance linguistics.
Another factor that has contributed to the increased relevance of cross-linguistic
research and that defines Romance linguistics today is the growth of experimental and
quantitative studies. The volume just cited is a good indication of this (Prieto, Mascaró€&
Solé 2007). The same conclusion can be reached by looking at recent LSRL volumes,
including the present one. As Prieto et al. argue, experimental studies and Romance
linguistics stand in a mutually beneficial relationship based on the need of the former
for multilanguage data and on the cross-linguistic interest of the latter, facilitated by
experimental techniques.
Before concluding this brief overview of Romance linguistics today, it is impor-
tant to observe that historical work is still very much a part of Romance linguistics,
although its presence may not be as obvious as it was over a century ago. As one will
notice by examining the table of contents of recent LSRL volumes, diachronic research
in Romance linguistics generally appears not as a separate subfield, but integrated with
other areas of linguistic analysis. The data under analysis may be diachronic, and it
may capture variation in time or describe older stages of various Romance languages,
yet the contributions are often included under syntax, phonology, or morphology. This
volume is a good illustration of this. Despite a parasession on Language Variation and
Change and sufficient contributions based on diachronic data and/or older stages of
the language, the need for a cohesive table of contents, as well as the nature of the
contributions, required that we place these papers in one of three sections: phonetics/
phonology, syntax, or morphology/interfaces, rather than a separate section on Varia-
tion and Change. We will comment on the historical contributions in Section€2.
 Editors’ introduction 

2.â•… Contributions to the present volume

The papers included here are organized in three main sections: phonetics/phonology,
syntax, and interfaces, including morphology. Within each section, they are grouped
in clusters according to common threads, such as subject matter and framework.

2.1â•… Phonetics/Phonology
All but two of the papers (those by Montreuil and Martínez-Gil) in this section are
experimental. The experimental papers reveal the relevance of this type of work for
the field of Romance linguistics: some serve to correct wrong assumptions about the
data (cluster one); others use the data to attain a better understanding of cross-language
and cross-dialectal issues, including diachronic evolution of Romance varieties; finally,
the third cluster investigates prosodic phenomena.
In the first cluster, two articles, Bullock & Toribio and Carignan & Fagyal present
empirical evidence that challenges previously held assumptions regarding two well-
known and much discussed processes, [s]-hypercorrection in Dominican Spanish
and Vowel Harmony in French. Theoretical linguistic treatments of Dominican
[s]-hypercorrection assume that lexical forms in popular Dominican Spanish no longer
contain any trace of coda or final /s/ and, thus, speakers randomly insert them into
syllable- and word-final position. Yet, on the basis of data from naturalistic recorded
interviews, Bullock & Toribio argue that [s]-hypercorrection is inaccurately described
as a rule that targets the syllable coda and is instead much more complex – phonetically
motivated in some contexts, and in others, potentially deployed to mark an intonation
boundary. Dominican [s]-hypercorrection is therefore not a random process and most
likely it is not a feature of lost-coda speakers, as speakers in the study who deleted
/s/ categorically also showed no occurrences of hypercorrect [s]. Carignan & Fagyal’s
contribution evidences the need to consider experimental data in determining the
phonetic or phonological nature of specific processes. Most descriptions of Vowel
Harmony (VH) in French have been based on disyllabic tokens. Under the hypothesis
that VH could be a local, gradient phenomenon due to the articulatory influence of
the final stressed vowel on the preceding vowel, rather than a phonological word-level
processs, Carignan & Fagyal investigate the temporal extent of VH in trisyllabic words.
Acoustic measurements taken at the midpoint of target vowels reveal that word-medial
vowels are significantly more fronted and less open when followed by a word-final /i/
than a word-final /a/, while word-initial vowels do not show this effect. This suggests
that VH in French might not be a word-level phenomenon, but rather the local coar-
ticulatory effect of the final vowel on the vowel immediately adjacent to it. Results also
show an increase of the degree but not the temporal extent of assimilatory influence as
the trigger becomes longer.
 Romance Linguistics 2009

In the second cluster of experimental papers, two papers relate experimental data
with historical explanations and cross-linguistic evidence: “The production and
provenance of palatal nasals in Portuguese and Spanish” by Shosted & Hualde, and
“Lenition and phonemic contrast in Majorcan Catalan” by Hualde, Nadeu & Simonet.
A third paper dealing with cross-linguistic/cross-dialectal data is “Alveolar later-
als in Majorcan Spanish” by Miquel Simonet. Shosted & Hualde present acoustic
evidence bearing on the articulation of the palatal nasal consonants in Portuguese
and Spanish. Historically, one class of Portuguese palatal nasals arose from a nasalized
palatal approximant – the phonetic leftovers of a historical process (typical of Galician-
Portuguese but not Spanish) that nasalized vowels and deleted nasal consonants. Some
descriptions of Brazilian Portuguese suggest that this nasalized approximant is still
extant; in order to investigate this possibility, the authors compare the articulation of
palatal nasals in Portuguese and Spanish in a variety of vocalic contexts to determine
the degree of occlusion. The findings suggest that that the nasalized approximant of
medieval Galician-Portuguese may have been retained in Brazilian Portuguese, thus
reflecting a conservative trait passed down from Galician-Portuguese. The matter
of the Brazilian Portuguese palatal nasal relates to the larger question of diachronic
merger and incomplete neutralizations, and to what extent synchronic variation may
persist in categories of sounds even after these categories have collapsed into a single
phonological unit.
Hualde, Nadeu & Simonet examine the /b/ and /v/ contrast in Majorcan Catalan.
It has been reported that in Catalan dialects with phonemic /v/, the phoneme /b/
tends to be realized as a stop in the same contexts where dialects without the contrast
(i.e. Central Catalan) exhibit approximant realizations (Rafel 1976:€62). This situation
has been explained in the literature as contrast preservation (Recasens 1991; Wheeler
2005), with weakening of /b/ being prevented in order to maintain its contrast with
/v/. However, the authors of this paper argue that no special preservation-of-contrast
constraint is involved in the phonetic facts, which follow from different dialects being
at different stages in the Western Romance historical weakening of intervocalic oral
stops. The experimental part of the paper seeks to confirm that /b/ is more constricted
in Majorcan Catalan than in Central Catalan. The results show that in Majorcan Catalan
intervocalic /b/ is realized with significantly greater constriction than in Central Cata-
lan, and /b/ is also more occluded than /d/.
Simonet analyzes the effects of contact with Catalan on the production of alveolar
laterals in the Spanish spoken in Majorca. Traditionally, Majorcan Catalan alveolar
laterals have been described as velarized (dark), while (Castilian) Spanish laterals have
been described as non-velarized (clear) (Recasens, 1987, 2004; Recasens & Espinosa,
2005). The existing sociolinguistic research has stated that dark laterals are a receding
feature while clear ones are diffusing across groups of speakers, including the Catalan-
dominant Catalan-Spanish bilinguals residing in Majorca (Pieras, 1999). Simonet’s
 Editors’ introduction 

results indicate that Spanish-dominant bilinguals produce clear lateral allophones

while Catalan-dominant bilinguals differ as a function of gender, with young females
using clear allophones and males using dark laterals. These findings suggest that
although dark allophones might be receding in favor of clear ones, they are not doing
so in as fast a pace as indicated in previous studies.
While the previous papers all treated segmental process (with the exception of
vowel harmony that can be considered suprasegmental), the third cluster of experi-
mental papers focuses on prosody, dealing with units of speech production (Vogel &
Wheeldon) and intonation (Lipski). Vogel and Wheeldon investigate the nature of the
units of speech production in Italian. The findings for Italian are essentially those pre-
viously reported for Dutch, despite substantial prosodic differences between Romance
and Germanic languages. It is shown that in order to account for these findings, a
prosodic unit between the Phonological Word and the Phonological Phrase is needed.
Given that there are problems with taking the Phonological Word as the unit of speech
encoding in Italian, as in Dutch, the authors argue for an alternative that appears to be
valid for both languages, the Composite Group.
Lipski’s “Pitch Polarity in Palenquero: A possible locus of H tone” is a second
experimental paper that deals with the phonology of contact varieties (Simonet exami�
ned Spanish in contact with Catalan). Palenquero, an Afro-Iberian creole language
of Colombia, exhibits intonational patterns that differ notably from the surrounding
regional dialects of Spanish. Authors like Hualde & Schwegler (2008) have observed
that “Palenquero is best analyzed as an accentual language, rather than as a language
with lexical tone” (p. 25). Lipski offers more detailed data on a particular configuration
that may signal a final vestige (or the incipient development) of a phonologically dis-
tinctive tone, directing his attention to the phrase-final negator -nú, the post-nominal
possessive clitics -mí “my” and -sí “your”, and the dative and accusative clitics -mí “me”
and -bó “you”. His data reveal a systematic process of tonal dissimilation between the
two tonic syllables, which in turn suggests the presence of a distinctively specified
High tone on these clitics.
The final two papers in the phonology section do not use experimental methods,
yet some of their concerns can be linked to those of the authors of the experimen-
tal contributions discussed above. Fernando Martínez-Gil also intends to correct the
record regarding the metrical system in Hispano Romance that was thought to be,
since the Hispanic Vulgar Latin stage and at least up to pre-apocope Old Spanish,
quantity-insensitive. He argues that the failure of two otherwise regular sound changes
to materialize in Hispano Romance, namely, final -m deletion, and identical vowel
simplification, provides strong support to the hypothesis of a minimal prosodic word
requirement in this branch of Ibero-Romance, and that the relevant metrical unit cor-
related with the minimal word was a bimoraic foot, not a disyllabic one, as had been
suggested in the recent literature on the topic. He proposes that the two sound changes
 Romance Linguistics 2009

failed to apply in order to avoid the reduction of a prosodic word below a minimum,
yielding a form smaller than a bimoraic foot. “Multiple Opacity in Eastern Regional
French” by Jean-Pierre Montreuil, offers an OT-CC (i.e. OT with so-called candidate
chains) account of the opaque interaction between lengthening and consonant devoic-
ing and schwa epenthesis in Eastern Regional French. It contributes to the theoretical
discussion on how to account for opacity in OT, arguing that a model like OT-CC
that reintroduces derivational evaluations in the model and performs local evaluations
rather than global ones, makes more accurate predictions than the alternatives (Stratal
OT, Harmonic Serialism or LP-OT) which fail to handle opacity. In line with the com-
parative nature of Romance linguistics, Montreuil’s article points to the desirability
of an analysis that finds new coherence in bringing together historical derivations,
synchronic chains and geographical variation. Both contributions in this last group
refer to historical data to some extent, although to a much greater extent in “Word-
minimality and Sound Change in Hispano-Romance”.

2.2â•… Syntax
As we did with the phonology papers, we group the articles in this section in clusters.
The first cluster contains two papers, by Méndez-Vallejo and Charnavelle, that address
issues in information structure. Information structure is an area of syntax that often
intersects with phonology, more specifically with prosody. This is the case with the
other two articles on information/grammatical structure, which therefore appear in
the Interfaces section (Cardinaletti & Repetti & Burnett).
The chapter “Syntactic variation in Colombian Spanish: the case of the Focalizing
Ser (FS) structure” by Dunia Catalina Méndez Vallejo, examines some of the syntactic
properties of the Focalizing ser (FS) structure in Colombian Spanish. Although the
FS shows some resemblance to the pseudo-cleft construction, the syntactic analysis
she proposes is based on the premise that the two forms are syntactically different.
Looking at the syntactic contexts in which this form may occur, Méndez Vallejo shows
(i)€that the FS may precede any kind of phrase, serving any kind of syntactic function,
and (ii), that the agreement relation between FS ser and other sentential constituents
is predictable. This strongly suggests, as claimed in her analysis, that the FS is in a
TP-internal Focus Phrase, generated below TP and above vP.
“Anaphoricity, Logophoricity and Intensification: The puzzling case of son propre
in French” by Isabelle Charnavel argues for the existence of an interaction between
binding and intensification in light of French possessive son propre (“his own”). Propre
(“own”) has two possible interpretations: its semantic effect consists of contrasting
either the possessor or the possessum with a set of contextually determined alter-
natives. Crucially, these double intensifying properties of propre correlate with the
binding properties of son propre. When son is associated with possessor propre, son
 Editors’ introduction 

propre behaves like an anaphor that can be long-distance bound if the antecedent is a
logophoric center. However, when son is associated with possessum propre, son propre
lacks both anaphoric and logophoric properties. This correlation shows that there is
an interaction between the modules of binding and intensification: it is only when the
referent of its antecedent is intensified that son propre needs to be bound.
The second syntax cluster contains five articles that focus on pronouns: clitics
and subject pronouns. Alba de la Fuente, Beavers & Nishida, and Giurgea treat clitics,
while Villa-García and de Prada Pérez focus on subject pronouns. Three of these
contributions, those by Giurgea, Villa-García, and de Prada Pérez, also address the
topic of null vs. overt distribution. It is also worth mentioning the cross-linguistic
element present in the chapters by Alba de la Fuente, Giurgea and de Prada Pérez.
Anahí Alba de la Fuente’s chapter, “More on the Clitic Combination Puzzle: Evidence
from Spanish, Catalan and Romanian”, tackles the difficult issue of the combination
of first and second person plural clitics in Romance. Its main contribution is to bring
new empirical data to the discussion, calling attention to the relevance of the feature
[number] on (im)possible combinations. Spanish, Catalan and Romanian accept, in
general, combinations of 1st and 2nd person clitics. However, not all 1st and 2nd person
combinations are possible: although two single clitics can be combined, the combina-
tion of two plural clitics results in ungrammaticality in Spanish and Romanian. On
the other hand, all three languages accept clitic combinations that include a plural
dative and a singular non-dative and systematically reject combinations that include
a singular dative and a plural non-dative. Alba de la Fuente argues that clitic restric-
tions are drawn by the degree of markedness of the dative clitic with respect to the
other clitic in the cluster and proposes a Plural Blocking Effect as a restriction that
captures those generalizations.
“The Spanish Dative Alternation Revisited” by John Beavers & Chiyo Nishida dis-
cusses an interesting set of data concerning alternating frames for ditransitive verbs in
Spanish, with or without a doubling clitic. They argue that, even though many of the
types of semantic effects observed for dative-shift alternations in English are also asso-
ciated with the clitic-doubled variant of the Spanish constructions, previous analyses
of English cannot be extended to Spanish. They propose a “Morphosyntactic Alignment
Principle”, stating that the dative alternant must have a monotonically strengthened
set of truth conditions in comparison to those of the corresponding oblique. This pro-
posal predicts both the similarities and differences between the languages.
Ion Giurgea, in “Romanian Genderless Pronouns and Parasitic Gaps”, argues that
the distribution of (overt) object clitics and null objects in Romanian can be explained
if we assume that the so-called ‘neuter pronouns’ of Romanian are genderless. Romanian
has a null object used as a bound variable with a neuter pronoun antecedent. This item
differs from parasitic gaps by the fact that it does not require an A-bar moved ante-
cedent and can only occur with neuter pronouns, while overt clitics are excluded in
 Romance Linguistics 2009

this context. Giurgea proposes that this is due to the fact that object clitics are always
marked for gender, while neuter pronouns are genderless. His analysis presents inde-
pendent evidence for the proposal that the so-called ‘neuter pronouns’ of Romanian
and other Romance languages, definite as well as indefinite and quantificational, lack
a value for Gender.
In “To Agree or Not to Agree: Beyond quintessentially syntactic agreement in
Spanish” Julio Villa-García examines the agreement relationship between the verb and
the (overt/null) subject in Non-Caribbean Spanish. In doing so, this chapter addresses
a developing area of syntactic theory concerning the interface of grammatical and per-
formance systems. Villa-García shows that, in some cases, certain acceptable sentences
contain an anomalously-agreeing subject in terms of one of the three agreement/φ-
features, namely person, number, and gender. He proposes that in Spanish there is a
condition on the output of the operation Agree, which posits that, if full/optimal agree-
ment between the verb and the subject does not obtain, at most one feature can be left
unchecked/unvalued syntactically. The remaining unvalued feature of T0 is deleted/
valued for convergence through alternative agreement mechanisms available in the
grammar, including default agreement and semantic/pragmatic agreement. From this,
it follows that agreement in Spanish is not a quintessentially syntactic phenomenon
handled exclusively by a probe-goal system like Agree.
In an important contribution to a well-studied variable, de Prada Perez analyzes
and compares patterned variation in the distribution of null and overt subjects in
Spanish and Catalan in “Variation in Subject Expression in Western Romance.” Based
on quantitative treatment of naturalistic data collected among Spanish monolinguals
in Valladolid and among Catalan-Spanish bilinguals in Minorca, the author analyzes,
ranks, and compares the weight of internal factors that condition null and overt pro-
nominal subjects in both languages. She finds that although there are no significant
differences between the languages regarding overall rates, multiple regression analyses
point to different variable and constraint hierarchies across them.
The two articles in the third and final syntax cluster focus on diachronic data,
both on Old French (Arteaga & Herschensohn and Vance, Donaldson & Steiner). In
“A Phase-based Analysis of Old French Genitive Constructions” Deborah Arteaga€&
Julia Herschensohn consider two types of genitive constructions in Old French, one in
which the possessor is introduced by a preposition and another, the so-called juxtapo-
sition genitive, in which no overt preposition is found and the possessor is invariably
in the oblique case. They propose a minimalist analysis for each construction. Their
analysis provides an account not only of the variation found in the determiner that
surfaces in the juxtaposition genitive, but also shows that this genitive is not really
different from the prepositional ones, there being no clear categorical distinctions
between genitive and prepositional possessive constructions. In “V2 Loss in Old French
 Editors’ introduction 

and Old Occitan: The role of fronted clauses” Vance, Donaldson & Steiner contribute to
the controversy regarding the analysis of Old French as a verb-second language (V2).
Contrary to Kaiser’s (2002) suggestion that Old French was not a true V2 language, the
authors show, based on quantitative analysis of thirteenth-century literary, historical,
and legal texts in French, that V2 was indeed present and beginning a gradual loss. By
analyzing data that capture the trajectory from V2 to non-V2 syntax, the authors find
that fronted subordinate clauses were an important factor favoring V2 loss. An original
data set from Old Occitan is also examined and compared with patterns found in the
evolution of V2 loss in French. “V2 Loss in Old French and Old Occitan” is one of two
quantitative studies in the syntax section (see also the chapter by de Prada Pérez).

2.3â•… Morphology and interfaces

We organize the articles in this section in three subgroups. The first one contains two
chapters on morphology (Luís and Bonet & Torres Tamarit); the second one deals with
the syntax-phonology interface, more specifically information structure, intonation
and vowel quality (Cardinaletti & Repetti and Burnett); finally, the third subgroup
analyzes diachronic data (Amaral & Howe and Brazeal).
In the morphology subgroup, we find two somewhat different papers. “The Loss
and Survival of Inflectional Morphology: Contextual vs. inherent inflection in creoles”
by Luís studies a contact variety, Korlai Creole Portuguese. This adds to the collection
of papers in this volume that study contact data (Lispki, Simonet, and de Prada Pérez).
The chapter by Bonet & Torres Tamarit, “Allomorphy in Pre-Clitic Imperatives in
Formenteran Catalan” is, on the other hand, an optimality-theoretic study of allomorphy
in Catalan. The phonology section of this volume also contains an optimality theoretic
analysis, in this case of Eastern Regional French.
Luís examines Korlai Creole Portuguese’s overt inflectional morphology in an
attempt to understand how creole inflections differ from non-creole inflections. She
finds that the use of purely formal inflections, such as theme vowels (interpreted as
instances of inherent inflection), is retained, while person-number agreement affixes
(interpreted as contextual inflection) are absent, thereby countering McWhorter’s
(2001) claim that the survival of creole affixes is determined by their semantic values.
Thus, Luís’s analysis corroborates the conclusions of Kihm (2003), Arends et al. (2006),
and Plag (2008) that creole languages are more likely to retain inherent morphology
than contextual morphology.
Bonet & Torres-Tamarit offer an output-based analysis of the extension patterns
exhibited by Formenteran Catalan in imperatives in enclisis. It is argued that the
appearance of the extension is caused by a phonological constraint requiring a right-
aligned moraic trochee. The form of the imperatives with the extension coincides with
 Romance Linguistics 2009

the inflectional stem that is found in other forms of the imperative paradigm. Lexical
Conservatism and Uniform Exponence play a key role in the analysis.
In the subgroup dealing with the phonology/syntax interface, Cardinaletti and
Repetti study the nature of the preverbal schwa in wh-questions and declarative
sentences in Northern Piacentine dialects of Italian. They show that these preverbal
vowels are two different syntactic entities, an interrogative vowel and a “subject field”
vowel whose realization depends on phonological considerations (the quality of the
following segment) as well as on the presence of other clitics (both types of vowels are
incompatible with other preverbal clitics). The second article on the phonology/syntax
interface is Heather Burnett’s “Pitch Accent, Focus, and the Interpretation of Non-WH
Exclamatives in French”. She examines the cross-linguistic realization of the class of
exclamatives in the Romance languages and makes the novel claim that exclamative
operators are not uniquely licensed by wh-morphology, but also by prosody (con-
trastive focus) on gradation quantifiers. Exclamations with focused gradation quan-
tifiers display many of the same semantic properties that have been observed with
wh-exclamatives. Using Quebec French data, she proposes that the semantic content
of this type of exclamative is a gradation construction with an implicit threshold clause
and focus on the quantifier: the exclamative operator binds the threshold yielding an
extreme degree reading, and the presence of this operator is licensed by the focus
structure of the sentence.
The final cluster in the Interfaces section addresses morphosyntactic topics from a
historical perspective: the diachrony of the present perfect in Portuguese and the gram-
maticalization of commencer/cominciare in French and Italian. Based on diachronic
data drawn from 16th- and 19th-century Portuguese corpuses, Amaral & Howe analyze
the process that led to the contemporary use of present perfect in Portuguese (pretérito
perfeito composto) to express iterative or durative actions extending to the moment of
speech. The paper questions the appropriateness of the prevailing model of the dia-
chrony of the perfect in Romance (first proposed by Harris in 1982) to explain the evo-
lution of this form in Portuguese, first, by pinpointing semantic differences between
Portuguese and Mexican Spanish perfects (lumped together in Harris’s continuum)
and, second, by revealing semantic details in the development of Portuguese not pre-
viously accounted for. In “Grammaticalization of Commencer/Cominciare ‘To Begin’ in
French and Italian” Brazeal examines aspectualizers. The author explains commencer/
cominciare as grammaticalized forms that developed from full verbs in Latin. Through
a cogent diachronic explanation based on grammaticalization theory, Brazeal reveals
these verbs’ current properties as having both lexical and functional categories, and
contributes to solving the ambiguous grammatical status of aspectualizers as both full
verbs and auxiliaries. He suggests, instead, that these verbs have undergone deseman-
ticization and decategorization and currently occupy intermediate positions in the
continuum between full verbs and auxiliaries.
 Editors’ introduction 

3.â•… Conclusion

The contributions in this volume point to a cohesiveness in Romance linguistics that

lies in the integration of up-to-date linguistic research with a comparative tradition and
the in-depth study of a language family. Much of Romance linguistics today, including
the papers in this collection, rests on the belief that multilingual, cross-linguistic
research, both synchronic and diachronic, has the ability not only to test but also to
advance current theoretical claims and models. We trust that these studies will con-
tinue the LSRL tradition of simultaneously defining and contributing to the advance
of the field of Romance Linguistics.

References

Arends, Jacques, Josje Verhagen, Eva H. Van Lier, Suzanne Dikker & Hugo Cardoso. 2006. “On
the Presence Versus Absence of Morphological Marking in Four Romance-based Creoles”.
The Structure of Creole Words: Segmental, syllabic and morphological aspects ed. by Parth
Bhatt & Ingo Plag, 223–241. Tübingen: Max Niemeyer.
Dworkin, Steven N. 2000. “Introduction”. New Aproaches to Old Problems: Issues in Romance
historical linguistics ed. by Steven N. Dworkin & Dieter Wanner, ix–xiv. Amsterdam &
Philadelphia: John Benjamins.
Harris, Martin. 1982. “The ‘past simple’ and ‘present perfect’ in Romance”. Studies in the Romance
Verb ed. by Martin Harris & Nigel Vincent, 42–70. London: Croom Helm.
Hualde, José Ignacio & Armin Schwegler. 2008. “Intonation in Palenquero”. Journal of Pidgin
and Creole Languages 23.1–31.
Kaiser, Georg A. 2002. Verbstellung und Verbstellungswandel in den romansichen Sprachen.
Tübingen: Max Niemeyer.
Kihm, Alain. 2003. “Inflectional Categories in Creole Languages”. Phonology and Morphology of
Creole Languages ed. by Ingo Plag, 333–63. Tübingen: Max Niemeyer.
Klausenburger, Jurgen. 2006. “Introduction: From Romance philology to (historical) Romance
linguistics?” Historical Romance Linguistics: Retrospective and perspectives [Selected
papers from the 34th Linguistic Symposium on Romance Languages (LSRL), Salt Lake City,
March 2004] ed. by Randall S. Gess & Deborah Arteaga, 1–15. Amsterdam & Philadelphia:
John Benjamins.
McWhorter, John H. 2001. “The World’s Simplest Grammars are Creole Grammars”. Linguistic
Typology 5.125–166.
Pieras, Francesc. 1999. Social Dynamics of Language Contact in Palma de Mallorca: Attitude and
Phonological transfer. Ph.D. thesis, Pennsylvania State University, College Park, Pa.
Plag, Ingo. 2008. “Creoles as Interlanguages: Inflectional morphology”. Journal of Pidgin and
Creole Languages 28: 1.101–130.
Posner, Rebecca. 1998. “Romance Linguistics in the Nineties”. Romance Philology 51.326–355.
Prieto, Pilar, Joan Mascaró & Maria-Josep Solé. 2007. “Introduction”. Segmental and Prosodic
Issues in Romance Phonology ed. by Pilar Prieto, Joan Mascaró & Maria-Josep Solé, v–xv.
Amsterdam & Philadelphia: John Benjamins.
 Romance Linguistics 2009

Rafel, Joaquim. 1976. “Fonologia diacrònica catalana: Aspectes metodològics”. Problemes de llengua
i literatura catalanes: Actes del II colloqui internacional sobre el català, 47–65. Montserrat:
Publicacions de l’Abadia de Montserrat.
Recasens, Daniel. 1987. “An Acoustic Analysis of V-to-C and V-to-V Coarticulatory Effects in
Catalan and Spanish VCV Sequences”. Journal of Phonetics 15.29–312.
Recasens, Daniel. 1991. Fonètica descriptiva del català. (Assaig de caracterització de la pronúncia
del vocalisme i consonantisme del català al segle XX). Barcelona: Institut d’Estudis Catalans.
Recasens, Daniel. 2004. “Darkness in [l] as a Scalar Phonetic Property: Implications for phonology
and articulatory control”. Clinical Linguistics and Phonetics 18: 6/8.593–603.
Recasens, Daniel & Espinosa, A. 2005. “Articulatory, Positional and Coarticulatory Characteris-
tics for Clear /l/ and Dark /l/: Evidence from two Catalan dialects”. Journal of the Interna-
tional Phonetic Association 35: 1.1–25.
Wheeler, Max. 2005. The Phonology of Catalan. Oxford: Oxford University Press.
part i

Phonetics/Phonology
 Correcting the record
on Dominican [s]-hypercorrection

Barbara E. Bullock & Almeida Jacqueline Toribio

The University of Texas at Austin

Theoretical linguistic treatments of Dominican [s]-hypercorrection all assume

the hypotheses, advanced by Terrell (1986), that lexical forms in popular
Dominican Spanish no longer contain any trace of coda or final /s/ and, thus,
speakers randomly insert them into syllable- and word-final position (Terrell
1986; Núñez-Cedeño 1988, 1989, 1994; Harris 2002; Bradley 2006; Vaux
2001, 2002). We demonstrate that Terrell’s premises ensue from an inadequate
description of Dominican [s]-insertion and, as a consequence, phonological
theories based on them cannot account for the range of actual, as opposed to
hypothesized, exemplars that occur in Dominican speech. Using natural data,
we argue that [s]-hypercorrection is inaccurately described as a rule that targets
the syllable coda and is instead much more complex.

1.â•… Background

A salient and stigmatized trait of popular Dominican Spanish is the sporadic realiza-
tion of an [s] in a lexical item in which it does not occur underlyingly (e.g. priesto <
prieto “very black”). In the Dominican Republic, speakers who produce an intrusive
[s] are said to be attempting to hablar fisno (<fino) “to talk in an elevated style”. As this
variety of Spanish is distinguished from most others by the complete effacement of
coda /s/, Dominican hypercorrect [s]-insertion (DHS) is traditionally envisioned as a
classic case of rule inversion where a synchronic process of epenthesis has replaced a
diachronic rule of deletion (Vennemann 1972). As shown in the examples in (1), DHS
co-occurs with /s/ deletion in (1a) or with either /s/ deletion/aspiration as in (1b).1
(1) a. El diale[s]to ma’ boni[s]to era el de Loma de Cabrera. (F, 34, illiterate)
“The prettiest dialect was the one from Loma de Cabrera.”
b. La familia mía es de a[s]quí mi’mo […] a mí, a mí[s]
me gu[h]ta la agricultura. (M, 67, illiterate)
“My family is from right here…Me, I like agriculture.”

.╅ Our speakers are coded for gender, age, and grade/literacy (F, 31, illiterate).
 Barbara E. Bullock & Almeida Jacqueline Toribio

From these examples, it indeed appears that both processes, deletion (or aspiration)
and [s] insertion, appear to target the same environment, coda position.
Data from DHS, such as those illustrated in (1), have served as empirical sup-
port for a number of diverse theoretical claims in both derivational and constraint-
based approaches to Spanish and general phonology. Núñez-Cedeño (1988, 1994)
and Harris (1983, 2002) each invoke data from DHS in support of their hypotheses
regarding Spanish constraints on the licensing of segments in syllable positions.
More recently, Vaux (2001, 2002) and Bradley (2006) have considered the theoretical
implications of DHS in accounting for hypercorrection within non-derivational
models of phonology.2 Each of these theoretical treatments derives empirically from
the observations of Terrell (1986), who advanced the notion that popular Dominican
Spanish speakers have reconstructed the underlying forms of lexical items so that
they no longer contain any trace of pre-consonantal or final /s/. As a consequence,
Terrell contends, these speakers, driven by linguistic insecurity, freely insert [s] into
any syllable- or word-final position.
Revising Terrell’s original account, generative phonologists hold that the applica-
tion of DHS is blocked when the targeted rhyme of a syllable is already branching.
Under Harris (1983), a form such as *muersto (<muerto) “dead person” would be
inadmissible with a hypercorrect [s] since the rhyme of the penultimate syllable is
already saturated. Núñez-Cedeño (1988, 1994), in turn, argues that data from DHS
provides crucial evidence for a geminate representation for the Spanish trill. As the
two halves of a geminate are arguably heterosyllabic, syllable rhymes containing the
first “half ” of a trill are already saturated so that *casrretera (<carretera) “road”, for
example, would be impossible.3 Harris (2002) also invokes DHS as support of the nar-
rowing conditions on his three-syllable window account of Spanish stress. For Harris,
antepenultimate stress in Spanish is well formed if and only if the penultimate syllable
does not branch. Consequently, ill-formed *teléfosno (< teléfono) “telephone” is ruled
out because the intrusive [s] closes the penult, potentially precluding a proparoxytonic
stress profile.
Recent works have focused on DHS as a source of data regarding general theoretical
concerns. DHS is problematic when viewed from the perspective of constraint-based
theories since a hypercorrect surface form always violates Faithfulness, by contain-
ing a segment that was not present in the input, and Markedness, by arguably adding

.╅ Bradley (2006:24) maintains that, according to the tent of minimal constraint violation,
“just one postnuclear [s] is necessary to satisfy output-output correspondence.” This remains
to be confirmed.
.╅ Bradley (2006) argues convincingly that the trill is not a geminate tap. Under his gestural
account, DHS is ruled out before the trill for co-articulatory, not prosodic, reasons.
 Correcting the record on Dominican [s]-hypercorrection 

a syllable coda. Vaux (2001, 2002) suggests that DHS is phonologically “unnatural”
because the presence of an intrusive [s] could never be potentially confused with its
absence. He argues that the only way to account for DHS is derivationally, through a
simple rule that inserts [s] into the syllable coda. Undeterred by Vaux’s arguments,
Bradley (2006) demonstrates how the facts of DHS can be accommodated within a
non-derivational account. Bradley argues that although popular Dominican Spanish
speakers are unaware of where coda [s] properly belongs in their own lexical entries,
they are aware that it surfaces in the lexical entries of others so that they produce DHS
when attempting to align their speech forms with the input produced by more conser-
vative speakers.
It bears reiterating that all the above referenced works depart from the same set
of assumptions made current by Terrell (1986), which have remained unchallenged in
the generative phonology literature. Terrell (1986) suggests that popular Dominican
speakers have no knowledge of where a coda /s/ belongs in a lexical item and that
DHS, then, targets any available syllable coda as a reflection of a speaker’s attempt to
restore a missing [s] in a lexical entry. This suggestion is asserted directly by Harris
(2002:97): “…lost-s speakers aren’t sure where the s’s are in standard dialects, so their
‘corrections’ are essentially random, missing the target as often as not.”
However intuitively appealing it may be to the theoretician, the romanticized
notion of Dominicans as “lost-s” speakers is a testable (and falsifiable) hypothesis. So,
too, is the idea that DHS is an “essentially random” process. Yet, none of the theoretical
works overviewed above tests these assertions or questions the evidence upon which
they are based; this in spite of the fact that existing sociolinguistic accounts confront
these issues with empirical data. Dominican sociolinguists directly contest the notion
that /s/ is absent from popular speakers’ lexical forms. Instead, /s/ deletion, like DHS,
is not categorical, rather conditioned by social and stylistic factors. Lopéz-Morales
(1990) and Alba (2004) each report that the rate of /s/ deletion in careful speech
decreases significantly with respect to spontaneous speech across speakers with varying
degrees of literacy and education. This implies that not even illiterate speakers could be
classified as “lost-s” speakers since they are clearly able to maintain /s/ when circum-
stances demand that they adjust their speech to a more formal register.
The only data-driven description of DHS of which we are aware is a case study by
Morgan (1998), who recorded and transcribed the speech of a tour guide. Remarkably,
in a ten-minute stretch of discourse, the guide produced approximately 200 hyper-
correct tokens.4 The results of Morgan’s analysis revealed that DHS for this speaker
occurred word finally over 80% of the time, primarily when the following word begins

.â•… López-Morales (1990) notes that the rates of hypercorrection in his corpus of both
spontaneous and careful speech are extremely low. This is true of our data as well.
 Barbara E. Bullock & Almeida Jacqueline Toribio

with a voiceless stop. And when it occurred in word-internal position, hypercorrect

[s] also appeared nearly exclusively before a voiceless stop. The data from this single
speaker, a prodigious user of DHS, suggest that [s] insertion is not random.
Note that the sociolinguistic literature, although sparse, belies the two main
assumptions of the theoretical treatments of DHS. Here, we aim to provide addi-
tional evidence, based on extensive data from the field, that Terrell’s assumptions are
unfounded and that the widespread interpretation of DHS as a random coda inser-
tion phenomenon is equally false. Instead, we find patterns in our data that support
and expand on the findings of Morgan (1998) and we interpret these as evidence that
hypercorrect [s] has a dual origin. We propose that one form of [s] insertion has a
plausible phonetic motivation, hence, it is “natural” (contra Vaux 2001, 2002). We sur-
mise that the second form of [s] insertion may arise to mark a phrase-final boundary,
a phenomenon also noted in some forms of contemporary French (Fagyal€& Moisset
1999; Smith 2003).

2.â•… The data

The data were elicited via naturalistic interviews of approximately 45 minutes each
with 35 adults (ages 18–77; 17 men, 17 women) and in semi-structured interviews of
15 minutes with 89 children (ages 6–17; 41 girls, 48 boys). Each participant also com-
pleted a story retelling task about an old man and his rooster (El Viejo Lino y su gallo,
Quiquiriquí), adapted from a short folktale by Zevallos (1997), a methodology that we
have found to be particularly useful for eliciting similar content from participants of
diverse ages and levels of education. The second author, a native Dominican speaker,
conducted all interviews. Each interview was recorded via a head-mounted micro-
phone onto a Marantz handheld digital recorder at 44 kHz, 16 bit quantization.
We selected these participants because they are representative of the rural poor.
The Spanish of this area is stigmatized on the island because of its prevalent use of vari-
ous linguistic features that characterize popular speech, including [s] hypercorrection
(see Bullock & Toribio 2009 for an overview of other structural features). All speakers
were born and raised in the rural Cibao in small villages located within close proxim-
ity of one another where basic necessities like food, water, and electricity are in short
supply and where children and adolescents only have access to two to three hours of
schooling per week day. The majority of our participants, with the notable exception
of adolescent females, are unable to read and write with any facility. The teenage girls
professed to be motivated to learn to read and write in order to have access to jobs in
the lucrative tourist industry. The majority of the male children, by contrast, aspired
to be baseball players and many of them, like the adults interviewed, were functionally
illiterate to the point of being unable to spell their own names when asked to do so by
 Correcting the record on Dominican [s]-hypercorrection 

the interviewer. The adults were all employed as day laborers, farmers, cooks, house-
keepers or caregivers.
We transcribed each interview and then conducted an auditory analysis and,
when necessary, a visual inspection for the presence of high frequency energy in a
spectrograph. We discarded from this analysis any ambiguous tokens in which there
was only a slight percept of turbulence but no clear visual evidence of a period of
frication. In this way, we coded the data conservatively, taking account only of each
perceptually salient occurrence of hypercorrect [s] and of fully realized lexical /s/. We
report on a subset of the data here: data from 5 adults (ages 29–69, 1 male, 4 females)
and 69 children (ages 6–16, 32 females, 37 males).

2.1â•… General sociolinguistic observations

The realization of hypercorrect [s] was very sporadic. Our most prolific user of intru-
sive [s], an illiterate adult female, manifested hypercorrection in 1.3% of her lexical
entries, a number that tallies exactly with Lopéz-Morales’ (1990) findings in the care-
ful speech of his participants. Many of our speakers failed to manifest any tokens of
intrusive [s]. In general, females employed more DHS as well as more instances of
lexical /s/ maintenance than did males and, not surprisingly, DHS appeared more
widely in the speech of the illiterate or semi-literate speakers. Among the children,
we note several trends that suggest that DHS usage is conditioned by social factors.5
High school girls (those in grade 8 and above) displayed almost no intrusive [s] in
their speech and a high degree of lexical /s/ maintenance. Illiterate boys (grade 4 and
under) also showed no occurrences of DHS but also deleted coda /s/ categorically.
This particular group of male children, then, produced no pre-consonantal or final
[s]. It was the semi-literate boys and girls (grades 4–8) who produced all the tokens of
DHS among the children. Significantly, they also produced tokens of correct lexical /s/,
unlike the illiterate males.
Although these observations invite a more in-depth quantitative analysis, taken
as a whole they demonstrate that the use of DHS is, in fact, categorically absent from
the speech of those participants, illiterate boys, who could be considered true “lost-s”
speakers. These facts directly contradict the thesis articulated by Harris (quoted above)
that intrusive [s] is a compensation strategy among “lost-s” speakers. We also suggest
that the sociolinguistic trends point to literacy, rather than ambient input, as a poten-
tially heavily-weighted factor in conditioning DHS since all the children are exposed
to correct lexical /s/ input as modeled for them by the literate, adolescent females.

.╅ We group the children by school grade level rather than by age because children who fail
to learn to read are repeatedly held back at grade 4 or under in spite of their age. This means
that it is not uncommon to find a 14-year old in grade 2 or 3.
 Barbara E. Bullock & Almeida Jacqueline Toribio

In this relatively isolated, rural area, these same girls are the neighbors, relatives, play-
mates, classmates, or caregivers of the semi-literate or illiterate children.

2.2â•… Phonological distribution of DHS in the data

As stated earlier, the data yielded vanishingly few instances of DHS (103 tokens
across the selected corpus of 45,659 words), precluding a statistical analysis. None-
theless, from the tokens that do occur clear patterns of the phonological distribution
of intrusive [s] emerge from the data and these are consistent with those described by
Morgan (1998) in his case study of a tour guide from the urban capital, Santo Domingo.
Our results are summarized in Table 1.

Table 1.╇ Dominican hypercorrect [s] insertion across contexts

Word-final Word-internal Total

Phrase-final Phrase-internal

/T /V /Other /T /Other

Token 24 45 7 2 24 1 103
Percent 23.3% 43.68% 6.79% 1.94% 23.3% .97% 99.98%

Word internally, DHS occurs almost exclusively before voiceless stops (abbre-
viated as T), a pattern consistent with Morgan’s observations.6 Examples from the
selected corpus are given in (2).

(2) Word-internal DHS

a. Mi nombre es Eri[s]ka (F, 6, grade 2)
“My name is Erika.”
b. Lo repi[s]ten (M, 15, grade 5)
“They repeat it.”
c. las hama[s]cas (F, 60, literate)
“the hammocks”

It is important to note that word-internal insertions occur in just under a quarter of the
tokens. Far more frequent are hypercorrections in word-final position: DHS appears

.╅ Our full corpus yielded three tokens that were exceptions to this generalization; in
each of these cases, intrusive [s] was realized before a nasal stop (N) (e.g. pre[s]ñada <
preñada ‘pregnant’).
 Correcting the record on Dominican [s]-hypercorrection 

overwhelmingly in word-final position (76% of the time, approximating Â�Morgan’s

results). Most commonly, word-final intrusive [s] spans an intonational phrase
boundary and occurs if the following word begins with a voiceless stop. A hyper�
correct [s] only occasionally appeared before an unstressed initial vowel (V) in phrase-
internal position. These types of word-final [s] insertion are shown in the examples in
(3) and (4).

(3) Word-final DHS __#T

a. Ya[s] tenía raíz (F, 51, illiterate)
“It already had roots.”
b. Yo estudio lo que la profe me[s] pone en la escuela y lo hago (F, 7, grade 2)
“I study what the teacher gives me in school and I do it.”
c. Y ‘damás lo que encontraba era[s] plumas (M, 11, grade 5)
“And the only thing he found were feathers.”

(4) Word-final DHS __#V

a. Un compadre Lino tenía[s] un gallito (F, 12, grade 7)
“Old man Lino had a rooster.”
b. de[s] animales (M, 10, grade 4)
“about animals”

For some individuals in our study, hypercorrect [s] occurs in phrase-final position,
accompanied by a distinct falling boundary tone. What is remarkable about this form
of DHS is that certain individuals use it repetitively when narrating, as demonstrated
in examples (5a, b) below.

(5) Phrase-final DHS

a. Como no tenía burro[s], decidió irse a pie[s], como hubiera llovido[s],
y había mucho lodo[s] (M, 19, literacy skills unknown)
“Since he did not have a donkey, he decided to go by foot, since it had
rained, and there was a lot of mud.”
b. Había un viejo[s], llamado Lino y había un gallo[s], llamado Quiquiriquí,
y abajo de una mata[s], había un lobo (M, 13, grade 4)
“There was an old man, named Lino and he had a rooster, named
Quiquiriquí, and under a tree, there was a wolf.”
c. Ella es grande, no tan grande[s] (F, 10, grade 4)
“It [the town] is big, but not so big.”

These observations regarding the patterns of DHS in our corpus provide an exhaustive
description of the environments in which we found intrusive [s].
 Barbara E. Bullock & Almeida Jacqueline Toribio

3.â•… Discussion: Implications for previous generative analyses

Generative phonologists have cast DHS as a coda phenomenon that operates essen-
tially in a random fashion constrained only by the weight of the penultimate syllable
(*muersto) (Harris 2002) and by minimality, where only one [s] can be inserted per
word (Bradley 2006). Given the rarity of hypercorrect [s] in our corpus, it is not sur-
prising that examples refuting their claims may not have occurred. However, their
absence does not mean that they cannot occur.
Given the patterns of [s] insertion that we (and Morgan (1998)) have unearthed
from naturally occurring speech, we believe that there are several reasons to ques-
tion the premise that pace certain restriction, speakers freely insert [s]. Note, first, that
the [s] insertion is far from random. Rather, the site in which the segment occurs is
systematically constrained by either segmental context (occurring primarily before T
both within and between words) or prosody (arising in phrase-final position). Thus,
Núñez-Cedeño’s hypothesized ill-formed example, *casrretera might not arise in
Dominican speech, not because of the representational nature of the rhotic, but simply
because DHS occurs in a highly constrained segmental environment, primarily before
stops. What this means is that DHS is not random coda insertion.
Can we characterize DHS as a coda-based phenomenon at all? Recall that hyper-
correction occurs most frequently between words rather than within or at the edge of a
lexical item. That is, it seems that DHS is, in part, a sandhi phenomenon and, just as is
the case with the more familiar case of Romance sandhi, liaison in French, the syllabic
affiliation of the epenthesized [s] is ambiguous. Yet unlike French, [s] insertion cannot
plausibly be argued to resolve any syllable-based markedness violations. Because DHS
is most likely to occur before a voiceless stop, it increases syllable markedness, either
by adding a coda to a well-formed open syllable or by creating a complex onset from a
simple one. The only form of DHS that would result in an unmarked syllable would be
before a word-initial vowel, as in (4), and yet, as our data and Morgan’s data demon-
strate, this is one of the least likely environments for intrusive [s].
It could be argued that the presence of hypercorrect [s] in absolute word-final
position is indicative of its status as a coda element. However, it is not clear that the
domain of application for this form of DHS (as illustrated in (5)) is the syllable. Instead,
the intrusive [s] used by some speakers for narration appears to function more as a
(redundant) phrase marking boundary and, as such, would be bound to a higher unit
of prosodic structure than the syllable. Phrase-final frication, such as this, also appears
in the speech of some speakers of Parisian French, arguably as a consequence of the
reintroduction of a downstream noise source on a final devoiced vowel (Smith 2003;
Fagyal & Moisset 1999).
Finally, the fact that DHS occurs overwhelmingly before voiceless stops cannot
be accounted for by Harris’ depiction of the process simply as a coda rule. Even if
 Correcting the record on Dominican [s]-hypercorrection 

the environment of a variable coda rule were adjusted to limit the segmental context
before which epenthetic [s] could occur, our corpus weakly attests to an even stronger
refutation of the claim that DHS targets codas. That is, intrusive [s] sporadically occurs
in utterance initial position before T, apparently defying the phonotactics of the Spanish
language where initial sT clusters have been historically remedied by an epenthetic
vowel (e.g. escuela < Latin, SCOLA).7
(6) Phrase-initial DHS before T
[s]tampoco. (F 51, illiterate)
“(me) neither”

In this example, hypercorrect [s] cannot be appearing in a coda. Notice, though, that
the voiceless stop again appears to be strong “attractor” for an epenthetic [s].
In our view, the frequent presence of DHS before voiceless stops suggests that the
process might be driven by phonetic misperception. That is, speakers may have come
to misparse the turbulence induced by the period of glottal abduction necessary for
the production of T as [hT]. This may have given rise to the reinterpretation of [hT] as a
variant of [sT] (Ohala 1993; Ohala & Solé 2008). This possibility can only be confirmed
through controlled studies of perception.
In conclusion, through the use of natural production data, rather than hypothesized
exemplars, we have demonstrated that the distribution of DHS is much more con-
strained yet, at the same time, more complicated than has been previously thought.
Simply put, DHS is not random coda insertion. Our sociolinguistic data also refute the
dubious proposition that DHS is a feature of “lost-s” speakers. Clearly, there is much
to (re)learn about this intriguing phenomenon. We hope to have provided future
researchers with a more solid empirical base from which to begin.

References

Alba, Orlando. 2004. Cómo Hablamos los Dominicanos. Santo Domingo: Ramon Jimenés.
Bradley, Travis. 2006. “Spanish Rhotics and Dominican Hypercorrect s”. Probus 18.1–33.
Bullock, Barbara E. & Almeida Jacqueline Toribio. 2009. “Reconsidering Dominican Spanish:
Data from the Rural Cibao”. Revista Internacional de Lingüística Iberoamericana 14.49–73.
Fagyal, Zsuzsanna & Christine Moisset. 1999. “Sound Change and Articulatory Release: Where
and Why are High Vowels Devoiced in Parisian French”. Proceedings of the 15th International
Congress of Phonetic Sciences, San Francisco, 309–312.
Harris, James. 1983. Syllable Structure and Stress in Spanish. Cambridge, MA: MIT Press.

.╅ We have unearthed two such examples of utterance initial DHS in our selected corpus but
it is unclear if the other is a false start.
 Barbara E. Bullock & Almeida Jacqueline Toribio

Harris, James. 2002. “Flaps, Trills and Syllable Structure in Spanish”. MIT Working Papers in
Linguistics ed. by Aniko-Csirmaz, Zhigiang Li, Andrew Nevins, Olga Vaysman & Michael
Wagner, 81–108. Cambridge, MA: Department of Linguistics, MIT.
LÓpez-Morales, Humberto. 1990. “El Torno de la /s/ Final Dominicana: Cuestiones Teóricas”.
Voz y Letras 1.129–138.
Morgan, Terrell. 1998. “The Linguistic Parameters of /s/-Insertion in Dominican Spanish: A Case
Study of Qualitative Hypercorrection”. Perspectives on Spanish Linguistics ed. by Javier
Gutiérrez-Rexach & José del Valle, J., 79–96. Columbus, OH: Ohio State University.
Núñez-Cedeño, Rafael. 1988. “Structure-preserving Properties of an Epenthetic Rule in Spanish”.
Advances in Romance Linguistics ed. by David Birdsong & Jean Pierre Montreuil, 319–335.
Foris: Dordrecht.
Núñez-Cedeno, Rafael. 1989. “La /r/, único fonema vibrante del español”. Anuario de Lingüistica
Hispánica 5. 153–71.
Núñez-Cedeño, Rafael. 1994. “The Alterability of Spanish Geminates and its Effects on the
Uniform Applicability Condition”. Probus 6.23–42.
Ohala, John J. 1993. “Sound Change as Nature’s Speech Perception Experiment”. Speech Com-
munication 13.155–171.
Ohala, John J. & Maria-Josep Solé. 2008. “Turbulence and Phonology”. UC Berkeley Phonology
Lab Annual Report. University of California: Berkeley.
Smith, Caroline L. 2003. “Vowel Devoicing in Contemporary French”. Journal of French Language
Studies 13.177–194.
Terrell, Tracy. 1986. La entonación del español; su estructura fónica, variabilidad y dialectología.
Madrid: Cátedra.
Vaux, Bert. 2001. “Consonant Epenthesis and Hypercorrection”. Ms., Harvard University.
Vaux, Bert. 2002. “Consonant Epenthesis and the Problem of Unnatural Phonology”. Ms.,
Harvard University.
Vennemann, Theo. 1972. “Rule Inversion”. Lingua 29.290–242.
Zevallos, Andrés. 1997. Los cuentos del Tío Lino. Lima: Lluvia Editores.
 V-to-V assimilation in trisyllabic words
in French
Evidence for gradience and locality*

Christopher Carignan & Zsuzsanna Fagyal

University of Illinois at Urbana-Champaign

We investigate the extent of regressive vowel-to-vowel assimilation in trisyllabic

words in French. Sixty existing words were inserted in a carrier sentence, each
containing /a/ in word-initial unstressed position, and /a/ or /i/ in word-medial
unstressed and word-final stressed positions. Acoustic measurements taken at
the midpoint of vowels showed that /a/ in word-initial position was less fronted
and more open than /a/ in word-medial position when the vowels were followed
by a final front vowel. The degree of fronting of /a/ word-initially was greater
when the vowel was followed by a front vowel in both medial and final positions.
These results suggest that vowel-to-vowel assimilation in trisyllabic words in
French is a gradient and local effect of the final front stressed vowel on the vowel
immediately adjacent to it.

Key words:╇ anticipatory assimilation; Vowel Harmony; gradience;

phonetics; phonology

1.â•… Introduction

Word-level vowel-to-vowel assimilation (henceforth, VVA) is considered the phonetic

basis of vowel harmony in French, which is defined as “the influence of the vowel of
a stressed syllable on the pronunciation of a mid vowel in a preceding open syllable”
(Walker 2001:€54–55). When the vowel of the word-final primary stressed syllable is a
high or a high-mid vowel, the preceding mid vowel is expected to assimilate in height

*We would like to thank students and professors of the French Linguistics and SLATE dis-
cussion groups, as well as two anonymous reviewers of this paper for their comments and
suggestions for revising an earlier version of this paper. All shortcomings are ours.
 Christopher Carignan & Zsuzsanna Fagyal

to the following stressed vowel, as in (1a) and (2a). Conversely, when the stressed vowel
is a low or mid-low, the preceding mid vowel becomes mid-low, as in (1b) and (2b):

(1) a. aim-er /eme/ “to love”

b. aim-able /εmabl/ “lovely”

(2) a. au-to /oto/ “car”

b. au-tomne /ftfn/ “fall”

The phonetic mechanism of this harmony pattern is acoustically real, and is best under-
stood as “variations in tongue height along with related displacement of the tongue
position along the front–back axis” (Nguyen & Fagyal 2008:€ 1). The effect of the
word-final stressed vowel on the preceding vowel is non-morphologically conditioned
(contrary to Dell 1973 and in support of Tranel 1987), likely unrelated to word frequency,
and represents an instance of fine phonetic variation encoded in native speakers’ mental
lexicon. Thus, in light of Examples€ (1a) through (2b), vowel harmony in French is
regressive assimilation anchored in gradient lingual coarticulatory patterns, with the
final vowel as the trigger and the preceding vowel as the assimilating vowel.
Previous studies, however, almost exclusively focused on harmony in disyllabic
tokens. Reports on VVA in longer words are rare and somewhat contradictory. In tri-
syllabic words, for instance, at least two assimilatory patterns seem to operate: (1) none
of the preceding vowels assimilates to the final vowel (Example€(3a)), and (2) assimila-
tion is local, affecting only the vowel of the penultimate syllable, but not the vowel of
the word-initial suffix (Examples€(3b, c, and d)). In addition, the second scenario seems
irregular, as in some cases, such as the word territoire “territory” (3e), the penultimate
rather than the final vowel seems to trigger the local assimilatory effect: “This word is
exceptional in that the vowel causing the harmony is not the stressed vowel, but the
penultimate /i/” (Walker 2001:€110).

(3) a. orgueil /ftgœj/ “pride” > orgueilleux /ftgœjø/ “proud” (Tranel 1987:€61)
b. clair /klεt/ “clear” > éclairer /eklere/ “to clarify”
c. code /kfd/ “code” > décoder /dekode/ “to decode”
d. aveugle /avœgl/ “blind” > aveugler /avøgle/ “to blind”
e. terre /tεt/ “land” > territoire/teritwat/“territory” (Walker 2001:€54)

In light of these examples, one can conclude that the extent of VVA in long words
remains little understood. As opposed to disyllabic words, morphological complexity
arising from the presence of a prefix in trisyllabic words is likely to play a role. In addi-
tion, local assimilatory effects might not be tied to word-final stress. It is possible that
the assimilatory effect of a front vowel, if any, operates regardless of the location of
the triggering vowel in the word. Since actual realizations of VVA in trisyllabic words
have not yet been studied acoustically, the present study is designed to investigate this
phenomenon in a restricted and controlled set of existing words in French.
 V-to-V assimilation in trisyllabic words in French 

2.â•… Research questions and hypotheses

In this study, we investigate the extent of regressive VVA in French, using acoustic
phonetic methods in existing trisyllabic words with comparable phonetic structure.
We follow previous work by Nguyen and Fagyal (2008) showing that VVA in disyllabic
words is grounded in gradient coarticulatory patterns (see also Ohala 1994; Benus &
Gafos 2007; Cole 2009).
Our first research question is whether there is extensive anticipatory coarticu-
lation between the word-final stressed, or trigger, vowel (henceforth, V1) and the
preceding target vowel (henceforth, V2) in a trisyllabic word, as has been shown for
disyllabic words. Our second research question targets the prosodic domain of VVA,
i.e. whether the assimilatory influence of the stressed vowel (V1) extends to the word-
initial vowel (henceforth, V3).
Our hypothesis for the first research question is that, in agreement with previous
findings on disyllabic words, there will be extensive local assimilation of V2 to V1
in longer words. Given that findings from previous studies do not authorize precise
predictions of the extent of VVA in long words, we consider the second research
question an empirical question that results from our experiment; nevertheless, there
are expected answers for a restricted set of target words. We expect that if the domain
of VVA is the word, we should find acoustic evidence of coarticulation triggered by
V1 on both V2 and V3. If VVA is local and non-subject to feature-spreading rules,
however, then vowels non-adjacent to V1 in long words would not show regres-
sive assimilation to V1. In this case, one would more likely conclude that VVA in
French is restricted to local coarticulatory assimilation, confirming observations in
Examples€(3b–e).

3.â•… Methodology

A corpus of sixty trisyllabic French words was generated using the phonetic search
feature of the electronic version of the Petit Robert dictionary (see Appendix for the
complete word list). Words were either vowel or consonant-initial and each of the
three syllables contained either /a/ or /i/ as their nucleus. Although we were primarily
interested in observing the results of VVA in i-i-a and a-a-i type words, there were no
entries of i-i-a type words in the dictionary. Therefore, we examined results from all
words which contained /a/ in V1 position, of which we found sixty entries in the dic-
tionary. Therefore, we are focusing in this study on words with /a/ in word-initial and
word-medial positions, in order to compare the effect of the word-final trigger vowel
(/a/ or /i/) on the same target vowel. Target words contained a variety of intervocalic
consonants to make sure no single consonant or consonant type could exert a biasing
 Christopher Carignan & Zsuzsanna Fagyal

coarticulatory influence on any of the target vowels. Additionally, although VVA has
been defined as applying to mid vowels in French, we expect the phonetic mechanisms
of VVA to be fundamentally the same when the target vowel is an open /a/ rather than
a mid vowel. Thus, this experiment is based exclusively on /a/ as target vowels (V3 and
V2) and /i/ vs. /a/ as trigger vowels (V1) in order to guarantee uniformity in acoustic
information about the vowels in each position in the word.
The participants recruited for this first study were three native speakers of European
French: 2 female speakers from the North East of France and 1 male speaker from the
North-West of France, all exchange students at the University of Illinois. The speak-
ers were recorded in the recording booth of the Phonetics Lab at the University of
Illinois at Urbana-Champaign using an AKG C 520 head-mount condenser micro-
phone, positioned 3–5 cm from the corner of the mouth. Words from the list were
presented visually to the participants as a series of slides in Power Point which they
were instructed to read aloud in a normal voice and volume, and at a normal speak-
ing rate. The target word was first presented in the carrier phrase “Il retape (target
word) parfois” (“He retypes (target word) sometimes”). The target word was then
presented in isolation in the next slide, followed by a blank slide (see Figure 1 for an
example of the series of slides). We will, henceforth, refer to the first presentation
type as in-sentence and the second as isolated conditions. Two of these series were
included for each of the 75 words, in randomized order, resulting in 4 tokens for each
target word (2 in the in-sentence condition and 2 in the isolated condition), and an
overall total of 300 combined tokens for the experiment. The experiment was self-
paced: participants could determine the amount of time they took between each slide.
One break was built into the presentation of the slides, and none of the participants
took more than one five to ten minutes break to complete his/her readings. The experi-
ment typically lasted about 15 minutes.
The acoustic analysis was performed in Praat. Syllable and phone boundaries were
segmented using a Praat plugin implementing Jean-Philippe Goldman’s EasyAlign
automatic segmentation algorithm for French,1 and were verified and corrected manu-
ally, if needed. A Praat script was used which measured the formant values in Hertz
(Hz) at the midpoint of each of the three vowels in the target word; the midpoint was
chosen as the location where the formant values would be measured in order to mini-
mize any coarticulatory acoustic effects on the formant values from the neighboring
consonants. These values were cross-checked manually for outliers, and the manually
measured values were used as replacements for the outliers, if the outlier values were
deemed to be incorrect. These changes amounted to less than one percent of all formant

.â•… Jean-Philippe Goldman’s webpage is available at: http://www.unige.ch/lettres/linguistique/

goldman/
 V-to-V assimilation in trisyllabic words in French 

values, and concerned primarily /i/ that was sometimes devoiced word-finally. For the
purposes of statistical analyses, the vowel formant frequencies in Hz were converted
to the perceptually more relevant auditory frequency scale, called Bark scale,2 using
Traunmüller’s formula (Traunmüller 1990, cited in Nguyen & Fagyal 2008:€7).

11 retape cagibi parfois Cagibi

Figure 1.╇ Example of the series of Power Point slides presented to the speakers to read aloud

4.â•… Results

4.1â•… Extent of V-to-V assimilation influence in French

In order to test whether coarticulatory influence from the trigger vowel can extend
to the left edge of the target word, words featuring /a/ as both target vowels (V3 and
V2) and the two different trigger vowels (V1) were selected. Figure 2 shows V1, V2
and V3 formant frequencies measured at the vowel midpoints in a-a-i type words for
each speaker and both conditions. F1 frequencies are plotted along the y-axis, F2 along
the x-axis,3 using the originally measured Hz scale. Ellipses represent productions of
each vowel within two standard deviations. What we can observe immediately from
these figures is that, for all speakers, the vowel space for V2 indicates, in general, lower
F1 and higher F2 values than the vowel space for V3. In other words, /a/ is more
closed and more fronted in word-medial than in word-initial positions before /i/ for
all speakers.
Differences between the distributions of V3 /a/ and V2 /a/ in the two word condi-
tions are shown in Figure 3. The corresponding distributions for the three speakers
are distinguished by different shades and patterns. In each case, the left-hand pairs

.â•… The human auditory system’s response to frequency is non-linear. A difference in 500 Hz
between two sounds is perceived differently, for instance, below and above 2 kHz. In order to
normalize for these discrepancies, the use of the Bark or Mel scales is recommended (Johnson
2003:€51; Stevens 2000:€237).
.╅ All formant plots were generated using JPlotFormants v1.4. The program can be down-
loaded from the following web site: http://www.linguistics.ucla.edu/people/grads/billerey/
PlotFrog.htm
 3100 2900 2700 2500 2300 2100 1900 1700 1500 1300 1100 900 700 500 2700 2500 2300 2100 1900 1700 1500 1300 1100 900 700 500
F2 191 F2 200
Speaker F1 291 Speaker M1 300
391
400
491
591 500

691 600
791
700
891
800
991
1091 900

F1 F1
 Christopher Carignan & Zsuzsanna Fagyal

2900 2700 2500 2300 2100 1900 1700 1500 1300 1100 900 700 500
F2 200

Speaker F2 300
400
500
* -V3 /a/
600
700 + -V2 /a/
800
o -V1 /i/
900
1000
1100

Figure 2.╇ Plotted formant values in Hz and 2-standard-deviation ellipses for all vowels in a-a-i type words. Measurements from the in-sentence
and isolated conditions are collapsed in for easier reading of the multiple data points
 V-to-V assimilation in trisyllabic words in French 

display the distribution for V3 /a/, with the leftmost plot of each pair corresponding
to the in-sentence condition. The general tendency of lower F1 and higher F2 is again
confirmed for V2 /a/ in comparison to V3 /a/. Given the negative correlation between
F1 frequency and tongue height, and positive correlation between F2 frequency and
fronting of the tongue body, these findings strongly suggest that V2 /a/ is more fronted
and closed than V3 /a/ and, therefore, that V1 /i/ has, at the very least, an extensive
coarticulatory influence on this word-medial vowel. However, the distribution of F1
values in isolated condition for speaker F2 and the distributions of F1 and F2 values
for speaker M1 in both word conditions show subtle but clear differences, which called
for inferential statistics.
We ran Paired Samples t-tests to verify statistically if /a/ in V2 was, on average,
significantly more /i/-like than its corresponding counterpart in V3 (Hypothesis 1).4
Statistics were calculated separately for each speaker, vowel type, formant, and word
condition. Results of individual t-tests and descriptive statistics shown in Table 1 (see
Appendix) confirm observations depicted in Figures 2 and 3: on average /a/ immediately
preceding /i/ had significantly lower F1 and higher F2 than /a/ in word-initial position
for each speaker and both word conditions. While in most paired comparisons, V3 and
V2 were not correlated, which means that realizations of /a/ could be assumed to be
independent in each syllable, in three out of four comparisons formant values in V3
and V2 showed weak but statistically significant correlation (p < 0.05) in speakers’
F1 and M1 speech. This indicated that formant values in the two word conditions could
not be collapsed and analyzed jointly, but rather word condition had to be kept as a
separate independent variable in further statistical analyses.5
A comparison of V3 and V2 in a-a-i type words, as performed above, can help
determine whether the word-final trigger vowel has a significant effect on the word-
medial vowel, as we have just seen by observing the greater closing and fronting influ-
ence of V1 /i/ on V2 /a/ than on V3 /a/. However, this type of comparison cannot
fully describe the extent of this influence, namely, whether V1 /i/ has any effect on the
production of V3 /a/. Therefore, a comparison of V3 /a/ and V2 /a/ in a-a-i type words
to those in a-a-a type words is needed, in order to observe possible differences due to
the trigger vowels.
The plots in Figure 4 display F1 and F2 formant frequency ellipses within two
standard deviations from mean values (denoted by the symbols indicated in the legend)
of V3 /a/ and V2 /a/ in a-a-a and a-a-i words. We can observe, for all speakers, that

.╅ Statistical analyses reported in this paper have been performed using the software package
SPSS 17.0.
.╅ Measurements from the in-sentence and isolated conditions are collapsed in Figures 2 and
4 for easier reading of the multiple data points and/or ellipses.
9 7.5 7.5
F1 F1 F1

8 7.0 7.0

7 6.5 6.5

6 6.0 6.0

5 5.5 5.5

4 5.0 5.0

V3-In-sentence V3-Isolated V2-In-sentence V2-Isolated V3-In-sentence V3-Isolated V2-In-sentence V2-Isolated V3-In-sentence V3-Isolated V2-In-sentence V2-Isolated
 Christopher Carignan & Zsuzsanna Fagyal

15 12.0 15
F2 F2 F2
12.0
14
14
11.5
13
13 11.0
12
10.5
12
11
10.0

11 0.5 10
V3-In-sentence V3-Isolated V2-In-sentence V2-Isolated V3-In-sentence V3-Isolated V2-In-sentence V2-Isolated V3-In-sentence V3-Isolated V2-In-sentence V2-Isolated

Speaker F1 Speaker M1 Speaker F2

Figure 3.╇ Boxplots of F1 and F2 formant frequency distributions in a-a-i type words for V3 /a/ (left boxplot pairs) and V2 /a/ (right boxplot
pairs) in two word conditions (in-sentence and isolated)
 V-to-V assimilation in trisyllabic words in French 

V2 /a/ in a-a-i type words has a lower F1 and higher F2 than any of the other three
instances of /a/. However, we can also observe that the mean of the word-initial V3
vowel in a-a-i words is not any higher or lower than the mean of the initial V3 vowel
in a-a-a words. Taken together, these observations highlight the likelihood that word-
medial V2 /a/ is more fronted and more closed than the other three vowels. This sug-
gests that there is no regressive assimilatory influence of the trigger vowel V1 /i/ on V3
/a/ in the speech of any of the three speakers.
To test for the statistical significance of these tendencies, we conducted Repeated
Measures ANOVAs6 with position of target vowel (V3 and V2), type of trigger (/i/ and
/a/), and word condition (in-sentence and isolated) as independent within-subject fac-
tors and F1 and F2 values as dependent variables. This 2x2x2 design yielded separate
comparisons for each formant and speaker. The relative contribution of each source
of variance tested is shown in Table 2 in the Appendix. Cells depicting statistically
significant (at p < 0.05) single or joint factors are shaded.
For speaker M1, tests of within-subject effects revealed that for the first formant,
there were significant main effects of position of /a/ in the word for the first formant
(F(1, 28) = 62.836, p < 0.001), type of triggering vowel (F(1, 28) = 14.197, p < 0.001),
and word condition (F(1, 28) = 17.288, p < .001). Among the interactions, only the
interaction of the position of /a/ in the word and the type of trigger was significant
(F(1, 28) = 7.240, p€ <€ 0.012), which means that the type of trigger had a different
effect depending on which position the target vowel was placed in the word (initial or
medial). It should also be noted that the joint interaction of position of /a/ in the word
and word condition was not significant, but with more data points, it might approach
significance (F(1, 28) = 4.172, p€<€0.050). For the second formant, there was a widely
significant main effect of the position of /a/ in the word (F(1, 28) = 19.198, p < 0.001),
but neither the type of trigger vowel (F(1, 28) = 3.646, p < 0.066) nor word condition
(F(1, 28) = 2.608, p < 0.117) contributed significantly to observed variance in the data.
For female speaker F1, the same tests revealed more significant interactions for
both formants. For the first formant, there were significant main effects of position of
/a/ in the word (F(1, 28) = 29.282, p < 0.001) and type of triggering vowel (F(1, 28)€=
12.140, p < 0.002), but word condition had no significant main effect (F(1, 28)€= 2.379,
p < .134). Similar to speaker M1, only the interaction of the position of /a/ in the
word and the type of trigger was significant (F(1, 28) = 25.762, p < 0.001), but in this

.â•… Prior to performing these analyses, we used Mauchly’s test (p > 0.05) to test for the RM
ANOVA version of equality of variances (see Field 2005:€427–482). As Mauchly’s test returned
non-significant results in each comparison, we assumed that the condition of sphericity was
met in each case.
 2700 2500 2300 2100 1900 1700 1500 1300 1100 900 700 500 2300 2100 1900 1700 1500 1300 1100 900 700 500
F2 200 F2 200
Speaker F1 300 Speaker M1 300
400
400
500
600 500
X
700 600
X
+O* 800 + 700
900 O*
800
1000
1100 900

F1 F1
 Christopher Carignan & Zsuzsanna Fagyal

2700 2500 2500 2100 1900 1700 1500 1300 1100 900 700 500
F2 200
300
Speaker F2 400
* -a-a-a
500
600 + -a-a-a
X 700
+ o -a-a-i
O* 800
900 X -a-a-i
1000
1100

Figure 4.╇ Two standard deviation ellipses and distribution means represented by the symbols listed in the legend in Hz for V2 /a/ and V3 /a/ in
a-a-a and a-a-i type words. Measurements from the in-sentence and isolated conditions are collapsed in for easier reading of the multiple data points
 V-to-V assimilation in trisyllabic words in French 

speaker’s speech both formants showed this effect. Again, type of trigger exerted a dif-
ferent effect on /a/ depending on its position in the target word.
In female speaker F2’s speech, position of /a/ in the word (F(1, 28) = 41.871, p <
0.001), type of trigger (F(1, 28) = 14.075, p < 0.001), and word condition (F(1, 28) =
7.241, p < .012) were also significant. As opposed to the other speakers, these factors
were significant for both formants. Besides the joint effect of position of /a/ and type
of trigger (F(1, 28) = 10.784, p < 0.003) for the second formant, there was a significant
three-way interaction between the three factors. The joint interaction of position of /a/
and word condition was also statistically significant (F(1, 29) = 13.560, p < 0.001) for
the same formant. This means that for this speaker the type of trigger had a signifi-
cantly different effect on /a/ in the front/back dimension depending on which position
the target vowel (initial or medial) was placed in the word, as well as in which word
condition (in-sentence or isolated).
To summarize, position of /a/ in the target word and type of trigger were signifi-
cant within-subject factors for all speakers and all formants but the second formant in
M1’s speech. The interaction of these two factors was also significant for most speak-
ers and dependent variables. Word condition has also contributed significantly to the
observed variances, but less often and less unequivocally than the two other factors.
Since the interaction between position of /a/ in the target word and type of trigger is
of particular interest for answering our second research question (does V-to-V assimi-
lation extend to the left edge of the word), we will now consider this interaction in
greater detail.
Figure 5 displays, in Bark for each speaker, the interaction plots of mean F1 values
in the top row and mean F2 values on the bottom row for the target vowel /a/ in two
positions (V3 or V2) and with two different trigger vowels (/a/ or /i/) in the target
words (a-a-a and a-a-i). For each pair of dots, the left-hand dot represents the mean of
V3 /a/, while the right-hand dot represents the mean of V2 /a/. The dotted lines con-
nect mean values of V3 /a/ and V2 /a/ when the trigger V1 is /i/ (a-a-i type words),
while the solid lines connect mean values of these same vowels when the trigger V1 is
/a/ (a-a-a type words).
In order to see a different effect of the trigger on /a/ in V3 than in V2 as a function
of the type of trigger (/i/ or /a/), we should look for non-parallel lines. This is what we
observe for both formants for all three speakers. As we can see, /a/ in V3 (word-initial)
position was not at all affected by the type of trigger in four out of six cases. The two
exceptions were the two female speakers, who seem to have fronted /a/ in both word-
initial and word-medial positions in a-a-i type words (which is why mean F2 is higher
for V3 /a/ in these words). However, the vowel /a/ was always significantly more closed
(lower F1) and more fronted (higher F2) in word-medial position (V2) than in word-
initial (V3) positions for all speakers.
 6.80
7.25 6.80
6.60
7.00 6.60
6.40
6.75 6.40

F1(Bark)
6.20

F1(Bark)
F1(Bark)
6.50 6.20
type_of_trigger 6.00 type_of_trigger type_of_trigger
6.25 /i/ /i/ /i/
/a/ 5.80 /a/ 6.00 /a/
V3(initial) V2(medial) V3(initial) V2(medial) V3(initial) V2(medial)
position of /a/ in target word position of /a/ in target word position of /a/ in target word
 Christopher Carignan & Zsuzsanna Fagyal

13.20 type_of_trigger 11.20 type_of_trigger 12.80 type_of_trigger

/i/ /i/ /i/
13.00 /a/ 11.10 /a/ 12.60 /a/
11.00
12.80 12.40
10.90
12.60 12.20

F2(Bark)
F2(Bark)
F2(Bark)

10.80
12.40 10.70 12.00

12.20 10.60 11.80

V3(initial) V2(medial) V3(initial) V2(medial) V3(initial) V2(medial)
position of /a/ in target word position of /a/ in target word position of /a/ in word
Speaker F1 Speaker M1 Speaker F2
Figure 5.╇ F1 and F2 distribution means in Bark for V3 /a/ (left in dot pairs) and V2 /a/ (right in dot pairs) in a╂a╂a and a-a-i type words
in both conditions
 V-to-V assimilation in trisyllabic words in French 

4.2â•… Coarticulatory strength

A notable finding that we did not anticipate was evidence that the trigger vowel may
have a gradient effect on the degree of regressive assimilation of the preceding vowel.
Figure 6 displays the formant frequency distributions and means of V3 /a/ in all word
types containing /a/ in their first syllable. As we can see, the fronting and closing influ-
ence of /i/ is slightly greater when the trigger is longer, in other words, when /i/ is both
in final and in medial positions in the word. This increased effect is observed for all
three speakers. However, speaker F2 seems more variable in terms of fronting of V3 /a/
across word types than the other two: we can see that there is some fronting and closing
of word-initial V3 /a/ when /i/ is adjacent to it in type a-i-a words when compared to
V3 /a/ in a-a-a type words. Nevertheless, these correlates are not as pronounced as
when the /i/ trigger is “longer” in a-i-i words (i.e. when both V1 and V2 are comprised
in the /i/ trigger).
Due to the small number of a-i-a type words in the corpus, we used non-parametric
Independent Samples Mann-Whitney (2-tailed) tests to verify the statistical signifi-
cance of these observations.7 Results show that the first formant was, on average, sig-
nificantly lower in V3 /a/ when it was followed by a “long” (a-i-i) front trigger than by
a “short” (a-i-a) front trigger in two out of three speakers’ speech. These were: speaker
F18 (Mdn aii = 6.46, Mdn aia = 7.42; U = 378.5, p < 0.001, r = −0.379) and speaker M1
(Mdn aii = 6.37, Mdn aia = 6.57; U = 373, p < 0.008, r = −0.311). Formant values were
not significantly different in the two contexts for speaker F2 (Mdn aii = 7.12, Mdn aia€=
7.27; U = 574, p < 0.28ns, r = −0.127).
Results for the second formant are even more convincing. The second formant
was, on average, significantly higher in V3 /a/ in a-i-i type words that in a-i-a type
words for all speakers. (Speaker F1: Mdn aii = 13.22, Mdn aia = 12.65; U = 430, p€<
0.005, r = −0.318; Speaker F2: Mdn aii = 12.667, Mdn aia = 12.235, U = 451, p < 0.017,
r = −0.270; Speaker M1: Mdn aii = 11.13, Mdn aia = 10.57; U = 343.5, p < 0.003,
r€=€−0.352). Although the effect size is small in each case, these results confirm overall
the gradient, essentially fronting, effect of a longer trigger on the preceding vowel in
our target words.

.╅ Statistical significance of the Kolmogorov-Smirnov tests confirmed this choice: first and
second formant values, indeed, showed non-normal distributions for both formant frequen-
cies in all three speakers’ speech.
.╅ Following conventions of reporting on the results of Independent Samples Mann-Whitney
tests, we give the median (Mdn), the U, p, and r values for each test. Note that it is more
appropriate to report on the median than the mean for non-parametric test, and r values allow
gauging the effect size (Field 2005:€532).
 2700 2500 2300 2100 1900 1700 1500 1300 1100 900 700 500 2300 2100 1900 1700 1500 1300 1100 900 700 500
F2 200 F2 200
Speaker F1 300 Speaker M1 300
400
400
500
600 500
O 700 600
+X* 800 O 700
900 X*+
1000 800
1100 900

F1
 Christopher Carignan & Zsuzsanna Fagyal

F1

2500 2300 2100 1900 1700 1500 1300 1100 900 700 500
F2 200
300
Speaker F2 400
* -a-a-a
500
600 + -a-i-a
700 O -a-i-i
O + 800
X* X -a-a-i
900
1000
1100

Figure 6.╇ 2-standard-deviation ellipses and distribution means (represented by the symbols listed in the legend) in Hz for V3 /a/ in a-a-a, a-i-a,
a-i-i and a-a-i type words
 V-to-V assimilation in trisyllabic words in French 

5.â•… Discussion

In summary, based on the results of this experiment, we can say that the regressive
assimilatory influence of the word-final front stressed vowel /i/ did affect the open
vowel /a/ immediately adjacent to it. Thus, our hypothesis for the first research question
is, confirmed: there is, indeed, extensive assimilation of V2 to V1 in trisyllabic words,
as has been shown by previous studies for disyllabic words. However, the regressive
assimilatory influence of the final vowel did not extend to the leftmost, word-initial
/a/ in the target words. We must, therefore, conclude that regressive V-to-V assimi-
lation appears to be a local phenomenon. Although further analyses of mid vowels
are warranted, results reported here call into question any representation of VVA by
feature-spreading rules or constraints relevant to the entire word as their domain of
application. Our results seem more concurrent with the few examples of fronting of
the penultimate vowel in trisyllabic words reported in Examples€(3b–d) in the Intro-
duction. The additional observation that regressive assimilation significantly increased
in strength when the trigger vowel had a greater temporal extent lends further support
to the interpretation that VVA is a gradient coarticulatory phenomenon (Nguyen &
Fagyal 2008:€19–22).
Exemplar-theoretical representations of VVA can, very likely, account for these
effects, as both local and long-distance assimilation can result in convergent regu-
larities between fine phonetic detail and lexical representations, and thus lead to the
incorporation of these details in native speakers’ mental lexicon (see Cole 2009 for a
discussion). Lack of evidence for long-distance assimilation in long words, such as we
found in this corpus, would argue against formal accounts of the VVA phenomenon
in French as the spreading of a single phonological feature to the leftmost edge of the
word (Dell 1973; Casagrande 1984), or markedness constraints associated with the
same prosodic domain (Durand & Lyche 2004).
How this local regressive assimilation interacts with prosodic structure, such as
initial accent,9 and how it fits in with other phonological phenomena in French, such
as pre-tonic vowel raising, remains to be investigated (see Casagrande 1984). It is pos-
sible that local hyperarticulation due to initial accent in French could exert an influence
on word-initial vowels in our target words.10 However, long words in French have
been shown to display variable initial accent placement (see Jun & Fougeron 2002),
which means that initial accent might or might not have been realized in all target
words used in this study. We have no evidence for systematic initial accent occurring

.╅ Initial accent (accent initial) in French refers to phrase-initial pitch prominence, whose
realization is extremely variable. When it is realized, initial accent typically falls on the first or
second syllable of a lexical word in an accentual phrase.
.╅ We thank one of the anonymous reviewers of this paper for this observation.
 Christopher Carignan & Zsuzsanna Fagyal

on all word-initial syllables in our corpus. Also, the prosodic status of initial accent in
Parisian French is different from final stress, as the most recent phonological analyses
represent the former at the level of the accentual phrase rather than the prosodic word
(see Welby 2006). Thus, at this point it is unclear what our predictions of a possible
effect of initial accent would be. Provided the initial accent is systematically realized,
one could predict the some sort of an enhancement of /a/, i.e. the word-initial vowel
becoming more /a/-like and therefore resisting the assimilatory effect of the word-final
front vowel. This is a possible effect, but it requires targeted investigations, in which
one of the test conditions is the absence or presence of initial accent. We leave the
exploration of this question for future research.
The same is true for possible vowel reduction word-medially. Although one would
not predict vowel reduction in word-medial unstressed syllables in this or other dia-
lects of French spoken in France (see Valdman 1993:€30–31), experiments designed
to separate out possible effects of position and assimilatory fronting would have to be
carried out in order determine if fronting of V2 in our target words is due entirely to
the following front vowel.

6.â•… Conclusion

The current study showed that the only vowel subject to vowel-to-vowel assimilation
in a trisyllabic word in French is very likely the word-medial vowel, the one adja-
cent to the word-final front trigger. Although important individual variations were
observed with regard to fronting that need to be further investigated, this assimila-
tory pattern was robust and general across the three speakers and multiple real words.
Results reported in this paper suggest that regressive vowel-to-vowel assimilation in
long words in French exists, but it is a local assimilatory phenomenon rather than a
word-level phonological process.

References

Benus, Stefan & Adamantios Gafos. 2007. “Articulatory Characteristics of Hungarian ‘Transparent’
Vowels”. Journal of Phonetics 35. 271–300.
Casagrande, Jean. 1984.€The Sound System of French. Washington, DC: Georgetown University Press.
Cole, Jennifer. 2009. “Emergent Feature Structures: Harmony systems in exemplar models of
phonology”. Language Sciences 31€: 2–3.144-160.
Dell, François. 1973. Les règles et les sons. Paris: Hermann.
Durand, Jacques & Chantal Lyche. 2004. “Structure et variation dans quelques systèmes vocaliques
du français: l’enquête Phonologie du français contemporain (PFC)”. Variation et francophonie
ed. by Aidan Coveney & Carol Sanders, 217–240. Paris: L’Harmattan.
Field, Andy. 2005. Discovering Statistics using SPSS. Second Edition. London, New Delhi: Sage
Publications.
 V-to-V assimilation in trisyllabic words in French 

Johnson, Keith. 2003. Acoustic & Auditory phonetics. Second Edition. Malden (MA): Blackwell.
Jun, Sun-Ah & Cécile Fougeron. 2002. “Realizations of Accentual Phrase in French Intonation.”
Probus 14: 1.147–172.
Nguyen, Noël & Zsuzsanna Fagyal. 2008. “Acoustic aspects of vowel harmony in French”. Journal
of Phonetics 36: 1. 1–27.
Ohala, John. 1994. “Towards a Universal, Phonetically-Based Theory of Vowel Harmony”.
Proceedings of the International Conference on Speech and Language Processing (ICSLP ’94),
Yokohama, Japan. 491–494.
Stevens, Kenneth. 2000. Acoustic Phonetics. Second edition. Cambridge (MA): MIT Press.
Tranel, Bernard. 1987. The Sounds of French. Cambridge: Cambridge University Press.
Traunmüller, H. 1990. “Analytical Expressions for the Tonotopic Sensory Scale”. Journal of the
Acoustical Society of America 88. 97–100.
Valdman, Albert. 1993. Bien entendu! Introduction à la prononciation française. Upper Saddle
River (NJ): Prentice Hall.
Walker, Douglas. 2001. French Sound Structure. Calgary: University of Calgary Press.
Welby, Pauline. 2006. “French Intonational Structure: Evidence from tonal alignment.” Journal
of Phonetics 34: 3. 343–371.

Appendix A

List of all trisyllabic words with /a/ in V1 position, and /a/ or /i/ in V2 and V3 positions
(Le Petit Robert électronique or electronic version of Le Petit Robert):

[a] [a] [a] panama padischah canitie agami nagari

abaca pataras rabbinat kamichi amathie panaris
ananas patata (7) lapilli apathie pahlavi
apparat tarama parisis aphasie paradis
baccara (15) [a] [i] [i] sashimi arrachis ramassis
balata achylie tassili avachi patati
baraka [a] [i] [a] alibi (13) avanie safari
canada assignat ascidie avarie salami
hamada califat athymie [a] [a] [i] canari tabagie
kamala habitat atypie abasie gabarit tatami
malaga harissa avili abattis hallali (25)
nacarat marina cagibi acabit maladie

Appendix B

Table 1. Means, standard deviations and results of Paired Samples t-test (two-tailed)
for F1 and F2 of /a/ in V3 (word-initial) and V2 (word-medial) positions and in two
word conditions; **p < = .001, *p < = .0.05, ns non-significant. Low F1 (in V2) and F2
(in V3) values are shaded in grey.
 Christopher Carignan & Zsuzsanna Fagyal

In-Sentence Isolated

V3 V2 V3 V2

Mean St.dev. Mean St.dev. Mean St.dev. Mean St.dev.

Speaker F1
F1 7.29 .68 6.23 .72 7.21 .69 5.98 .79
t-test t(48) = 7.408**, r = −.049ns t(48) = 9.765**, r = −.281*
F2 12.44 .52 13.19 .51 12.60 .53 13.25 .54
t-test t(48) = −9.957 , r = .142*
** t(48) = −9.991 , r = .268*
**

Speaker F2
F1 7.44 .48 6.54 .68 7.31 .47 6.42 .67
t-test t(48) = 6.968**, r = −.200ns t(48) = 7.397**, r = −.021ns
F2 12.22 1.05 12.67 .53 11.95 .56 12.68 .53
t-test t(48) = 2.875**, r = .142ns t(48) = 7.177**, r = .161ns
Speaker M1
F1 6.81 .37 6.06 .44 6.65 .42 5.98 .43
t-test t(48) = 8.345 , r = −.183ns
** t(48) = 6.657 , r = −.292*
**

F2 10.65 .44 11.11 .53 10.71 .53 11.19 .56

t-test t(48) = −5.35**, r = .254ns t(48) = −4.926**, r = .240ns

Appendix C

Table 2. Within subject interactions in the Repeated Measures ANOVA design (position
of target vowel x type of trigger x word condition) for F1 and F2 for each speaker.
Statistically significant (p < .05) single and joint factors are shaded in grey.

Speaker F1 Speaker M1 Speaker F2

factors F1 F2 F1 F2 F1 F2

position_in_word p < 0.001 p < 0.001 p < 0.001 p < 0.001 p < 0.001 p < 0.029
type_of_trigger p < 0.002 p < 0.036 p < 0.001 p < 0.066 p < 0.001 p < 0.001
sentence_position p < 0.134 p < 0.013 p < 0.001 p < 0.117 p < 0.012 p < 0.004
position_a_in_word* p < 0.001 p < 0.001 p < 0.012 p < 0.071 p < 0.003 p < 0.016
type_of_trigger
position_a_in_word* p < 0.246 p < 0.204 p < 0.050 p < 0.333 p < 0.617 p < 0.016
sentence_position
sentence_position* p < 0.314 p < 0.214 p < 0.463 p < 0.077 p < 0.591 p < 0.374
type_of_trigger
position_a_in_word* p < 0.356 p < 0.272 p < 0.562 p < 0.908 p < 0.435 p < 0.001
type_of_trigger*
sentence_position
 The production and provenance of palatal
nasals in Portuguese and Spanish

Ryan Shosted & José Ignacio Hualde

University of Illinois at Urbana-Champaign

We present acoustic evidence bearing on the articulation of the palatal nasal

consonants in Portuguese and Spanish. We outline the etymological origins of
these consonants, some of which are shared by the two languages. One class
of Portuguese palatal nasals arose from a nasalized palatal approximant – the
phonetic leftovers of a historical process (typical of Galician-Portuguese but
not Spanish) that nasalized vowels and deleted intervocalic alveolar nasal
consonants. Some descriptions of Brazilian Portuguese suggest that this
nasalized approximant is still extant and perhaps even dominant among
speakers of the modern language. We compare the articulation of palatal
nasals in a variety of vocalic contexts to determine the degree of impedance.
We find evidence suggesting that the nasalized approximant of medieval
Galician-Portuguese may have been retained in Brazilian Portuguese. This
highly conservative trait appears to have spread throughout the lexicon even to
etymologically unrelated forms and may be the reason for attested asymmetries
in coarticulation among BP nasal consonants.

1.â•… Introduction

1.1â•… The problem of the Portuguese palatal nasal

When examining the diachronic “merger” of speech sounds, the experimental phonoÂ�
logist must now consider whether and to what extent synchronic variation persists in
categories of sounds even after these categories have nominally collapsed into a single
phonological unit. Indeed, mounting evidence suggests that sub-phonemic distinc-
tions may be both systematic and persistent (Warner et al. 2004; Hay et al. 2006;
Kilpatrick et al. 2007, inter alia).
The question of phonetic differences in diachronically merged phonemes has
immediate relevance in Brazilian Portuguese (BP), where a uniform description of
the phonemic palatal nasal obstruent (spelled 〈nh〉) has not been reached and where
historical evidence suggests the phoneme has a number of articulatory origins. The
two sides of the debate on the contemporary articulation of 〈nh〉 are joined on the one
hand by Perini (2002) and on the other by Azevedo (1981). Perini (2002) states that
 Ryan Shosted & José Ignacio Hualde

the phoneme is realized as a palatal nasal glide [J̃] while Azevedo (1981) refers to the
same consonant as a palatal nasal occlusive [\].1 Lipski (1975:€ 75) has commented
briefly on the articulation of the sound in BP, noting that it frequently “reduce[s] to a
mere nasalized glide”. In addition, Hajek (1997:€168) writes that the palatal nasal “may
be variably deoccluded in some Portuguese dialects”.
In articulatory terms, the difference between the two articulations is slight, having
to do with the degree and perhaps duration of linguopalatal contact. Variation in the
realization of 〈nh〉 among BP speakers has not been examined systematically2 nor is
it presently clear whether the difference between the two articulations is perceptually
salient in any linguistic sense. We do, however, have our own anecdotal evidence that
some native BP speakers acquiring Spanish have difficulty producing the palatal nasal
stop of that language, opting for the palatal nasal glide instead (which may result in a
foreign accent in Spanish).
Whether BP’s phonemic palatal nasal is realized as a nasal stop3 or a glide relates
to the larger question of diachronic merger and incomplete neutralization because
Portuguese 〈nh〉 is a reflex of different Latin sequences (see Section 1.2), only some of
which gave rise to the phonemic palatal nasal of Spanish (spelled 〈ñ〉).4 While there may
be some debate over the phonetic realization of BP’s palatal nasal, the Spanish palatal
nasal is uncontroversially characterized as a fully occlusive nasal stop (e.g. Hualde
2005:€173), although a tendency to neutralize this segment with a sequence of alveolar
nasal stop followed by glide has been noted for some dialects (Colantoni & Kochetov
2010).5 As we will see, both Portuguese and Spanish also have exclusive sources for
their palatal nasal consonant(s). We will argue that stop and approximant realizations
of the palatal nasal may have coexisted in Portuguese at some time with different lexical
distributions (Williams 1975) and that both articulations may have persisted up to the
present day, albeit as allophonic variants.
Hence, a number of questions both practical and theoretical emerge: (1) Do reali�
zations of the phonemic palatal nasals of Spanish and Portuguese (both BP and EP)

.╅ See Cruz-Ferreira (1997) for the same account in European Portuguese (EP). While we
find these discrepancies worthy of attention, these authors do not convey any controversy
surrounding the articulation of the palatal nasal in either variety of Portuguese.
.╅ By contrast, the articulation of the palatal nasal in Catalan has been researched thoroughly,
demonstrating, e.g. that it is a simple segment composed of one lingual gesture, unlike the
palatalized alveolar /nj/, composed of two (Recasens & Romero 1997).
.â•… For reference to nasal consonants as “stops” see e.g. Reetz & Jongman (2009:€11, 15).
.╅ As we will see, both Portuguese and Spanish also have exclusive sources for their palatal
nasal consonant(s).
.╅ Deocclusivization has been reported in a couple of local varieties of Cuban Spanish, as well
as in Equatorial Guinea, where Spanish is spoken as a second language (Quilis 1993:€243).
 Palatal nasals in Portuguese and Spanish 

differ in articulatory terms? (2) Can various historical origins account for various
realizations of the phonemic palatal nasal in Portuguese (i.e. is there evidence of
incomplete neutralization)? (3) Is there something about the articulation of palatal
nasal stops that makes them more vulnerable to de-occlusivization than alveolar or
bilabial nasals? Our hope is that an instrumental study of this sound’s articulation will
lead to a clearer hypothesis about its origins in Galician-Portuguese (G-P), including
any intermediate steps in its evolution. We also hope to contribute to a better under-
standing of the phonological patterning and historical development of the palatal
nasal in other languages.

1.2â•… The path to the palatal nasal in Portuguese and Spanish

Latin did not have a palatal nasal consonant. The palatal nasal consonant(s) of modern
Portuguese, Galician, and Spanish arose from four etymological sources attested in
Latin. For convenience we will refer to these as Classes I–IV (Table 1). Class I and II
palatalizations are common to Spanish, Portuguese, and Galician. Of the three lan-
guages we are concerned with, the Class III palatalization occurred exclusively in
Portuguese and Galician while the Class IV palatalization occurred only in Spanish.
One source (Class I) of Portuguese 〈nh〉 and Spanish 〈ñ〉 is a sequence of alveolar
nasal plus palatal glide (from the gliding of Latin i or e before another vowel): vinea€>
Port. vinha, Sp. viña “vineyard”, senior > Port. senhor, Sp. señor “sir”). Despite this
change, which would seem to remove any nasal plus palatal glide sequences, palatalized
alveolar nasals do exist in both modern Portuguese and Spanish. Forms like Port.
gardênia, Sp. gardenia “gardenia” entered the language after the palatalization of Class
II words had already occurred. Their phonetic realization is uncontroversially [nj]
(if€the following /i/ does not receive lexical stress). The phonemic palatalized alveolar
thus stands in contrast to the phonemic palatal nasal, e.g. in Port. insânia “insanity” vs.
sanha “fury”, Sp. uranio “uranium” vs. huraño “unsociable”.
Another origin (Class II) of Portuguese 〈nh〉 and Spanish 〈ñ〉 is the Latin sequence
-gn- or -ng-, e.g. pugnu > Port. punho, Sp. puño “fist” and ungula > Port. unha,
Sp. uña “fingernail”.6

.╅ The phonetic bases of the Class II palatalization deserve greater attention than we can
provide here. The -ng- sequence would normally result in a velar nasal rather than a palatal
(particularly if followed by a non-front vowel, as in ungula). The divergent formant transitions
associated with velar and palatal places of articulation are well known (Fant 1960) making
a listener-based error improbable as the source of the sound change. However, Williams
(1975:€94) proposes that the change only occurred before front vowels, a context where /g/
systematically palatalizes, and moreover that in Portuguese these were Spanish borrowings,
e.g. ringire > (Sp. *[renjLir] >). [re\ir] reñir > Port. renhir “dispute, contend”. For forms like
ungula, without a front vowel, the standard account is that the lateral had previously palatalized
 Ryan Shosted & José Ignacio Hualde

Class III words gave rise to palatal nasals in G-P (but not Spanish) and involved
original sequences where an alveolar nasal was preceded by the palatal vowel /i/ and
followed by another vowel. Examples include gallina > galinha “hen”, molinu€ >
moinho “mill”, regina > rainha “queen”, caminu > caminho “road”, vinu > vinho “wine”,
vicinu > vizinho “neighbor”, etc. vs., Sp. gallina, molino, reina, camino, vino, vecino,
with an alveolar nasal as in Latin. The simplest hypothesis for this development is that
as the alveolar nasal stop occlusion weakened, the resulting nasal approximant was
colored by the preceding palatal vowel, e.g. vinu > [vĩjo]. To the extent that the nasal
approximant subsequently “hardened”7 to a palatal nasal stop it would be by con-
vergence with the palatal nasal stop that arose in other contexts (Williams 1975:€71).
A parallel development can be observed in the feminine article, where assimilation has
produced a bilabial nasal: una > *ũw̃↜渀屮↜a > Port. uma “one.fem” (in Galician we find a
velar nasal instead, /u]a/).
In other Portuguese VNV sequences, the nasal consonant has disappeared leaving
at most vowel nasalization as its only trace.8 Examples include lana > lã “wool”, manu€>
mão “hand”, luna > lua “moon”, bona > boa “good.fem.sg”. Exceptionally, we also find
〈nh〉 in the word mea > minha “mine, fem”, by progressive nasalization.
Teyssier (1980:€35–36, 51) notes that the hiatus of nasal and oral vowels was “by
nature very unstable”; accordingly it was resolved by either excrescence (dégagement)
of a nasal consonant or by oralization of the first vowel.
Spanish 〈ñ〉 arose from one Class of words which did not give rise to 〈nh〉 in
Portuguese. We will categorize these as Class IV etyma. Here, the intervocalic geminate
-nn- palatalized in Spanish but merely degeminated in G-P, where as we just saw,
singleton -n- was lost: annu > Sp. año, Port. ano “year” cf., manu > Sp. mano, Port. mão
“hand” (Penny 2002:€71, for Spanish). The change in point of articulation in Spanish,
from alveolar to palatal, may be attributed a larger or more palatalized contact area in
the production of a geminate (Payne 2006).
The modern reflexes of the four types of Latin etyma are summarized in Table 1.
One class of words uniquely gave rise to a phonemic palatal nasal in G-P and Spanish
(Class III in G-P; Class IV in Spanish). All languages share Classes I and II. Based
on the absence of Class III palatal reflexes in Spanish (which uniquely gave rise to
�nasalized glides in G-P) it is perhaps unsurprising that no one claims the existence of

by analogy with singulariu > senlheiro (ibid.). Palatalization of the -gn- sequence in words
like pugnu > punho is similarly perplexing from an acoustic-perceptual standpoint.
.╅ We borrow the term from Hajek (1997) who uses it to describe the development of Bolog-
nese palatals. Ferreiro (1999) has suggested that the emergence of the palatal consonant was
somehow geared towards preventing vowel hiatus, referring to it as “[…]a consoante palatal
anti-hiática” (39, fn. 149) (our thanks to Fernando Martínez-Gil for pointing this out).
.â•… Nasal vowels were later lost in Galician (García de Diego 1984).
 Palatal nasals in Portuguese and Spanish 

Spanish nasalized glides. At no stage in the development of the Spanish palatal nasal
has /↜J̃/ been posited.

Table 1.╇ Etymological classes that gave rise to a phonemic palatal nasal in Portuguese,
Spanish, and Galician. Starred forms represent classes without a phonemic palatal
nasal reflex

Etymological Class Portuguese Spanish Galician

Class I: -nj- vinea > vinha > viña > viña

Class II: -gn-, -ng- pugnu > punho > puño > puño
Class III: -in- vinu > vinho > vino* > viño
Class IV: -nn- annu > ano* > año > ano*

For Class III (vinu) reflexes in Portuguese, the original alveolar nasal consonant
was deleted after nasalization of the high vowel had occurred (Sampson 1999). The
palatal nasal stop reemerged or “hardened” thereafter and has until lately gone unchal-
lenged as the canonical realization of this phoneme (Azevedo 1981; Cruz-Ferreira
1997). In articulatory terms, this so-called “hardening” means that linguopalatal contact
increased in degree and/or duration. If we are to believe that hardening occurred at
some point, along with reports that the palatal approximant articulation is currently
favored (Perini 2002) then the lenition /\/ > /↜J̃/ must be posited, as well.
We offer an alternative, more parsimonious account9 whereby the merger of early
G-P Class III nasal approximants with palatal nasal stops from Class I–II is still incom-
plete, with perhaps substantial variation between the two articulations (although there
is no indication that the etymological lexical distribution of these two articulations has
been preserved). We submit that Class I and II forms developed a palatal nasal stop
(as evidenced by Spanish) early in Ibero-Romance and that only afterwards did G-P
develop the nasalized glide in Class III words. We claim that this sound is still present
in Portuguese phonology, particularly in Brazil, where it was perhaps transported in the
15th and 16th centuries, before the palatal nasal stop was settled upon as the preferred
articulation by speakers of EP (and presumably Galician). Thus, we argue that the
palatal nasal approximant with reduced linguopalatal contact in BP is a conservative
linguistic trait of Portuguese rather than a New World innovation.
However, our account does not assume that Class I and II reflexes were also realized
as nasal approximants at an earlier stage in Portuguese. Instead, we believe that leni-
tion of Classes I and II (/\/ > /↜J̃/) has resulted in at least a partial merger with the
Class III nasal approximant /↜J̃/. In other words, we propose that at some stage there
was a phonemic contrast between Class III /↜J̃/ on the one hand and Class I–II /\/ on

.╅ Besides its parsimony, our account has the additional advantage of not singling out the palatal
nasal (to the exclusion of other nasal consonants) as a target of lenition on an ad hoc basis.
 Ryan Shosted & José Ignacio Hualde

the other (as Williams 1975 also claims). The contrast was later lost (perhaps very early
on), but [J̃] and [\] have both survived as allophonic variants in BP.
According to Williams (1975), the distinction between nasal approximant and
stop is found in the 14th-century Cancioneiros (collections of G-P troubadour poetry)
where Class I and III reflexes do not rhyme (Williams 1975:€82). Speakers in the Middle
Ages produced the reflexes of e.g. vinu and vinea with two different nasal consonants,
the approximant for vinu (Class III) and the stop for vinea (Class I). Soon after-
wards, however, both the palatal consonants of Class I and Class III were written as
〈ĩ〉 or 〈 ỹ 〉 in Medieval manuscripts, sometimes alternating with 〈nn〉, 〈ñ〉 suggesting
that the sounds were quickly confounded (Maia 1986:€489–490, García de Diego 1984
[1909]: 57). The collapse of the Class I/III distinction could have come about for two
reasons: either Class III “hardened” to /\/ or Class I “softened” to /↜J̃/. A third pos-
sibility is that the merger is still only partial and both [\] and [↜J̃] exist today in some
as-yet-undetermined variation.10
Because the synchronic facts of BP nasal palatal occlusion are largely unknown and
because instrumental data in this area are scarce, our goal is to collect evidence through
a study examining acoustic and aerodynamic differences between BP 〈nha〉 (palatal)
and 〈nia〉 (palatalized alveolar followed by glide) sequences, contrasting these with
Spanish 〈ñ〉 and 〈nia〉. Because of limitations inherent to the present methodology,
we intend to follow up with a comprehensive electropalatographic (EPG) study of the
sounds in the future.

2.â•… Methods

2.1â•… Speakers
Three speakers of Puerto Rican Spanish (PRS; P1–3) and three speakers of BP (two
from São Paulo (B2 and B3), one from Rio de Janeiro (B1)), aged between 25 and

.╅ There is some dialectal evidence that European Portuguese (EP) has developed and
even reinforced a fully occluded palatal nasal stop: palatal [\] may still vary with alveolar [n]
in certain villages of the Algarve, including Santa Luzia and Monte Gordo (Hammarström
1953:€ 176). More modern evidence on this point has yet to be uncovered. An anonymous
reviewer has suggested two other relevant differences between EP and BP. The word camião
“truck/lorry” comes from French camion, so no intervening palatal nasal stage may be posited.
In EP hiatus or an optional rising diphthong separates 〈i〉 and 〈ão〉. In BP either a palatal
nasal stop or a palatal nasal glide intervenes. (This may, however, be a case of folk etymology
whereby speakers form an augmentative based on caminho “path”.) The second case is the
realization of nasality in words like põem “put.3pl.” and têm “have.3pl.” In EP these words
have two diphthongs while in BP there is only one, such that, e.g. têm and tem “have.3sg” are
non-contrastive. These observations make it clear that our study could be usefully extended to
an examination of differences between EP and BP, as well.
 Palatal nasals in Portuguese and Spanish 

30€participated in the study. The Brazilians had all been in the continental United States
for less than three years at the time of the experiment, as had the speakers of PRS.
As we mentioned above, some Spanish dialects, including Buenos Aires Spanish
(Quilis 1993:€243, Colantoni & Kochetov 2010), show a tendency to merge the palatal
nasal and the sequence alveolar nasal plus palatal glide so that the contrast between
uranio “uranium” and huraño “unsociable” or unión “union” and uñón “big fingernail”
is lost or unstable. This tendency has not been reported for Puerto Rican Spanish,
which therefore seemed like an appropriate choice for our study.

2.2â•… Materials
We set out to build a list of natural tokens in Portuguese and Spanish that included
the phonemic palatal nasal (whether stop or approximant), palatalized alveolar stop,
and palatal approximant in comparable positions (e.g. Port. canha “left hand”, Tânia
“(personal name)”, saia “skirt”). We selected [a] as the word-final vowel and then
sought to balance the quality of the preceding vowel in the VCa sequences between
[i e a o u] (e.g. Port. minha “my.fem”, manha “cunning”, grenha “mop of hair”) using
reverse dictionaries (Bosque & Pérez Fernández 1987; d’Andrade Pardal 1993). We were
able to complete sets of five (5) -aCa, -oCa, and -eCa tokens (where C ranged between
the palatal approximant, palatalized alveolar and palatal nasal; = 45 tokens) with addi-
tional sets of five (5) -iCa and -uCa forms (where C was limited to the phonemic palatal
nasal; = 10 tokens). Portuguese and Spanish cognates were used whenever possible.
In total, 55 lexical items were chosen. The tokens are listed in Appendices A and B.
During the aerodynamic recordings (see Section 2.3) a partial list was inadvertently
presented to speaker B1 who therefore pronounced only 46/55 tokens (five 〈nh〉 three
〈i〉, and one 〈ni〉 token are missing). Items not pronounced by B1 are marked with an
asterisk in Appendix A.

2.3â•… Procedures
Tokens were repeated three times in randomized order in the carrier phrase, Digo ____
para ti “I say ___ to you” (the same phrase worked for both languages with only
minor, and for our purposes irrelevant, differences in pronunciation of the non-test
material). Audio was recorded using an AKG c520 head-mounted condenser micro-
phone in an audiometric booth at the Illinois Phonetics & Phonology Laboratory.
The audio recordings were made at a sampling rate of 44.1 kHz using a Marantz
PMD570 solid-state recorder and a Grace preamplifier. The audio data were labeled
and measured in Praat 5.0.03. On a separate occasion, tokens were recorded using a �Glottal
Enterprises oral flow mask fitted with a heated Fleisch pneumotachometer (Fleisch
1925; Farquharson & Anthony 1970:€ 813). Cannulae from the pneumotachometer
head were attached to the leads of a Biopac TSD-160 differential pressure transducer.
Resulting signals were digitized using software developed for the Signal Processing
 Ryan Shosted & José Ignacio Hualde

Toolbox in Matlab 7.0.1. Airflow signals were labeled and measured using routines
written by the first author for use in Matlab R2007b (7.5.0.342).

2.4â•… Measures
Our primary interest lies in the occlusive characteristics of the BP phonemic palatal
nasal consonant: whether it is realized as [\] or [J̃]. Based on the assumption that
greater oral impedance will lower the intensity of the sound radiated from the vocal
tract, we sought to measure the difference in acoustic intensity between the test con-
sonants (palatal nasal, palatalized alveolar nasal, and palatal glide) which have varying
degrees of oral impedance. We could not determine beforehand whether a BP palatal
nasal would be realized as a stop or an approximant. To the extent that the palatal nasal
patterns with /j/, we might say that it is an approximant (low oral impedance); to the
extent that it patterns with /nj/ we may call it a stop (high oral impedance).
Measurements of relative intensity as an acoustic correlate of degree of constric-
tion have been used by Cole et al. (1999), Soler & Romero (1999), Ortega-Llebaria
(2004) and Carrasco (2008), among others, for Spanish /b d g/. The same has been
done using oral flow (Shosted & Willgohs 2006).
Acoustic intensity curves were generated for the test tokens using standard set-
tings11 in Praat 5.0.03 (Boersma & Weenink 2008). Next, the intensity curves corres�
ponding to [VCap] (e.g. the underlined portion in the Portuguese phrase: digo senha
para ti “I say password to you”) were extracted and analyzed. The depression in intensity
associated with the test consonant was measured with respect to the following vowel [a].
We measured the maximum intensity value during [a] and the minimum value during
the test consonant. Figure 1 shows how these measures were taken in three BP tokens
demonstrating the three test consonants. We anticipated that the depression would be
deepest for an occluded test consonant due to increased oral impedance. We call this an
“over-the-curve” (OTC) measure. Where OTC is the difference between the maximum
intensity of [a] and the minimum intensity of the test consonant, OTCstop > OTCglide.
One limitation of our acoustic method is that nasal sound emission may com-
pensate for reduced oral sound emission in nasal consonants. This will be discussed
in greater detail in Section 4.2. To validate results based on acoustic measures we also
recorded oral flow. As above, oral flow during the vowel following the test consonant
was taken as a maximum and the difference between this value and the minimum
during the consonantal depression was measured. This resulted in a measure compa-
rable to OTC acoustic intensity difference, i.e. OTC oral flow difference.

.╅ Minimum pitch = 100 Hz; Time step = 8 ms; mean pressure subtracted (cf. Praat 5.0.03
online manual, “Sound: To Intensity…”).
 Palatal nasals in Portuguese and Spanish 

5 47
Frequency (kHz)

Intensity (dB)
2.5 44.5

0 42

a \ a p

min max

0 0.398
Time (s)

5 47
Frequency (kHz)

Intensity (dB)
2.5 44.5

0 42

a n ja p

min max

0 0.3581
Time (s)

5 60
Frequency (kHz)

Intensity (dB)

2.5 57.5

0 55

a j a p

min max

0 0.398
Time (s)

Figure 1.╇ Examples of annotated tokens from BP Speaker B1: (top) banha ‘bathe.3p.sg’;
(middle) insânia ‘insanity’; (bottom) saia “skirt” (stop closure from the next word in the
carrier phrase is included for reference). The position of minima during the test consonant
and maxima during the following vowel were collected through custom-written scripts
in Praat 5.0.03 once consonantal boundaries had been demarcated
 Ryan Shosted & José Ignacio Hualde

2.5â•… Statistical tests

Data were fitted to linear mixed-effects models with speaker as random effect (using
the NLME package’s LME function in R 2.8.1; Baayen 2008). The models were submit-
ted to repeated measures ANOVA with measures from each speaker’s three word rep-
etitions averaged. The independent variable was test consonant and the acoustic and
airflow measures were dependents. Repeated measures ANOVAs were also conducted
with language as independent variable and acoustic and airflow measures of the palatal
nasal as dependent variables, using a generalized linear model (R’s GLM function).
Tukey HSD tests for significant differences between consonants were implemented
using the Multcomp package’s GLHT (General Linear Hypotheses) function in R.

3.â•… Results

Boxplots for acoustic intensity (Figure 2) and oral flow (Figure 3) show differences
between consonants for each speaker. Larger values for these measures may be inter-
preted as a greater degree of oral impedance. Details of the statistical models, including

PRS
P1 P2 P3
OTC Int diff (dB)

15
10
5
0
j nh nj j nh nj j nh nj

BP Palatal Nasal
B1 B2 B3
OTC Int diff (dB)

5
OTC Int diff (dB)

5 4
4 3
3
2 2
1 1
0 0
j nh nj j nh nj j nh nj BP PRS

Figure 2.╇ Boxplots of OTC acoustic intensity differences (between the test consonant and
following [a]) for consonants in PRS (right) and BP (lower left). Boxplots of acoustic intensity
differences for the palatal nasal in both languages appear at lower right. A higher value is to be
interpreted as greater degree of oral impedance
 Palatal nasals in Portuguese and Spanish 

significant differences between test consonants, are given in Tables 2, 3, and 4. For
convenience and consistency in the results section we will represent the palatal nasal
consonants of both Spanish and Portuguese with the digraph nh, the palatal glide with
j, and the palatalized alveolar with nj.

PRS

OTC OF Diff (ml/s)

P1 P2 P3
600
500
400
300
200
100
j nh nj j nh nj j nh nj

BP
Palatal Nasal
OTC OF Diff (ml/s)

B1 B2 B3
OTC OF Diff (ml/s)

600 600
500 500
400 400
300 300
200 200
100 100

j nh nj j nh nj j nh nj BP PRS

Figure 3.╇ Boxplots of OTC oral flow (OF) differences (between the test consonant and
following [a]) for consonants in PRS (right) and BP (lower left). Boxplots of oral flow
differences for the palatal nasal in both languages appear at lower right. A higher value is
to be interpreted as greater degree of oral impedance

Table 2.╇ ANOVA results for PRS and BP based on a linear mixed effects model with test
consonant as fixed effect, speaker as random effect, and measure as dependent variable.
*** = p < 0.001, ** = p < 0.01, * = p < 0.05, n.s. = not significant; nDF = numerator
degrees of freedom; dDF = denominator degrees of freedom

Language Measure nDF dDF F p(F)

PRS Oral Flow 2 160 93 ***

Intensity 2 160 150 ***
BP Oral Flow 2 151† 18 ***
Intensity 2 160 7 **
† There are only 156 observations in this model. Recall that nine tokens were missing from Speaker
B1’s data. The value for denominator degrees of freedom (dDF) is calculated by subtracting the
number of groups in fixed effects (Consonant = 2) and random effects (Speaker = 3) from the number
of observations. For recording sessions that went as planned there are 55 × 3 = 165 observations and
therefore a dDF of 160.
 Ryan Shosted & José Ignacio Hualde

Table 3.╇ ANOVA results for the palatal nasal consonant based on a generalized linear
model with language as independent variable and measure as dependent

Consonant Measure nDF dDF F p(F)

Palatal nasal Oral Flow 1 143‡ 0.01 n.s.

Intensity 1 148 58 ***
‡ There are 145 observations for this model. Recall that five nh tokens were missing from Speaker B1’s
data. Since speakers were supposed to produce 25 nh tokens the total number of observations is
(25 × 6) – 5 = 145 for Speaker B1. Appendix A marks with an asterisk those tokens not produced by B1.

Table 4.╇ Statistically significant differences (Tukey HSD) between test consonants in
terms of two measures (oral flow difference and acoustic intensity difference). In general
“x > y” may be read, “x is more occluded than y”

Language Measure j–nh j–nj nh–nj

PRS Oral Flow j > nh *** j > nj *** nh > nj ***

Intensity j > nh *** j > nj *** n.s.
BP Oral Flow j < nh *** j < nj *** nh < nj *
Intensity n.s. j < nj ** nh < nj ***

In general, the oral flow measure corroborated and refined the results of the
acoustic intensity measure. We find, for example, that while no difference between
j and nh was uncovered for BP using acoustic intensity alone, it can be argued that
nh is more occluded than j by reference to oral flow. In general, we can say that j
is the most occluded of the three consonants in PRS, more so than both nh and nj.
Furthermore, In PRS, nh is more occluded than nj. In BP, the opposite is true: nj is
more occluded than nh and nh is more occluded than j. This is not true for speaker B3,
for whom the difference between nh and nj in terms of both aerodynamic and acoustic
measures failed to achieve significance at an alpha-level of 0.05.

4.â•… Discussion and conclusions

4.1â•… Summary and significance of results

If intensity as measured here is taken as an acoustic correlate of occlusion, we may
state the following: (1) PRS /\/ is realized with an occlusion comparable to that of the
PRS palatalized alveolar /nj/; (2) BP /\/ is less occluded than BP’s palatalized alveolar
/nj/ while sharing characteristics of occlusion with the palatal approximant; (3) PRS /\/
is more occluded than BP /\/. Finding (2) is confirmed by the more direct correlate of
oral airflow while findings (1) and (3) are not. Analysis of oral flow suggests that PRS /\/
is actually more occluded than /nj/ but fails to suggest that there is a significant differ-
ence between PRS /\/ and BP /\/.
 Palatal nasals in Portuguese and Spanish 

Confirmation of finding (2) using both acoustic and aerodynamic evidence sup-
ports Perini’s (2002) assertion that BP may have a palatal nasal glide /↜J̃/ instead of a
palatal nasal stop /\/. We observe that for one of three BP speakers (B3) this was not
true. We cannot ascribe this to dialectal variation, since B3 is one of two speakers from
São Paulo. B2, the other paulista, also seems to produce a nasal approximant.
Our results are consistent with the hypothesis that the BP palatal nasal approximant
is a conservative linguistic trait passed down from medieval G-P. While it is indeed
Â�possible that the sound underwent a “round-trip” change from palatal glide to stop back
to glide, the evidence that all Portuguese palatal nasals were at some point realized as
full palatal stops (the intermediate stage) is supported, so far as we can tell, by assump-
tion only. We posit a simpler, more elegant solution that is grounded in experimental
evidence. Of course, further study is necessary to corroborate the physiological validity
of this hypothesis and judge the diffusion of the sound in the Lusophone world.
The high OTC acoustic intensity and OTC oral airflow differences associated with
PRS /j/ bespeak the fact that this phoneme is often realized as a fully occluded [j] or
[jë N] (Hualde 2005:€165–172). For our PRS speakers, the realization of /j/ varied unpre-
dictably between occlusive and approximant but with enough occlusive realizations
to raise the OTC acoustic intensity substantially (i.e. indicative of greater oral imped-
ance). This phenomenon also varied between speakers, with P1 producing the most
frequent occlusions.
Initially it seems puzzling that PRS /j/ should have a higher degree of impedance
than /nj/, since both presumably have equivalent degrees of oral occlusion. If the sta-
tistically significant differences for the PRS occlusives /j/ and /nj/ are in error this of
course calls into question our results regarding differences between BP /nj/ and 〈nh〉.
For now, we present some alternative explanations for these results and plan further
articulatory evidence to shore up our case. For the acoustic measures of PRS /j/ and /
nj/ we might explain the discrepancy by citing the fact that the intensity under obser-
vation is derived from both oral and nasal sound pressure. During a nasal consonant
the intensity depression occasioned by oral occlusion is mitigated by the nasally-emitted
sound wave (cf. the different intensities of voiced stops and nasals). Ideally, oral flow
could resolve the issue, insofar as it truly eliminates the ambiguity caused by nasal
acoustic emission. Perplexingly, however, the results of the oral flow measure mirror
those of the intensity measure: /j/ is still more occluded than /nj/ in PRS. Because
the airflow results corroborate the patterns seen in the acoustic results, the discrep-
ancy between /j/ and /nj/ can be explained only by reasoning that PRS /nj/ actually
manifests less oral occlusion than occluded /j/, a claim we did not intend to test at the
outset. We note, however, that intervocalic nasal lenition in Iberian Spanish has been
demonstrated (Honorof 2003) and may be operative in PRS or other varieties.
The airflow differences between /nj/ and /\/ in PRS may have an alternative expla-
nation. In Catalan and Russian, Recasens & Romero (1997:€43) observe a “transitional
event resulting from alveoprepalatal closure release for [\] occurring at the alveolar
 Ryan Shosted & José Ignacio Hualde

zone earlier than at the prepalatal zone.” If the tongue indeed peels away from the hard
palate in an anterior-to-posterior direction during the palatal nasal, the aerodynamic
consequence will be a slightly negative pressure in the anterior cavity as the volume
circumscribed by the mask increases. This will lower oral flow minima during the
palatal nasal consonant and thereby increase the measured differences. A “peel-back”
articulation might explain why /\/ appears to have greater oral impedance for PRS
Speakers P1 and P2 (though not P3). On the other hand, we have no reason to believe
that negative airflow can account for the observed differences in impedance between
BP /\/ and /nj/, since in this case /nj/ manifests greater impedance. Ultimately these
questions must be addressed by investigating the linguopalatal contact patterns and
lingual dynamics of the palatal nasal in dialects of Spanish and Portuguese. For nasal
consonants, at least, it appears that there is ample reason to be wary of both acoustic
and aerodynamic measures of oral impedance.
Ideally, the phonetic aspects of speech should be measured using a variety of
channels (e.g. acoustic, aerodynamic, articulatory, etc.) to ensure the reliability of the
data gathered and to avoid over-speculating on the implications of one’s experimental
findings. In the present study, we have used acoustic intensity and oral flow as proxies
for the characteristic we truly wish to measure: degree of oral impedance. The inference
that acoustic intensity may stand in for degree of impedance has been put to good use,
particularly in studying the differences between oral approximant and stop articula-
tions (Soler & Romero 1999; Cole et al. 1999; Ortega-Llebaria 2004; Carrasco 2008). To
our knowledge, ours is the first study to use acoustic intensity as a means of inferring
oral occlusion during nasal approximants and stops. However, here the circumstances
are somewhat different. To wit, there is a substantial limitation in our acoustic method
when applied to nasals: the acoustic energy sampled using a conventional microphone
setup is a combination of both oral and nasal output. Let us digress for a moment to
consider the implications of this method.
In the present study, we found, for example, that acoustic intensity was similarly
high for the oral palatal approximant /j/ and palatal nasal 〈nh〉 of BP. This brought
us to the conclusion that the palatal nasal of this language is likely an approximant.
However, any loss of acoustic energy in the oral cavity may be compensated by shunt-
ing air through the nasal cavity. Nasal stops have higher intensity than voiced oral
stops for precisely this reason. Thus, the explanation for the similar intensity of /j/ and
〈nh〉 may not stem from their similar oral occlusion (as we have inferred) but from
velopharyngeal port opening during /\/, which compensates for the high impedance
of the oral cavity.
While this methodological impediment necessarily restrains our conclusions, we
do not believe it invalidates them. This is particularly true of those conclusions that
are confirmed by using the more direct oral airflow method, which, though prone
to unique experimental artifacts, reassures us by replicating the intensity patterns.
Recall, moreover, that our experimental design allowed for the comparison of three test
 Palatal nasals in Portuguese and Spanish 

consonants: one oral and two nasals. If our conclusions were based merely on the com-
parison of an oral and a nasal consonant, methodological problems may have undercut
them indeed. However, we present acoustic evidence that two nasals differ significantly
from one another, e.g. in BP where the OTC acoustic intensity difference is higher for
/nj/ than for /\/, i.e. acoustic intensity is lower for /nj/ than for /\/ (see Figure€ 2).
While without physiological data we cannot be certain, we know of no reports sug-
gesting that velopharyngeal port size varies systematically with place of articulation
for nasal consonants. Hence, we must assume that the simplest explanation of the dif-
ference we have discovered is dependent on oral occlusion rather than increased nasal
sound emission. Similarly, we present evidence that the Portuguese nasal has greater
acoustic intensity than the Spanish nasal. We feel safe in regarding this difference as a
function of oral occlusion, not velopharyngeal aperture. Because of the inherent limi-
tations of acoustic speculation on articulation, we recommend that future studies
of this phenomenon use other physiological methodologies like EPG, for which we
present preliminary evidence in Figure 4.12 These data, collected from Speaker B1
using a WinEPG system (copyright 2005; Articulate Instruments Ltd.), confirm our
hypothesis that complete palatal closure does not obtain during the BP palatal nasal.
Further quantitative analysis of these data is forthcoming. We believe that it will also
be fruitful to consider familiarity and frequency effects for the tokens examined.

Tróia fronha acrimônia

Figure 4.╇ Linguopalatal contact configurations during /j/, /\/ (realized as [ ȷ̃]), and /nj/
(in the words Tróia “Troy”, fronha “pillowcase”, and acrimônia “acrimony”) produced by
Speaker B1. Shading is proportional to the absolute duration during which a particular
contact was maintained (darker shading indicates longer period of contact)

4.2â•… Phonological implications

In BP it has been observed that before the nasal consonants [m n] vowels must be
stressed in order to undergo allophonic nasalization (e.g. sino [sı̃´nu] “bell” but pinoia
[pinf́ja] “bad deal”) (Wetzels 2006:€ 28–29). However, Abaurre & Paggotto (1996)

.╅ In their report on the consonants and vowels of BP, Reis & Espesser (2006:€194, Fig. 15)
present a single palatogram of /\/ from the word canhada “a narrow plain between moun-
tains”. The postpalatal occlusion in the consonant is complete, though covering only a single
horizontal row of electrodes.
 Ryan Shosted & José Ignacio Hualde

found that before the palatal nasal, vowel nasalization was general and independent
of lexical stress. Wetzels (2006:€ 29) accounts for this asymmetry by reasoning that
palatal nasals occur in the syllable coda, since Portuguese tautosyllabic VN sequences
are said to result in vowel nasalization. In order to simultaneously place the palatal
nasal in the onset of the following syllable (obeying the Maximal Onset Principle, cf.
Blevins 1995 inter alia), Wetzels argues that BP /\/ must be geminate. Hence, canhoto
“left-handed” is not underlyingly /ka.\f́.tu/ but /ka\.\f́.tu/ with an ambisyllabic nasal,
otherwise /a/ would not surface as [ã].13 Wetzel’s analysis singles out the palatal nasal
as a unique nasal consonant in an ad hoc fashion. Our results suggest that the palatal
nasal can be singled out based on its phonetic content, viz. it is an approximant, not a
stop. We might expect a nasalized glide to influence the nasality of preceding vowels
more heavily by contributing to the length of the preceding nucleus.14 Indeed, a host
of experimental studies affirm that “the perception of vowel nasalization is favored by
increasing vowel duration,” a percept that may be more readily imparted by a nasal
approximant than a nasal stop (Hajek 1997:€89).

References

Abaurre, Maria Bernadete & Emílio Gozze Pagotto. 1996. “Nasalização no Português do Brasil”.
Gramatica do Português Falado VI ed. by Ingedore G. Villaça Koch, 495–526. Campinas:
Editora da UNICAMP.
d’Andrade Pardal, Ernesto. 1993. Dicionário inverso do português. Lisboa: Edições Cosmos.
Azevedo, Milton. 1981. A Contrastive Phonology of Portuguese and English. Washington, D.C.:
Georgetown University Press.
Baayen, R.H. 2008. Analyzing Linguistic Data: A practical introduction to statistics using R.
Cambridge: Cambridge University Press.
Blevins, Juliette. 1995. “The Syllable in Phonological Theory”. The Handbook of Phonological
Theory ed. by John A. Goldsmith, 206–244. Oxford: Blackwell.
Boersma, Paul & David Weenink. 2009. Praat 5.0.03.
Bosque, Ignacio & Manuel Pérez Fernández. 1987. Diccionario inverso de la lengua española.
Madrid: Gredos.

.╅ It has been argued that European Portuguese intervocalic oral glides may be ambisyllabic,
as well (Vigário & Falé 1993; Mateus & Andrade 2000).
.╅ An anonymous reviewer has correctly pointed out that our methodology does not allow
us to test whether a nasal approximant is followed by a nasal occlusive. We have measured
the minimum oral airflow and oro-nasal acoustic intensity which are at best indicative of
maximum oral occlusion only. We might assume that the transition from a vowel to a nasal
consonant will inevitably yield some short period in which the tongue is close enough to
the hard palate to produce a nasalized glide, given sufficient velopharyngeal opening. More
fine-grained observations of the length, nasality, lingual posture, and linguopalatal contact of
BP’s palatal nasal should come from EPG or EMA (electromagnetic articulometry) signals
sampled in sync with nasal flow.
 Palatal nasals in Portuguese and Spanish 

Carrasco, Patricio. 2008. An Acoustic Study of Voice Stop Allophony in Costa Rican Spanish. Ph.D.
dissertation, University of Illinois at Urbana-Champaign.
Colantoni, Laura & Alexei Kochetov. 2010. “Palatal nasals or nasal palatalization?” Paper
presented at LSRL 40, Seattle, WA, March 26–28, 2010.
Cole, Jennifer, José I. Hualde & Khalil Iskarous. 1999. “Effects of Prosodic Context in /g/ Lenition in
Spanish”. Proceedings of LP’89 ed. by Osamu Fujimura, 575–589. Prague: The Karolinum Press.
Cruz-Ferreira, Madalena. 1997. “Portuguese (European)”. Handbook of the International Phonetic
Association ed. by International Phonetic Association, 126–130. Cambridge: Cambridge
University Press.
Fant, Gunnar. 1960. Acoustic Theory of Speech Production. The Hague: Mouton.
Farquharson, I.M. & J. K. F. Anthony. 1970. “Research Techniques in Voice Pathology”. The Journal
of Laryngology and Otology 84. 809–817.
Ferreiro, Manuel. 1999. Gramática histórica galega. Santiago de Compostela: Edicións Laiovento.
Fleisch, A. 1925. “Der Pneumotachograph–ein Apparat zur beischwindigkeitregistierung der
Atem luft”. Pflügers Archiv für die gesamte Physiologie des Menschen und der Tiere 209. 713.
García de Diego, Vicente. 1984 [1909]. Elementos de gramática histórica gallega (= Verba,
Appendix 23). Santiago de Compostela: Universidade de Santiago de Compostela.
Hajek, John. 1991. “The Hardening of Nasalized Glides in Bolognese”. Certamen Phonologicum€II:
Papers from the 1990 Cortona Phonology Meeting ed. by Pier Marco Bertinetto, Michael
Kenstowicz & Michele Loporcaro, 259–278. Turin: Rosenberg & Sellier.
Hajek, John. 1997. Universals of Sound Change in Nasalization. Oxford: Wiley-Blackwell.
Hammarström, Göran. 1953. Phonetique auditive sur les parlers de l’Algarve. Uppsala: Almqvist€&
Wicksells.
Hay, Jennifer, Paul Warren & Katie Drager. 2006. “Factors Influencing Speech Perception in the
Context of a Merger-in-Progress”. Journal of Phonetics 34. 458–484.
Honorof, Douglas N. 2003. “Articulatory Evidence for Nasal De-occlusivization in Castilian”.
Proceedings of the XVth International Congress of Phonetic Sciences (ICPhS XV) ed. by
Maria-Josep Solé, Daniel Recasens & Joaquin Romero, 1759–1762. Barcelona: Universitat
Autònoma de Barcelona.
Hualde, José Ignacio. 2005. The Sounds of Spanish. Cambridge: Cambridge University Press.
Kilpatrick, Cynthia, Ryan Shosted & Amalia Arvaniti. 2007. “On the Perception of Incomplete
Neutralization”. Proceedings of the XVIth International Congress of Phonetic Sciences (ICPhS XVI)
ed. by Jürgen Trouvain & William Barry, 653–656. Saarbrücken: Universität des Saarlandes.
Lipski, John M. 1975. “Brazilian Portuguese Vowel Nasalization: Secondary aspects”. Canadian
Journal of Linguistics 20. 59–77.
Maia, Clarinda de Azevedo. 1986. História do galego-português: estado linguístico da Galiza e do
noroeste de Portugal desde o século XIII ao século XVI: com referência à situação do galego
moderno. Coimbra: Instituto Nacional de Investigação Científica.
Mateus, Maria Helena & Ernesto d’Andrade. 2000. The Phonology of Portuguese. Oxford: Oxford
University Press.
Ortega-Llebaria, Marta. 2004. “Interplay between Phonetic and Inventory Constraints in the
Degree of Spirantization of Voiced Stops: Comparing intervocalic /b/ and intervocalic /g/ in
Spanish and English”. Laboratory Approaches to Spanish Phonology ed. by Timothy L. Face,
237–253. Berlin: Mouton de Gruyter.
Payne, Elinor M. 2006. “Non-durational Indices in Italian Geminate Consonants”. Journal of the
International Phonetic Association 36. 83–95.
Penny, Ralph J. 2002. A History of the Spanish Language. 2nd Ed. Cambridge: Cambridge
University Press.
 Ryan Shosted & José Ignacio Hualde

Perini, Mário A. 2002. Modern Portuguese: A Reference grammar. New Haven, Conn.: Yale
University Press.
Quilis, Antonio. 1993. Tratado de fonología y fonética españolas. Madrid: Gredos.
Recasens, D. & Joaquín Romero. 1997. “An EMMA Study of Segmental Complexity in Alveo-
palatals and Palatalized Alveolars”. Phonetica 54. 43–58.
Reetz, Henning & Allard Jongman. 2009. Phonetics: Transcription, Production, Acoustics, and
Perception. Oxford: Wiley-Blackwell.
Reis, César & Robert Espesser. 2006. “Estudo eletropalatográfico de Fones Consonantais e
Vocálicos do Português Brasileiro”. Estudos da Lingua(gem) 3. 181–204.
Sampson, Rodney. 1999. Nasal Vowel Evolution in Romance. New York: Oxford University Press.
Shosted, Ryan & Beatriz Willgohs. 2006. “Nasals Unplugged: The aerodynamics of nasal de-
occlusivization in Spanish”. Selected Proceedings of the 2nd Conference on Laboratory
Approaches to Spanish Phonetics and Phonology ed. by Manuel Díaz-Campos, 14–21.
Somerville, Mass.: Cascadilla Proceedings Project.
Soler, Antonia & Joaquín Romero. 1999. “The Role of Duration in Stop Lenition in Spanish”.
Proceedings of the XIVth International Congress of Phonetic Sciences (ICPhS XIV) ed. by
John J. Ohala, Yoko Hasegawa, Manjari Ohala, Daniel Granville & Ashley C. Bailey, 483–486.
Berkeley, CA: University of California Berkeley.
Teyssier, Paul. 1980. Histoire de la langue portugaise. Paris: Presses Universitaires de France.
Vigário, Maria & Isabel Falé. 1993. “A sílaba do português: Uma descrição e algumas considerações
de ordem teórica”. Actas do IX Encontro Nacional da Associação Portuguesa de Lingüística,
465–478. Coimbra, Portugal: APL.
Warner, Natasha, Allard Jongman, Joan Sereno & Rachèl Kemps. 2004. “Incomplete Neutraliza-
tion and Other Sub-Phonemic Durational Differences in Production and Perception:
Evidence from Dutch”. Journal of Phonetics 32. 251–276.
Wetzels, W. Leo. 2007. “Primary Word Stress in Brazilian Portuguese and the Weight Parameter”.
Journal of Portuguese Linguistics 5: 1. 9–58.
Williams, Edwin B. 1975. Do Latim ao Português. 3rd Ed. Rio de Janeiro: Tempo Brasileiro.

Appendix A. Portuguese tokens

The nine tokens (five /\/, three /j/, and one /nj/ token) which Speaker B1 did not produce
are marked with an asterisk.

praia urânia cronha* cheia* Eugênia

saia* melânia fronha areia vinha
raia insânia monha lenha tinha
maia cizânia colônia penha minha
vaia jóia Polônia* grenha linha
manha cóia peônia senha pinha
sanha Tróia cerimônia brenha unha
canha sequoia* acrimônia ténia empunha*
banha* póia meia vénia cunha
aranha ronha* veia gardênia testemunha
România dronha* teia Armênia gadunha
 Palatal nasals in Portuguese and Spanish 

Appendix B. Spanish tokens

playa Alemania carroña bella Eugenia

saya insania zampoña sella viña
raya Romania bisoña leña tiña
maya tania colonia peña riñaguiña
vaya joya Polonia greña aliña
maña boya peonia seña uña
saña Troya ceremonia Breña Coruña
caña secuoya acrimonia tenia empuña
baña hoya plebeya venia acuña
araña roña epopeya gardenia pezuña
ucrania doña omeya Armenia
 Lenition and phonemic contrast
in Majorcan Catalan*

José Ignacio Hualde1, Marianna Nadeu1 & Miquel Simonet2

1University of Illinois at Urbana-Champaign/2University of Arizona

It has been claimed that in Majorcan Catalan /b/ tends to be realized as a stop in
contexts where Central Catalan /b/ has approximant realizations. Here we test
this claim using two continuous dimensions that we take as acoustic correlates
of degree of constriction. We also examine the relative degree of constriction of
Majorcan /b/ with respect to /d/ and /g/. We argue that Majorcan Catalan is less
advanced in the historical weakening of /b/ than other dialects. Against other
proposals, we maintain that a principle of contrast preservation is not required
in order to explain the phonetic facts.

1.â•… Introduction

In most Catalan varieties, like in Spanish, the voiced obstruent phonemes /b d g/ have
approximant realizations intervocalically and in many postconsonantal prevocalic
contexts (Mascaró 1984, among others). In a few other Catalan dialects, however,
plosive allophones, especially of /b/, are common in all contexts, including the inter-
vocalic position. Rafel (1976:€62) points out that those dialects where stop realiza-
tions of /b/ are found are precisely those that preserve an independent phoneme /v/.
These dialects are Balearic (including Majorcan, Minorcan and Eivissenc), parts of
Valencian, and Algherese, spoken in Sardinia (Recasens 1991). The contrast is also
alive to a certain extent, but receding, in the Camp de Tarragona area (Recasens
1971/75, 1985, 2009).
Wheeler (2005) appeals to a principle of preservation of contrast to explain why
/b/ fails to lenite in dialects with phonemic /v/, such as Balearic (see also Recasens
1991). According to Wheeler (2005:€320–321),

*We are grateful to Ryan Shosted, to Jesús Jiménez, to members of the audience at LSRL 39,
and to two anonymous reviewers for this volume for comments that have helped us improve
the quality of this paper.
 José Ignacio Hualde, Marianna Nadeu & Miquel Simonet

There can be little doubt that the much greater frequency of the bilabial plosive
[b] in lenitable contexts in Balearic reflects a principle of maintenance of contrast
between this phoneme and the voiced labio-dental fricative, a phoneme which is
lacking in the eastern continental variety.

Wheeler argues that the high rate of occurrence of [b] in contexts of lenition in
Majorcan cannot be attributed simply to higher conservatism, because the dental and
velar voiced plosives are less frequently realized as stops in these contexts.
A first objective of this paper is to obtain a more accurate view of the vari-
ation in the realization of /b/ in contexts of lenition in Majorcan and in Central
Catalan. Existing dialectological work is based on impressionistic transcription that
forces the transcriber to choose between one of two symbols: a stop or a fricative/
approximant. Such transcriptions, although valid as a first approximation, simplify
phonetic reality in nontrivial ways. Arguably, the spirantization of /b d g/ in lan-
guages like Catalan and Spanish involves a continuum of weakening. Realizations
of these phonemes across phonological contexts can range from very constricted to
vocalized realizations, rather than representing an alternation between two distinct
allophones, stop and approximant (or ‘fricative’), in complementary distribution.
In this paper, instead of classifying tokens as either stops or approximants, we use
two continuous acoustic features derived from the intensity curve, which we take
to be reliable correlates of the degree of constriction of the consonant (see Soler &
Romero 1999; Cole, Hualde€& Iskarous 1999; Lavoie 2000; Ortega-Llebaria 2004;
Carrasco 2008; Kingston 2008; Colantoni & Marinescu 2008; Eddington 2009, for
similar approaches).
In a first study we compare the degree of constriction of the consonants ortho-
graphically represented as b and v in Majorcan and in Central Catalan. In Majorcan
these two symbols correspond to two different phonemes /b/ and /v/, whereas in
Central Catalan these are two orthographic representations for the same phoneme /b/.
The hypothesis is, thus, that Majorcan Catalan speakers will produce more con-
stricted realizations of /b/ than Central Catalan speakers and that this will be
reflected in a reduction of acoustic energy in the portion of the sound wave corres�
ponding to the consonant.
In another study, limited to Majorcan and using a different corpus, we examine the
degree of constriction of /b/ relative to other segments, seeking to confirm whether in
fact /b/ is more constricted than /d/ and /g/ in this Catalan variety. As we will see, our
acoustic results by and large agree with previous impressionistic characterizations.
The second goal of this paper is to offer an explanation of the facts. Here we
depart from other authors and argue that no contrast preservation principle needs to
be invoked to explain the facts.
 Lenition and phonemic contrast in Majorcan Catalan 

2.â•… Majorcan vs. Central Catalan /b/1

2.1â•… Experimental methods and research hypotheses

For this study we have analyzed unscripted speech data from oral interviews that were
conducted for the Atles interactiu de l’entonació del català (Prieto & Cabré 2007–2010).
In particular, we have analyzed map task data from this source. A map task is a collab-
orative task in which a pair of native speakers participate. One of the participants has a
map with a marked route and must orally indicate how to draw this route to the other
participant, whose map only shows landmarks. From this corpus we have selected the
map tasks data of four towns in Majorca (Manacor, Pollença, Sineu, and Sóller) and
four towns from the Central Catalan dialectal area (Barcelona, Banyoles, Berga, and
Ripoll). We thus have data from a total of 16 speakers, two for each town. All speakers
are female, between 19 and 55 years of age (average age is€32.6 years).
From these recordings, we have segmented all instances of orthographic b and v
in traditional lenition contexts, namely between vowels, and after a liquid or /s/ and
before a vowel, both word-internally and across a word boundary. We did not include
tokens where b or v was utterance initial or was preceded by a nasal or a plosive.
Using the intensity display in PRAAT (Boersma & Weenink 2008), we have manu-
ally marked the minimum in the intensity curve within the duration of the consonant,
b or v, and the maximum during the following vowel. We have taken two measure-
ments, based on the intensity curve, as correlates of degree of constriction.
One measurement, “Intensity Difference” (IntensDiff) is the difference in deci-
bels between the target consonant and the following vowel. This is calculated by sub-
tracting the intensity minimum during /b/ or /v/ from the intensity maximum in the
portion corresponding to the following vowel. The more open the constriction, the
smaller the difference in intensity with respect to the following vowel is expected to
be. Although several other studies have used a measurement of intensity difference or
ratio in CV transitions (Soler & Romero 1999; Carrasco 2008; Colantoni & Marinescu
2008; Eddington 2009) or in a larger window (Cole, Hualde & Iskarous 1997; Ortega-
Llebaria 2004), our measurement adds certain controls and focuses exclusively in
measuring energy above the voice bar in the transition between the target consonant
and the following vowel. Intensity was calculated after applying a pass Hann band filter

.╅ The data that we present in this section (but not those in Section€3, which are from a
different corpus) are also discussed in Hualde & Nadeu (to appear), which is written in
Catalan. However, here we have completely redone the analysis, introducing what we believe
are important improvements both in the acoustic analysis and in the statistical treatment of
the results. In Section€4 we also expand on the argumentation in that earlier paper.
 José Ignacio Hualde, Marianna Nadeu & Miquel Simonet

between 500 and 10,000 Hz to the signal in order to exclude from the calculation of the
intensity curve possible effects of high frequency and low frequency noise. This filter-
ing procedure excludes the effects of the voice bar (F0). Exclusion of low frequency
sound is hypothesized to maximize the intensity difference between approximants and
more constricted voiced obstruents.
Our second measurement is “Maximum Rising Velocity” (MaxVel). This is based
on the procedure proposed in Kingston (2008), albeit with some important modifi-
cations. The intensity curve is extracted with the same settings as in our IntensDiff
measurement. After that, we calculate the first difference of the intensity curve in 0.001€s
steps. Finally, we extract the maximum value of the first difference that is found between
the intensity minimum corresponding to /b/ or /v/ and the maximum corresponding
to the following vowel. The maximum of the first difference indicates the maximum
rising velocity between the midpoint of the consonant and the midpoint of the fol-
lowing vowel. This measurement reduces possible effects of variation in the intensity
of the vowel and focuses on the abruptness of the transition between consonant and
vowel. More lenited consonants have a less abrupt transition irrespective of the time
from the intensity minimum to the intensity maximum or the absolute intensity values
of these two points.
Importantly, we deviate from Kingston’s (2008) procedure in not distinguishing
frequency bands. Kingston’s method includes extracting six frequency bands per token
and then three dependent variables per frequency band per speaker, which requires
3€×€6€=€18 statistical models per speaker. For a large number of speakers, like we have
in this experiment and in the one reported in Section€3, the number of statistical models
can be unmanageable, increasing the likelihood of chance results.
A first hypothesis concerns the relative intensity values of b and v in each of the
two dialects separately. The hypothesis is that Majorcan b and v will differ in their
relative intensity, since these are two different phonemes in this dialect. For Central
Catalan, on the other hand, no difference is expected, since these are two graphemes
for a single phoneme.
Our second, and more crucial, hypothesis is that Majorcan /b/ will also be more
constricted than Central Catalan /b/ (which corresponds to Central b and v).

2.2â•… Results
The IntensDiff and MaxVel results are plotted in Figure 1 and Figure 2, respectively.
In both figures, the left panel corresponds to the Majorcan data and the right figure to
the Central Catalan data.
The data were submitted to several Mixed Effects regression models (Baayen
2009; Johnson 2008) with IntensDiff and MaxVel as response, respectively. The only
fixed factor was a combination of consonant (b vs. v) and dialect (Majorcan vs.
Central), which resulted in a four-level factor: Majorcan b (Mb), Majorcan v (Mv),
 Lenition and phonemic contrast in Majorcan Catalan 

Majorcan Central

22

22
20

20
18

18
Intensity Differance

Intensity Differance
16

16
14

14
12

12
10

10

n = 128 n = 102 n = 67 n = 38
b v b v
consonant consonant

Figure 1.╇ Intensity difference of 〈b〉 and 〈v〉 in Majorcan and Central Catalan

Majorcan Central
1000
1000

900
900

Maximum Rising Veloctiy

Maximum Rising Veloctiy

800
800

700
700

600
600

500
500

400
400

n = 128 n = 102 n = 67 n = 38
b v b v
consonant consonant

Figure 2.╇ Maximum Rising Velocity of 〈b〉 and 〈v〉 in Majorcan and Central Catalan
 José Ignacio Hualde, Marianna Nadeu & Miquel Simonet

Central b (Cb) and Central v (Cv). This allowed for checking several pair-wise com-
parisons without sending the data to an initial two-factor analysis of variance and a
subsequent post-hoc analysis. The Mixed Effects model considered individual speaker
as a random factor. This is a very significant conceptual improvement over simple
linear regression models, on which traditional analyses of variance are based, since the
latter do not allow for the inclusion of random factors.
Regarding MaxVel as response, the model, with Mb (Majorcan b) as intercept
(i.e. the level against which other levels are contrasted in pair-wise comparisons),
revealed Mb (β€=€812.5) tokens to be significantly different from all other levels: Mv
(β€ =€ −262.6, t = −5.83, p < 0.001), Cb (β = −247.3, t = −2.53, p = 0.01), and Cv
(β€=€−256.1, t = −2.45, p = 0.01). In order to find out whether Cb and Cv differ from
each other, the Mixed Effects model was refitted with Cb as the intercept. The latter
model revealed that Cb (β = 565.1) was not different from Cv (β = −8.7, t = −0.12, ns)
or Mv (β = −15.3, t = −0.15, ns), but it was different from Mb (β = 247.3, t = 2.53,
p€=€0.01), as already revealed by the previous model. In sum, MaxVel data indicated
that Mb tokens differ from all other tokens while the other three levels do not differ
from each other.
Regarding IntensDiff, the model in which Mb was used as intercept revealed no
significant differences, although two of the pair-wise comparisons approached signifi-
cance at the 0.05 alpha level: Mb-Mv (β = −1.97, t = −1.86, p = 0.06), Mb-Cb (β€=€−3.86,
t = −1.83, p = 0.06), Mb-Cv (β = −3.53, t = −1.54, ns). A second model with IntensDiff
as response but Cb as the intercept also revealed no significant differences, although
the Cb-Mb comparison approached significance (β = 3.86, t = 1.83, p = 0.06). In sum,
the clear patterns revealed by MaxVel arise as mere tendencies when IntensDiff is used
as the dependent value. This might be due to the fact that, when acoustic measures
are used as correlates of constriction degrees in the comparison of alleged fricatives
(Mv), approximants (Cb, Cv) and voiced stops (Mb), measures based on rate of change
(MaxVel) are more reliable than those based on absolute intensity minima and max-
ima (IntensDiff).
Examples of spectrograms of Central Catalan and Majorcan /b/, showing different
degrees of constriction, are given in Figures 3 and 4.2

.â•… In Figures 3–7, the dark line represents the intensity curve expressed in decibels. In all
four spectrograms, the intensity minimum is set to 30 dB (the lowest point along the y axis),
and the maximum value is set to 100 dB. All four curves are represented within this range
(30–100 dB).
 Lenition and phonemic contrast in Majorcan Catalan 

6 ö a n s

0 0.4389
Time (s)

Figure 3.╇ Spectrogram of a token of abans “before” produced by a speaker of Central Catalan

s 6 b � b 6 ò i 6

0 1.037
Time (s)

Figure 4.╇ Token of sa barberia “the barber shop” produced by a speaker of Majorcan Catalan

To summarize, our results indicate that, as predicted, in Majorcan Catalan b is

more constricted than v (as revealed by the acoustic measurements that we have taken
as correlates of articulatory constriction), whereas there is no difference in constriction
between b and v in Central Catalan, where these are two orthographic representations
 José Ignacio Hualde, Marianna Nadeu & Miquel Simonet

for the same phoneme. The data also show that Majorcan /b/ is more constricted than
Central Catalan /b/ (orthographic b and v).

3.â•… Is /b/ more occluded than /d/ and /g/ in Majorcan?

3.1â•… Methods and hypotheses

As mentioned before, Wheeler (2005:€320–321) points out that, although all of /b d g/
in lenition contexts are transcribed as stops in Majorcan dialectological texts (Veny &
Pons i Griera 1998) much more often than in texts for other Catalan dialects, this is
particularly so in the case of /b/, which shows much higher rates of stop allophones
in contexts of lenition than /d/ and /g/ in such texts. Wheeler attributes this phonetic
behavior to the need to maintain the contrast between /b/ and /v/.
To investigate the degree of constriction of /b/ when compared to other relevant
segments, we have employed a second data source of spontaneous speech. This source
is a collaborative task in which two speakers engage in a game known as “20 questions”,
where one of the participants has to guess the identity of a celebrity by asking fewer
than 20 yes/no questions. From this corpus, which contains speech of 20 native speak-
ers of Majorcan Catalan (balanced for gender), we have segmented all tokens of /b/,
/d/, /g/ and /v/ as well as, for comparison, all tokens of intervocalic /p/, /t/ and /k/. As
before, IntensDiff (the difference in intensity between the target consonant and the fol-
lowing vowel) and MaxVel were calculated with PRAAT using a script. The hypothesis
would be that /b/ would show higher values for these measurements than /d/ and /g/,
indicating a more constricted realization. A total of 866 tokens have been analyzed.
We have excluded from the study three tokens of poble “village” and one of futbolista
“soccer player”, which contain geminate [bb].3

3.2â•… Results
The results are visually presented in Figure 5 and Figure 6 for IntensDiff and MaxVel,
respectively. The number of tokens analyzed for each consonant is also indicated in the
figures. The levels are organized from left to right in alphabetical order.
The validity of the energy measurements that we have taken is confirmed by
the fact that the voiceless stops /p t k/ have greater values for both measurements of
constriction than /b v d g/. It can be seen in the figures that, according to both mea-
surements, /b/ occupies an intermediate position between the voiceless stops /p t k/
and /v d g/, which is consistent with our hypothesis.

.╅ This corpus was used in Simonet (2008) for a study of intonational contours. The reader
is referred to this source for details.
 Lenition and phonemic contrast in Majorcan Catalan 

For the statistical analysis of the data, the results of the two intensity measure-
ments were fitted with Mixed Effects regression models (Baayen 2009; Johnson 2008),
with /b/ tokens as intercept (i.e. the other levels are compared to /b/ in pair-wise com-
parisons). Markov Chain Monte Carlo sampling was used to calculate the reported
p-values. Both individual speaker and word (i.e. lexical item) were used as random fac-
tors. Only Mixed Effects regression models allow for the use of several random factors,
which results in a very significant improvement over other statistical methods.
The statistical results revealed /b/ to be significantly different from all other con-
sonants in its MaxVel values, having significantly smaller MaxVel values than the
voiceless stops /p t k/, but significantly greater ones than the voiced obstruents /d g v/.
In particular, /b/ (β = 746.28) was significantly more occluded than /d/ (β = −267.62,
t€=€−4.98, p < 0.01), /g/ (β = −196.48, t = −2.97, p < 0.01) and /v/ (β = −242.75, t€=€−3.73,
p < 0.01), and less occluded than /p/ (β = 349.02, t = 5.95, p < 0.01), /t/ (β = 309.83,
t€=€5.57, p < 0.01) and /k/ (β = 302.19, t = 5.45, p < 0.01).
An identical pattern is revealed by the results of the model in which IntensDiff
was taken as the response or dependent variable. In particular, /b/ (β = 20.25) was
significantly more occluded than /d/ (β = −6.98, t = −4.78, p < 0.01), /g/ (β = −4.64,
t€=€−2.61, p < 0.01) and /v/ (β = −7.74, t = −4.39, p < 0.01), and less occluded than
/p/ (β = 8.7, t = 5.48, p < 0.01), /t/ (β = 3.34, t = 2.22, p = 0.02) and /k/ (β = 4.96,
t€=€3.31, p < 0.01).
30
25
Intensity Difference
15 10 20

n = 57 n = 265 n = 52 n = 144 n = 103 n = 171 n = 74

5

b d g k p t v
phone

Figure 5.╇ Intensity Difference between consonant and following vowel in Majorcan Catalan
 José Ignacio Hualde, Marianna Nadeu & Miquel Simonet

1200 1000
Maximum Rising Velocity
400 600 800
200

n = 57 n = 265 n = 52 n = 144 n = 103 n = 171 n = 74

b d g k p t v
phone

Figure 6.╇ Maximum Rising Velocity between consonant and following vowel
in Majorcan Catalan

Note that, while MaxVel values provided robust and interpretable results in both
experiments, IntensDiff values were only usable in the second experiment, since no ten-
dencies reached statistical significance in the first experiment (although the tendencies
were the same as for MaxVel). In the discussion of the results of the first experiment,
we suggested that acoustic approaches based on rate of change could be more reliable
than those based on absolute maxima and minima. It is possible that the larger amount
of data of the second experiment might have affected the IntensDiff values, making
them usable, or that the fact that two random factors instead of only one were used in
the statistical models might have improved the predictive power of the models.

4.â•… Against preservation of contrast as an explanatory factor

The results of our first experiment show that Majorcan Catalan intervocalic /b/ is
realized with significantly greater constriction than in other Catalan dialects, such as
Central Catalan, that do not have a /b/-/v/ contrast. The results of the second experi-
ment show that /b/ is also more occluded than /d/ and /g/ (and also than /v/). This is in
agreement with the frequent presence of phonetic [b] in dialectological transcriptions
of Majorcan Catalan texts in contexts where an approximant [β] is normally found in
 Lenition and phonemic contrast in Majorcan Catalan 

other dialectal variants. As mentioned in the introduction, Wheeler (2005) attributes

the presence of [b] in Majorcan Catalan in intervocalic and other contexts of lenition
to a principle of contrast preservation. The preservation of the contrast between the
phonemes /b/ and /v/ would act against the weakening of the plosive. This principle
of contrast preservation between the phonemes /b/ and /v/ as an explanation for the
resistance of /b/ to lenite is also mentioned by Recasens (1991). Ortega-Llebaria (2004)
also appeals to preservation of contrast with /v/ to explain why intervocalic /b/ under-
goes less lenition in English than in Spanish.
Our interpretation is somewhat different. Historically, intervocalic /b/ was a
stop in all Catalan dialects, as it was in Spanish. This phoneme /b/ contrasted with
either bilabial /β/ or with labiodental /v/, depending on the area. Thus, in the original
Castilian area where bilabials were the norm instead of labiodentals, intervocalic /b/
in cabeça /kabétsa/ “head” < capitia, sabe /sábe/ “s/he knows” < sapit, etc. contrasted
with phonemic /β/ in, for instance, cavallo /kaβáˆo/ “horse” < caballu, deve /déβe/
“s/he should, owes” < debet (see Alonso 1962:€189–190; Penny 2002:€72–74; Lapesa
1981:€39 n.27; Lloyd 1987:€237–241; but see Martínez-Gil 1998 for a different view).
Lenition of /b/ led to the loss of the contrast. The neutralization can be hypothesized
to have started in those contexts that favor the weakest realizations of /b/: in positions
of weak stress and between open vowels, spreading from there to other contexts. The
same process took place in areas of the Iberian Peninsula where /v/ was found instead
of /β/, which originally may have included most of the Catalan-speaking area, per-
haps excluding only varieties in contact with Aragonese. The difference in point of
articulation between [v] and [β] < /b/ may have acted to slow down the complete
merger between the two historical phonemes. Although in Catalan the only varieties
that have preserved [b] in intervocalic contexts are found in areas where this segment
contrasts with [v], it should be noticed that in the Castilian Spanish area a contrast
between /b/ and /β/ has been preserved up to the present day in Serradilla, Cáceres
(Ariza 1993 [1990, 1992]).
We may conclude that the more constricted realizations of /b d g/ that are found
in Majorcan Catalan simply reflect a more conservative stage in the evolution of the
language. Wheeler (2005:€320–321) also affirms that the greater frequency of stop allo-
phones of /b d g/ in Majorcan Catalan than in other varieties may have this cause, but,
as mentioned above, argues that the higher frequency of plosive realizations of /b/ is
due to a principle of maintenance of phonemic contrast.
If we consider the evolution of intervocalic obstruents in Western Romance in a
little more detail, it is clear that contrast preservation fails to explain the facts. Instead
there must be other reasons that explain why /b/ tends to show more resistance to
weakening than /d/ and /g/.
Stop and approximant configurations of voiced obstruents have been around for
almost two thousand years in Latin and Ibero-Romance, at some stages as allophones
 José Ignacio Hualde, Marianna Nadeu & Miquel Simonet

of the same phonemes and at other stages in phonemic opposition. Their evolution
shows little evidence of maintenance of contrast as a pertinent factor.
Regarding the labials, the first relevant change is the lenition of intervocalic /b/,
which subsequently results in merger with /w/. Whereas we cannot be sure when
/b/ started to admit approximant realizations or even when these approximant real-
izations of intervocalic /b/ became the usual articulatory target, we know that /b/
merged with /w/ by the first century of our era. This is because around this time the
graphemes b and v start getting confused (Allen 1978:€ 41). We also know that all
Romance languages have merged the results of the Classical Latin intervocalic -v-
and -b- phonemes. The details of this evolution appear to have been the following.
First, /b/ acquired approximant allophones in intervocalic position: [b] 〉 [β]/V__V,
as in habe↜渀屮re [abe:re]€> [aβe:re]. At this point, we would still have a contrast between
/b/ and /w/, since habe↜渀屮re [aβe:re] would contrast with lavaâ•›re [lawa:re]. Then the
contrast was lost, through another change affecting [w], which became labiodental,
[w] > [v], as in [lawa:re] > [lava:re], thus reducing the distance between the two pho-
nemes. A subsequent process was the merger between [β] and [v], which is found in
all Romance languages:

Latin Italian French Spanish

habe↜渀屮re avere avoir [aβer] haber, Old Sp. aver
lav↜↜are
 lavare laver [laβar] lavar

Eventually, thus, -v- and -b- merged in /v/ in French and Italian, and in all or most of
Catalan, and in /β/ in (northern) old Castilian Spanish.4 No Romance language has
preserved this contrast.
The next relevant change is the lenition of Latin voiced-internal intervocalic -p-
(as well as -t- and -c-) in Western Romance. In French this resulted in merger with
existing /-v-/ in word-internal intervocalic contexts. In Ibero-Romance, on the other
hand, -p- > [-b-], which at this point contrasted with the results of -v-, -b-. In Central
and Southern Portuguese, this contrast has been preserved as /-b-/ vs. /-v-/. In Castilian
Spanish as well, a contrast between /β/ and /b/ was created by the voicing of -p-:

Latin French Portuguese Old Spanish

habe↜渀屮re avoir haver /aβer/ aver
lav↜↜are
 laver lavar /laβar/ lavar
sape↜渀屮re savoir saber /saber/ saber

.╅ Notice that if we assume that northern Castilian had both voiced and voiceless bilabial
fricatives instead of labiodentals the subsequent change to [h] of the voiceless bilabial fricative
in this area appears rather more natural.
 Lenition and phonemic contrast in Majorcan Catalan 

This contrast between /β/ and /b/ was maintained in Old Spanish, but was later lost by
the subsequent weakening of /-b-/, in a process that probably lasted several centuries.
In fact, the contrast between /b/ and /β/ has been preserved to the present day in the
town of Serradilla in Extremadura (Ariza 1993 [1992]).
The /b/-/v/ contrast in Majorcan Catalan mirrors the evolution in Central and
Southern Portuguese, whereas other Catalan varieties have undergone subsequent
merger, like in Spanish.
The different historical developments that we have briefly traced show that there
is little reason to believe in a principle of maintenance of contrast. The merger of all of
Latin -v-, -b- and -p- as /v/ in French and as /b/ [β] in Spanish rather argues against
that principle.
Why is it then that /b/ shows somewhat more resistance to weaken than /d/ and
/g/ in Majorcan Catalan if it is not a desire to preserve the contrast with /v/? What we
notice is that the same tendency is found elsewhere. Labials show greater resistance to
weakening than consonants with other places of articulation (see Foley 1970, although
not in every language, Hyman 1975:€ 164–169, Kirchner 2001:€ 6). For instance, in
French, Latin intervocalic -t- and -c- were regularly lost, as in vita > vie, secu↜渀屮↜渀屮ru >
sûr, but labial -p- has been preserved as /-v-/, as we have just noted (Nyrop 1899). In
Catalan as well, labials are more resistant to deletion than dentals and velars. Although
we find instances of both preservation and of loss of all of Latin -b-, -d-, -g-, the labial
has been preserved more regularly than the other two consonants. Latin intervocalic
-d- and -g- were frequently lost: sudaâ•›re > suar, leâ•›gaâ•›le > lleial (although -d- gives
/z/ in some contexts, Badia Margarit 1981:€187–189, Coromines 1973:€209–216, and
-g- is also sometimes preserved, especially in posttonic position), but -b- and -v- have
been normally preserved, caballu > cavall, nova > nova, except in some specific
morphological contexts, such as the imperfect past and sometimes in contact with a
rounded vowel.
What we see is a greater resistance of intervocalic labial obstruents to weaken,
regardless of the possible consequences of this lenition for the system of phonological
oppositions.
Ohala & Riordan (1979) provide evidence from other, non-Romance, languages
in which /b/ is immune to other processes affecting voiced plosives (see also Solé,
Sprouse & Ohala 2009). For example, in Nubian the voiced plosives /d, 3, g/ lengthen
and devoice in certain morphological contexts. The labial /b/, on the other hand,
increases its duration without devoicing. In Bantu languages, as well, /b/ is kept voiced
in certain contexts where /d/ and /g/ devoice.
According to Ohala & Riordan (1979), these asymmetries between labial and
non-labial voiced plosives are due to aerodynamic conditions needed for the produc-
tion and maintenance of voicing. In order for the vocal folds to vibrate, they must
be adducted and the airflow through them must be sufficient. For the vibration to be
maintained, there must be a certain difference between subglottal and oral pressure.
 José Ignacio Hualde, Marianna Nadeu & Miquel Simonet

In the case of the voiced plosives, the constriction that blocks the passage of air
increases the pressure in the vocal tract, reducing the difference between oral and
subglottal pressure. This explains the tendency of voiced plosives to either devoice
or weaken. Ohala (1983, 1997) and Ohala and Riordan (1979) have referred to this
phenomenon as the Aerodynamic Voicing Constraint (AVC).
The AVC explains the difference between labial and other voiced plosives (Ohala€&
Riordan 1979). Labial /b/ permits the continuation of voicing during a longer period
than the other plosives. This is not due to the existence of a larger oral volume but
rather to the compliance of the surface of the part of the vocal tract involved in the
production of the consonant. Labials have a larger surface of compliance than dentals
or velars (the entire surface of the tongue and part of the cheeks), which allows reduc-
tion in oral pressure.
Although Ohala’s AVC is most directly concerned with differences in the likeli-
hood of devoicing of plosives depending on place of articulation, we believe that this
account can be extended to our case, explaining the place asymmetries that we find in
the weakening of /b d g/.
In our interpretation of the AVC, a spirantization process would affect /d/ and /g/
before it affects /b/ for aerodynamic reasons, since [b] offers less aerodynamic diffi�
culties. This would be independent of whether or not the language has a phoneme /v/
with which /b/ is in contrast.
Let us now consider the fact that those Catalan varieties where noncontinuant
realizations of /b/ in intervocalic position and other contexts of lenition have been
observed are precisely those that also preserve a phoneme /v/. That is, it would seem
that we need to be able to explain why in Catalan varieties we find the two states of
affairs in (a) and (b) but not those in (c) or (d) in Table 1:

Table 1.╇ Possible realizations of b and v in intervocalic position in Catalan dialects

roba nova

a. [-β-] [-β-] Central Catalan, etc.

b. [-b-] [-v-] Majorcan, some Valencian, Algherese
c. [-b-] [-β-] ?
d. [-β-] [-v-] Camp de Tarragona?

As we have noticed, the situation in (c) is in fact essentially the one that appears
to have obtained in Old Spanish, later changed to that in (a) through the lenition of
[-b-], and that has been preserved in Serradilla up to our days.5 Old Catalan — or, at

.â•… Outside of Ibero-Romance, but within the Iberian Peninsula, a contrast [-b-] vs. [-β-],
albeit limited to the context after [u], is nowadays found in the Basque dialect spoken in
Arbizu, Navarre (Hualde 1996).
 Lenition and phonemic contrast in Majorcan Catalan 

least, most of the Old Catalan territory —, on the other hand, had the distribution
in€(b), still preserved in Majorca. The contrast in (d), which, in principle, could have
also resulted from (b) through lenition, seems more difficult to maintain through time,
also because of aerodynamic reasons. Rafel (1976) points out that in those Catalan
varieties that have a phoneme /v/, its allophonic range includes frictionless bilabial
realizations in intervocalic position. This weakening would appear to also obey aerody-
namic conditions. The production of voiced fricatives requires a strong airflow inside
the oral cavity in order to produce friction, but, at the same time, the air pressure must
be low in the oral cavity in order to maintain vibration of the vocal folds (as expressed
in the AVC). The difficulty to satisfy both requirements simultaneously leads to either
devoicing or loss of friction, giving rise to approximants (Johnson 2003; Ohala 1983,
1997). A contrast between /-β-/ and /-v-/ would be difficult to preserve through time if
the second phoneme is actually realized as [-v-] ~ [-β-]. Rafel (1976) observes that the
weakening of /b/ leads to the confusion of the two phonemes, since there is too much
overlap between their dispersion fields (see also Alonso 1962:€187–188). That is, the
situation in (d) is not impossible, but it seems clear that a contrast between a phoneme
realized as [-β-] and another phoneme variably realized as [-v-] ~ [-β-] would tend to
lead to phonemic merger (as we saw happened with the successors of Latin -b- and
-v- in all Romance languages).6
We thus conclude that the different allophonic realizations of /b/ in Majorcan
Catalan and in Central Catalan can be explained as different stages in a historical evo-
lution involving the weakening of the voiced plosives and that no special principle of
maintenance of contrast needs to be invoked.7

References

Allen, W. Sidney. 1978. Vox Latina: The pronunciation of Classical Latin, 2nd ed. Cambridge:
Cambridge University Press.
Alonso, Dámaso. 1962. La fragmentación fonética peninsular (Enciclopedia Lingüística Hispánica,
vol. I, supplement). Madrid: Consejo Superior de Investigaciones Científicas.
Ariza, Manuel. 1990. “Diacronía de las consonantes labiales sonoras en español”. El cambio
lingüístico en la Romania, 11–26. Tarragona. (Repr. in Ariza 1993:€47–63.)

.╅ A reviewer points out that the situation in (d) was the one found in the Camp de Tarragona
dialect in the 20th century. As noted above, the phonemic contrast is in fact disappearing
from this dialectal area. Martínez-Gil (1998) also assumes the contrast in (d) for Old Spanish,
against the majority opinion.
.╅ As a reviewer points out, the view that sound change is purely conditioned by phonetic
factors, without reference to phonological notions such as preservation of contrast, is indeed
the Neogrammarian position (see, e.g. Hock 1991). Our view is thus somewhat different from
that of Martínez-Gil (this volume).
 José Ignacio Hualde, Marianna Nadeu & Miquel Simonet

Ariza, Manuel. 1992. “/b/ oclusiva y /(/ fricativa en Serradilla, Cáceres”. Anuario de Letras 30.
173–176. (Repr. in Ariza 1993:€65–70.)
Ariza, Manuel. 1993. Sobre fonética histórica del español. Madrid: Arco/Libros.
Baayen, Harald. 2009. Linguistic Data Analysis. Cambridge: Cambridge University Press.
Badia Margarit, Antoni. 1981. Gramàtica històrica catalana. Barcelona: Tres i Quatre.
Boersma, Paul & David Weenink. 2008. PRAAT: Doing phonetics by computer (Version 5.0.35),
University of Amsterdam. Downloaded 15 Oct. 2008 from http://www.praat.org/.
Carrasco, Patricio. 2008. An Acoustic Study of Voice Stop Allophony in Costa Rican Spanish.
Ph.D. dissertation, University of Illinois at Urbana-Champaign.
Cole, Jennifer, José I. Hualde & Khalil Iskarous. 1999. “Effects of Prosodic Context in /g/ Lenition
in Spanish”. Proceedings of the 4th International Linguistics and Phonetics Conference ed. by
Osamu Fujimura, Brian D,. Joseph & B. Palek 575–589. Prague: The Karolinum Press.
Colantoni, Laura & Irina Marinescu. 2008. “The Scope of Stop Weakening in Argentine Spanish”.
Paper presented at the 4th Laboratory Approaches to Spanish Phonology Conference, Austin,
Texas, September 2008.
Coromines, Joan.1973. Lleures i converses d’un filòleg. 2nd ed. Barcelona: El Pi de les Tres
Branques.
Eddington, David. 2009. “A Gradient Analysis of Spanish Voiced Approximants: New data
undermine some traditional notions”. Poster presented at the 4th Phonetics and Phonology
in Iberia Conference, Las Palmas de Gran Canaria, Spain 2009.
Foley, James A. 1977. Foundations of Theoretical Phonology. Cambridge: Cambridge Univ. Press.
Hyman, Larry. 1975. Phonology: Theory and Analysis. New York: Holt, Rinehart & Winston.
Hock, Hans Henrich. 1991. Principles of Historical Linguistics, 2nd revised ed. Berlin: Mouton
de Gruyter.
Hualde, José I. 1996. “Arbizuko euskararen zenbait soinu bereizgarriz”. Uztaro 18.49–60.
Hualde, José I. & Marianna Nadeu. to appear. “Lenició i contrast fonèmic en català central i en
mallorquí”. Homenatge a Max Wheeler (provisional title) ed. by Clàudia Pons & Maria Rosa
Lloret. Barcelona: Publicacions de l’Abadia de Montserrat.
Johnson, Keith. 2003. Acoustic & Auditory Phonetics. Oxford: Blackwell.
Johnson, Keith. 2008. Quantitative Methods in Linguistics. Oxford: Blackwell.
Kingston, John. 2008. “Lenition”. Selected Proceedings of the 3rd Conference on Laboratory
Approaches to Spanish Phonology ed. by Laura Colantoni & Jeffrey Steele, 1–31. Somerville,
Mass.: Cascadilla Proceedings Project.
Kirchner, Robert M. 2001. An Effort Based Approach to Consonant Lenition. New York & London:
Routledge.
Lapesa, Rafael. 1981. Historia de la lengua española. 8th ed. Madrid: Gredos.
Lavoie, Lisa. 2000. Phonological Patterns and Phonetic Manifestations of Consonant Weakening.
Ph.D. dissertation, Cornell University, Ithaca, N.Y.
Lloyd, Paul M. 1987. From Latin to Spanish. Philadelphia: American Philosophical Society.
Martínez-Gil, Fernando. 1998. “On the Spelling Distinction b vs. u/v and the Status of Spiran-
tization in Old Spanish”. Theoretical Analyses on Romance Languages: Selected papers from
the 26th Linguistic Symposium on Romance Languages ed. by José Lema & Estela Treviño,
283–316. Amsterdam & Philadelphia: John Benjamins.
Martínez-Gil, Fernando (this volume). “Word-Minimality and Sound Change in Hispano-
Romance”.
Mascaró, Joan. 1984. “Continuant Spreading in Basque, Catalan, and Spanish”. Language Sound
Structure ed. by Mark Aronoff & Richard T. Oehrle, 110–121. Cambridge, Mass.: MIT Press.
 Lenition and phonemic contrast in Majorcan Catalan 

Nyrop, Kristoffer. 1899. Grammaire historique de la langue française. Copenhague: Det Nor-
diske Forlag.
Ohala, John. 1983. “The Origin of Sound Patterns in Vocal Tract Constraints”. The Production of
Speech ed. by Peter F. MacNeilage, 189–216. New York: Springer.
Ohala, John. 1997. “Aerodynamics of Phonology”. Proceedings of the 4th Seoul International
Â�Conference on Linguistics [SICOL], 92–97. Seoul: Linguistic Society of Korea.
Ohala, John & Carol J. Riordan. 1979. “Passive Vocal Tract Enlargement During Voiced Stops”.
Speech Communication Papers ed. by Jared L. Wolf & Dennis H. Klatt, 89–92. New York:
Acoustical Society of America.
Ortega-Llebaria, Marta. 2004. “Interplay between Phonetic and Inventory Constraints in the
Degree of Spirantization of Voiced Stops: Comparing intervocalic /b/ and intervocalic /g/
in Spanish and English”. Laboratory approaches to Spanish phonology ed. by Timothy Face,
237–253. Berlin: Mouton de Gruyter.
Penny, Ralph. 2002. A History of the Spanish Language, 2nd ed. Cambridge: Cambridge
University Press.
Prieto, Pilar & Teresa Cabré, coords. 2007–2010. Atles interactiu de l’entonació del català. http://
prosodia.upf.edu/atlesentonacio/
Rafel, Joaquim. 1976. “Fonologia diacrònica catalana: Aspectes metodològics”. Problemes de
Â�llengua i literatura catalanes: Actes del II col·loqui internacional sobre el català, 47–65.
Montserrat: Publicacions de l’Abadia de Montserrat.
Recasens, Daniel. 1971/1975. “Notes entorn del rendiment fonològic de l’oposició /b/-/v/ al
Tarragonès: Estat actual”. Estudis Romànics 16. 163–183.
Recasens, Daniel. 1985. Estudi lingüístic sobre la parla del Camp de Tarragona. Montserrat: Pub-
licacions de l’Abadia de Montserrat.
Recasens, Daniel. 1991. Fonètica descriptiva del català. (Assaig de caracterització de la pronún-
cia del vocalisme i consonantisme del català al segle XX). Barcelona: Institut d’Estudis
Catalans.
Recasens, Daniel. 2009. “Sobre la rellevància del detall articulatori i acústic en l’explicació
del canvi fonètic”. Diachronic Linguistics ed. by Joan Rafel, 115–134. Girona: Editorial
Documenta Universitaria.
Simonet, Miquel. 2008. Language Contact in Majorca: An experimental sociophonetic approach.
Ph.D. dissertation, University of Illinois at Urbana-Champaign.
Solé, Maria-Josep, Ronald Sprouse & John Ohala. 2009. “Adjustments in Voicing Control and
Cross-linguistic Patterns”. Paper presented at the 4th Phonetics and Phonology in Iberia
Conference, Las Palmas de Gran Canaria, Spain 2009.
Soler, Antonia & Joaquín Romero. 1999. “The Role of Duration in Stop Lenition in Spanish”.
Proceedings of International Congress of Phonetic Sciences 99, San Francisco ed. by John
Ohala, Yoko Hasegawa, Manjari Ohala, Daniel Granville & Ashlee C. Bailey, 483–486.
Berkeley, CA: Univ. of California, Berkeley.
Veny, Joan & Lídia Pons i Griera. 1998. Atles lingüístic del domini català. Etnotextos del català
oriental. Barcelona: Institut d’Estudis Catalans.
Wheeler, Max. 2005. The Phonology of Catalan. Oxford: Oxford University Press.
 Alveolar laterals in Majorcan Spanish
Effects of contact with Catalan?

Miquel Simonet
University of Arizona

This paper offers a sociophonetic profile of the production of alveolar laterals

in Majorcan Spanish, a dialect of Spanish spoken in the island of Majorca,
where Catalan is also spoken. Traditionally, Catalan alveolar laterals have been
described as being “dark”, i.e. velarized, while (Peninsular) Spanish laterals have
been described a “clear”, i.e. non-velarized. Thus, it could be hypothesized that
Catalan-dominant bilinguals speaking Spanish would tend to use velarized
laterals. However, some recent literature has shown that velarization is a
receding feature in Majorcan Spanish with young Catalan-dominant bilinguals
leaning towards the use of clear variants. This paper discusses acoustic data
gathered from several groups of Majorcan speakers, classified as a function
of their dominant language, age and gender. The results indicate that all these
three factors affect the degree of velarization of laterals, with Catalan-dominant
bilinguals using more velarized variants, and younger subjects (especially
females) using the least velarized variants.

1.â•… Introduction

Variationist studies whose goal is to assess the effects of sociolinguistic contact on the
Spanish spoken in the Catalan linguistic area are rather scarce (Boix & Vila 1998). Spe-
cifically referring to the Balearic Islands, much more common are the studies that deal
with language attitudes or issues of language choice, such as what are the social factors
and/or communicative situations that lead to the choice of Catalan instead of Spanish
or vice versa (Querol 2005; Villaverde 2005). The purpose of the present paper is to
investigate one of the sociophonetic characteristics of the Spanish variety spoken in
Majorca, where a conservative dialect of Catalan is spoken alongside a dialect of Spanish.
Particularly, the present paper analyzes the potential effects of linguistic contact on the
production of alveolar laterals.
Majorcan Spanish has undergone dramatic sociodemographic restructuring in
the last decades. During the 1950s and 1960s, massive numbers of Spanish-speaking
monolinguals immigrated to the island of Majorca from the Iberian Peninsula, mostly
 Miquel Simonet

from the South, e.g. Andalusia, Murcia, Extremadura, Castilla-la-Mancha (Salvà 2005).
Consequently, the use of Spanish increased dramatically during those decades and,
most importantly, was extended to the informal registers. In Majorca, prior to the
1950s, Spanish was used only in the highly formal registers, mostly by bilinguals whose
first language was Catalan and who tended to use Spanish only in the public spheres.
After the arrival of the migrants, the social norms of the 1950s and 1960s considered
Spanish as the de facto language to be used with Spanish speakers; that is, there was an
obvious situation of asymmetrical bilingualism according to which Catalan-dominant
bilinguals would unexceptionally switch to Spanish in the presence of Spanish speakers
(Boix & Vila 1998). Furthermore, due to the arrival of immigrants from the main-
land, non-Catalan-accented Spanish became an integral part of the pool of variation
present in Majorcan Spanish. Majorcans who were born during the 1960s grew up
in a bilingual society in which Spanish was the only language used in all spheres of
communication (both formal and informal) while Catalan was exclusively used in the
informal settings and only amongst Majorcans of islander origin. This situation has
been progressively changing since the early 1980s (Querol 2005; Villaverde 2005).
The degree to which the sociolinguistic situation in Majorca resembles that of
other Catalan-speaking regions is still largely unknown. This is important because
the social, political and demographic fabric of Majorca is different from that of both
Catalonia and Valencia. According to Querol (2004), the amount of use and positive
evaluation of Catalan in Majorca lies somewhere in between those on Catalonia and
those in Valencia. Thus, while in Catalonia Catalan is largely transmitted to the follow-
ing generation (45%), its transmission lags significantly behind in both the Balearic
Islands (38%) and Valencia (26%) (Querol 2004:€46). When subjects are asked if they
like or would like to speak Catalan, 65% of the respondents answer in the affirmative
in Catalonia, while only 40% and 35% do in the Balearic Islands and Valencia, respec-
tively (Querol 2004:€48). Finally, we may add that, in the same way that the fact that
Catalan receives a positive evaluation and is largely transmitted to the next genera-
tion does not imply that Spanish-like linguistic features will not affect it, as the spread
in Barcelona of the so-called xava sociolect of Catalan shows, the fact that Spanish is
widely used and positively evaluated in Majorca does not imply that Catalan-like fea-
tures will not be found in it. The goal of the present paper is to investigate one of the
phonetic characteristics of the Spanish spoken in Majorca and possibly relate its use to
the sociolinguistic background of the participants.
Alveolar laterals were selected as the focus of the present sociophonetic investiga-
tion because previous phonetic descriptions have suggested that there are robust
differences in the production of /l/ between (Peninsular) Spanish (Chafcouloff 1972;
Fernández Planas 2000; Recasens 1987) and (Majorcan) Catalan (Recasens 2004;
Recasens & Espinosa 2005). In particular, Spanish laterals have been described as
being “clear”, while Majorcan Catalan laterals have been characterized as being “dark”
 Laterals in Majorcan Spanish 

(Recasens & Espinosa 2005). Recasens (2004: 593) explains that “darkness” in laterals
refers to a phonetic property “according to which an alveolar lateral may sound darker
(and thus, more [u]-like) or clearer (and thus, more [i]-like).” This difference makes
alveolar laterals a potential feature for cross-linguistic transfer in an intensive Spanish-
Catalan contact situation, such as the one in Majorca.
Pieras (1999) conducted the only other variationist investigation to date on lateral
production in Majorca. Pieras recorded semi-spontaneous speech of a total of thirty-one
Catalan-Spanish bilinguals from Palma, the capital city of Majorca. Fourteen speakers
considered themselves native speakers of Spanish and seventeen considered them-
selves native speakers of Catalan, even though they were all highly proficient bilinguals
who were exposed to their non-dominant language on a daily basis. The bilinguals were
all recorded in Spanish. The analysis consisted of the binary classification of laterals
based on auditory and visual (spectrographic) inspections. It was found that, of the
14 Spanish-dominant bilinguals, only one pronounced dark laterals, while the rest
consistently used clear laterals. Interestingly, the only participant using dark laterals
had been heavily exposed to Catalan-accented Spanish during childhood, since his
parents were Catalan-dominant bilinguals but chose to speak exclusively in Spanish
with him. Of the 17 Catalan-dominant bilinguals, 10 systematically used clear laterals.
The other 7 speakers variably used dark and clear alveolar laterals. Interestingly, there
was a strong effect of age within the Catalan-dominant data subset, according to which
the older participants were more likely to use dark laterals than the younger ones.
According to Pieras’ data, the Catalan-dominant speakers born before 1960 remained
frequent users of dark, Catalan-like laterals, while those born after 1960 systematically
used clear, Spanish-like laterals, although they remained Catalan-dominant bilinguals
in adulthood.
What might have caused this age difference? Why are younger speakers, according
to Pieras’ data, less likely to transfer the dark laterals of their native language (Catalan)
to their non-native language (Spanish)? Bibiloni (1985) claims that young Catalan
speakers residing in the large metropolitan areas are likely to prefer clear, Spanish-like
laterals instead of the traditional Catalan dark laterals even when using their native
language, Catalan. That is, according to Bibiloni, the dark lateral allophone is receding
while the clear one is spreading. Taking these two observations into account, one
might hypothesize, as Pieras (1999) does, that a factor that may have favored the bor-
rowing of clear laterals by Catalan-dominant bilinguals in both their first (Catalan)
and second (Spanish) languages is the possibility that dark laterals are stigmatized
in Majorca. Indeed, according to Pieras, the production of dark laterals is stereotypi-
cally imitated when Spanish speakers joke about Catalan-accented Spanish. In other
words, dark laterals seem to be salient in indexing the Catalan-speaking origin of a
speaker of Spanish. Furthermore, according to Pieras, dark laterals may be perceived as
characteristic of rural Majorcan speakers. As a consequence, Pieras claims that Catalan
 Miquel Simonet

speakers are likely to adopt clear laterals in their speech, including both their Spanish
and Catalan. However, while anecdotal observations of both Pieras and myself point
towards the possibility that velarized laterals are stigmatized in Majorca, experimental
evidence is still lacking.
A decade has passed since Pieras conducted his sociolinguistic interviews. In
opposition to what Pieras found for alveolar laterals, recent sociophonetic research has
shown that some Catalan phonetic features are spreading to the Spanish variety spoken
in Majorca (Simonet 2008). The present paper addresses the following research ques-
tions: Do Catalan-dominant bilinguals use clear allophones when speaking Spanish
or are they likely to transfer dark allophones from Catalan? Do Spanish-dominant
bilinguals use clear laterals or are some likely to borrow dark allophones from Catalan?
Are there age or gender effects? One important improvement of the present paper
over previous work is that a detailed acoustic analysis is carried out here. Existing
research (Pieras 1999) is based on impressionistic transcription that only acknowledges
two possibilities for which phonetic symbols to employ. Phonetic symbols, however,
simplify phonetic reality. Importantly, there are reasons to believe that darkness in
alveolar laterals is a gradient (and not discrete) property (Recasens 2004). Thus, only
an acoustic analysis can reveal evidence that considers the velarization degree inherent
in laterals.

2.â•… Method

2.1â•… Participants
A language background questionnaire (LBQ) was administered to 40 Catalan-Spanish
bilingual participants. The LBQ was based on the one used in Flege & MacKay (2004).
The participants were asked questions regarding demographic information such as
gender, age (year of birth), time spent outside of Majorca, and place of birth. Addition-
ally, they were asked the following language-background questions: (1) “Which language
do you consider to be your native language?” (2) “Which language did you use to
speak at home, as a child, with your family?” (3) “Which language do you use mostly in
your daily life?” (4) “Rate the percentage of use of Spanish/Catalan in your daily life: (i)
with family, (ii) with friends, (iii) at work/school, (iv) when shopping, (v) with strangers.”
The LBQ was administered in the language of choice of the participants.
Two groups were created as a function of their dominant language, a group of
Catalan-dominant and a group of Spanish-dominant subjects. This classification was based
on the responses to the LBQ. A bilingual speaker was classified as Catalan-dominant
if s/he answered “Catalan” to the first three questions and had higher percentages of
use of Catalan in all of the communicative situations about which the questionnaire
inquired. The opposite was true for the bilinguals classified as Spanish-dominant.
 Laterals in Majorcan Spanish 

Within each dominant-language group, two age groups were created, one mainly
consisting of speakers in their fifties or early sixties (older [O]) and another
mainly consisting of speakers in their twenties or early thirties (younger [Y]). Only
the younger speakers were recorded in both their dominant and non-dominant
languages, since several of the older speakers resisted being recorded in their non-
dominant language. Consequently, Spanish data is available for only 30 participants,
divided into three main groups: (i) 10 Catalan-dominant speakers in their twenties or
early thirties, (ii) 10 Spanish-dominant speakers also in their twenties or early thir-
ties and, finally, (iii) 10 Spanish-dominant speakers in their fifties or early sixties.
Five males and five females were recruited in each of the three groups. In sum,
there were a total of six sub-groups: SPMY (Spanish-dominant males, younger),
SPFY (Spanish-dominant females, younger), SPMO (Spanish-dominant males,
older), SPFO (Spanish-dominant females, older), CTMY (Catalan-dominant males,
younger), and CTFY (Catalan-dominant females, younger). (See Simonet (2008) for
further details.).

2.2â•… Recordings
The 30 speakers were asked to read a list of Spanish sentences. The speakers read aloud a
total of nine (9) target sentences five times from a printed list (9 *5 = 45). Each speaker
recorded five randomizations. Each list included many (50) distracters, which had dif-
ferent prosodic configurations so as to minimize the use of repetitive list intonation
(Simonet, 2008). A total of 1350 sentences were recorded: 9 [sentences] *30 [speakers]
*5 [iterations]. There was a theoretical ceiling of 225 data points per cell or subgroup.
The materials controlled for: (1) prosodic context [utterance-initial, utterance-
medial, utterance-final], and (2) surrounding vowels [/i, a, u/]. The complete list of
sentences is provided in Table 1. The materials were adapted from the Catalan materials
used in Recasens & Espinosa (2005). The general design also follows Recasens &
Espinosa (2005).

Table 1.╇ Spanish materials read aloud by the 30 Catalan-Spanish bilingual speakers
Sentence Translation Context

litros de leche “liters of milk” li-

laca muy buena “very good hair spray” la-
lupa francesa “French lens” lu-
pedí líquido “s/he ordered liquid” -ili-
se pondrá laca “s/he will put hair spray on” -ala-
habla zulú bien “s/he speaks Zulu fluently” -ulu-
venden dos mil “they sell two thousand” -il
el menor mal “the smaller damage” -al
hecho de tul “made out of tulle” -ul
 Miquel Simonet

The speakers were recorded in their home or workplace using a solid-state digital
recorder (Marantz PMD660) and a head-worn dynamic microphone (Shure SM10A).
The productions of the participants were digitized at 44.1 kHz (16-bit quantization)
and later down-sampled at 22.05 kHz in order to save disk space.

2.3â•… Analysis
The acoustic analysis of the laterals was based on an investigation of the second formant
(F2). F2 is generally taken as an optimal acoustic correlate of degree of darkness in
alveolar laterals since it “is positively related to tongue dorsum raising and fronting and
dorsopalatal contact size, and inversely related to the length of the back cavity behind
the primary constriction and to back constriction narrowing” (Recasens & Espinosa
2005:€10). Formant tracks were calculated with Praat software (Boersma 2001), which
applies a Gaussian-like window, and computes the LPC (linear predictive coding)
coefficients with the algorithm by Burg (cf. Praat’s help manual).
For each sentence, two or three temporal landmarks were hand-marked by inspecting
synchronized displays of sound waves and spectrograms. Specifically, the steady state of
the lateral (based on F2 trajectories) was identified, together with the midpoint or steady
state (also based on F2 trajectories) of the flanking vowels. Thus, in VCV sequences, three
time points were marked, while in VC and CV sequences, only two were.
Formant values from the selected time points were extracted in Hz and then
converted into Bark units (Zwicker 1961) using the Hz-to-Bark function available in
Praat (cf. Praat’s help manual for the formula used by the function). Therefore, F2 values
were extracted from all laterals, as well as from their flanking vowels /i, a, u/. The Bark
scale is a logarithmic psychoacoustic scale believed to reflect human perception. Due
to the fact that formant values were to be compared across speakers, they were further
transformed (normalized). The S-procedure was used (Watt & Fabricius 2002; Fabricius
2007). This procedure calculates the speaker’s F2 centroid (and F1, if necessary) and
expresses all F2 values with reference to the centroid. First, the mean F2 for /i/, which
is the most fronted vowel, and the mean F2 for /u/, which is the less fronted vowel, are
calculated for each individual speaker, from the vowels surrounding the target lateral
consonants. This provides the theoretical limits of the vowel space for each speaker.
Second, the F2 centroid, which is the mean of the /i/ and /u/ F2 means, is calculated
for each speaker. Third, all the F2 values of the laterals are divided by the F2 centroid
of each speaker. This normalization procedure results in F2 values expressed with
reference to “1”, which represents the speaker’s F2 centroid of their vowel system.
This normalization procedure has been used in research on the acoustics of vow-
els (Watt & Fabricius 2002; Fabricius 2007). In the present investigation, however,
we use it for the study of lateral consonants. The procedure allows for an acoustic
analysis of laterals that relates them with the vowel space of each individual speaker
(Recasens 2004).
 Laterals in Majorcan Spanish 

Prior to the analysis of the laterals, which is provided in the following section, we
analyzed the formant values of the flanking vowels. This is especially important in this
case because the S-normalization procedure expresses F2 values in the laterals with
respect to F2 values in the flanking vowels. Table 2 provides the F2 differences between
the mean F2 values of the cardinal vowels /i/ and /u/ as a function of the six groups
(cf. Simonet 2008). A cursory examination of Table 2 reveals that males have narrower
vowel spaces than females; that is, the distance in Bark units between the mean F2 for
/i/ and the mean F2 for /u/ is greater for females than for males. This is not understood
here as a social, but as a physiological difference.

Table 2.╇ Mean F2 difference in Bark between /i/ and /u/ as a function of group
Group F2 difference

SPMO 5.687
SPFO 6.251
SPMY 5.048
SPFY 6.057
CTMY 5.590
CTFY 6.287

The statistical analyses of the F2 values of the laterals were performed on both
S-normalized and Bark F2 values. There were 29 missing data points, due to vowel
devoicing or formant tracking errors. Therefore, the analyses discussed below are
based on a total of 1321 data points.
At this juncture, it needs to be added that a high F2 value (both Bark and
S-normalized) indexes a clear alveolar lateral, while a low F2 lateral (both Bark
and S-normalized) indexes a dark value (Recasens 2004). However, absolute F2 values
need to be interpreted with care since higher/lower values also index physiological
differences (related to vocal tract size) between, for instance, males and females.

3.â•… Results

Three extra-linguistic factors were considered: age, gender, and dominant language.
Since the experimental design is not fully orthogonal, two data subsets were extracted in
order to investigate the effects of each one of the factors within a balanced data set. First,
in order to investigate the effects of age and gender, a subset was extracted from the main
corpus consisting of only the data gathered from the Spanish-dominant speakers (SP
subset). In other words, age is considered a factor exclusively for the Spanish-dominant
speakers of Spanish, for which we have a balanced number of older and younger par-
ticipants. Second, in order to investigate the effects of dominant language and gender,
a subset was extracted consisting of only the data gathered from the younger speakers
(Y subset). That is, dominant language is considered a factor exclusively for the younger
 Miquel Simonet

Spanish-dominant and Catalan-dominant speakers, who have comparable ages.

Gender is considered a factor in both data subsets since we have balanced numbers in
both. For each of the two subsets, F2 Bark and F2 S-normalized values extracted from
the steady state of laterals were submitted to two independent two-way ANOVAs, one
for each of the two main dependent variables or normalization procedures.
Regarding the SP subset, the ANOVA carried out on the F2 Bark values revealed
significant effects of gender (F[1,873] = 74.2; p < 0.001), and age (F[1,873] = 15.3;
p < 0.001), but no significant interaction (F[1,873] = 2.2; ns). These effects arise from
the following two facts: (1) younger Spanish-dominant speakers tend to have higher
F2 Bark values than the older Spanish-dominant speakers, and (2) females have much
higher F2 Bark values than males. The gender difference obtains in both age groups,
which is revealed by the lack of a significant interaction. Figure€1 displays the direction
of the effects and their relative strength.
The ANOVA carried out on the S-normalized F2 values also revealed significant effects
of gender (F[1,873] = 36.4; p < 0.001), and age (F[1,873] = 23.1; p < 0.001), and no sig-
nificant interaction (F[1,873] = 2.4; ns). These effects arise from the following two facts: (1)
younger Spanish-dominant speakers tend to have higher S-normalized F2 values than the
older Spanish-dominant speakers, and (2) males have higher S-normalized F2 values than
females. Figure€1 displays the direction of the effects and their relative strength.
1.00
11.6

0.95
F2 Normalized
11.2
F2 Bark

0.90
10.8

0.85
10.4

SPFO SPFY SPMO SPMY SPFO SPFY SPMO SPMY

Group Group

Figure 1.╇ Means and 95% confidence intervals of F2 (Bark and S-normalized) values as a function
of age and gender of the speakers, Spanish-dominant data subset (SPFO Spanish-dominant Females
Older, SPFY Spanish-dominant Females Younger, SPMO Spanish-dominant Males Older, SPMY
Spanish-dominant Males Younger)

Notice that the factor gender triggers contradictory effects for each of the two
normalization procedures, i.e. F2 Bark and S-normalized F2. Thus, while the F2 Bark
values reveal that females use clearer laterals than males, the S-normalized F2 values
seem to show that it is males who use clearer allophones than females. We offer an
interpretation of this apparent contradiction in the Discussion section.
 Laterals in Majorcan Spanish 

Regarding the Y subset, the ANOVA carried out on the F2 Bark values revealed
significant effects of gender (F[1,874] = 154.3; p < 0.001), and dominant language
(F[1,874] = 174.1; p < 0.001), and a significant interaction (F[1,874] = 23.3; p < 0.001).
These effects arise from the following two facts: (1) females have higher F2 Bark values
than males, (2) the Spanish-dominant speakers have higher F2 Bark values than the
Catalan-dominant ones, and (3) the difference between Catalan-dominant males and
Catalan-dominant females is much larger than that between Spanish-dominant males
and Spanish-dominant females. The latter finding causes the interaction revealed by
the ANOVA. Figure 2 displays the direction of the effects and their relative strength.
Overall, it could be affirmed that Catalan-dominant males use dark lateral allophones
while the other three groups of speakers use clear allophones. Catalan-dominant
females and Spanish-dominant males use laterals with F2 Bark values that are inter-
mediate between the two extremes, represented by the Spanish-dominant females (on
the “clear” side) and the Catalan-dominant males (on the “dark” side).
The ANOVA carried out on the S-normalized F2 values also revealed signifi-
cant effects of gender (F[1,874] = 6.73; p < 0.01), and dominant language (F[1,874] =
188.01; p < 0.001), and a significant interaction between the two factors (F[1,874] = 30.9;
p < 0.001). These effects arise from the following two facts: (1) the Spanish-dominant
speakers have much higher S-normalized F2 values than the Catalan-dominant speak-
ers, and (2) males differ from females, but this difference varies as a function of lan-
guage dominance, which causes the interaction. That is, Catalan-dominant females
have slightly higher S-normalized F2 values than Catalan-dominant males while
Spanish-dominant males have higher S-normalized values than Spanish-dominant
females. Figure 2 displays the direction of the effects.
0.80 0.85 0.90 0.95 1.00
10.0 10.5 11.0 11.5

F2 Normalized
F2 Bark
9.5

CTFY CTMY SPFY SPMY CTFY CTMY SPFY SPMY

Group Group

Figure 2.╇ Means and 95% confidence intervals of F2 (Bark and S-normalized) values as a function
of dominant language and gender of the speakers, younger-speaker data subset (CTFY Catalan-
dominant Females, CTMY Catalan-dominant Males, SPFY Spanish-dominant Females, SPMY Spanish-
dominant Males)
 Miquel Simonet

Once again, notice that the factor gender triggers contradictory effects for each of
the two normalization procedures, i.e. F2 Bark and S-normalized, but only within the
Spanish-dominant data subset. Thus, while the F2 Bark values reveal that the Spanish-
dominant females use clearer laterals than the Spanish-dominant males, the S-normalized
F2 values seem to show that it is males who use clearer allophones than females. An
interpretation is offered below.

4.â•… Discussion

The present paper reported on acoustic data obtained from six groups of Catalan-Spanish
bilinguals residing in the Mediterranean island of Majorca, where Spanish and Catalan
are spoken. Data were gathered from Catalan-dominant as well as Spanish-dominant
bilinguals, from both males and females and also from two age groups. The acoustic
analyses revealed the findings discussed below.
The Spanish-dominant younger speakers seemed to use slightly clearer laterals
(higher F2 values) than the Spanish-dominant older participants. The age difference
was seen with both normalization procedures (Bark and S-normalized units). On the
one hand, the Spanish-dominant females presented higher F2 Bark values than the
Spanish-dominant males. On the other hand, the Spanish-dominant males presented
higher S-normalized F2 values than the Spanish-dominant females. In other words,
while one normalization procedure suggests that females use clearer allophones than
males, the other one suggests that males use clearer allophones than females. The
Catalan-dominant speakers use darker allophones than the Spanish-dominant ones.
This is more obvious for the Catalan-dominant males than for the Catalan-dominant
females. In other words, the Catalan-dominant males use dark allophones while the
Catalan-dominant females use somewhat clearer allophones that resemble (or have
similar F2 Bark values to) the ones used by the Spanish-dominant males. The Spanish-
dominant females use extremely clear allophones, as compared to the other groups.
First, it is unclear what might have caused the age differences amongst the
Spanish-dominant speakers. The fact that both normalization procedures reveal the
same effect suggests that the age difference is robust, though small. This indicates
that the younger speakers not only use laterals with higher overall (F2 Bark) values,
but also with values that situate them in a more fronted position within their own
vowel space (S-normalized F2). It appears that the older speakers analyzed here have
slightly wider vowel spaces than the younger ones (see above). Thus, the fact that the
older speakers use lateral allophones with less fronted values (F2 Bark) may not be due
to the fact that they have narrower vowels systems and therefore cannot front their
laterals more, but to an intentional centralizing gesture. We might speculate, regard-
ing this finding, that the younger speakers are using clearer laterals than the older
 Laterals in Majorcan Spanish 

ones probably because they attempt to dissimilate from Catalan-accented speakers

who tend to use darker lateral allophones. Recall that Pieras (1999) suggested that dark
laterals are stigmatized in Majorca. Thus, it is possible that young Spanish-dominant
speakers try to make sure their laterals are not perceived as dark. A way to test this
hypothesis would be to compare the laterals used by the young Spanish-dominant
speakers residing in Majorca with those produced by monolingual speakers of Spanish
residing in the Iberian Peninsula. Our speculation would predict that the laterals produced
by young Spanish-dominant, Majorcan bilinguals are clearer than the ones produced by
monolinguals. This is left for future research.
Second, we will offer an interpretation of the fact that the two normalization pro-
cedures seem to reveal contradictory findings for the Spanish-dominant males and
females. Recall that the females use alveolar laterals that are overall (F2 Bark) clearer
than the ones used by the males. However, since males have much narrower vowel
spaces than females (less F2 difference between /i/ and /u/) (Simonet 2008), a normali�
zation procedure that expresses the position of a lateral with respect to its distance
from an F2 centroid (calculated from /i/ and /u/ means) will likely show that males
need to approximate their laterals to their /i/ more than females do in order for their lat-
erals to “sound” clear. On the other hand, since females have a wider vowel space, their
laterals do not need to be so fronted (with respect to their own /i/) for them to sound
very clear, i.e. with a high F2 value. In other words, males and females use two different
strategies to achieve a very similar goal: to use a clear alveolar lateral in their Spanish.
In our opinion, this explains the gender differences within the Spanish-dominant group
and suggests that these are not due to sociolinguistic factors. Our interpretation is that
the two findings (extrapolated from the two normalization procedures) cancel each
other out. Thus, a relevant gender difference within the Spanish-dominant group does
not need to be postulated.
Third, the results of the analyses revealed strong effects of the dominant language,
with the Catalan-dominant speakers using overall darker allophones than the Spanish-
dominant ones. These effects were strong, but more obvious for the Catalan-dominant
males than for the females. In other words, the Catalan-dominant females used
lateral allophones that were somewhere between the clear allophones used by the
Spanish-dominant females and the dark allophones used by the Catalan-dominant
males. Recall that we have argued that the gender differences encountered within the
Spanish-dominant group obey two different strategies to achieve a similar goal and
thus do not need a sociolinguistic interpretation. However, the gender differences
found within the Catalan-dominant group are wide and robust. This suggests that a
sociolinguistic interpretation of the effects of gender within the Catalan-dominant
group is needed.
Pieras (1999) found that both Catalan-dominant males and females residing in
Palma (Majorca’s major city) who were born after 1960 were likely to use clear allophones
 Miquel Simonet

instead of dark ones, which were used only by the older speakers. From this find-
ing, Pieras inferred that the young Catalan-dominant speakers in Palma had made an
effort to learn the clear lateral allophone from the Spanish-dominant bilinguals and
were making sure to use it in their Spanish. Bibiloni (1985) claimed that Catalan-like,
dark laterals are receding in Catalan and Spanish-like, clear laterals are being adopted
by Catalan speakers in Majorca. The Catalan-dominant speakers who participated in
the present experiment, however, did not all use clear allophones even though they are
all in their twenties or early thirties, and thus born after the 1960s. The evidence pre-
sented here suggests that young Catalan-dominant males are “resisting” the adoption
of clear lateral allophones even when speaking Spanish, their non-dominant language.
Catalan-dominant females, on the other hand, were found to borrow clear laterals
from Spanish, although their laterals were not as clear as the ones used by the Spanish-
dominant females in their age group. Thus, if a change in progress exists, as Pieras
(1999) affirms, it is in a less advanced stage than claimed.

5.â•… Conclusion

The present paper has offered a sociophonetic profile of the production of alveolar
laterals in Majorcan Spanish, a dialect of Spanish in intensive contact with Catalan.
Traditionally, Majorcan Catalan alveolar laterals have been described as being dark,
i.e. velarized, while (Peninsular) Spanish laterals have been described as being clear, i.e.
non-velarized (Recasens 1987, 2004; Recasens & Espinosa 2005). The Spanish produc-
tions of six groups of Catalan-Spanish bilinguals (30 individuals) were analyzed. The
individuals were classified as a function of their dominant language, gender and age.
The existing sociolinguistic research has suggested that young Majorcan speakers
tend to use clear laterals irrespective of whether they are dominant speakers of Spanish
or Catalan (Pieras 1999). That is, while older generations of Catalan-dominant bilinguals
have a tendency to transfer dark laterals from Catalan and thus use dark laterals in
Spanish, young Catalan-dominant bilinguals tend not to do so (assuming they use
dark laterals in Catalan) (Pieras 1999).
The results of the experiment reported here indicate that Spanish-dominant
bilinguals produce clear lateral allophones while Catalan-dominant bilinguals differ as
a function of gender, with young females using clear-like allophones and males using
dark laterals. The data presented here do not allow us to verify whether a change in
progress affecting laterals is in effect or whether the situation obeys a phenomenon of
gender differentiation. While there is reason to believe that Catalan-dominant speak-
ers are borrowing Spanish-like, clear laterals, our data also suggest that dark later-
als are (still?) a reality of Majorcan Spanish. Interestingly, the age difference revealed
within the Spanish-dominant group suggests that Spanish-dominant speakers may be
 Laterals in Majorcan Spanish 

progressively diverging from (rather than converging with) Catalan speakers. This is
an unexpected contact effect, although not an unattested one.
Finally, it is shown here that an acoustic analysis is able to provide more infor-
mation than an analysis based on phonetic transcription and frequency counts,
especially since lateral velarization is a scalar property (Recasens 2004). In particu-
lar, the acoustic data discussed here showed that some Catalan-dominant speakers,
who could be impressionistically perceived as borrowing clear laterals from Spanish,
are still using allophones that differ acoustically from the ones used by the Spanish-
dominant bilinguals.

Acknowledgements

Thanks are due to the editors and reviewers of this volume for their comments, which
have significantly improved the quality of the paper. The usual disclaimers apply.

References

Bibiloni, Gabriel. 1983. La llengua dels mallorquins. Anàlisi sociolingüística. Ph.D. dissertation,
Universitat de Barcelona.
Boersma, Paul. 2001. “Praat, a Aystem for Doing Phonetics by Computer.” Glot International
5.341–345.
Boix, Emili & Francesc X. Vila. 1998. Sociolingüística de la llengua catalana. Barcelona: Ariel.
Chafcouloff, Michelle. 1972. “Recherches sur la structure acoustique de /l/ et ses corrélations
articulatoires”. Travaux de l’Institut de Phonétique d’Aix 1.101–110.
Fabricius, Anne. 2007. “Variation and Change in the ‘trap’ and ‘strut’ Vowels of RP: A real time
comparison of five acoustic data sets”. Journal of the International Phonetic Association
37.293–320.
Fernández Planas, Ana M. 2000. Estudio electropalatográfico de la coarticulación vocálica en
estructuras VCV en castellano. Ph.D. dissertation, Universitat de Barcelona.
Flege, James E. & Ian R. MacKay. 2004. “Perceiving Vowels in a Second Language”. Studies in
Second Language Acquisition 26.1–34.
Pieras, Francesc. 1999. Social Dynamics of Language Contact in Palma de Mallorca: Attitude and
phonological transfer. Ph.D. dissertation, Pennsylvania State University.
Querol, Ernest. 2004. “Comparació dels usos i les representacions de les lengües a les Illes Balears,
Catalunya, el País Valencià i Andorra”. Treballs de Sociolingüística Catalana 18.43–62.
Querol, Ernest. 2005. Els balears i el català: usos i representacions socials. Palma: Documenta
Balear.
Recasens, Daniel. 1987. “An Acoustic Analysis of V-to-C and V-to-V Coarticulatory Effects in
Catalan and Spanish VCV Sequences”. Journal of Phonetics 15.299–312.
Recasens, Daniel. 2004. “Darkness in [l] as a Scalar Phonetic Property: Implications for phonology
and articulatory control”. Clinical Linguistics and Phonetics 18.593–603.
 Miquel Simonet

Recasens, Daniel & Aina Espinosa. 2005. “Articulatory, Positional and Coarticulatory Char-
acteristics for Clear /l/ and Dark /l/: Evidence from two Catalan dialects”. Journal of the
International Phonetic Association 35.1–25.
Salvà, Pere. 2005. “La nova realitat geodemogràfica de les Illes Balears al començament del segle
XXI: creixement de la població i fluxos migratoris”. Treballs de Sociolingüística Catalana
18.131–141.
Simonet, Miquel. 2008. Language Contact in Majorca: An experimental sociophonetic approach.
Ph.D dissertation, University of Illinois at Urbana-Champaign.
Villaverde, Joan A. 2005. “L’enquesta sociolingüística 2003. Principals resultats”. Treballs de
Sociolingüística Catalana 18.83–96.
Watt, Dominic & Anne Fabricius. 2002. “Evaluation of a Technique for Improving the Mapping
of Multiple Speakers’ Vowel Spaces in the F1-F2 plane”. Leeds Working Papers in Linguistics
9.159–163.
Zwicker, Eberhard. 1961. “Subdivision of the Audible Frequency Range into Cricital Bandwidths
(Frequenzgruppen)”. Journal of the Acoustical Society of America 33.248.
 Units of speech production in Italian

Irene Vogel1 & Linda Wheeldon2

1University of Delaware/2University of Birmingham, UK

When speakers produce a prepared utterance, the amount of time required to

initiate the utterance reflects the number of units in the utterance. In this paper,
we investigate the nature of this unit on the basis of Italian experimental data,
and compare our findings to previous studies of Dutch. We demonstrate that
the findings for Italian and Dutch are essentially the same, despite substantial
prosodic differences between Romance and Germanic languages. It is shown that
in order to account for these findings, a prosodic unit between the Phonological
Word and the Phonological Phrase is needed. Two possibilities are considered,
a recursive Phonological Word and a more recently proposed constituent, the
Composite Group.

1.â•… Introduction1

It has been proposed that when speakers produce a prepared utterance, the number
of units contained in the utterance will determine the amount of time required to
initiate the production (Sternberg et al. 1978; Sternberg et al. 1980). This raises the
question of what type of units must be counted in this process. On the basis of previous
research, it has been suggested that the relevant unit is a prosodic one, as opposed to a
morphological or syntactic one, specifically the Phonological Word (e.g. Levelt 1989),
and experimental studies of Dutch have been conducted to provide psycholinguistic
support for this position (Wheeldon & Lahiri 1997, 2002). It appears, however, that
in some cases the results reveal a complication with regard to the claim that the Phono�
logical Word is the appropriate unit of speech encoding. In the present paper, we fur-
ther investigate the nature of this unit on the basis of Italian experimental data, and
demonstrate that the findings for Italian are essentially the same as those for Dutch,
despite substantial prosodic differences between Romance and Germanic languages.
In the following sections, we first present a brief summary of the issues related
to speech encoding and production that are relevant for the present paper. We then
discuss the experimental design and main findings of the Dutch experiments, since the

.╅ We would like to thank the two anonymous reviewers for their very helpful comments.
 Irene Vogel & Linda Wheeldon

Italian experiment we conducted was based on this previous research. Subsequently,

we discuss the methodology and present the results of the Italian experiment. Given
that there are problems with taking the PW as the unit of speech encoding in Italian,
as in Dutch, we suggest an alternative that appears to be valid for both languages, a
constituent in the prosodic hierarchy between the Phonological Word and the Phono-
logical Phrase – the Composite Group (Vogel 2009, submitted).

2.â•… Background

2.1â•… Speech production

When a speaker wishes to produce an utterance, the articulation of this utterance is
planned in groupings or units of speech. Thus, there is a correlation between the num-
ber of units that compose an utterance and the amount of time needed to begin to
articulate it. Different properties of an utterance may contribute to the amount of time
needed to encode it. For example, an utterance with more syllables would take longer
to encode than one with fewer syllables. Thus, (1a) will be encoded more quickly than
(1b) (e.g. Meyer, Roelofs & Levelt 2003).
(1) a. catσ1 b. caσ1taσ2maσ3ranσ4
Beyond the number of syllables, it appears that other aspects of the complexity of an
utterance play a role in the time required for its encoding. Thus, in the hypothetical
case in (2), it can be expected that (2a) would require less time to encode than (2b),
although they both contain the same number of syllables.
(2) a. [caσ1taσ2maσ3ranσ4]? b. [caughtσ1 inσ2 theσ3]? [rainσ4]?
The question that arises at this point is what determines the additional complexity of
some structures with respect to others. It has been proposed that in the final stages of
preparation prior to articulation, the relevant property is phonological – as opposed to
morphological or syntactic. More specifically, when producing a prepared utterance,
complexity is determined by the number of Phonological Words (PWs) the utterance
contains (Wheeldon & Lahiri 1997, 2002). Thus, if the question marks used as bracket
labels in the examples in (2) are taken to be PWs, it is predicted that the presence of
two such constituents in the second item will result in a longer preparation time than
the presence of a single PW unit in the first item. This possibility was tested experi-
mentally with Dutch, as described in the next section.

2.2â•… Speech encoding in Dutch

Wheeldon and Lahiri (1997, 2002) present a series of experiments conducted with
speakers of Dutch aimed at testing the claim that the PW is the constituent that
 Units of speech production in Italian 

determines the relevant type of complexity in the phonological encoding of speech for
production. They demonstrated that in a prepared speech paradigm, the more PWs an
utterance contains, the more time a speaker will require prior to beginning the actual
production of the utterance. In this paradigm, subjects saw a stimulus word or phrase
on a computer screen, and then used this item as part of their response to a recorded
question played through headphones (e.g. wat zoek je? = “what do you seek?”; wat
doe je? “what do you do?”), or simply as part of a generic response (Het was… = “It
was…”). Following the stimulus presentation, three beeps were heard, the last of which
was preceded by a variable pause duration, to prevent the development of an automatic
response rhythm. The subjects were instructed to prepare their utterance and to wait
until the last beep to begin their answer.
In Wheeldon and Lahiri (1997) (Study 1) the five stimulus structures shown in (3)
were investigated. The brackets indicate the PW structures that were posited. It should
be noted that in this study, the placement of the clitic in the first PW with the verb was
one of the experimental conditions, however, in the present study, what is crucial is
that there is a single PW following the first portion of the reply in this case.

(3) Visual Display Response

(Structure) (Example)
a. V + Clitic
het water [Ik zoek het] [water]
“the water” “I seek the water”
b. A + N Phrase
vers water [Ik zoek] [vers] [water]
“fresh water” “I seek fresh water”
c. Single word
water [Ik zoek] [water]
“water” “I seek water”
d. Pronoun
het [Ik zoek] [het]
“it” “I seek it”
e. Null (control)
zoek [Ik zoek] [Ø]
“seek” “I seek”

In Wheeldon and Lahiri (2002) (Study 2), some of the same structures were included,
along with some new ones. The stimulus structures tested in this study are those in (4),
where the last two differ in the position of primary stress.
 Irene Vogel & Linda Wheeldon

(4) Structure Example Number of PWs

a. N + N compound [oog] [lid] “eye lid” 2 PW
b. A + N phrase [oud] [lid] “old member” 2 PW
c. single word (sw) [orgel] “organ” 1 PW
d. single word (ws) [orkaan] “hurricane” 1 PW
The procedure followed was similar to that in Study 1. The responses were all given in
the generic response structure shown in (5).
(5) a. Response – part 1: Het was… (“It was…”)
b. Response – part 2: ooglid/oud lid/orgel/orkaan
As Wheeldon and Lahiri hypothesized that the number of PWs in an utterance is what
determines the amount of time needed to prepare to produce it, it was expected that
the stimuli would fall into three groups, as shown in (6), which combines the struc-
tures of the two studies. The control was predicted to exhibit the shortest lag or latency
as the subjects prepared to produce it, while the items with two PWs (i.e. the phrases
and the compounds) could be expected to exhibit the longest response latencies. The
remaining items, all consisting of a single PW, were predicted to require an intermediate
amount of preparation time.
(6) a. 2 PWs SLOW (longest latencies) (items 3b; 4a, b)
b. 1 PW FAST (shorter latencies) (items 3a, c, d; 4c, d)
c. Ø FASTEST (shortest latencies) (item 3e – control)
The results confirmed most of these predictions, however, the response latency observed
with the compounds was not what was expected if we simply count the number of PWs,
as shown in Table 1. It should be noted that what is relevant for the present comparison
is the number of PWs following the first portion of the response in Study€1, so in the
V+clitic structures, we list here only the PW that followed this portion.

Table 1.╇ Response rates for different structures in Wheeldon and Lahiri (1997) = Study 1;
(2002) = Study 2

Study 1 Study 2
Response Rate Structure Example Structure Example
Slow A + N phrase [vers][water] A + N phrase [oud][lid]
“fresh water” “old member”
Fast (V + clitic) +N [water] N+N [oog][lid]
“water” Compound “eye lid”
Pron [het] Word [orgel],
“it” [orkaan]
“organ, hurricane”
Word [water]
“water”
Fastest Control Ø
 Units of speech production in Italian 

As can be seen, the compounds exhibited a faster response time than the phrases,
despite the fact that both constructions contain two PWs. In fact, the compounds
behaved like single words in the amount of time required to produce a prepared
response containing these stimuli.
The fact that the number of PWs did not accurately predict the response pat-
terns with the compounds led Wheeldon and Lahiri (2002) to introduce an additional
statement to their proposal regarding the units of speech encoding. First, they speci-
fied that compounds involve an additional layer of prosodic structure, such that the
individual PWs are grouped into a larger, recursive PW. By contrast, the PWs of a
phrase are directly grouped into a Phonological Phrase (PPh), as shown in (7a) and
(7b), respectively.

(7) a. [[oog]PW [lid]PW]PW b. [[oud]PW [lid]PW]PPh

With this difference in structures, it was then proposed that in compounds it is only
the outer PW that counts in the determination of the response latencies. Aside from
requiring an additional specification, the recursive PW analysis potentially involves
several more general drawbacks, including the introduction of recursion, a property
generally associated with syntax, into the phonological component. It turns out more-
over that the structures that are claimed to be recursive do not actually conform to the
usual definition of recursion, which requires that a constituent of a particular type be
embedded within a larger constituent of the same type, that is, one that exhibits the
same characteristic properties (e.g. Pinker & Jackendoff 2005; Jackendoff & Pinker
2005; Neeleman and van de Koot 2006). In the case of the recursive PWs, the outer
level is crucially different in its properties from the inner level, and thus does not meet
the requirement of recursion that constituents must be embedded in larger ones with
the same properties (Vogel 2009, submitted). As will be seen below, an alternative pro-
posed to account for the Italian data also accounts for the Dutch findings, and avoids
the problem with the definition of recursion.

3.â•… Experiment – speech encoding in Italian

A study similar to the Dutch research was conducted with Italian, both to determine
if compounds pattern like single words or like phrases, and to investigate the cross-
linguistic properties of speech encoding. That is, since Germanic and Romance lan-
guages exhibit substantial differences in their overall prosodic structure and patterns,
the question was whether these differences would lead to different results in speech
encoding, or whether the phenomenon of speech encoding is so basic that it is not
sensitive to such cross-linguistic variation.
 Irene Vogel & Linda Wheeldon

3.1â•… Procedure
As in the Dutch studies, a prepared speech paradigm was used in the Italian experi-
ment. The subjects saw the target word or phrase on the computer and then, after
three beeps, responded with the item they had seen. Again, the third beep followed
a pause of variable duration, so as to avoid anticipatory responses. In this study, only
the simplest form of response was used, that with just the stimulus previously seen on
the screen. The experiment was presented using e-prime software, which recorded the
response time following the third beep. Prior to the actual experiment, practice items
were presented so the subjects could become familiar with the task.

3.2â•… Subjects
The participants were 17 students at the University of Bologna. They were tested indi-
vidually in a quiet room at the University. None of the subjects reported a history of
hearing or language problems.

3.3â•… Stimuli
3.3.1â•… Structure of stimuli
The target stimuli were pairs of Verb + Noun compounds and matched V + Noun
phrases. In addition, control stimuli consisting of a single word were used for com-
parison. Words with internal structure (other than an inflectional marker) referred
to here as “derived” items, and words without such structure were used. The control
structures were included to permit us to examine the possibility that morphological
complexity plays a significant role in speech encoding in Italian, although this was not
found for Dutch.
All stimuli consisted of four syllables, with primary prominence on the penul-
timate syllable. The words in each condition were also matched for their onset seg-
ments to avoid differences in the triggering of the voice key. The compounds and
phrases, as well as the two types of words were organized in groups with as much
phonological consistency as possible, as illustrated in (8).2 The brackets indicate PWs;

.╅ It should be noted that the prosody of the compounds and phrases is essentially identical
since phrasal prominence as well as compound prominence are on the last stressed syllable
in Italian. Other lexical prominences to the left are somewhat reduced in both cases, although
this does not lead to vowel raising of [ε, f] to [e, o] in either the compounds or phrases (cf. Ber-
tinetto€& Loporcaro 2005; Nespor & Vogel 1986). A similar stress pattern is observed in the
individual lexical items tested here since they have penultimate stress and an earlier, weaker,
secondary prominence; however, as single words, they are susceptible to vowel raising in syl-
lables without primary stress.
 Units of speech production in Italian 

all suffixes in Italian form part of the PW with their root (Nespor & Vogel 1986; Vogel
2009, among others). Each condition contained twelve items, for a total of 48 responses
per subject.

(8) Structure Example Number of PWs

a. V + N compound [cava][tappi] “cork screw” 2 PW
b. V + N phrase [caccia][talpe] “(he) hunts moles” 2 PW
c. Non-derived word [calamari] “calamari” 1 PW
d. Derived word3 [calzolaio] “shoe maker” 1 PW

Prior to the actual experiment, a practice session was conducted with 12 items (3 sets
of compounds, phrases and words). These were similar to the experimental stimuli
although not as closely matched, and were not used in the experiment itself. (See
Appendix for the full list of stimuli.)

3.3.2â•… Frequency of stimuli

It is well known that the frequency of a word has an effect on how quickly people
respond to it in different types of tasks. It would therefore be desirable for the stimuli
in any experiment, including the present one, to contain words of relatively similar
frequencies. This is problematic, however, when comparing words, even compounds,
with phrases, since phrases – unless they are idiomatic expressions – do not tend to
arise in precisely the same form with any degree of frequency. Moreover, a commonly
cited reference on Italian word frequency, Juilland and Traversa (1973), is based on
written texts, and thus does not list many items that are very frequent in the spoken
language, such as scolapasta “pasta colander”. Indeed, the dictionary contains relatively
few compounds, and those that are included (e.g. aeroplano “airplane”, capolavoro
“masterpiece”) are generally not the type used in the present study (i.e. Verb + Noun).
Some of the individual words (derived and underived) were found, but, of course, no
phrases were listed in the dictionary.
We thus also consulted an on-line source of word frequencies in Italian based on
the spoken language, BADIP (Banca dati dell’italiano parlato). This source was found
to contain more of the Verb + Noun type of compounds used in our study, as well as
individual word stimuli. We therefore made use of the BADIP counts for the purpose
of evaluating the relative frequency of the items used in the present study. Specifically,

.╅ Not all of the items with internal morphological structure were derived by productive
word formation rules, however, they were nevertheless considered distinct from those in
which no such structure was present at all. Since it was crucial to keep stress and syllable
structure as similar as possible across matched items, it was necessary to be somewhat more
flexible with regard to internal morphological structure.
 Irene Vogel & Linda Wheeldon

we considered the word form frequencies of (a) the compounds, (b) the derived words,
and (c) the non-derived words. In addition, we considered the frequencies of the two
word components of the compounds, and the two words in the phrases. The mean
frequency counts for each condition are given in Table 2.

Table 2.╇ Comparison of word form frequencies from the BADIP database

Condition Word Frequency First word/morpheme Second word/morpheme

Compounds 0.3 31.0 27.0

Phrases 41.2 17.8
Non-derived 14.9
Derived 7.0

It should be noted that not all of our stimulus items appeared in the BADIP list,
and in these cases the frequency was considered to be zero. We chose to include less
frequent items in certain cases since it was crucial that the stimuli conform to specific
phonological characteristics with regard to their segmental, syllabic and stress proper-
ties. Since the purpose of controlling for frequency is to ensure that any differences we
observe in speech latency can be attributed to prosodic structure rather than frequency
of use, the use of less frequent items is only an issue to the extent that the latency data
pattern with mean word frequency for each condition. Based on the word frequencies
for the single words, this would predict that the production latencies for the non-
derived words should be faster than for the derived words and that the compounds
should be the slowest. In contrast, the component frequencies for the compounds and
phrases do not differ greatly, predicting similar production latencies for these condi-
tions. As we show below neither of these frequency-sensitive patterns occurs.

4.â•… Results and analysis

While seventeen students participated in the study, it was necessary to exclude the
data from three of these individuals from our analysis. This was done since these sub-
jects had a mean response time, or response latency, of more than one second, which
indicated either that they did not respond immediately after the third beep, or did not
respond at all to a substantial number of items. In addition, we removed one of the
single words with internal morphological structure across all subjects: lamentele “com-
plaints”. This item was frequently misread by the subjects as a construction ending in
a clitic sequence, lamen-tele, analogous to lavar- tele “wash them for yourself ”. Since
lamen does not in fact exist, many subjects showed false starts and then corrections
with this item, and it thus seemed most appropriate to remove it for all subjects.
 Units of speech production in Italian 

We excluded several additional data points from the reaction time analysis. Specifi-
cally, we excluded error trials which were the result of false triggers, or responses that
were initiated before the third beep and timeouts, and failures to respond within the
maximum window allowed, 2 seconds. These exclusions resulted in the loss of 7% of
the data. The percentage error rate was similar across all conditions (Compounds 7%,
Phrases 8%, Non-derived, 6%, Derived 8%). An analysis of variance on percentage
error rates yielded no significant difference between conditions.
Finally, all data points that fell two standard deviations from the mean were
counted as outliers and excluded from the analysis. These exclusions resulted in the
loss of a further 5% of the data points.
Thus, on the basis of the responses of fourteen subjects, with the exclusion of
the items mentioned, we observe the following response patterns. Figure 1 shows the
mean response times for each of the four types of stimuli.

850.00

800.00
Mean onset latency (ms)

750.00

700.00

650.00

600.00
Compound Phrase Nonderived Derived

Figure 1.╇ Mean response latencies for four stimulus categories

As this figure shows, the amount of time required for encoding the phrases is
approximately 30 ms. longer than the time required for all of the other types of stimuli.
Thus, it can be seen that, as with the Dutch data, the compounds patterned with the
single words, not with the phrases.
An ANOVA was run on the reaction time data for the four experimental condi-
tions with subjects as the random factor. The main effect of condition was significant,
F(3,39) = 3.99, p < .05. Bonferroni post hoc pairwise comparisons were conducted to
determine which conditions differed significantly from each other. Response latencies
 Irene Vogel & Linda Wheeldon

for the phrases were significantly slower than for the compounds (p < .05) and the non-
derived words (p < .05). The difference between the phrases and the derived words was
marginally significant (p = .053). No significant differences were observed between the
response latencies for the compounds, non-derived and derived words.

5.â•… Discussion

As the results show, the response patterns for Italian were similar to those for Dutch.
Crucially, it was found in both languages that the response latencies for phrases are
longer than for compounds and for single lexical items. That is, despite the fact that
compounds contain two Phonological Words, the time required to encode them for
speech production is similar to that of grammatical words consisting of a single
Phonological Word.
The fact that the response patterns for Italian and Dutch are the same is particularly
interesting given that these languages, as representatives of the Romance and Germanic
families, otherwise exhibit quite different prosodic properties. Such properties include,
but are not limited to (a) the fact that Germanic languages are generally considered
“stress timed” as opposed to the “syllable timing” of Romance languages, and (b) the
presence of substantial marking of word boundaries in Germanic languages, while in
Romance languages the tendency is to syllabify and apply phonological phenomena
across words, typified by liaison in French.4 Thus, the finding that compounds pattern
like single words rather than phrases with regard to the amount of time needed for
their encoding for production appears to be independent of the basic prosodic struc-
ture of a language. More generally, it seems that the processes involved in speech pro-
duction are not sensitive to differences among languages, but rather may be seen as
constituting universal principles of linguistic organization.
With regard to the question of the unit involved in speech encoding, the original
proposal that this is the Phonological Word is challenged by the Italian findings, as
by the Dutch findings involving compounds. Since compounds pattern with individual
lexical items, and not with phrases, the fact that compounds and phrases both comprise

.╅ Of course, languages that exhibit additional prosodic differences will ultimately need
to be examined to further evaluate the proposal that language-specific prosodic differences
do not affect the overall process of speech encoding. In the meantime, preliminary results
with another Romance language, but one that is rather different prosodically from Italian,
Romanian, reveal a similar pattern of response latencies to those reported thus far (cf. Vogel€&
Spinu 2010).
 Units of speech production in Italian 

two Phonological Words appears not to be the relevant property in determining the
response latencies. As was seen above, Wheeldon and Lahiri (2002) originally pro-
posed for Dutch that the difference between compounds and phrases meant that a
recursive PW was needed for compounds. It must then be specified that for the pur-
poses of speech encoding only the outer PWs are counted. Given that Italian showed
the same response pattern, a similar interpretation could be offered for this language
as well. Aside from problems associated with introducing recursion into the phono-
logical component of grammar, and other problems with recursive PWs in particular
(cf. Vogel 2009, submitted), this solution relies on a stipulation regarding the relevant
level of PW structure in the determination of response latencies. As with any stipula-
tion, one must then ask why this particular requirement exists, and not some other one
(see Lahiri and Wheeldon, submitted).
Since the number of PWs does not allow us to predict the response latency patterns
without a distinction between two levels of PW structure, we must raise the ques-
tion as to whether the PW is indeed the appropriate unit for speech encoding. At first
glance, it might appear that what is relevant instead is whether or not the items coin-
cide with a single lexical entry available for lexical access. This seems uncontroversial
with regard to the non-compounds, and it also seems plausible for the compounds
since they involve structures and meanings that are not necessarily morphologically
and/or syntactically transparent. This would distinguish the two types of words and
compounds from the phrases, as the latter would involve the lexical accessing of two
individual components. However, the findings of Wheeldon & Lahiri (1997, 2002
summarized above) have demonstrated that the prepared speech task is sensitive to
the number of prosodic units not the number of lexical items or indeed the number of
content words that must be produced.
Thus, the question remains as to the nature of the unit involved in speech encoding.
Since the process of encoding treats compounds and individual grammatical words in
the same way, regardless of the number of PWs they contain, a unit is required that
does not crucially distinguish between these structures.5 In recent analyses of con-
structions involving affixes and clitics, as well as compounds, it has been proposed that
a phonological constituent is required between the Phonological Word and Phono-
logical Phrase levels in the prosodic hierarchy, the Composite Group (cf. Vogel 2009,

.╅ As one of the reviewers suggests, it would be particularly interesting to test longer com-
pounds, which abound in Germanic, but not Romance, languages. Since the number of syllables
in the response is relevant, it would be crucial that any longer compounds be matched with
single words and phrases with similar syllabic properties (e.g. teapot lid/tabulate).
 Irene Vogel & Linda Wheeldon

submitted).6 This unit replaces the Clitic Group in the prosodic hierarchy of Nespor
and Vogel (1986), and differs from it in several ways. For example, it allows the skip-
ping of prosodic levels which was not originally permitted by the Strict Layer Hypoth-
esis, and crucially for the present purposes, it not only groups together the members
of compounds, it also groups together words and associated clitics. The Composite
Group thus provides a unified account for the Italian and Dutch patterns in which lexi-
cal items, compounds and phrases behave in the same way. That is, it can be seen that
the number of Composite Groups in the various constructions in Dutch and Italian,
shown in (9)–(11), predicts precisely which items exhibit longer and shorter response
latencies.7

(9) Dutch – Study 1

a. (V+clitic) + N (…)[water] PW/CompG “water”
1 CompG
b. A + N phrase [vers]PW/CompG [water]PW/CompG “fresh water”
2 CompG
c. Single word [water]PW/CompG “water”
1 CompG
d. Pronoun [het]PW/CompG “it”
1 CompG
e. Null (control) --- Ø

(10) Dutch – Study 2

a. N + N compound [[oog]PW [lid]PW]CompG “eye lid”
1 CompG
b. A + N phrase [oud]PW/CompG [lid]PW/CompG “old member”
2 CompG

.╅ It should be noted that even if one does not accept the introduction of the Composite
Group into the prosodic hierarchy, the analysis must include a constituent that is greater than
the single PW and smaller than the PPh, such as the recursive PW (cf. Lahiri and Wheeldon
submitted). By the same token, an approach such as Itô and Mester’s in which minor and
major words may be used to differentiate compounds from phrases – despite other potential
differences – crucially requires a level of phonological organization that is neither the Phono-
logical Word nor the Phonological Phrase (cf. Itô & Mester 2009).
.╅ It will be recalled that in Wheeldon and Lahiri (2002), the verb and clitic were also of
relevance. They are not shown here since what is crucial for the present study is what follows
this first portion of the response.
 Units of speech production in Italian 

c. Single word (sw) [orgel]PW/CompG “organ”

1 CompG
d. Single word (ws) [orkaan]PW/CompG “hurricane”
1 CompG

(11) Italian Study

a. V + N compound [[cava]PW [tappi]PW]CompG “cork screw”
1 CompG
b. V + N phrase [caccia]PW/CompG [talpe]PW/CompG “(he) hunts
moles”
2 CompG
c. Non-derived word [calamari]PW/CompG “calamari”
1 CompG
d. Derived word [calzolaio]PW/CompG “shoe maker”
1 CompG

Only those structures with two Composite Groups require longer times for encoding.
All the other structures have faster response times since they contain a single Composite
Group (or none in the case of the control structure in the first Dutch study). Thus, not
only are the findings in the Italian study accounted for without stipulations regarding
the counting of inner and outer PW structures, the full set of findings in Dutch, shown
above in Table 1, is accounted for without additional specifications.

6.â•… Conclusions

In this paper, we have investigated the question of what linguistic unit is relevant for
speech encoding. We first considered two studies involving Dutch, and then presented
our research on Italian. The Italian study used the same type of prepared speech para-
digm as the Dutch studies, and measured the response latencies with different types of
structures: compounds, phrases, and derived and non-derived words.
The main finding – that Italian showed the same patterns as Dutch – is par-
ticularly interesting in that it demonstrates that the basic process of speech encoding
appears to be the same cross-linguistically, since the prosodic systems of Germanic
and Romance languages are quite different in general. With regard to the question
of what the relevant unit for speech encoding is, it was shown that simply counting
the number of PWs in a string does not yield the correct predictions. One possibility
is that proposed in Lahiri and Wheeldon (2002), where it is specified that for com-
pounds a recursive PW structure is needed, and it is the outer PW in this case that is
 Irene Vogel & Linda Wheeldon

relevant as the unit for speech encoding. We also considered an alternative analysis in
which it was proposed that the appropriate unit for speech encoding is a constituent
between the Phonological Word and the Phonological Phrase in the prosodic hierar-
chy – the Composite Group. In both types of analysis, what is crucial is that there is
a grouping of elements that is larger than the basic PW, but smaller than the Phono-
logical Phrase.

References

BADIP (Banca dati dell’Italiano Parlato = “Database of Spoken Italian”):

http://languageserver.uni-graz.at/badip/
Bertinetto, Pier Marco & Michele Loporcaro. 2005. “The Sound Pattern of Standard Italian,
as Compared with the Varieties Spoken in Florence, Milan and Rome”. Journal of the Inter-
national Phonetic Association 35.131–151.
Itô, Junko & Armin Mester. 2009. “The Onset of the Prosodic Word”. Phonological Argumentation:
Essays on Evidence and Motivation ed. Steve Parker, London: Equinox.
Jackendoff, Ray & Stephen Pinker. 2005. “The Nature of the Language Faculty and its Implica-
tions for the Evolution of Language”. (Reply to Fitch, Hauser, and Chomsky). Cognition
97.211–225.
Juilland, Alphonse & Vincenzo Traversa. 1973. Frequency Dictionary of Italian Words. The
Hague: Mouton.
Lahiri, Aditi & Linda Wheeldon. submitted. “Phonological Trochaic Grouping in Language
Planning and Language Change”.
Levelt, Willem J. M. 1989. Speaking: From Intention to Articulation. Cambridge, MA: MIT Press.
Meyer, Antje S., Ardi Roelofs & Willem J.M. Levelt. 2003. “Word Length Effects in Picture Naming:
The role of a response criterion”. Journal of Memory and Language 47.131–147.
Neeleman, Ad & Jan van de Koot. 2006. “On Syntactic and Phonological Representations”.
Lingua 116.1524–1552.
Nespor, Marina & Irene Vogel. 1986. Prosodic Phonology. Dordrecht: Foris.
Pinker, Stephen & Ray Jackendoff. 2005. “The Nature of the Language Faculty and its Implica-
tions for the Evolution of Language”. (Reply to Fitch, Hauser, and Chomsky). Cognition
97.211–225.
Sternberg, Saul, Stephen Monsell, Ronald L. Knoll & Charles E. Wright. 1978. “The Latency
and Duration of Rapid Movement Sequences: Comparisons of speech and typewriting”.
Information processing in motor control and learning ed. George E. Stelmach, 117–152. New
York: Academic Press.
Sternberg, Saul, Charles E. Wright, Ronald L. Knoll & Stephen Monsell. 1980. “Motor Programs
in Rapid Speech: Additional evidence”. The Perception and Production of Fluent Speech,
ed. Ronald A. Cole. 507–534. Hillsdale, NJ: Erlbaum.
Vogel, Irene. 2009. “The Status of the Clitic Group”. Phonological Domains: Universals and Devi-
ations ed. Janet Grijzenhout & Baris Kabak, 15–46. Berlin: Mouton de Gruyter.
Vogel, Irene. submitted. “Recursion in phonology”.
 Units of speech production in Italian 

Vogel, Irene & Laura Spinu. 2010. “The Composite Group as the Domain of Speech Encoding in
Romanian”. Paper presented at 8th Annual Hawaii International Conference on Arts and
Humanities. Honolulu, HI.
Wheeldon, Linda & Aditi Lahiri. 1997. “Prosodic Units in Speech Production”. Journal of Memory
and Language 37.356–381.
Wheeldon, Linda & Aditi Lahiri. 2002. “The Minimal Unit of Phonological Encoding: Prosodic
or lexical word”. Cognition 85.B31–B41.

Appendix: Stimuli

a.â•… Experimental Stimuli

Compound Phrase Simple Word Complex Word

cavatappi caccia talpe calamari calzolaio
“cork screw” “(he) hunts moles” “calamari” “shoe maker”
ficcanaso faccio nodi fazzoletto fidanzato
“busy body” “(I) make knots” “handkerchief ” “fiancé”
giradischi gioca calcio generale giardiniere
“record player” “(he) plays soccer” “general” “gardener”
guardaroba gusta vini gorgonzola godimento
“wardrobe” “(he) tastes wines” “gorgonzola” “enjoyment”
guastafeste guida treni Guatemala guarnizione
“kill joy” “(he) drives trains” “Guatemala” “decoration”
lavapiatti lascio Pisa labirinto lamentele
“dishwasher” “(I) leave Pisa” “labyrinth” “complaints”
reggiseno rode piante rubinetto rotazione
“brassiere” “(it) gnaws plants” “faucet” “rotation”
rubacuori rompe vetri religione relazione
“heart breaker” “(he) breaks “religion” “relation”
windows”
scaldabagno scopro perle scarafaggio scadimento
“water heater” “(I) discover pearls” “cockroach” “expiration”
scolapasta spendo soldi scarabocchio scocciatura
“colander” “(I) spend money” “scribble” “annoyance”
segnalibro serve lepre secolare sentimento
“book mark” “(he) serves rabbit” “secular” “sentiment”
spazzaneve spengo luci scaramuccia scatoletta
“snow plough” “(I) turn out lights” “skirmish” “can”
 Irene Vogel & Linda Wheeldon

b.â•… Practice Stimuli

Compound Phrase Simple Word Complex Word

grattacielo mangio mele maresciallo cambiamento
“sky scraper” “(I) eat apples” “marshal” “change”
lanciafiamme legge libri elefante legatura
“flame thrower” “(he) reads books” “elephant” “binding”
schiaccianoci cerca funghi argomento durazione
“nut cracker” “(he) seeks “argument” “duration”
mushrooms”
 Pitch polarity in Palenquero
A possible locus of H tone

John M. Lipski
The Pennsylvania State University

In the Afro-Iberian creole language Palenquero, tonic syllables receive a level

H€tone, and lexical words have at most one H tone per word. According to
previous studies, the final H tone of a phrase is usually either maintained as
a level tone with no L% boundary tone, or is downstepped to a mid tone. The
present study examines phrase-final combinations of words ending in tonic
vowels followed by one or more negative, possessive, or object clitics, all of which
receive an H tone. Field data reveal a systematic process of tonal dissimilation
between the tonic syllables, most frequently involving pitch upstepping of
the clitic, and less frequently downstepping of the clitic. This systematic pitch
dissimilation, not found elsewhere in Palenquero (including other phrase-final
combinations of successive tonic syllables that do not involve clitics), suggests the
operation of the Obligatory Contour Principle, in turn pointing to the emergence
of an H tone lexically attached to Palenquero enclitics.

1.â•… I ntroduction: Level and downstepped phrase-final high pitch

in Palenquero

The Afro-Iberian creole language Palenquero, spoken in San Basilio de Palenque,

Colombia, exhibits intonational patterns that differ notably from the surrounding
regional dialects of Spanish. In their analysis of Palenquero intonation, Hualde and
Schwegler (2008) observe that accented syllables receive a level H tone, and lexical
words have at most one H tone per word. The authors conclude (p. 25) that “Palenquero
is best analyzed as an accentual language, rather than as a language with lexical tone.”
The present study, while essentially grounded in Hualde and Schwegler’s conclusion,
offers more detailed data on a particular configuration that may signal a final vestige
(or the incipient development) of a phonologically distinctive tone. Attention will be
directed specifically at the phrase-final negator -nú, the post-nominal possessive clitics
-mí “my” and -sí “your,” and the dative and accusative clitics -mí “me” and -bó “you.”
 John M. Lipski

The data show a process of tonal dissimilation, which in turn suggests the presence of
a distinctively specified High tone on these clitics.
Among the many ways in which Palenquero intonation differs from patterns
found in Spanish is the behavior of phrase-final oxytonic words. Hualde and Schwegler
(2008:13–15) observe the frequent absence of an L% boundary tone: “a level high tone
on the final syllable, without the final fall that is nearly obligatory […] in final declara-
tive sentences.” An alternative resolution is downstepping of the phrase-final H tone
to a mid tone: “[…] the H tone of the last syllable is sometimes downstepped, so that
it is pronounced at the same level tone as the preceding (toneless pretonic) syllable”
(p. 15) The authors note (p. 15) that “from the data available it is not yet entirely
clear what pragmatic or other factors trigger this downstepping of utterance-final
highs.” They document a similar phenomenon (p. 17) when the sentence ends in two
stressed monosyllables.1

2.â•… Palenquero tonic clitics

Palenquero negation is carried out by the negative particle -nú affixed to the end of
the negated clause. The negator -nú has been described as a stressed element having
a high pitch, and the pitch tracks offered by Hualde and Schwegler (2008:11, 17) are
consistent with this observation. The data collected for the present study reveal that
other enclitics also receive a high tone (i.e. always pronounced with a higher pitch
than immediately preceding atonic syllables, and at least as high as preceding tonic
syllables),2 and therefore should be considered as stressed items. This group includes
the postnominal possessives -mí “mine,” -sí “your (s.),” -éle “his, her,” -súto “our,” and
the corresponding dative and accusative clitics (identical to the possessives except
for the 2-s -bó “your”). It should be noted from the outset that whereas in Palenquero
(as in Spanish) a pitch accent may be phonologically distinctive in polysyllabic words
(e.g. mína “mine” vs. miná “to see,” cása “house” vs. casá “to marry”), there are no

.â•… Moniño (2003:524) asserts that the difference in pitch between accented and unaccented
syllables remains constant throughout a phrase: five semitones or two and a half whole musical
tones. The data collected for the present study do not confirm this observation; the difference
between high and low pitch, normalized by F0 to equivalent intervals on the musical scale, is typi-
cally on the order of 1–3 semitones, although inter- and intra-speaker variation is considerable.
.╅ Throughout this study I use a heuristic determination of high and low pitch similar to that
stated by Prieto et al. (2005:375): “In order for a syllable to be perceived as high, the pitch level
needs to stay high or rise for a good portion of the accented syllable,“and conversely.
 Pitch polarity in Palenquero 

corresponding atonic/low pitched clitics and therefore no true minimal-pair verifica-

tion of the high pitch on the aforementioned clitics.3 The claim of a high tone is based
primarily on their consistent behavior similar to demonstrably phonological high
pitch on other words, and with their participation in pitch-polarity phenomena.4
When following an oxytonic word, phrase-final negative, possessive, and object
clitics reveal a nuanced behavior. In particular, the pitch of the final clitic is almost
always distinct from the pitch of the immediately preceding tonic syllable (at least by
the average F0 difference between other tonic and atonic syllables) and this difference
is often expressed as upstep with respect to the preceding syllable. The same obser-
vation does not hold across the board for other sequences of phrase-final non-clitic
tonic syllables. Moreover, phrase-final monosyllabic enclitics often depart from the
aforementioned level high tone ending, and may exhibit a L% (or HL%) boundary
tone, often with upstepped ¡H*on the final clitic. As with final downstepping, the pre-
cise factors responsible for the boundary tone are not yet known, but in the present
corpus this configuration is found most often in utterances that appear to be emphatic
or contrastive.5

.â•… Quasi-minimal pairs are easy to find, e.g. queléba “want-IMP” vs. quelé bó “to want you,”
etuléno “let’s study” vs. etulé nú “to not study,” with the clitics consistently higher in pitch than
the corresponding final atonic syllables.
.╅ It may be that the mora rather than the syllable is the attachment point for the H tone
(cf. the discussion in Prieto et al. 2005). For example phrase-final mái [mai ] mí “my mother”

is nearly always realized with a high tone on the nuclear vowel [a] followed by a sharp dip to a
very low tone on the offglide and a sharp rise to a high tone on mí. Similarly, phrase-final bái
[bai ] “go” always ends with a sharp tonal dip, unlike the usual absence of a low boundary tone

with other verbs ending in a single tonic vowel.
.╅ Despite their assertion of the absence of boundary tones following phrase-final oxytones,
Hualde and Schwegler (2008:12) present a pitch track of utterance-final majaná mí “my chil-
dren” in which the upstepped possessive clitic -mí is followed by a L% boundary tone; the
authors note without further explanation that -mí is “emphasized.” In Prieto (2009)’s analysis
of Catalan intonation, the HL% boundary tone is associated with obviousness statements and
requests. The fact that the Palenquero data were produced in response to a seemingly arbitrary
request may have triggered an “obviousness” response, although I have frequently observed
the same pattern in spontaneous speech. The (H)L% boundary tone is even more common
when phrase-final clitics follow an atonic syllable, and not always correlated with emphatic
or contrastive usage, but simply in line with the L% boundary tone following phrase-final
atonic syllables: éle a sendá abuélo mí “he is my grandfather” or í téng cása nú “I don’t have
a house.” The exact distribution of (H)L% boundary tones following phrase-final clitics
remains to be determined.
 John M. Lipski

3.â•… Data collection for the present study

In order to obtain a reasonable number of samples, including phrase-final combina-

tions of two or three clitics, data were obtained by means of a questionnaire. Inter-
views were conducted with twenty five young Palenquero speakers (fifteen male,
ten female), all of whom had learned Palenquero in school; some also had learned
the rudiments of the language at home by listening to older relatives, but none had
routinely spoken Palenquero before attending school. The ages of the interviewees
ranged from eighteen to twenty-two. A second set of interviews was conducted with
twenty five traditional Palenquero speakers (thirteen men and twelve women), all
considered true native speakers, with ages ranging from forty five to over ninety.6
Both sets of interviewees were recommended by the same colleagues who had assisted
in the initial data collection. Each interviewee was asked to translate a number of
sentences from Spanish into Palenquero, designed to produce the required clitic
sequences. They were also asked specific questions to elicit the appropriate clitic
combinations. Phrase-final position was chosen for analysis, to observe the interac-
tion between high pitch and final downstep, and to remove from consideration any
possible sandhi or assimilatory effects. Many of the phenomena described in the
following sections can also be observed in phrase-internal contexts, but the patterns
are not always as clear-cut due to the effects of following syllables. The bisyllabic1-pl
súto and the 3-s éle were excluded from consideration, except in the rare double
object constructions of the type ndá bó éle “to give it to you,” which normally occur
only with éle. Possible differences between traditional and younger speakers are the
subject of ongoing research and will be reported elsewhere. The following sections,
necessarily brief because of space limitations and subject to the vicissitudes of field
data collection under less than ideal conditions, present the initial findings and offer
a possible interpretation.

.â•… All Palenquero speakers were recorded in San Basilio de Palenque in 2008–2009, with the
able assistance of Víctor Simarra Reyes and Bernardino Pérez Miranda. These two colleagues
recommended the interviewees on the basis of their known fluency in Palenquero. The format
was free conversation supplemented by directed questions. I conducted all interviews, using
both Spanish and Palenquero; during each interview I was assisted by one of the aforemen-
tioned colleagues. All recordings were made on a Marantz PMD620 digital recorder using an
Audio-Technica ATR-55 directional microphone. The digitized tokens were analyzed using
PRAAT software.
 Pitch polarity in Palenquero 

4.â•… The behavior of Palenquero phrase-final clitics

In Palenquero phrase-final oxytones followed by a high-pitch clitic, three possible

configurations of adjacent tonic syllables are possible: (1) two identical high pitches
(i.e. no tonal differential); (2) an upstepped second pitch; (3) a downstepped second
pitch. In an initial rough tally, upstep or downstep was defined as a pitch differential
at least equal to the average pitch difference between word-internal tonic and atonic
syllables for a given speaker. Using this heuristic criterion, nearly all of the phrase-
final post-oxytonic enclitics undergo systematic tonal dissimilation (upstep or down-
step) with respect to the preceding tonic vowel. To give a more detailed account of the
behavior of phrase-final sequences of tonic syllables, individual pitch measurements
were taken at the midpoint of each tonic vowel,7 and the pitch value of the second
vowel was subtracted from the pitch of the first vowel. The data revealed no consistent
differences in behavior between the corresponding object and possessive clitics, and
therefore the two groups of clitics were analyzed as a single category.8 The negator
-nú showed greater variability, in particular a higher proportion of downstep, and was
analyzed separately.9 Since the average F0 of male voices is lower than the average F0
for female voices and the average difference in Hz between high and low pitch is cor-
respondingly smaller, male and female data were computed separately. Qualitatively,
male and female speakers treat phrase-final sequences of tonic syllables identically; for
purposes of illustration, only data from male speakers are included below. Figures€1
and 2 show the behavior of phrase-final sequences of two tonic vowels, the first of
which belongs to a (non-clitic) noun or verb. The data represent the 1-s object or pos-
sessive -mí (in the utterances é a sendá posá mí “that is my house” and bó quelé ablá mí
“you want to speak to me”), the negator -nú (in í a teng posá nú “I don’t have a house”
and í quelé ablá nú “I don’t want to speak”), and the non-clitic má “more” (in í quelé
ablá má “I want to speak more”).10

.╅ In those instances in which the phrase-final clitic was followed by a L% boundary tone,
the vowel of the clitic was also lengthened, and the pitch drop occurred well after the mid-
point (e.g. Figure 7), thus not affecting the measurement of pitch differential with respect to
the preceding syllable.
.╅ Two-tailed t-test: df = 90, t = 1.99, p =.90.
.â•… Two-tailed t-test comparing the behavior of object/possessive clitics vs. nú: df = 282,
t€= 1.97, p <.0005.
.â•… One-way ANOVA comparing the behavior of object/possessive clitics, nú, and non-clitic
final tonic monosyllables: F(2,178) = 7.43; p <.001.
 John M. Lipski

30

25
20
F0 difference (Hz)

15
10

5
0

–5
–10
-mi -nu no clitics

Figure 1.╇ Palenquero male voices, pitch differential of phrase-final tonic syllables with respect
to preceding tonic syllable; N = 283 in each category

250

200
-mi
-nu
150 no clitics
Frequency

100

50

0
–26 –12 1 14 28 41 54 67 81 94
F0 difference (Hz)

Figure 2.╇ Palenquero male voices, pitch differential of phrase-final tonic syllables with respect
to preceding tonic syllable; N = 283 in each category

Figure 3 shows the distribution of pitch differentials in the phrase-final combina-

tion -mí-nú (ending the utterances é a sendá posá mí nú “that is not my house” and
bó quelé ablá mí nú “you don’t want to speak to me”), demonstrating the predominant
phrase-final higher+lower pitch pattern when two clitics follow an oxytone.11

.â•… Two-tailed t-test comparing the behavior of mí and nú in -mí-nú clusters (df = 150,
t€ = 1.98, p <.0001). The higher+lower tonal pattern for -mí-nú following a high-pitched
syllable is consistent with the cross-linguistic tendency summarized by Hyman (2007:3–4) for
a high tone to be additionally raised when preceding a low tone.
 Pitch polarity in Palenquero 

15
10
F0 difference (Hz)
5
0
–5
–10
–15
–20
mi nu

Figure 3.╇ Palenquero male voices, pitch differential of phrase-final -mí -nú with respect to
preceding tonic syllable; all responses (N = 151)

5.â•… Pitch polarity of phrase-final single enclitics: Examples

A typical example of a raised pitch on a phrase-final post-oxytonic clitic is (1) and

Figure 4:12
(1) (í quelé) ndrumí nú
1-s want sleep neg
“I don’t want to sleep”

300
Pitch (Hz)

ndu mi nu
0
0 0.6001
Time (s)

Figure 4.╇ Palenquero ndrumí nú ‘not sleep’

.╅ In the transcribed examples, tonic vowels are indicated by a written accent mark. All
pitch tracks represent male speakers, unless otherwise indicated.
 John M. Lipski

Possessive clitics show a similar pattern, as in (2) and Figure 5.

(2) (éle a sendá) mamá mí

3-s asp be mother poss-1s
“she is my mother”

250
Pitch (Hz)

ma ma mi
0
0 0.6795
Time (s)

Figure 5.╇ Palenquero mamá mí ‘my mother’

Object clitics behave similarly to possessive clitics, as in (3) and Figure 6.

(3) tó ané ablá mí

all 3-pl speak 1-s
“all of them speak to me”

300
Pitch (Hz)

tua ne a βla mi
0
0 0.9171
Time (s)

Figure 6.╇ Palenquero tó ané ablá mí ‘all of them speak to me’

 Pitch polarity in Palenquero 

The phrase (í quelé) ndrumí nú “I don’t want to sleep” of Figure 4, uttered by

another speaker in response to the question bó quelé ndrumí? “do you want to sleep?”,
is shown in Figure 7; the ¡H*+ L and L% boundary tone are apparently due to the
emphatic/contradictory response.

300
Picth (Hz)

ndu mi nu
0
0 0.8878
Time (s)

Figure 7.╇ Palenquero ndrumí nú ‘not sleep’ with L% boundary tone

Figure 8 shows the downstep to a mid tone that occurs when two non-clitic tonic
syllables occur phrase finally.

(4) (ané) gobbé a miní má

3-pl return prep come more
“they come back again”

200
Pitch (Hz)

go bbe a mi ni ma
0
0 0.8449
Time (s)

Figure 8.╇ Palenquero gobbé a miní má ‘to come back again’

 John M. Lipski

6.â•… Sequences of two or more clitics: Examples

When sequences of two or more phrase-final clitics follow an oxytonic word, the situa-
tion becomes more complex. Rather than continual upstepping between a word-final
tonic vowel and a sequence of enclitics, an option rarely observed in Palenquero, what
occurs is non-monotonic pitch polarity among the successive tonic vowels. The most
common configuration is upstepped clitic + downstepped clitic. This combina-
tion represents roughly ¾ of all instances of two clitics following a final tonic syllable.
There are very few cases where the second clitic is upstepped with respect to the
first clitic, suggesting that Palenquero speakers are frequently employing the normal
phrase-final downstep to effect the pitch differential between the two clitics. A typical
example is (5) and Figure 9:

(5) (éle a sendá) tatá mí nú

3-s asp be father poss-1s neg
“he is not my father”

250
Pitch (Hz)

ta ta mi nu
0
0 0.7796
Time (s)

Figure 9.╇ Palenquero tatá mí nú ‘not my father’

Two-clitic combinations not containing the negator -nú are relatively uncommon,
the most frequent involving the combination of the 1-s. or 2.s pronouns mí or bó and
the 3-s. pronoun éle; in these configurations, the first clitic is usually upstepped with
respect to a preceding tonic syllable, while éle is downstepped, as in Figure 10, repre-
senting the phrase:

(6) (í tan) kitá bó éle

1-s fut remove 2-s 3-s
“I will take it off you”
 Pitch polarity in Palenquero 

250
Pitch (Hz)

ki ta βo e le
0
0 0.59
Time (s)

Figure 10.╇ Palenquero kitá bó éle ‘to take it off you’

When two phrase-final object clitics are followed by the negator nú, final nú may
be upstepped, as in Figure 11 (pronounced by a female speaker), representing:
(7) (í) tán vendé bó éle nú
1-s fut sell 2-s 3-s neg
“I will not sell it to you”

250
Pitch (Hz)

tam ben de βo e le nu
0
0 1.613
Time (s)

Figure 11.╇ Palenquero tán vendé bó éle nú ‘will not sell it to you’

7.â•… The OCP and a constraint-based analysis of Palenquero tonal polarity

The tonal differential correlated with the morphosyntactic status of the clitics sug-
gests that these clitics are underlyingly specified with a H tone, whereas the high pitch
normally associated with Palenquero tonic syllables in non-clitic words is a purely
 John M. Lipski

phonetic realization, as proposed by Hualde and Schwegler (2008). This is similar to

the polarity phenomena in Papiamentu as described by Römer (1977, 1980, 1983;
also Rivera-Castillo 1998; Rivera-Castillo & Faraclas 2006; Rivera-Castillo & Pickering
2004; Remijsen & van Heuven 2005), in which prepositions unspecified for tone
exhibit tonal polarity before syllables with underlying high or low tones.13
The robustness of tonal dissimilation involving Palenquero enclitics argues in
favor of a distinctive H tone lexically associated with these items. This behavior bears
the signature of the Obligatory Contour Principle (OCP), disallowing adjacent iden-
tical elements (McCarthy 1986; Yip 1988; Frisch, Pierrehumbert & Broe 2004).14 By
extension, then, non-final tonic syllables could be analyzed as lexically specified for H,
or alternatively, as having H associated through an implementation rule that interprets
the stressed syllable as H. The latter is more likely, given the absence of minimal pairs
based entirely on tone15 and also the observed variability of the pitch of phrase-internal
stressed syllables. Using an Optimality Theory approach, the following constraints will
be invoked:

OCP: This is the familiar Obligatory Contour Principle, disallowing adjacent identi-
cal phonological specifications. Here, as in Pulleyblank (2004), the OCP is assumed
to be operative only in the combinations final stressed syllable + clitic or
clitic + clitic.
Max [h]: This constraint requires that an underlying H tone be preserved. As has been
postulated for other African languages as well as for pitch accent languages (de Lacy
2002; Pulleyblank 2004), H tone is assumed to be more harmonic than other tones
(in the case of Palenquero, a default and phonologically underspecified L tone). This
constraint will only operate on clitics, phonologically specified as H. Other stressed
vowels receive a H pitch through phonetic implementation rules, but do not contain
this specification in their lexical representation.

.╅ Rivera-Castillo and Pickering (2004:262) also observe that Papiamentu syllables carrying
both tonic stress and H tone are realized with a higher pitch than syllables only specified for
H. If Palenquero phrase-final clitics are analyzed as carrying H, then the tendency to realize
these clitics with a higher pitch than other tonic syllables is consistent with Papiamentu.
.╅ A homologous case occurs in Yoruba, where postverbal clitics routinely participate in
phonologically-motivated tonal dissimilation with respect to the immediately preceding
syllable. (Akinlabi & Liberman 2000; Pulleyblank 2004). In other circumstances, Yoruba does
not exhibit OCP effects; the host+clitic situation is unique.
.â•… Except for a few isolated examples such as ma ngómbe “the cows” vs. má ngómbe
“more cows.”
 Pitch polarity in Palenquero 

*Hd/L: This constraint disallows a L tone on the head of a prosodic foot (de Lacy 2002).
This constraint deals with pitch, not a phonologically specified tone, and will therefore
apply to the pitch accent associated with tonic vowels in non-clitic Palenquero words.

Given a combination such as ablá nú “not speak,” the preferred choice with the tone of
nú raised above that of the final vowel of ablá is selected as in Tableau 1. Rather than
representing an H tone and a “Higher” tone, upstepping of the clitic is graphically
represented in the tableau with no written pitch accent, in effect delinking the phoneti-
cally implemented H tone resulting from the word-final pitch accent.

Tableau 1.╇ Palenquero ablá nu ‘not to talk’

/ablá nú/ speak-NEG OCP M[H] *H/L
ablá nú !*
� abla nú *
{�}ablá nu {!}*
abla nu !* *

The data in the corpus show that less frequently OCP violations are avoided
through lowering the pitch of final Â�-nú rather than the final stressed vowel of the pre-
ceding word. This shows that the ranges of the constraints max [h] and *hd/l partially
overlap (e.g. Hayes 2000) in a fashion that quantitatively prefers the ranking max [h]
>> *hd/l but which occasionally produces the opposite ranking.16
Two-clitic combinations following word-final tonic syllables are more complex,
although the most common pattern across various grammatical configurations is
upstep + downstep or upstep + dip with final downstep. Whereas downstep of a
tonic clitic is preferred in the final position of tonic vowel + two clitic combinations,
downstep is a less preferred option when only a single clitic follows a word-final tonic
syllable. This suggests that the immediate transition between a phonetically realized
high tone on a tonic syllable and a phonologically H tone on an enclitic is the prefer-
ential environment for upstep, whereas downstep takes precedence in the second of
two phonologically H tone clitics. An additional constraint is required to depict this
dependence, and will be provisionally designated as cliticupstep. Tableau 2 shows
the results for the combination ngutá mí nú “not please me.”

.â•… Hayes defines “strictness bands” for each constraint, which can in principle overlap par-
tially or totally. Each actually occurring token is defined by “selection points” which are “the
particular value of strictness taken on by a constraint on a given speaking occasion” (Hayes
2000:€90). This approach explicitly allows for transitivity effects during speech production: “It
can be seen that as long as the strictness bands of two constraints overlap, then both rankings
of the two constraints will be available for the generation of outputs” (Hayes 2000:€90).
 John M. Lipski

Tableau 2.╇ Palenquero ngutá mí nú ‘not please me’

/ngutá mí nú/ OCP C-U M[H] *H/L
please-me-NEG
ngutá mí nú !**
ngutá mí nu !* *
ngutá mi nú !* *
ngutá mi nu !* **
nguta mí nú !* *
nguta mi nu !* ** *
� nguta mí nu * *
nguta mi nú !* * *

8.â•… Against a floating-tone analysis

The fact that the Palenquero tonal shifts occur at clitic boundaries suggests that the tonal
shift may perform a morphosyntactic function, such as has been proposed for high-
tone raising in the Afro-European (English and Portuguese-derived) creole language
Saramaccan by Good (2003, 2004); also Ham (1999), Kramer (2007). In Saramaccan, a
raised H tone at either edge of a serial verb construction potentially enhances sentence
processing: “The HR morpheme [H tone at the right edge of a word: JML] fills the
role of an indicator that there will be at least one other verb in the predicate, in effect
“warning” that there is more to come in the sentence than might be expected. The HL
[H tone at the left edge of a word: JML] morpheme […] can be understood as “remind-
ing” the hearer that a serial verb phrase is being constructed” (Good 2003:€111). The
morphemes in question are analyzed as floating H tones that attach to the appropriate
edge of verbs in the serial constructions. The tonal shift between a Palenquero word-
final tonic vowel and a following clitic could be similarly construed as a “warning”
or “reminder” of morphosyntactic status, by means of a floating tone. Although the
Palenquero data are not entirely inconsistent with such an approach, several facts mili-
tate against adopting this solution. First of all, although the term “clitic” is used in the
present study, the morphosyntactic role of the negator -nú (which can occur at the end
of any type of phrase irrespective of the internal syntactic configuration of the phrase)
is very different than the possessive morphemes -mí and -sí (which attach directly to
the possessed noun), and the object clitics -mí, -bó, éle, and súto, which attach directly
to verbs. The syntactic role of -nú is similar to that of the sentential adverbs má “more”
and yá “already, now,” both of which carry high pitch and neither of which undergoes
the tonal shifts associated with -nú. Similarly, -mí and -sí are in effect postnominal
adjectives, just like high-pitched ngánde “big” and ngólo “fat,” which do not partici-
pate in tonal shifts when following a noun ending in a tonic vowel. Object nouns
beginning with tonic vowels also do not exhibit tonal dissimilation with respect to
 Pitch polarity in Palenquero 

a verb-final tonic vowel, as in miná ngómbe “see cow(s).” Moreover, when occurring
after word-final atonic vowels, negative, possessive, and object clitics simply emerge
with high tone, with no effect on the preceding vowel, which might be expected if a
floating H€tone were present. Finally, the fact that phrase-final post-oxytonic clitics
sometimes lower in pitch with no corresponding upstep of the preceding tonic vowel
does not suggest the presence of a floating H tone. Despite these considerations, the
possibility of floating tones in Palenquero deserves further investigation, since if dis-
tinctive H€tones are present in the language, floating tones may develop, e.g. during
fast-speech vowel reductions.

9.â•… Palenquero tonal polarity: African leftover or innovation?

Remaining to be addressed is origin of the tonal behavior of Palenquero enclitics, in

particular whether this represents a carryover from one or more African languages
(or from an earlier stage of the creole language) or a more recent innovation. Hualde
and Schwegler (2008) draw on data from older speakers, born in the early years of
the 20th€ century, and taken as representative of the most traditional form of the
Palenquero language. The present study also included some speakers from the same
age range, but also speakers in their teens. All produced the tonal displacement of
phrase-final clitics, but with some qualitative differences. Many of the examples of
downstepped post-oxytonic phrase-final clitics were produced by the oldest speakers,
speaking non-emphatically. For the same older speakers, upstepped negative -nú was
most often found in emphatic speech, with either implicit or explicit contradiction, i.e.
when the opposite response is presumed. Upstepped -nú would have an obvious prag-
matic advantage, with the raised pitch highlighting the unexpected negative element at
the end of a phrase which in absence of the negator could be construed with the oppo-
site meaning. Younger speakers may be in the process of generalizing upstepped -nú to
non-emphatic utterances, in effect creating a stable tonal distinction that is essentially
independent of pragmatic criteria. The same would be true for possessive clitics and
object pronouns, all of which are realized by younger speakers predominantly with an
upstepped tone when occurring phrase-finally. Among older speakers the behavior
of phrase-final possessive and object clitics is more variable, but there is some cor-
relation between upstepped tone and emphatic or contrastive environments. These
preliminary observations suggest that younger speakers may prefer an upstepped pitch
for phrase-final clitics in all pragmatic contexts, representing an innovation from the
traditional dialect, in which upstepped final high pitch is correlated with emphatic or
contrastive contexts. Further research is required in order to clarify this conjecture.
The preceding considerations notwithstanding, a possible substrate contribu-
tion cannot be totally excluded, at least in the case of the phrase-final negator -nú.
 John M. Lipski

The presence of numerous Kikongo lexical items in Palenquero suggests that Kikongo
was once spoken in the emerging maroon community, although Hualde and Schwegler
(2008:26) find no evidence of the complex Kikongo tonal sandhi phenomena (e.g.
Odden 1994) in Palenquero. However Kikongo uses the negative particles ke … kó or
ka … kó. The first element of the pair bears a low tone, as does Palenquero preverbal no
in the instances where double negation is found, while Kikongo phrase-final -kó bears
a high tone (Lumwamu 1973:€213–15; Odden 1994:€184). A calque of the Kikongo
pattern – both morphosyntactically and tonally – cannot be ruled out, especially in
view of Schwegler’s (1986:174; also 1983, 1988) analysis which suggests that double
negative patterns eventually evolve to postverbal negation only. This pattern could
have eroded partially as the proto-Palenquero preverbal negative gradually fell into
disuse, ultimately resulting in the somewhat ambiguous use of -nú among the oldest
Palenquero speakers: always high-toned when emphatic, but sometimes carrying a
high tone when no emphatic or contrastive function can be plausibly proposed.17

References

Akinlabi, Akinbiyi & Mark Liberman. 2000. “The Tonal Phonology of Yoruba Clitics.” Clitics in
Morphology and Syntax, ed. by Birgit Gerlach & Janet Grijzenhout, 31–62. Amsterdam &
Philadelphia: John Benjamins.
De Lacy, Paul. 2002. “The Interaction of Tone and Stress in Optimality Theory.” Phonology 19.1–32.
Frisch, Stefan, Janet Pierrehumbert & Michael Broe. 2004. “Similarity Avoidance and the OCP.”
Natural Language and Linguistic Theory 22.179–228.
Good, Jeff. 2003. “Tonal Morphology in a Creole: High-tone raising in Saramaccan serial verb
constructions.” Yearbook of Morphology 2002, ed. by Geert Booij & Jaap van Marle, 105–134.
Dordrecht: Kluwer.
Good, Jeff. 2004. “Tone and Accent in Saramaccan: Charting a deep split in the phonology of a
language.” Lingua 114.575–619.
Ham, William. 1999. “Tone Sandhi in Saramaccan: A case of substrate transfer?” Journal of
Pidgin and Creole Languages 14.45–91.
Hayes, Bruce. 2000. “Gradient Well-Formedness in Optimality Theory.” Optimality Theory:
Phonology, Syntax, and Acquisition, ed. by Joost Dekkers, Frank van der Leeuw & Jeroen van
de Weijer, 88–120. Oxford: Oxford University Press.
Hualde, José Ignacio & Armin Schwegler. 2008. “Intonation in Palenquero.” Journal of Pidgin
and Creole Languages 23.1–31.

.╅ Kikongo pronominal possessives, while postposed as in Palenquero, exhibits a more

complex morphophonemic behavior. Postposed pronouns take a monosyllabic class prefix,
some of which carry a high tone, while others are low-toned (e.g. Lumwumu 1973:€133–136).
It is therefore less likely that the tendency to realize Palenquero phrase-final -mi and -si with
upstepped high tones can be directly attributed to a carryover from Kikongo.
 Pitch polarity in Palenquero 

Hyman, Larry. 2007. Universals of Tone Rules: 30 years later. Tones and tunes: volume 1, ed. by
Tomas Riad & Carlos Gussenhoven, 1–34. Berlin & New York: Mouton de Gruyter.
Kramer, Maarvin. 2007. “Tone on Quantifiers in Saramaccan as a Transferred Feature from
Kikongo.” Synchronic and Diachronic Perspectives on Contact Languages, ed. by Magnus
Huber & Viveka Velupillai, 43–66. Amsterdam & Philadelphia: John Benjamins.
Lumwumu, François. 1973. Essai de morphosyntaxe systématique des parles Kongo. Paris:
Klincksieck & Brazzaville: Université de Brazzaville.
McCarthy, John. 1986. “OCP Effects: Gemination and antigemination.” Linguistic Inquiry
17.207–63.
Moniño, Yves. 2003. “Lengua e identidad afro-americana: el caso del criollo de Palenque de
San Basilio (Colombia).” Caminos cruzados: ensayos en antropología social, etnoecología y
etnoeducación, ed. by Catherine Alès & Jean Chiappino, 517–531. Paris: IRD-Éditions &
Mérida, Venezuela: Universidad de los Andes, GRIAL.
Odden, David. 1994. “Syntactic and Semantic Conditions in Kikongo Phrasal Phonology.”
Perspectives on Phonology, ed. by Jennifer Cole & Charles Kisseberth, 161–201. Stanford: CSLI.
Prieto, Pilar. 2009. “The Intonational Phonology of Catalan.” Prosodic Typology 2, ed. by Sun-Ah
Jun. Oxford: Oxford University Press.
Prieto, Pilar, Mariapaola D’Imperio & Barbara Gili Fivela. 2005. “Pitch Accent Alignment in
Romance: Primary and secondary associations with metrical structure.” Language and
Speech 48.359–396.
Pulleyblank, Douglas. 2004. “A Note on Tonal Markedness in Yoruba.” Phonology 21.209–425.
Remijsen, Bert & Vincent J. van Heuven. 2005. “Stress, Tone, and Discourse Prominence in the
Curaçao Dialect of Papiamentu.” Phonology 22.205–235.
Rivera-Castillo, Yolanda. 1998. “Tone and Stress in Papiamentu: The contribution of a constraint-
based analysis to the problem of Creole genesis.” Journal of Pidgin and Creole Languages
13.1–38.
Rivera-Castillo, Yolanda & Nicolas Faraclas. 2006. “The Emergence of Systems of Lexical and
Grammatical Tone and Stress in Caribbean and West African Creoles.” Sptrachtypologie
und Universalienforschung 59: 2.148–169.
Rivera-Castillo, Yolanda & Lucy Pickering. 2004. “Phonetic Correlates of Stress and Tone in a
Mixed System.” Journal of Pidgin and Creole Languages 19.261–284.
Römer, Raúl. 1977. “Polarization Phenomena in Papiamentu.” Amsterdam Creole Studies
1.69–79.
Römer, Raúl. 1980. “Proclisis y enclisis en una lengua tonal: algunas comparaciones entre el
papiamento y el español.” Diálogos Hispánicos de Amsterdam 1.113–123.
Römer, Raúl. 1983. “Papiamentu Tones: How speakers use them and how they bind speakers.”
Papiamentu problems and possibilities, 85–96. Zutphen: De Walburg Pers.
Schwegler, Armin. 1983. “Predicate Negation and Word-Order Change: A problem of multiple
causation” Lingua 61.297–334.
Schwegler, Armin. 1986. Analyticity and Syntheticity: A diachronic perspective with special reference
to Romance languages. Berlin: Walter de Gruyter.
Schwegler, Armin. 1988. “Word-Order Changes in Predicate Negation Strategies in Romance
Languages.” Diachronica 5.21–58.
Yip, Moira. 1988. “The Obligatory Contour Principle and Phonological Rules: A loss of identity.”
Linguistic Inquiry 19.65–100.
 Word-minimality and sound change
in Hispano-Romance*

Fernando Martínez-Gil
The Ohio State University

This paper proposes that word minimality, correlated with a bimoraic foot,
provides a unitary account of three separate historical developments in the
Hispano-Romance languages, to date considered unrelated: the preservation
of a word-final nasal, the change /ee/ > /ej/, and yod augmentation in four
present 1st person verb forms: soy, doy, voy and estoy. It is suggested that two
otherwise regular sound changes, -m deletion and identical vowel simplification
(/ee/€>/e/), failed to occur in a small number of lexical items in order to avoid
their impeding reduction to a subminimal size, and that yod accretion in the
Old Spanish monosyllabic forms do, so, vo, and estó was also driven by
compliance with a minimal word requirement. In support of this analysis is the
empirical evidence from the historical evolution of Hispano-Romance, which
points to a weight-sensitive metrical system in which the bimoraic foot can be
identified as the minimal prosodic word.

1.â•… Introduction

It is well known that many languages impose a minimal size restriction on lexical
words, which can be identified with the canonical metrical foot. This paper explores the
role of word minimality on sound change in the historical development of Hispano-
Romance (hereafter, HR), focusing primarily on its role in altering the regular outcome
of two HR sound changes: (a) the deletion of Latin word-final -m; and (b) the simpli-
fication of a sequence of identical vowels.1 An analysis in terms of word minimality is

*I am grateful to three anonymous reviewers, whose valuable comments and criticism have
significantly improved the final version of this paper. All errors are my own.
.╅ In this paper, the term Hispano-Romance is used to designate the Central and Western
branches of Ibero-Romance, consisting of the three major Ibero-Romance varieties: Galician-
Portuguese, Leonese, and Castilian (= Spanish).
 Fernando Martínez-Gil

shown to provide a simple and direct explanation for the failure of both word-final -m
deletion and identical vowel simplification to take place in the transition from Latin
to HR whenever the result would have been subminimal lexical word. Instead of the
regular outcome, certain HR forms exhibit alternative historical developments, which
can be readily analyzed as strategies to avoid their impending reduction to a sub-
minimal prosodic size. Thus, whenever these two sound changes would have resulted
in a monomoraic word, a word-final nasal was preserved, not deleted, in a number of
monosyllabic words, and instead of identical vowel simplification, the second member
in a sequence of identical mid-front vowels underwent desyllabification, becoming a
glide: /ee/ > /ej/. Another strategy to deal with identical vowel sequences is found in
Old Galician-Portuguese, where such sequences were maintained for a relatively long
period in disyllabic words, when otherwise regular simplification would have resulted
in violations of prosodic minimality. To this date, such exceptional developments have
remained unexplained.
In addition to the blocking effects on the two sound changes just outlined, it is sug-
gested that word minimality requirements are also at work in promoting, rather than
blocking, another HR sound change, the accretion of a word-final glide in a handful of
monosyllabic 1st person singular present indicative verb forms, a sound change that
took place in the transition from Old to Modern HR, turning the early medieval HR
forms so “I am”, do “I give”, vo “I go”, and estó “I am, stand”, into Modern Spanish soy,
doy, voy, estoy, and Galician-Portuguese sou, dou, vou, estou.
The facts just outlined are reminiscent of a common manifestation of minimal pro-
sodic word requirements in synchronic grammars; namely, blocking effects on some
phonological process emerge when such process would truncate the word below a
minimum size, thereby yielding a form smaller than a binary foot. The proposed
analysis claims that the word-minimal prosodic requirement that impelled such
alternative historical developments can be formally identified as a bimoraic foot, not
disyllabic one. If the minimal word coincides with the metrical foot, as alleged by the
theory of Prosodic Morphology, such a result refutes the hypothesis of a disyllabic
quantity-insensitive foot as the basic metrical unit throughout the historical develop-
ment of HR, and up to the emergence of medieval apocope (Holt 1997; Lipski 1997).
The implications of the present analysis are twofold. From the point of view of the
linguistic history of the HR languages, it presents a unified prosodic account of three
historical developments that traditionally have been viewed as unconnected sound
changes. From a more general perspective, it presents an illustration of how sound
change can be driven by prosodic constraints. Thus, a general thrust of the account put
forth in this paper is that while word minimality may not have the clear-cut status of an
absolute restriction on the size of lexical items throughout the historical development
of HR, it still represents a force to be reckoned with in historical change.
 Word-minimality and sound change in Hispano-Romance 

The rest of this paper is organized as follows. Section 2 presents a brief introduction
to the minimal word and how it relates to the metrical foot in Prosodic Morphology.
Section 3 surveys the impact of word minimality on three historical developments
in HR, inhibiting word-final -m deletion and identical vowel simplification in some cases,
and promoting yod augmentation in others, providing crucial empirical evidence for
the minimal word as a bimoraic foot. Section 4 surveys the basic facts of Latin prosody,
and argues for the preservation of the moraic trochee as the canonical metrical foot in
the transition to HR. Section 5 discusses some apparent exceptions to word-minimal-
ity in modern HR. Finally, Section€6 contains some concluding remarks.

2.â•… The minimal word in Prosodic Morphology

The notion of word minimality attempts to capture the empirical observation that
morphological units in many languages must fulfill a minimal-size requirement, which
can be readily identified as prosodic in nature. A principled account of word minimality
within the theory of Prosodic Morphology crucially appeals to the universal Prosodic
Hierarchy (1a) (cf. Nespor & Vogel 1986; McCarthy & Prince 1990, 1993, 1995a, 1995b,
1998, 1999; Inkelas 1990; Itô 1990; Nespor 1999, and much related work), in conjunc-
tion with strict Foot Binarity (1b) (McCarthy & Prince 1990, 1993, 1995a; Mester 1994;
Hayes 1995). The category Prosodic Word embodies the correlation between phono-
logical and morphological structure.

(1) a. Prosodic Hierarchy: b. Foot Binarity:

PrWd (Prosodic Word) Feet are binary under either a
syllabic or a moraic analysis
F (Foot)

σ (Syllable)

μ (Mora)

In the moraic theory of the syllable (Hyman 1985; Hayes 1989a; Broselow 1995), the
mora is the timing unit that mediates between the syllabic and the segmental levels. As a
stress-bearing unit, each vowel is associated with a mora; moreover, in some languages,
such as Latin, a consonant or a glide located in the syllable coda receives a mora by posi-
tion (Allen 1973; Hayes 1989a; Mester 1994; Sherer 1994).
As shown in (2), moras encode syllable weight. In quantity-insensitive languages
all syllables are monomoraic (i.e. light). In languages sensitive to syllable weight, a light
syllable (i.e. an open syllable containing a short nuclear vowel), as in (2a), is monomoraic;
 Fernando Martínez-Gil

a heavy syllable (one with a long nuclear vowel or closed by a consonant or glide), as in
(2b), is bimoraic:2

(2) a. Light syllable: b. Heavy syllables:

σ σ σ σ

μ μ μ μ μ μ μ

p a p a˜ p a s p a j

Three major types of binary metrical feet, shown in (3), have been proposed in the pro-
sodic phonology and morphology literature (cf. Hayes 1995; McCarthy & Prince 1993,
1995a, 1995b,1998):
(3) Foot types under strict Foot Binarity (H = heavy syllable; L = light syllable)
a. Moraic Iamb: (μ μ́) = LH, H, LL L (quantity-sensitive systems)
b. Moraic trochee: (μ́ μ) = H, LL
c. Syllabic trochee: (σ́ σ) (quantity-insensitive systems)

The formal distinction between the two types of prosodic feet relevant for our purposes
is illustrated in (4) with the Spanish words casa “house”, pan “bread”, and rey “king”. As
shown, all three can be parsed as a bimoraic trochee, as shown in (4a), but, for obvious
reasons, only casa would qualify as a syllabic trochee (4b):

(4) a. Bimoraic foot b. Disyllabic foot

(= moraic trochee): (= syllabic trochee):
F F

μ μ σ σ

ká sa ká sa
pá n * pán
r- é j r- é j

.╅ In some languages (e.g. Modern Spanish), lexical diphthongs are counted as heavy (Harris
1983, 1995; Dunlap 1991; Roca 2006). Whether this was the case also in the medieval period
remains to be determined. As an anonymous reviewer points out, under this hypothesis, and
given the binary limit on foot size, it is unclear what the precise moraic status of a triphthong,
as in buey “ox”, would be.
 Word-minimality and sound change in Hispano-Romance 

In Prosodic Morphology the notion of “minimal word” is derived directly from the
combination of the Prosodic Hierarchy and strict Foot Binarity (see, e.g. McCarthy€&
Prince 1995b: 321). Thus, given the premises: (a) the prosodic word immediately domi-
nates the foot in the Prosodic Hierarchy; and (b) the foot is universally binary, it follows
that the minimal prosodic word must coincide with the metrical foot, and that mono-
moraic feet are either highly marked or prohibited outright, since they do not belong
to any of the metrical types of the universal inventory in (3). The restriction that the
minimal prosodic word be coextensive with the metrical foot, dubbed here the Prosodic
Minimality Condition, is formulated schematically in (5) following (Orie€& Pulleyblank
2002:€108):

(5) Prosodic Minimality Condition (PMC):â•… PrWdmin = Foot

(the minimal prosodic word coincides with a metrical foot)

It is important to point out that while the PMC is an absolute and inviolable restriction
in some languages (Latin and Modern English are well-known examples), in other lan-
guages it is only a strong tendency which reflects the unmarked case. The presence of a
handful of subminimal forms in Modern Spanish, such as fe, va “(s)he goes”, ve “(s)he
sees”, dé “I give-subj.”, di “I gave”, etc. would seem to indicate that Spanish belongs to the
latter category. In Section€5 below, I discuss some evidence from the contemporary HR
languages which suggests that prosodic word minimality may not be infringed at all in
such apparently monomoraic forms.
For over two decades now, word minimality has played a critical role in the study of
a wide range of phenomena in the phonology-morphology interface, including templa-
tic morphology, reduplication, prosodic circumscription, word truncation, word aug-
mentation by epenthesis, and blocking of apocope (cf. McCarthy & Prince 1990, 1993,
1995a, 1995b; Orie & Pulleyblank 2002; Downing 2005, 2006, and references therein).
While such work has significantly advanced our understanding of prosody-related phe-
nomena in synchronic grammars, relatively little research has been devoted to the role
of word minimality in sound change. The present paper represents an attempt to fill
this gap.

3.â•… The evidence for the minimal prosodic word as a bimoraic foot in HR

We turn now to the three HR sound changes claimed here to have been affected by
a prosodic constraint that can be identified with a bimoraic minimal word, but not
with a disyllabic one. The first of these changes is the loss of word-final -m, which
typically marked the Latin accusative, the case generally assumed to be the source
 Fernando Martínez-Gil

of most Romance non-verb forms. The nasal was subject to regular deletion in the
transition to HR, as illustrated by the following representative examples from the
five Latin conjugational classes: a↜r a neam “spider” > Gal., Sp. araña, Port. aranha;
amicum “friend” > Gal., Port., Sp. amigo vinum “wine” > Gal. viño, Port. vinho, Sp.
vino; noctem “night” > Gal., Port. noite, OLeon. nueyte, nue(y)che, Sp. noche; ma num 
“hand” > Gal. man, mao, Port. mão, Sp. mano.3 There are, however, a handful of excep-
tions to word-final nasal deletion in the HR, listed in (6), which consistently involve
monosyllabic forms.4
(6) quem “who” > Gal. quen, Port. quem [kẽ↜J̃], Sp. quien
re m “thing” > OGal. ren, OPort. rem “nothing” (from rem (natam)
lit. “born thing”)
cu m “with” > Gal., Sp. con, Port. com [kõ];
tam “so”> Gal., Sp. tan, Port. tão [tãw̃↜] (but cf. iam “already” > Gal. xa,
Port. ja, Sp. ya)
While most treatises on the history of the HR languages mention the failure of the final
consonant to delete in such monosyllables, in our knowledge no attempt to explain
such exceptional behavior has been made to date. If a bimoraic foot is proposed as the
minimal prosodic word, the preservation of the word-final nasal simply follows from
the fact that deletion would have rendered these words monomoraic, and thus prosodi-
cally subminimal.
One might suspect that deletion of final -m was morphologically conditioned;
namely, that it applied only just in case the nasal was the accusative marker or the first
person marker in verbs (e.g. amaba-m “I used to love”). Since the final nasal in con
and tan does not derive from the accusative suffix, the failure of -m to delete would be
thereby explained. Notice, however, on the one hand, that rem and quem are indeed

.╅ Here and in subsequent examples throughout this paper, the following abbreviations are
used: VLat. = Vulgar Latin; Gal. = Galician; Leon. = Leonese, Port. = Portuguese; Sp. = Spanish;
M = Modern; O = Old.
.â•… The forms ren, rem, are well attested in Old Galician-Portuguese (García de Diego 1984
(1909): 15, 43, 101; Ferreiro 1999:€144, 270; Williams 1947:€97, 121, 158; but cf. the deletion of
the nasal in OSp. so. The variants soo, soon are documented in Old Galician (García de Diego
1984 (1909): 43), and so are sõ, sõo, and são in Old Portuguese (Williams 1938:€233). That some
function words can occasionally exhibit prosodic word behavior is well documented (Selkirk
1995). In both Gal. and Port. there are some examples of non-lexical words that have under-
gone some sort of moraic augmentation, which is likely to be the result of a word-minimality
effect, since deletion of the final consonant would have rendered them monomoraic: nec
“no(t)” > *ne > Gal. nen, Port. nem [nẽ↜J̃], sic “thus” > *si > Port. sim [sĩ] “yes”, pro “for” >
Sp., Gal., Port. por, que > dial. Leon. quéi [kéj] “what” (García Arias 2003:€97). Interestingly,
although OSp. had two interrogative pronouns qui (< qui) “who” and quien (< quem)
“who(m)”, monomoraic qui was dropped around the 15th century, and only bimoraic quien
remains in Modern Spanish.
 Word-minimality and sound change in Hispano-Romance 

accusatives, and on the other, there are some polysyllabic nonlexical words, such as
numquam (= ne + umquam) “never” > Sp. nunca, in which the final -m was not a case
marker, and yet it still underwent deletion.
The second sound-change claimed in this paper to have been inhibited by word-
minimality is the simplification of two adjacent identical vowels, a sound change
arguably driven by hiatus resolution throughout the history of HR. The process is
well attested in Vulgar Latin, including the Appendix Probi. Grandgent (1907:€94–96)
provides abundant examples of the simplification of i i > i and u↜u > u in Vulgar Latin:
audii (for audivi) > audi “I heard”; julii > Juli “Julius-gen.”; consilii > consili
“consultation-gen.”; e qu us > ĕcus “horse”; antiqu us > anticus “old”; co qu us > cŏcus
“cook”, etc. (see also Väänänen 1968:€84–88, among others). Identical vowel simplifica-
tion was widespread, both throughout the historical development of HR (where such
sequences quite generally came about by the deletion of an intervening consonant),
and in the transition from the old to the modern HR languages: fastidiu “loathe,
disgust” > *fastiio > OSp. fastío > MSp. hastío “weariness”; digitu “finger” > *degedo >
*deedo€> Sp., Gal., Port. dedo; sigillu “small figure, image” > OSp. seello, MSp., MGal.
sello, Port. sêlo “seal”; prehendere “to seize, grasp” > preendere > prendere > Sp. prender;
fide > *fe[ð]e > OSp., OGal.-Port. fee > MSp., MGal., MPort. fe. “faith”; co o perire “to
cover” > OSp., OGal.-Port. cobrir > MSp. cubrir; articu lu “small joint” > OGal-Port.
artigoo > MGal, MPort. artigo “article”; do lo re “pain” > OGal-Port. door > MGal.,
MPort. dor; viginti “twenty” > OGal.-Port. viinte > MGal, MPort. vinte; vigila re
“to watch” > *vegelare > *veelare > Sp., Gal. Port. velar; palatiu “palace” > OGal.-Port.
paaço> MGal., MPort. paço; frigidu “cold”> *friidu > fridu (attested in Pompeii;
Menéndez Pidal 1980:€ 94) > Sp., Gal.-Port. frío, etc… Simplification of identical
vowels is also pervasive in modern colloquial Spanish, especially when such sequences
are intramorphemic (alcohol [alkól] “alcohol”, zoología [boloxía] “zoology”), or when
they straddle word boundaries (Navarro Tomás 1977:€152).
Significantly, however, identical vowel simplification in HR did not take place in
sequences of identical front mid vowels, whenever the result would have been a mono-
moraic word; instead, the second vowel underwent desyllabification, becoming a glide,
as shown in (7):5
(7) lege “law” > lee > OLeon. llei, Sp. ley (y = [j]), Gal, Port. lei
rege “king” > ree > Sp. rey, OLeon., Gal., Port. rei
grege “herd” > gree > Sp. grey, Leon., Gal., Port. grei
bo ve “ox” > OLeon. buee > buee MLeon., Sp. buey (OGal.-Port. boe > boi)
sede “seat” > OGal-Port. see, sei > MGal., MPort. sé “headquarters”

.╅ The intermediate forms lee, ree, gree, and buee, attested in the earliest and Galician-
Portuguese Leonese documents, reflect the early loss of the intervocalic consonant prior to
gliding of the final vowel (Menéndez Pidal 1980:€79; García de Diego 1984 (1909): 31, 41, 43).
 Fernando Martínez-Gil

It is instructive to note that no other known instances of the change /ee/ > /ej/ are
otherwise found in the history of HR. On the other hand, the outcome in (7) is par-
ticularly striking for Spanish, a language in which falling diphthongs quite regularly
underwent monophthongization in its early (i.e. preliterary) history, not only /ej/ >/e/
(e.g. laicu “layman” > *laigo > Gal., Port., OLeon. leigo, Sp. lego; cant avi “I sang” >
*cantai > Gal.-Port. cantei, OLeon. cantey, Sp. canté; ba seu “kiss” > *baiso > Gal. beixo,
Port. beijo, OLeon. beiso, beixu, Sp. beso; ca seu “cheese” > *caiso > Gal. queixo, Port.
queijo, OLeon. queiso, Sp. queso; materia “matter” > *mateira > Gal., Port. madeira,
Sp. madera “wood”, etc.), but also /oj/ > /o/ (no cte “night” > Gal., Port. noite, OLeon.
nueche, nueyche, Sp. noche; tru cta “trout” > OGal.-Port. truita, MGal. troita, truita,
MPort. truta, Sp. trucha; mu ltu “much” > OGal.-Port. muito, MGal. moito, muito,
MPort. muito, Sp. mucho), and /ow/ > /o/ (alteru “(an)other” > *aut(e)ro > Gal., Port.
outro, Sp. otro; falces “sickle, scythe” > *fauce > Gal., Port. fouce, Sp. hoz; saltu “forest
pasture” > *sauto > Gal., Port. souto, Sp. soto “grove, thicket”; causa “cause, reason” > Gal.,
Port. cousa, Sp. cosa “thing”; auru “gold” > Gal., Port. ouro, Sp. oro). Such diphthongs
have been preserved in modern Galician and Portuguese (indeed, they are the most
common type of diphthongs), but they are extremely rare in Spanish (in fact, /ow/
does not occur at all); a conspicuous exception are the items in (7).6 Apparently, the
need for prosodic words to preserve a minimal size effectively impeded the notorious

Of course, an alternative to both desyllabification and simplification would be to preserve

both vowels unaltered. In fact, this outcome is found in several verb forms, in which para�
digmatic factors appear to have favored the preservation of both vowels. Thus, an Old
Spanish (heterosyllabic and heteromorphemic) identical vowel sequence has survived in
two common modern Spanish verbs, both in the infinitive: creer “to believe” (< credere),
leer (< legere) “to read”, and in the 3rd person singular of the present indicative: cree “(s)he
believes” (< credit), and lee “(s)he reads” (< legit) (oddly, simplification did take place in
another common verb: OSp. veer > MSp. ver). In addition, the preservation of an identical
vowel sequence is attested in certain modern Asturian dialects of Leonese: cruu (beside
cruo) (< cru↜渀屮du) “raw”, too (for todo) (< to↜渀屮tu) “all-masc.” (García Arias 2003:€24–25). The
period in which monophthongization of HR /ej/ came about in Proto-Spanish is not known.
The HR dipthong /aj/, however, is documented in 10th century Mozarabic forms such as
carrayra, carraira (< carra↜渀屮ria; Zamora Vicente 1979:€ 35; Lapesa 1981:€ 176), a historical
stage not too distant from the change /ee/ > /ej/.
.╅ There are only two known exceptions to such early monophthongization of /ej/ in
Spanish: peine “comb” (< pectine), and seis “six”(< sex), for which there is no straightforward
explanation. Interestingly, there is another class of MSp. words, three in total, that contain
the diphthong /ej/, namely: veinte “twenty” (< viginti), treinta “thirty” (< triginta), reina
“queen” (< regina). However, they are not exceptions to monophthongization; here, the
diphthong derives from the resolution of the OSp. hiatus /e.í/ by means of syllable merger
(ve.ín.te > véin.te, tre.ín.ta > tréin.ta, re.í.na > rei.na, etc.), a sound change that occurred in
the transition to Modern Spanish (c. the 14th or 15th centuries; cf. Menéndez Pidal 1980:€84).
Notice, incidentally, that given the prohibition against long vowels in HR, identical vowel
 Word-minimality and sound change in Hispano-Romance 

aversion for falling diphthongs in Spanish throughout its history. In sum, a bimoraic
minimality condition on lexical words provides a straightforward explanation for
the failure of identical vowel hiatus in (7) to be resolved historically by the otherwise
regular process of simplification. Instead, the identical vowel hiatus was changed into
the unusual falling diphthong /ej/, thus effectively avoiding the impending creation of
monomoraic words, had hiatus avoidance been carried out by identical vowel simplifi-
cation, while still satisfying the bimoraic word minimality requirement.
Of course, one could easily imagine a number of different alternative ways to avert
a imminent violation of word minimality in such disyllabic words. A logically pos-
sible strategy, for example, would be to do nothing, namely, to maintain the identical
vowel sequence and endure the potentially undesirable hiatus configuration. In fact,
this solution was adopted in two Spanish verb forms: cree “(s)he believes” and lee
“(s)he reads”, but for reasons which are not clear, it somehow fell through in OSp.
vee “(s)he sees” > MSp. ve (see n. 5; cf. also early OSp. fee > MSp. fe). A much more
substantial instance of this seemingly procrastinating strategy is abundantly attested
in Old Galician-Portuguese disyllabic words containing a sequence of identical vowels
created by the deletion of an intervening /n/ or /l/, a sample of which is given in (8)
(García de Diego 1984 (1909): 30, 41, 43, Williams 1938:€94–95, 125–127, 170–171,
Ferreiro 1999:€44–45, 62–63, 66–67, 84–90).
(8) Latin OGal.-Port. Latin OGal.-Port.
so↜渀屮lu “alone-masc.” soo mulu “donkey” muu
bo↜渀屮nu “good-masc.” boo pede “foot” pee
no↜渀屮du “knot” noo lana “wool” laa
nu↜渀屮du “naked-masc.” nuu sana “healthy-fem.” saa
so↜渀屮nu “noise, sound” soo ala “wing” aa
do↜渀屮nu “gift” doo rana “frog” raa

It was only in the transition to modern Galician and Portuguese that the identical
vowel sequences in the disyllabic words in (8) underwent simplification, resulting in
vowel-ending monosyllables, apparently in contravention of word minimality. This
issue is briefly addressed in Section€ 5, where we suggest that such words undergo
moraic lengthening, and or laxing, and suggest that they may not be legitimate counter-
examples to word minimality at all.7

simplification can be formally interpreted within a general process of hiatus resolution by

syllable merger.
.â•… With regards to items such as those in (8), García de Diego notes that in Modern Galician
the vowel is realized as long if speakers analyze it as a combination of two identical vowels:
“la lengua actual las hace todas monosílabas; con vocal larga si dura la conciencia de la com-
posicion…” (1984 (1909): 31).
 Fernando Martínez-Gil

To summarize so far, the evidence provided by the failure of both -m deletion

and identical vowel reduction to materialize in HR also lends strong support to the
hypothesis that the prosodic unit that satisfied word-minimal requirements in HR is a
bimoraic foot, not a disyllabic one.
There is one final historical puzzle that can be given a straightforward account
under a bimoraic minimal-word analysis, but does not fit under a disyllabic minimal-
word counterpart. Early HR exhibited a handful of monosyllables for the 1st person
singular of the present indicative in some very common verbs, shown in (9), all of
which end in a high offglide in the modern HR languages: a front glide /-j/ in Spanish,
and a round one /-w/ in Galician and Portuguese.8

(9) Latin OGal-Port. OSp. Msp. Gal./Port.

do do (> dou) do doy dou “I give”
sto *estó (> estóu) estó estoy estou “I am”
suâ•›m son/som so soy son (sou)/sou “I am”
va↜渀屮do (> vau >) vou vo voy vou “I go”

Many attempts have been made in the past to explain the historical emergence of the
/-j/ marker in Spanish (see Wanner 2006 for a comprehensive review). While these
accounts may differ significantly in some relevant aspects, they all share one overarching
aim, namely, that of establishing the origin of the accretive yod, which according to
Wanner (2006) initially appeared in soy around the 13th century, and subsequently
extended analogically to the other monosyllabic forms. As for Old Galician-Portuguese,
the traditional view is that the /-w/ in vou, created by regular sound change from Latin
va↜渀屮do↜渀屮, was adopted by the other forms also through analogical extension (Williams
1938:€219–223, 233).
All former accounts suffer from a significant shortcoming: they fail to estab-
lish any formal connection between the final yod and the prosodic structure of the
verb form to which it was agglutinated, and therefore they are unable to explain why
the yod was not also added (analogically or otherwise) to polysyllabic verb forms
morphologically and semantically related to those in (8); compare, for example, doy
“I€give” with entrego “I hand in” and otorgo “I grant” (not *entregoy, *otorgoy), or voy

.â•… The Spanish existential present indicative form hay “there is, there are” (OSp. ha) should
probably be added to the set in (8). On the other hand, modern theoretical approaches to
Spanish phonology analyze both OSp. estó and MSp. estoy as monosyllabic at the underlying
level, /sto/ and /stoj/, respectively (cf. Harris 1983, 1987). In addition, because it ends in a
stressed vowel, under the analysis that Old Spanish metrical structure is based on a moraic
trochee, as proposed in this paper, the stressed final syllable in estó is monomoraic just as the
monosyllabic forms in (9).
 Word-minimality and sound change in Hispano-Romance 

“I go” with salgo “I leave”, parto “I depart” (not *salgoy, *partoy). Thus, while the search
for the elusive source of the accretive glide is in itself an interesting topic of research,
one can argue that the nature of glide augmentation in (9) cannot be fully understood,
nor adequately explained, without taking into account minimal-word constraints on
prosodic structure. Yet, it is clear that the early Old Spanish forms so, do, vo, estó and
ha were prosodically marked, since they failed to reach the bimoraic minimal size. As
shown in (10), such requirement is satisfied by a monosyllabic word such as soy, but
not by one like so, which consists of only one mora. If we compare the prosodic struc-
ture of OS. so in (10a) with that of MSp. soy in (10b), the former evidently violates the
bimoraic minimal word requirement, while the latter satisfies it.

(10) a. PrWd b. PrWd

F F

σ σ

μ μ μ

s ó s ó j

In short, glide augmentation in (9) is fully consistent with an analysis of the minimal
prosodic word in HR as a bimoraic foot, but not as a disyllabic one. Whatever its origin
and the analogical mechanism by which it reached the other monosyllabic forms, the
primary motivation for yod augmentation in our analysis lies in the drive to satisfy the
prosodic requirement that lexical words consist minimally of two moras; the accretive
yod is the mechanism that restored prosodic well-formedness to such HR subminimal
verb forms. In this view, then, the crucial event is the insertion of a glide increment;
the disputed origin of the yod, while an interesting topic in its own right, becomes
epiphenomenal. Space constraints prevent us from exploring here the grounds for the
failure of yod augmentation to apply to a handful of monomoraic lexical items in Mod-
ern Spanish, briefly discussed in Section€ 5 below. They include a few defective verb
forms, which either remained monomoraic throughout the Old Spanish period and up
to contemporary Spanish, such as da (< dat) “(s)he gives”, va “(s)he goes”, etc., or were
bimoraic in Old Spanish but uncharacteristically underwent identical vowel simplifica-
tion in the transition to the modern language, such as OSp. vee > MSp. ve “(s)he sees”,
as well as one vernacular noun, early OSp. fee > MSp. fe “faith”. The existence of such
exceptions presents a problem for any previous analyses, as much as it does for the
present one. Conceivably they are the result of a variety of factors, both linguistic (e.g.
paradigmatic conditions in verb forms) and sociolinguistic (perhaps involving dialect
contact), which are not well understood at present. Although we cannot offer an illumi-
nating solution here to the problem posed by such handful of exceptions, it is still the
 Fernando Martínez-Gil

case that, unlike previous analyses, the prosodic account of yod augmentation put forth
in this paper provides a principled explanation for the incremental glide.

4.â•… On the metrical system of Latin and its historical evolution into HR

In order to demonstrate our claim that the three sound changes discussed in the
previous section came about in order to avoid a potential infringement of the PMC
(cf. (5) above), namely, one in which the minimal word equals a bimoraic foot, one
needs to show that the phonological system of Proto-HR was indeed sensitive to
syllable weight, and that the canonical metrical foot in the proto-language was the
moraic trochee. Such a demonstration is crucial because it has been explicitly argued in
the literature that: (a) sensitivity to syllable weight disappeared in spoken Latin; (b)€dur-
ing its formative period, and up to the apocope stage, the canonical metrical unit in
Spanish was the syllabic trochee (Lipski 1997; Holt 1997; see also Roca 1990).
There are compelling reasons to suspect that syllable weight (i.e. the light vs. heavy
syllable distinction) was indeed an active property in the phonology of Proto-HR. The
phonological conditioning of two historical changes in Spanish, pre-tonic vowel syn-
cope and obstruent voicing, provide solid support for such a claim. First, pre-tonic
vowel syncope in Spanish took place when the vowel in question was located in an
open (i.e. light) syllable (ve.r(i).ta .te > ver.dad “truth”, de.l(i).ca .tu > del.ga.do “thin”;
ci.v(i).ta .te >cib.dad > ciu.dad “city”; for clarity, syllable boundaries are indicated by
periods), but failed to occur when the preceding syllable was closed by a consonant,
a fact pointed out by Pensado Ruíz (1984:€311) (tem.pes.ta .te > tem.pes.tad “storm”;
*im.pig.na .re > em.pe.ñar “to pawn”; mi.nis.te.ri.u > me.nes.ter “need, want”; li.ber.
ta .te > li.ber.tad “freedom”). And second, voicing affected obstruents whey they were
immediately preceded by a light syllable (i.e. one ending in a nuclear vowel: lu.pu >
lo.bo “wolf ”; vi.ta > vi.da “life”; fo .cu > fue.go “fire”), but not when it was preceded by
a heavy syllable, i.e. a syllable ending in a consonant (co r.pu > cuerpo; po n.te > puente,
al.tu > alto; cir.ca > cerca), or a glide (au.tum.nu >otoño “autumn”; pau.co > poco
“little”; sal.tu > *sa[w].to > soto “grove”; al.t(e).ru > *a[w].tro > otro “(an)other”;
sa.pi.at > *sa[j].pa > sepa “I know-subj.”; ca.pu.it > *ca[w].pe > cope > cupe “I fitted”).
As is well known, the Latin metrical system included the following properties:
(a)€ primary stress is based on a left-headed bimoraic foot (= a moraic trochee)
placed on the word’s right margin, coupled with extrametricality of the final sylla-
ble; (b)€Weight-to-Stress: stress falls on the penult if heavy (i.e. bimoraic), and on the
antepenult if the penult is light; and (c) word-minimality is an absolute restriction:
a prosodic word consists minimally of a bimoraic foot (cf., Allen 1973; Mester 1994;
Hung 1995; Apoussidou & Boersma 2003; Jacobs 2000, 2003; Roca 2006). In addition,
monosyllabic (though bimoraic) lexical words lexical words were common in Latin in
 Word-minimality and sound change in Hispano-Romance 

the nominative case of 3rd declension imparisyllabic nouns (e.g. mons “mountain”, rex
“king”, l ac “milk”, mel “honey” etc.), but they invariably added a syllable in the other
cases, including the accusative (cf. montem, regem, l actem, mellem, respectively),
the source of the overwhelming majority of HR nouns and adjectives. Thus, when
the word final accusative marker -m was dropped in Vulgar Latin, these words still
remained disyllabic.9
A series of historical events have undoubtedly led scholars to suspect that in
the transition to HR, the quantity-sensitive system of Latin must have turned into a
quantity-insensitive one, at least up to the emergence of medieval apocope. The fol-
lowing four are among the most consequential: (1) the loss of phonemic quantitative
distinctions among vowels, which obliterated (but by no means entirely; see below)
the systematic interdependence between stress locus and a heavy/light penult in a vast
number of trisyllabic or larger lexical items; (2) Vulgar Latin desyllabification of a short
front vowel in the penultimate syllable, which turned a large number of proparoxytones
into paroxytones (vi.ne.am > VLat. ví.nia > Sp. ví.ña “vineyard”; plu.vi.am > VLat.
plú.via > Sp. llú.via “rain”; stress marks supplied for clarity); (3) the analogical right-
ward shift of the stress from the antepenult to the penult in a considerable number of
verb forms from the Latin 3rd conjugation in -ere: tra here > *traére > MSp. traér
“to bring”; bibere > *bebére > MSp. bebér “to drink”; fugere > *fugére > MSp. huír
“to flee”; renego > *renégo > reniégo “I deny” > MSp. reniégo “I renounce”; recipit >
*recípit > Sp. recíbe “(s)he receives”); and (4) syncope of the post-tonic vowel, which
effectively shifted the stress pattern of a large number of lexical items from a propar-
oxytonic to a paroxytonic one: do m(i)nu “lord” > duéño “owner”; fra x(i)nu > frésno
“ash (tree)”; o c(u)lu > ójo “eye”, etc. Because of their contribution to the regulariza-
tion of penultimate stress, these events lend strong support to a scenario in which
during a relatively long period in its historical transition from Hispanic Latin, an over-
whelming majority of items in the HR lexical vocabulary were minimally disyllabic,
and stress almost invariably fell on the penult. The mayority of exceptions comprise a
handful of words with proparoxytonic stress, in which the vowel /a/ failed to undergo

.╅ A vast majority of monosyllabic functional words in Latin were also bimoraic: demon-
strative pronouns, such as hic, hoc, is, id, adverbs, such as iam, sīc, hic, nunc, prepositions,
such as ad, ex, ob, cu m, comparative particles, such as tam, plus, interrogative particles, such
as num, negative particles ne, non, etc. Monomoraic words were not lexical, and they had to
be enclitic to a lexical word (i.e. they were prosodically dependent on lexical words), as is the
case, for example, with the coordinative conjunction -que (see Mester 1994). On the other
hand, the vast majority of Latin verb forms were also minimally disyllabic, with very few ex-
ceptions, including some present and preterit indicative singular forms of īre “to go”: īs “you
go”, it “(s)he goes”, etc. (which were quickly replaced in Hispanic Latin by those of vado “to
walk”), and esse “to be”: sum “I am”, es “you are”, est “(s)he is”.
 Fernando Martínez-Gil

post-tonic vowel syncope, and thus preserved their original antepenultimate stress:
o rpha nu > Sp. huérfano, OGal. órfao, órfoo > MGal. orfo, Port. órfão; ra pha nu > Sp.,
Gal. rábano, Port. rábano, rabão, etc. Thus, the combination of a lexical vocabulary
containing almost exclusively disyllabic or larger words in combination with nearly
systematic penultimate stress appears to have led to the inevitable conclusion that
during its formative period HR must have had exhibited a quantity-insensitive sys-
tem, one in which the syllabic trochee was the canonical metrical foot, as explicitly
argued in Holt (1997), and Lipski (1997). Presumably keeping with this line of reason-
ing, Lipski (1997) surmises that “[q]uantity sensitivity disappeared in late Latin, and
played no role in Old Spanish stress assignment (p. 573),” and he concludes that in the
pre-apocope period: “only bisyllabic feet were acceptable; and therefore the quantity of
individual syllables was irrelevant” (1997:€579).
It can be shown, however, that the Spanish stress system prior to apocope is fully
amenable to an analysis based on the moraic trochee, by simply assuming weight by
position, which we motivated on independent grounds earlier with the pre-tonic vowel
syncope and obstruent voicing facts. Consider the prosodification of three represen-
tative Proto HR items: *pane “bread” (< pane), with a light penult vs. campo “field”
(<€campu) and *laite “milk” (< lacte), with a heavy penult. As shown in (11), all other
things being equal (i.e. in the absence of crucial evidence bearing on one or the other
option) it is difficult to pin down the particular prosodic unit that underlies the metrical
structure of such words (a fact already pointed out by Mester 1994); namely, they can be
parsed as either syllabic trochees (11a), or as moraic trochees (11b):

(11) a. Parsed as moraic trachees:

F F F

μ μ μ μ μ μ μ μ

pá ne ká m p o lá j t e

b. Parsed as syllabic trochees:

F F F

σ σ σ σ σ σ

p á n e k á m p o l á j t e

The equivalence of both analyses for the item pane is obvious and deserves no further
comment. As for campu, and laite, a moraic analysis would be straightforward by
simply assuming two properties already active in the Latin stress system: (a) Weight
by Position; and (b) Weight-to-Stress (“a heavy syllable bears stress”, thus inducing
extraprosodicity of the final mora). If the analyses in (11) are equivalent in that both
are capable of accounting for the proto HR stress system, then the choice between the
 Word-minimality and sound change in Hispano-Romance 

two obviously becomes an empirical question. For the analyst the temptation to opt for
the disyllabic foot is almost inevitable in view of Trubetzkoy’s renowned implicational
universal that a language’s stress system cannot be quantity-sensitive if it does not have
also have a phonemic contrast between short and long vowels (see Hyman 1985).
It is reasonably clear by now, however, that there are languages which lack vowel
length contrasts, and yet their stress system is, at least partially, sensitive to syllable
weight. Modern Spanish is such a case, because of two basic properties: (a) unmarked
stress falls in the last syllable of a word if heavy (e.g. papél “paper”, canción “song”; the
inflexional consonants /-s/, /-n/ are extraprosodic), and on the penult if the final is light
(e.g. humílde “humble”, etápa “stage”, mordísco “bite”); and (b) antepenultimate stress
is ruled out if the penult is heavy (this is what Roca 1990, 2006 calls the Salamanca
constraint: cf. pálido “pale” vs. hypothetical *pálindo, *pálaido, *páliado, etc., which are
systematically prohibited; see Harris 1983:€ 10–11, cap. 4, 1995; Dunlap 1991; Lipski
1997; Roca 1999, 2006; but cf. Roca 1990, who argues that sensitivity to syllable weight
in the Spanish synchronic grammar is limited to a lexicalized residue of the general
role this property played in Latin). Moreover, sensitivity to syllable weight appears to
be a gradual property, not a categorical one, as suggested by recent studies on prosodic
structure (Kager 1992, 1999:€145–147, De Lacy 1997; Gordon 1999, 2004).
Quantity-sensitivity in Latin essentially meant that penultimate stress was depen-
dent on a heavy penult, and a heavy penult was either a long vowel or a closed syllable.
When vowel length distinctions were lost, sensitivity to a heavy penult in HR became
limited to a syllable closed by a consonant, as in campo, or a glide, as in Gal.-Port. leite.
Moreover, it is an indisputable fact that, although to a lesser extent than Latin,
the pre-apocope HR metrical system, just as Modern Spanish, still exhibited some
degree of quantity-sensitivity, as witnessed by the impossibility of antepenultimate
stress in the presence of a heavy penult: sobérbia “haughtiness” (< superbia), maéstro
“teacher” (< magistru), hiniésta “broom plant” (< genesta), donáire “wit” (< *donáiro
< do na ri u) (cf. *sóberbia, *máestro, *híniesta, *dónaire, clearly ruled out). In other words,
there is no reason to suspect that the Salamanca constraint has not been inherited by
the HR languages in a seemingly direct and unbroken line from the mother language. A
fact sometimes overlooked in historical linguistics is that language history sometimes
involves linguistic continuity to a larger degree than linguistic change. An analysis in
terms of the bimoraic foot (11a) is appealing, in the sense that, the loss of vowel quantity
aside, it postulates an essential continuity of the Latin prosodic system in its transition
to Romance.
Lets us recapitulate at this point two fundamental and unquestionable historical
facts: (a) the canonical metrical foot in Latin was the moraic trochee; (b) in the time
span that stretches from Hispanic Latin to the medieval apocope stage, the vast majority
of lexical items were disyllabic or larger, a clear indication that during this period the
proto-language was subject to some sort of word-minimality. The critical question that
arises, then, is whether in the long period before the emergence of apocope the minimal
 Fernando Martínez-Gil

prosodic word in HR was the syllabic trochee or its moraic counterpart. Interestingly,
apocope (deletion of a word-final /-e/), a sound believed to have taken place towards the
end of the preliterary period (around the 10th or 11th centuries), constitutes a critical
piece of empirical evidence bearing on this issue, because it cast serious doubts on the
viability of the syllabic trochee as the minimal prosodic word in the proto-language.
As illustrated in (12) with data from Spanish and Galician-Portuguese, apocope
took place after a single anterior coronal consonant, whether a fricative or an affri-
cate obstruent (12a), or a sonorant (12b). Unlike Spanish, Galician-Portuguese did not
undergo apocope after /d/ (derived in most cases by voicing of intervocalic /t/), as seen
in (12c). In some instances, Spanish (but not Galician-Portuguese) underwent apo-
cope after the medieval dental africate /ts/, as shown in (12d), as well as in some 3rd
person singular present indicative verb forms (12e) (the absence of apocope in Spanish
verb forms is due to paradigmatic factors). Finally, in the infinitive of some common
verbs, apocope yielded monosyllabic forms in Galician and Portuguese, and disyllabic
in Spanish (12f).

(12) Apocope in Hispano-Romance (P-HR = Proto-HR):

Lat. P-HR Sp. Gal. /Port.
a. tu↜渀屮sse *tose tos tos “cough”
cru ↜渀屮ce *cruce cruz cruz “cross”
lu
↜渀屮ce *luce luz luz “light”
b. pa↜渀屮ne *pane pan pan/pão “bread”
so↜渀屮le *sole sol sol “sun”
ma↜渀屮re *mare mar mar “sea”
c. vite *vide vid vide “grapevine”
site *sede sed sede “thirst”
rete *rete red rede “net”
d. calce *cauce coz couce “kick”
fasce *fasce haz feixe “bundle”
pisce *pesce pez peixe “fish”
e. tenit *tene tiene ten/tem “(s)he has”
ponit *pone pone pón “(s)he puts”
dicit *dice dice di “(s)he says”
f. po↜渀屮ne↜re *ponere poner pór “to put”
venire *venire venir vir “to come”
cre↜de↜re *creere creer crer “to believe”

Apocope constitutes an event of the utmost significance for the prosodic evolution of
HR, as it has been singled out as the single factor responsible for reintroducing quantity-
sensitivity into the metrical system of Spanish (Lipski 1997; similar considerations
would apply to Galician-Portuguese). Lipski (1997:€580) summarizes the consequence
 Word-minimality and sound change in Hispano-Romance 

of apocope for the prosodic system of post-apocope Spanish in this manner: “The
(re-)introduction of quantity sensitivity into Spanish meant that the foot template
shifted from the bisyllabic…back to the moraic template…which had existed in Latin….
The only readjustment occasioned by the creation of new stress configurations was
the substitution of a moraic foot template for a syllabic template.” In reality, however,
although apocope may have (re)introduced sensitivity to syllable weight within the
final syllable, the heavy vs. light distinction in the penult had been active in the lan-
guage since Latin, as we have already seen.
More significantly, apocope is incompatible with an analysis in terms of a disyllabic
foot on two counts. First, the process reduced the size of a large number of disyllabic
words to consonant-ending monosyllables; if the minimal size of the prosodic word
coincides with the metrical foot, if follows that in the stage prior to apocope the mini-
mal prosodic word in the language could not have been a disyllabic foot, and there-
fore the syllabic trochee could not have been the canonical metrical unit either, since
otherwise the output of apocope in disyllabic words, such as those in (12), would have
resulted in what amounts to a massive violation of prosodic word minimality. And
second, apocope in trisyllabic or larger words with paroxytonic stress would have ren-
dered them oxytonic (cf. Proto-HR *ponére > Sp. ponér in (12f)); final stress on a closed
(i.e. heavy) syllable correlates with a moraic trochee, not with its syllabic counterpart.10
In sum, the evidence from medieval apocope clearly indicates that this sound change is
entirely compatible with an analysis of the minimal prosodic word based on a bimoraic
trochee, and thus with a quantity-sensitive analysis in which the bimoraic foot is the
minimal prosodic word, but it cannot be reconciled with a quantity-insensitive sys-
tem based on the syllabic trochee. In fact, one can argue that the description of the
Spanish stress system at any stage of its historical evolution, before or after medieval
apocope, is fully consistent with moraic trochee as the unit identifiable with the minimal
prosodic word.

5.â•… The status of prosodic word minimality in modern HR

We have seen that prosodic minimality, correlated with the canonical bimoraic foot,
was an absolute restriction on lexical words in Latin, and overwhelmingly strong
tendency in the HR languages up to and throughout the medieval period, though
not strictly an absolute restriction. An inspection of the lexical vocabulary of the
modern HR languages confirms the presence of a small number of vowel-ending
monosyllables; however, word minimality, identified with a bimoraic foot, continues

.╅ I thank an anonymous reviewer for bringing this point to my attention.

 Fernando Martínez-Gil

to be an extremely strong tendency. While the existence of minor exceptions is rela-

tively common across languages and should not invalidate an otherwise evident and
well-grounded linguistic generalization, there is some robust phonetic evidence in
support of the claim that the minimal prosodic word in the modern HR languages is
indeed the bimoraic foot, and that the few apparent violations of prosodic minimality
are not true counterexamples to this generalization. This conclusion is based on the
following facts.
In all three major modern HR languages, Spanish, Galician, and Portuguese there
is a relatively large number of monosyllabic words that either contain a nuclear (rising
or falling) diphthong or are closed by a consonant: Spanish pie “foot”, fui “I went”, hoy
“today”, ley “law”, mar “sea”, sal “salt”, or Galician/Portuguese dou “I give”, vai “(s)he
goes”, dor “pain”, mel “honey”, etc. (the monosyllabic noun pie “foot”, as well as verb
such as forms fui “I was/went”, fue “(s)he was/went”, dio “(s)he gave”, and vio “(s)he saw”,
would be bimoraic under the assumption that rising diphthongs make heavy syllables;
see n. 2). According to Navarro Tomás’ survey (1966:€ 54) on the relative frequency
of the Spanish vocabulary according to prosodic type (any combination of stress and
syllable count) monosyllabic words closed by a consonant are relatively common;
they make up around 13% of all the Spanish words that bear primary stress, although
still considerably less frequent than disyllabic items such as casa “house”, which com-
prise around a 30% of all lexical words. By contrast, vowel-ending monosyllables are
extremely rare. We have found 15 such words in Spanish: 2 nouns, of which, only one,
fe “faith”(< Lat. fide), is from the native vocabulary (the other, té “tea” is a Chinese
loanword), the personal pronouns yo “I” and tú “you”, and a total of 11 mostly defec-
tive verb forms (i.e. verbs that either lack a root vowel, such as dar “to give” and ver “to
see”, or are heavily suppletive, such as ir “to go”, and ser “to be”); they include 7 pres-
ent and preterite forms (1st or 3rd p. sg.): da “(s)he gives”; dé “I/(s)he give-subj.”; di
“I gave”; ve “(s)he sees”; vi “I saw”; va “(s)he goes”; sé “I know” (from saber “to know”),
and 4 imperatives: di “say!”; ve “go!”; da “give!”; sé “be!” (the monosyllabic forms of
haber “to have” have been disregarded here due to its auxiliary status).
As for modern Galician and Portuguese, their lexical stock also contains a hand-
ful of monomoraic nouns and adjectives. The following 17 tokens include all nonverb
forms I have been able to find, some of them are rarely used in colloquial language
(the phonetic form given for each item is common to both Galician and Portuguese
unless indicated otherwise): nú [nú] “naked” (< nu du), mú [mú] “donkey” (< mu lu);
crú [krú] “raw” (< cru du); cú [kú] “buttocks” (<cu lu); sé [sέ] “headquarters” (< se
de “seat”); fé [fέ] “faith” (< fide); té [tέ] “tea” (cf. te [te] = pronominal clitic); pé [pέ]
“foot” (< pede); Port. mó [mf́] “molar tooth” (Gal. moa) (< mo la “millstone”) (cf. Gal.
mo [mo] < me + o = pron. clitic + masc. sing. def. article); só [sf́ “alone” (< so lu); dó
[df́] “grief, sorrow” (< do lu); nó [nf́] “knot” (< no du); pó [pf́] “dust” (< *pu l(v)u for
pu lvis); pá [pá] Gal. “baker’s peel”, Port. “shovel” (< pa la); Gal. lá [lá˛] (more com-
monly, lan), Port. lã [l!˜´] “wool” (< l a na); Gal. rá [rˉá˛] (more commonly, ran), Port. [rˉ!˜´]
 Word-minimality and sound change in Hispano-Romance 

“frog” (< ra na); Gal. sá [sá˛] (more commonly, san), Port. sã [s!˜´] “healthy-fem.” (< sa na).
Monomoraic verb forms in Galician and Portuguese do not seem to differ significantly
in number from those listed above for Spanish; notice that for some of such Spanish
monomoraic verb forms, the corresponding forms in Galician and Portuguese are
bimoraic: e.g. cf. Sp. va, sé, dí, vi vs. Gal., Port. vai, sei, dei (dial. Gal. din), vin (Port.
vim), respectively; and vice versa: cf. Gal.-Port. monomoraic rí “(s)he laughs”; é [έ]
“(s)he is”; dí “(s)he says”; cre “(s)he believes”; le “(s)he reads” vs. Spanish bimoraic ríe,
es, dice, cree, lee, respectively.
Interestingly, there is synchronic evidence in the modern HR languages that the
vowel in such monosyllables is subject to certain phonetic modifications, such as
lengthening and changes in vowel quality, which can be readily interpreted as mani-
festations of a process of moraic lengthening, whose most significant outcome is the
satisfaction of word minimality in surface representations. We take this evidence to
provide phonological motivation for the prosodic representations in (13), in which an
item such as Sp. fe “faith” (13a) is a structurally a heavy syllable, just as a vowel-final
oxytone (13b), or a consonant-final monosyllable (13c):

(13) a. σ b. σ c. σ

μ μ μ μ μ μ

f é k a f é p á n

Two phonetic facts related to vowel duration and vowel quality lend compelling support
to the bimoraic representation in (13):

1. Vowel duration. As is well known, the HR languages exhibit a robust phonetic

connection between stress and duration: stressed vowels tend to be longer than
unstressed ones. Significantly, according to Navarro Tomás (1977:€201, vowel dura-
tion in Spanish stressed monosyllables, such as tú “you”, fe, “faith”, té “tea”, etc., is
around 200 milliseconds, about twice the duration of a stressed vowel closed by
a consonant in disyllabic words such as cerca, torta (80–95 mls.), and consider-
ably longer than stressed vowels in open penults, such as cebo, pasa, puro, mora
(120–150 mls.) (cf. also Monroy Casas 1980; Martínez Celdrán 1984:€244–245).
Although we have not been able to find similar studies on stressed-vowel duration
in Galician and Portuguese, a priori there is no reason to suspect they would differ
significantly from Spanish. In fact, the spectrographic evidence presented in Porto
Dapena’s (1977) detailed study of a northwestern Galician dialect indicates that
the low vowel /a/ in the items pá [pá˛] “baker’s peel” and rá [rˉá˛] “frog”, not only
exhibits an extremely open quality that clearly distinguishes it from stressed /a/
in polysyllabic words, but its duration is approximately equal to the (bimoraic)
falling diphthong in pau [páw] “stick”, and considerably longer than the stressed
 Fernando Martínez-Gil

vowel in casa [kása] “house.” Freixeiro Mato (1998:€73) explicitly alludes to the
lengthening of /a/ in Galician rá “frog” and lá “wool”, and transcribes it as long
in these items: [Ära˜], [Äla˜]. Finally, Sá Nogueira (1958:€37) observes that nasalized
vowels in Portuguese are perceived as longer than oral vowels.
2. Vowel quality. A striking feature that can be observed in the 17 vowel-ending
monosyllabic nouns and adjectives in modern Galician and Portuguese mentioned
earlier is that when the vowel is mid, its quality is predictably lax (for laxing in
Galician, see Carballo Calero 1979:€28–29, Porto Dapena 1977:€14, 18, García de
Diego 1984 (1909): 8–9, Álvarez, Regueira, and Monteagudo 1986:€19–20, Freixeiro
Mato 1998:€73; for Portuguese, see Williams 1938:€33; Sá Nogueira 1938:€59–60,
1958:€37). While in some of such items the lax mid vowels can be traced directly to
a Latin short mid vowels, in the remaining instances the Latin ancestor contains a
long mid vowel, and thus an explanation for the lax quality is not readily available
in terms of regular sound change. Not surprisingly, Williams (1938:€33) observes
with regards to fé “faith” in Portuguese that “[t]here is no apparent reason for the
opening of ẹ in fé (from fĭdem).” In Galician laxing also affects the low vowel, whose
extremely open articulation is explicitly alluded to in several studies of the language
(see, e.g. Porto Dapena 1977:€14, 18). Even in Spanish, which lacks a phonemic
tense-lax contrast, at least for some speakers the vowel in té, fe, is noticeably more
open than that of /e/ in polysyllabic words. To my knowledge, the only exception
to this systematic correlation is Gal. bó [bó], Port. bom [bõ] “good-masc.”; here,
tensing of the mid lax vowel when followed by a nasal is a rather regular develop�
ment in the history of Galician-Portuguese. Laxing, on the other hand, does
not occur in monosyllabic verb forms; instead, we find both tense and lax vowels
(cf. é [έ] “(s)he is” vs. le [lé] “(s)he reads”), presumably due to requirements of
paradigm uniformity. The phonetic correlation between vowel quality and dura-
tion is well established, although the particular vowel quality that combines with
longer duration appears to be subject to interlinguistic variation. In English, for
example, tense vowels are significantly longer than their lax counterparts. On these
grounds, Demuth et al. (2006:€140) analyze the tense vowels of English as bimoraic,
and the lax ones as monomoraic. In contrast, Martínez Celdrán (1984:€245) sug-
gests that the opposite correlation is more common: lax vowels tend to be longer
than their tense counterparts. Indeed, one is led to conclude that in Galician and
Portuguese, in addition to increased duration, laxness is a property associated
with the prosodification of monosyllabic vowel-ending words as bimoraic feet.

6.â•… Conclusion

This paper has argued that a minimal prosodic word requirement provides a simple
unified account of three seemingly exceptional developments in the history of HR
 Word-minimality and sound change in Hispano-Romance 

which traditionally had not been considered related sound changes. We have identi-
fied the relevant metrical unit correlated with the minimal word as a bimoraic foot,
counter to recent proposals that postulate a disyllabic one. It has been shown that two
otherwise regular sound changes, -m deletion and identical vowel simplification, failed
to occur when the prosodic word would have been reduced to a form smaller than
a bimoraic foot; instead of the expected changes, alternative historical developments
took place that can be analyzed as strategies to avoid the reduction of the target words
to a subminimal prosodic size. This account has obvious implications for the historical
analysis of the HR metrical system, especially because it runs counter to the standard
view, which holds that since the Hispanic Vulgar Latin stage, and at least up to pre-
apocope Old Spanish, stress was quantity-insensitive, and based on a syllabic trochee.
The analysis presented here casts serious doubts on the validity of such a claim. Further
evidence countenancing the moraic trochee as the word-minimal prosodic constitu-
ent in HR is provided by the early medieval process of apocope, whereby word-final -e
underwent deletion when preceded by a single coronal consonant, thus turning many
disyllabic words into monosyllabic consonant-ending ones. If the minimal prosodic
word indeed coincides with the metrical foot, as alleged by Prosodic Morphology, a
proposal that identifies such minimal prosodic unit as a disyllabic foot in HR makes the
counterfactual prediction that apocope should not have occurred; on the other hand
the facts on apocope are entirely compatible if the bimoraic foot is identified with the
minimal word in HR. Similar considerations have been advanced to explain glide aug-
mentation as a word minimality effect. In modern HR, on the other hand, the rare cases
of monosyllabic vowel-ending words avoid prosodic word subminimal size by either
moraic lengthening (Spanish), and both lengthening and/or change in vowel quality
(Galician and Portuguese). The facts considered in this paper naturally raise the ques-
tion: given a variety of strategies, all of which fulfill the goal of avoiding the reduction of
a lexical word below the required minimum, what determines the choice of a particular
mechanism? For example, why did 1st person monosyllabic verb forms resort to yod
augmentation in order to satisfy minimality (as in soy or doy) in the transition to modern
Spanish, but other monosyllabic verb forms, such as va or ve, etc., stayed monosyllabic
and undergo moraic lengthening instead? These are intriguing issues, but space limita-
tions prevent us from addressing them here in any meaningful way, and thus they must
await future research.

References

Allen, W. Sidney. 1973. Vox Latina: the Pronunciation of Classical Latin. New York: Cambridge
University Press.
Álvarez, Rosario, X.L. Regueira & H. Monteagudo. 1986. Gramática galega. Vigo: Galaxia.
Apoussidou, Diana & Paul Boersma. 2003. “The Learnability of Latin Stress.” Proceedings of the
Institute of Phonetic Sciences 25.101–148. Amsterdam: University of Amsterdam.
 Fernando Martínez-Gil

Broselow, Ellen. 1995. “Skeletal Positions and Moras.” The Handbook of Phonological Theory, ed.
by John Goldsmith, 175–205. Cambridge, Massachusetts: Blackwell.
Carballo Calero, Ricardo. 1979. Gramática elemental del gallego común. Vigo: Galaxia.
De Lacy, Paul. 1997. Prosodic Categorization. M.A. thesis, University of Auckland.
Demuth, Katherine, Jennifer Culbertson & Jennifer Alter. 2006. “Word-Minimality, Epen-
thesis, and Coda Licensing in the Early Acquisition of English.” Language and Speech
49.137–174.
Downing, Laura J. 2005. ìMorphological Complexity and Prosodic Minimality.” Catalan Journal
of Linguistics 4.83–106.
Downing, Laura J. 2006. Canonical Forms in Prosodic Morphology. New York: Oxford University
Press.
Dunlap, Elaine. 1991. Issues in the Moraic Structure of Spanish. Doctoral dissertation, University
of Massachusetts, Amherst (distributed by University of Massachusetts GLSA).
Ferreiro, M. 1999. Gramática histórica galega. Santiago de Compostela: Edicións Laiovento.
Freixeiro Mato, Xosé Ramón. R. 1998. Gramática da lingua galega, Vol. I: Fonética e Fonoloxía.
Vigo: Edicións A Nosa Terra.
García Arias, Xosé Luis. 2003. Gramática histórica de la lengua asturiana. Uviéu (Oviedo):
Academia de la Llingua Asturiana.
García de Diego, Vicente. 1984 (1909). Elementos de gramática histórica gallega. (Anexo 23 de
Verba.) Santiago de Compostela: Universidade de Santiago de Compostela.
Gordon, Mathew. 1999. Syllable Weight: Phonetics, phonology, and typology. Doctoral dissertation,
UCLA.
Gordon, Mathew. 2004. “Syllable Weight.” Phonetically-Based Phonology, ed. by Bruce Hayes,
Robert Kirschner & Donca Steriade, 277–312. New York: Cambridge University Press.
Grandgent, C.H. 1907. An Introduction to Vulgar Latin. Boston: D. C. Heat & Co.
Harris, James W. 1983. Syllable structure and Stress in Spanish: A nonlinear analysis. Cambridge,
Mass.: MIT Press.
Harris, James W. 1995. “Projection and Edge Marking in the Computation of Stress in Spanish.”
The Handbook of Phonological Theory, ed. by John Goldsmith, 867–887. Cambridge, Mass.:
Blackwell.
Hayes, Bruce. 1989a. “Compensatory Lengthening in Moraic Phonology.” Linguistic Inquiry
20.253–306.
Hayes, Bruce. 1989b. “The Prosodic Hierarchy in Meter.” Rhythm and Meter, ed. by Paul Kiparsky€&
Gilbert Youmans, 201–260. Orlando: Academic Press.
Hayes, Bruce. 1995. Metrical Stress Theory. Chicago: University of Chicago Press.
Holt, Eric. 1997. The Role of the Listener in the Historical Phonology of Spanish: An optimality-
theoretic account. Doctoral Dissertation, Georgetown University, Washington D.C.
Hung, Henrietta H. 1995. The Rhythmic and Prosodic Organization of Edge Constituents: An opti-
mality theoretic account. Bloomington, Indiana: I.U.L.C.
Hyman, Larry. 1985. A Theory of Phonological Weight. Dordrecht: Foris.
Inkelas, Sharon. 1990. Prosodic Constituency in the Lexicon. New York: Garland Press (Stanford
University Doctoral Dissertation, 1989).
Itô, Junko. 1990. “Prosodic Minimality in Japanese.” Chicago Linguistic Society 26: Parasession
on the Syllable in Phonetics and Phonology, ed. by K. Deaton, M. Noske & M. Ziolkowski,
213–239. Chicago: Chicago Linguistic Society.
Jacobs, Haike. 2000. “The Revenge of the Uneven Trochee: Latin main stress, metrical constitu-
ency, stress-related phenomena and OT.” Analogy, Levelling, Markedness, ed. by Aditi Lahiri,
333–352. Berlin: Mouton de Gruyter.
 Word-minimality and sound change in Hispano-Romance 

Jacobs, Haike. 2003. “The Emergence of Quantity-Sensitivity in Latin.” Optimality Theory and
Language Change, ed. by D. Eric Holt, 229–247. Dordrecht: Kluwer.
Kager, René. 1992. “Are There Any Truly Quantity-Insensitive Systems?” Berkeley Linguistics
Society 18.123–132.
Kager, René. 1999. Optimality Theory. New York: Cambridge University Press.
Lapesa, Rafael. 1981. Historia de la lengua española (9a ed.). Madrid: Gredos.
Lipski, John M. 1997. “Spanish Word Stress: The interaction of moras and minimality.” Issues in
the Phonology and Morphology of the Major Iberian Languages, ed. by Fernando Martínez-Gil€&
Alfonso Morales Front, 559–593. Washington D.C.: Georgetown University Press.
Martínez Celdrán, Eugenio. 1984. Fonética (con especial referencia a la lengua castellana).
Barcelona: Editorial Teide.
McCarthy, John & Alan Prince. 1990. “Foot and Word in Prosodic Morphology: The Arabic
broken plural.” Natural Language and Linguistic Theory 8.209–283.
McCarthy, John & Alan Prince. 1993. Prosodic Morphology: Constraint interaction and satisfaction.
Ms., Technical Report, Rutgers University Center for Cognitive Science.
McCarthy, John & Alan Prince. 1995a. “Prosodic Morphology.” The Handbook of Phonological
Theory, ed. by John Goldsmith, 318–366. Cambridge, Massachusetts: Blackwell.
McCarthy, John & Alan Prince. 1995b. “Faithfulness and Reduplicative Identity.” University
of Massachusetts Occasional Papers in Linguistics 18: Papers in Optimality Theory,
ed. by Jill Beckman, Laura Walsh Dickey & Suzanne Urbanczyk, 249–384. Amherst,
Massachusetts: GLSA.
McCarthy, John & Alan Prince. 1998. “Prosodic Morphology.” The Handbook of Morphology,
ed. by Andrew Spencer & Arnold Zwicky, 283–305. Cambridge, Mass.: Blackwell.
McCarthy, John & Alan Prince 1999. “Faithfulness and Identity in Prosodic Morphology. The
Prosody-Morphology Interface, ed. by René Kager, Harry van der Hulst & Wim Zonneveld,
218–309. New York: Cambridge University Press.
Menéndez Pidal, Ramón. 1980. Manual de gramática histórica española (16th ed.). Madrid:
Espasa Calpe.
Mester, R. Armin. 1994. “The Quantitative Trochee in Latín.” Natural Language and Linguistic
Theory 12.1–61.
Monroy Casas, Rafael. 1980. Aspectos fonéticos de las vocales españolas. Madrid: SGEL.
Navarro Tomás, T. 1966. Estudios de fonología española. New York: Las Américas Publishing Co.
Navarro Tomás, Tomás. 1977. Manual de pronunciación española (19th ed…). Madrid: C.S.I.C.
Nespor, Marina. 1999. “Stress Domains.” Word Prosodic Systems in the Languages of Europe, ed. by
Harry van der Hulst, 117–159. New York: Mouton de Gruyter.
Nespor, Marina & Irene Vogel. 1986. Prosodic Phonology, Dordrecht: Foris.
Orie, Olanike Ola & Douglas Pulleyblank. 2002. “Yoruba Vowel Elision: Minimality effects.”
Natural Language and Linguistic Theory 20.101–156.
Pensado Ruíz, Carmen. 1984. Cronología relativa del castellano. Salamanca: Ediciones Universidad
de Salamanca.
Porto Dapena, J. Álvaro. 1977. El gallego hablado en la comarca ferrolana. Anexo 9 de Verba.
Santiago de Compostela: Universidade de Santiago de Compostela.
Roca, Iggy. 1999. “Stress in the Romance Languages.” Word Prosodic Systems in the Languages of
Europe, ed. by Harry van der Hulst, 658–811. New York: Mouton de Gruyter.
Roca, Iggy. 1990. “Diachrony and Synchrony in Spanish Word Stress.” Journal of Linguistics
26.133–164.
Roca, Iggy. 2006. “The Spanish Stress Window.” Optimality-Theoretic Studies in Spanish Phonology,
ed. by Fernando Martínez-Gil & Sonia Colina, 239–277. Philadelphia: John Benjamins.
 Fernando Martínez-Gil

Sá Nogueira, Rodrigo de. 1938. Tratado de fonética portuguesa. Lisbon: Imprensa Nacional.
Sá Nogueira, Rodrigo de. 1958. Tentativa de explicação dos fonómenos fonéticos em português.
Lisbon: Livraria Clásica Editora.
Selkirk, Elisabeth. 1995. “The Prosodic Structure of Function Words.” University of Massachusetts
Occasional Papers in Linguistics 18: Papers in Optimality Theory, ed. by Jill Beckman, Laura
Walsh Dickey & Suzanne Urbanczyk, 439–469. Amherst, Mass.: GLSA.
Sherer, Tim D. 1994. Prosodic Phonotactics. Doctoral dissertation, University of Massachusetts,
Amherst (Distributed by University of Massachusetts GLSA).
Väänänen, Veiko. 1968. Introducción al latín vulgar. Madrid: Gredos.
Wanner, Dieter. 2006. “An Analogical Solution for Spanish soy, doy, voy, and estoy.” Probus
18.267–308.
Williams, Edwin B. 1938. From Latin to Portuguese. Philadelphia: University of Pennsylvania Press.
Zamora Vicente, Alonso. 1979. Dialectología española (2a ed.). Madrid: Gredos.
 Multiple opacity in Eastern Regional French*

Jean-Pierre Montreuil
The University of Texas at Austin

Several processes collaborate in Eastern Regional French to create a case of

multiple overapplication which challenges current treatments of opacity. Their
interaction has not been discussed in the literature. All processes – lengthening,
devoicing and epenthesis – are clearly synchronic, productive, and apply beyond
the level of phonetic implementation. An OT-CC analysis is presented which
explains in detail how valid candidate chains are constituted, why the valid chains
not leading to the winner are eliminated, and how precedence constraints account
for opacity.

1.â•… Introduction

Recent developments in Optimality Theory (OT) have sought to reintroduce deriva-

tional evaluations into the model and it has been argued (McCarthy 2006, 2007, 2008;
Jesney 2009; Kimper 2008; Pruitt 2008) that versions of OT which perform local evalu-
ations rather than global ones make more accurate predictions regarding the range
of variation and the typology of phonological phenomena. Models like Harmonic
Serialism (HS), Lexical Phonology-OT (LP/OT) or Stratal OT fail to handle opacity
adequately, as McCarthy (2007) demonstrates. OT with Candidate Chains (OT-CC) is
a faithfulness-based version of derivational OT which, like HS, constrains the grammar
by imposing at each iteration a requirement of harmonic improvement, thus limiting
the candidate set to just a few competing forms. McCarthy (2008) discusses at great
length the extent to which it can provide a unified account of opacity (see also Lee
2007, but Riggs 2008).
In this context, this paper presents an OT-CC treatment of multiple opacity
and dialectal variation in the regional French spoken in Lorraine and Alsace, areas
which have long been in contact with German and whose substratal dialect is most
often Germanic. In Lorraine, vowel lengthening and consonant devoicing interact
opaquely and in Alsace, schwa epenthesis adds another layer of opacity in final CL

*I would like to thank the LSRL audience for their questions and several anonymous reviewers
for their comments and suggestions.
 Jean-Pierre Montreuil

clusters. These are synchronic processes which, by themselves, are all productive and
exceptionless; opacity arises only when they interact.1 The OT-CC account devel-
oped here shows how opacity can be accounted for and how variation can result from
�Precedence constraints.
In Sections€2 and 3, I present the basic markedness constraints that regulate moraic
association in Lorrain French and examine the interaction between devoicing and
lengthening. The geographic repartition of forms in atlases shows the opaque form as
being the most prevalent.
In Sections€4 and 5, I move to Alsatian French, where epenthesis comes into
play. I discuss the nature of the candidate chains, their LUM sequences (sequences of
unfaithful mappings) and the Precedence constraints required in OT-CC.

2.â•… Vocalic length and consonantal voice

In the regional French of Lorraine (Lanher et al. 1979–1989) and Alsace (Carton et al.
1983; Philipp 1965; Wolf 1983), as is common in many languages of the world, vowels
lengthen before voiced consonants. From the literature on the topic, it appears that
lengthening is more than just a matter of phonetic implementation. This contextual
lengthening is usually reflected in transcriptions of words such as [ba˜:] bague “ring”,
not *[ba:]. In parallel, there is clear evidence of lexical length.2 The data in (1) are
illustrative of Western Lorrain, an area which comprises just a small sliver of villages
in the maps of the Atlas Linguistique de la Lorraine Romane (ALLR), representing less
than 15% of Lorraine.
(1) [o˜Š] auge “trough” [ba˜v], [bc˜v] bave “drool”
[∫mi˜z], [∫mε̃˜z] chemise “shirt” [no˜Š], [nf↜˜â†œŠ] neige “snow”
[vc˜z], [va˜z] vase “mud” [a∫], [c∫] hache “axe”
[ti˜d], [tje˜d] tiède “tepid” [ft], [hft] hotte “sack”

Autosegmental theory has contributed important insights on the proper structural

representation of the relationship between length and voice and suggested that it best
be translated through the concepts of weight-bearing codas and of mora-sharing.
In Moraic Phonology (Hayes 1989), Weight-by-Position (WbyP) allows some coda

.╅ These facts constitute a case of multiple overapplication. All technical terms, e.g. opacity,
overapplication, transparent, counterbleeding are used in their traditional meaning. McCa-
rthy (2007, esp. Ch. 1) provides a useful overview of related terms and issues.
.╅ Lexically long vowels will thus be underlying bimoraic, but it does not follow that unmarked
vowels should be viewed as underlyingly moraic, since the grammar can provide their weight
by default. As a result, there can be dep-µ constraints, but no max-µ constraints (except in
possible cases of long vowel shortening, which do not concern us here).
 Multiple opacity in Eastern Regional French 

consonants to project a mora and mora-bearing is often sonority related. It is now well
substantiated in contemporary research, notably in Zec (2003), that the more sonorous
voiced consonants tend to bear weight while voiceless consonants do not. Defining
WbyP is the first step; the second step consists in spreading the mora associated to the
coda consonant onto the vowel, which results in vocalic length.
To benefit from the insights of moraic theory, constraints that refer to mora asso-
ciation will be preferred to constraints like *vC[vce] (= No short vowels before voiced
consonants), or ident-long, which merely restate distributional facts and lack explan-
atory power. WbyP is represented here by a constraint on the positional weight of d
(=€voiced obstruents). The grammar fragment related to weight states the following:

a. vowel weight (intrinsic markedness): Vµ 〉〉 depµ(v), so vowels bear weight (Vµ is

undominated and not included in Tableau 1.)
b. coda weight (positional markedness): dµ]σ 〉〉 depµ(c), so final voiced codas bear
weight.
c. lengthening: in this framework, lengthening results from mora-sharing. When a
mora is shared, a faithfulness violation occurs since a new association is created.
Following Shaw (2008), we call the violated constraint dep-link-µ; a lengthy
discussion regarding the status of dep-link-µ (and, more generally, the relation-
ship between moras and faithfulness) is included in Section€ 5. What triggers
mora-sharing is the dominance of µhead over dep-link-µ. This ensures, at the
expense of creating a new association, that moras anchor preferentially to vowels.
In Tableau€1, candidate c., phonetically interpreted as [ba˜:], is the winner because
it satisfies µhead, both moras being associated to the vowel.

Given these premises, the OT grammar unfolds as shown in Tableau 1.

Tableau 1.╇ Length and weight

/ ba: / dµ]σ µhead dep-µ(c) dep-µ(v) dep-link-µ *Vµµ
a. µ
| *! *
b a :
b. µµ
| | *! * *
b a :
 c. µµ
|â•›/â•›| * * * *
b a :
Markedness: *Vµµ = no bimoraic (long) vowels
dµ]σ (= wbp): voiced coda consonants are moraic
µhead = moras link to vowels
Faithfulness: dep-µ(v): do not insert moras on vowels
dep-µ(c): do not insert moras on consonants
dep-link-µ = do not insert new associations to µ
 Jean-Pierre Montreuil

3.â•… Lengthening and devoicing: Simple opacity

There is no interaction of rules in Western Lorrain that can create opacity. However, in
most other areas of Lorraine and in all of Alsace, devoicing comes into play and inter-
acts opaquely with lengthening. Final consonants are devoiced, as they tend to be in
Germanic and many other Indo-European languages (Brockhaus 1995; Itô & Mester
2003; Rubach 1990; Wiese 1996 for German, van Oostendorp 2008 for Dutch), includ-
ing several Romance languages e.g. Old French, Occitan, Catalan, Romansch. Alterna-
tions like [frεs] or [frε˜s] fraise “strawberry” vs. [frεzje] fraisier “strawberry bush” are
created.3 For instance, Pt 54, Pierre-la-Treiche, shows the following forms:
(2) bave [bε˜f] “drool” chemise [∫mẽ˜∫] “shirt”
tiède [tje˜t] “tepid” blouse [blu˜s] “overshirt”
fraise [frε˜s] “strawberry”

The simplest positional markedness analysis in OT (Prince & Smolensky [1993] 2004)
will interpret devoicing as resulting from a *vcecoda 〉〉 ident-vce 〉〉 *vce/obs constraint
ranking. We see no reason to diverge from this assumption here.

Tableau 2.╇ Devoicing

/ ba: / *vcecoda ident-vce *vce/obs
a. a. bag *! *
 b. b. bak *
Markedness: *vcecoda = codas are voiceless
*vce/obs = obstruents are voiceless
Faithfulness: ident-vce = faithfulness to input voice

In (1), lengthening occurred without devoicing. In all other areas, devoicing does
occur. In the devoicing zone, large areas show the same long vowels we find in some
voicing zones, and this of course leads to phonological opacity. In such areas, length
contrasts occur and minimal pairs can be observed, as in [vit] vite “quickly” vs.[vi˜t]
vide “empty” or [vœf] veuf “widower” vs. [vœ˜f] veuve “widow”. This is somewhat
reminiscent of the Friulan situation discussed in Hualde (1990) where lengthening

.╅ In most variants of Alsatian French, a [p], [t], [k] transcription represents a simplifica-
tion of the data. For instance, Philipp (1965) notes that in the French of Blaesheim devoiced
fricatives are properly represented as voiceless, ex. [f] rather than [v] but that such is not the
˚
case for plosives. Blaesheim has no voiced plosives /b d :/ and displays exclusively unvoiced
aspirated plosives in strong position: [b↜渀屮h], [d↜h] and [:h] and unvoiced non-aspirated plosives
˚ ˚ ˚
in weak position: [b], [d] and [:]. Carton et al. (1983:€62–65) make similar remarks for the
˚↜ ˚ ˚
French spoken in Riquewihr, with a slightly different notation. These facts do not invalidate
our analysis.
 Multiple opacity in Eastern Regional French 

and devoicing create gender alternations such as [lo˜f] vs. [lovә] “wolf ” m./f., [ru˜t] vs.
[rudә] “pure” m./f., [fre˜t] vs. [fredә] “cold” m./f., [bea˜t] vs. [beadә] “blessed” m./f.4
The geographical distribution of these forms needs to be clearly understood, as we
observe that the forms which are generated by opaque rule orders are in fact the most
prevalent across Lorraine. The lengthening isogloss cuts Lorraine diagonally in two
roughly equal halfs. Thus four areas emerge:
zone A (Southern Lorrain): lengthening, no devoicing. Ex: [ba˜:]
zone B (Eastern Lorrain): lengthening, devoicing. Ex: [ba˜k]
zone C (Western Lorrain): no lengthening, no devoicing. Ex: [ba:]
zone D (Northwestern Lorrain): no lengthening, devoicing. Ex: [bak]

Two examples, tiède “lukewarm” and lessive “wash”, are shown in (3), which illustrate
the relative areal importance of each manifestation of final-VC.
(3)
V:D (zone A) 16.4%
V:T (zone B) 29.1%
VD (zone C) ╇ 5.1%
VT (zone D) 49.4%
Example of area breakdown for Map€692 tiède (final occlusive)

V:D (zone A) 11.4%

V:T (zone B) 57%
VD (zone C) 10.1%
VT (zone D) 21.5%
Example of area breakdown for Map€701 lessive (final fricative)

The devoicing isogloss is relatively stable, while the length isogloss fluctuates and may
respond to the featural structure of the final consonant (and possibly the preceding
vowel). Namely, fricatives encourage lengthening. Accordingly, a number of points
that are found in zone D for occlusives belong to zone B with fricatives.
The challenging grammar is that of zone B, where opacity occurs. To the extent
that we normally observe vowel lengthening before voiced consonants, a [frε˜s]/[frεzje]
alternation seems paradoxical. The issue is not that we lack phonetic explanations to
correlate vocalic lengthening and consonantal devoicing: consonantal voice has been

.╅ In his often-cited analysis, Hualde explains this phenomenon by appealing to the sonority
associated with moraic elements: as codas devoice, they lose their ability to bear weight. Con-
sequently, their association with µ is severed, and µ is free to reassociate with the preceding
vowel, causing lengthening. Our discussion in Section€ 4 will exploit the concept of mora-
sharing, rather than Hualde’s mora-stranding; see also Blumenfeld (2006), Kavitskaya (2002)
and Morén (2001).
 Jean-Pierre Montreuil

shown to decay at the end of long syllables. Here, length and voicelessness appear
clearly distinct and occur in all possible combinations, as different grammatical
choices are made. In zone B, vowel-lengthening and consonant devoicing interact in
an opaque fashion: lengthening overapplies. If devoicing had occurred first, the moti-
vation for lengthening would have been non-existent. The counterbleeding scenario,
in which /ba:/ bague first lengthens to [ba˜:], then devoices to [ba˜k], is a familiar one
in derivational theory (DT).
The absence of intermediate forms in OT makes such situations problematic for
this theory. In the simplest case, each process is the result of a markedness contraint, or
set of contraints, dominating a faithfulness constraint, and while this allows for several
possible constraint rankings, none of them succeeds in writing a grammar in which
the opaque candidate is optimal. For instance, Tableau 3 offers a ranking consistent
with the M 〉〉 F scenario to evaluate the candidates which we encounter in the four dia-
lectal zones defined earlier. These four forms also correspond to the sympathetic (S),
opaque (O), faithful (F), and transparent (T) candidates in a standard opacity analysis
tableau. The grammar in Tableau 3 generates the transparent form as the winner, i.e.
with the devoiced consonant but no length on the vowel, and could apply to zone D.

Tableau 3.╇ Unsuccessful tableau for /ba:/: [ba˜k] (largest zone)

/ ba: / µhead *vcecoda ident-vce dep-link-µ
S zone A[ba˜:] *! *
O zone B[ba˜k] * *!
F zone C[ba:] *! *
T zone D[bak] *

Other constraint rankings presumably apply in other zones. For instance a ranking
like µhead, identvce 〉〉 dep-link-µ, *vcecoda would select [ba˜:] as in zone A, while a
ranking like dep-link-µ, identvce 〉〉 µhead, *vcecoda would select [ba:] as in zone€C.
The crucial point is that no ranking of these four constraints can generate [ba˜k], since
the transparent candidate will always violate a subset of the violations incurred by the
opaque candidate. The grammar of zone B, – again, the largest zone€– where the opaque
candidate surfaces, requires that enhancements to the theory be properly defined.

4.â•… Interaction of multiple processes: Double opacity in Alsatian French

As far as lengthening and devoicing are concerned, Alsace mostly patterns after
zone B of Lorraine: both processes interact opaquely and the favored form for /ba:/
is [ba˜k]. However, Alsace adds to this mix a third process which is not common in
Lorraine: schwa epenthesis (and subsequent resyllabification.) While Lorrain final
 Multiple opacity in Eastern Regional French 

obstruent€+ liquid clusters tend to simplify through deletion of the liquid: /kupl/ [kup]
“couple”, Alsatian repairs these sonority-defying clusters with schwa-epenthesis: /sikl/
[sikәl] “cycle”, /kupl/ [kupәl] “couple”. The [i] vowel in [sikәl] is short, which is what
is expected; recall that the source of length is the mora generated by a voiced coda in
the input, and this does not apply to /sikl/. The interesting interaction occurs of course
in words that end in a voiced consonant cluster, like /tabl/ “table”. The surface form is
[ta˜pәl]. The DT sequence runs as follows:
(4) 1. input: / tabl / “table”
2. length: [ta˜bl]
counterbleeding
3. devoicing: [ta˜pl]
counterbleeding
4. epenthesis: [ta˜pәl]
feeding
5. (resyllabification) [ta˜. pәl]

Epenthesis clearly needs to follow devoicing. In the reverse order, the labial consonant
would no longer be in coda position and would fail to devoice: the output form would
be *[ta˜bәl]. Forms like [ta˜pәl] show that devoicing does not occur simply because of
some phonetic correlation between greater vowel length and lesser consonantal voice
in word-final position, since [p] is word-internal in the output. The facts concern-
ing resyllabification are not entirely clear, and precise phonetic measurements and
psycholinguistic experiments will need to be performed. There is no suggestion in
the literature that the output is anything other than the expected Romance pattern,
with resyllabification into [si.kәl] and [ku.pәl]. There is also no indication that the
epenthetic vowel might be a low-level, intrusive vocalic element with no syllabic value,
as can sometimes be demonstrated from the timing of gestures, notably in the case
of metathesis (Hall 2006); here no pre-existing source of vocalicity exists. These are
important points to keep in mind since the second round of overapplication is induced
not by epenthesis itself, but by the resyllabification that it entails. A turbidity treatment
(Goldrick 2001) could be contemplated.5
Since devoicing and epenthesis are also in a counterbleeding relationship, the
total result is double opacity within the root. In OT, epenthesis means that sonseq 〉〉
dep-ә, where the markedness constraint penalizing reverse sonority codas dominates
phonotactic faithfulness.

.╅ Goldrick (2001) argues for an OT model with enriched outputs in which segments and
mora are associated by directional Projection and Pronunciation lines. This results in what
he terms ‘turbid’ representations which can account for a number of cases where opacity is
generated by misalignment or ambiguous alignment (ambisyllabicity, late resyllabification…)
 Jean-Pierre Montreuil

A Stratal OT analysis would require three different strata, even though there is
no morphology involved. In (5), I only consider a subset of the constraints, and omit
dµ]σ, which dominates in all strata, and dep-µ(v), vµµ and dep-link-µ which are
dominated in all strata.6

(5) Stratum One links input /tabl/ to [ta˜bl]

µhead, dep-ә 〉〉 identvce, dep-µ(v) 〉〉 *vcecoda, sonseq
Stratum Two links /ta˜bl/ to [ta˜pl]
µhead, dep-ә 〉〉 *vcecoda 〉〉 dep-µ(v), identvce, sonseq
Stratum Three links /ta˜pl/ to [ta˜pәl]
µhead, 〉〉 *vcecoda, sonseq 〉〉 dep-µ(v), identvce, dep-ә

5.â•… OT-CC and legitimate paths

While Stratal OT, as a multi-grammar serial model, makes a major contribution to

the analysis of alternations in morphologically complex words, it fails to convince
in the present case, where opacity affects even the simplest environments, roots and
monomorphemic words. In 2006, 2007, 2008, McCarthy presented OT-CC, a single-
grammar, faithfulness-based, version of OT which evaluates not single forms or even
sets of forms, but rather candidate chains. In candidate chains, each link differs from
the previous one by a lum (localized unfaithful mapping); from the initial form to the
terminal link, gradualness and harmonic improvement are required. As in Harmonic
Serialism, chains keep growing until convergence is reached. Again, the ranking order
remains stable.
The requirements of the theory (informally stated) are that (1) the initial form be
the most harmonic faithful parse of the input, (2) gradualness prevail: each new link
has all the lums of the preceding link, plus (only) one, and (3) harmonic improvement
obtain: each new link must be more harmonic than the preceding one and the most
harmonic option for that lum.
For instance, in the Lorrain matching of /ba:/ to [ba˜k], the following chains
would be invalid, even though they lead to the winner:
<ba:, ba˜k>skips lums, i.e. an intermediate lum is missing
<ba˜:, ba˜k>is not initiated by the faithful candidate
<ba:, bak, ba˜k>the last step is not harmonically improving

.â•… Depending on one’s analysis of resyllabification, a further stratum might be required

involving the reranking of alignment constraints. This point will not be pursued here.
 Multiple opacity in Eastern Regional French 

The following chains, on the other hand are valid, but not all of them lead to the most
prevalent winner. In fact they correspond to the four zones we delineated earlier:

Valid chains
faithful: <ba:> zone C
transparent <ba:, bak> zone D
sympathetic <ba:, ba˜:> zone A
opaque but winning <ba:, ba˜:, ba˜k> zone B

lum sequences need to be formulated strictly in terms of faithfulness. The first steps
are not controversial and for zones C and D, moraic structure unfolds in the way indi-
cated in (6):

(6) a. <b a :>

 sequence: <>

b. µ

<b a :, b a :>
 sequence: <-µ()>

c. µ µµ

<b a :, b a :, b a :> zone C

 sequence: <-µ(), -µ()>

d. µ µ

<b a :, b a :, b a k> zone D

 sequence: <-µ(), vce>

The controversial item is the terminal link in (8a), which calls for discussion. We have
adopted the view that a matching between an input VC and output A represents two
violations of faithfulness, namely dep-µ(v) and dep-µ(c), as shown in (7), while the
matching between the same input and output B represents three violations of faithful-
ness, namely dep-µ(v), dep-µ(c) and dep-link-µ, since in the latter matching there
exists in the output a µ-to-V association which is not present in the input.

(7) µ µ µ µ

input: V C output A: V C output B: V C

two F-violations two F-violations
 Jean-Pierre Montreuil

Shaw (2008) expresses a slightly different view. He argues that although a candidate
with a shared mora does not violate dep-link-µ with respect to an input that has no
mora on the final consonant, dep-link-µdoes play a role in the lum sequence, since the
mora and the segment are in correspondence in the structure of the previous link.

(8) a. µ µµ µµ

<b a :, b a :, b a :, b a :> zone A

 sequence: <-µ(), -µ(), --µ>

b. µ µµ µµ µµ

<b a :, b a :, b a :, b a :, b a k> zone B

 sequence: <-µ(), -µ(), --µ, vce>

In acknowledging that moras have always had a somewhat difficult status in faithful-
ness theory, McCarthy (2007:€ 76–77) argues that the “problem lies in the fact that
moras perform two main functions: they mark the distinctions between long and short
and between syllabic and non syllabic segments; and they “make position”, designating
some CVC syllables as heavy. Quantity and syllabicity are matters of faithfulness, since
they are both contrastive in some languages, but weight-by-position is probably not,
since the weight of codas never seems to be contrastive within a language (though see
Elfner 2006). It follows that changing a segment’s quantity or syllabicity is a lum, but,
say, resyllabifying a moraic coda as a nonmoraic onset is not”. In this view, adding or
deleting moras should come at no cost to faithfulness, but changes in quantity violate
ident constraints.
Lengthening could have been presented as resulting solely from the interaction of
markedness constraints, for instance from having µhead dominate such constraints
as *µshare and *V˜, thus suggesting that (6c) and (8a) represent not two different lum
sequences, but a single one. OT-CC however requires that each change (1) correspond
to an M 〉〉 F scenario and (2) alter the lum sequence. Our approach (as well as Shaw’s
original dep-linkµ proposal) is consistent with this OT-CC’s requirement. It provides
the path within which precedence can intervene to deal with opacity.7

.â•… One anonymous reviewer agrees that “in a chain such as (8a), and with respect to the
second mora, there is a definite Faith violation between the initial link [ba:] and the final link
[ba˜:]” but suggests that “the formalism using lines to connect moras and segments gets in
the way of expressing things properly.” This may indeed be a case where the expository conve-
nience of autosegmental associations could be challenged.
 Multiple opacity in Eastern Regional French 

The forms in zones A and C where devoicing does not come into play, since
identvce dominates *vcecoda, are not problematic. The challenge is to explain why in
the areas where *vcecoda dominates identvce, the transparent forms are not always
preferred to the opaque forms: why is there a zone B? This is where Precedence comes
into play.
In McCarthy (2007), Precedence constraints “favor certain relations among the
constituent lums of a sequence and penalize others”. In zone B, the opaque form wins
over the transparent form because Prec(dep-link-µ, identvce), named prec1 in
Tableau 4, requires that any violation of identvce be preceded by a violation of dep-
link-µ. While [ba˜k] satisfies this requirement (8b), [bak] fails to do so (6d), and thus
incurs a fatal violation of Prec(dep-linkµ, identvce).
Precedence refers to the lum sequence, not to the tableau: it is the lum sequence
in (6d) which fails to show the right precedence, while the lums in (8b) are properly
sequenced. In addition, a metaconstraint B 〉〉 Prec (A,B) ensures that B is never
affected by the precedence constraint. This means that in Tableau 4, Prec(dep-linkµ,
identvce) must find its place between identvce and dep-µ. In Tableau 4, only the ter-
minal links (TL) are indicated and lum sequences are not repeated. Again, the (now
optimal) opaque candidate is the zone B winner with the lum sequence given in (8b).

Tableau 4.╇ The role of precedence

/ ba: / *vcecoda identvce prec1 dep-µ
 opaque: TL [ba˜k] * *
transparent: TL [bak] * *!

(9) Prec 1 = prec(dep-linkµ, identvce), after identvce

do not devoice unless lengthening has occurred

As we move to Alsatian and consider the additional opacity generated by the epen-
thesis process, we will not observe a sudden increase in the complexity level of the
grammar, as we would in a Sympathy analysis of multiple opacity. Simply, a second
Precedence constraint will need to come into play.
An OT-CC analysis of the Alsatian facts can be sketched as follows: here again,
some potential chains will be declared invalid and remain out of consideration, for
reasons that were discussed earlier. For instance, sequences like <tabl, tapl, ta˜pәl> or
<tabl, ta˜bәl> are guilty of skipping links and/or failing the gradience test. Others show
a lack of harmonicity, even as they lead to the winner:

<tabl, ta˜bl, ta˜bәl, ta˜pәl> last step unharmonic

<tabl, tapl, ta˜pl, ta˜pәl> third step unharmonic
<tabl, tabәl, ta˜bәl, ta˜pәl>third and last steps unharmonic
 Jean-Pierre Montreuil

The grammar of the zone in which the optimal mapping /tabl/: [ta˜pәl]€is generated
will show two distinct parts:

a. the strongest constraints remain the regular markedness constraints, namely the
constraints on moraic association, as well as sonseq and *vcecoda. Together they
eliminate chains ending in schwaless terminal links: <tabl, tapl> and <tabl, ta˜bl,
ta˜pl> violate sonseq, while the faithful <tabl> violates all three.
b. Precedence constraints, in interaction with faithfulness constraints, evaluate
chains ending in outputs with a schwa. Just as prec1 responded to the opacity in
the length/voice interaction, a second precedence constraint, prec2, will reflect
the opacity of voice and sonority. It will penalize any chain where epenthesis
occurred without prior devoicing, i.e. where the dep-ә lum is not preceded by a
identvce lum. The grammar is summarized in (10) and (11).

(10) Constraint order: µ-head, sonseq, *vcecoda 〉〉 dep-ә 〉〉 prec2 〉〉 identvce 〉〉

prec1 〉〉 dep-link-µ
(11) Prec 2 =prec(identvce, dep-ә), after dep-ә
do not epenthesize unless devoicing has occurred

Tableau 5 presents a close-up on the appropriate grammar fragment, as it applies to

several chains. The (non-pertinent) higher constraints (µ-head, sonseq, *vcecoda) are
omitted to the left, while the various (non-determining) faithfulness constraints to
moras are implied on the right side.

Tableau 5.╇ Two precedence constraints in Alsatian

/ tabl/“table” dep-ә prec2 identvce prec1 dep-µ
 a. <tabl, ta˜bl, ta˜pl, ta˜pәl> * * *
b. <tabl, ta˜bl, ta˜bәl ta˜pәl> * *! * *
c. <tabl, ta˜bl, ta˜bәl> * *! *
d. <tabl, ta˜bl, ta˜bәl, ta˜pәl> * *! * *
e. <tabl, tapl, ta˜pl, ta˜pәl> * * *! *
f. <tabl, tabәl, ta˜bәl, ta˜pәl> * *! * *

6.â•… Conclusion

While transparent orders do not need special grammatical stipulations, opaque orders
do, but they must be accounted for by a mechanism which is itself constrained. OT-CC
reserves intervention to opaque orders. The same grammar stays in place, while the
candidate set is tailored to just a few plausible chains. More importantly, OT-CC forces
 Multiple opacity in Eastern Regional French 

a rethinking of what we evaluate: not mere phonetic forms, but paths. Not just what
segments are, but how they synchronically come to be.

References

Blumenfeld, Lev. 2006. Constraints on Phonological Interactions. Stanford: Ph.D. dissertation.

Brockhaus, Wiebke. 1995. Final Devoicing in the Phonology of German. Linguistische Arbeiten,
336. Tübingen: Niemeyer.
Carton, Fernand et al. 1983. Alsace. Les Accents des français. pp. 14–18. Paris: Hachette.
Elfner, Emily. 2006. “Contrastive Syllabification in Blackfoot”. West Coast Conference on For-
mal Linguistics 25. ed. by Donald Baumer, David Montero & Michael Scanlon, 141–149.
Somerville, MA: Cascadilla Press.
Goldrick, Matthew. 2001. “Turbid Output Representations and the Unity of Opacity”. North-
eastern Linguistics Society 30. ed. by M. Hirotani, A. Coetzee, N. Hall & J.-Y. Kim. Amherst,
231–245. MA: GLSA.
Hall, Nancy. 2006. “Cross-linguistic Patterns of Vowel Intrusion”. Phonology 23: 3. 387–429.
Hayes, Bruce. 1989. “Compensatory Lengthening in Moraic Phonology”. Phonology 7. 31–71.
Hualde, José I. 1990. “Compensatory Lengthening in Friulian”. Probus 2: 1. 31–46.
Itô, Junko & Armin Mester. 2003. “On the Sources of Opacity in OT: Coda processes in German”.
The Syllable in Optimality Theory, ed. by Caroline Féry & Ruben van de Vijver, 271–303.
Cambridge: Cambridge University Press.
Jesney, Karen. 2009. “Positional Faithfulness, Non-locality and the Harmonic Serialism Solution”.
Rutgers Optimality Archives: ROA-1018.
Kavitskaya, Darya. 2002. Contemporary Lengthening: Phonetics, phonology, diachrony. London:
Routledge.
Kimper, Wendell. 2008. “Local Optionality and Harmonic Serialism”. Rutgers Optimality
Archives: ROA-988.
Lanher, Jean et al. 1979–1989. Atlas linguistique et ethnographique de la Lorraine Romane.
Paris: CNRS.
Lee, Minkyung. 2007. OT-CC and Feeding Opacity in Javanese. Studies in Phonetics, Phonology
and Morphology 13: 2. 333–350.
McCarthy, John J. 2006. “Slouching towards Optimality: Coda reduction in OT-CC”. Phonological
Studies7. 89–104.
McCarthy, John J. 2007. Hidden Generalizations: Phonological opacity in Optimality Theory.
London: Equinox. Advances in Optimality Theory.
McCarthy, John J. 2008. “The Gradual Path to Cluster Simplification”. Phonology 25: 2. 271–319.
Morén, Bruce. 2001. Distinctiveness, Coercion and Sonority: A unified theory of weight. Routledge:
New York.
Philipp, Marthe. 1965. Le système phonologique du parler de Blaesheim: étude synchronique et
diachronique. Université de Nancy doctoral thesis.
Prince, Alan & Paul Smolensky [1993] 2004. Optimality Theory: Constraint interaction in genera-
tive grammar. Oxford: Blackwell.
Pruitt, Kathryn. 2008. “Iterative Foot Optimization and Locality in Stress Systems”. Rutgers
Optimality Archives: ROA-999.
 Jean-Pierre Montreuil

Riggs, Daylen. 2008. “Opacity in Icelandic: Transparency and OT with candidate chains”. North-
eastern Linguistics Society. New York: Ithaca.
Rubach, Jerzy. 1990. “Final Devoicing and Cyclic Syllabification in German”. Linguistic Inquiry
21: 1. 79–94.
Shaw, Jason. 2008. “Compensatory Lengthening via Mora Preservation in OT-CC: Theory and
predictions”. Rutgers Optimality Archives: ROA-916.
van Oostendorp, Marc. 2008. “Incomplete Devoicing in Formal Phonology”. Lingua 118.
1362–1374.
Wiese, Richard. 1996. The Phonology of German. Oxford: Clarendon Press.
Wolf, Lothar. 1983. Le français régional d’Alsace. Etude critique des alsacianismes. Paris: Klincsieck.
Zec, Draga. 2003. “Prosodic Weight”. The Syllable in Optimality Theory. ed. by Caroline Féry &
Ruben van de Vijver, 123–143. Cambridge: Cambridge University Press.
part ii

Syntax
 Syntactic variation in Colombian Spanish
The case of the Focalizing Ser (FS) structure*

Dunia Catalina Méndez Vallejo

West Virginia University

This article examines some of the syntactic properties of the Focalizing ser (FS)
structure in Colombian Spanish. The FS has been reported in only a few dialects
of Spanish (Venezuelan, Colombian, Ecuadorian, Dominican, and Panamanian),
and it is not stigmatized, despite being dialectally marked. Although the FS
shows some resemblance to the pseudo-cleft construction, the syntactic analysis
included here is based on the premise that the two forms are syntactically
different. Looking at the syntactic contexts in which this form may occur, I
show that the FS may precede any kind of phrase, serving any kind of syntactic
function, as long as it is post-verbal. Furthermore, I illustrate that the agreement
relation between FS ser and certain sentential constituents is not random. At the
end, I present compelling evidence to suggest that the FS is in a TP-internal Focus
Phrase, generated below TP and above vP.

Key words:╇ TP-internal focus; functional projection; information structure;

dialectal variation

1.â•… Introduction

Pseudo-cleft constructions are often used in Spanish to focalize certain constituents

within a sentence. A sentence such as (1), for example, can be transformed into a
pseudo-cleft sentence such as (2), where the direct object (lingüística “linguistics”)

*I am grateful to Iván Ortega Santos, Mary Kato, and Chad Howe for stimulating feedback.
I am also grateful to the audience at LSRL 39 for their comments and interest on this topic.
All errors and shortcomings are my own.
 Dunia Catalina Méndez Vallejo

is the focused constituent and is c-commanded by the copula verb fue “was” (> ser
“to be”):1
(1) Juan estudió lingüística
Juan study.3sg.pret linguistics
“Juan studied linguistics”
(2) Lo que Juan estudió fue lingüística
the.neut comp Juan study.3sg.pret be.3sg.pret linguistics
“What Juan studied was linguistics”

In certain Spanish dialects (i.e. Venezuelan, Ecuadorian, Panamanian, Colombian, and

Dominican), a sentence involving the same focus interpretation can also be created
without the relative clause (lo que “what”), as shown in (3):
(3) Juan estudió fue lingüística
Juan study.3sg.pret be.3sg.pret linguistics
“It was linguistics that Juan studied”

The structure in (3), here referred to as the Focalizing Ser structure (henceforth FS),
has not been widely studied. It has been reported to occur in Brazilian Portuguese
(Mikolajczak 2003; Oliveira & Braga 1997), Venezuelan Spanish (Sedano 1988, 1990,
1994, 1995, 2003a, 2003b), Colombian Spanish (Albor 1986; Curnow & Travis 2003),
Caribbean Spanish (Bosque 1999; Camacho 2006), and Dominican Spanish (Toribio
1992, 2002). In terms of the syntactic study of the FS, Bosque, Toribio, and Camacho
have each proposed their own theoretical analyses of this dialectally-marked form.
However, all of these proposals fail to account for many empirical facts and present
some serious conceptual problems.
The aim of this paper is to provide an account of the syntactic properties of the
FS, specifically in terms of the kind of constituents that can (or cannot) be FS-focused
and the morphology of FS ser. Based on the data here presented,2 I briefly discuss my

.╅ Pseudo-clefts have been extensively studied in Spanish (e.g. Guitart 1989; Sedano 1990;
Toribio 1992, 2002). Toribio (2002:€139), in particular, suggests that pseudo-clefts in Spanish
comprise an FP (Focus Phrase) complement whose head selects a CP or IP. Hence, the focused
constituent is generated in [Spec, FP], ser (‘to be’) moves from F to I, and the embedded CP
raises to [Spec, IP].
.╅ Some of the examples presented here are based on documented utterances from naturally-
occurring conversations that took place in Bucaramanga, Colombia, in the summer of 2008.
Other examples come from acceptability judgments on 125 sentences (84 containing the FS
and 41 containing distracters such as clefts and pseudo-clefts). These sentences were evalu-
ated by 45 university students from Bucaramanga after having completed a series of tutorial
tasks. The sentences were part of a mini-dialogue which was presented in audio format,
and the participants were asked to rate them using a scale from 1 (least acceptable) to 5
(most acceptable). The participants heard the mini-dialogue and were prompted to choose
 Syntactic variation in Colombian Spanish 

theoretical proposal for the FS, as I show that the FS needs to be analyzed as a TP-
internal focus structure.
This paper is organized as follows: Section€2 provides some evidence supporting
the idea that the FS is not a pseudo-cleft and must be analyzed as an independent
syntactic structure; Section€3 describes what can (and cannot) be FS-focused and the
agreement relation found between FS ser and certain sentential constituents; Section€4
introduces my theoretical proposal for the FS; and, Section€5 states some conclusions
and generalizations.

2.â•… The FS as an alternative focus structure

According to some studies (Albor 1986; Sedano 1988, 1990, 1994, 1995, 2003a, 2003b;
Toribio 1992, 2002), the FS is viewed as structurally equivalent to the pseudo-cleft.
The common claim established in all these studies is that while CP elements (a rela-
tive pronoun and a complementizer) are pronounced in the pseudo-cleft, they are not
pronounced in the FS.
Albor, for example, describes the FS as a syntactic derivation of the pseudo-cleft.
Hence, an FS-sentence such as (4) would result from a series of transformations, as
shown in (5):

(4) Él necesita es descansar

he need.3sg.pres be.3sg.pres rest.inf
“It is resting that he needs”

(5) Deep structure: DESCANSAR ES DESCANSAR Y ÉL

NECESITA DESCANSAR
“To rest is to rest and he needs to rest”
Transposition of the goal: Descansar es descansar y
DESCANSAR él necesita
“To rest is to rest and to rest he needs”
Relativization of the goal: *Descansar es descansar QUE él necesita
“To rest is to rest that he needs”
Transposition of the antecedent: *Es descansar DESCANSAR
que él necesita
“It is to rest to rest that he needs”
Pronominalization of the antecedent: Es descansar LO que él necesita
“It is to rest what he needs”

a number (from 1 to 5) by clicking on a computer screen. See Méndez Vallejo (2009) for a
discussion of methodological issues and for a complete description of the data collection.
 Dunia Catalina Méndez Vallejo

The goal of ser becomes a subject: LO QUE ÉL NECESITA es descansar

“What he needs is to rest”
The relative pronoun is omitted: QUE él necesita es descansar
“That he needs is to rest”
Que is omitted: El necesita es descansar
“He needs is to rest” [Albor 1986:€184]

Sedano (2003a, 2003b), based on her statistical analysis of the FS in Caracas Spanish,
suggests that although the FS and the pseudo-cleft are practically equivalent, the FS is
structurally simpler (it lacks a relative clause), which facilitates the production of focus
(i.e. it allows last-minute focalization).
Finally, Toribio proposes a unified syntactic analysis of the FS and the pseudo-
cleft, and claims that the FS “is the null operator counterpart of the traditional pseudo-
cleft.” (Toribio 2002:€134). Thus, according to her analysis, Examples (2) and (3) above
would be structured as shown in (6) and (7), respectively:

(6) Lo que Juan estudió fue

[lo (pro) [cp op que [ip Juan estudió t…
the.neut that Juan study.3sg.pret be.3sg.pret
lingüística
linguistics
“What Juan studied was linguistics”

(7) Lo que Juan estudió fue

[lo (pro) [cp op que [ip Juan estudió t…
the.neut that Juan study.3sg.pret be.3sg.pret
lingüística
linguistics
“What Juan studied was linguistics” [Toribio 2002:€134]

Although analyzing the FS as an incomplete form of the pseudo-cleft seems to be a

simple and appealing solution, later research shows that the FS is indeed an inde-
pendent syntactic structure. Following Curnow and Travis (2003), for example, it is
clear that the pseudo-cleft is sensitive to clitic climbing (8)–(9), whereas the FS is
not (10)–(11):

(8) okLo que quiero es irme

the.neut comp want.1sg.pres be.3sg.pres go.inf-cl
“What I want is to leave”

(9) *Lo que mei quiero es iri

the.neut comp cl want.1sg.pres be.3sg.pres go.inf
“What I want is to leave”
 Syntactic variation in Colombian Spanish 

(10) okQuiero es irme

want.1sg.pres be.3sg.pres go. inf-cl
“It is leaving that I want to do”
(11) okMei quiero es iri
cl want.1sg.pres be.3sg.pres go.inf
“It is leaving that I want to do”

In addition, both Curnow and Travis and Bosque (1999) claim that the pseudo-cleft
cannot focus negative polarity items (12), whereas the FS can (13):

(12) *El que no vino fue nadie

the.masc comp not come.3sg.pret be.3sg.pret nobody
“The one who did not come was nobody”

(13) okNo vino fue nadie

not come.3sg.pret be.3sg.pret nobody
“It was nobody who came”

Bosque also shows that focused constituents can be extracted out of a pseudo-cleft
structure (14), but not out of an FS structure (15):

(14) ok¿Quién fue el que salió?

who be.3sg.pret the.masc comp leave.3sg.pret
“Who was it that left?”

(15) *¿Quién salió fue?

who leave.3sg.pret be.3sg.pret
“Who was it that left?”

Finally, I have observed that progressives cannot be focalized by the pseudo-cleft (16),
but they can by the FS (17):

(16) *Lo que ha estado es haciendo frío

the.neut comp have.3sg.pres be.perf be.3sg.pres do.progr cold
“What it has been doing is cold”

(17) okHa estado es haciendo frío

have.3sg.pres be.perf be.3sg.pres do.progr cold
“It is being cold that it has been”

Looking at the data in (8)–(17), it becomes evident that the FS and the pseudo-cleft
do not react similarly to certain syntactic phenomena (e.g. occurring in cases of clitic
climbing and focusing negative polarity items, extracted constituents, and progres-
sives). Thus, the fact that the FS and the pseudo-cleft are not interchangeable in all
syntactic contexts strongly suggests that they do not share the same structural configu-
ration. Given this, the analysis that I present in this paper takes as a starting point the
 Dunia Catalina Méndez Vallejo

assumption that the FS is not a pseudo-cleft but a completely independent syntactic

structure. In fact, as I will propose in Section€4, the FS is a TP-internal functional projec-
tion, which clearly suggests that the FS does not correlate with a separate clause (as it
is the case in the pseudo-cleft structure).

3.â•… Syntactic properties of the FS

Previous analyses on the FS have discussed some of the syntactic contexts in which
this form may and may not occur (e.g. Sedano 1990, 2003a, 2003b; Bosque 1999;
Curnow€& Travis 2003). Sedano, for example, has conducted several statistical studies
of the FS in Caracas Spanish and has concluded that it is more commonly found when
it precedes prepositional phrases, adverbs, and adjectives. Curnow and Travis also pro-
vide a quantitative analysis of the FS in Colombian Spanish and suggest that it tends
to occur more often before adjuncts (i.e. prepositional phrases and adverbs). Bosque,
following a more theoretical approach, shows some syntactic contexts in which the FS
may (or may not) occur, which he takes as evidence against the idea that the FS is a
pseudo-cleft.
Although these are some interesting findings, it is still unclear exactly where the
FS can be placed and which constituents can (and cannot) be FS-focused. Further-
more, the morphology of FS ser has been loosely mentioned in some studies (Bosque
1999; Curnow & Travis 2003), but it has not been properly described and explained.
In this section, I will define not only the type of constituents that can be FS-
focused (Section€3.1), but also the agreement pattern found between FS ser and the
matrix verb, and between FS ser and the focused element (Section€3.2). The analysis
that I offer in this paper is purely syntactic and does not intend to draw any kind of
quantitative results.

3.1â•… What gets focused by the FS?

The FS is used in Colombian Spanish to focalize a wide variety of phrases. For example,
it can be found preceding prepositional phrases (PPs), as shown in (18)–(19):
(18) Vamos es para la casa
go.1pl.pres be.3sg.pres to the house
“It is home that we are going to”

(19) Nevó fue en Berlín

snow.3sg.pret be.3sg.pret in Berlin
“It was in Berlin that it snowed”

The FS can also be used to focus single adjectives (20), or entire adjectival phrases (21):
 Syntactic variation in Colombian Spanish 

(20) La blusa me quedó fue grande

the blouse cl fit.3sg.pret be.3sg.pret big
“It was big that the blouse was on me”
(21) Los visitantes están es súper cansados
the visitors be.3pl.pres be.3sg.pres super tired
“It is super tired that the visitors are”

Furthermore, the FS may precede single adverbs (22), or entire adverbial phrases (23):
(22) Llegaron fue ayer
arrive.3pl.pret be.3sg.pret yesterday
“It was yesterday that they arrived”
(23) En ese entonces ganaba era casi siempre lo mismo
in that time win.1sg.imp be.3sg.imp almost always cl same
“At that time, it was almost always that I used to win the same thing”

The adverbs in (22)–(23) have been referred to in the literature as “VP-adverbs”

(cf. Zagona 2002) or “lower adverbs” (cf. Cinque 1999). As shown, this kind of adverbs
is felicitously FS-focused. However, as illustrated in (24)–(25), “IP-adverbs” (cf. Zagona
2002) or “higher adverbs” (cf. Cinque 1999) must not be FS-focused:
(24) *No veo es francamente nada
not see.1sg.pres be.3sg.pres frankly nothing
“It is frankly that I do not see anything”
(25) *Estás es naturalmente loca
be.2sg.pres be.3sg.pres naturally crazy
“It is naturally that you are crazy”

In addition to VP-adverbs, the FS may precede determiner phrases (DPs) serving as

direct objects (26), indirect objects (27), and even subjects (28):3
(26) La tía Carmen nos trajo fue una torta
the aunt Carmen cl bring.3sg.pret be.3sg.pret a cake
“It was a cake that my aunt brought for us”
(27) Nos mostraron las fotos fue a nosotros
cl show.3pl.pret the pictures be.3sg.pret to we
“It was we to whom they showed the pictures”

.╅ Toribio (1992, 2002) claims that post-verbal subjects must not be focused by the FS in
Dominican Spanish. However, based on the acceptability judgments that I conducted and data
that I have gathered from T.V. programs, I have found that FS-focused post-verbal subjects are
completely acceptable in Colombian Spanish.
 Dunia Catalina Méndez Vallejo

(28) Esta mañana habló fue el alcalde

this morning speak.3sg.pret be.3sg.pret the mayor
“It was the mayor who spoke this morning”

Importantly, the FS may not precede DPs that have moved to positions higher than
T’, such as CP or Spec, TP. For example, in sentences (29)–(30), the object DPs have
moved from vP to a Topic Phrase above TP, and the subject DP has moved from vP to
Spec, TP:4

(29) *Una tortai la tía Carmen nos trajo fue ___

a cake the aunt Carmen cl bring.3sg.pret be.3sg.pret
“It was a cake that my aunt brought for us”

(30) *A nosotros nos mostraron las fotos fue ___

to we cl show.3pl.pret the pictures be.3sg.pret
“It was we to whom they showed the pictures”

(31) *El alcalde esta mañana habló fue ___

the mayor this morning speak.3sg.pret be.3sg.pret
“It was the mayor who spoke this morning”

Besides DPs, the FS may precede in-situ interrogative words, specifically those uttered
in echo-questions:

(32) A: Carlitos y Marinita aprendieron inglés en Cambridge

“Carlitos and Marinita learned English in Cambridge”
B: ¿Aprendieron inglés fue dónde?
learn.3pl.pret English be.3sg.pret where
“It was where that they learned English?”

(33) A: El profesor dijo que había examen y trabajo final

“The teacher said that there was an exam and a final paper”
B: ¿Dijo fue qué?
say.3sg.pret be.3sg.pret what
“It was what that he said?”

.╅ Much debate still remains regarding the final placement of pre-verbal subjects in
Spanish. It has been proposed, for example, that pre-verbal subjects check nominative case
within VP and later move to Spec, IP (TP) to check EPP features (e.g. Contreras 1991). Other
authors argue that pre-verbal subjects are placed in a Topic Phrase and move to Spec, IP
(e.g. Mejías-Bikandi 1992) or above IP (e.g. Ordóñez 1997). Given that the placement of
pre-verbal subjects does not affect my theoretical proposal for the FS, I will simply assume
here that pre-verbal subjects move to Spec, TP.
 Syntactic variation in Colombian Spanish 

However, as initially pointed out by Bosque (1999), the FS may never precede inter-
rogative words that have been extracted to CP:

(34) *¿Dóndei aprendieron inglés fue ____

where learn.3pl.pret English be.3sg.pret
Carlitos y Marianita?
Carlitos and Marianita
“It was where that Carlitos and Marianita learned English?”

(35) *¿Quéi dijo fue ____ el profesor?

what say.3sg.pret be.3sg.pret the teacher
“It was what that the teacher said?”

Finally, I have observed that the FS may focus perfectives (36), progressives (37), or a
combination of a perfective and a progressive (38):

(36) La porción la han es aumentado

the portion cl have.3pl.pres be.3sg.pres increase.perf
“It is increased that the portion has become”

(37) Voy es saliendo

go.1sg.pres be.3sg.pres leave.progr
“It is leaving that I am doing”

(38) Había era estado estudiando toda la noche

have.3sg.imp be.3sg.imp be.perf study.progr all the night
“It was studying that he had been doing all night”

Interestingly, the FS cannot focus auxiliary verbs that have moved to T:

(39) *Era había estado estudiando toda la noche
be.3sg.imp have.3sg.imp be.perf study.progr all the night
“It was studying that he had been doing all night”

So far, I have shown that the FS may focus a wide variety of phrases (e.g. PPs, AdjPs,
AdvPs, DPs, PerfPs, ProgrPs). Looking in particular at FS-focused DPs, it is clear
that a phrase may be FS-focused, irrespective of its syntactic or semantic role within
the sentence (e.g. subject, direct object, indirect object). However, as illustrated in
Examples€(24)–(25), (29)–(31), (34)–(35), and (39), the FS never precedes a constitu-
ent that has moved to a pre-verbal position, or a verb placed in T (i.e. a matrix or an
auxiliary verb).
Based on these empirical observations, the generalization that I will establish here
is that the FS may focus any type of constituent as long as it is post-verbal. This finding
will become particularly crucial for my syntactic analysis of the FS. As I will explain
in more detail in Section€4, having the FS preceding only post-verbal elements clearly
suggests that it constitutes a TP-internal functional projection.
 Dunia Catalina Méndez Vallejo

3.2â•… The morphology of FS ser

The morphological characteristics of FS ser have been very little discussed in the litera-
ture. Although Bosque (1999) and Curnow and Travis (2003) have mentioned a few
peculiarities regarding FS ser morphology, there seems to be certain tendency to evade
this topic or to treat it as an unimportant issue. In this paper, I would like to emphasize
the complexity of FS ser morphology and to describe two agreement processes that I
have unveiled: one, observed between FS ser and the verb in T, and another, between
FS ser and the focused constituent.5
In terms of the first agreement process, FS ser must agree in tense with the verb in
T. As shown in (40)–(41), for example, the sentence becomes felicitous only when both
the matrix verb and FS ser agree in tense:
(40) okEsa actriz está es loca
that actress be.3sg.pres be.3sg.pres crazy
“It is crazy that that actress is”
(41) *Esa actriz estaba es loca
that actress be.3sg.past be.3sg.pres crazy
“It is crazy that that actress was”

Furthermore, FS ser and the verb in T must also agree in aspect:6

(42) okEsa actriz estaba era haciéndose la loca
that actress be.3sg.imp be.3sg.imp do.progr-cl the crazy
“It was pretending to be crazy that that actress was doing”
(43) *Esa actriz estaba fue haciéndose la loca
that actress be.3sg.imp be.3sg.pret do.progr-cl the crazy
“It was pretending to be crazy that that actress was doing”

.╅ Given the brevity of this article, I will only focus on describing the agreement pattern
between FS ser and the verb in T, and between FS ser and the focused constituent. For a more
complete account of FS ser morphology and a theoretical explanation of such agreement,
please see Méndez Vallejo (2009).
.╅ One of the reviewers reports a case of non-agreement (Yo dije a las doce era por la radio)
occurring in natural discourse from a corpus of Colombian Spanish. In this case, it seems
that the verb in T (dije [1.SG.PRET]) does not agree with FS ser in Aspect (era [3SG.IMP]).
Personally, I find this sentence completely unacceptable and it is difficult to understand its
meaning in isolation. Looking at the results of my acceptability judgment tests, I find that
sentences similar to this one are ruled out by my informants. For example, a sentence such
as Ellos eran es buceadores profesionales has an average score of 1.7 and a sentence such as El
comité olímpico había premiado es a los atletas has an average score of 2.1 (the acceptability
judgment scale goes from 1 [least acceptable] to 5 [most acceptable]).
 Syntactic variation in Colombian Spanish 

However, FS ser does not agree with the verb in T in either person or number:

(44) okLes traje fue unas empanadas de queso

cl bring.1sg.pret be.3sg.pret some empanadas of cheese
“It was some cheese empanadas that I brought for them”

(45) *Les traje fui unas empanadas de queso

cl bring.1sg.pret be.1sg.pret some empanadas of cheese
“It was some cheese empanadas that I brought for them”

(46) okLes trajimos fue unas empanadas de queso

cl bring.1pl.pret be.3sg.pret some empanadas of cheese
“It was some cheese empanadas that we brought for them”

(47) *Les trajimos fuimos unas empanadas de queso

cl bring.1pl.pret be.1pl.pret some empanadas of cheese
“It was some cheese empanadas that we brought for them”

As for the second agreement process, FS ser shows a more complicated pattern as it
appears to hold different agreement relations with post-verbal subjects and objects
(DOs and IOs). First, when the FS focuses post-verbal subjects, FS ser agrees with
them in both person and number:7

(48) okSaqué la basura fui yo

take-out.1sg.pret the garbage be.1sg.pret I
“It was I who took the garbage out”

(49) *Saqué la basura fue yo

take-out.1sg.pret the garbage be.3sg.pret I
“It was I who took the garbage out”

(50) *Saqué la basura fuimos yo

take-out.1sg.pret the garbage be.1pl.pret I
“It was I who took the garbage out”

.╅ Notice that the verb in T must agree with the post-verbal subject in person and number,
regardless of FS ser. Although it is possible to suggest that FS ser agrees with both the verb in
T and the post-verbal subject in person and number, I argue here that it only maintains this
kind of agreement with the latter, given that in cases such as (44)–(46) FS ser does not agree
with the verb in T in either person or number. I recognize that it is difficult to show that FS
ser only agrees with the post-verbal subject in person and number, as there are no cases of
disagreement between the subject and the verb in T. For a lengthier discussion on FS-focused
post-verbal subjects, see Méndez Vallejo (2009).
 Dunia Catalina Méndez Vallejo

Second, when the FS precedes direct objects, it optionally agrees with them in number.
For example, when the DO is plural FS ser may remain in singular (51), or become
plural too (52):8
(51) okTengo es dos gatos
have.1sg.pres be.3sg.pres two cats
“It is two cats that I have”
(52) okTengo son dos gatos
have.1sg.pres be.3pl.pres two cats
“It is two cats that I have”

Importantly, FS ser does not establish any person agreement with the DO. As shown in
(53)–(54), the sentence is felicitous only when FS ser is conjugated in third person:
(53) okA Hilda le habías comprado era dos periquitos
to Hilda cl have.2sg.imp buy.perf be.3sg.imp two parakeets
“It was two parakeets that you had bought for Hilda”
(54) *A Hilda le habías comprado eras dos periquitos
to Hilda cl have.2sg.imp buy.perf be.2sg.imp two parakeets
“It was two parakeets that you had bought for Hilda”

The agreement variation shown between FS ser and DOs (in (51)–(52)) may reside
entirely on contrast. That is, if the FS is used in a non-contrastive context (55) number
agreement is not required. However, if the FS is used in a contrastive context (56),
number agreement is required:9
(55) A: ¿Tienes mascotas?
“Do you have any pets?”
B: Sí, tengo (ok es/*son) dos perros y un gato
yes have.1sg.pres be.3sg/3pl.pres two dogs and a cat
“Yes, it is two dogs and a cat that I have”
(56) A: ¿No tenías dos gatos y un perro?
“Did you not have two cats and a dog?”
B: No, tengo (*es/ok son) dos perros y un gato
No have.1sg.pres be.3sg/3pl.pres two dogs and a cat
“Yes, it is two dogs and a cat that I have”

.╅ In cases of FS-focused singular DOs, FS ser always takes singular morphology ([3SG]). As
discussed in Méndez Vallejo (2009), I argue that this is a case of default morphology rather
than a case of agreement with singular DOs.
.╅ A more detailed discussion concerning the relation between contrastive focus and FS ser
morphology can be found in Méndez Vallejo (2009).
 Syntactic variation in Colombian Spanish 

Interestingly, FS ser never establishes person or number agreement with IOs (57), or
with DOs that have Differential Object Marking (58):

(57) Les habíamos comprado un videojuego

cl have.1pl.imp buy.perf a videogame
(ok era/*eran) a mis primos
be.3sg/3pl.imp to my cousins
‘It was my cousins to whom we had bought a videogame’

(58) Besó (ok fue/*fueron) a mis primos

kiss.3sg.pret be.3sg/3pl.pret to my cousins
‘It was my cousins whom he kissed’

In fact, this lack of agreement between FS ser and IOs (and DOs with Differential
Object Marking) holds in both non-contrastive (59) and contrastive (60) contexts:

(59) A: ¿A quién le diste tus joyas?

“To whom did you give your jewelry?”
B: Se las di (ok fue/*fueron) a los revendedores
cl cl give.1sg.pret be.3sg/3pl.pret to the scalpers
“It was the scalpers to whom I gave them”

(60) A: ¿No les diste tus joyas a los fabricantes?

“Did you not give your jewelry to the manufacturers?”
B: No, se las di (ok fue/*fueron) a los revendedores
no cl cl give.1sg.pret be.3sg/3pl.pret to the scalpers
“No, it was the scalpers to whom I gave them”

Until this point, the data here presented shows that FS ser establishes two separate
agreement relations: one the one hand, it agrees with the verb in T in both tense and
aspect; on the other, it agrees in person and number with post-verbal subjects, and in
number with DOs (with no Differential Object Marking). It is important to clarify that
the FS does not establish any kind of agreement with IOs or with DOs that have Differ-
ential Object Marking. Also, the number agreement variability found between FS ser
and DOs may be explained in terms of contrastive focus. That is, if the FS is produced
in a non-contrastive context, it does not agree with the DO in number; but if the FS is
uttered in a contrastive context, it must agree with the DO in number.
To summarize this section, I will now point out my main findings: (a) Any con-
stituent may be FS-focused as long as it is post-verbal; (b) FS ser agrees with the verb
in T (the matrix verb or the auxiliary verb) in both tense and aspect; (c) FS ser agrees
with post-verbal subjects in person and number, and with DOs (without Differential
Object Marking) only in number; and (d) FS ser does not establish any agreement with
IOs or with DOs that have Differential Object Marking.
 Dunia Catalina Méndez Vallejo

4.â•… What is the FS?

Starting from a view of the FS as an independent structure (not as an incomplete form

of the pseudo-cleft), Bosque (1999) and Camacho (2006) have proposed their own
syntactic accounts of this form. Bosque, for example, argues that the FS is a Focus
Phrase generated inside VP, where the non-copulative verb ser functions as the head
of the focus projection. Hence, a sentence such as (61) is analyzed by Bosque as shown
in (62):
(61) Juan comía era papas
Juan eat.3sg.imp be.3sg.imp potatoes
“It was potatoes that Juan used to eat”
(62) [ip Juani [vp ti [v comía [fp [[f0 era ] papas ]]]]] [Bosque 1999:€4]
Although this proposal correctly predicts most of the cases shown in Section€3, it fails
to account for FS-focused auxiliaries (see (36)–(38)), as these would not be located
inside VP, but immediately above it.10
Camacho, on the other hand, claims that the FS is an equative structure (i.e. a
Copula Phrase) that originates as an adjunct of VP. In this kind of construction, the
subject is null and the predicate is the focused VP. Thus, following this proposal, a
sentence such as (63) is derived as illustrated in (64):
(63) Los pájaros se comieron fue las migas
the birds cl eat.3pl.pret be.3sg.pret the crumbs
“It was the crumbs that the birds ate”
(64) [ip Los pájaros [I’ se comieron [vp [vp [v] ei] [ip cop x
[i’cop fue [las migasi]]]]]] [Camacho 2006:€19]
Under this view, Camacho also argues that the FS must only focus single, domain-
final constituents. Given this, sentences such as (65)–(66) are ruled out by Camacho
because in the former case there are more than two constituents being focused, and in
the latter the focused constituent is not placed at the end of the relevant domain (the
TP clause):
(65) *Marta le compró fue pan a su abuela
Marta cl buy.3sg.pret be.3sg.pret bread to her grandmother
“It was bread to her grandmother that Marta bought”
(66) *Marta le compró fue pan a su abuela
Marta cl buy.3sg.pret be.3sg.pret bread to her grandmother
“It was bread that Marta bought for her grandmother” [Camacho 2006:€20]

.╅ As suggested in the literature (e.g. Klein 1968; Zagona 2002), auxiliary verbs are generated
inside Auxiliary Phrases (Perfective Phrases, or Progressive Phrases) above vP.
 Syntactic variation in Colombian Spanish 

Although Camacho’s proposal seems to correlate with certain informational proper-

ties found in Spanish (e.g. focus aligns with the nuclear peak of the sentence, which is
typically domain-final), it runs into some serious empirical problems. First of all, his
theoretical proposal (exemplified in (64)) cannot account for FS-focused perfectives
and progressives (see (36)–(38)). Furthermore, contrary to Camacho’s claim, I have
found that the FS can focus more than one constituent, or a constituent that is not
domain-final. Thus, sentences such as (65)–(66) above are, in fact, perfectly acceptable
in Colombian Spanish.
Based on the observations that I have made so far, and following previous
research evidencing the existence of TP-internal functional projections (cf. Horvath
1986; Yanagida 1995; Belletti 2004; Hsu 2008), I propose here that the FS generates as
a Focus Phrase (FocP), below T and above vP. Thus, a sentence such as (67) should be
analyzed as sketched in (68):
(67) Carolina nació fue en Málaga
Carolina born.3sg.pret be.3sg.pret in Málaga
“It was in Málaga that Carolina was born”
(68) [tp Carolinai [t′ naciój [focp fue [vP ti [v tj [pp en Málaga ]]]]]]
Analyzing the FS as a TP-internal Focus Phrase correctly accounts for the data shown
in Section€3, as it predicts that: (a) only post-verbal subjects may be FS-focused; and (b)
constituents placed above vP and below T (e.g. PerfPs, ProgrPs) may be FS-focused.
Hence, having the FS in a TP-internal Focus Phrase placed below T allows us to
explain why the FS may not occur sentence-initially (69), before pre-verbal subjects
(70) or above matrix verbs (71):
(69) *Fue a Cecilia le entregaron el diploma
be.3sg.pret to Cecilia cl deliver.3pl.pret the diploma
“It was Cecilia to whom they delivered the diploma”
(70) *En esa época era Ximena iba a las clases de ballet
in that time be.3sg.imp Ximena go.3sg.imp to the classes of ballet
“It was Ximena who went to Ballet classes at that time”
(71) *Es estudia en Medellín
be.3sg.pres study.3sg.pres in Medellín
“It is in Medellín where he studies”
In turn, placing the FS above vP (not inside vP) further allows us to account for FS-
focused perfectives and progressives:
(72) [ip Valentina [i′ había [focp era
Valentina have.3sg.imp be.3sg.imp
[perf soñado [vP mucho]]]]]
dream.perf much
“It was dreaming that Valentina had done a lot”
 Dunia Catalina Méndez Vallejo

(73) [ip Valentina [i′ estaba [focp era [progrp soñando]]]]

Valentina be.3sg.imp be.3sg.imp dream.progr
“It was dreaming that Valentina was doing”

Finally, based on the assumption that both focus and topic projections occur TP-
internally (cf. Belletti 2004; Hsu 2008), and taking into account cases such as (74),11
I will further claim that the FS may be preceded by a Topic Phrase (TopP) where
discourse-old constituents can move to:

(74) A: ¿Si te contaron que Horacio está podrido en plata?

“Did they not tell you that Horacio is filthy rich?”
B: Lo dudo… tendrá plata será la familia
cl doubt.1sg.pres have.3sg.fut money be.3sg.fut the family
“I doubt it…It might be his family who has money”

In brief, the syntactic structure that I propose for the FS can be illustrated as in
(75) below:
(75) TP

T′

T TopP

FocP

SER PerfP

ProgrP

nP

n′

υ V

V XP

.╅ Notice that I have already shown similar cases in Section€3 (see Examples (32), (48),
and (57)).
 Syntactic variation in Colombian Spanish 

5.â•… Conclusions

The FS has been reported and described in the literature (Albor 1986; Sedano 1990),
but its formal properties have not been completely clarified, despite several efforts
(Bosque 1999; Toribio 1992, 2002; Curnow & Travis 2003; Camacho 2006). This
dialectally-marked form has been previously examined as an incomplete form of a
cleft structure (Albor 1986; Sedano 1990; Toribio 1992, 2002), but later research
shows that it is syntactically unrelated to clefts (Bosque 1999; Curnow & Travis 2003;
Camacho 2006). The study presented in this paper maintains this latter claim, according
to which the FS and the pseudo-cleft are different syntactic structures.
Based on data presented in Section€3, I have drawn some important generaliza-
tions regarding the syntactic configuration of the FS: (1) the FS may focus any type of
constituent as long as it is post-verbal; (2) FS ser agrees with the verb in T (a matrix or
an auxiliary verb) in both tense and aspect; (3) FS ser agrees with post-verbal subjects
in both person and number, and with DOs (with no Differential Object Marking) only
in number; (4) the agreement variability found between FS ser and DOs directly cor-
relates with contrastive focus; and (5) FS ser does not establish any kind of agreement
with IOs and with DOs showing Differential Object Marking.
Given these empirical observations, and taking into account certain problems
observed in other syntactic accounts (Bosque 1999; Camacho 2006), I have proposed
that the FS should be examined as a functional projection placed outside vP. Specifi-
cally, following Belletti’s (2004) research according to which both topic and focus pro-
jections may be found within the internal periphery of IP, I have claimed that the FS
is generated inside a Focus Phrase (FocP), placed below T and above vP. The struc-
ture that I propose in this paper is innovative and it correctly predicts what has been
observed in the data.

References

Albor, H.R. 1986. “Uso e interpretación de ser en construcciones galicadas y en Él necesita es

descansar”. Thesaurus XLI: 173–186.
Belletti, A. 2004. “Aspects of the Low IP Area.” In The structure of IP and CP. The Cartography of
Syntactic Structures. Ed. by L. Rizzi. New York: Oxford University Press. 16–51.
Bosque, I. 1999. “On Focus vs. Wh-movement: The Case of Caribbean Spanish.” Sophia Linguistica
44–45: 1–32.
Camacho, J. 2006. “In Situ Focus in Caribbean Spanish: Towards a Unified Account of Focus.”
Proceedings of the 9th Hispanic Linguistics Symposium. Somerville, MA: Cascadilla
Proceedings Project.
Cinque, G. 1999. Adverbs and Functional Heads: A Cross-linguistic Perspective. New York: Oxford
University Press.
 Dunia Catalina Méndez Vallejo

Contreras, H. 1991. “On the Position of Subjects.” Syntax and Semantics: Perspectives on Phrase
Structure 25: 63–79.
Curnow, T. & C. Travis. 2003. “The Emphatic es Construction of Colombian Spanish.” 2003
Conference of the Australian Linguistic Society.
Guitart, J. 1989. “On Spanish Cleft Sentences.” Studies in Romance Linguistics: Selected Papers
from the Seventeenth Linguistic Symposium on Romance Languages XVII: 129–154.
Horvath, J. 1986. Focus in the Theory of Grammar and the Syntax of Hungarian. Dordrecht: Foris.
Hsu, Y.-Y. 2008. “The Sentence-internal Topic and Focus in Chinese.” 20th North American con-
ference on Chinese linguistics (NACCL-20), Columbus, OH: The Ohio State University.
Klein, P. 1968. Modal Auxiliaries in Spanish. Ph.D. dissertation, University of Washington.
Mejías-Bikandi, E. 1992. “The VP-internal Subject Hypothesis and Spanish Sentence Structure.”
Contemporary Research in Romance Linguistics. Papers from the 22nd Linguistic Symposium
on Romance Languages. 123: 275–289.
Méndez Vallejo, D.C. 2009. Focalizing ser (“to be”) in Colombian Spanish. Ph.D. dissertation.
Indiana University.
Mikolajczak, S. 2003. “Os tipos das construções com clivagem em portugués.” Studia Romanica
Posnaniensia: 187–196.
Oliveira, M.A. & M.L. Braga. 1997. “On Focusing Sentences in Brazilian Portuguese.” Towards
a Social Science of Language: Papers in honor of William Labov, ed. by G. Guy, C. Feagin,
D.€Shiffrin & J. Baugh. Amsterdam/Philadelphia: John Benjamins.2: 207–221.
Ordóñez, F. 1997. Word Order and Clause Structure in Spanish and other Romance Languages.
Ph.D. dissertation, CUNY.
Sedano, M. 1988. “Yo vivo es en Caracas: Un cambio sintáctico.” Studies in Caribbean Spanish
Dialectology. Washington, DC: Georgetown University Press. 115–123.
Sedano, M. 1990. Hendidas y otras construcciones con ser en el habla de Caracas. Caracas:
Universidad Central de Venezuela.
Sedano, M. 1994. “Presencia o ausencia de relativo: Explicaciones funcionales.” Thesaurus
49.3: 491–518.
Sedano, M. 1995. “A la que yo escribí fue a María vs. Yo escribí fue a María: El uso de estas dos
estructuras en el español de Caracas.” Boletín de Lingüística 9: 51–80.
Sedano, M. 2003a. “Más sobre seudohendidas y construcciones con verbo ser focalizador en
el habla de Caracas.” Lengua, variación y contexto: Estudios dedicados a Humberto López
Morales. F. Moreno Fernández et al. (eds.), Madrid: Arco Libros. 2: 823–847.
Sedano, M. 2003b. “Seudohendidas y oraciones con verbo ser focalizador en dos corpus del español
hablado de Caracas.” Revista Internacional de Lingüística Iberoamericana 1. 1: 175–204.
Toribio, A.J. 1992. “Proper Government in Spanish Subject Relativization.” Probus 4: 291–304.
Toribio, A.J. 2002. “Focus on Clefts in Dominican Spanish.” Structure, Meaning, and Acquisition of
Spanish, ed. by J. Lee, K. Geeslin & C. Clements, Somerville, MA: Cascadilla Press. 130–146.
Yanagida, Y. 1995. Focus Projection and Wh-head Movement. Ph.D. dissertation, Cornell University.
Zagona, K. 2002. The Syntax of Spanish. Cambridge, Cambridge University Press.
 Anaphoricity, logophoricity
and intensification
The puzzling case of son propre in French*

Isabelle Charnavel
UCLA/IJN-ENS

The goal of this paper is to show the existence of an interaction between binding
and intensification in light of French possessive son propre (“his own”). Propre
(“own”) has two possible interpretations: its semantic effect consists in contrasting
either the possessor (possessor propre) or the possessum (possessum propre) with
a set of contextually determined alternatives. Crucially, these double intensifying
properties of propre correlate with the binding properties of son propre. When son
is associated with possessor propre, son propre behaves like an anaphor that can be
long distance bound if the antecedent is a logophoric center. However, when son
(“his”) is associated with possessum propre, son propre lacks both anaphoric and
logophoric properties. This correlation shows that there is an interaction between
the modules of binding and intensification: it is only when the referent of its
antecedent (i.e. the possessor) is intensified that son propre needs to be bound.

1.â•… Introduction

The goal of this paper is to explore the properties of French possessive son propre (“his
own”) to shed light on the interaction between binding and intensification.
At first glance, son propre exhibits a puzzling behavior with respect to binding:
depending on the cases, it behaves like an anaphor (e.g. 1), a logophor (e.g. 2) or lacks
anaphoric properties altogether (e.g. 3) as shown by the differences of anaphoric
domains in (1)–(3). In (1), son propre has to be locally bound by its antecedent; in€(2),
it can be long distance bound if the antecedent is a center of perspective; but in (3),
son propre exhibits neither of these requirements: the antecedent does not locally
c-command son propre and is not a center of perspective either.

*I would like to thank Dominique Sportiche for very useful advice and discussions.
 Isabelle Charnavel

(1) a. [Cet hôtel]k protège sak (propre) plage sans se préoccuper

this hotel protects its own beach without se care
des plages des hôtels voisins.
of_the beaches of_the hotels neighboring
“This hotel protects its (own) beach without caring about the beaches
of the neighboring hotels.”
b. Les clients de [cet hôtel]k préfèrent sak (*propre) plage à
the guests of this hotel prefer its ╇╛own beach to
celles des hôtels voisins.
the_ones of_the hotels neighboring
“The guests of this hotel prefer its (*own) beach to the beaches of the
neighboring hotels.”

(2) a. [Ce pont]j a bénéficié du fait que les autorités

this bridge has benefited of_the fact that the authorities
ont donné plus d’ avantages à sonj (*propre) architecte
have given more of benefits to its ╇╛own architect
qu’ à celui du musée.
than to the_one of_the museum
“This bridge benefited from the fact that the authorities provided more
benefits to its (*own) architect than to the architect of the museum.”
b. [Le patron de cette entreprise]j a bénéficié du fait que les
the boss of this company has benefited of_the fact that the
autorités ont donné plus d’ avantages à sesj (propres)
authorities have given more of benefits to his own
employés qu’ à ceux de son concurrent.
employees than to the_one of his competitor
“The boss of this company benefited from the fact that the authorities
provided more benefits to his (own) employees than to the employees
of his competitor.”

(3) [Ce pont]i a l’ air très fragile. Soni (propre) architecte

this bridge has the air very fragile its own architect
a demandé un contrôle de sécurité.
has asked a control of security
“This bridge looks very fragile. Its (own) architect asked for a safety check.”

I argue that this apparently mixed behavior of son propre with respect to binding can
be understood if it is correlated with the interpretive properties of propre (“own”).
It will first be shown that propre behaves like an intensifier which can have two dif-
ferent targets for intensification. Then, I will present the binding properties of son
propre and show that they crucially correlate with the double intensifying properties
 On French son propre 

of propre. This demonstrates that the modules of binding and intensification interact
with each other.
This empirical result should have important consequences on linguistic theory,
given that binding and intensification are not supposed to apply at the same level: the
syntactic principles of binding theory deal with the distribution of pronominal and
anaphoric elements (cf. Chomsky 1981; Pollard & Sag 1992; Reinhart & Reuland 1993;
Huan & Liu 2001…etc) while the semantic and pragmatic principles of intensification
deal with the distribution of intensifiers, which belong to focusing devices (cf. König
& Siemund 2000; Eckardt 2001; Bergeton 2004…etc). Therefore, this paper leads to
question the locus and the principles of binding theory and the intensification mod-
ule. However, these crucial theoretical issues cannot be addressed in detail here, since
this would go far beyond the scope of this paper. The aim of this article is mainly to
establish empirical facts: it presents new data and correlations that raise crucial issues
for binding theory and intensification.

2.â•… The interpretive properties of propre: Two cases of intensification

First, I show that propre (“own”)1 behaves like a flexible intensifier specialized in
possessive DPs: its semantic effect consists in contrasting either the possessor2 (pos-
sessor propre) or the possessum (possessum propre) with a set of contextually deter-
mined alternatives.

2.1â•… Possessor propre

Let’s compare the two following sentences:
(4) a. Clairei a pris sai voiture pour aller au travail.
Claire has taken her car for go to_the work
“Claire took her car to go to work.”
b. Clairei a pris sai propre voiture pour aller au travail.
Claire has taken her own car for go to_the work
“Claire took her own car to go to work.”

.â•… The adjective propre has other uses in French (in particular ‘clean’, ‘peculiar to’, ‘liable to’),
but I will concentrate on propre meaning ‘own’ here. This propre is identifiable by its DP-
internal distribution: it only occurs in the prenominal position of a definite possessive DP
(cf. Charnavel 2009b).
.╅ Here, I assume as is standard that the relation of possession denotes a broad range of rela-
tions. The possessor corresponds to the possessing entity and the possessum to the possessed
entity (for references about possessives, see for example Storto 2003).
 Isabelle Charnavel

Both sentences are true in the same situation where Claire has a car and she took this
car to go to work: the presence of propre does not change the truth-conditions of (4b)
as compared to (4a).
However, the two sentences do not have the same felicity conditions: (4b) is felici-
tous only if there is some other referent in the discourse background whose car is or
has been under discussion with respect to its use by Claire to go to work. For example,
(4b) could be felicitous in the following context: Claire usually takes her husband’s car.
Thus, propre requires some other contextually salient referent(s) that play(s) the role of
alternative(s): propre imposes a contrastiveness condition.
More specifically, the alternatives induced by propre in this case target the
possessor;3 that’s why I call this first case possessor propre. Thus in (4b), the referent of
Claire is contrasted with another contextual possessor, i.e. Claire’s husband.
This means that possessor propre has an effect similar to focusing the possessor by
stressing it:
(5) Clairei a pris SAi voiture pour aller au travail.
Claire has taken her car for go to_the work
“Claire took her car to go to work.”

However, as opposed to focus particles which trigger focal stress on their associates,
propre bears focal stress itself; it is in this sense that I call it an intensifier (cf. Eckardt
2001 for further details about intensification, and Charnavel 2009a, 2009b for a semantic
analysis of possessor propre).

2.2â•… Possessum propre

In the first case called possessor propre, the semantic effect of propre consists in con-
trasting the referent of the possessor with a contextually determined set of alternatives.
We observe a second case in which the alternatives target the possessum, as illustrated
by the following example. I call it possessum propre.

.╅ Note that propre can also target the possessor if it is expressed by a prepositional phrase de
X (“of X”), although it is not judged as good as the other case by all native speakers of French.
(i) Donc me voilà débarquant dans un appartement plus grand que le
so me here turning_up in a apartment more big than the
propre appartement de mes parents en France! [attested on google]
own apartment of my parents in France
“And then, I was turning up at an apartment that was bigger than my parents’ own
apartment in France!”
 On French son propre 

(6) a. Arnaudi est devenu si insupportable que sai fille

Arnaud is become so unbearable that his daughter
a cessé de lui rendre visite.
has stopped of him visit
“Arnaud has become so unbearable that his daughter stopped visiting him.”
b. Arnaudi est devenu si insupportable que sai propre fille
Arnaud is become so unbearable that his own daughter
a cessé de lui rendre visite.
has stopped of him visit
“Arnaud has become so unbearable that his own daughter stopped
visiting him.”

As in the case of possessor propre, both sentences are true in the same situation, but
they have different felicity conditions: alternatives come into play in (6b).
However, it is not the referent of the possessor that is targeted in this sentence:
Arnaud – the possessor – is not contrasted with other fathers. Rather, it is Arnaud’s
daughter – thus the possessum – that is contrasted with other individuals. For example,
(6b) would be felicitous in the following context: Arnaud’s friend and Arnaud’s cousin
have already stopped visiting Arnaud because he is too bad-tempered. Thus, propre
targets the possessum in this case4 since it is the referent of the whole possessive DP
sa fille (“his daughter”) that is contrasted with other individuals.5 Furthermore, as
opposed to possessor propre, possessum propre requires an ordering of the alternatives

.╅ As in the previous case, the possessum can also be targeted when the possessor is expressed
by a prepositional phrase de X (“of X”): here, the referent of the victim’s son is contrasted with
other individuals:
(ii) Le meurtrier présumé qui a été placé en hôpital psychiatrique
the murderer presumed who has been placed in hospital psychiatric
n’ est autre que le propre fils de la victime. [attested on google]
ne is other than the own son of the victim
“The presumed murderer who has been placed in a psychiatric hospital is no other
than the victim’s own son.”
.╅ The example (6b) could suggest that it is not the possessum individual, but rather the rela-
tion (‘daughter’) that is contrasted with other relations (‘friend’ or ‘cousin’ in the context). But
this is incorrect: it is not necessary that the alternatives be related to the possessor as shown
by the following example. In (iii), at least one of the salient alternatives – the witness – does
not stand in a specific relationship to John. Therefore, the relation of motherhood cannot be
contrasted with other relations; it is rather the individual referent of John’s mother that is
contrasted with other individuals.
 Isabelle Charnavel

on a scale of likelihood: the individual intensified by propre corresponds to an unlikely

one in the context: in (6b), Arnaud’s daughter is less likely than his cousin or his friend
to stop visiting him.
This means that in this case, propre has an effect comparable to focusing the pos-
sessum by stressing it as shown in (7). But as in the first case, propre is an intensifier in
that it bears focal stress itself.6

(7) Arnaudi est devenu si insupportable que sai FILle

Arnaud is become so unbearable that his DAUGHter
a cessé de lui rendre visite.
has stopped of him visit
“Arnaud has become so unbearable that his daughter stopped visiting him.”

To sum up this section, propre has two possible interpretations: it can contrast either
the possessor (possessor propre) or the possessum (possessum propre) with a contextu-
ally determined set of alternatives. That’s in this sense that propre can be considered as
a flexible intensifier specialized in possessive DPs.7

3.â•… The binding properties of son propre

In this section, I argue that the intensifying properties of propre correlate with the
binding properties of son propre: it is only in the case of possessor propre – as opposed
to possessum propre – that son propre exhibits anaphoric properties.

(iii) Ce n’ est pas la victime qui a dénoncé Jeani, ni un témoin,

it ne is not the victim who has denounced John nor a witness
c’est sai propre mère qui l’ a dénoncé!
it is his own mother who him has denounced
“It’s not the victim who denounced John, nor a witness, it’s his own mother who
denounced him!”
.╅ This is at least the case in my dialect of French. But in the case of possessum propre, some
French speakers prefer to put stress on the possessum itself (as it seems to be the case in
English and German).
.╅ For space reasons, I cannot provide a precise analysis of propre as a flexible intensifier in
this paper (see Charnavel 2009a, 2009b for a semantic analysis of propre). But it is sufficient
for my purposes here (i.e. to show the correlation between binding and intensification) to
pinpoint the double intensifying properties of propre.
 On French son propre 

3.1â•… Possessor son propre: Anaphoric and logophoric properties

First, I show that son propre exhibits anaphoric or/and logophoric properties when
son is associated with possessor propre (possessor son propre, henceforth): when it is the
possessor that is intensified by propre, son propre behaves like an anaphor that can be
long distance bound if the antecedent is a logophoric center.

3.1.1â•… First case: Anaphoric son propre

When the referent of the possessor, i.e. the antecedent, is inanimate, possessor son pro-
pre has anaphoric properties, unlike the pronoun son. As stated by principle A of bind-
ing theory, this means that son propre needs to be locally bound, i.e. it requires a locally
c-commanding and coindexed antecedent.

(8) Principle A of Binding Theory (cf. Chomsky 1981, 1986 and subsequent
revisions of it): an anaphor must be bound in its domain.

The following sentences, which involve possessor propre, illustrate the c-command
requirement.

(9) [=1] a. [Cet hôtel]k protège sak (propre) plage sans se

this hotel protects its own beach without se
préoccuper des plages des hôtels voisins.
care of_the beaches of_the hotels neighboring
“This hotel protects its (own) beach without caring about the beaches
of the neighboring hotels.”
b. Les clients de [cet hôtel]k préfèrent sak (*propre) plage à
the guests of this hotel prefer its own beach to
celles des hôtels voisins.
the_ones of_the hotels neighboring
“The guests of this hotel prefer its (*own) beach to the beaches of the
neighboring hotels.”

In (9a), both sa propre plage (“its own beach”) and sa plage (“its beach”) license cet
hôtel (“this hotel”) as antecedent. However, in (9b), cet hôtel (“this hotel”) is only a
possible antecedent for sa plage (“its beach”), not for sa propre plage (“its own beach”).
Since the crucial difference between the two sentences is that cet hôtel (“this hotel”)
does not c-command sa (propre) plage (“its (own) beach”) in (9b), but does in (9a),
this means that sa propre plage as opposed to sa plage needs to be c-commanded by
its antecedent.
 Isabelle Charnavel

Moreover, the binder must be local, as exemplified by the following sentence.

(10) [=2a] [Ce pont]j a bénéficié du fait que les autorités
this bridge has benefited of_the fact that the authorities
ont donné plus d’ avantages à sonj (*propre) architecte
have given more of benefits to its ╇ own architect
qu’ à celui du musée.
than to the_one of_the museum
“This bridge benefited from the fact that the authorities provided more
benefits to its (*own) architect than to the architect of the museum.”

In (10), son architecte (“its architect”) licenses the long-distance antecedent ce pont
(“this bridge”), but son propre architecte (“its own architect”) does not.
Therefore, the following generalization holds:
(11) In the case of inanimate possessors, possessor son propre is a complex possessive
anaphor obeying principle A of Binding Theory (as formulated by Chomsky
1981, 1986 and subsequent revisions).

3.1.2â•… Second case: Logophoric son propre

However, the generalization (11) does not hold for animate possessors, as illustrated
by the following contrast:
(12) [=2] a. [Ce pont]j a bénéficié du fait que les autorités
this bridge has benefited of_the fact that the authorities
ont donné plus d’ avantages à sonj (*propre) architecte
have given more of benefits to its ╇ own architect
qu’ à celui du musée.
than to the_one of_the museum
“This bridge benefited from the fact that the authorities provided more
benefits to its (*own) architect than to the architect of the museum.”
b. [Le patron de cette entreprise]j a bénéficié du fait que
the boss of this company has benefited of_the fact that
les autorités ont donné plus d’ avantages à sesj (propres)
the authorities have given more of benefits to his own
employés qu’ à ceux de son concurrent.
employees than to the_one of his competitor
“The boss of this company benefited from the fact that the authorities
provided more benefits to his (own) employees than to the employees
of his competitor.”

(12b) shows that ses propres employés (“his own employees”) licenses a long-distance
antecedent le patron de cette entreprise (“the boss of this company”) as opposed to
 On French son propre 

son propre architecte (“its own architect”) in (12a). Since the crucial difference is that
the possessor is animate in (12b), this means that son propre does not require a local
binder when the possessor antecedent is animate.8
Similarly, it is not always true that son propre must be c-commanded by its ante-
cedent in the case of animate possessors:
(13) a. L’ opinion de Sébastienj portait autant sur saj (propre)
the opinion of Sébastien was_about as_much on his own
mère que sur la mère de sa femme.
mother than on the mother of his wife
“Sébastien’s opinion was as much about his (own) mother than about his
wife’s mother.”
b. Le sujet de [l’ article]j portait autant sur sonj (*propre)
the topic of the article was_about as_much on its ╇╛own
titre que sur le titre du film en question.
title than on the title of_the movie in question
“The topic of the article was as much about its (*own) title than about the
title of the movie in question.”

(13a) contrasts in this respect with (13b) since sa propre mère (“his own mother”)
licenses the animate non c-commanding antecedent Sébastien in (13a), while son pro-
pre titre (“its own title”) cannot have the inanimate non c-commanding l’article (“the
article”) as antecedent in (13b). So, in the case of animate antecedents, son propre does
not require a locally c-commanding antecedent.
Therefore, son propre seems to fall into the class of long-distance anaphors such as
Mandarin Chinese ziji (cf. Huang & Liu 2001), which pose a challenge to the standard
theory of anaphor binding. The hypothesis that has been proposed in such cases is
the theory of logophoricity (cf. Huang & Liu 2001; Giorgi 2007…etc): long-distance
anaphors are logophoric, i.e. they do not have to obey the syntactic constraints of bind-
ing, but the constraints of logophoricity requiring that the antecedent be a center of
perspective of the sentence. This idea is based on the fact that some West African
languages have specific pronouns used to express the perspective of the person they
refer to. The term logophor has been originally coined for such cases (cf. Hagège 1974)
and has then been extended to situations in other languages where the usual rules of
binding do not apply, that is in the case of long distance anaphors, which have their
antecedents outside their binding domains (e.g. Mandarin Chinese ziji).

.╅ It would be worth defining the notion of locality and the anaphoric domain in detail; but
since I do not have space to investigate all the relevant examples here, I simply assume for now
that the anaphoric domain is the clause; this approximation is sufficient for my purposes here.
 Isabelle Charnavel

I propose that possessor son propre supports this hypothesis: possessor son propre
can be long distance bound if it is logophoric. This means that in such cases, son pro-
pre refers to a specific type of antecedent, namely a logophoric center: the antecedent
refers to a person whose words, thoughts or point of view are being reported.9 More
specifically, I argue that son propre belongs to the class of logophors that require a de
se reading.10
The distinction between de re and de se readings corresponds to the distinction
between the report of the knowledge of the speaker and that of the knowledge of the
referent of the antecedent (cf. Chierchia 1989). This means that the antecedent of son
propre corresponds to a logophoric center if and only if its referent is aware of the
reflexivity of the possession, i.e. if and only if its referent could knowingly say mon
propre (“my own”).
Thus, I propose that the de se reading is the primitive property defining son propre
as a logophor. This property is therefore sufficient as a diagnostic for logophoricity.
However, for methodological reasons, I will also use two other properties that derive
from this one to identify logophoric son propre, because they are clearer diagnostics,
i.e. animacy and consciousness of the referent of the antecedent. De se reading entails
consciousness of the referent of the antecedent since it is necessary to be conscious to
be able to knowingly say mon propre. Moreover, consciousness entails animacy, and
therefore, by transitivity, animacy of the referent of the antecedent is also entailed by
the de se reading. That’s why following Huang & Liu (2001), I will use the following
three criteria as diagnostics for the logophoricity of possessor son propre: (a) animacy
of the referent of the antecedent; (b) consciousness of the referent of the antecedent;
(c)€de se reading.

a. Animacy of the referent of the antecedent.

As already suggested in the pair (2) repeated here as (14), the referent of the ante-
cedent has to be animate to license logophoric son propre. Put another way, possessor
son propre does not require a locally c-commanding antecedent if the referent of the
antecedent is a center of perspective, and this is possible only if it is animate.

.╅ Sells (1987) proposes three primitive roles for the antecedent of logophors and he sug-
gests that these roles characterize certain cross-linguistic variations:
a. Source: the one who is the intentional agent of the communication,
b. Self: the one whose mental state or attitude the proposition describes,
c. Pivot: the one with respect to whose (time-space) location the content of the proposition
is evaluated.
.â•… Mandarin Chinese ziji in Huang & Liu’s dialect (2001:19) or Italian proprio (cf. Giorgi
2007:€333) also belong to this class of logophors.
 On French son propre 

(14) [=2] a. [Ce pont]j a bénéficié du fait que les autorités

this bridge has benefited of_the fact that the authorities
ont donné plus d’ avantages à sonj (*propre) architecte
have given more of benefits to its ╇ own architect
qu’ à celui du musée.
than to the_one of_the museum
“This bridge benefited from the fact that the authorities provided more
benefits to its (*own) architect than to the architect of the museum.”
b. [Le patron de cette entreprise]j a bénéficié du fait que
the boss of this company has benefited of_the fact that
les autorités ont donné plus d’ avantages à sesj (propres)
the authorities have given more of benefits to his own
employés qu’ à ceux de son concurrent.
employees than to the_one of his competitor
“The boss of this company benefited from the fact that the authorities
provided more benefits to his (own) employees than to the employees
of his competitor.”

Ses propres employés (“his own employees”) in (14b) licenses a long distance antecedent
le patron de l’entreprise (“the boss of the company”), but the long distance antecedent ce
pont (“this bridge”) in (14a) for son propre architecte (“its own architect”) is ungram-
matical. This is so because “the boss of the company” can be a perspective-holder in
(14b) as opposed to “this bridge” in (14b). This difference can be easily diagnosed by
the animacy of the referent of le patron de l’entreprise vs ce pont.

b. Consciousness of the referent of the antecedent.

Similarly, the center of perspective of a sentence has to be conscious; therefore, if the

referent of the antecedent is not conscious, logophoric son propre is not possible, as
shown by the following contrast:
(15) a. [Le pharaon]i a beaucoup aimé les embaumeurs qui à
the Pharaoh has a_lot liked the embalmers who at
présent prennent soin de soni (*propre) corps.
present take care of his ╇ own body
“The Pharaoh had liked a lot the embalmers who are now taking care
of his (*own) body.”
b. [L’ esprit du pharaon]i devait penser que les embaumeurs
the spirit of_the Pharaoh must think that the embalmers
prenaient bien soin de soni (propre) corps.
took well care of his own body
“The Pharaoh’s spirit was probably thinking that the embalmers were
taking great care of his (own) body.”
 Isabelle Charnavel

In (15a), the Pharaoh is dead, therefore not conscious, and this diagnostic shows that
the Pharaoh cannot be the center of perspective of the sentence. Thus, son propre corps
(“his own body”), which is not locally c-commanded by le pharaon (“the Pharaoh”),
is not possible, as predicted by the logophoricity hypothesis. However in (15b), son
propre corps (“his own body”) can be long distance bound by l’esprit du pharaon (“the
Pharaoh’s spirit”) because the referent of this antecedent is conscious, thus a possible
center of perspective.

c. De se reading.

The de se reading is the strictest criterion to define the logophoric center in the case
of possessor son propre. The context of Beaumarchais’s Marriage of Figaro can exem-
plify this property: in this setting, the maid Marceline knows that Suzanne will marry
Figaro, but she does not know until the end of the play that Figaro is her son. In this
context, the following contrast holds:
(16) a. Marcelinei disait que Suzanne allait épouser soni (#propre) fils.
Marceline said that Suzanne was_going_to marry her ╇ own son
“Marceline said that Suzanne would marry her (#own) son.”
b. Marcelinei disait que Suzanne avait épousé soni (propre) fils.
Marceline said that Suzanne had married her own son
“Marceline said that Suzanne had married her (own) son.”
If (16a) is uttered at the beginning of the play, the de se reading is not available since
Marceline does not know yet about her motherhood. Therefore, as predicted by the
logophoricity hypothesis, she cannot be the center of perspective and Marceline can-
not long-distance bind son propre fils (“her own son”): son propre cannot be logophoric
in this case. However, if (16b) is uttered at the end of the play, the sentence is appropri-
ate because Marceline knows at that time that Figaro is her son; thus, Marceline is the
center of perspective according to the criterion that I propose, which licenses the long
distance anaphor son propre fils (“her own son”). This contrast demonstrates that the
de se diagnostic appears to be the most relevant one to define the notion of logophoric
center in the case of possessor son propre. Conversely, this means that if the de se read-
ing is not available, possessor son propre cannot be logophoric and has therefore to be
an anaphor requiring a locally c-commanding antecedent.
To sum up, the following generalization holds for possessor son propre:
(17) Possessor son propre is either an anaphor obeying the syntactic constraints
of anaphoricity (local c-commanding antecedent) or a logophor obeying
the discourse-related constraints of logophoricity (antecedent as
perspective holder).11

.╅ Note that the sets of anaphoric and logophoric uses of son propre are not in comple-
mentary distribution, but overlap since their properties are not exclusive of each other. Thus,
 On French son propre 

3.2â•… Possessum son propre: No anaphoric properties

While possessor son propre exhibits anaphoric properties, I show in this section that
possessum son propre does not. This argues for the presence of an interaction between
binding and intensification: when the possessor, i.e. the referent of the antecedent of
son propre, is intensified, anaphoric properties arise, but it is not the case when it is the
possessum that is intensified.
As illustrated by the following examples, possessum son propre lacks both ana-
phoric and logophoric properties:

(18) [=3] a. [Ce pont]i a l’ air très fragile. Soni (propre)

this bridge has the air very fragile its own
architecte a demandé un contrôle de sécurité.
architect has asked a control of security
“This bridge looks very fragile. Its (own) architect asked for a
safety check.”
b. [Ce pont]i a l’ air très fragile. Soni (*propre) architecte
this bridge has the air very fragile its ╇╛own architect
a reçu moins de moyens que tous les autres architectes
has received less of means than all the other architects
des ponts de la région.
of_the bridges of the area
“This bridge looks very fragile. Its (*own) architect got less means than
all the other architects of the bridges of the area.”
c. [Cet enfant]i a l’ air très perturbé. Sai (propre) mère
this child has the air very disturbed his own mother
passe moins de temps à la maison que toutes les autres
spends less of time at the house than all the other
mères des enfants de la classe. (m’ a-t-il dit.)
mothers of_the children of the class me has he said
“This child looks very disturbed. His (own) mother spends less time at
home than all the other mothers of the children in the class. (he said)”

In (18b), son propre architecte (“its own architect”) is a case of possessor propre: this
bridge is contrasted with other bridges as “possessors” of an architect. In (18a) however,
propre intensifies the possessum: the bridge’s architect is opposed to other individu-

possessor son propre can be both anaphoric and logophoric if its antecedent both locally
c-commands it and is the center of perspective (de se reading).
 Isabelle Charnavel

als who would ask for a safety check too, and he is an unlikely individual among
the alternatives to express such a request since he designed the bridge himself. Cru-
cially, this difference in intensification correlates with a difference in binding: (18b) is
ungrammatical if it includes propre because son propre architecte (“its own architect”)
is an anaphor requiring a local antecedent, but (18a) is grammatical because son propre
architecte (“its own architect”) does not exhibit binding properties. In other terms, this
contrast shows that possessum propre does not require a local antecedent and therefore
argues for the non anaphoric status of possessum son propre.
Moreover, the same example shows that possessum son propre also lacks logo-
phoric properties. Recall that possessor son propre may be long distance bound if the
antecedent is a logophoric center and we established that a logophoric center has to
be animate. That’s why (18b), which presents the inanimate ce pont (“this bridge”) as
antecedent of possessor son propre, is ungrammatical with propre, while (18c), in which
possessor son propre has the animate cet enfant (“this child”) as antecedent, is gram-
matical: it is because a child, unlike a bridge, can be a center of perspective that (18c),
unlike (18b), is well-formed; this is further suggested by the fact that the parenthesis
in (18c), which explicitly makes the child the source of the sentence, improves the
grammaticality of the sentence. However, the sentence (18a), in which propre does
not intensify the possessor, but the possessum, is crucially grammatical, even if son
propre architecte (“its own architect”) has the inanimate ce pont (“this bridge”) as long
distance antecedent. This demonstrates that possessum son propre, unlike possessor son
propre, lacks anaphoric and logophoric properties altogether.
So as opposed to possessor son propre, possessum son propre does not obey any
binding constraints:12 its antecedent does not have to c-command it, nor to be local.
Moreover, it does not have to be non c-commanding or non local either, as shown by
the following example:

.╅ Note that the fact that possessum son propre obeys neither the constraints of anapho-
ricity nor the constraints of logophoricity does not mean that possessum son propre is not
constrained at all. In particular, as shown in (iv), it seems that possessum son propre requires
that the closest c-commanding DP be its antecedent when there is one available. But the
important point for our purposes is that possessum son propre does not exhibit any anaphoric
or logophoric properties as opposed to possessor son propre.
(iv) [Ce pont]i a donné lieu à la rumeur que [le maire]j se
this bridge has given rise to the rumor that the mayor se
méfie de sonj/*i propre architecte.
distrusts of its/his own architect
“This bridge has given rise to the rumor that the mayor distrusts his/*its
own architect.”
 On French son propre 

(19) Dans un moment de folie, après avoir tué les voisins,

in a moment of madness after have killed the neighbors
Micheli a tué sesi propres enfants.
Michel has killed his own children
“In a moment of madness, after he killed the neighbors, Michel killed his
own children.”

To sum up, possessor son propre obeys the constraints of anaphoricity or/and the con-
straints of logophoricity while possessum son propre does not. As shown in the second
section, possessor son propre intensifies the possessor, i.e. the referent of the antecedent
of son propre, while possessum son propre intensifies the possessum. Crucially, this
correlation therefore shows that there is an interaction between the modules of binding
and intensification:13 it is only when the referent of its antecedent is intensified that
son propre needs to be bound.

4.â•… Conclusion

At least three empirical results come out of the study of the properties of son pro-
pre. First, it documents the existence of an anaphor in French (where otherwise, ana-
phoric properties are mainly encoded by the reflexive clitic se). Moreover, it reveals
the �existence of a logophor in French, which supports the theory of logophoricity.
Thirdly – and this is the main result of this paper, – it sheds light on the existence of
an interaction between binding and intensification. Further investigation should show
how the connection works and what the consequences of this finding are for the archi-
tecture of the grammar.

References

Bergeton, Uffe. 2004. The Independence of Binding and Intensification. Ph.D. dissertation,
University of Southern California.
Charnavel, Isabelle. 2009a. “On the Intensifier propre (“own”) in French: a Counterpart of même
(“self ”) in Possessive DPs?”, to appear in Proceedings of SALT 19.
Charnavel, Isabelle. 2009b. Linking Binding and Focus: on Intensifying son propre in French.
Master Thesis, UCLA.

.â•… Some typological studies (cf. König & Siemund 2000) suggest the existence of a link
between binding and intensification, but very few theoretical studies do; moreover, when they
do, they argue for the independence between binding and intensification (cf. Bergeton 2004).
So to my knowledge, the result provided in this paper is new.
 Isabelle Charnavel

Chierchia, Gennaro. 1989. “Anaphora and Attitudes De Se”. Semantics and Contextual Expression,
ed. by R. Bartsch, J. van Benthem & P. van Emde Boas, 1–31, Dordrecht: Foris.
Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, Noam. 1986. Knowledge of Language: Its Nature, Origin, and Use. New York: Praeger.
Eckardt, Regine. 2001. “Reanalyzing selbst”. Natural Language Semantics 9: 4. 371–412.
Giorgi, Alessandra. 2007. “On the Nature of Long-Distance Anaphors”. Linguistic Inquiry
38:€2.€321–342.
Hagège, Claude. 1974. “Les Pronoms Logophoriques”. Bulletin de la Société Linguistique de Paris
69.287–310.
Huang, C.-T. James & C.-S. Luther Liu. 2001. “Logophoricity, Attitudes and ziji at the Inter-
face”. Long Distance Reflexives, ed. by Peter Cole et al., Syntax and Semantics 33, 141–195.
New York: Academic Press.
König, Ekkchard & Peter Siemund. 2000. “Intensifiers and Reflexives: A Typological Perspective”.
Reflexives: Forms and Functions, ed. by Zygmunt Frajzyngier & Traci Curl, 41–74.
Amsterdam: Benjamins.
Pollard, Carl & Ivan A. Sag. 1992. “Anaphors and the Scope of Binding Theory”. Linguistic
Inquiry 23. 261–303.
Reinhart, Tanya & Eric Reuland. 1993. “Reflexivity”. Linguistic Inquiry 24: 4. 657–720.
Sells, Peter. 1987. “Aspects of Logophoricity”. Linguistic Inquiry 18. 445–479.
Storto, Gianluca. 2003. Possessives in Context: Issues in the Semantics of Possessive Constructions.
Ph.D. dissertation, University of California, Los Angeles.
 More on the clitic combination puzzle
Evidence from Spanish, Catalan and Romanian

Anahí Alba de la Fuente

University of Ottawa

Languages like Spanish, Catalan and Romanian accept combinations of 1st and
2nd person clitics. However, not all 1st and 2nd person combinations are possible.
Although two single clitics can be combined, the combination of two plural clitics
results in ungrammaticality (Rivero 2008; Nevins & Săvescu 2008). This is true for
Spanish and Romanian, but only partially for Catalan. On the other hand, all three
languages accept clitic combinations that include a plural dative and a singular non
dative and systematically reject combinations that include a singular dative and a
plural non-dative. In view of this, I argue that there is a number-case restriction
that parallels the Person Case Constraint, first proposed by Bonet (1991), and,
in more general terms, I defend that clitic restrictions are drawn by the degree of
markedness of the dative clitic with respect to the other clitic in the cluster.

1.â•… Introduction

Clitics and clitic combinations have been subject to a great deal of research. However,
the restrictions that arise when clitics are combined constitute a puzzle that continues
to intrigue and challenge linguists. Many explanations have been proposed in order to
account for the combinatorial possibilities of clitics, including linear ordering templates
(Perlmutter 1971), syntactic accounts (Anagnostopoulou 2003, 2005; Ormazabal€ &
Romero 2007; Adger & Harbour 2007), morphologic accounts (Rivero 2008; Heap
1998, 2005), Optimality Theory accounts (Grimshaw 1997, 2001), etc. This paper aims
to contribute to the ongoing discussion with novel data from Spanish, Catalan and
Romanian, and a new analysis of the effect of marked features on clitic combinations.
More specifically, this paper deals with combinations of 1st and 2nd person clitics
and the effect of the [Plural] feature in such combinations, which has recently begun to
receive more attention by researchers (Anagnostopoulou 2005; Rivero 2008; Nevins &
Săvescu 2008).
The paper is structured as follows: Section€1 presents the Person Case Constraint,
as first proposed by Bonet (1991), and proposals by Anagnostopoulou (2005) and
 Anahí Alba de la Fuente

Rivero (2008), with respect to 1st/2nd person clitic combinations. Section 2 provides
novel data from Spanish, which show that there is an asymmetry with respect to the
combination of singular and plural clitics. In addition, two possible accounts for this
asymmetry are presented in this section: Heap’s (1998, 2005) Least Leafy to the Left
constraint, and Alba de la Fuente’s (2007) Plural-Blocking Effect.
Section 3 includes a discussion of novel data from Catalan, Romanian and Italian,
as well as the implications for the proposals presented in Section€3.
Section 4 presents a revised version of the PBE, which implies that two aspects –
namely, case and feature markedness – are responsible for the combinatorial patterns
of object clitics in Spanish, Catalan and Romanian.
Finally, Section€5 provides the general conclusions of the paper.

2.â•… The effect of person and number in clitic combinations

The Person Case Constraint, proposed by Bonet (1991), divides languages into two
types: Strong and Weak PCC languages. The Strong version of the constraint, which
refers to languages like French, is reproduced in (1) and the Weak version, which refers
to languages like Spanish, is reproduced in (2).
(1) Strong PCC: “In a combination of a weak direct object and an indirect object
[clitic, agreement marker or weak pronoun], the direct object has to be third
person.”
(2) W
 eak PCC: “In a combination of a weak direct object and an indirect object
[clitic, agreement marker or weak pronoun], if there is a third person it has to
be the direct object.”

Thus, both Strong and Weak PCC languages allow for 3rd person clitic combinations
and reject combinations in which the indirect object is 3rd person and the direct
object is 1st or 2nd, as we can see in (3) and (4), respectively. On the other hand, Weak
PCC languages accept combinations of 1st/2nd person clitics, whereas Strong PCC
languages reject them, as shown in (5).
(3) a. Je vais le lui donner.
I go 3.sg.acc 3.sg.dat give
“I’m going to give it to him”
b. Yo se lo voy a dar
I 3.sg.acc 3.sg.dat go to give
“I’m going to give it to him”
(4) a. *Il me lui presentera à la fête.
He 1.sg.acc 3.sg.dat introduce-fut at the party
“He will introduce me to him”
 More on the clitic combination puzzle 

b. *Él me le presentará en la fiesta.1

He 1.sg.acc 3.sg.dat introduce-fut in the party
“He will introduce me to him at the party”
(5) a. Tú te me presentaste en la fiesta.2
You 2.sg.acc 1.sg.dat introduced at the party
“You introduced yourself to me at the party”
b. *Tu te m’est presenté à la fête.
You 2.sg.acc 1.sg.dat-pst introduced at the party
“You introduced yourself to me at the party”

Anagnostopoulou (2005) claims that 1st/2nd person clitic combinations are possible
because Weak PCC languages allow for Multiple Agree. Anagnostopoulou (2003, 2005)
proposes a competition analysis for these contexts, in which both the DO and the IO
try to check features (person and number) against the same head, which is transitive v.
In these cases, the high IO moves to transitive v first and checks features against it. In
Strong PCC languages, the derivation crashes when both the IO and DO are 1st or 2nd
person, since we have two elements trying to check features against one head. Weak
PCC languages, on the other hand, allow for Multiple Agree, which implies that both
the DO and the IO can check features simultaneously against transitive v and, there-
fore, the combination of 1st/2nd person clitics becomes possible. This explains the
grammaticality of the Spanish sentence in (5a) (Weak PCC) and the ungrammaticality
of the French sentence in (5b) (Strong PCC).
However, not all combinations are possible and Anagnostopoulou (2005) argues
that there is a condition for Multiple Agree to take place:

(6) A Condition on Multiple Agree: “Multiple Agree can take place only under
non- conflicting feature specifications of the agreeing elements.”

Independent evidence for this condition comes from Icelandic long distance agreement
in expletive constructions: the dative and the nominative elements can only agree
when they are both plural and when they are both singular. It is impossible for the

.â•… Sentences like the one in (4b) are considered to be grammatical in leísta dialects of
Spanish. In Standard Spanish, le is an IO clitic. However, in these cases, le is not functioning as
an IO, but as a DO (which in Standard Spanish would be the form lo). Thus, the acceptance of
these sentences does not pose problems to the PCC since, functionally, the 3rd person clitic is
still the DO, as required by the Weak PCC.
.╅ The clitic te in (5a) can be both analyzed as accusative and reflexive. In fact, as Nicol
(2005) indicates, the reflexive reading seems to be the favoured one among Spanish speakers.
In line with this preference, this type of clitic will be labelled refl – rather than acc – in the
remainder of the paper.
 Anahí Alba de la Fuente

nominative and dative to enter into a Multiple Agree relation when the nominative is
[+plural] and the dative is [–plural].3,4
With respect to Spanish, Rivero (2008) notes, in her discussion about Quirky Person
Restrictions (QPR), that there is an asymmetry between singular and plural clitic clus-
ters, which is not accounted for by the PCC. According to the Weak version of the
constraint, the two sentences in (6) should be acceptable. However, the sentence with
two singular clitics (6a) is correct and the sentence whose clitics are both plural is not
possible in Spanish (6b).
(7) a. Te me antojaste tú
2.sg.refl 1.pl.dat fancied you
“I took a fancy on you”
b. *Os nos antojasteis vosotros
2.sg.refl 1.pl.dat fancied you-pl
“We took a fancy on you (pl)”

This contradicts Anagnostopoulou’s Condition on Multiple Agree, according to which

(7b) should also be grammatical, since there are no conflicting feature specifications
(both clitics are [+Person] and [+Plural]).
Rivero (2008) defends that 1st/2nd person clitics are marked with the [+Partici-
pant] feature and, additionally, 1st plural and 2nd plural clitics also have a [+Plural]
feature. Thus, 1st plural and 2nd plural are the most marked clitics in terms of features,
which prevents them from combining.
However, examples (8a) and (8b) show that there is yet another asymmetry with
respect to 1st and 2nd person clitic combinations and that the ungrammaticalities
found in these examples cannot be due to markedness alone since the clusters in both
(8a) and (8b) carry the same amount of marked features – namely, two [+Participant]
and one [+Plural] – and, yet, (8a) is grammatical and (8b) is ungrammatical.
(8) a. Te nos presentaste en la fiesta.
2.sg.refl 1.pl.dat introduced at the party
“You introduced yourself to me at the party”
b. *Os me presentasteis en la fiesta.
2.pl.refl 1.sg.dat introduced-pl at the party
“You introduced yourself to me at the party”

.╅ Anagnostopoulou (2005: 223) argues that the opposite case cannot be tested, since these
verbs surface as 3rd person singular when they do not enter long-distance Agree with the
nominative.
.â•… Anagnostopoulou (2005) refers to Holmberg and Hróarsdottir (2002) for further discussion
on this matter.
 More on the clitic combination puzzle 

3.â•… A Singular-Plural asymmetry in Spanish

3.1â•… Th
 e Plural-Blocking Effect (preliminary version)
(Alba de la Fuente 2007)
As it is the case in (8a) and (8b), the clusters in (9b) and (9d) both contain one singular clitic
and one plural clitic. However, one sentence is grammatical and the other one is ungram-
matical. This indicates that the asymmetry cannot be caused by a mere accumulation
of marked features in the cluster.
(9) a. ¡Huy, que te me ensucias!
Oops that 2.sg.refl 1.sg.dat stain
“You are staining yourself ” (and I am affected by it)
b. ¡Huy, que te nos ensucias!
Oops that 2.sg.refl 1.pl.dat stain
“You are staining yourself ” (and we are affected by it)
c. *¡Huy, que os nos ensuciáis!
Oops that 2.pl.refl 1.pl.dat stain
“You are staining yourselves” (and we are affected by it)
d. *¡Huy, que os me ensuciáis!
Oops that 2.pl.refl 1.sg.dat stain
“You are staining yourselves” (and I am affected by it)
Moreover, the same restrictions surface with different structures, such as accusative-
dative and reflexive-dative, as in (8),5 and reflexive-ethical dative,6 as in (9).
Crucially, in all the ungrammatical examples the first clitic in the cluster is plural.
In fact, the grammaticality status of sentences like the ones in (9) seems to depend on
the position of the plural clitic in the cluster.7 In view of this, I proposed the Plural-
Blocking Effect (preliminary version) (Alba de la Fuente 2007):
(10) In a combination of 1st and 2nd person clitics, the first clitic cannot be plural.

.╅ As indicated in Footnote€2, the clitics te (5a, 8a) and os (8b) can be both analyzed as
accusative and reflexive.
.╅ Ormazabal and Romero (2007) argue that ethical datives are not affected by the PCC, since
they are non-argumental. However, we can see from the examples in (9) that they are both
subject to the PCC and to a restriction regarding the combination of singular and plural clitics.
.╅ Reversing the order of the clitics in the cluster would not improve the ungrammatical
sentences, since this would violate the ordering template of Spanish clitics. As defended
by Perlmutter (1971), clitics in Spanish are ordered by person and not by case. Specifically,
Spanish clitics must be arranged according to the following template: se > 2 > 1 > 3. Thus, if the
order of the clitics is reversed in (9c) and (9d), the sentences will still be ungrammatical.
 Anahí Alba de la Fuente

Further evidence for this proposal comes from feature impoverishment and the case
of “spurious se” (Perlmutter 1971). As shown in (11), the spurious se appears in com-
binations of 3rd person clitics and it implies the replacement of the dative clitic le or
les by the form se.8
(11) a. *Los libros les los dimos a mis padres
The books 3.pl.dat 3.pl.acc gave to my parents
b. Los libros se los dimos a mis padres
The books se 3.pl.acc gave to my parents
“The books, we gave to my parents”

Se is underspecified for person and, therefore, it creates no conflicts with other adjacent
clitics, since it has poor feature content (Jackubowicz et al. 1998; Rivero 2008). In fact,
se is the most underspecified pronoun of the entire inventory in Spanish, since it is
underspecified for both number and case.
Moreover, along the lines of feature impoverishment, we have a similar situation
with the use of usted in Spanish. In (12) we have two sentences, each containing a clitic
cluster whose first clitic refers to a 2nd person plural entity. In contrast with (9c) and
(9d), the two sentences in (12) are grammatical.
(12) a. Pasen ustedes, que se nos van a enfermar si
Pass all that se 1.pl.dat go to get_sick if
siguen en la calle.
continue on the street
“Come in, you (pl.,formal) are going to get sick (on us) if you keep waiting
out in the street”
b. Pasen ustedes, que se me van a enfermar si
Pass all that se 1.sg.dat go to get_sick if
siguen en la calle.
continue on the street
“Come in, you (pl.,formal) are going to get sick (on me) if you keep waiting
out in the street”
This is possible because, even though the first clitic is equivalent to a 2nd person plu-
ral, the actual clitic form used in the sentence is underspecified for number. As in the
case of (11b), the selected form is se. In (13), on the contrary, we have the same pair
of sentences, this time with a cluster whose first clitic is fully specified for person and
number. As expected, the two sentences are ungrammatical.

.â•… Nevins (2007) also deals with the notion of spurious se- and leísmo- in his discussion of
the effects that the representation of 3rd person has in clitic combinations.
 More on the clitic combination puzzle 

(13) a. *Pasad todos, que os nos váis a enfermar

Pass all that 2.pl.refl 1.pl.dat go to get_sick
si seguís en la calle.
if continue on the street
“Come in, you (pl) are going to get sick (on us) if you keep waiting
out in the street”
b. *Pasad todos, que os me váis a enfermar si
Pass all that 2.pl.refl 1.sg.dat go to get_sick if
seguís en la calle.
continue on the street
“Come in, you (pl) are going to get sick (on me) if you keep waiting
out in the street”

In sum, the contrast between (12) and (13) can be taken as further support for the
Plural Blocking Effect (henceforth PBE), since the language is using a strategy that
chooses to impoverish the first clitic of the cluster. This choice eases the way for the
second clitic and, therefore, the resulting combination is grammatical.

3.2â•… The Least Leafy to the Left constraint (Heap 1998, 2005)
In an attempt to account for clitic ordering in both standard and non-standard Spanish,
Heap (2005) proposes a Feature Geometric analysis applied to clitic combinations.
More specifically, he proposes a model based on Harley and Ritter (1998) with elements
from Bonet (1991). The structure proposed by Heap (2005) is reproduced in (14):

(14) CL

PARTICIPANT OTHER

[speaker] [group] CLASS

GENDER CASE

[feminine] [dative]
[Extracted from Heap 2005: 90]

Heap (1998, 2005) proposes that the linear ordering of clitics follows the Least Leafy
to the Left (LLL):

(15) Arrange clitics from the morphologically least specified to most specified.
 Anahí Alba de la Fuente

According to this constraint, the least specified clitics must precede the most specified
clitics.9 In other words, when it comes to number, singular clitics must precede plural
clitics, as indicated by the PBE.
In sum, both the PBE and the LLL constraint account for the clitic combination
patterns in Spanish.

4.â•… Evidence from other Romance languages

4.1â•… Romanian
Romanian presents exactly the same restrictions as Spanish and in the same contexts,
which may lead us to assume that the PBE – and the LLL – is active in Romanian the
same way it is in Spanish.
However, the linear ordering of clitics in Spanish and Romanian are not the same.
Thus, in Romanian, the dative clitic must precede the accusative one and, crucially, the
1st person clitic must precede the 2nd person one (see Săvescu 2007; Nevins & Săvescu
2008). In Spanish, on the other hand, 2nd person must precede 1st person, as we have
seen in (5), (7), (8), (9) and (13). This implies that the ordering of clitic clusters in
Spanish mirrors that of Romanian.10
(16) a. Mi te-ai prezentat la petrecere.
1.sg.dat 2.sg.acc-past introduce at the party
“You introduced yourself to me at the party”
b. Ni te-ai prezentat la petrecere.
1.pl.dat 2.sg.acc-past introduce at the party
“You introduced yourself to us at the party”
c. *Ni v-aţi prezentat la petrecere.
1.pl.dat 2.pl.acc-past introduce at the party
“You introduced yourselves to us at the party”
d. *Mi v-aţi prezentat la petrecere.
1.sg.dat 2.pl.acc-past introduce at the party
“You introduced yourselves to me at the party”

.â•… With respect to the Feature Geometry, specificity – or markedness- correlates with structure
and, so, more complex structures are more marked (Heap 2005:€91).
.╅ Given the fact that clitics are ordered by person and not by case, in Spanish, a sentence such
as Te me presentaron is potentially ambiguous, since it could be interpreted as “They introduced
you to me” (teACC meDAT) or as “They introduced me to you” (teDAT meACC). However, in these
cases, the acc dat order is preferred over dat acc, so Spanish effectively mirrors Romanian with
respect to clitic ordering.
 More on the clitic combination puzzle 

Thus, examples like (16b) and (16d) contradict both the LLL constraint and the PBE
as defined above. Contrary to the predictions of these two constraints, (16b) is gram-
matical, despite containing a plural clitic in first position, and (16d) is ungrammatical,
even though the first clitic is singular.
However, it is essential to remark that the ungrammaticalities surface in the same
contexts in both Spanish and Romanian, so the same restrictions must hold for the
two languages, regardless of the fact that the offending plural clitic is in first position
in Spanish and in second position in Romanian.
Nevins & Săvescu (2008) argue that the singular/plural asymmetry found in
Romanian is an effect of case syncretism. In Romanian, 1st and 2nd singular clitics
show an accusative-dative distinction, whereas 1st/2nd plural clitics are case syncretic.11
However, as we can see in Table 1, Spanish 1st /2nd clitics, both singular and plural,
are case syncretic, and, yet, the restrictions are the same in both languages. Therefore,
the fact that Spanish shows case syncretism both in the singular and in the plural
necessarily indicates that case syncretism cannot explain the asymmetry observed
when combining singular and plural clitics.

Table 1.╇ Accusative and dative clitics in Romanian and Spanish

Romanian Spanish

Accusative Dative Accusative Dative

1sg mă mi me me
2sg te ţi te te
3sg l/o i lo/la le
1pl ne ne/ni nos nos
2pl vă vă/v/vi os os
3pl i/le le/li los/las les

4.2â•… Catalan
In general, Catalan behaves like Spanish and Romanian with an important excep-
tion. Catalan marginally accepts the combination of two plural clitics, in contrast with
Spanish and Romanian.

(17) a. Tu te’m vares presentar a la festa

You 2.sg.refl-1.sg.dat past introduce at the party
“You introduced yourself to me at the party”

.â•… Nevins & Săvescu (2008) explain that, in order to diagnose syncretism, dative clitics must
be looked at in isolation.
 Anahí Alba de la Fuente

b. Tu te’ns vares presentar a la festa

You 2.sg.refl-1.pl.dat past introduce at the party
“You introduced yourself to us at the party”
c. ?Vosaltres us ens vàreu presentar a la festa
You.Pl 2.pl.refl 1.pl.dat Past introduce at the party
“You introduced yourselves to us at the party”
d. *Vosaltres us me vàreu presentar a la festa
You.Pl 2.pl.refl 1.sg.dat Past introduce at the party
“You introduced yourselves to me at the party”
This marginal acceptance of two plural clitics contrasts with the observations in
Spanish and Romanian discussed above, and constitutes a challenge for the PBE. On
the other hand, it corresponds with the acceptance of the Italian cluster vi ci (18), as
defended by authors such as Bonet (1991).
(18) Vi ci manderà
2.pl.dat 1.pl.acc send-fut
“S/he will send us to you”
The sentence in (18) combines two 1st/2nd plural clitics and, even though this is a
violation of the PBE, the sentence is Grammatical in Italian. However, ci can also be a
locative, which means that the sentence could be interpreted as “S/he will send us to
you”, but also as “S/he will send us there”. In fact, the analysis of ci as a locative – rather
than a 2nd person plural clitic – seems to be the preferred interpretation provided by
Italian speakers when presented with this example.
However, even though the example in (18) is potentially ambiguous, given that ci
can either be an object clitic or a locative, both clitics in the Catalan clusters in (17) are
unambiguously object clitics. So, whereas the ambiguity of Italian ci may not be consid-
ered as a robust piece of evidence against the PBE, the Catalan examples in (17) pose a
real challenge for the proposal.

5.â•… The Plural-Blocking Effect (PBE)

As discussed in Section€2, Anagnostopoulou (2005) argues that 1st/2nd clitic combi-

nations are allowed in Weak PCC languages because such languages allow for Multiple
Agree. In addition, in order for Multiple Agree to take place, the two pronouns must
not have conflicting feature specifications.
At the same time, Icelandic shows a blocking effect regarding plural in long
distance agreement in expletive constructions. More specifically, Anagnostopoulou
(2005: 224) explains that “Multiple Agree in Icelandic is possible when both the dative
and the nominative are [–plural] and impossible when the dative is [–plural] and the
nominative [+plural].”
 More on the clitic combination puzzle 

When 1st and 2nd person clitics are combined in Spanish and Romanian, the
dative can be either singular or plural, whereas the non-dative element can only be
singular when combined with a dative.
In view of all this, it becomes obvious that the restrictions presented in Section€3
must be reformulated in order to successfully account for all the data from Spanish,
Romanian and Catalan discussed in Sections€3 and 4. These data show that the accept-
ability of the clitic combinations presented in this paper depends on two notions: case
and feature markedness. The data, thus, show that there is a blocking effect regarding
number and that the acceptance of a cluster also depends on the case of the clitic carrying
the [+Plural] feature.
On the other hand, we have seen that Catalan and Italian also seem to accept com-
binations in which both elements are plural, which is also the case in Icelandic long
distance agreement, according to Anagnostopoulou (2005). Therefore, if we consider
the PBE again, we may rephrase it as follows:
(19) General Plural-Blocking Effect: In a combination of 1st and 2nd person clitics
which includes a dative, the non-dative clitic of the cluster cannot outrank the
dative in number.

This general restriction is common to all the languages discussed in this paper. How-
ever, languages such as Spanish and Romanian reject combinations in which the
non-dative clitic contains a marked [Plural] feature. Thus, we may argue that these
languages adopt a more rigorous version of the restriction, stated in (20):
(20) Narrow Plural-Blocking Effect: In a combination of 1st and 2nd person clitics
which includes a dative, the non-dative clitic cannot be plural.

It is crucial to note that this restriction in two versions actually parallels the PCC, by
which the direct object (i.e. the non-dative element) always has to be third person
(Strong PCC) or, alternatively, when there is a third person in the cluster, it has to be
the direct object (Weak PCC). In other words, the Strong version of the PCC states
that the non-dative clitic cannot be 1st or 2nd person and the Weak PCC states that the
non-dative clitic cannot be more marked than the dative in terms of person. Thus, we
have two parallel restrictions on clitic combinations which deal with case and feature
markedness and always affect the same part of the cluster (the non-dative clitic); the
PCC refers to person and the PBE to number.12,13 In fact, this analogy points towards

.╅ With respect to person restrictions, authors such as Ormazabal & Romero (2007) and
Nevins & Săvescu (2008) defend that, rather than person, we are, in fact, dealing with an
animacy constraint, since they assume that 1st and 2nd person carry the [+animate] feature.
.╅ The parallelism between the PBE and the PCC was also noted and pointed out to me by
Eulàlia Bonet (p.c.) and Oana Săvescu (p.c.).
 Anahí Alba de la Fuente

the possibility that both the PCC and the PBE may actually be part of a larger, more
general constraint drawn by case and feature markedness.

6.â•… Conclusion

There is an asymmetry regarding the combination of singular and plural clitics that
surfaces in a number of Romance languages, such as Spanish, Romanian and Catalan.
In these languages, a combination of 1st/2nd clitics with a dative is always ungram-
matical if the non-dative clitic outranks the dative.
This number-case restriction, the PBE, parallels the person-case restriction pro-
posed by Bonet (1991) (the PCC), it is not due to case syncretism (contra Nevins &
Săvescu 2008), and it does not depend on linear ordering (contra Heap 1998, 2005).
Further research is now required in order to investigate the validity of the PBE
with combinations including 3rd persons. More specifically, given the parallelism
between the PCC and the PBE, it seems necessary to investigate whether the PBE
and the PCC can be unified into one general restriction, and, more generally, to what
extent case and feature markedness are responsible for all the restrictions that arise
when two clitics are combined.

References

Academia Română, Institutul de linvistică “Iorgu Iordan – Al. Rosetti”. 2005. Gramatica limbii
române I, Cuvântul. Bucarest: Editura Academiei Române.
Adger, David & Harbour, Daniel. 2007. “Syntax and Syncretisms of the Person Case Constraint”.
Syntax 10: 1. 2–37.
Alba de la Fuente, Anahí. 2007. “As Time goes by… A Diachronic Analysis of Clitic Restrictions
in Spanish”. Paper presented at the Hispanic Linguistics Symposium 2007, University of
Texas at San Antonio, November 1–4.
Anagnostopoulou, Elena. 2003. The Syntax of Ditransitives: Evidence from Clitics. Berlin, New
York: Mouton de Gruyter.
Anagnostopoulou, Elena. 2005. “Strong and Weak Person Restrictions: A Feature Checking
Analysis”. Clitic and Affix Combinations ed. by Lorie Heggie & Francisco Ordóñez, 199–235.
Amsterdam, Philadelphia: John Benjamins.
Bonet, Eulàlia. 1991. Morphology after Syntax: Pronominal Clitics in Romance Languages. Doc-
toral Dissertation, MIT, Cambridge, Mass.
Grimshaw, Jane. 1997. “The Best Clitic: Constraint Conflict in Morphosyntax.” In Elements of
Grammar ed. by Liliane Haegeman, 169–196 Dordrecht: Kluwer.
Grimshaw, Jane. 2001. “Optimal Clitic Position and the Lexicon in Romance Clitic Systems.”
In Optimality Theoretic Syntax ed. by Géraldine Legendre, Jane Grimshaw & Sten Vikner,
105–240. Cambridge, Mass: MIT Press.
 More on the clitic combination puzzle 

Harley, Heidi & Elizabeth Ritter. 1998. “Meaning in Morphology: Motivating a Feature-Geometric
Analysis”. Ms. University of Calgary and University of Pennsylvania.
Harley, Heidi & Elizabeth Ritter. 2002. “Person and Number in Pronouns: a Feature-Geometric
Analysis”. Language, 78: 3. 482–526.
Heap, David. 1998. “Optimalizing Iberian Clitic Sequences”. Theoretical Advances in Romance
Linguistics ed. by José Lema & Esthela Treviño, 227–248. Amsterdam, Philadelphia: John
Benjamins.
Heap, David. 2005. “Constraining Optimality: Clitic Sequences and Feature Geometry”. Clitic
and Affix Combinations ed. by Lorie Heggie and Francisco Ordóñez, 81–102. Amsterdam,
Philadelphia: John Benjamins.
Holmberg, Anders & Thorbjörg Hróarsdottir. 2002. “Agreement and Movement in Icelandic
Raising Constructions”. Working Papers in Scandinavian Syntax 69. 147–168.
Jakubowicz, Celia, Lea Nash, Catherine Rigaut & Christophe-Loic Gérard. 1998. “Determiners
and Clitic Pronouns in French-Speaking Children with SLI”. Language Acquisition 7, 113–160.
Nevins, Andrew. 2007. “The Representation of Third Person and its Consequences for Person-
Case Effects”. Natural Language and Linguistic Theory 25. 273–313.
Nevins, Andrew & Oana Săvescu. 2008. “An Apparent Number Case Constraint in Romanian:
The Role of Syncretism”. Paper presented at the 38th Linguistic Symposium on Romance
Languages (LSRL 38), University of Illinois, Urbana-Champaign, April 4–6.
Nicol, Fabrice. 2005. “Romance Clitic Clusters: On Diachronic Changes and Cross-linguistic
Contrasts”. Clitic and Affix Combinations ed. by Lorie Heggie & Francisco Ordóñez, 199–235.
Amsterdam, Philadelphia: John Benjamins.
Ormazabal, Javier & Juan Romero. 2007. “The Object Agreement Constraint”. Natural Language
and Linguistic Theory 25, 315–347.
Perlmutter, David M. 1971. Deep and Surface Structure Constraints in Syntax. New York:
Rinehart and Winston.
Rivero, María Luisa. 2008. “Oblique Subjects and Person Restrictions in Spanish: A Morpho-
logical Approach”. Agreement Restrictions ed. by Roberta D’Alessandro, Susan Fischer &
Gunnar Hrafn Hrafnbjargarson, 215–250. Berlin: De Gruyter.
Săvescu, Oana. 2007. “Challenging the Person Case Constraint: Evidence from Romanian”.
Romance Linguistics 2006. Selected Papers from the 36th Linguistic Symposium on Romance
Languages (LSRL 36), New Brunswick, March-April 2006 ed. by José Camacho, Nydia
Flores-Ferrán, Liliana Sánchez, Viviane Déprez & María José Cabrera, 255–269. Amsterdam,
Philadelphia: John Benjamins.
 The Spanish dative alternation revisited*

John Beavers & Chiyo Nishida

The University of Texas at Austin

We examine Spanish dative alternations, and argue that although there are parallels
to English, Harley’s (2003) analysis of English cannot be extended to Spanish,
contra Bleam (2001). We propose an alternative based on the Morphosyntactic
Alignment Principle of Beavers (2006, in press b), wherein the thematic role of
the dative argument is a truth conditional strengthening of the thematic role
of the corresponding oblique, which follows from the dative vs. oblique case
alternation. We support this analysis by showing that it accounts for a range of
less-oft discussed Spanish dative/oblique alternations. We then subsume Harley’s
phrase structural analysis of English and our case-based analysis of Spanish under
a more general notion of relative morphosyntactic prominence, predicting both
the similarities and differences between the languages.

1.â•… Introduction

We examine dative DP/PP alternations in Spanish, as in (1), where the PP a María

“to Mary” in (1a) alternates with dative a María in (1b) doubled by the dative clitic
le. We refer to the theme in (1) as the direct object (DO) and the dative in (1b) as the
indirect object (IO). We refer to (1a) as the Prepositional Dative Construction (PDC)
and (1b) as the Indirect Object Construction (IOC). This alternation corresponds to
a semantic contrast: (1a) can have a caused motion reading but (1b) has only a caused
possession reading.
(1) a. Juan envió la carta a María.
Juan sent the letter to María
“Juan sent the letter to María.”
b. Juan lei envió la carta a Maríai.
Juan cl.dat.3sg sent the letter to María
“Juan sent María the letter.”

*We are grateful to two native speakers of Spanish, Jocelly Guie-Meiners and Iera Zinkunegi
Uzkudun, who helped us with the data. We also thank Ricardo Bermúdez-Otero, Carmen
Dobrovie-Sorin, and Heidi Harley for their feedback.
 John Beavers & Chiyo Nishida

Due to similarities with the English glosses, many researchers (Demonte 1995; Bleam
2001; Cuervo 2002, 2003; de Pedro Munilla 2004, inter alia) have proposed that vari-
ants of the standard English shell analysis (Larson 1988) also apply to Spanish: in both
(1a) and (1b) the complements form a constituent projected by a null head that gener-
ates the appropriate semantics and ensures that the first XP asymmetrically c-com-
mands the second.
We agree that the alternations share a common core. However, we review the data
supporting the shell analysis and conclude that it is not appropriate for Spanish, since
the predicted c-command asymmetries do not obtain. Rather, what is common is not
the syntactic relationship between the complements within each clause, but how the
goal/recipient is realized across clauses: oblique vs. direct argument. Following Beavers
(2006, in press b), when this contrast is semantically significant, it conforms to (2).

(2) Morphosyntactic Alignment Principle (MAP) (Version 1): In an argument/

oblique alternation of argument x of some verb, the direct argument realization
of x has monotonically stronger truth conditions associated with it than the
corresponding oblique realization.

We show that (2) underlies the Spanish and English data in (1), reflecting a contrast
between an oblique goal vs. an IO recipient, where recipient is a semantic subtype of
goal. We also show that Spanish datives alternate with more obliques than just a PPs,
but always conforming to (2), thus providing more substantial support for (2) than
English dative alternations.
In §1 we discuss the basic facts of the alternation, and review evidence for a shell
analysis. In §2 we discuss some syntactic problems with applying this analysis to Spanish
and propose an alternative that does not take the languages to be wholly parallel.
In §3 we discuss some novel semantic generalizations about the alternation and suggest
that the shell analysis does not account for them either. In §4 we propose an alternative
based on (2) that preserves the universal syntax-to-semantics mapping as a relation of
thematic roles to various argument-coding properties, of which phrase structure is
just one. We argue that the differences between the languages fall out of a contrast
in how IOs are coded – positionally in English but morphologically in Spanish.
We conclude in §5.

2.â•… Background

Harley (2003:€33) (building on Larson 1988; Pesetsky 1995:€135–136; see also Rappaport
Hovav & Levin 2008) proposes that the English alternation in the translation to (1)
follows from two different syntactic event decompositions. The analysis of English (1a)
is in (3a), where the two arguments form a small clause-type predication headed by a
 The Spanish dative alternation revisited 

null P that takes the theme the letter as a specifier and the goal to María as a complement.
The analysis of English (1b) is in (3b), where the small clause is instead headed by a
null P indicating possession, taking the recipient Mary as specifier and the theme as
complement.

(3) a. [vP Juan [v’ vCAUSE [PP the letter [P’ PLOC [PP to María ] ] ] ] ]
b. [vP Juan [v’ vCAUSE [PP María [P’ PHAVE [DP the letter ] ] ] ] ]

This predicts that the variants should differ in whether caused possession is entailed
(following Oehrle 1976) and c-command asymmetries between the complements
(following Larson 1988). Based on apparent parallelisms therein with English, Bleam
(2001) extends this analysis to Spanish data as in (1); we outline these arguments
in parallel.
Semantically, in both languages there is a meaning contrast. In Spanish a Londres
“to London” is felicitous in (4a), indicating the inanimate goal of the theme, but not in
(4b), unless Londres refers to a possessor (e.g. Scotland Yard); this also holds in English
(cf. the glosses):

(4) a. Juan envió la carta a Londres.

John sent the letter to London
“John sent the letter to London.”
b. #Juan le envió la carta a Londres.
John cl sent the letter to London
#“John sent London the letter.” (OK on Scotland Yard reading)

Second, Bleam notes that both languages exhibit asymmetric c-command. For example,
in the PDC a DO can bind into the oblique PP but not vice versa (5), while an IO can
bind into a DO, but not vice versa (6). (Note that in Bleam’s data the IO precedes the
DO; we return to this below).

(5) a. El editor envió [cada libro]i a sui autor.

the editor sent each book to its author
“The editor sent [each book]i to itsi author.”
b. *El editor envió sui libro [a cada autor]i.
the editor sent his book to each author
*“The editor sent hisi book to [each author]i.”

(6) a. El editor le envió [a cada autor]i sui libro.

the editor cl sent to each author his book
“The editor sent [each author]i hisi book.”
b. *El editor le envió a sui autor [cada libro]i.
the editor cl sent to its author each book
*“The editor sent itsi author [each book]i.”  (Bleam exx. (14)–(17))
 John Beavers & Chiyo Nishida

However, there are two issues with applying this analysis to Spanish, one syntactic and
one semantic. We start with the syntactic issue.

3.â•… Problems with shell approach and an alternative syntactic analysis

As noted, Bleam’s Spanish IOCs crucially are IO DO, replicating the English order.
However, the neutral order in Spanish is DO IO. (See below for corpus data findings.)
Yet as Bleam herself admits the English/Spanish parallels disappear in this order, e.g.
in (7) the DO can bind into the IO.

(7) El editor le envió [cada libro]i a sui autor.

the editor cl sent each book to its author (Bleam ex. (18))

To account for this, Bleam argues that DO IO is derived from IO DO by moving the
DO across IO, yielding two possible IOCs in Spanish:

(8) a. [ cl V [ IO … [ DO … ] ] ]
b. [ cl V [ DOi … [ IO … [ ti … ] ] ] ]

She gives (9) to argue for this, where the IO backwards binds into the DO, suggesting
it c-commands it at some stage:

(9) (?) El editor le envió sui libro a cada autori.

the editor cl sent his book to each author (Bleam ex. (19))

Bleam does not discuss how this comes about, but de Pedro Munilla (2004) proposes
that it results from optional reconstruction at LF. Thus the Spanish pattern is as in
(10), which is not parallel to English: Spanish has an extra IOC showing symmetric
c-command (cp. (7) and (9)).

(10) a. PDC: V DO PP: PP goal; DO binds into PP

b. IOC: cl V IO DO: IO possessor; IO binds into DO
c. IOC: cl V DO IO: IO possessor; IO/DO bind into each other

Yet (8) is problematic for two reasons. First, neither author discusses what motivates
DO movement. Cuervo (2002) proposes that ditransitives have uninterpretable EPP/
ACC(usative) features that attract the DO to its specifier position. But on the assumption
that ACC must be obligatorily checked, it is not clear how she handles IO DO word
order. Second, more fundamentally, (8) assumes that the neutral word order is derived.
However, the null hypothesis is that it is basic. This conforms to data gathered from
the Spanish Royal Academy corpus CREA (written and oral) constituting 518 tokens
of clitic doubled constructions for enviar “send”, entregar “give/turn in”, poner, “put”,
quitar “remove”, and vender “sell”, all past perfective with 3SG clitic le and 3SG/3PL
 The Spanish dative alternation revisited 

subject. In these data DO IO order (393 tokens) is clearly unmarked, whereas IO DO

order (125 tokens found) is used overwhelmingly (81.6%) when the DO (underscored)
is heavy (see Wasow 2002), as in the following:
(11) …se le quitó a Landívar el derecho a hablar a nombre de esa comisión.
“they took away from Landívar the right to speak on behalf of that comission”

This suggests that the IO comes before the DO for processing. In other cases, DO IO is
the only possibility, as when the verb and the DO form a semantic unit, as in (12).
(12) a. Juan le puso fin a la relación/#a la relación fin.
John cl.3sg put end to the relation/to the relation end
“John put an end to the relationship.”

Thus (10b,c) are not free variants, contra previous authors. Given the marked contexts
of IO DO order, (13) is a preferable analysis, where the neutral order (13a) is underived,
and the marked order (13b) is an alternate base-generated/scrambled option constrained
to certain contexts.
(13) a. [ cl V [ DO … [ IO … ] ] ]
b. [ cl V [ IOi … [ DO … [ (ti) … ] ] ] ]

Prima facie, though, (13a) does not capture symmetric binding. However, we argue
that a simple, independently motivated assumption generates the correct results: the
dative clitic can be a binder in Spanish. Evidence for this comes from (14), where
the reflexive DP inside the PP complement must be interpreted as coreferent with
the clitic, thus, the latter binds the former.
(14) Los fracasos lei arrebataron la confinza en sí mismai.
the failures cl.3sg took.away the confidence in herself
“The failures wrenched from her (her) confidence in herself.”

With this assumption symmetric c-command follows for free from (13a). In (13a)
the clitic c-commands the DO, which c-commands the IO, and thus the clitic can
bind into the DO, which can bind into the IO, predicting symmetric c-command as
the byproduct of two asymmetric c-command relations. Conversely, in (13b) both the
clitic and the IO DP c-command the DO, ruling out symmetric binding. Thus (13),
with the independently motivated assumption that clitics can bind, captures (10c).
However, Cuervo (2003) argues against such an analysis. First, she states that “…
such an approach would predict that in [DO IO] constructions, the accusative should
also be able to bind into the lower dative DP.” (p.42) She gives (15) (her (51)) to show
that this is impossible.
(15) */??La policía les entregó los bebés a sus (respectivos) padres.
the police cl.3pl gave the babies to their respective parents
“The police gave the babies to their respective parents.”
 John Beavers & Chiyo Nishida

However, (15) is independently marginal for having two 3PL animate objects competing
to be linked to les. If we have an inanimate DO, the DO can bind into the IO, contra
Cuervo. (See also Bleam’s (7) above):

(16) La policía les entregó las carteras a sus (respectivos) dueños.

the police cl.3pl gave the wallets to their (respective) owners
“The police gave the wallets back to their respective owners.”

Next, Cuervo argues that a DO IO analysis cannot account for the ungrammaticality of
(17a). Cuervo suggests that (17a) is explained on the IO DO analysis as a Weak Crossover
(WCO) violation, where the DO has been extracted across the coindexed possessive as
in (17b), an explanation not available in (17c).

(17) a. *[Qué (libro) ]i le entregamos a sui dueño?

what book cl we.gave to its owner
“Intended: What (book) did we give to its owner.”
b. [ [ Qué (libro) ]i le entregamos a sui dueño ti ]
c. [ [ Qué (libro) ]i le entregamos ti a sui dueño]

However, the facts are not this clear cut, since (17a) is acceptable to our consultants
(when given contextual information), supporting (17c). Thus WCO does not argue
against a DO IO analysis.
Finally, Cuervo argues that the clitic cannot be a binder since this would require
it to be a pronoun and stand for an argument. We agree that clitics are not DPs and
assume that they are verbal affixes (Miller & Sag 1996), adjoined to a verb. However,
we argue that clitics can stand for an argument. Unlike French and Italian, Spanish
allows clitic doubling, which has posed difficulty for defining what clitics are and how
the IO is realized. In the generative tradition, Spanish IOs are always defined as lexical
XPs and clitics as functional categories, e.g. case absorbers (Jaeggli 1982), agreement
markers (Suñer 1988), heads of AGRIO (Franco 1993), heads of ApplP (Cuervo 2003).
In contrast, we take the more empirically transparent position: the IO can be instanti-
ated by a clitic alone (without postulating a phonologically-null element pro) or clitic/
DP pair, both contributing to the interpretation of the argument, as in (18).

(18) El profesor nos/os/les dio la A a la clase entera.

the professor cl.1pl/2pl/dat.3pl gave the A to the class entire
“The professor gave us/you all/them the entire class an A.”

In (18) the choice of clitic gives critical information about the referent of the IO not
provided by the DP, and the clitic is thus a partial instantiation of the IO. For space
reasons we do not exhibit how the IO is syntactically instantiated by the clitic alone or
clitic + DP; however, e.g. an analysis proposed by Steele (1988) for Luiseño subjects
can be extended to Spanish IOs.
 The Spanish dative alternation revisited 

To summarize, with a simple, independently-motivated assumption – that the

clitic can be a binder – we can capture (10) without ad hoc movement. We turn next to
a semantic issue regarding shell analyses.

4.â•… The morphosyntactic alignment principle

The Harley-style analysis also fails to fully capture the semantics of the dative alternation.
On this analysis the semantic contrast is between caused possession and PLOC, though
Harley never defines the latter relation. However, there is an interesting property under-
lying the contrast between IOCs and PDCs. Consider (19) with an inanimate goal/
recipient. As noted above, the IOC is ruled out since an inanimate cannot be a pos-
sessor (except on a repair reading). But crucially, the PDC allows both caused motion
and caused possession readings, the latter again on a repair reading.

(19) a. John sent a book to London. (Caused motion or possession)

b. #John sent London a book. (Only caused possession)

Thus (19b) encodes a subset of the readings of (19a). How is this possible? We suggest
that the PLOC-type relation in (19a) is a type of abstract co-location subsuming multiple
types of coincidence relationships between entities, which we equate with Hale and
Keyser’s (2002: 208) central coincidence relation (which underlies literal and metaphorical
give and with). A recipient is something that enters into a central coincidence rela-
tionship with the theme and comes to possess it (see also Pesetsky 1995:€ 135–136;
141); a literal goal involves central coincidence plus co-location. Relevant to our data,
the truth conditions that constitute possessor-hood entail central coincidence, but not
conversely, so that, representing each thematic relation as a relationship between two
arguments and an event, the implicational thematic hierarchy in (20) arises.

(20) ∀x∀y∀e[possessor’(x,y,e) → central.coincidence’(y,x,e)]

Thus (19b) entails (19a) but not conversely, explaining the interpretative patterns in
(19). However, Harley (2003: 41ff.) suggests that PDCs do not always encode every-
thing IOCs can, giving light verb and idiom pairs such as gave John a kick/*gave a kick
to John and threw Mary a glance/*threw a glance to Mary (ibid.: ex. (13)) as evidence.
However, in all of her cases IOCs are possible, provided the goal/recipient is heavy (cp.
She gave a kick (in the pants) to every boy who ever did her wrong). This shows that PDCs
can encode everything IOCs encode (see Rappaport Hovav & Levin 2008:€ 150–160
for an extensive discussion); we offer an alternative explanation for cases where IOCs
are preferred below.
Interestingly, as Beavers (2006, in press b) shows, implicational relationships between
variants are not restricted to datives; similar effects are also found with DO/oblique and
 John Beavers & Chiyo Nishida

subject/oblique alternations (albeit with thematic roles not having to do with

possession). Of course, not every alternation shows a truth conditional contrast (as
Rappaport Hovav & Levin 2008 show for the dative alternation with verbs of giving such
as give and hand). But the generalization Beavers proposes is that when an argument/
oblique alternation is semantically contentful, it encodes a monotonic strengthening/
weakening of the truth conditions that comprise the thematic role of the alternating
participant along an independent implicational hierarchy of the sort in (20):

(21) Morphosyntactic Alignment Principle (MAP) (Version 2): In an argument/

oblique alternation of argument x, in the direct argument realization x bears
stronger truth conditions along some implicational thematic hierarchy than in
the oblique realization.

Note that we view dative alternations as lexicalized rather than derived, and more gen-
erally (21) is a lexicalization principle predicting what kinds of lexicalized alternations
will be possible. Furthermore, this principle does not predict which verbs will alter-
nate, e.g. it does not explain why certain verbs (costar “to cost”) lack one of the two
variants, since that is due to the idiosyncrasy of these verbs. It simply predicts what
semantic contrast will be instantiated when an alternations is lexicalized.
The MAP underlies Spanish dative alternations as well: (22a) is also compatible
with both motion/possession readings, but (22b) has only a possession reading,
mirroring (19) and strengthening the parallelism.

(22) a. Juan envió la carta a María/Londres.

Juan sent the letter to María/London
“Juan sent the letter to M./London.” (London or Scotland Yard)
b. Juan lei envió la carta a Maríai/#Londresi.
Juan cl gave the letter to María/London.
“Juan sent María/#London the letter.” (Only Scotland Yard)

In fact, we show that not only does the MAP apply in (22), it is significantly more
pervasive than in English. We discuss several IO/oblique alternations in Spanish that
do not have English parallels and encode contrasts other than goal vs. recipient. Yet
in each case the MAP holds, suggesting its generality in Spanish. Partly for space rea-
sons and partly because these results are tentative, we leave the relevant roles vague,
although greater or lesser affectedness (perhaps modeled as in Beavers in press a) is
usually key. The crucial point is that the conditions for the IO always subsume those of
the oblique, but additional constraints also necessarily obtain.
De “from/off ” Alternations. Consider first the IO/de alternation in (23), where
in (23a) the tablecloth is removed (weak reading), while in (23b) the table has addi-
tionally been made bare (e.g. has lost an integral part; strong reading), reflecting monotonic
strengthening (this also occurs with sacar “take out”, arrancar “pull out”, confiscar
“consfiscate”, extraer “extract”, etc.).
 The Spanish dative alternation revisited 

(23) a. Juan quitó el mantel de la mesa.

Juan removed the tablecloth from the table
“John removed the tablecloth from the table.”
b. Juan le quitó el mantel a la mesa.
Juan cl removed a chair to the table
“John removed the tablecloth from the table.” (affecting it)

To illustrate the semantic contrast we need a way of ruling out readings to see which
reading each variant allows. Consider taking an instrument out of a glass case, where
(presumably) an instrument never constitutes an integral part of a glass case. Here the
PDC is acceptable but crucially the IOC is unacceptable as in (24).

(24) a. El dentista extrajo el instrumento de la vitrina.

the dentist extracted the instrument from the glass case
“the dentist took out the instrument from the glass case.”
b. #El dentista le extrajo el instrumento a la vitrina.
the dentist cl extracted the instrument to the glass case

This mirrors the IO/a facts. However, we also expect the PDC to be compatible with
both readings. To see this, we need a context where the strong reading must hold, for
example extracting a molar from the mouth, assuming a molar is an integral part of
the mouth. As expected, the IOC is acceptable. But here the PDC is unacceptable,
contra the MAP:

(25) a. #El dentista extrajo la muela del niño.

the dentist extracted the molar from the boy
b. El dentista le extrajo la muela al niño.
the dentist cl extracted the molar to. the boy
“The dentist extracted the boy’s molar.”

This suggests that the MAP does not hold here – rather, the PDC encodes the weak
reading and a denial of the strong one. Is this problematic?
We suggest not. Rather, it could be that the existence of a dedicated variant for
encoding the strong reading in (25b) blocks this reading from obtaining in (25a).
Indeed, for distinct, competing variants, we might expect three types of relationships:

(26) a. The oblique variant is compatible with weak and strong readings.
b. The argument variant blocks the strong reading in the oblique.
c. The oblique variant implicates a denial of the strong reading.

As Beavers (2006, in press b) shows, English alternations generally have the property in
(26a). But so far Spanish has some alternations with the property in (26a) (IO/a) and
some with (26b) (IO/de). Indeed, English may also exhibit some blocking: this would
account for the light verb and idiomatic data above where IOCs are preferred (but not
 John Beavers & Chiyo Nishida

categorically) over PDCs (thanks to Heidi Harley, p.c., for pointing this out to us). If
(as Harley 2003 suggests) these are cases where strong readings are obligatory, and the
IOC is the canonical way to encode this, it may block the PDC unless other factors
(heaviness) intervene (providing a cleaner analysis than the Heavy-NP shift + to –
insertion analysis of Harley 2003: fn.9). Finally, it is not hard to imagine languages
where (26c) might hold (Beavers 2006:€226ff. suggests that certain accusative/dative
alternations in Spanish hacer causatives may instantiate this). Thus IO/de alternations
do follow the MAP if blocking is taken into account, suggesting commonality with
IO/a alternations. We do not propose here an explanation for when blocking does or
does not occur, leaving this for future work.
En “in/on/into/onto” Alternations. Other verbs show IO/en alternations. An
example is in (27), where in the PDC the entailment is that the tablecloth is on the table
(weak reading), but in the IOC there is an additional entailment that it has entered into
the appropriate setting relationship with the table, namely covering it (strong reading)
(other such verbs are meter “insert/put in(to)”, colocar “place”, etc.).
(27) a. Juan puso un mantel en la mesa.
Juan put the tablecloth in the table
“John put the tablecloth on the table.”
b. Juan le puso un mantel a la mesa.
Juan cl put the tablecloth to the table
“John put the tablecloth on the table.” (covering it)
To test the MAP we must force weak/strong readings. For putting fruit in a fridge
(which presumably does not affect the fridge) we predict only the PDC to be possible.
Alternatively, when stuffing a turkey, i.e. turning it into a dish, we expect the IOC to
be fine and the PDC to either also allow this reading or be blocked. These are shown
in (28)–(29), with blocking.
(28) a. Juan metió la fruta en el frigo.
John inserted the fruit in the frige
“John put the fruit into the fridge.”
b. #Juan le metió la fruta al frigo.
John cl inserted the fruit to.the frige
(29) a. #Metes el relleno en el pavo y lo cueces en el horno.
You.put the stuffing in the turkey and it you.cook in the oven
b. Le metes el relleno al pavo y lo cueces en el horno.
cl you.put the stuffing to.the turkey and it you.cook in the oven
“You put the stuffing in the turkey and you cook it in the oven.”
Another such alternation has a stative meaning: in (30a) Juan discovers something
strange in the box (weak reading), while in (30b) the box is characterized by strange-
ness (strong reading) (encontrar “find”, detectar “detect”, reconocer “recognize”, oír
“hear”, sentir “feel” also show this).
 The Spanish dative alternation revisited 

(30) a. Juan vio algo extraño en la caja.

Juan saw something strange in the box
“John saw something strange in the box.”
b. Juan le vio algo extraño a la caja.
Juan cl saw something strange to the box
“John found the box strange.”

This again follows the MAP, if containing something strange means having a strange
quality, while being characterized by strangeness means having a strange quality
characteristic of the entire participant. Examples (31)–(32) illustrate the meaning
differences: IDs do not characterize wallets, but arrogance characterizes people (and
again we see blocking).

(31) a. Mi padre siempre encuentra el carnet en la cartera.

My father always finds the id in the wallet
“My father always finds his id in his wallet.”
b. #Mi padre siempre le encuentra el carnet a su cartera.
My father always cl finds the id to his wallet

(32) a. #Juan encuentra cierto aire de arrogancia en su nuevo jefe.

John finds certain air of arrogance in/about his new boss.
b. Juan le encuentra cierto aire de arrogancia al nuevo jefe.
John cl finds certain air of arrogance to.the new boss
“John finds certain air of arrogance in/about his new boss.”

Para “for” Alternations. Finally we have IO/para alternations. Consider (33), where (33a)
possession may but does not necessarily obtain (i.e. a benefactive reading is possible),
while in (33b) possession must also obtain. Thus in the PDC hijo primogénito may
be a non-existing or existing person, while in IOC the first son must exist. Therefore,
the MAP is met (and no blocking occurs; this also occurs with dibujar “draw”, hacer
“make”, encontrar “find”, comparar “buy”, buscar “search”, etc.).

(33) a. Juan construyó una casa de campo para su hijo primogénito.

Juan built a country house for his son first-born.
“John built a country house for his first son.” (who may not exist)
b. Juan le construyó una casa de campo a su hijo primogénito.
Juan cl built a country house to his son first-born
“John built his first son a country house.” (who exists)

5.â•… The Syntax/Semantics mapping

What these alternations bear in common is the weak/strong to PP/IO correlation

predicted by the MAP, which varies by the factors in (34).
 John Beavers & Chiyo Nishida

(34) a. Whether the PP is goal (a, en, para), source (de), or location (en);
b. The verb-specific extra io entailment, e.g. type of affectedness.

Thus the MAP links Spanish and English. However, the PHAVE/PLOC analysis does not
capture this, at least not at an appropriate level of generality. First, it only predicts a dif-
ferent predication relationship between the two arguments, but not necessarily that the
roles assigned to the PP and IO should be related, much less by monotonic strengthening.
It also does not extend to the other dative alternations in Spanish, since more MAP-
type relationships than just possession vs. central coincidence are involved. One could
posit more heads that stand in more alternations like (34), though the MAP only
emerges as an ad hoc fact.
Of course, we can say that the MAP is a general constraint on possible alternations
of theta-role assigning heads, which is the analysis we propose to capture the above
data. But there is a bigger problem with the shell-type analysis, namely the morpho-
syntax. The MAP relates a semantic contrast to different realization options for just one
participant (even if the semantic relations are dyadic). Yet in the PHAVE/PLOC alternation
the only obligatory morphosyntactic correlate is a change in specifier/complement
relations between the arguments. While it may be appropriate for English, this corre-
lation does not hold for Spanish, where the contrast is between a clitic doubled dative
DP and a PP, regardless of the positional relation to the DO (see §2). So how can we
preserve the MAP – or any uniform syntax-to-semantics mapping – while maintaining
cross-linguistic applicability? We need a grammatical distinction that is (a) a priori to
phrase structure and morphology and (b) able to compare realization options for a
single argument, not just co-arguments.
Following Beavers (2006) (who builds on Levin and Rappaport Hovav 2005), we
suggest that the relevant distinction is the more general one of morphosyntactic promi-
nence. It is well known that various argument coding devices – case, grammatical
function, and syntactic position – can be ranked relative to one another by crite-
ria appropriate for each category. For example, cases are ranked according to their
relative morphological markedness, grammatical functions are ranked according
to behavioral properties such as (but not limited to) accessibility to operations such
as relativization, and syntactic positions are ranked by c-command asymmetries, as
summarized in (35).
(35) a. Case Markedness – Nom > Acc > Dat > Obl (Croft 2003:€142ff.)
b. NP - Accessibility – SU > DO > IO > OBL (Keenan & Comrie 1977)
c. C-command – SU > IO > DO > OBL (Pesetsky 1995:€160–163)

Crucially, we can link monotonic strengthening to an increase in an argument’s

morphosyntactic prominence along one or more hierarchies. In both English and
Spanish obliques are coded as PPs. The difference – by which the word order and
c-command facts follow – is how IO is encoded: in English it is coded positionally,
 The Spanish dative alternation revisited 

relative to other DP positions, while in Spanish it is coded by case, i.e. a dative clitic with
optional doubling. This difference has an effect on the syntactic diagnostics expected
of dative alternations. In English, the positional fact is that the IO always c-commands
the DO (since English does not allow VP-internal scrambling of two DPs). In Span-
ish the positional relationship of IO and DO DPs is not part of IO or DO encoding and
is not restricted by any other categorical factors. This is not to say that IO coding in
Spanish is not also partly positional, only that it is not relative to other DPs. Thus no
particular c-command relationship must obtain in the alternation; rather, the alternating
heads constrained by the MAP in Spanish vary instead in the case they assign the rel-
evant argument (and are not necessarily syntactically dyadic). Thus by generalizing the
universality claim from categorical c-command relationships between co-arguments
to relative morphosyntactic prominence of ways of encoding a single argument (which
may manifest as c-command), we can maintain a universal syntax/semantics mapping
across languages, with the Harley-type analysis being one of many instantiations.

6.â•… Conclusion

We have supported the parallelism view that English and Spanish dative alternations.
However, the similarity goes only as far as the relationship between semantics and
morphosyntactic prominence (a cross-linguistic variable) for one argument, with a
single Morphosyntactic Alignment Principle governing both alternations. This raises
the question of what we expect cross-linguistic parallels to be: can we apply analyses
appropriate for one language wholesale to another, or should we use cross-linguistic
comparison to isolate the factors needed to state universal principles?

References

Beavers, John. 2006. Argument/Oblique Alternation and the Structure of Lexical Meaning. Ph.D.
thesis, Stanford University.
Beavers, John. In press a. “On Affectedness”. Natural Language and Linguistic Theory.
Beavers, John. In press b. “The Structure of Lexical Meaning: Why Semantics Really Matters”.
Language.
Bleam, Tonia. 2001. “Properties of the Double Object Constructions in Spanish.” A Romance
Perspective of Language Knowledge and Use, ed. by Rafael Nuñez-Cedeño, Luis López€&
Richard Cameron. 233–253.Amsterdam: John Benjamins.
Croft, William. 2003. Typology and Universals. Cambridge: CUP.
Cuervo, María C. 2002. “Structural Asymetries but Same Word Order.” (Asymmetry in Grammar
Volume€ 1: Syntax and Semantics, ed. by Anna M. Di Sciullo. 117–144. Amsterdam:
John Benjamins.
 John Beavers & Chiyo Nishida

Cuervo, María C. 2003. Datives at Large. Ph.D. thesis, MIT.

Demonte, Violeta. 1995. “Dative Alternation in Spanish”. Probus 7.5–30.
De Pedro Munilla, Mónica. 2004. “Dative Doubling Structures in Spanish: Are they Double
Object Constructions?” WCCFL 23, pp. 169–182. Somerville: Cascadilla Press.
Franco, Jon. 1993. On Object Agreement in Spanish. Ph.D. thesis, University of Southern California.
Hale, Kenneth & Samuel J. Keyser. 2002. Prolegomenon to a Theory of Argument Structure.
Cambridge, Mass.: MIT Press.
Harley, Heidi. 2003. “Possession and the double object construction, Linguistics variation Yearbook,
ed. by P. Pica & J. Rooryck 2. 31–70.
Jaeggli, Osvaldo. 1982. Topics in Romance Syntax. Dordrecht: Foris.
Keenan, Edward. & Bernard Comrie. 1977. “Noun Phrase Accessibility and Universal Grammar”.
Linguistic Inquiry 8.62–100.
Larson, Richard. 1988. “On the Double Object Construction”. Linguistic Inquiry 2.333–378.
Levin, Beth & Malka Rappaport Hovav. 2005. Argument Realization. Cambridge, UK: Cambridge
University Press.
Miller, Philip H. & Ivan Sag. 1996. “French Clitic Movement Without Clitics or Movement”. Natural
Language and Linguistic Theory 15.573–639.
Oehrle, Richard T. 1976. The Grammatical Status of the English Dative Alternation. Ph.D. thesis,
MIT.
Pesetsky, David. 1995. Zero Syntax: Experiencers and Cascades. Cambridge, Mass: MIT Press.
Rappaport Hovav, Malka & Beth Levin. 2008. “The English Dative Alternation: A Case for Verb
Sensitivity”. Journal of Linguistics 44.129–167.
REAL ACADEMIA ESPAÑOLA, Corpus de referencia del español actual (CREA), http://www.rae.es.
Steele, Susan. 1988. “Subject Values”. Language 65.537–577.
Suñer, Margarita. 1988. “The Role of Agreement in Clitic-doubled Constructions”. Natural Language
and Linguistics Theory 6.391–434.
Wasow, Thomas. 2002. Postverbal Behavior. Stanford: CSLI Publications.
 Romanian genderless pronouns
and parasitic gaps

Ion Giurgea
Institute of Linguistics of the Romanian Academy
& University of Constance

I argue that the distribution of (overt) object clitics and null objects in Romanian
can be explained if we assume that the so-called “neuter pronouns” of Romanian
are genderless. I show that Romanian has a null object used as a bound variable
with a neuter pronoun antecedent. This item differs from parasitic gaps by the
fact that it does not require an A-bar moved antecedent and can only occur with
neuter pronouns, while overt clitics are excluded in this context. I propose that
this is due to the fact that object clitics are always marked for gender, while neuter
pronouns are genderless. I present independent evidence for the proposal that the
so-called “neuter pronouns” of Romanian and other Romance languages, definite
as well as indefinite and quantificational, lack a value for Gender.

1.â•… Introduction

In this paper, I will show that the distribution of (overt) object clitics and null objects
in Romanian can be explained if we assume that the so-called “neuter pronouns” of
Romanian are genderless. In addition to the null anaphors licensed by verbs which
take propositional objects, Romanian has a null object which is only used as a bound
variable with a neuter pronoun antecedent. This very restricted use explains the fact
that this type of object has gone unnoticed until now, being misinterpreted, in some
of its contexts, as a parasitic gap. However, this item differs from parasitic gaps by
the fact that it does not require an A-bar moved antecedent but instead requires its
antecedent to be a neuter pronoun. On the other hand, overt clitics are excluded in
the context where this item appears. This distribution can be explained by assuming
that Romanian object clitics are always marked for gender, while neuter pronouns
are genderless.
After presenting the evidence for anaphoric object pro in Romanian (Section€2),
I will argue for the existence of genderless pronouns in Romanian as well as other
Romance languages (Section€3). Crucial evidence for the lack of gender comes from
definite “neuter” pronouns. In languages with a binary masculine/feminine gender
 Ion Giurgea

opposition, genderless pronouns are used for reference to entities which do not fall
under a nominal concept – either uncategorized perceptual objects, or propositions
and state-of-affairs introduced by clausal projections. The genderless pronouns of
Romanian are pro, the demonstratives asta/aceasta and aia/aceea (formally identi-
cal to the feminine singular, but distinguished from the feminine singular by their
syntactic behavior with respect to agreement, clitic-doubling and accusative marking)
and the so-called neuter indefinite and quantificational pronouns (ceva “something”,
ce “what”, nimic “nothing”, orice “anything”, tot(ul) “everything”). In Section€ 4 I will
briefly address the implications of the findings in this article for the analysis of the
Romanian so-called “neuter nouns”, a class of nouns which trigger masculine agree-
ment in the singular and feminine agreement in the plural and have sometimes been
analyzed as constituting a third gender.

2.â•… Null objects in Romanian vs. parasitic gaps

As known at least since Dobrovie-Sorin (1994), Romanian does not have arbitrary
object pro. The correspondent of (1)a in Romanian is agrammatical:1
(1) a. Questa musica rende pro allegri (It.) (Rizzi 1986)
this music makes happy.mpl
“This music makes people happy”
b. *Muzica asta face fericiţi (Ro.)
music-the this makes happy.mpl

However, I will argue that Romanian has anaphoric object pro. The evidence for this
type of pronoun comes from a construction which resembles parasitic gaps. The
received view on parasitic gaps in Romanian is that they exist in the language, but
are restricted to non-clitic-doubled Ā-chains (Dobrovie-Sorin 1990, 1994; Cornilescu
2002; Alboiu 2002):
(2) a. Ce-ai aruncat fără să citeşti?
what have.2sg thrown without subj read.2sg
“What did you throw away without reading?”

.╅ The following abbreviations are used in this paper

1,2,3 1st, 2nd, 3rd person cl clitic neg negative clitic (French)
acc accusative dat dative neut neuter
art genitival article f feminine pro-n-cl pro-N-clitic
(a genitive and possessive imper imperative sg singular
marker agreeing inf infinitive subj subjunctive particle
with the possessee) m masculine
 Romanian genderless pronouns and parasitic gaps 

b. *Pe care l-ai aruncat fără să citeşti?

obj which it-have.2sg thrown without subj read.2sg

However, for me and many other people I consulted, the contrast does not oppose
clitic-doubled and non-clitic-doubled fronted elements. What looks like a parasitic
gap is only possible with neuter pronouns (ce “what”, nimic “nothing”):
(3) a. Ce-ai aruncat fără să citeşti?/ fără a citi?
what have.2sg thrown without subj read.2sg/ without to read
b. Ce-ai mâncat fără să tai?/ fără a tăia?
what have.2sg eaten without subj cut.2sg/ without to cut
“What did you eat without cutting?”
(4) NIMIC n-am mâncat fără să tai
nothing not-have.1sg eaten without subj cut.1sg
“I ate nothing without cutting it”

All other types of non-clitic-doubled fronted phrases – DPs of the form [ce NP] “what
NP”, fronted bare NPs, the animate wh-pronoun cine –, although non-D-linked,
exclude an object gap:
(5) a. Ce carte ai aruncat fără să ??(o) citeşti?/ fără
what book have.2sg thrown without subj(it) read.2sg/ without
a *(o) citi?
to (it) read
“What book did you throw away without reading?”
b. Ce aliment ai mâncat fără să â•›*(-l) tai?/ fără
what aliment have.2sg eaten without subj(it) cut.2sg/ without
a *(-l) tăia?
to ↜(it) cut
“What aliment did you eat without cutting?”
(6) MACAROANE am mâncat fără să â•›*(le) tai
pasta(fpl) have.1sg eaten without subj (themFPL) cut.1sg
“It is pasta that I ate without cutting”
(7) Pe cine ai admirat înainte de a%*(-l) cunoaşte?
obj who have.2sg admired before of to â•›(him) know
“Whom have you admired before meeting?”

The explanation I propose for this distribution is that Romanian (or at least the idiolect
in which the contrast in (3)–(7) is found) does not have parasitic gaps at all, and what
looks like a parasitic gap in constructions with ce or nimic is in fact a genderless object
pro. The idea is that every time the accusative object can have a value for the category
gender, it will appear in the form of a clitic. In (5)–(6), where there is a nominal ante-
cedent, the pronoun takes the gender of the noun (pronouns anaphoric to expressions
 Ion Giurgea

which contain a noun can always take the grammatical gender of the noun of their
antecedent). In case the antecedent does not contain a noun but is animate, like in€(7),
the pronoun can take the masculine as a “natural” (interpretable) gender, since in
Romanian, like in the other Indo-European languages which have inflectional gen-
der, the masculine as a natural gender is interpreted as +animate (and +male by an
implicature). In (3)–(4), the antecedent is a neuter pronoun. As will be shown in the
next section, neuter pronouns are arguably genderless, so the anaphoric pronoun can-
not take the gender of its antecedent. Moreover, since natural gender is restricted to
animates (the masculine is interpreted as +animate, and the feminine is interpreted as
+female),2 the anaphoric pronoun cannot appear with a gender feature interpreted as
natural gender. It follows that the anaphoric object must be genderless. The fact that we
find in this case null objects instead of clitics can be explained if we assume that clitic
forms are always marked for gender in Romanian (i.e. there are no morphological
defaults for the category gender in the paradigm of accusative clitics), while object pro
is genderless. Notice indeed that overt pronouns are excluded in (3)–(4):

(3′) a. *Ce-ai aruncat fără să-l/o citeşti?

what have.2sg thrown without 3rdmsg.acc/3rdfsg.acc read.2sg
b. *Ce-ai mâncat fără a-l/o tăia?
what have.2sg eaten without to 3rdmsg.acc/3rdfsg.acc cut

(4′) *NIMIC n-am mâncat fără să-l/o tai

nothing not-have.1sg eaten without subj 3rdmsg.acc/3rdfsg.acc cut.1sg

The data presented so far allow an alternative explanation: one may say that the para-
sitic gap construction exists in the language but is just dispreferred, and the speakers
only use it as a last resort when no gender is available for the object pronoun. We
may decide between the two explanations using contexts where the antecedent of the
pronoun has not undergone movement. If the object in (3)–(4) is a genderless pro,
we expect it to appear also in these cases, while if it is a parasitic gap, it should not be
allowed if its binder has not undergone A-bar movement. The following examples sup-
port the pro hypothesis, showing null objects anaphoric to indefinite pronouns which
have not undergone A-bar movement:

(8) a. Au adus ceva ca să monteze mâine

have.3pl brought something in-order-to subj fix/mount.3pl tomorrow
“They brought something to mount tomorrow”
b. Au adus o sculă ca s- ↜*(o)
have.3pl brought an equipment in-order-to. subj (it)

.â•… On the so-called ‘neuter gender’ of Romanian, see Section€4.

 Romanian genderless pronouns and parasitic gaps 

monteze mâine
fix/mount.3pl tomorrow
“They brought a device which they should mount tomorrow”

(9) a. N-atinge nimic fără să strice

not-touches nothing without subj breaks
“(S)he doesn’t touch anything without breaking it”
b. N-atinge nici o jucărie fără s-╇ *(o) strice
not-touches no toy without subj (it) breaks
“(S)he doesn’t touch any toy without breaking it”
c. Încearcă, te rog, să atingi ceva fără să strici
try.imper please subj touch.2sg something without subj break.2sg
“Would you try to touch something without breaking it?”

These examples, together with the sharp deviance of object gap with A-bar antecedents
containing an N or an animate pronoun ((5)–(7)) provide sufficient evidence against
analyzing the null objects in (3)–(4) as parasitic gaps.
However, this type of null object does not behave like regular pronouns either.
Thus, the antecedent cannot be in another sentence:

(10) Au adus cevai. *O să monteze proi mâine

have.3pl brought something fut subj fix/mount.3pl tomorrow

I conclude that anaphoric genderless objects are restricted to a bound variable

use. This idea can be formalized using Kratzer’s (1998, 2006) proposal that at least
some instances of pronouns with a bound variable reading represent bare indices
with φ-features inherited via Agree from the binder. Adopting this theory, what
we called null object pro can be considered to be the spell-out of a bare index with
unvalued Gender.
Kratzer uses this theory to explain the existence of bound variable readings for 1st
and 2nd person pronouns, as reflected in the sloppy reading of an example such as:

(11) I’m the only one who takes care of my children

(sloppy reading = the others do not take care of their children)

Notice that the null objects in (3)–(4) and (8)–(9) are inside subjunctive adjunct clauses
(introduced by “without”, “before”, “in order to”). If the assimilation of these objects to
Kratzer’s bare indices is correct, we predict that these environments should allow for
bound variable readings of 1st and 2nd person pronouns. The following example shows
that this prediction is borne out:

(12) Numai eu am plecat fără să ştie

only I have left without subj knows
supraveghetorul meu (sloppy reading)
supervisor-the my
 Ion Giurgea

In conclusion, Romanian null objects are used as bound variables which have neuter
pronouns as antecedents.3 This can be explained by the fact that neuter pronouns do
not have gender, while object clitics are always marked for gender.

3.â•… Genderless pronouns in Romance and Romanian

In this section, I provide evidence for the proposal that the so-called “neuter pronouns”
of Romanian and other Romance languages are genderless.
I will first consider definite neuter pronouns. In Romance languages, including
Romanian, definite neuter pronouns are used for referents which do not fall under
a nominal concept. There are two situations of reference to entities which do not fall
under a nominal concept: (i) the referent is a perceptual object which has not been
categorized (“identified”) yet (see (13)) or (ii) the referent is a propositional object,
introduced in the discourse by a clausal projection (see (14)):
(13) a. Ce-i asta? (Rom.)
what is that
b. Qu’est-ce que c’est ça? (Fr.)
c. Qué es esto? (Sp.)
(14) a. Nu cred asta. (Rom.)
not believe.1sg this
b. Cela je ne le crois pas (Fr.)
this i neg it believe not

.╅ Null object pronouns must be distinguished from the sequence null D + noun-ellipsis
(i.e. nominal ellipsis in bare nouns). As Giannakidou and Merchant (1996) and Panagiotidis
(2002) have shown for Greek, and Giurgea (2008) for Romanian, what looks like an indefinite
null object in examples such as (i) is to be analyzed as the null D of bare nouns (cf. Longobardi
1994) followed by noun ellipsis. One argument for this analysis is the possibility of
having overt modifiers of the noun, like in the other instances of noun ellipsis as
shown (ii)–(iii):
(i) Nu mai sunt pahare. – Lasă că aduce [Ø] Maria
not more are glasses let.imper that brings Maria
‘There are no more glasses’ ╛╛– ‘Don’t worry, Mary will bring some’

(ii) Ai luat trandafiri galbeni? Eu aş fi vrut [Ne] roşii.

have.2sg bought roses yellow I would have liked red
‘Did you buy yellow roses?’ ‘I would have preferred red’

(iii) Aici se construiesc numai case din lemn. Noi avem şi [Ne] de cărămidă
here refl build.3pl only houses in wood we have also of brick
‘Here people build only wooden houses. We also have brick houses’
 Romanian genderless pronouns and parasitic gaps 

c. Esto no lo creo (Sp.)

this not it believe.1sg

Gender on definite pronouns can reflect either the gender of their antecedent (“ana-
phoric gender”) or a property of the referent (“natural gender”). Romance languages
have a binary gender opposition on pronouns between masculine and feminine, and as
these names suggest, these genders, as natural genders, reflect properties of animates
(i.e. sex; the masculine is the unmarked term, see above). Gender on pronouns can
also be anaphoric (this being the only option for inanimates). In this case, the gender
of the pronoun reflects the gender of the nominal concept under which the referent
falls, if the pronoun is referential, or the gender of its binder, if the pronoun has a
bound variable reading. If the pronoun has an antecedent in the discourse, it will take
the gender of (the noun of) its antecedent. Since gender is a property of nouns, this
indicates that besides co-reference the pronoun has a relation of identity-of-sense
anaphora with its antecedent, which we may call “nominal anaphora” (see Corblin
1995 on this notion).4

(15) a. Am pus paltonuli pe scaun. Peste eli am

have.1sg put coat(m)-the on chair over 3rd.msg have.1sg
pus umbrela. (Rom.)
put umbrella-the
b. J’ai laissé mon manteaui là-bas. Ili doit
I have left my coat(m) over there 3rd.msg must
être nettoyé (Fr.)
be cleaned

If the pronoun is used deictically, it will have the gender of the nominal concept under
which the referent falls:

(16) [before a bill fallen on the ground]

a. Ia-o, ce mai aştepţi (Rom.) (hârtie “bill” – fem.)
take-3rd.fsg what still wait.2sg.
“Take it, what are you waiting for?”
b. Prends-le, tu hésites encore? (Fr.) (billet “bill” – masc.)
take-3rd.msg you hesitate still

.╅ Therefore it has been proposed that personal pronouns contain an empty N, which pro-
vides the gender (see Panagiotidis 2002, a.o.). There are also pronouns whose only relation
with their antecedent is nominal anaphora – the so-called ‘laziness pronouns’. For an overview
of the various cases in which the only relation between the pronoun and their antecedent is
nominal anaphora, see Elbourne (2005).
 Ion Giurgea

Since noun ellipsis can also involve a concept which is salient in virtue of its presence in
the communication situation rather than in the discourse (what has been called “prag-
matic antecedent” by Hankamer and Sag (1976)), as shown in (17) below, the facts in
(16) confirm the idea that gender in pronouns may come from nominal anaphora.5

(17) [before a hat on a shop display]

a. Am şi eu una aşa (Rom.) (pălărie “hat” – fem.)
have.1sg also I one.f like-this
b. Moi aussi j’en ai un
me too I pro-n-cl have one.m
comme ça (Fr.) (chapeau “hat” – masc.)
like this
“I too have one like this”

But, as we have seen in (13)–(14), there are cases in which pronouns must refer to enti-
ties for which there is no nominal concept available (either they are perceptual objects
not yet categorized, or propositional objects introduced into the discourse by CPs).
What gender can these pronouns have? Nominal anaphora cannot provide gender,
since there is no nominal concept under which the referent falls, and natural gender
cannot be used either, because it is restricted to animates. Then we expect to find forms
lacking gender.
Before providing evidence that the forms with this use – which I will call anomi-
nal€ – are indeed genderless, I would like to point out that languages which have a
neuter gender typically use the neuter in this case:

(18) a. Ich glaube es nicht (Germ.)

I believe it not
b. Nonne mauis illud credere (…) (Latin)
not-inter prefer.2sg that.neut believe.inf
“Don’t you prefer to believe that…” (Cicero, De Natura Deorum, III.12)

.╅ Discourse anaphora and deixis are arguably two facets of the same phenomenon: refer-
ence to a contextually salient entity, or, in the case of identity-of-sense anaphora, recovery
of a contextually salient concept. An entity or concept may be salient either by having been
mentioned in the discourse (discourse anaphora) or by its presence in the utterance context
(deixis). This explains why there are no demonstratives specialized for contextually salient
non mentioned entities, but languages consistently use the same expressions for reference to
previously mentioned entities and to contextually salient non mentioned entities.
 Romanian genderless pronouns and parasitic gaps 

(19) a. Nescio id quid est (Latin)

not-know.1sg that.neut what is
b. Was ist das? (Germ.)
what is that.neut
The use of the neuter may be explained by the fact that it is the semantically unmarked
gender, used both for inanimates and for maximal generality. In other words, the
−animate interpretation is the result of an implicature, so that the neuter can be said
to be devoid of any descriptive content in its use as natural gender:
(20) Ex nihilo nihil (Latin)
from nothing nothing
(21) mens, quae (…) supponit ea omnia non existere de quorum
mind which supposes those.neut all.neut not exist.inf of whose
existentia vel minimum potest dubitare (lat.)
existence even very-little can doubt.inf
 (Descartes, Meditationes de prima philosophia, Synopsis, 12)
“The mind, which supposes that everything it can doubt does not exist”
Now I will proceed to the discussion of “anominal” pronouns in Romance, arguing
that they are genderless.
As anominal pronouns, we sometimes find special forms (see Meyer Lübke, Rom.
Gr. III, § 87, 98–99, II § 98). Iberic languages use a special inflection, -o, restricted to
the singular:
(22) Sp.: masc.sg. fem.sg. anominal (“neuter” pronoun):
3rd person él ella ello
demonstratives: este esta esto (close to the speaker)
ese esa eso (close to the hearer)
aquél aquella aquello (remote)
Port.: demonstratives: este esta isto (close to the speaker)
esse essa isso (close to the hearer)
aquele aquela aquilo (remote)
In French, Catalan and Italian we find a special root: French ce/ça (demonstrative and
weak pronoun), ceci, cela vs. celui-ci/celle-ci, celui-là/celle-là (demonstratives), it. ciò,
cat. això (demonstratives), ho (clitic), prov. ço€:
(23) a. C’est impossible (Fr.)
that/it is impossible
b. Ho crec (Cat.)
3rd.neuter believe
‘I believe it’
 Ion Giurgea

Under the hypothesis that anominal pronouns lack gender, the existence of special
forms is expected: the difference between these forms and the other pronominal forms
corresponds to a difference in gender. For Spanish, Picallo (2002) already argued that
-o- pronouns are not marked for Gender.
But we may also find forms from the paradigm of one of the genders:

i. masculine accusative clitics in western Romance (except Catalan, see (23)):

(24) a. Je le sais (Fr.)

I 3rd.m know
b. Lo so (It.)
c. Lo sé (Sp.)

ii. pro in null subject Romance languages:

(25) a. Ce-i asta? pro e un cal/*El e un cal (Romanian)

what is that is a horse 3rd.m is a horse
b. pro e imposibil
is impossible.m.sg.

(26) Decidieron [pro producir aquellos documentales]i aunque proi

decided.3pl produce.inf those documentaries although
no les proporcionara nunca ningún beneficio (Picallo 2002: Note€13, (i)c)
not them provide.3sg never no benefit
‘They decided to produce those documentaries although it wouldn’t ever
provide them with any benefit’

iii. PP clitics:

(27) a. Nous y pensons (Fr.)

we to-it think
b. Ci pensiamo (It.)
c. Hi pensem (Cat.)

iv. Romanian doesn’t productively use object clitics as anominal pronouns. The femi-
nine form o appears in anominal use only with some of the verbs which take
propositional objects and states of affairs (see (28)). In most cases where Western
Romance uses a neuter object clitic, in Romanian there is no overt object at all:

(28) a. Ţi-am spus-o de mult

you.dat-have.1 told-3rd.fsg of/since much
“I told you long ago”
 Romanian genderless pronouns and parasitic gaps 

b. E, acum am făcut-o
well now have.1 done-3rd.fsg
“Well, now I/we did it”

(29) a. Nu (*o) ştiu

not (3rd.fsg) know.1sg
/sper/ (?o) cred (without a nominal antecedent for o)
/hope.1sg/(3rd.fsg) believe.1sg
b. Je ne le sais /espère/crois pas (Fr.)
i neg 3rd.msg know/hope/believe not

v. As demonstratives, Romanian uses forms identical to the feminine singular:

(30) a. Ce e aia?
what is that.fsg
“What’s that?”
b. Nu cred asta
not believe.1sg that.fsg

We may suppose that the fact that some forms with an adnominal use are identical with
forms of the paradigm of one of the genders is due to morphological underspecifica-
tion.6 The crucial evidence for this hypothesis comes from Romanian, where anominal
demonstratives and the homonymous feminine demonstratives have a different syn-
tactic behavior. First, and most importantly, singular anominal demonstratives do not
trigger feminine agreement on a predicative adjective, but masculine agreement:

(31) Asta e imposibil /*imposibilă (in anominal use)

this.fsg is impossible.msg impossible.fsg

The most likely explanation for this agreement mismatch is that the apparent mas-
culine agreement represents a morphological default, used when the controller is
unmarked for gender (Cornilescu 2000; Giurgea 2008). The idea that the masculine
singular form of adjectives is a morphological default is supported by the fact that this

.â•… This applies to all the forms listed in (i)–(v). Therefore, I analyze the apparent masculine
clitics used ‘anominally’ in Romance languages other than Romanian (ex. (24)) as genderless
forms, assuming that le/lo is a form used both for masculine and for genderless clitics. In the
Distributed Morphology formalization, one may use an underspecified vocabulary insertion
rule: [clitic, 3rd, sg.]→ le/lo.
 Ion Giurgea

form is used with clausal subjects (see (32)) and, for most adjectives, may also be used
adverbially (see (33)):7

(32) a. [A-ţi iubi duşmanii] e imposibil

to you.dat love enemies-the is impossible
“To love one’s enemies is impossible”
b. [Să-ţi iubeşti duşmanii] e imposibil
subj-you.dat love.2sg enemies-the is impossible

(33) Scrie greu /încet/ frumos

writes difficult.msg /slow.msg/ beautiful.msg
“He writes with difficulty /slowly/beautifully”

This idea is confirmed by the special behavior of the predicate “good”. When applied
to propositional objects or state of affairs, the adjective “good” has the special form
bine, which also appears as an adverb (“well”). This form has a further restriction:
it cannot appear with nominal subjects (the form bine used with nouns has a different
meaning€ – “respectable” –, normally applied to humans). This restriction cannot
be explained by semantics, because it applies even if the nominal subject refers to a
proposition or state of affairs (see (34)c). The explanation I propose is that bine lacks
gender, and an adjectival predicate must copy the gender of its subject. The only DPs
which may appear as subjects of bine are neuter pronouns (see (34)a), confirming the
hypothesis that these pronouns are genderless:

(34) a. Asta/pro e bine/*bun

this is bine/bun
“That’s good”
b. [Să-ţi iubeşti duşmanii] e bine
subj-you.dat love.2sg enemies-the is bine
“To love one’s enemies is good”
c. *Întoarcerea noastră /*Iubirea de duşmani e bine
returning-the our love-the of enemies is bine

Note moreover that neuter indefinite and quantificational pronouns may appear as
subjects of bine, confirming our proposal in Section€2 that these pronouns lack gender
(ex. (35)). When referring to concrete objects, these pronouns take bun (see (36)),
which shows that bine is not simply the genderless form of bun, but has in addition a
semantic restriction to propositional objects:

.╅ Romanian does not have a productive adverbial suffix, like western Romance -ment(e),
except in the case of the suffix -esc, which adds an -e in the adverbial use (resulting in -eşte).
 Romanian genderless pronouns and parasitic gaps 

(35) a. Ce-i mai bine? Nimic nu-i bine

what is more bine nothing not-is bine
“What’s better? Nothing is good”
b. Ceva e bine în ce-a făcut
something is bine in what has done
“There IS something good in what he did”
(36) Adu-mi ceva bun
bring.imper-me.dat something good
Another difference between feminine and anominal demonstratives is that while the
former take the differential object marker (pe) even if they refer to objects, in case of
noun ellipsis,8 the latter never take pe:
(37) a. Ia(-l pe) ăsta ! (e.g. pantof “shoe” – masculine)
take(3rd.m obj) this.msg
“Take this one!”
b. Ia(-o pe) asta ! (e.g. pălărie “hat” – feminine)
take(3rd.f obj) this.fsg
“Take this one!”
c. Ia asta! (without a nominal antecedent)
take this
d. N-am spus(*-o pe) asta (without a nominal antecedent)
not-have.1sg said(3rd.fsg-obj) this
“I didn’t say this!”

Another peculiarity of anominal demonstratives is that they are never clitic-doubled

when fronted (as noticed by Cornilescu (2000)). They are in fact the only instance of a
definite DP which is not clitic-doubled when fronted – in Romanian, clitic doubling is
obligatory with definites and partitive indefinites, whether they are topics or foci:

(38) a. Asta aşteptam!

this waited.1sg
“That’s what I was waiting for”
b. Ocazia asta *(o) aşteptam!
opportunity-the this 3rd.fsg.cl.acc waited.1sg
“That’s the opportunity I’ve been waiting for”

The hypotheses in Section€2 provide a straightforward explanation for this behavior:

anominal demonstratives lack gender, while accusative clitics always spell-out gender.

.╅ pe is impossible with inanimates with an overt noun. With ellipsis, absence of pe is mar-
ginally possible with inanimates, and obligatory with animates.
 Ion Giurgea

An accusative bare index with unvalued gender will have a null spell-out. If we assume
that bare indices are the same thing as clitics or represent a pro associated with a clitic,
the null object found with genderless antecedents indicate that a genderless clitic has
a null spell-out. By recognizing the existence of genderless clitics with a null spell-out,
we may keep the generalization that definite and partitive indefinites are clitic-doubled
when fronted in Romanian: anominal demonstratives are not an exception, but are
clitic-doubled by a null clitic.
Note that anominal neuters can be doubled by the feminine clitic o exactly with
those verbs which allow a feminine clitic denoting a state of affairs or proposition:

(39) Asta n-am făcut-o/spus-o

this not-have.1sg done-3rd.fsg/said-3rd.fsg

This seems to suggest that these verbs allow an anominal pronoun marked as feminine.
The fact that anominal demonstratives do not allow feminine adjectives (except in the
affective idiomatic expression asta-i bună lit. “that’s good.fsg”, meaning “I can’t believe
that!”) (see (31)) can be explained by assuming that the anominal interpretation of
feminines can only be licensed by the verb (perhaps via a sort of contextual recovery
of a null N), so that feminines in an anominal use are only possible in the object posi-
tion of certain verbs. In the absence of the licensing verb, the anominal interpretation
is only possible with genderless pronouns, therefore singular anominal demonstratives
cannot trigger feminine agreement on predicative adjectives (except in the aforemen-
tioned expression, where the same contextual recovery of an N can be invoked).9

.â•… In the plural, Romanian allows a null N with the interpretation −animate – e.g. multe
‘many.fpl’ = ‘many things’, altele ‘other things’, toate ‘everything’, cele ce… ‘the.fpl that…’ =
‘the things that’ etc. (see Giurgea 2008 for discussion). As expected, this N can also combine
with demonstratives, giving the impression of plural forms of anominal pronouns – astea
‘these (things)’, alea ‘those (things)’. Since no nominal content is recovered by ellipsis and the
meaning is −animate, these forms may indeed have what I called ‘anominal use’. Note however
that in this case the interpretation comes from the properties of the feminine plural null N
and not from the absence of gender (the existence of this null N is shown by the combination
with adnominal determiners and modifiers, e.g. cele din cer şi de pe pământ ‘the.fpl of-in sky
and of on earth’ = ‘the things in the sky and on the earth’). Therefore we predict clitic doubling
to be possible, and indeed these DPs are doubled by feminine plural clitics when the condi-
tions for doubling are fulfilled:
(i) Toate le ştie
all.fpl 3rd.fpl knows
‘(S)he knows everything’

(ii) Astea le uitase

these.fpl 3rd.fpl had forgotten
‘These things, (s)he had forgotten’
 Romanian genderless pronouns and parasitic gaps 

Another peculiarity of genderless pronouns appears in relative clauses (Al. Grosu,

p.c.). While DPs containing a (lexical or elliptical) N as well as animate pronouns only
allow the care strategy of object relativization in contemporary Romanian, a strategy
which involves obligatory clitic doubling, neuter indefinite and quantificational pro-
nouns only resort to the ce- strategy, which allows lack of clitic doubling:
(40) a. o carte [pe care am cumpărat-o la târg]
a book(f) obj which have.1sg bought-3rd.f at market
“a book I bought at the market”
b. *o carte [ce-am cumpărat la târg]
a book what have.1sg bought at market
(41) a. ceva [ce am cumpărat la târg]
something what have.1sg bought at market
“something I bought at the market”
b. *ceva [pe care l-am cumpărat la târg]
something obj which 3rd.m-have.1sg bought…

The most likely explanation of this contrast is that pe- marking requires the presence of
gender, as can also be inferred from the absence of pe- marking on anominal demon-
stratives, shown in (37) above. The problem in (41)b does not come from care: when ce
relativization cannot be used, like in PP-relativization, we may find care with a neuter
pronoun. We may also find it in subject position, although in this case ce seems to be
more frequent:10
(42) a. ceva cu care să scriu
something with which subj write.1sg
“something I can write with”
b. ceva ce/care s-a stricat
something what/which refl-has broken
“something which got broken”

.â•… A similar phenomenon has been used as an argument for the idea that ‘neuter pronouns’
are unmarked for gender by Picallo (2002), for Spanish. She notes that the interrogative cuál
‘which’ is compatible only with masculine or feminine nominals, but not with neuter pro-
nouns or sentences (in this case, only the neuter interrogative qué ‘what’ is allowed):
(i) *Cuál quieres, esto o aquello? (Picallo 2002:€22)
which want.2sg this.neut or that.neut

(ii) *Cuál prefieres, salir a dar una vuelta o quedarte en casa

which prefer.2sg go-out.inf for a walk or remain.inf-refl at home
‘What do you prefer, to go out for a walk or to stay home?’

She explains this contrast by assuming that cuál is always marked for gender.
 Ion Giurgea

To conclude, we have argued that definite anominal pronouns (i.e. definite pronouns
referring to entities which do not fall under a nominal concept), as well as indefinite
and quantificational neuter pronouns (which may also be qualified as “anominal”) are
genderless in Romance languages.

4.â•… Consequences for the analysis of Romanian “neuter” nouns

Romanian has a large class of nouns, traditionally called “neuter”,11 which trigger mas-
culine agreement in the singular and feminine agreement in the plural:
(43) un scaun/două scaune
a.m chair/two.f chairs

Pronouns anaphoric to these nouns also have masculine forms in the singular and
feminine forms in the plural.
In traditional grammar, these nouns have either been analyzed as ambigeneric
(masculine in the singular and feminine in the plural) or as constituting a third gender,
the neuter.12

.╅ This class continues the Latin neuter (although some Latin words have changed their
gender class during the evolution leading to Romanian, it has been shown that most inherited
Latin neuters have been included in the Romanian ‘neuter’ class (86% according to Windisch
1973), while inherited masculine inanimates have become ‘neuter’ in a percentage of around
50% (see the counting in Windisch 1973), and almost a half of them kept their gender; the
endings characterizing the Romanian ‘neuter’ continue the forms of the Latin neuter – see
especially the plural -uri, old Romanian -ure <lat. -ora –, with the adaptation of the plural
endings to the feminine; this adaptation already appeared in late Latin in agreeing forms,
as witnessed by inscriptions and by the Italian pairs of the type uovo/uova, braccio/braccia
etc.) Romanian neuter nouns almost entirely refer to inanimates. There are only a handful
of exceptions – class names such as animal ‘animal’, mamifer ‘mammal’, and words with a
secondary meaning applied to animate beings – model ‘model’.
.â•… For the first view (the ‘ambigeneric’ analysis), see, a.o., Jacotă (1905), Bazell (1937), Hořejši
(1957), Nandris (1961), Manoliu-Manea (1971). For the second (the ‘neuter gender’ analysis),
see, a.o., Graur (1937), GLR (1963), GALR (2005).
The only novel analysis in the generative framework I know about is Farkas’s (1990),
according to which neuter nouns are unmarked for gender, and the masculine singular and
feminine plural are default forms on targets of agreement. This analysis is at odds with the
proposals made in this paper: if the special behavior of neuter pronouns with respect to
pronominal resumption and doubling is due to the fact that they are unmarked for gender,
neuter nouns, which behave like all the other nouns and not like neuter pronouns, cannot be
said to be unmarked for gender. For a possible reconciliation between Farkas’s analysis and the
facts concerning the pronominal system, see Croitor&Giurgea (2009)
 Romanian genderless pronouns and parasitic gaps 

The facts presented in this paper provide evidence against the neuter gender
analysis€– more precisely, against the idea that pronouns have three values for the cateÂ�
gory neuter: We have seen that languages which have a neuter gender use this gender
for reference to entities which do not fall under a nominal concept (the “anominal
use”; see Section€3 above, (18)–(19)). If Romanian pronouns had three genders, with
the neuter having forms homonymous with the masculine in the singular and with the
feminine in the plural, we would have expected to find masculine singular forms used
for uncategorized perceptual objects and propositional objects. But, as we have seen,
we find either null pronouns or the genderless demonstratives, formally identical with
the feminine, and with some verbs the feminine clitic o.
A thorough evaluation of the other possible analyses of Romanian neuter nouns
goes beyond the scope of this paper. Here, I would only like to point out two alterna-
tive solutions, which are discussed at length in Croitor & Giurgea (2009): (i) Using
Corbett’s (1991) distinction between controller genders (or agreement classes) and target
genders, it can be assumed that Gender is introduced on the Number head (as proposed
by Ritter (1993), for Romance languages) and that this head selects for noun classes –
Number /−pl −fem/ selects class I and class III nouns, Number /−pl +fem/ selects class
II nouns, Number /+pl −fem/ selects class I nouns, Number /+pl +fem/ selects class II
and class III nouns –; (ii) if one adopts an analysis in which number is not introduced
by a functional head, but rather is marked in the lexical entries of nouns as a feature
with optional values, “neuter” nouns may be described as marked for choosing a value
for gender depending on the value chosen for number – more precisely, they would
choose one pair from the set of pairs {〈gender = masc., number€=€sg.〉, 〈gender = fem.,
number = pl.〉}.13
To conclude, the fact that Romanian uses genderless pronouns in the same condi-
tions as the other Romance languages supports the idea that Romanian has only two
values for the category Gender – masculine and feminine.

References

Alboiu, G. 2002. The Features of Movement in Romanian. Editura Universităţii Bucureşti,

Bucharest.
Aquaviva, P. 2008. Lexical Plurals. A Morphosemantic Approach. Oxford University Press.
Bazell, Ch. E. 1937. The Rumanian Neuter. Neuphilologische Mitteilungen 38, 273–275.

.â•… These analyses express the intuition behind the term ‘ambigeneric’ without positing two
lexical entries for each ambigeneric noun (one for the singular, marked as masculine, and one
for the plural, marked as feminine). Evidence against assuming lexical pairs comes from ellipsis
and from the productivity of the type (see Aquaviva (2008) and Croitor & Giurgea (2009)).
 Ion Giurgea

Corbett, G.1991. Gender. Cambridge University Press.

Corblin, F. 1995. Les formes de reprise dans le discours. Presses Universitaires de Rennes.
Cornilescu, A. 2000. Notes on the Interpretation of the Prepositional Accusative in Romanian.
Bucharest Working Papers in Linguistics, 91–106.
Cornilescu, A. 2002. Rhematic Focus at the Left Periphery, the Case of Romanian. Romance Lin-
guistics and Linguistic Theory 2000, ed. by Claire Beyssade et al. John Benjamins, 77–92.
Croitor, B. & I. Giurgea (2009) On the so-called Romanian “Neuter”. University of Bucharest
Review. Bucharest Working Papers in Linguistics, 11, 2, 21–39.
Dobrovie-Sorin, C. 1990. Clitic Doubling, Wh-movement, and Quantification in Romanian.
Linguistic Inquiry 21: 351–397.
Dobrovie-Sorin, C. 1994. The Syntax of Romanian. Berlin, Mouton de Gruyter.
Elbourne, P. 2005. Situations and Individuals. Cambridge, Mass., London: MIT Press.
Farkas, D. 1990. Two Cases of Underspecification in Morphology. Linguistic Inquiry 21, 539–550.
GALR. 2005. Gramatica limbii române. Editura Academiei Române, Bucharest.
Giannakidou, A. & J. Merchant. 1996. On the Interpretation of Null Indefinite Objects in
Modern Greek. Studies in Greek Linguistics. Proceedings of the 10th Annual Meeting of
the Department of Linguistics, Faculty of Philosophy, Aristotle Univeristy of Thessaloniki.
Thessaloniki: Kyriakidis.
Giurgea, I. 2008. Recherches sur la structure interne des pronoms et des expressions nominales
sans nom exprimé. Ph.D. Thesis, University of Paris 7.
GLR.1963. Gramatica Limbii Române, Editura Academiei RSR, Bucharest.
Graur, A.1937. Sur le genre neutre en roumain. Bulletin linguistique 5, 5–11.
Hall Jr., R. 1965. The “Neuter” in Romance: A Pseudo-problem. Word XXI, 421–427.
Hankamer, J. & I. Sag. 1976. Deep and Surface Anaphora. Linguistic Inquiry 7: 3, 391–428.
Hořejši, V. 1957. Problema substantivelor aşa-zise “neutre” în limba romînă în lumina legăturilor
cu alte limbi. Studii şi cercetări de lingvistice, 8, 415–429.
Jacotă, M. 1905. Încercări asupra eterogenului român. Arhiva (Iaşi) XVI, 282–286.
Kratzer, A. 1998. More Structural Analogies Between Pronouns and Tenses. Proceedings of SALT
VIII, 92–109, ed. by Strolovitch, D. & A. Lawson. Ithaca: Cornell University.
Kratzer, A. 2006. Minimal Pronouns. Fake Indexicals as Windows into the Properties of Bound
Variable Pronouns. Ms. University of Massachusetts at Amherst.
Longobardi, G. 1994. Reference and Proper Names: A Theory of N-Movement in Syntax and
Logical Form. Linguistic Inquiry, 25, 4, 609–665.
Manoliu-Manea, M. 1971. Gramatica comparată a limbilor romanice. Bucharest, Editura
Didactică şi Pedagogică.
Meyer-Lübke, W. 1890–1906. Grammatik der romanischen Sprachen. Leipzig.
Nandris, O. 1961. Le genre, ses réalisations et le genre personnel en roumain. Revue de Linguistique
Romane 25, 47–74.
Rizzi, L. 1986. Null Objects and the Theory of pro. Linguistic Inquiry 17, 501–557.
Panagiotidis, P. 2002. Pronouns, Clitics and Empty Nouns, John Benjamins, Amsterdam,
Philadelphia.
Picallo, C. 2002. Abstract Agreement and Clausal Arguments. Syntax 5: 2, 116–147.
Ritter, E. 1993. Where’s Gender? Linguistic Inquiry 24, 795–803.
Windisch, R. 1973. Genusprobleme im Romanischen. Das Neutrum im Rumänischen. Tübinger
Beiträge zur Lin
 To agree or not to agree
Beyond quintessentially syntactic
agreement in Spanish*

Julio Villa-García
University of Connecticut

Non-Caribbean Spanish is a prototypical null-subject language. Yet,

verbs normally enter into an agreement relationship with the (overt/null)
subject. In some cases, however, certain acceptable sentences contain an
anomalously-agreeing subject in terms of one of the three agreement/φ-features,
namely person, number, and gender. Assuming that the nominals in question
establish an Agree relationship with the verb à la Chomsky (2000, 2001,
2004, 2008), I submit that in Spanish there is an operative condition on the
output of Agree, which posits that if full/optimal agreement between the verb
and the subject does not obtain, at most one feature can be left unchecked/
unvalued syntactically. The remaining unvalued feature of T0 is deleted/valued
for convergence through alternative agreement mechanisms available in the
grammar, including default agreement and possibly semantic/pragmatic
agreement. From this, it follows that agreement in Spanish may not be a
quintessentially syntactic phenomenon handled exclusively by a probe-goal
system like Chomsky’s Agree.
Keywords:╇ Agree; subjects; disagreement; extra-grammatical agreement; Spanish

*This paper has benefited enormously from personal communications with Martin Atkinson,
Violeta Demonte, Ángel Jiménez, Irina Monich, Andrew Radford, Juan Romero, and William
Snyder. I would particularly like to thank Jonathan David Bobaljik, Željko Bošković, Jairo
Nunes, Ana Ojea, Antxon Olarrea, Susi Wurmbrand, and two anonymous reviewers for helpful
comments and suggestions. I am also grateful to the audiences at UConn Ling Lunch and
LSRL-39, especially, John Beavers, Mary Kato, Mario Montalbetti, and Karen Zagona. I am also
thankful to my consultants, Marc Artiga Galindo, Elena Castellari, Jhon Fredy Díaz Arévalo,
Francisco José Fernández-Rubiera, Rosi Ana Gregis, Ioana Hanciu, Anna Sala Bonada, Viviana
Maggioni, José Riqueros-Morante, Johanna Rivera, and Maria Stournara, for their judgments
and examples. This research has been funded by a full graduate fellowship awarded to the
author by the savings bank “la Caixa” (Barcelona, Spain). The usual disclaimers apply.
 Julio Villa-García

1.â•… Introduction

Non-Caribbean Spanish (henceforth, Spanish) is a paradigmatic null-subject language

characterized by the possibility of unpronounced subjects and the property of “free
subject inversion.” Still, subjects normally establish an agreement relationship with the
verbal predicate, as evidenced by the morphological endings of the verb, which gener-
ally display person and number agreement:
(1) a. María llegó
Mary-3.sg arrive-pst-3.sg
“Mary arrived.”
b. *María llegaron
Mary-3.sg arrive-pst-3.pl

However, there exist certain cases when full subject-agreement (i.e. matching in terms
of all agreement phi-/φ-features) does not obtain, namely with regard to person, as in
(2a), or number, as in (2b):
(2) a. Los estudiantes llegamos
the students-3.sg arrive-pst-1.pl
“We students arrived.”
b. ?Me llamó mientras estaba en la playa mis primos
cl. call-pst-3.sg while be-pst-1.sg on the beach my cousins-3.pl
“My cousin called me while I was on the beach.”

In parallel fashion, null-subject sentences show that it may be necessary to resort to

extra-grammatical strategies (e.g. the discourse-pragmatic context) to determine the
gender of the silent subject, as shown by the English paraphrases of sentences like the
following:
(3) Llegó
arrive-pst-3.sg
“S/he arrived.”

Set against this background, the paper aims to discuss some murky issues surrounding the
relationship that Spanish null and lexical subjects establish with the grammatical subject
expressed through the inflectional endings of the verb. In particular, this paper attempts
to explore the boundaries of “disagreement” in Spanish, with an eye on hinting at the
ways in which this phenomenon is constrained. The general framework adopted here
is the Minimalist Program for Linguistic Theory, and more specifically Chomsky’s
(2000, 2001, 2004, 2008) Agree system.
The paper is organized as follows: Section 2 provides a cursory look at subject-verb
agreement mismatches in Spanish as well as the identification of the gender feature
of silent subjects in finite contexts; Section 3 focuses on the challenges raised by the
Spanish data for the Agree system; Section 4 proposes a generalization aimed at
 To agree or not to agree 

capturing the Spanish facts and discusses possibilities for future research; Section 5
offers some concluding remarks.

2.â•… Discordant subjects in Spanish

As mentioned in the introduction, in Spanish the verb normally agrees with the subject,
as shown by the grammaticality contrasts in (4) (see also (1)):
(4) a. Llamó la chica
call-pst-3.sg the girl-3.sg
“The girl called.”
b. *Llamaste la chica
call-pst-2.sg the girl-3.sg
c. Esther compró un perro ayer
Esther-3.sg buy-pst-3.sg a dog yesterday
“Esther bought a dog yesterday.”
d. *Esther compramos un perro ayer
Esther-3.sg buy-pst-1.pl a dog yesterday
Abstracting away from the debate over the grammatical or left-peripheral status of
explicit subjects in Spanish, I will pursue the hypothesis that subjects enter into an
Agree relationship with the verb à la Chomsky (2000 et seq.). Nonetheless, Spanish
allows for a certain degree of “disagreement,” an issue to which I turn immediately.

2.1â•… Person
In Spanish, certain plural subjects trigger agreement in first or second person when
they refer to a group which includes the first (5b) or the second person (5c), as indicated
by verbal inflection (cf. Hurtado 1984; Fernández-Soriano 1989, among others):
(5) a. Los ciudadanos están hartos de tanto papeleo
the citizens-3.pl be-prs-3.pl fed up of so much paperwork
“The citizens are fed up with so much paperwork.”
b. Los ciudadanos estamos hartos de tanto papeleo
the citizens-3.pl be-prs-1.pl fed up of so much paperwork
“We citizens are fed up with so much paperwork.”
c. Los ciudadanos estáis hartos de tanto papeleo1
the citizens-3.pl be-prs-2.pl fed up of so much paperwork
“You citizens are fed up with so much paperwork.”

.╅ The reader should note that this is present only in varieties like Peninsular Spanish, which
employ the second-person plural form vosotros (‘you’).
 Julio Villa-García

In languages like English, the meaning expressed by (5b) and (5c) is conveyed by
employing the first- and second-person plural pronouns as determiners. In this
paper, I will assume that plural NPs such as ciudadanos in (5) are inherently third-
person plural, following Olarrea (1996) and Camacho (2006).2 Thus, in Spanish there
seems to be a mismatch in terms of the feature person in the structures at issue. Note
that the pattern in (5) is not confined to (left-dislocated) preverbal subjects (see also
Ortega-Santos 2005):

(6) a. El verano es lo que los estudiantes llamábamos “vida” [SVO]

the summer is cl. what the students call-pst-1.sg life
b. El verano es lo que llamábamos los estudiantes “vida” [VSO]
the summer is cl. what call-pst-1.sg the students life
c. El verano es lo que llamábamos “vida” los estudiantes [VOS]
the summer is cl. what call-pst-1.sg life the students
All: “The summer is what we students used to call “life.””

Prima facie, it is tempting to assume that the construction illustrated in (5) is a feature
of null-subject languages of the Spanish type. However, other Romance pro-drop lan-
guages like Italian, Portuguese, and Romanian, alongside non-Romance null-subject
languages like Serbo-Croatian, do not manifest the structure in (5); they behave like
English in this regard:

(7) a. Noi studenti sappiamo ciò che vogliamo (Italian)

we students know-prs-1.pl this that want-prs-1.pl
“We students know what we want.”
b. Voi studenti sapete ciò che volete
you students know-prs-2.pl this that want-prs-2.pl
“You students know what you want.”
c. Gli studenti sanno ciò che vogliono
the students know-prs-3.pl this that want-prs-3.pl
“The students know what they want.”

However, sentences like (7a) and (7b) are not acceptable in Spanish:

(8) a. *Nosotros estudiantes sabemos lo que queremos

we students know-prs-1.pl cl. that want-prs-1.pl
“We students know what we want.”

.â•… See Corbett (2006) for the proposal that nominals are endowed with φ-features. On this
view, ciudadanos in (5b) would bear the feature first-person plural. However, as mentioned in
the main text, I will not entertain this possibility here and instead pursue the hypothesis that
plural nominals are intrinsically third-person plural.
 To agree or not to agree 

b. *Vosotros estudiantes sabéis lo que queréis

you students know-prs-2.pl cl. that want-prs-2.pl
“You students know what you want.”

In Spanish, however, it is possible to have a structure akin to an appositive on condition

that the definite article precedes the noun estudiantes:3

(9) a. Nosotros los estudiantes sabemos lo que queremos

we, the students know-prs-1.pl the that want-prs-1.pl
“We students know what we want.”
b. Vosotros los estudiantes sabéis lo que queréis
you the students know-prs-2.pl the that know-prs-2.pl
“You students know what you want.”
c. Ellos los estudiantes saben lo que quieren
they the student know-prs-3.pl the that want-prs-3.pl
“They, the students, know what they want.”

It has been suggested in the literature that constructions like those in (5b) and (5c)
may be explained by postulating a pro + appositive analysis, the empty category pro
being responsible for establishing an Agree relationship with the verb (much in the
spirit of Ortega-Santos 2008, among others). However, a question which immedi-
ately arises is what the relationship between pro and the appositive is. Likewise, if
pro is a pronoun merely lacking phonological content, we expect sentences like (10)
to be grammatical in languages which behave like English in the relevant respects,
contrary to fact:

(10) *They students know what they want (cf. They, the students, know…)

I take the ungrammaticality of sentences like (10) to provide prima facie evidence
against a [pro + appositive] analysis.4 Similarly, Spanish plural nouns like ciudadanos

.╅ This pattern seems to require the overt pronominals to be emphatic or contrastive, at
least in Peninsular Spanish. Note that these sentences sound more natural if the sequence los
estudiantes appears between commas.
.╅ As pointed out by an anonymous reviewer, however, substandard varieties of English
exhibit constructions where a third-person plural pronoun is allowed in this context, as in
Them doctors didn’t treat you properly. Speakers of these varieties – and speakers of main-
stream English – would use the form us instead of we as in Us students, though, suggesting
that the pronominal gap could be treated in terms of clitic (I, we, they…) vs. nonclicit (me, us,
them…) subject pronouns within a still unstable set of competing pronominal forms. I leave
this task for future research.
 Julio Villa-García

in (5) can be preceded not only by the definite article, but also by quantifiers, in parallel
fashion to regular (i.e. non-discordant-person) subjects:

(11) Muchos estudiantes no sabéis lo que queréis

many students-3.pl not know-prs-2.pl that that want-prs-2.pl
“Many students (including you) don’t know what they want.”

This constitutes solid evidence against an appositive analysis, since the equivalent
construction in English is not possible with quantifiers in any dialect:

(12) *You many students don’t know what you want

Note that although the pattern in (5) is a language-specific characteristic of Spanish, it

is not limited to this language. In fact, other Romance and non-Romance null-subject
languages such as Asturian, Catalan, and Greek, display an identical construction, as
the following examples from Asturian illustrate:

(13) a. Los llingüistas nun tienen que falar munches llingües

the linguists-3.pl not have-prs-3.pl that speak-inf many languages
“Linguists don’t have to speak many languages.”
b. Los llingüistas nun tenemos que falar munches llingües
the linguists-3.pl not have-prs-1.pl that speak-inf many languages
“We linguists don’t have to speak many languages.”
c. Los llingüistas nun tenéis que falar munches llingües
the linguists-3.pl not have-prs-2.pl that speak-inf many languages
“You linguists don’t have to speak many languages.”

Nonetheless, discordant person subjects are not a syntactic property of the Null Subject
Parameter, as has been seen. The brief cross-linguistic survey summarized above sug-
gests that there is a one-way implication, though: if the language has discordant person
subjects, then it is a [+null subject] language.

2.2â•… Number
In certain dialects of Peninsular and Latin American Spanish, cases of number disagree-
ment are attested in which a singular verb appears with a plural postverbal subject,
suggesting some sort of default agreement operation, as illustrated in (14).5

.╅ This pattern is by no means confined to Spanish, and in fact it is part of a well-documented
phenomenon across the languages of the world (cf. Greenberg’s 1966 Universal 33). In English,
this situation can be replicated in the case of existential sentences like There’s people who care
about you (cf. There are people who…). For further discussion of the [there’s + PL] construction
 To agree or not to agree 

(14) a. Me falta varias piezas del puzzle

cl. lack-prs-3.sg several pieces-3.pl of + the puzzle
“I am missing several pieces of the puzzle.”
cf. Me faltan-prs-3.pl varias piezas del puzle.
b. Se me cae los lagrimones
cl. cl. fall-prs-3.sg the tears + huge-3.pl
“I cry a lot.”
cf. Se me caen-prs-3.pl los lagrimones.
On certain occasions, mainly when a postverbal subject has been separated from the
verb by a number of constituents, sentences like the following – albeit marginal – are
frequently found in the casual speech of native speakers of Peninsular Spanish:
(15) a. ?Compró un coche en esa zona de Madrid el otro día
buy-pst-3.sg a car in that area of Madrid the other day
mis primos
my cousins-3.pl
“The other day my cousins bought a car in that area of Madrid.”
cf. Compraron-pst-3.pl un coche en esa zona de Madrid el otro día mis primos.
b. ?Me llamó el otro día mientras estaba en la
cl. call-pst-3.sg the other day while be-pst-1.sg on the
playa tomando el sol con un calor terrible mis hermanas
beach taking the sun with a heat horrible my sisters-3.pl
“My sisters called me while I was on the beach sunbathing the other day,
and it was so hot.”
cf. Me llamaron-pst-3.pl el otro día mientras estaba en la playa tomando el
sol con un calor terrible mis hermanas.
The sentences in (14) seem to be clear instances of default (i.e. third-person singular)
agreement, (15) possibly being due to processing limitations in real time. A priori, these
could be sentence planning errors, or afterthoughts, as argued by Villalba (2000).
The opposite situation (i.e. singular subject but plural verb) occurs in cases of
questions with singular quién “who” (intead of plural quiénes) and third-person plural
verbal morphology. This pattern, illustrated in (16), seems to be on the rise in North-
ern Peninsular Spanish.
(16) ¿Quién vinieron?
who-sgv come-pst-3.pl
“Who came?”
cf. ¿Quiénes vinieron?

in English, see Sobin (1997). On agreement patterns displayed by existential constructions in

Spanish, whose nominal displays object-like properties, see Rodríguez-Mondoñedo (2007).
 Julio Villa-García

Examples like (16) cannot readily be accounted for by the processing hypothesis,
and they pose serious difficulties for a default agreement (i.e. third-person-singular)
analysis.6,7

2.3â•… Where is the gender of null subjects?

It is well known that Spanish is a language where adjectives, demonstratives, determin-
ers, nouns, clitic and strong personal pronouns, interrogative and relative pronouns,
quantifiers, reflexives and reciprocals, along with absolutive past and passive par-
ticiples, have (masculine or feminine) gender (Zagona 2002), as illustrated by the
following example (where reference to person and tense has been omitted for the
sake of exposition):
(17) Las niñas de esta aldea son altas
the-fem.pl children-fem.pl of this-fem.sg village-fem.sg are tall-fem.pl
“The girls of this village are tall.”

Yet, by looking at the English paraphrases of the following sentences involving null
subjects, it is easy to observe that the status of the gender feature (as opposed to
person and number) is not clear, since in Spanish the feature gender does not have a
morphological reflex realized as a verbal ending:
(18) a. Ha llamado
have-prs.prf-3.sg phone-prt
“S/he has called.”
b. Quiere que vengan
want-prs-3.sg that come-prs-3.pl
“S/he wants them-masc/fem(?) to come.”

The above amounts to saying that the inflected verb itself does not express any informa-
tion regarding the gender of the silent subject. Nevertheless, I will assume that the
head T0 in Spanish is endowed with person, number, and gender φ-features, the latter
being present on an abstract level without a concomitant morphological reflex. This is
partially motivated by the fact that nominals in Spanish bear various features, among
which three are uncontroversial φ-features, namely person, number, and gender.

.╅ See Camacho (2003) and Ortega-Santos (2005) for cases of first-conjunct agreement with
postverbal subjects.
.╅ An important issue left for future research is whether the appearance of number disagreement
in certain dialects suggests default operations, as noted by an anonymous reviewer. In this
connection, a logical question to pose is whether preverbal subjects in Caribbean Spanish can
be informative as to default operations in non-Caribbean Spanish.
 To agree or not to agree 

Furthermore, features in recent conceptions of morphology (cf. Distributed Morphology)

are presumed to form part of feature bundles, so it is far from implausible that Spanish
T0 carries more than two φ-features. A note of caution is in order at this point, though.
Since gender is never expressed morphologically in the verb, we cannot refer to cases
like (18) as instances of “disagreement” per se. The only claim that can be made on the
basis of (18) is that the recovery of the gender feature of null subjects cannot be the
result of a strictly grammatical operation such as Agree, at least in its strong form, a
matter to which I turn immediately.
In relation to sentences like those in (18), Andrew Radford (p.c.) points out that
two hypotheses can be formulated regarding gender: on the one hand, it could be
argued that the fact that the gender of the subject is not expressed in the sentence
leads to ambiguity, which presupposes that the feature gender is present, but has a
null spell-out. On the other hand, it could be contended that the situation in (18) is
associated with vagueness (i.e. the feature is not present in the syntax, and is derived
via pragmatic inferencing; that is, it is the context that helps determine the gender of
the “missing” pronoun). The second hypothesis raises the question of whether finite
null subjects (i.e. instances of pro) bear exactly the same features as a non-null pro-
nominal. Juan Romero (p.c.) disagrees with the hypotheses above in that the features
of a nominal (but not the features of T0) are inherent and interpretable in the Mini-
malist approach. Thus, if agreement determined the features of the null subject,
we would be introducing new elements in the derivation, in violation of the Inclu-
siveness Condition (Chomsky, 2001: 2). This observation is similar to Holmberg’s
(2005) claim that Rizzi’s (1982) classical analysis of pro-drop is problematic in light
of the Minimalist assumption that agreement features are uninterpretable and thus
incapable of recovering the content of pro. Providing a solution to this problem is
far beyond the scope of this paper, though for the purposes of this article, I will
assume that the gender features of the null subject and of the head T0 are not dealt
with by means of the Agree mechanism.

3.â•… Agree revisited

Under the Agree system of Chomsky (2000, 2001, 2004, 2008), T0 and v0 have a set of
uninterpretable agreement φ-features, namely person, number, and, in some languages,
gender, which are unvalued, though see fn. 8. The operation of Agree(ment) “establishes
a relation (agreement, Case checking) between an LI [lexical item] α and a feature F”
(Chomsky, 2000: 101). These features are morphosyntactic features which are not
interpretable at LF. Chomsky has also suggested that these features are often not realized
visibly at PF (e.g. gender in Spanish verbs), flagrantly violating the Interpretability
Condition, which posits that “the language faculty, including lexical items, operates
 Julio Villa-García

exclusively with features that are interpretable at the interfaces” (Chomsky, 2000: 27).
Uninterpretable features (intimately associated with inflectional morphology) must
then be deleted in the course of the derivation. Traditionally, verbal endings have been
said to agree in person and number with subjects, as noted above. In order to abandon
spec-head accounts of agreement, Chomsky introduces the operation Agree, which he
defines as “the erasure of uninterpretable features of probe and goal” (2000: 37). An
active probe (i.e. one that contains an uninterpretable feature, e.g. T0) can agree with
an active goal (i.e. one comprising an uninterpretable feature, e.g. a noun) provided
that both probe and goal are related (i.e. as long as a matching condition is met, and
match should be taken to be “identity,” as argued by Chomsky, 2001: 4). As Chom-
sky (2000: 39) emphasizes, feature identity must be understood as “choice of feature,
not of value.” Thus, an interpretable feature [2Person] is identical to an uninterpretable/
unvalued feature [uPerson]. On the basis of Chomsky’s (2000) reflections on certain fea-
tures of verbs and nouns being interpretable and other features being uninterpretable
(and for this reason requiring deletion in the course of the derivation, as has been
seen), the following assumptions can be made:

–â•fi Agreement (φ-) features of nominals are interpretable at LF8

–â•fi Agreement (φ-) features of verbs are uninterpretable at LF
–â•fi Inherent Case features of nominals are interpretable at LF
–â•fi Structural Case features of nominals are uninterpretable at LF

In this paper I will not explore the issue of Case in any systematic fashion, the focus
being on φ-features, although it is important to keep in mind that both agreement
and Case are closely related in the Agree framework. In order to illustrate the opera-
tion Agree, let us consider a sentence where full/optimal agreement obtains (that is,
when the three φ-features presumably borne by T0 are valued and deleted through the
mechanism of Agree):
(19) Llegó el cura
arrive-pst-3.sg the priest-masc-3.sg
“The priest arrived.”

.╅ There is a potential problem regarding the gender feature in inanimate nominals like
mesa (‘table’), which lack natural gender, when the inanimate nouns in question have been
replaced by unpronounced subjects. In these cases, the gender feature is a mere grammatical
marker, and therefore an uninterpretable feature (by virtue of lacking semantic content). A
way of circumventing this problem would be to adopt a system along the lines of Pesetsky &
Torrego (2007) and Bošković (to appear). The latter author assumes that grammatical gender
without semantic import is a (lexically) valued uninterpretable feature, unlike semantically
contentful gender, which is an interpretable feature.
 To agree or not to agree 

In (19), when T0 enters the derivation via the operation of Merge, it carries the unin-
terpretable φ-features [uPerson], [uNumber], and [uGender], making it active for
searching and agreeing with a goal. The nominal el cura, for its part, contains an unvalued
structural Case feature [uCase] (or [uK]) alongside three interpretable features, namely
[3Person], [SgNumber], and [MascGender]. Because the features of the goal match
the pertinent features of the probe T0 (i.e. [uPerson], [uNumber]), and [uGender]),
el cura is an appropriate goal. The operation Agree deletes the uninterpretable features
of both probe and goal and values them. Thus, Agree for the pair (T0, el cura) can be
represented as follows:9
(20) Agree (T0, el cura) = (T0, el cura)
[uNumber] [SgNumber] [uNumber] [SgNumber]
[uPerson] [3Person] [uPerson] [3Person]
[uGender] [MascGender] [uGender] [MascGender]
[uCase] [SgNumber] [uCase]
[3Person] [NomCase]
[MascGender]

From this, it follows that “verbs agree with nouns, not conversely” (Chomsky, 2000: 124).
Let us now return to the cases of anomalous agreement discussed in the previous
section, some of which are repeated here for the sake of illustration.
(21) Los ciudadanos estamos hartos de tanto papeleo (=(5b))
Agree (T0, los ciudadanos) = (T0, los ciudadanos)
[uNumber] [PlNumber] [uNumber] [PlNumber]
[uPerson] [3Person] [uPerson] [3Person]
[uGender] [MascGender] [uGend.] [MascGend.]
[uCase] [PlNumber] [uCase]
=> [?Person] [NomCase]
[MascGender]

(22) Se me cae los lagrimones (=(14b))

Agree (T0, los lagrimones) = (T0, los lagrimones)
[uNumber] [PlNumber] [uNumber] [PlNumber]
[uPerson] [3Person] [uPerson] [3Person]
[uGender] [MascGender] [uGender] [MascGender]
[uCase] => [?Number] [uCase]
[3Person] [NomCase]
[MascGender]

.╅ The typographical convention used throughout is thus: strikethrough signals deleted features;
bold indicates valued features.
 Julio Villa-García

(23) Ha llamado (=(18a))

Agree (T0, null subject) = (T0, null subject)
[uNumber] [SgNumber] [uNumber] [PlNumber]
[uPerson] [3Person] [uPerson] [3Person]
[uGender] [?Gender] [uGender] [?Gender]
[uCase] [SgNumber] [uCase]
[3Person] [NomCase]
=> [?Gender]

Hence, it is not at all clear how the purely syntactic Agree system can deal with the
features under consideration. Moreover, the Spanish facts challenge Chomsky’s
(2001: 15) Maximize Matching Effects Condition.10

4.â•… A constraint on the output of Agree

The generalization gleaned from the Spanish data discussed hitherto is that in the
absence of optimal agreement (i.e. person, number, and gender features of T0 being
deleted/valued by means of Agree) between the verb and the (overt/null) subject, two
φ-features seem to be handled via the Agree operation, the remaining φ-feature of T0
being deleted/valued by alternative mechanisms. As a matter of fact, the derivations
under consideration ultimately converge, which assumes that the feature left syntacti-
cally unchecked must be dealt with by some operation other than Agree, if we wish to
retain the Agree system, mutatis mutandis. Therefore, I propose the following condi-
tion on the output of Agree, which is operative (at least) in Spanish, and possibly in
related languages like Asturian and Catalan:
(24) At most one φ-feature may be left syntactically unchecked/unvalued, and this
φ-feature may be handled by alternative mechanisms.

I submit that Spanish is a split-phi probe language. Similarly, Agree is deemed to be

an operation (rather than a configuration, e.g. spec, head). Thus, it is possible for two
φ-features to probe without the third one, in the spirit of Béjar (2003).

.╅ As has been pointed out, Corbett (2006) has proposed an analysis according to which
a nominal like los ciudadanos in (5b) is not necessarily third-person plural, as is standardly
assumed. This intuition has been implemented syntactically by Sauerland (2004), who assumes
that the DP has a Phi/Ф Phrase layer on top of it, which is responsible for agreeing with T0. In
cases of number disagreement, however, it is difficult to motivate the stipulation that a nominal
morphologically marked as plural bears singular features in sentences like those in (14)/(15).
Instead, I will entertain the hypothesis that los ciudadanos in examples like (5) is third-person-
plural, along the lines of Olarrea (1996) and Camacho (2006), as mentioned above.
 To agree or not to agree 

At this point, however, a non-trivial question which arises concerns the way in
which the feature which does not probe is handled (except in the cases of default agree-
ment). As far as the feature person is concerned in cases like (5b) and (5c), where the
subject includes the speaker or the hearer, the person feature is not dealt with by means
of Agree, but possibly by a mechanism of semantic/pragmatic agreement. Although this
possibility will not be explored in detail here for reasons of space, it is in accordance
with the recent view that certain manifestations of agreement may be determined in
the post-syntactic component (see Bobaljik 2008 and references therein). With respect
to gender in finite sentences with null subjects, I assume that the null subject (cf. pro)
is a full DP, though it is underspecified for gender. This is motivated by the observation
that in sentences like (18a), it is clear that the subject is third-person singular, but no
information is disclosed about its gender. Therefore, the gender feature is handled
by a mechanism other than Agree, which deals both with the gender feature of the
unpronounced nominal, and with the corresponding feature carried by T0.
This type of approach seems reasonable provided that we adopt a Bobaljik/
Brody-style Single Output Syntax model whereby Agree is assessed at LF, and the two
interfaces communicate (see, for instance, Fox & Nissembaum 1999 and Bobaljik &
Wurmbrand 2005). Other authors who assume that agreement is not a purely syntactic
phenomenon in Spanish include Demonte (in preparation), who discusses concord within
the nominal phrase. Likewise, drawing on work by Miyagawa (2005) and Chomsky
(2008), who contend that C0 is the locus of agreement features, which are then handed
over to T0, Jiménez (2008) claims that in terms of agreement, Spanish is situated in
the middle of a continuum between agreement-prominent languages like English,
and discourse-prominent languages like Japanese, which lends further credence to
the analysis advocated here.
This move raises a number of complications, however, including the assumption
that the syntax-pragmatic interface is beyond the syntax-semantics/LF interface. Thus,
I leave this as an open possibility for future research.11

.╅ Despite the fact that I have assumed throughout the paper that Spanish overt/null subjects
enter into an Agree relationship with the verb, it is important to note that a prominent line of
research in Spanish hypothesizes that the actual subject is either pro (cf. Baker 1996; Olarrea
1996), or the inflectional endings of the verb (cf. Alexiadou & Anagnostopoulou 1998, among
many others). On this view, overt subjects do not establish an agreement relationship with T0.
However, questions arise as to why the subject and the verb conventionally display agreement
(cf. (1)), and, granting that pro/the verbal endings are the ‘actual’ subject, how the relationship
between these and the overt subject (e.g. clitic doubling or connectivity à la Cinque 1990) can
be characterized. If this turns out to be the correct analysis, the proposal advocated here can
be salvaged by assuming that a relationship along the lines of Agree occurs between the overt
subject and the actual subject.
 Julio Villa-García

Lastly, with regard to number, a mechanism of default agreement is at work in cases

where the subject is plural and the verb is singular (cf. (14)), while a semantic/pragmatic
agreement explanation is in order for cases where the opposite is apparent (cf. (16)).
It is important to note that in a system like the one proposed in (24), if valuing
only two features were enough for a derivation to converge, we would encounter the
problem of overgeneration. For this reason, it appears that it is necessary to have
the generalization in (24) as well as a list of which mismatches are (not) attested. In
other words, (24) seems to be a necessary but not a sufficient condition. Some possible
outcomes permitted by (24) which are clearly ungrammatical in Spanish include the
following, in addition to those in (1b) and (4b):
(25) a. *Se venden libro
cl. sell-prs-3.pl book-3.sg
“*A book are sold.”
b. *Nosotros / vosotros saben …
we-1.pl / you-2.pl know-prs-3.pl …
“We/you know …”
c. *El ciudadano estoy / estás harto de tanto papeleo
the citizen-3.sg be-prs-1.sg/ be-prs-2.sg fed up of so much paperwork
“*The citizen am/are fed up with so much paperwork.”
d. *Vinisteis a verme tú
come-pst-2.pl to see + cl. you-2.sg
“You came to see me.”

Nevertheless, cases like (25) can receive an independent explanation. By way of

illustration, utterances like (25a) are ruled out because default agreement is never
third-person plural. Ill-formed outcomes like (25c), for their part, can be excluded
by arguing that first-person plural refers to a group that includes the speaker, while
second-person plural is a set that comprises the hearer. However, third-person sin-
gular carries the featural make-up [-speaker, -hearer], which implies that neither the
speaker (first-person) nor the hearer (second-person) can be included.
Overall, Spanish seems to be a composite of two types of languages, those with
discordant person subjects (cf. (5b)/(5c)), a point of independent variation observed in
other languages, and those exhibiting mismatches in number with postverbal subjects
(cf. (14)), a pervasive phenomenon across the world’s languages (cf. Greenberg 1966;
Ortega-Santos 2008). A logical question to pose at this juncture, however, is whether
the generalization in (24) is true or spurious. This issue is principally posed by the
observation that gender is never expressed in the verbal inflection, so it is not possible
to determine whether a person or number mismatch is in reality a case of two features
out of three being valued via Agree, as opposed to, say, total semantic agreement in
the NP…V-first-/second-plural cases, or total absence of agreement in the postverbal
plural NP cases.
 To agree or not to agree 

The data and analysis provided in this paper cannot tackle this question straight-
forwardly, and for this reason cross-linguistic research is needed at this point. To this
end, consideration of the realm of mismatches found across the board in the world’s
languages is imperative.

5.â•… Conclusion

This paper took as its point of departure the presumption that subjects (whether overt
or null) in Spanish enter into an Agree relationship with the verb (Chomsky 2000,
2001, 2004, 2008). Spanish is a standard textbook example of a null-subject language.
Overt and null subjects conventionally enter into an agreement relationship with the
verb, whose inflectional endings bear person and number morphemes. Since Spanish
nominals carry different features, and three of those are person, number, and gender
φ-features, I have pursued the possibility that the head T0 in Spanish hosts person,
number, and (abstract) gender φ-features.
It has been seen that in non-Caribbean Spanish, certain grammatical sentences
exhibit mismatches/lack of agreement between the verb and the subject in terms of
person or number. Further, it has been shown that the status of the gender feature in
the case of null subjects in finite contexts is not obvious, showing that pro is not just a
pronominal deprived of phonological content.
In light of the above, this paper has advanced the hypothesis that agreement (in
Spanish) is not a quintessentially syntactic phenomenon (i.e. it is not always the result of
a probe-goal relationship). In much the same way, it is not at all clear that (the strongest
version of) Chomsky’s (2001) Maximize Matching Effects Condition is operative in
Spanish, assuming that the operation Agree underlies agreement processes in natural
language. It seems that in Spanish (at most) one feature can be handled by a mecha-
nism other than Agree, so I postulated a condition on the output of Agree which states
that if optimal agreement (i.e. person, number, and gender φ-features being deleted/
valued syntactically by means of Agree) fails to obtain, at most one feature can be left
syntactically unchecked/unvalued, and this remaining feature is handled by alternative
mechanisms. These include default agreement in cases of verbs carrying third-person
singular morphemes and plural nouns (cf. (14)), and possibly semantic/pragmatic
agreement to account for cases of singular quién “who” with a plural verb (cf. (16)),
discordant person subjects (cf. (5b)/(5c)), as well as the “missing” gender of null
subjects (cf. (18)). Another consequence of this paper is that pro – assumed to be
the empty category representing null subjects – is not exactly a pronominal without
phonological material, but a full DP whose gender feature is underspecified.
Throughout the paper, a number of questions were raised, including whether the
generalization put forward in this article is true or spurious, an issue which future
 Julio Villa-García

comparative research should care to address. Likewise, although this paper has focused
solely on subject-verb agreement, the issue of subject-predicate agreement is also
worthwhile exploring, since this will definitely shed light on the checking of gender
features. For the time being, however, the fact still remains that Spanish exhibits a num-
ber of acceptable cases of anomalous agreement between grammatical and lexical/null
subjects which any theory of language should endeavor to explain. The preliminary
suggestions made in the present paper are intended to be a small step toward this end.

References

Alexiadou, Artemis & Elena Anagnostopoulou. 1998. “Parameterizing AGR: Word Order,
V-Movement and EPP-Checking”. Natural Language and Linguistic Theory 16. 491–539.
Baker, Mark. 1996. The Polysynthesis Parameter. New York: Oxford University Press.
Béjar, Susana. 2003. Phi syntax. Ph.D. dissertation, University of Toronto.
Bobaljik, Jonathan David & Susi Wurmbrand. 2005. “The Domain of Agreement”. Natural Language
and Linguistic Theory 23. 4. 809–865.
Bobaljik, Jonathan David. 2008. “Where’s Phi? Agreement as a Post-Syntactic Operation”. Phi-Theory:
Phi Features Across Interfaces and Modules, ed. by D. Harbour, D. Adger & S. Béjar. Oxford:
Oxford University Press. 295–328.
Bošković, Željko. To appear. “On Unvalued Uninterpretable Features.” Proceedings of NELS 39.
Camacho, José. 2003. The Structure of Coordination: Conjunction and Agreement Phenomena in
Spanish and Other Languages. Dordrecht: Kluwer Academic Publishers.
Camacho, José. 2006. “Do Subjects Have a Place in Spanish”. New Perspectives in Romance
Linguistics, ed. by J-P. Montreuil & C. Nishida. Amsterdam-Philadelphia: John Benjamins.
51–66.
Chomsky, Noam. 2000. “Minimalist Inquiries: The Framework”. Step by Step: Essays on Minimal-
ist Syntax in Honor of Howard Lasnik, ed. by R. Martin, D. Michaels & J. Uriagereka.
Cambridge, Mass.: MIT Press. 89–156.
Chomsky, Noam. 2001. “Derivation by Phase”. Ken Hale: A Life in Language, ed. by M. Kenstowicz.
Cambridge, Mass.: MIT Press. 1–52.
Chomsky, Noam. 2004. “Beyond Explanatory Adequacy.” Structures and Beyond. The Cartography
of Syntactic Structures, ed. by A. Belletti. Oxford: Oxford University Press. 104–131.
Chomsky, Noam. 2008. “On Phases”. Foundational Issues in Linguistic Theory: Essays in Honor
of Jean-Roger Vergnaud, ed. by R. Freidin, C. Otero & M.L. Zubizarreta. Cambridge, Mass.:
MIT Press. 133–166.
Cinque, Guglielmo. 1990. Types of A′-dependencies. Cambridge, Mass.: MIT Press.
Corbett, Greville. 2006. Agreement. Cambridge: Cambridge University Press.
Demonte, Violeta. In preparation. “Number Agreement and Conjoined Nouns in (Spanish) DPs
with and without Adjectives. AGREE at the Syntax-semantics Interface”. Ms., CCHS-Consejo
Superior de Investigaciones Científicas, Spain.
Fernández Soriano, Olga. 1989. Rección y Ligamiento en Español: Aspectos del Parámetro del
Sujeto Nulo. Ph.D. dissertation, Universidad Autónoma de Madrid.
Fox, Danny & Jon Nissenbaum. 1999. “Extraposition and Scope: A Case for Overt QR”. Proceedings
of the 18th West Coast Conference on Formal Linguistics (WCCFL 18), ed. by S. Bird, A. Carnie,
J.D. Haugen & P. Norquest. Somerville, Mass.: Cascadilla Press. 132–144.
 To agree or not to agree 

Greenberg, Joseph. 1966. “Some Universals of Grammar with Particular Reference to the Order of
Meaningful Elements”. Universals of Language, ed. by J. Greenberg. Cambridge, Mass.: MIT
Press. 73–113.
Holmberg, Anders. 2005. “Is There a Little pro? Evidence from Finnish”. Linguistic Inquiry 36.4.
533–564.
Hurtado, Alfredo. 1984. “La Hipótesis de la Discordancia”. Los Clíticos del Español y la Gramática
Universal. Supplement of Revista Argentina de Lingüística.
Jiménez, Ángel. 2008. “A Discourse/Agreement-based Approach to Peripheries”. Paper presented
at the Mediterranean Syntax Meeting 2, Boğaziçi University, Istanbul, October 2008.
Miyagawa, Shigeru. 2005. “On the EPP”. Perspectives on Phases, ed. by M. McGinnis & N. Richards.
Cambridge, Mass.: MIT Working Papers in Linguistics. 201–236.
Olarrea, Antxon. 1996. Pre and Postverbal Subjects in Spanish: A Minimalist Account. Ph.D.
dissertation, University of Washington, Seattle.
Ortega-Santos, Iván. 2005. “On Locative Inversion and the EPP in Spanish”. Actas del VIII
Encuentro Internacional de Lingüística del Noroeste de la Universidad de Sonora, ed. by
R.M. Ortiz Ciscomani. Universidad de Sonora, México. 131–150.
Ortega-Santos, Iván. 2008. Projecting Subjects in Spanish and English. Ph.D. dissertation, University
of Maryland, College Park.
Pesetsky, David & Esther Torrego. 2007. “The Syntax of Valuation and the Interpretability of
Features”. Phrasal and Clausal Architecture: Syntactic Derivation and Interpretation, ed. by
S. Karimi, V. Samiian & W. Wilkins. Amsterdam: John Benjamins. 262–294.
Rizzi, Luigi. 1982. Issues in Italian Syntax. Dordrecht: Foris.
Rodríguez-Mondoñedo, Miguel. 2007. The Syntax of Objects: Agree and Differential Object
Marking. Ph.D. dissertation, University of Connecticut, Storrs.
Sauerland, Uli. 2004. “A Comprehensive Semantics for Agreement”. Paper presented at the
Phi-Workshop, McGill University, Montreal, Canada, August 2004.
Sobin, Nicholas. 1997. “Agreement, Default Rules, and Grammatical Viruses”. Linguistic Inquiry 28.
318–343.
Villalba, Xavier. 2000. The Syntax of Sentence Periphery. Barcelona: Universitat Autònoma de
Barcelona, Servei de Publicacions.
Zagona, Karen. 2002. The Syntax of Spanish. Cambridge: Cambridge University Press.
 Variation in subject expression
in Western Romance*

Ana de Prada Pérez

University of Florida

Syntactic-theoretic accounts report variation across languages on the availability

of null pronominal subjects. As a result, languages are classified as null and
non-null subject languages. However, the homogeneity or heterogeneity of
null subject languages is not discussed. Variationist research, on the other
hand, indicates that variation is attested across different varieties of null subject
languages. This paper expands on this research comparing the distribution of null
and overt pronominal subjects in two null subject Western Romance languages:
Spanish and Catalan. Naturalistic data collected via sociolinguistic interviews in
Valladolid and Minorca, Spain, were explored using a variationist approach, with
a total of 7,025 tokens. The weight of eleven different internal factors relevant to
this distribution were analyzed and ranked, revealing differences between the two
languages. This variationist analysis succeeds in locating the contexts where the
languages differ in their favoring of overt and null forms.

1.â•… Introduction

Syntactic theoretic accounts differentiate between languages that require the presence
of an overt subject (non-null subject languages) and those that allow for an empty
category, known as pro (null subject languages). This typological distinction was first
introduced by Perlmutter (1971). A null subject is a syntactically present subject that

*This work was funded by the NSF dissertation research improvement grant 0746748. Thanks
to Almeida Jacqueline Toribio, Diego Pascual Cabo, the editors of this volume, and two
anonymous reviewers for her helpful comments and to Analia Alcolea, Pepe Álvarez, Mark
Amengual, Ernesto Carriazo-Osorio, Carmen Castro, Eva Florit Pons, Lorena Cuya Gavilano,
César Giraldo, Natalia Guzmán, Bonnie Holmes, Elizabeth Finanger, Araceli de Prada, Britton
Smith, and Arthur Wendorf for their invaluable assistance. Finally, a special thanks to the
participants of this study. All errors are my own.
 Ana de Prada Pérez

is phonetically unpronounced. It can be thematic (1a) or expletive (1b). Crucially, only

thematic subjects can be overt.1

(1) a. Ø Llevaré impermeable.

“I’ll wear a raincoat.”
b. Ø Parece que Ø va a llover.
“It seems it will rain.”

While both thematic and expletive null subjects must be licensed, only thematic sub-
jects must also be identified, i.e. associated with phi-features to establish the reference
of the argument (Rizzi 1986). Licensing is regulated by the morphological uniformity
principle (Jaeggli & Safir 1989:€29), which states that only languages with homogeneous
(only underived or only derived) paradigms can license null pronominal subjects.
Identification is achieved locally in Spanish through strong agreement (an Agr-S+T
complex with overt phi-features and a strong nominative case feature), which provides
the null subject with referential value.
Syntactic theoretic accounts focus on the parametric variation attested across lan-
guages regarding the availability of null pronominal subjects. However, the alternation
between null and overt pronominal subjects, which is available only with thematic sub-
jects, goes largely unexplored. This distribution is a variable phenomenon, i.e. subject
expression cannot be reduced to a rule. Variationist studies examine this variability,
mainly with naturalistic corpora collected via sociolinguistic interviews. They exam-
ine subject expression (the dependent variable) as a function of multiple variables or
factor groups simultaneously, be they language-internal or language-external.2 This
line of research has revealed that differences in pronominal expression may emerge
as a function of regional dialect (Cameron 1996; Otheguy et al. 2007), speech genre
(Travis 2007), extent of language contact (Otheguy et al. 2007), or language-internal
factors (Bentivoglio 1987; Enríquez 1984; Morales 1997; Silva-Corvalán 1982, 1994,
2001), to name but a few variables.
Interestingly, this methodology is amenable to comparisons between null sub-
ject languages of the same type. Thus, this project applies this methodology to the

.╅ Notice, however, the occurrence of overt ello in Dominican Spanish (Toribio 2000).
.╅ For instance, the language-external variable or factor group Gender has two possible con-
straints or factors, Male and Female. As can be discerned from the example, constraints are the
different levels in a variable. A variationist approach would examine whether Gender is signif-
icant in subject expression and whether being a Male or a Female correlates with a higher or
lower incidence of overt subjects. Likewise, language-internal variables, such as Co-reference,
with the constraints Co-referent (same subject as the previous clause) or Non-co-referent
(different subject from the previous clause), are examined.
 Variation in subject expression 

comparison of Spanish and Catalan, two null subject languages that identify thematic
pro locally.
In summary, the large body of existent research on null subject languages com-
pares parametrically-different languages (and null subject languages that identify
thematic pro locally with those which do so non-locally) from a theoretical point of
view, or compares different varieties of the same language from a variationist approach.
The present project expands on both approaches comparing two parametrically-
similar languages, Spanish and Catalan, regarding the distribution of null and overt
pronominal subjects.

2.â•… Factors involved in subject expression in Spanish

Numerous variationist studies have examined the factors that predict the distribu-
tion of null and overt pronominal subjects. These factors are mainly discourse – or
morphology-related. Some studies additionally address language-external factors,
such as age and gender. The role of the latter, however, remains largely controversial.
The literature on the variable distribution of pronouns in Catalan is scant; to the best
of my knowledge, it is reduced to Casanova (1999). In general, Catalan is assumed to
behave like Spanish.
Null thematic subjects are licensed and identified syntactically in Spanish and
Catalan. In contrast, the alternation between null and overt pronominal subjects is
largely constrained by the discourse context. Null subjects are associated with notions
of continuity, while overt subjects introduce new information or add emphasis to a pre-
viously mentioned topic. Among these discourse-related factors or variables are topic
continuation, distance from referent, co-reference (i.e. switch reference in
the work of Cameron and his colleagues), Tense/Aspect/Mood (TAM) continuity,
speech connectivity, person, and clause type. There are some lexico-semantic
and morphology-related variables that play a role in the distribution of null and overt
subjects in Spanish. Among these are person and number, semantic and syntactic
verb type, animacy, and verb form ambiguity.
Subject expression has been widely proven to be regulated by the variable topic
continuation, that is whether the subject has been mentioned (topic continuation),
it is new, or it has been mentioned but it is expressing contrast (e.g. Bayley & Pease-
Álvarez 1996, 1997; Bentivoglio 1987; Enríquez 1984; Otheguy et al. 2007; Silva-Corvalán
1982, 1994). Null subjects are favored in contexts of topic continuation while new
topic and contrast contexts favor overt subjects (Bentivoglio 1987; Cameron 1994,
1995; Cameron & Flores 2004; Flores-Ferrán 2004; Silva-Corvalán 1994; Travis 2005,
2007). Co-referentiality or switch reference (in Cameron’s terminology) deter-
mines whether the subject is co-referential with the preceding subject. Therefore,
 Ana de Prada Pérez

co-referentiality is different from topic continuation in that the former takes

the syntactic function of the referent into account. Consider the following example,
from Silva-Corvalán (2001):
(2) Me vine con Pepe hoy a la oficina. Pepe/él/Ø vive cerca de mi casa.
“I came to the office with Pepe today. Pepe/he/Ø lives close to my house.”

In this example there are two clauses. The subject of the first clause is the first person
singular, while in the second clause the subject is Pepe; thus, the subjects are non-
co-referential. But Pepe is mentioned in the first clause, thus, there is topic continuity.
Co-reference (3) has been found to favor null subjects (Bayley & Pease-Álvarez 1996,
1997; Flores-Ferrán 2004; Otheguy et al. 2007; Silva-Corvalán 1994).
(3) Pepe vino hoy a la oficina. Pepe/él/Ø vive cerca de mi casa.
“Pepe came to the office today today. Pepe/he/Ø lives close to my house.”

The variable Distance from previous mention is measured by the number of

intervening subjects between the subject of the clause and its referent. For instance,
Examples (2) and (3) have a distance of 0 intervening subjects, since the referent is
found in the previous clause. In (4), however, the distance is 1 intervening subject:
(4) Me vine con Pepe hoy a la oficina. Ø Tengo mucho trabajo.
Pepe/él/Ø vive cerca de mi casa.
“I came to the office with Pepe today. I have a lot of work to do.
Pepe/he/Ø lives close to my house.”

Studies conclude that the further the referent is, the more probable it is to induce
an overt form (Cameron 1995; Cameron & Flores-Ferrán 2004; Flores-Ferrán 2004;
Travis 2005, 2007).
The variable clause type distinguishes between main and embedded clauses.
Returning to Example (2), the target subject is in a main clause. An example of an
embedded clause subject is given in (5).
(5) Me vine con Pepe hoy a la oficina. Ø Pensaba que Pepe/él/Ø vivía cerca
de mi casa pero no.
“I came to the office with Pepe today. I thought that Pepe/he/Ø lived close
to my house but he doesn’t.”

Embedded clauses favor null subjects (Margaza & Bel 2006:€92; Morales 1997:€157),
while main clauses do not favor either form (see, however, Silva-Corvalán 1994:€152).
The variable Tense-Aspect-Mood (TAM) continuity indicates whether the
tense of the target clause and that of the preceding clause are the same or different. In
Example (2) the tense is different; the first clause (Me vine con Pepe hoy a la oficina) is
in the preterit while the target clause (Pepe/él/Ø vive cerca de mi casa) is in the present
 Variation in subject expression 

indicative. In Example (5), however, the target clause (Pepe vivía cerca de mi casa) is
in the imperfect and so is the preceding verb form (pensaba), thus, it is a case of same
TAM. TAM continuity has been found to correlate with the distribution of null and
overt subjects (Ávila-Shah 2000; Bayley & Pease-Álvarez 1997; Paredes Silva 1993):
TAM continuity favors null subjects (Otheguy et al.2007; Silva-Corvalán 1982, 1994;
Travis 2007). Paredes Silva (1993) proposes to combine the variables TAM continuity
and co-reference, distinguishing between six degrees of speech connectedness, in
her work on Brazilian Portuguese. Bayley & Pease-Álvarez (1997) and Otheguy et al.
(2007) simplify Paredes Silva’s (1993) six degrees in their studies of Spanish in the US.
Otheguy et al. (2007), for instance, use three degrees: same referent and same TAM,
same referent and different TAM, and different referent.
Person can be considered a discourse variable because first and second person
subjects are always given information (Chafe 1994). The predictions would, therefore,
be that third person would exhibit more overt subjects than first and second person.
Results, however, show the opposite trend: first person singular favors overt subjects
(Bentivoglio 1987; Davidson 1996; Enríquez 1984; Morales 1997; Otheguy et al. 2007;
Silva-Corvalán 1982, 1994; Travis 2005, 2007). In particular, the use of overt forms
is increased in conjunction with estimative verbs, which express the opinion of the
speaker (Bentivoglio 1987; Enríquez 1984; Silva-Corvalán 1994; Travis 2005). Bayley€&
Pease-Álvarez (1996) also find that first person singular forms favor overt pronominal
subjects. However, they report the same trend for third person singular and, therefore,
they argue that it is due to the lack of saliency of singular endings. Singular forms
have been reported to favor overt subjects more than plural forms (Cameron 1992;
Enríquez 1984; Flores-Ferrán 2004; Silva-Corvalán 1994).
Spanish does not exhibit an overt pronominal form for inanimates. As a con-
sequence of the categorical distribution of null and overt pronominal subjects with
inanimate referents, variationist studies typically do not consider the variable animacy.
Nonetheless, null pronominal subjects alternate with overt lexical subjects with
inanimate referents. Lozano (2009) analyzes written data from a large corpus of L2
learners of Spanish (CEDEL) and finds an overproduction of overt pronominal subjects
with third person singular animate subjects. Importantly, this overproduction is not
attested with third person singular inanimate subjects.
The variable semantic verb type classifies verbs according to their meaning.
Numerous authors have shown that verbs that express the speaker’s opinion favor overt
subjects more than other verb types (Bentivoglio 1987; Enríquez 1984; Silva-Corvalán
1994; Travis 2007). Similarly, the variable syntactic verb type has been shown to be
significant in the distribution of subjects. Casanova (1999) finds that copulative verbs
favor overt subjects more than predicative verbs in Catalan, an effect that she ascribes
to the lack of semantic content of this type of verb. But Serratrice & Sorace (2003) and
 Ana de Prada Pérez

Serratrice et al. (2004) found that syntactic verb type was not a significant factor
group in the distribution of null and overt subjects in Italian.
The variable verb form ambiguity characterizes verb forms as ambiguous or
unambiguous (i.e. whether the verb inflection is exclusive to one person or not). In
many studies, overt subjects have been found to appear more frequently with verbal
forms that are ambiguous than with unambiguous ones (Bayley & Pease-Álvarez 1996,
1997; Silva-Corvalán 1994; Travis 2005). However, this effect has not always been
found, leading some researchers to conclude that other variables override verb form
ambiguity (Casanova 1999; Morales 1997; Ranson 1991). Ranson, for instance, found
that ambiguous forms exhibited fewer overt pronominal subjects than unambiguous
ones in the Andalusian dialect she examined. She attributes this to the fact that the
subject of ambiguous forms were identifiable in the context (the morphological ambi-
guity was resolved contextually).
These variables have been examined in contexts of dialectal variation and lan-
guage contact. The following section reviews the results reported in these studies.

3.â•… Dialectal variation and language contact

Differences across dialects with respect to subject expression are widely present in
the variationist literature. Initially, the difference that garnered most attention was the
overall percentage of overt subjects (see Cameron 1992, 1993; Lipski 1994; Otheguy€&
Zentella 2007; Otheguy, Zentella & Livert 2007; Travis 2007). The following table
(adapted from Silva-Corvalán 2001) presents an overview.

Table 1.╇ Dialectal variation in overall rates of pronominal subject expression

(adapted from Silva-Corvalán 2001)
Study Place Percentage of overt subjects

Enríquez (1984) Madrid 21%

Otheguy et al. (2007) Mainland (Mexico, 24%
Colombia and Ecuador)
Newcomers in New York
Miró Vera & Pineda (1982) Sevilla 27%
Silva-Corvalán (1982) Los Ángeles 35%
Barrenechea & Alonso (1977) Buenos Aires 36%
Otheguy et al. (2007) Caribbean Newcomers 36%
in New York
Horchberg (1986) Boston 37%
Cifuentes (1980) Santiago 38%
Bentivoglio (1987) Caracas 40%
Morales (1982) Puerto Rico 40%
Cameron (1992) San Juan 45%
 Variation in subject expression 

In general, Caribbean and Santiago de Chile Spanish exhibit the highest rates of
overt subjects, and Peninsular Spanish the lowest. Nonetheless, these percentages
are difficult to compare. As Silva-Corvalán (2001) explains, even though most of the
studies draw on data from sociolinguistic interviews, not all the studies include the
same factors, and they aim to answer different questions. Thus, comparisons across
different studies are not reliable. Exploring differences in patterns, on the other hand,
offers more viable comparisons. Cameron (1996) reveals patterning differences between
Puerto Rican and Madrid Spanish regarding non-specific second person singular;
Puerto Rican Spanish favors overt pronominal subjects while Madrid Spanish disfavors
them. Otheguy et al. (2007) report differences between two varieties of New€York
Spanish, Caribbean and Mainland. They found that second person singular is the con-
straint that favors overt pronominal subjects the most in Caribbean varieties while it
is third person singular in Mainland varieties. Thus, contrasts in patterns constitute a
better comparative tool than contrasts in overall percentages.
Contact has been argued to act as an accelerator of linguistic change, targeting
areas where variation exists in monolingual speech (Silva-Corvalán 1994). The discus-
sion on rates of pronominal expression across contact vs. non-contact varieties is as
unproductive as it is for comparisons across regional varieties. Bayley & Pease-Álvarez
(1996, 1997), Flores-Ferrán (2004), Silva-Corvalán (1994), Torres-Cacoullos & Travis
(forthcoming) and Travis (2007) reveal no increase in subject expression in Spanish
in the U.S. Lapidus & Otheguy (2005a, b), Lipski (1996), Montrul (2004), Otheguy€&
Zentella (2007), Otheguy et al. (2007), Toribio (2004), on the other hand, report an
increase. Otheguy et al. (2007), for instance, report a 36% overall pronominal subject
rate in recently arrived immigrants from the Caribbean (Dominican Republic, Puerto
Rico, and Cuba) in New York. New York born and raised Spanish speakers of Caribbean
heritage, however, exhibited a 42% rate. In the case of Mainland (Mexico, Colombia,
and Ecuador) Spanish speakers, those recently arrived displayed 24% overt pronomi-
nal subjects, while those born and raised in New York produced 33% overt pronominal
subjects. Due to these inconsistencies across studies, some researchers have resorted
to comparisons based on patterns. By way of illustration, consider Silva-Corvalan’s
(1994) study, where she employed the same methodology with diverse communities;
monolingual speakers in Mexico and bilingual Spanish speakers in L.A.. She did not
report an increase in overt pronominal subjects in contact varieties. She, did, how-
ever, find patterning differences between monolingual and bilingual speakers regarding
the variable person. While monolingual speakers favor overt pronominal forms the
most in first person singular, bilingual speakers favor overt subjects the most in third
person singular.
To conclude this section, research indicates that a large number of variables play
a significant role in the distribution under study here. Crucially, comparison between
different regional varieties or monolingual and bilingual varieties are better reflected
in patterning differences than in differences in rate of overt subjects.
 Ana de Prada Pérez

4.â•… The present project

The syntactic theoretic literature does not report differences across null subject lan-
guages, possibly due to the lack of interest in the alternation between null and overt
subjects. This alternation has been thoroughly studied in the variationist literature,
with an emphasis on regional variation. The present project, thus, extends the appli-
cation of variationist methodologies to the comparison of two null subject Western
Romance languages: Spanish and Catalan.

4.1â•… The sociolinguistic context

The data for this project was collected in Spain: the Spanish data was collected in
Valladolid and the Catalan data in Minorca. Valladolid is a largely monolingual com-
munity with little immigration. Data from el Instituto Nacional de Estadística (2006
Census) indicates that the monolingual regions with the smallest rates of immigration
are Cantabria and Castilla y León. Since no important differences exist between these
two regions, Valladolid, in Castilla y León, was selected for convenience. Minorca
was selected because of its little immigration (as compared to other Catalan-speaking
communities) and the extensive use of Catalan. Importantly, Catalan monolinguals
do not exist. Thus, data was collected from Catalan-dominant bilinguals that live in
Catalan-dominant communities. Among the Catalan Countries, Minorca reports
the lowest number of foreign-born among its population (2006 Census). Blas Arroyo
(2007) reports that immigration rates correlate with knowledge of Catalan, i.e. places
with more immigrants have lower numbers of Catalan speakers. In addition, Minorca
also exhibits extensive use of Catalan (cf. 2003 Sociolinguistic Survey, carried out by
the Conselleria d’Educació i Cultura del Govern de les Illes Balears). Minorca exhibits
three distinct varieties: Maó, Ciudadella and ‘pobles’, the villages in the center of the
island (Mascaró i Pons 1987). The participants for this study are from three villages in
the center of the island: Ferreries, Fornells, and Alaior.

4.2â•… Method
In order to compare Valladolid Spanish and Minorcan Catalan distributions of null
and overt subjects, 12 Valladolid Spanish speakers, two males and two females in each
of the three age groups, and 12 Minorcan Catalan speakers, also with two males and
two females in each of the three age groups, were recorded during an oral interview.
The oral interview included a language history, used for participant profiling, a socio-
linguistic interview, the target materials for this study, and a survey of language atti-
tudes and ideologies. The sociolinguistic interview was based on Tagliamonte (2006)
and adapted to the target culture. Participants were asked about their personal experi-
ences (their studies, jobs, families, trips, hobbies) as well as about their home town,
traditions, celebrations, typical dishes, etc. Some sample items appear in (6).
 Variation in subject expression 

(6) Ethnolinguistic interview, sample items

a. ¿Cómo se celebran las fiestas patronales aquí?
“How do you celebrate your patron saint here?”
b. ¿Has tenido la oportunidad de ir a la escuela? Háblame un poco de
cómo era tu vida cuando eras pequeño/a. ¿Cómo era un día normal en la
escuela? ¿Recuerdas algún profesor que te haya marcado positivamente?
¿Cómo era? ¿Y uno que no te gustara mucho? ¿Por qué? ¿Cómo era?¿En
qué lengua eran las clases? ¿Tenías un grupo de amigos en la clase con los
que jugaras en el recreo o por la tarde? ¿Cuántos erais? ¿Cómo eran tus
amigos? ¿A qué jugabais? ¿Cómo se jugaba?
“Have you had the opportunity to go to school? Tell me a little bit about
your life when you were a child. How was a regular day at school? Do you
remember a teacher that marked you in a positive way? What was s/he
like? And did you have one that you did not like that much? Why? What
was s/he like? What language were classes held in? Did you have a group
of friends in class to play during recess or in the evening? How many
were there? What were your friends like? What did you play? How did you
play that?”

The full interviews were transcribed and the sociolinguistic interview was coded for
subject form and 10 language-internal variables. The following table summarizes the
variables and their associated constraints:

Table 2.╇ Coding scheme: Variables and Associated Constraints

Variables Constraints Example

Subject Form Overt lexical Overt lexical: Me vine con

Overt pronominal Pepe hoy a la oficina. Pepe
Null subject vive cerca de mi casa.
Overt pronominal: Me
vine con Pepe hoy a la
oficina. Él vive cerca de
mi casa.
Null subject: Me vine con
Pepe hoy a la oficina. Ø
vive cerca de mi casa.
Topic Continuation Topic continuation Topic continuation: Me
New topic vine con Pepe hoy a la
Contrast oficina. Ø vive cerca de
mi casa.
New topic: Pepe vino hoy
a la oficina.
Contrast: Me vine con
Pepe y María hoy a la
oficina. Él vive cerca de
mi casa.
(Continued)
 Ana de Prada Pérez

Table 2.╇ Coding scheme: Variables and Associated Constraints╇ (Continued)

Variables Constraints Example

Co-referentiality/ Co-referential Co-referential: Pepe vino

switch reference Non-co-referential hoy a la oficina. Ø vive
cerca de mi casa.
Non-co-referential: Me
vine con Pepe hoy a la
oficina. Ø vive cerca de
mi casa.
Distance from 0 clauses 0 clauses: Me vine con
previous mention 1 clause Pepe hoy a la oficina. Ø
2 clauses… vive cerca de mi casa.
1 clause: Me vine con
Pepe hoy a la oficina.
Ø Tengo mucho trabajo.
Pepe vive cerca de mi casa.
2 clauses: Me vine con
Pepe hoy a la oficina. Ø
Tengo mucho trabajo.
Mi jefe lo quiere para
mañana. Pepe vive cerca
de mi casa.
Clause type Main Main: Me vine con Pepe
Embedded hoy a la oficina. Ø vive
cerca de mi casa.
Embedded: Me vine con
Pepe hoy a la oficina
porque Ø vive cerca de
mi casa.
TAM continuity Same TAM Same TAM: Me vine con
Different TAM Pepe hoy a la oficina. Ø
me llamó esta mañana
porque se le había
estropeado el coche.
Different TAM: Me vine
con Pepe hoy a la oficina.
Ø vive cerca de mi casa.
Speech connectivity Same referent, same TAM Same referent, same
(connect) Same referent, different TAM TAM: Me vine con Pepe
Different referent hoy a la oficina. Ø Tuve
un problema con el coche
esta mañana.
Same referent, different
TAM: Me vine con Pepe
hoy a la oficina. Ø Tengo
mucho trabajo.
Different referent: Me
vine con Pepe hoy a la
oficina. Ø vive cerca de
mi casa.

(Continued)
 Variation in subject expression 

Table 2.╇ (Continued)

Variables Constraints Example

Person First (Singular/Plural), First person singular: Ø
Second (Singular/Plural) Tengo mucho trabajo.
Third (Singular/Plural) Second person singular:
Ø Tienes mucho trabajo.
Third person singular: Ø
Tengo mucho trabajo.
Etc.
Animacy Animate Animate: Me vine con
Inanimate Pepe hoy a la oficina. Ø
vive cerca de mi casa.
Inanimate: Me he
comprado un jarrón
nuevo. Ø Es muy bonito.
Semantic verb type Psychological Psychological: Quiero que
Speech vengas.
Motion Speech: Dice que viene.
Copulative Motion: Viene.
Other Copulative: Es muy alto.
Other: Vive cerca de mi
casa.
Syntactic verb type Transitive Transitive: Compró un
Unergative regalo
Unaccusative Unergative: Vive cerca de
Copulative mi casa.
Unaccusative: Llegó ayer.
Copulative: Es alto.
Verb form ambiguity Ambiguous, Ambiguous: Comía
Unambiguous mucho entonces.
Unambiguous: Vive cerca
de mi casa.

The data was coded for all these variables. Example (2), repeated here for conve-
nience, exemplifies how the data was coded:
(7) Me vine con Pepe hoy a la oficina. Pepe/él/Ø vive cerca de mi casa.
‘I came to the office with Pepe today. Pepe/he/Ø lives close to my house.’

The variable subject form has been mostly studied with respect to the alternation
between null and overt pronominal forms. We also included lexical overt subjects in
our study. In the second clause in (7) the subject can be null or overt. When it is
overt, it can be pronominal (él) or lexical (Pepe). Example (7), for instance, would be
coded as topic continuation, since the referent is already mentioned in the context
and non co-referential, since the subject of the preceding clause is different. There are
0 intervening subjects, since the referent is mentioned in the preceding clause. It is
produced in a main clause. The TAM (present indicative) is different from the TAM
in the preceding clause (preterit). Regarding speech connectivity, it is a different
 Ana de Prada Pérez

referent. The subject is in the third person singular and it is animate. The verb falls
under the semantic category other and the syntactic category unergative. The verb
form is morphologically unambiguous.
A total of 3,439 Spanish tokens and 3,586 Catalan tokens were submitted to statis-
tical analysis using SPSS to compare overall rates and Goldvarb X to compare patterns.

4.3â•… Results and discussion

The overall rates of overt lexical, overt pronominal, and null subjects in Valladolid
Spanish and in Minorcan Catalan were examined. A 3 (subject form: null, overt pro-
nominal, overt lexical) x 2 (language: Spanish, Catalan) repeated-measures ANOVA
indicated a significant main effect for subject form (F(2,46) = 751.744, p < .01), no main
effect for language (F(1,23) = .470, p > .05) and no interaction (F(2,46) = .772, p€> .05).
Bonferroni post-hoc tests revealed that all participants produced more null than overt
subjects and more overt lexical than overt pronominal subjects (all p€ <€ .01). Thus,
there are no significant differences between the languages regarding overall rates.

80
70
60
50 Null
40 Pronominal
30 Lexical
20
10
0
Spanish Catalan

Figure 1.╇ Overall Rates of Subject Expression in Valladolid Spanish and Minorcan Catalan

The patterns were examined through multiple regression analyses. In these analy�
ses, contrasts are binary. Since the dependent variable distinguishes between null, overt
pronominal and overt lexical subjects, four contrasts were run: null vs. overt (lexical
and pronominal), null vs. overt pronominal, null vs. overt lexical, and overt pronominal
vs. overt lexical subjects. language was included as one of the variables, and it was
returned as significant. Thus, Valladolid Spanish and Minorcan Catalan differ in the
patterns of the distribution under study here. The following table summarizes the
variables that were returned as significant and their ranking:
 Variation in subject expression 

Table 3.╇ Cross-linguistic contrasts in variables implicated in subject expression

Contrasts Valladolid Spanish Minorcan Catalan

Null vs Overt Topic continuity Person

Distance Distance
Person Topic continuity
Co-reference Connect
Semantic verb type Semantic verb type
Age Animacy
Ambiguity
Null vs. Pronominal Animacy Person
Topic continuity Animacy
Person Distance
Co-reference Connect
Distance Syntactic verb type
Ambiguity Age
Gender Ambiguity
Gender
Null vs Lexical Person Person
Topic continuity Distance
Distance Co-reference
Semantic verb type Age
Co-reference Ambiguity
Age
Pronominal vs Lexical Person Distance
Topic continuity Person
Animacy Animacy
Age
Co-reference

In line with previous research, contexts of topic continuity favor null subjects,
topic shift contexts favor overt lexical subjects, and contexts of emphasis favor overt
pronominal subjects. Null subjects are favored when they are co-referential with
the preceding subject and overt subjects when they are non-co-referential. There is
an increase in the probability of an overt subject as distance increases. The variable
connect, which is significant only in Minorcan Catalan, indicates that the less con-
nected the subject is to the preceding context, the more overt subjects are used. As
in preceding research, ambiguous verbal forms favor overt pronominal forms. While
first person singular favors overt pronominal subjects and first person plural favors
null subjects, the third person favors overt lexical subjects. The variable syntactic
verb type reveals that transitive and copulative verbs favor null and overt pronom-
inal subjects, unergatives exhibit no preference and unaccusatives favor null and
overt lexical subjects. Finally, (semantically) copulative verbs favor null and overt
lexical subjects.
The two languages differ in some contexts. In non-co-referential contexts Minorcan
Catalan favors overt pronominal subjects and Valladolid Spanish overt lexical subjects.
 Ana de Prada Pérez

While both languages favor more overt forms as distance increases, the distance that
favors overt forms in Minorcan Catalan is 10 intervening subjects or more and in
Valladolid Spanish it is 5 intervening subjects or more. The second person singu-
lar favors a null subject in Minorcan Catalan and an overt pronominal subject in
Valladolid Spanish. The rest of the differences emerge in the variable semantic verb
type. Speech verbs favor overt pronominal subjects in Minorcan Catalan and null
subjects in Valladolid Spanish. Psychological verbs favor overt pronominal subjects in
Minorcan Catalan, while Valladolid Spanish disfavors them. Lastly, movement verbs
exhibit no preference in Minorcan Catalan while they favor overt lexical subjects in
Valladolid Spanish.
In summary, differences between Valladolid Spanish and Minorcan Catalan are
attested in the variable and constraint hierarchies. More meaningful contrasts in
patterns emerge from comparisons of constraint hierarchies. Nonetheless, no differ-
ences in overall rates emerge. Thus, the differences in patterns do not transcend to
differences in overall rates.

5.â•… Conclusion and directions for future research

Variationist research reveals differences across regional varieties in the overall rate of
overt pronouns, variable hierarchy, and constraint hierarchy (Otheguy et al. 2007).
This paper applies the same methodology to the comparison of Valladolid Spanish and
Minorcan Catalan. Results indicate differences in variable hierarchy and, most impor-
tantly, in constraint hierarchy, as was the case in regional variation. Nonetheless, no
differences are found in overall rates, unlike in regional variation between Caribbean
and Mainland Spanish. Since overall rates are difficult to compare because contexts
vary in sociolinguistic interviews, these data are not reliable to explain why differences
were found between regional varieties and not between languages. More research
is necessary to explain what contexts allow for the “overuse” of overt pronouns in
Caribbean Spanish. In sum, differences between Caribbean Spanish and Mainland
Spanish are more evident regarding overall rates of overt subjects than between
Valladolid Spanish and Minorcan Catalan. It is important to explore what a higher
overall rate implies for the use of overt and null subjects (i.e. is there a weakening of
constraints?). Comparisons of variable hierarchies, which depend on the contexts pro-
duced, are less reliable than constraint hierarchies with naturalistic data collected via
sociolinguistic interviews (Tagliamonte 2006). Thus, we have focused our discussion
on the favoring or disfavoring of constraints (i.e. constraint hierarchies). As in the case
of comparisons between different varieties of Spanish, Minorcan Catalan and Valladolid
Spanish differ in the distribution of null and overt subjects in a number of constraints,
 Variation in subject expression 

namely non-co-referential contexts, the distance at which overt forms start to be

favored, the second person singular form, speech, psychological, and movement verbs.
Using naturalistic data, comparisons across languages can reliably be stated in terms
of constraint hierarchies. However, differences in overall rates and variable hierarchies
would be better explored with a more experimental design, assuring that the contexts
produced for each participant are the same.
An additional limitation present when exploring Catalan is the lack of mono-
lingual speakers. Thus, the role of contact with Spanish remains a question. In work
in progress, we explore Minorcan Spanish data by Spanish-Catalan bilinguals at dif-
ferent points in the bilingual continuum. This allows us to determine different degrees
of contact with Spanish. We selected the participants that have the least amount of
contact with Spanish for this paper. Even though contact with Spanish is a confound,
Minorcan Catalan is a contact variety and is worth studying as it exists. In any case, we
can hypothesize that the differences discussed in this paper can only be attenuated by
convergence with Spanish.
This paper argues for the extension of variationist methodologies to the compari-
sons of languages that from a generative point of view are the same (i.e. they share
the same feature specifications) to be able to capture more finely grained differences.
Future research should examine other features widely studied in the variationist tradi-
tion (e.g. subject position) in typologically-similar languages.

References

Ávila-Shah, Bárbara I. (2000). “Discourse Connectedness in Caribbean Spanish”. Research on

Spanish in the United States, ed. by Ana Roca. Somerville, MA: Cascadilla Press. 238–251.
Barrenechea, Ana María & Alicia Alonso. 1977. “Los pronombres personales sujetos en el
español hablado en Buenos Aires”. Estudios sobre el español hablado en las principales
ciudades de América, ed. by Juan M. Lope Blanch, 333–349. Mexico City: Universidad
Nacional Autónoma de México.
Bayley, Robert & Lucinda Pease-Álvarez. 1997. “Null Pronoun Variation in Mexican-descent
Children’s Narrative Discourse”. Language Variation and Change 9.349–371.
Bayley, Robert & Lucinda Pease-Álvarez. 1996. “Null and Expressed Pronoun Variation
in Mexican-descent Children’s Spanish”. Sociolinguistic Variation: Data, theory, and
analysis, ed. by Jennifer Arnold, Renée Blake & Brad Davidson, 85–99. Stanford, Calif.:
CSLI Publications.
Bentivoglio, Paola. 1987. Los sujetos pronominales de primera persona en el habla de Caracas.
Caracas: Central University of Venezuela.
Blas Arroyo, José Luis 2007. “El contacto de lenguas como un factor de retención en procesos
de variación y cambio lingüístico. Datos sobre el español en una comunidad bilingüe
peninsular”. Spanish in Context 4: 2.263–291.
 Ana de Prada Pérez

Cameron, Richard & Nydia Flores-Ferrán. 2004. “Perseveration of Subject Expression across
Regional Dialects of Spanish”. Spanish in Context 1.41–65.
Cameron, Richard. 1996. “A Community-based Test of a Linguistic Hypothesis”. Language in
Society 25.61–111.
Cameron, Richard. 1995. “The Scope and Limits of Switch Reference as a Constraint on Pronominal
Subject Expression”. Hispanic Linguistics 6/7.1–27.
Cameron, Richard. 1994. “Ambiguous Agreement, Functional Compensation, and Nonspecific
Tú in the Spanish of San Juan, Puerto Rico, and Madrid, Spain”. Language Variation and
Change 5.305–334.
Cameron, Richard. 1992. Pronominal and Null Subject Variation in Spanish: Constraints, dialects,
and functional compensation. Ph.D. dissertation, University of Pennsylvania.
Casanova Seuma, Lourdes. 1999. “El sujeto en catalán coloquial”. Revista Española de Lingüística
29: 1.105–131.
Chafe,Wallace. 1994. Discourse, Consciousness and Time: The flow and displacement of conscious
experience in speaking and writing. Chicago: University of Chicago Press.
Cifuentes, Hugo. 1980. “Presencia y ausencia del pronombre personal sujeto en el habla culta de
Santiago de Chile”. Ms., University of Southern California, Los Angeles.
Davidson, Brad. 1996. “‘Pragmatic Weight’ and Spanish Subject Pronouns: The Pragmatic
and Discourse Uses of ‘tú’ and ‘yo’ in Spoken Madrid Spanish”. Journal of Pragmatics
26:€4:€10.543–565.
Enríquez, Emilia. 1984. El pronombre personal sujeto en la lengua española hablada en Madrid.
Madrid: Consejo Superior de Investigaciones Científicas.
Flores-Ferrán, Nydia. 2004. “Spanish Subject Personal Pronoun Use in New York City Puerto
Ricans: Can we rest the case for English contact?” Language Variation and Change
16.49–73.
Instituto Nacional de Estadística. (Censo del 2006). Immigration by Autonomous Communities
from 1998 to 2006. www.ine.es
Jaeggli, Osvaldo & Kenneth J. Safir. 1989. “The Null Subject Parameter and Parametric
Theory”. The Null Subject Parameter, ed. by Osvaldo Jaeggli & Kenneth J. Safir. Dordrecht:
Kluwer. 1–45.
Lapidus, Naomi & Ricardo Otheguy. 2005a. “Contact Induced Change? Overt Nonspecific ellos
in Spanish in New York”. Selected Proceedings of the 2nd Workshop on Spanish Sociolinguistics,
ed. by Lotfi Sayahi & Maurice Westmoreland, 67–75. Somerville, Mass.: Cascadilla.
Lapidus, Naomi & Ricardo Otheguy. 2005b. “Overt Nonspecific ellos in Spanish in New York”.
Spanish in Context 2.157–174.
Lipski, John M. 1996. “Patterns of Pronominal Evolution in Cuban-American bilinguals”. Spanish
in Contact: Issues in bilingualism, ed. by Ana Roca & John B. Jensen, 159–186. Somerville,
Mass.: Cascadilla.
Lipski, John M. 1994. Latin American Spanish. London: Longman.
Lozano, Cristóbal. 2009. “Selective Deficits at the Syntax-discourse Interface: Evidence from
the CEDEL2 corpus”. Representational Deficits in SLA: Studies in honor of Roger Hawkins
(Language Acquisition and Language Disorders), ed. by Neal Snape, Yan-kit Ingrid Leung &
Michael Sharwood-Smith. Amsterdam: John Benjamins, 127–166.
Margaza, Panagiota & Aaurora Bel. 2006. “Null Subjects at the Syntax-Pragmatics Interface:
Evidence from Spanish Interlanguage of Greek Speakers”. Proceedings of the 8th Generative
Approaches to Second Language Acquisition Conference (GASLA 2006), ed. by Mary Grantham
O’Brien, Christine Shea & John Archibald, 88–97. Somerville, Mass.: Cascadilla.
 Variation in subject expression 

Mascaró i Pons, Ignasi. 1987. “Ciutadella-Maó. Greu vs. Agut en dos parlars menorquins.
Plantejament de la quesito”. Randa 21.197–211.
Miró Vera, Ramona & Miguel Ángel de Pineda. 1982. “Determinación sociolingüística de la
presencia/ausencia del pronombre personal sujeto”. Sociolingüística andaluza 5: Habla
de Sevilla y hablas americanas, ed. by M. Teresa Palet Plaja, 37–34. Sevilla: Universidad
de Sevilla.
Montes Miró, Rosa Graciela. 1986. “Factores discursivos en el análisis de los pronombres per-
sonales sujeto en español”. Morphé 1.45–71.
Montrul, Silvina. 2004. “Subject and Object Expression in Spanish heritage Speakers: A case of
morphosyntactic convergence”. Bilingualism: Language and Cognition 7.125–142.
Morales, Amparo. 1997. “La hipótesis funcional y la aparición del sujeto no nominal: El español
de Puerto Rico”. Hispania 80.153–165.
Otheguy, Ricardo & Ana Celia Zentella. 2007. “Apuntes preliminares sobre el contacto lingüístico
y dialectal en el uso pronominal del español en Nueva York: Variación, cambio e identidad en
el uso variable del pronombre en seis comunidades hispanohablantes de la Gran Manzana”.
Spanish in Contact: Policy, Social and Linguistic Inquiries, ed. by Richard Cameron &
Kimberly Potowski, 239–257. Amsterdam, Philadelphia: John Benjamins.
Otheguy, Ricardo, Ana Celia Zentella & David Livert. 2007. “Language and Dialect Contact in
Spanish in New York: Towards the formation of a speech community”. Language 83.1–33.
Paredes Silva, Vera L. 1993. “Subject Omission and Functional Compensation: Evidence from
written Brazilian Portuguese”. Language Variation and Change 5.35–49.
Perlmutter, David M. 1971. Deep and Surface Structure Constraints in Syntax. New York: Holt,
Rinehart, and Winston.
Ranson, Diana L. 1991. “Person Marking in the Wake of /s/ Deletion in Andalusian Spanish”.
Language Variation and Change 3: 2.133–152.
Rizzi, Luigi. 1986. “Null Objects in Italian and the Theory of pro”. Linguistic Inquiry
17.501–557.€
Serratrice, Ludovica & Antonella Sorace. 2003. “Overt and Null Subjects in Monolingual and
Bilingual Italian Acquisition”. Proceedings of the Annual Boston University Conference on
Language Development 27: 2, ed. by Barbara Beachley, Amanda Brown & Frances Conlin.
739–750.
Serratrice, Ludovica, Antonella Sorace & Sandra Paoli. 2004. “Crosslinguistic Influence at
the Syntax-pragmatics Interface: Subjects and objects in English-Italian bilingual and
monolingual acquisition”. Bilingualism: Language and Cognition 7: 3.183–205.
Silva-Corvalán, Carmen 2001. Sociolingüística y pragmática del español. Washington, DC:
Georgetown University Press.
Silva-Corvalán, Carmen 1994. Language Contact and Change: Spanish in Los Angeles. Oxford:
Oxford University Press.
Silva-Corvalán, Carmen 1982. “Subject Expression and Placement in Mexican-American Spanish”.
Spanish in the United States: Sociolinguistic aspects, ed. by Jon Amastae & Lucía Elías-Olivares,
93–120. Cambridge: Cambridge University Press.
Tagliamonte, Sali A. 2006. Analysing Sociolinguistic Variation. Cambridge: Cambridge University
Press.
Toribio, Almeida Jacqueline. 2004. “Convergence as an Optimization Strategy in Bilingual Speech:
Evidence from code-switching”. Bilingualism: Language and Cognition 7. 165–173.
Toribio, Almeida Jacqueline. 2000. “Setting Parametric Limits on Dialectal Variation in Spanish.”
Lingua 110. 315–341.
 Ana de Prada Pérez

Torres-Cacoullos, Rena & Catherine Travis. Forthcoming. “Variable yo Expression in New€Mexico:

English Influence?” Spanish of the Southwest: A Language in Transition, ed. by Susana
Rivera-Mills & Daniel Villa. Iberoamericana/Vervuert.
Travis, Catherine. 2007. “Genre Effects on Subject Expression in Spanish: Priming in narrative
and conversation”. Language Variation and Change 19.101–135.
Travis, Catherine. 2005. “The Yo-yo Effect: Priming in subject expression in Colombian Spanish”.
Selected Papers from the 34th Linguistic Symposium on Romance Languages (LSRL),
Salt Lake City, 2004 ed. by Randall Gess & Edward J. Rubin, 329–349. Amsterdam:
John Benjamins.
 A phase-based analysis
of Old French genitive constructions

Deborah Arteaga & Julia Herschensohn

University of Nevada, Las Vegas & University of Washington

This paper considers two types of genitive constructions in Old French (OF), one
in which the possessor is introduced by a preposition (type un ami à moi “my
friend”/l’ami du roi “the king’s friend”) and another, in which no overt preposition
is found (type la niece le duc “the duke’s niece”), the so-called juxtaposition
genitive (JG). Although not preceded by a preposition, the possessor in the JG
is invariably in the oblique case (Arteaga 1995 and Delfitto & Paradisi 2009).
Assuming recent minimalist analyses of DPs (Lin 2009; Adger 2003; Hartmann€&
Zimermann, 2003; Carstens 2000, 2003), a phase-based account of OF genitive
constructions is proposed. It is assumed that in the JG, a null preposition checks
oblique case on the possessor. The feature [def], required in OF of possessors in
the JG, allows for the derivation of the various JG structures.

1.â•… Introduction

The structure and derivation of DPs has been the subject of several recent studies
(i.e. Adger 2003; Hartmann & Zimermann, 2003; Carstens 2000; Longobardi 1994).
Building on these works, this paper considers genitive constructions in Old French
(OF), including those with à “to” (type l’ami à Paul “Paul’s friend”), those with de “of ”
(type: l’ami de Paul “Pauls’ friend”), and the prenominal and postnominal juxtaposi-
tion genitive (JG) (type li amis le roi/li roi amis/le roi amis, “the friend of the king”).
Within a minimalist framework, following D’Alessandro & Roberts (2008), Chomsky
(2006), and Carstens (2003), we argue that the derivation of OF JG constructions is
phase-based.1

.╅ In a related article (Arteaga & Herschensohn 2011), we discuss the diachronic develop-
ment of French genitive constructions from Latin to Modern French, using more traditional
grammatical terminology.
 Deborah Arteaga & Julia Herschensohn

In Section€2 we begin by presenting an overview of OF genitive structures, using

the data presentation of Arteaga (1995) as a point of departure. We then turn in
Section€3 to the theoretical framework that we will draw on, and in 4 we propose a
new analysis of OF genitive structures. We show that the prepositionally marked
DPs operate as in Modern French (MF), unlike the JG, which, we argue, is determined
by a null preposition (in a K/PP) with a T-feature (Pesetsky & Torrego, 2009, hence-
forth P€& T) that checks oblique case. The crucial feature of this latter JG construction,
in our analysis, is that of definiteness (Lin 2008; Delfitto & Paradisi 2009, henceforth,
D & R), which is linked to the T-feature of K.

2.â•… OF genitive constructions

OF was a V2 language with a dual case (nominative/oblique) system, illustrated by

(1a) and (1b):
(1) a. Si fut uns sires
so was-3sg a-nom-m-sg lord-nom-m-sg
de Rome la citet
from Rome-obl-m-sg the-obl-f-sg city-obl-f-sg
“He was a lord from the city of Rome.” (La vie de Saint Alexis 13)
b. Donc li remembret de son
thus to.him remember-3sg of his-obl-m-sg
seignour celeste
lord-obl-m-sg heavenly-obl-m-sg
“Then he remembered his heavenly lord.” (La vie de Saint Alexis 57)
In (1a) the masculine subject DP uns sires “a lord” is in the nominative case, marked
by the canonical final -s. It is a so-called imparisyllabic noun (i.e. the nominative
form and the oblique form do not have the same number of syllables). The oblique
case of the DP appears in (1b) son seignour “his lord” (nominative: ses sires) as it is
the object of the preposition de “of.” In both examples, the verb occupies the second
position (V2).
In this paper, we focus on OF genitive structures. Arteaga (1995) notes that OF had
a wide array of such constructions, some of which still exist in Modern French (MF).2

.â•… An anonymous reviewer asks why “the JG was the only construction lost in the evolution
to Modern French.” It is indeed not the only one; V2, null subjects, two-Case system among
many other syntactic constructions were lost.
 A phase-based analysis of Old French genitive constructions 

For example, genitive structures where possession was marked by à were very com-
mon in OF (2a)–(2b):

(2) a. Fille [ad un comte]

daughter-nom-f-sg to a-obl-m-sg count-obl-m-sg
“A count’s daughter.”
(cf. MF “la fille à un comte”) (La vie de St. Alexis 42)
b. Por [la teste au
by the-nom-f-sg head-nom-f-sg to.the-obl-m-sg
serpant felon]
snake-obl-m-sg traitorous-obl-m-sg
“By the head of the traitorous snake.” (Yvain 3378)

Note that the possessor in à genitives does not have to be definite. In (2a), the pos-
sessor un comte is indefinite, whereas the possessor au serpant felon (2b) is definite.
The genitive with à is still found in MF in popular speech (e.g. un ami à moi “a friend
of mine” see Joseph 1995 for discussion). Another genitive structure shared by OF and
MF is the genitive with de (3):

(3) a. Sans la mort [de maint preudome]

without the-obl-f-sg death-obl-f-sg of many.a knight-obl-m-sg
“Without the death of many a nobleman.”
(cf. MF “Sans la mort de beaucoup de chevaliers.”)
 (La Queste del Saint Graal 18:27)
b. qui aus piez [du lit]
who at.the-obl-m-pl feet-obl-m-pl of.the-obl-m-sg bed-obl-m-sg
se gisoit
himself lay
“Who at the foot of the bed lay”
(Cf. MF “au pied du lit”) (Vergi 727)â•… (Herslund 1980:€85)

Such genitive constructions are still possible in MF.3 Like the genitive with à, the pos-
sessor in the genitive with de can be indefinite (3a) or definite (3b).

.╅ Due to space limitations, we do not address the differences between the distribution of
genitives with à in OF and MF. See Arteaga & Herschensohn (2010) for discussion.
 Deborah Arteaga & Julia Herschensohn

Despite the similarity of these two types of genitive constructions (à or de intro-

ducing the complement) in OF and MF, it is not the case that MF inherited all of OF’s
genitive structures. Consider (4a)–(4b) below:

(4) a. [la volenté Nostre Seignor]

the-nom-f-sg will-nom-f-sg our-obl-m-sg lord-obl-m-sg
“The will of Our Lord”
(cf. “la volonté de notre seigneur”) (La Queste 9:3–4)
b. [La niece le duc] manoit
the-nom-f-sg niece-nom-f-sg the-obl-m-sg duke-obl-m-sg remained-3sg
“The duke’s niece remained.”
 (La Chasteleine de Vergi 376)╅ (Foulet 1982:€14)

These structures have been labeled the juxtaposition genitive (JG) by philologists
(see Arteaga 1995 for discussion). The head noun (i.e. la volenté in (4a) and la niece
in (4b)) is the possessed. The second noun (i.e. nostre seignour, le duc) is the pos-
sessor DP. While oblique case is always assigned to the possessor DP, case assignment
in the JG varies for the head DP, according to its syntactic function. This is illustrated
by (5a)–(5b):

(5) a. C’ est [li filz le roi d’ Arragon]

it is-3sg the-nom-m-sg son-nom-m-sg king-obl-m-sg of Arragon
“He was the son of the king of Aragon.”
 (Lancelot 5780)╅ (Herslund 1980:€138)
b. Il jure Dieu,
he swears God-obl-m-sg
[le fil Sainte Marie]
the-obl-m-sg son-obl-m-sg sainte-obl-m-sg Mary-obl-f-sg
“He swears to God, the son of Saint Mary.”
 (Aspremont 946)╅ (Palm 1977:€90)

In (5a) the possessed li filz receives nominative case, as it follows the verb est. In (5b),
however, le fil is the object of the verb jurer, and is therefore assigned OBL case.
Note that traditional philologists have maintained that the JG is found where
the possessor was an individual, a proper noun, or a noun preceded by a definite
article (inter alia, Jensen 1990; Moignet 1976; Ménard 1988). It is recognized that the
possessor must be [+definite, +human] (cf. Herslund 1980; Moignet 1988; Ménard
1988; Jensen 1990; Delfitto & Paradisi 2009, among others). The possessed in the JG,
however, can be either [+def] or [-def], although the former is more common. Consider
(6)–(7) below:
 A phase-based analysis of Old French genitive constructions 

(6) des chevaliers le roi Artu

some knights-obl-m-pl the-obl-m-sg king-obl-m-sg Arthur.-obl-m.sg
“(Some) knights of King Arthur.” (Mort 5.6)â•… Herslund (1980:119)
(7) Une image Mahon
an-obl-f-sg image-obl-m-sg Mahon-obl-m-sg
“An image of Mahon.” (Ogier 9634)â•… Herslund (1980:104)
In addition to genitive structures in which the order head noun + possessor with
no introducing preposition is found, there is another possibility, namely the pre-
nominal position of the possessor with no intervening preposition (order: article of
head noun + possessor + head noun]. Consider (8a) and (8b):
(8) a. en [la rei prisun]
in the-obl-f-sg king-obl-m-sg prison-obl-f-sg put
“In the prison of the king.” (Becket, 1121)â•… (Palm 1977:€54)
b. [li Deu ami]
the-nom-m-pl God-obl-m-sg friends-nom-m-pl
“The friends of God” (Eust prose 38.3)â•… (Herslund 1980:€119)

In both (8a) and (8b) above, the genitive possessor (rei, Deu) has been preposed. Note
that the article in (8a) agrees in case and gender with the nominative singular head
noun prisun and not with the genitive rei (nominative: reis). In other words, the order
is article of head noun + possessor noun + head noun. Similarly, in (8b) li is assigned
nominative case, as is ami. In both examples, the preposed possessor receives oblique
case. Occasionally, however, the determiner is that of the prenominal genitive itself, as
in (9a)–(9b) below:4
(9) a. Par [le dieu d’ amours voulenté]
by the-obl-m-sg god-obl-m-sg of love-obl-f-sg will-obl-f-sg
“By the will of the god of love.” (Cace as mesdis 235)â•… (Tobler 1902:€69)
b. [Son pere kierue] menoit
his-obl-m-sg father-obl-m-sg plow-obl-f-sg drove-3sg pro drove
“He drove his father’s plow.”
 (Chronique rimée de Philippe Mouskes 17047)â•… (Tobler 1902:€70)

Recall that in (8a)–(8b) above, the order is article of head noun + possessor noun +
head noun. However, in (9a) and (9b) we see a different pattern. In (9a), the mas-
culine singular article le clearly refers to dieu (masculine oblique) and not voulenté
(feminine). In (9b), the masculine singular possessive adjective son, modifies pere

.â•… Cited in Arteaga (1995). Note also that the entire NP le dieu d’amours raises; similarly,
NPs with adjectives raise of a piece.
 Deborah Arteaga & Julia Herschensohn

and not kieure, which is feminine. The syntactic order of both examples is article of
possessor + possessor noun phrase+ head noun).
To summarize our discussion thus far, in OF, in the postnominal and prenominal
JG, the case assigned to the possessed head noun depends on its syntactic function,
whereas the case of the possessor is always oblique. There is no intervening preposi-
tion between the juxtaposed DPs. The possessor DP is always [+definite, +human],
whereas the possessed can be either [±definite]. Further, the definite article present in
such constructions can either be that of the head noun or its possessor.
One final note about all genitive constructions in OF: they can be iterated. Con-
sider (10):
(10) [[[le prei] [les oirs]
the-obl-m-sg meadow-obl-m-sg the-obl-m-sg heirs-obl-m-pl
[le Pelletier]]]
the-obl-m-sg Pelletier-obl-m-sg
“Pelletier’s heirs’meadow” Lanher (1975:117.5)â•… Holman (1992:142)
The example in (10) above shows iteration of the genitive structure. The example in (11)
illustrates the fact that in combinations of genitive structures, the preposed forms
could alternate with the JG, provided that the possessor was definite:
(11) les cors des amis
the-obl-m-pl hearts-obl-m-pl of.the-obl-m-pl friends-obl-m-pl
Nostre Seignor
our-obl-m-sg lord-obl-m-sg
“The hearts of our Lord’s friends.” (Eust. prose 38.15)â•… (Herslund 1980:€119)
The example in (11) above combines the prepositional genitive les cors des amis with
the JG amis Notre Seignor.
Therefore, a synchronic analysis of the OF data leaves us with the following three
research questions:
i. What is the derivation within the Minimalist framework of the prepositional
genitive?
ii. What is the derivation of the postnominal and prenominal JGs?
iii. How does the possessor noun get its oblique case?
We next turn to the theoretical framework that we assume.

3.â•… Theoretical framework

In this section we first review relevant assumptions of minimalism and recent analyses
of DP within this framework. We then discuss two earlier treatments of OF genitives,
those of Arteaga (1995) and D & R (2009).
 A phase-based analysis of Old French genitive constructions 

3.1â•… Minimalist analyses of DP

Recent work has advocated a phase-based approach to a variety of phenomena, includ-
ing DP (cf. Chomsky 2006; D & R 2009). For example, D & R, in a discussion of Italian
past participle agreement, assume an Agree relationship between a Probe and Goal in
which an interpretable feature matches and deletes an uninterpretable feature within
the minimal phase head as in (12) (from D & R 2008:€482).

(12) a. Given an Agree relation A between probe P and goal G,

morphophonological agreement between P and G is realized iff P and G
are contained in the complement of the minimal phase head H.
b. XP is the complement of a minimal phase head H iff there is no distinct
phase head H’ contained in XP whose complement YP contains P and G.

Pesetsky & Torrego (2004, henceforth, P & T) describe this P–G relation as Merge,
External Merge combining two independent elements to form a larger one and Inter-
nal Merge moving an element in the domain of another. Merge is driven by checking of
features between the two elements as D&R also indicate. Only unvalued features may
probe, and the converging check may result in agreement between the two internally
merged elements. In some cases – such as subject-verb agreement where the subject
carries uninterpretable case and the verb uninterpretable person – the two elements
carry complementary interpretable and uninterpretable features. Just as iterated v
checks features related to verbal arguments as V may raise and merge with v, iterated n
checks nominal features such as Case, gender, etc. as N raises and merges with n.
We adapt the idea of a K feature relating to abstract Case (Bittner & Hale 1996).5
Kayne (2005: 138) uses K to describe an agreement feature, “K-of,” a feature close to
genitive or dative. We consider that K is a T-feature in the spirit of P & T (2009:9). They
extend Bittner and Hale’s comparison of the nominal and verbal domains by exploring
further similarities and differences in patterns of complementation and clausal modi-
fication. In P & T (2004), they argue that prepositions bear T-features similar to tense,
features “which, when interpretable, situate or characterize an entity in either time or
space (P)” (2009:26). T is not simply an indication of temporality, but is an abstract
grammatical feature that grounds the DP in reference while syntactically licensing
the complement. P&T note that complements of N must be headed by interpretable T,
a feature we designate as [K].
As for the structure of DP, Carstens (2003) proposes as a complement to D an
iterated nP (Carstens 2003), potentially containing possessors of the head N, as well
as other functional projections (e.g. number, quantifier). The nP acts much as the vP,

.â•… Bittner and Hale (1996:4) propose that KP is “the maximal extension of the nominal
projection, whereas C represents the maximal extension of the verbal projection.”
 Deborah Arteaga & Julia Herschensohn

carrying the functional nominal features that the lexical noun needs to be grammati-
cally implemented in the DP and eventually the sentence. The iterated nP can serve
functions designated as separate functional projections such as gender and number in
earlier analyses of DP (e.g. Bernstein 1991; Longobardi 1994; Mallén 1997), accom-
modating raising of the head noun or of nominal modifiers.
Lin (2008) considers DPs in English and Mandarin. Following Pereltsvaig’s (2007)
defense of the universality of DP, Lin claims that the syntactic structure for the nominal
phrase DP is universal regardless of the presence of determiners in a given language. Thus
the syntactic structure of DPs in English, Russian and Mandarin would be identical. Lin
notes that DPs in English and Mandarin differ in terms of definiteness, gender, and num-
ber. For example, the phrase in English in (13a) must have a definite article to express the
feature Definite, whereas the same in Mandarin (13b) does not (Lin 2008:€776):

(13) a. English: The dog wants to cross the road

b. Mandarin: Dog want cross road.

In order to account for this syntactic difference, Lin proposes that there is invariably
an interpretable [+Definite] feature on D in Mandarin, which is only present in English
if the noun has an overt definite article. The [Def] feature on N is uninterpretable,
meaning that D and N are in a Probe-Goal relationship (cf. P & T), in which the [Def]
feature on D matches and deletes the uninterpretable [Def] of N. Lin postulates N to
n movement (Carstens 2000), followed by movement to the Spec of DP. We next turn
to prior accounts of the JG.

3.2â•… Previous analyses of the JG

3.2.1â•… Arteaga (1995)â•… Within a Government and Binding framework, Arteaga
(1995) claims that DPs in OF have lexical AGR (cf. Contreras 1992). This lexical AGR
can discharge features to the right, directly case-marking genitive complements in the
JG without the need for a preposition. The prenominal JG has the same structure as
the postnominal, in Arteaga’s analysis. With respect to the prenominal JG, she assumes
fronting of the lower NP which then adjoins to AgrP of the higher DP. To account for
this movement, she assumes that null AGR0 fronts to null D0, thereby obviating the
barrierhood of A′ by L-marking. If the lower D is null, the order la roi prison is obtained,
whereas if it is the higher D that is null, le roi prison may be found.
Weaknesses of Arteaga’s account include stipulative movement of null D to null
AGR only in the event that no D is expressed. One serious problem for her analysis
is the prenominal JG of the type la roi prison. She assumes movement of the lower
NP through the SPECs of AgrP and DP. Yet if the final landing site for the lower NP
is, indeed, the SPEC of DP, then the order *roi la prison would be obtained. In other
words, her analysis cannot yield the correct word order in cases like la roi prison.
 A phase-based analysis of Old French genitive constructions 

Further, Arteaga does not address the fact that the possessor complement in the JG is
invariably definite, nor does she relate the prepositional genitive constructions directly
to the JG.

3.2.2â•… Delfitto & Paradisi (2009)â•… D & P (2009) provide an analysis of genitive struc-
tures in OF, Old Italian, and Romance in general.6 They note that in OF, the possessor
in genitives may be preceded by the prepositions à or de (cf. (2–3)) above. Regarding
the distribution of these structures, they argue that à genitives are found only with
indefinite possessors, whereas de may be found in any and all genitive environments.
With respect to the JG, D & P limit their discussion to the postnominal JG (cf.
(4–7)) and relate the OF construction to similar genitives in other Romance and
Semitic languages. They claim that JGs are like genitives with à, in that they cannot be
iterative (unlike de genitives). They adopt Kayne’s (1993) antisymmetric structure for
DP in which the possessor precedes the head noun and in English can raise to spec
DP, giving the Saxon genitive (cf. Adger 2003, 257–258). Their structure of possessive
constructions is as in (14), in which the head N moves from an IP to a position in
spec D/PP.7

(14) la[D/PPvoiturej [de [IP Jean [AGR0 [e]j …

Following Kayne, D & P assume preposition insertion in the above structure. They
argue that in MF, possessive morphology is “completely silent,” which is why they do
not postulate any category, such as K, that would assign objective case. For the OF JG,
D&P (2009:€299) propose the structure in (15) whereby the head N raises from IP, as
does AGR/K, the “agreement-case morphology associated with the possessor con-
stituent” into D0.

(15) la [D/PP niècej [[AGR/K0k -D0][IP le duc [ek [e]j …

In (15), D & P assume an AGR/K phrase that assigns objective case to the possessor in
the OF JG. As the example in (15) illustrates, they further assume the underlying order
le duc la niece. They propose that first nièce, the head noun, raises as “NP-raising to
Spec-D0,” although nièce visibly follows the D la. Subsequently the AGR/K head raises
and adjoins to D before the possessor DP located in IP. They further claim that in the JG,

.╅ They also consider Semitic construct-state compounds, a construction beyond the scope
of this paper.
.â•… Kayne (1993, 102) uses the symbol D/P “to represent a prepositional determiner de (com-
parable to a prepositional complementizer).”
 Deborah Arteaga & Julia Herschensohn

both the possessed and the possessor must be definite. The JG cannot have multiple
iterations, they argue, which means that it is “an instance of a functional genitive.”
While the account of D & P is a serious attempt to link genitive structures in various
Romance languages, both synchronically and diachronically, it does have limitations.
First, the authors do not explain their choice of categories (D/PP, IP, K), nor do they
motivate the raising of the head N, or even clarify the landing site. This is crucial,
because the order *le duc la niece “the duke’s niece” (literally “the duke the niece”) is
not attested. They assume AGR/K in the JG but not in prepositional genitives, in part,
because there is no overt instantiation of possessive in MF. However, they overlook
instances such as ma voiture à moi (“my car,” literally “my car to me”), in which the
pronoun moi is clearly in the objective case. Prepositional insertion in OF is prob-
lematic, because there is overlap between all three genitive structures. Moreover, their
analysis does not extend to the prenominal JG, as in (8)–(9).8 In the next section, we
propose an alternative analysis of OF genitive constructions within the Minimalist
Program that addresses shortcomings of the two articles discussed above.

4.â•… A new proposal

While the analyses of Arteaga (1995) and D&P (2009) can explain many aspects of
OF genitives, from empirical and explanatory standpoints, there are gaps that need
to be addressed, as discussed in the preceding section. We propose instead a phase-
based account of the genitive which only employs a minimal number of independently
motivated projections (DP, nP, NP, PP), features ([def], [hum], [K]), and operations
(Merge, Agree), all well established and accepted components of minimalism.
For all genitive structures (prepositional and JG) in OF, we assume the structure
in (16):

(16) [DP D [nP … [NP N [KP[K/P [DP D [nP … [NP N]]]]]]]]

Our analysis assumes the Probe-Goal model of feature checking as well as an iter-
ated nP (Carstens 2000, 2003), needed independently for structures such as (10)–(11)
above. The Probe-Goal relationship of n/N and D is complex. Gender and number are

.â•… Furthermore, many of their claims, such as the lack of iteration of JG or genitive with à,
the notion that à genitives are almost always indefinite, and their observation that the pos-
sessor in the JG is almost always masculine, can simply not be reconciled with the data, as
there are counterexamples. Discussion of these issues is beyond the scope of this paper.
 A phase-based analysis of Old French genitive constructions 

uninterpretable on D (Probe) while they are interpretable on n/N (Goal). Conversely,

[def] is uninterpretable (Probe) on n/N yet interpretable on D (Goal). Following P & T
(2009), we assume that in OF prepositional genitive structures, the relational character
of à/de will carry an interpretable T-feature [K] that can value and delete the ucase fea-
ture of the oblique N. In prepositional genitive structures, as we have seen in (2a) and
(3a), the possessor may be indefinite, in which case D carries a [-def] feature.
The derivation, therefore, of these structures is as follows: N raises to n to check
features and case (cf. Carstens 2000, 2003) The oblique case of the possessor noun is
checked and deleted by K/P, whose interpretable T-feature allows it to license com-
plement DPs. The n/N carries an uninterpretable feature [udef] which is valued and
deleted by the [def] feature on D. The features of gender and number are checked and
valued by n.
Genitive constructions with à or de persist in MF. In contrast, the derivation of the
OF JG appears to be problematic since there is no obvious overt Goal to check [uCase]
of the OBL Probe possessor DP. Moreover, we have to account for both the prenominal
and postnominal word orders, in addition to the fact that the definite article can be
that of either the head noun or the complement.
For the canonical JG such as (4b) la nièce le duc, we assume that the JG also has
the structure above in (16). Recall that such structures have interpretable [+def] on the
D and interpretable [+hum] on the possessor N. We propose that the K/P head here
has no overt reflex (as in overt P), but must also carry the features [uhum] and [udef]
since those two features of the JG possessor provide sufficient reference to “ground”
the DP complement (cf. P&T). The [hum] and [def] features of K/P and DP check
and delete, while the interpretable T-feature of the null K head deletes the [ucase-obl]
feature of the possessor DP. For the postnominal JG, no movement other than n to N is
postulated, an independently motivated analysis, quite simpler than those of Arteaga
(1995) and D&P (2009).
How are prenominal JG structures derived? In our analysis, we derive the prenomi-
nal juxtaposition from the postnominal one. Recall, that in prenominal JG structures
(cf. (8–9)), the possessor precedes the possessed, and the definite determiner can be
that of either noun. We propose that in these constructions one of the DPs (possessor
or possessed) constitutes a defective phase (Chomsky 2001, 9), with a missing D. Bare
nouns exist in a variety of constructions in OF (unlike MF, which prohibits bare NP),
so that a defective phase is needed independently (see Arteaga & Herschensohn 2010).
In the case at hand of the JG, we assume that the defective phrase is not a complete
generic NP, but rather an N missing its D (hence defective). The rather unique (and
soon to be lost) JG provides a means for saving these defective phases. Clearly, a
similar construction in MF would not have a means for saving them, and the deriva-
tion would crash. This proposal, in turn, means that there is a missing D in either
 Deborah Arteaga & Julia Herschensohn

the possessor or the possessed DP, triggering movement of either the lower nP (la rei
prisun, as in (8a)), or the lower DP (le dieu d’amours in (9a) to the higher position:
To illustrate, consider (17), which is the structure of (8a); strikethroughs on ufea-
tures indicate features which have been matched and deleted:

(17) = (=8a) la rei prisun

DP

D nP1

la
[def, ugen,
unum, ucase-obl] nP2 nP1

reij
[udef, gen-m,
num-sg, ucase-obl] nP1

n NP

prisuni
[udef, gen-f, num-sg, ucase-obl]
N K/PP

[e]i K/P nP2

[T] NP

[e]j

With respect to (17), la rei prisun, we assume that the higher and lower N’s front to n1/
n2, respectively. This movement allows for the checking of number and gender in both
N’s, as well as deleting ucase of rei by interpretable T in K. Further, the [udef] feature of
both the upper N/n, prison, but not the lower, is checked; the oblique case of prisun is
checked outside the DP by the D la. rei is in a defective phrase, which lacks D and this
possessor noun is therefore forced to raise to an intermediate upper nP1 (between the
upper D and upper N). By means of this fronting, the lower N rei gets its uninterpre-
table [def] feature interpreted and valued by the upper D, la. It is this upper D, la, that
prevents the derivation from “crashing.” Crucially, our analysis provides for a landing
site for roi (contra Arteaga 1995 and D & P 2008).
In the structure in (9a), illustrated by (18) below, the article that is found is that of
the lower DP (dieu) and not voulente, which is feminine:
 A phase-based analysis of Old French genitive constructions 

(18) (=9a) = le dieu d’amours voulente

DP1

DP2j nP1

D2 nP2 n NP

le voulentéi N DP2

[def, uge [udef-f-sg, uobl] [e]i [e]j

unum, uease-obl] n2 PP

dieu d’amours
[udef-m-
sg-hum uobl]

Deleted ufeatures are indicated by strikethroughs in (18). In structures like (18), in

which the article of the possessor, le, is found, we again propose that as a first step, the
higher and lower N’s front to n1/n2, respectively. By fronting to n, the nouns can check
the gender and number of N. Moreover, by this movement, the lower N gets its unin-
terpretable [def] feature interpreted and valued by the lower D and its case checked
by interpretable T in K. While the lower DP is a licit genitive phase, the upper one is
defective, missing a determiner. Therefore, the entire DP2 moves upward to remedy
the defective higher DP (missing D cannot check off udef of voulente). By fronting, the
structure is saved, as the raised D le can value and interpret the definite feature of both
the head noun and the possessor.9 In either instance (17 or 18), if the D is unable to
check the two [udef] features, the derivation would crash.
Our analysis informs our current understanding of the Minimalist Program for
several reasons. We have provided independent evidence that the use of interpretable
and uninterpretable features, a hallmark of the MP, can account for a wide variety of
syntactic structures, in this case the OF genitive. Our analysis further confirms that of
Carstens (2000, 2003) in which a light shell nP is postulated. We show that this nP is
needed independently to account for iteration of genitive structures in OF (type [[[le
prei] [les oirs] [le Pelletier]]] “the meadow of Pelletier’s heirs.” Finally, we show that by
the use of a defective phase, we can account for preposed juxtaposition genitives (type
le rei prison/la rei prison “the king’s prison”.

.╅ See Carstens (2000:327) who also argues that syntactic phrases can check more than
one feature.
 Deborah Arteaga & Julia Herschensohn

In this section, we have made a number of claims regarding the structure and
derivation of OF genitive constructions. Iteration of genitive structures has been
accounted for by incorporating Carsten’s nP shell analysis. All D’s, within this analysis,
have interpretable [def] and uninterpretable number and gender. N’s, on the other
hand, have uninterpretable [def], uninterpretable case, as well as interpretable number
and gender. The genitive constructions with and without introducing prepositions are
very similar, in our analysis. Those introduced by à or de are assigned oblique case by
K/P, which is a T feature. JGs have a null P in a possessor K/PP that carries a T-feature,
[uhum] and [udef], features that match and delete with those of the complement N
and D. Finally, fronting in the JG is motivated by a defective DP (missing D) in either
the possessor or the possessed.

5.â•… Conclusion

This article has argued that OF genitives – those with prepositional à and de and the
JG – all have an identical underlying construction. In the former, P’s carry a T-feature
allowing them to check the oblique case of the possessor through the T-feature. In our
analysis, all overt D’s carry interpretable [def] and uninterpretable case, number,
and gender, whereas N’s carry uninterpretable [def] and case, as well as interpretable
number and gender.
In the case of JGs, we have posited a null P (K/PP). Our analysis can account for
the derivation of both postnominal and prenominal juxtaposition structures. Our gen-
eralization is that where D is missing, in either the possessor or possessed, n is found
in a defective phase. The lower n fronts when it lacks a determiner that can value and
interpret the feature [udef], whereas the entire lower DP, including the determiner,
fronts in the event that the upper n lacks a determiner. We have shown that a phase-
based analysis can derive the various OF genitive constructions in a principled way.

References

Adger, David. 2003. Core Syntax: A Minimalist Approach. Oxford: Oxford UP.
Arteaga, Deborah. 1995. “On Old French Genitive Constructions,” Contemporary Research in
Romance Linguistics, ed. by Jon Amastae, Grant Goodall, Mario Montalbetti & Marianne
Phinney, 79–90. Amsterdam: Benjamins.
Arteaga, Deborah & Julia Herschensohn. 2010. “Diachronic Considerations of Genitives from
Old to Modern French,” Ms.
Arteaga, Deborah & Julia Herschensohn. “Les noms-têtes vides en ancien français”, Ms.
Bernstein, Judith. 1991. “DPs in French and Walloon: Evidence for Parametric Variation In
Nominal Head Movement”. Probus 3: 101–126.
 A phase-based analysis of Old French genitive constructions 

Bittner, Maria & Ken Hale. 1996. “The Structural Determination of Case and Agreement”.
Linguistic Inquiry 27: 1–68.
Carstens, Vicki. 2000. “Concord in Minimalist Theory”. Linguistic Inquiry 31: 319–355.
Carstens, Vicki. 2003. “Rethinking Complementizer Agreement: Agree with a Case–Checked
Goal”. Linguistic Inquiry 34: 393–412.
Chomsky, Noam. 2001. “Derivation by Phase”. Ken Hale: A Life in Language, ed. by M. Kenstowics,
1–52. Cambridge, Mass.: MIT Press.
Chomsky, Noam. 2006. “Approaching UG from Below”. Unpub. Ms., MIT.
D’Alessandro, Roberta & Ian. Roberts. 2008. “Movement and Agreement in Italian Past Participles
and Defective Phases”. Linguistic Inquiry 39: 477–491.
Delfitto, Denis & Paola Paradisi. 2008. “Towards a Diachronic Theory of Genitive Assignment
in Romance”. Historical Syntax and Linguistic Theory ed by P. Paola Crisma & Giuseppe
Longobardi, 292–310. Oxford: Oxford UP.
Hartmann, Katharina & Malta Zimmermann. 2003. “Syntactic and Semantic Adnominal Geni-
tives”, (A)symmetrien – (A)symmetric ed. by Claudia Maienborn, 171–202. Tübingen:
Stauffenburg Verlag.
Herslund, Michael. 1980. Problèmes de syntaxe de l’ancien français. Compléments datifs et génitifs.
Uppsala: Akademisk Forlag.
Holman, Robyn. 1992. “The Syntax of the Genitive Structure in Thirteenth Century Vosgian
Charters”. Romance Notes 23.141–149.
Jensen, Frede. 1990. Old French and Comparative Gallo-Romance Syntax. Tubingen: Max
Niemeyer Verlag.
Joseph, John. 1995. “New French: a Pedagogical Crisis in the Making”. The Modern Language
Journal 72.31–36.
Kayne, Richard. 1993. The Antisymmetry of Syntax. Cambridge, Mass.: MIT Press.
Kayne, Richard. 2005. Movement and Silence. Oxford: Oxford€UP.
Lin, Yi-An. 2008. “A Probe-Goal Approach to Parametric Variation in English and Mandarin
Chinese Nominal Phrases”. Proceedings of the 20th North American Conference on Chinese
Linguistics (NACCL-20) ed. by Marjorie Chan & Hana Kang, vol II, 775–784. Columbus,
Ohio: The Ohio State University.
Longobardi, Giuseppe. 1994. “Reference and Proper Names: A Theory of N-Movement in Syntax
and Logical Form”. Linguistic Inquiry 25: 609–665.
Mallén, Enrique. 1997. “A Minimalist Approach to Concord in Noun Phrases”. Theoretical
Linguistics 23: 49–77.
Ménard, Pierre. 1988. Syntaxe de l’ancien français. Paris: Bordeaux Éditions Bière.
Moignet, Guy. 1976. Grammaire de l’ancien français. Paris: Klincksieck.
Pereltsvaig, Asya. 2007. “The Universality of DP: A View from Russian”. Studia Linguistica
6: 59–94.
Pesetsky, David & Esther. Torrego. 2004. “Tense, Case and the Nature of Syntactic Categories”.
The Syntax of Time ed. by Jacqueline Guéron & Jacqueline Lecarme, 495–537, Cambridge,
Mass.: MIT Press.
Pesetsky, David & Esther. Torrego. 2009. “Probes, Goals and Syntactic Categories”. Proceedings
of the 7th Annual Tokyo Conference on Psycholinguistics ed. by Y. Otsu. Tokyo: Hituzi Syobo
Publishing Company.
 V2 loss in Old French and Old Occitan
The role of fronted clauses*

Barbara Vance1, Bryan Donaldson2 & B. Devan Steiner1

1Indiana University/2University of Texas at Austin

Traditional analyses of Old French as a verb-second (V2) language

(e.g. Thurneysen 1892; Adams 1987) have recently been challenged by Kaiser
2002 and related work. At issue is the treatment of situations in which a particular
initial non-subject element can participate in either V2 or V3 order. The current
paper focuses on sentences in which the initial element is a fronted subordinate
clause. Using a diachronic data base of 13th-century prose, we argue first for
a revision in the criteria for identifying V3 that reduces the number of such
examples considerably. We then show that the rate of V2 vs. V3 with fronted
clauses varies with respect to date, genre, and fronted-clause type in ways that
suggest a syntactic change in progress rather than the absence of V2 effects. Our
conclusion is supported by a parallel study of a closely related (but minimally
syntactically different) language, Old Occitan.

1.  Introduction: V2 and fronted subordinate clauses

Although a sizeable literature has grown up around the claim that Old French (OFr)
was a verb-second (V2) language (e.g. Thurneysen 1892; Vanelli, Renzi & Benincà
1985; Adams 1987; Roberts 1993; Vance 1997, among many others), Kaiser 2002 has
suggested that this claim is unwarranted (see also Ferraresi & Goldbach 2002; Rinke
2003; Rinke & Meisel 2009). Kaiser points out that, compared to a living V2 language like
Modern German, OFr is unexpectedly tolerant of descriptively non-V2 clauses. This
contrast is given in (1) and (2): in German, a fronted subordinate clause1 obligatorily

*We thank the LSRL 39 audience and three anonymous reviewers for helpful discussion.
Richard Janda kindly helped us with the German examples. Naturally, we take full responsi-
bility for any remaining errors or shortcomings.
.  We take no theoretical stand here on whether such clauses are moved or base-generated
in initial position.
 Barbara Vance, Bryan Donaldson & B. Devan Steiner

triggers subject inversion like other initial constituents. In OFr, the fronted clause may
either count as the initial constituent in a V2 clause (2a) or not (2b).
(1) Modern German (adapted from Kaiser 2002)
a. Wenn die Frau Zeit gehabt hätte, hätte sie ein Buch gelesen.
if the woman time had had, would-have she a book read
“If the woman had had time, she would have read a book.”
b. *Wenn die Frau Zeit gehabt hätte, sie hätte ein Buch gelesen.
(2) Old French
a. Et se vous l’avez entendu, ne l’avez vous pas detenu.
And if you it-have heard, neg it-have you not retained
“And (even) if you have heard it, you have not retained it.”
 (Cassidorus p. 375)
b. se il nous en vuelent mener, nous irons volentiers.
if they to-us prt want to-bring we will-go willingly
“if they want to take us along, we will go willingly.” (Villehardouin l. 181)

Kaiser interprets this difference as partial evidence that OFr was not a true V2 language.
We argue, on the contrary, that examples like (2b) show that the OFr V2 constraint was
already breaking down in the language represented by 13th century prose and that the
absence of inversion after a fronted subordinate clause represents the earliest stage of
a major change in French clause structure completed – for all types of fronted con-
stituents – by the Renaissance.2 More precisely, we show that, during the 13th century,
the ratio of V2 to V3 word orders with fronted subordinate clauses varies according to
(a)€date, (b) fronted clause type, and (c) genre, in ways that suggest a language change
in progress. For comparative purposes, we also examine the history of Occitan, the
language of southern France, which has also lost apparent V2 properties over time.
The paper is organized as follows. Section 2 provides background on the contro-
versy regarding the analysis of OFr as a V2 language. Section 3 presents our empirical
study of the effect of fronted subordinate clauses on following word order. Section€4
presents a preliminary investigation of fronted subordinate clauses in Old Occitan
(OOc) and a comparison with OFr, and Section€5 concludes the paper.

2.  Old French as a V2 Language

Descriptively speaking, a V2 language is one in which the finite verb must occupy the
second position in main clauses. The first position of the clause may be filled either by

.  This idea is also expressed briefly in Vance (1997:€64–65) and Becker (2005:€357).
 V2 loss in Old French and Old Occitan 

the subject or by a variety of other constituents (e.g. adverbs, prepositional phrases,

direct objects, subordinate clauses);3 if the subject is not chosen as the initial con-
stituent it must follow the verb. This system has been neatly captured theoretically by
e.g. den Besten (1983) as the result of verb movement to the complementizer node€(C)
and an accompanying filling of its specifier; the fact that C is blocked by a subordinator
in embedded clauses explains the root nature of the phenomenon. Vanelli, Renzi€&
Benincà (1985) and Adams (1987) show that a similar analysis can be applied to OFr,
and various subsequent studies on both Germanic and medieval Romance have
revealed many fine details that are strikingly similar in the two language groups.4
Adams’ structure (slightly updated) for OFr is given in (3), where (a) represents V2 in
a main clause and (b) represents embedded clauses:
(3) a. [CP XP [C′ Vi [+fin] [TP (subject) [T′ ti [VP … ti …]]]]]
b. [CP [C′ que [TP subject [T′ V[+fin] [VP …. ti …]]]]]

One difference between OFr and the modern Germanic languages is that OFr allows
referential subjects to be null; however, the V2 grammar interacts with this property in
a very specific way. Null subjects are only possible in OFr if the subject would be post-
verbal if expressed (Foulet 1928) and therefore only in main clauses where V moves
to€C (Adams 1987; Vanelli et al. 1985). Another property of OFr is that it has two types
of subject inversion. Here we consider (following Roberts 1993 and Vance 1997) that
pronominal subjects in OFr cannot occur lower than Spec TP and therefore, when
postverbal, provide unambiguous evidence of V-to-C movement (i.e. “Germanic” or
“V2” inversion). Inverted nonpronominal subjects can occur lower than TP (as in
“Romance” inversion), even though V also moves to C, and therefore do not clearly
mark the raised position of V on their own.5
The recent controversy in the generative literature over the V2 status of OFr
appears to turn mostly on the interpretation of so-called verb-third (V3) clauses, since
it is agreed that verb-initial clauses are legitimate under special discourse conditions.

.  Neither coordinating conjunctions nor object clitics are counted in the calculation of V2.
.  We will not have room to discuss here the very interesting proposal of Ferraresi & Goldbach
2002 to treat OFr as a VSO language similar to Celtic.
.  In Germanic inversion, postverbal subjects usually occur between the auxiliary and the
past participle, whereas in Romance inversion, they normally occur after the past participle.
However, exceptions arise due to independent possibilities for leftward movement of par-
ticiples (as pointed out by Rinke & Meisel 2009) and variation in the possible positions of
subjects. Furthermore, clauses with only one verb are ambiguous. Vance (1997) claims that
information structure nevertheless disambiguates the two types of inversion. Rinke & Meisel’s
recent use of information structure to argue against V2 status for all OFr subject inversion will
need to be examined in future work.
 Barbara Vance, Bryan Donaldson & B. Devan Steiner

For adherents to Kaiser’s approach, the presence of any type of V3 examples in a corpus
is significant and problematic for a V2 analysis. In most other works, such examples
are considered marginal and are treated as consistent with a theoretically-driven
approach to V2 even if they do not, strictly speaking, adhere to the descriptive stan-
dard. The most extreme of these latter approaches is that of Benincà (1995, 2006), who
considers V>2 orders to be a natural part of V2 in Medieval Romance because of the
combination of V-to-C movement and a complex left-periphery allowing for the filling
of multiple specifier positions. However, since even Benincà acknowledges that OFr
has a rather strict descriptive V2 order relative to other medieval Romance varieties,
we will delay treatment of her position until Section€4, where it becomes relevant for
our comparison of OFr to OOc. For purposes of the present discussion, we need to
distinguish between two very frequent types of descriptively V3 clauses: (1) [[fronted
clause] SV] constructions that alternate with [[fronted clause]V(S)] and therefore may
reflect the beginnings of a non-V2 grammar, as in (2b) above; and (2) examples in
which a fronted clause is followed by the adverb si (“thus”, “then”) or a similar adverb
(e.g. donc/adonc “then”, lors “then”) and V(S):
(4) a. Quant je avrai les dois d’une main ars, si ardrai
when I will-have the fingers of-one hand burnt, then will-burn
je les autres.
I the others
“When/if I burn the fingers of one hand, then I will burn the others.”
 (Cassidorus p. 341)
b. si tost conme il virent le jour, si cueillirent leur voiles …
so soon as they saw the day then gathered-3pl their sails
“As soon as they saw daylight, they gathered their sails …”
 (Villehardouin l. 1623)

In purely descriptive terms, both (4a) and (4b) appear to be V3 clauses.6 However, the
postverbal subject pronoun in (4a) can be generated only by a V-to-C grammar; (4b)
likely has parallel structure. The “resumptive” adverb si appears to sum up the content
of the fronted clause and then function as the actual initial element of the main clause.
From this point of view, such sentences, which could not be generated by a simple
SVO grammar, confirm the general V2 nature of OFr rather than providing evidence
of a non-V2 grammar. Significantly, similar examples are also found in Modern German,
even if they are not particularly common. Kaiser himself gives two examples from
modern German Bible translations (2002:142), in which he attributes the use of the

.  Note that these examples could be considered descriptively V2 under analyses such as
Ferraresi & Goldbach 2002 or Ledgeway 2008 in which si is treated as a special grammatical X0
element. However, we prefer the traditional adverbial analysis of si (see also Poletto 2005) for
reasons we cannot go into in this paper.
 V2 loss in Old French and Old Occitan 

resumptive adverb dann ‘so’ to the unusual length of the fronted material. Our own
internet search turned up many examples (with dann, so, or da) from a variety of con-
texts and without especially long fronted subordinate clauses:
(5) Wenn man seinen Wohnsitz im Ausland anmeldet, so muss man
if one his residence in abroad declares thus must one
ja auch sein Auto umschreiben lassen.
surely also his car re-register let
“If you declare your (primary) residence as being abroad, then you must of
course also have your car [registration] transferred.”
 (from the financial advice site “Finanzfrage.net”)

In previous studies of fronted subordinate clauses in OFr, two main approaches have
been taken to classifying data such as these. Skårup 1975 considers the true initial con-
stituent in examples like (4) to be the adverb si; the preceding fronted clause is simply
irrelevant for V2. On the other hand, adherents to Kaiser’s approach (Kaiser 2002;
Rinke 2003; Rinke & Meisel 2009) as well as Becker’s independent study (2005), count
clauses like (4) in OFr as V3, exactly as they count clauses like (2b). We propose here to
use a third approach. From the point of view of the innovative non-V2 grammar that
will eventually triumph, XSV orders (as in (2b)) are permitted but XsiV(S) orders (as
in (4a–b) are not. Furthermore, even Modern German, a strict V2 language, resorts to
the resumptive adverb strategy occasionally. We therefore classify our clauses accord-
ing to whether they are compatible with the innovative grammar or not. XSV is a
progressive order, but XVS and XsiV(S) are conservative and hence – theoretically
speaking – V2. Applying this approach to the data reported in the previous literature
could change some interpretations significantly. In Becker’s study, for example, where
the statistics are broken down very finely, two thirds of the total examples of fronted
quant clauses are of the resumptive adverb type but are counted as V3. We will see in
Section€3 that counting this type of example as V2 allows a robust pattern of language
change to emerge from our statistics.
Before presenting our study, let us complete our inventory of attested word orders
after fronted subordinate clauses in OFr. Thus far we have examined three possibilities:
inversion (VS), non-inversion (SV), and the resumptive adverb strategy (si V(S)).
In addition to these possibilities, we find verb-initial orders in which the subject is
unexpressed, as in (6), and non-verb-initial orders in which the first constituent is
neither a subject nor a resumptive adverb, as in (7).7

.  We have excluded from consideration, at this stage of our research, examples in which
the fronted clause is (a) itself preceded by material which is likely to be independently
or jointly responsible for word order choices in the following main clause; (b) embedded
(together with following main-clause material), usually under a bridge verb; or (c) part of an
imperative or interrogative.
 Barbara Vance, Bryan Donaldson & B. Devan Steiner

(6) VX (null subject)

et pour ce que il sevent que nule gent n’ont si grant
and for this that they know that no people neg-have such great
pooir conme vous et la vostre gent, vous prient por dieu
power as you as you and your people you beg-3pl for God
que vous aiez pitie de la terre d’outremer.
that you have pity on the land overseas
“and because they know that no people have such great power as you and your
people, they beg you for God’s sake to have pity on the lands overseas.”
 (Villehardouin l. 22)

(7) XV (non-resumptive)
et se l’espee ne li fust tornee en sa main, ocis
and if the-sword neg to-him were turned in his hand killed
l’eust sanz faille.
him-would-have without fail
“and if the sword hadn’t been turned in his hand, he would have certainly
killed him.” (Queste 56, 10)

Null-subject examples such as (6) could be problematic to treat statistically, since a

missing subject cannot be counted with certainty as coming either before or after the
verb when both orders are possible with overt subjects, as is true of OFr just in the case
of fronted clauses (but not generally). However, since null subjects are overwhelmingly
linked with postverbal position at this stage of OFr, and since we have relatively few
examples of this type, we will take the step of counting them as V2.
Examples like (7) appear to demonstrate a true V3 pattern. However, since the
element that immediately precedes the verb is not the subject, such patterns cannot
be produced by a non-V2 SVO grammar. We therefore consider these examples V2,
noting that a clausal structure with two slots before the verb, as in Benincà (1995,
2006), may be necessary to treat them satisfactorily. Such examples are relatively rare
in our data.8
To summarize, we give an inventory in Table 1 of possible word orders following a
fronted clause in our texts, along with our interpretation of their status in a V2 gram-
mar. Here, and in the presentation of our statistics below, we place the construction
[fronted clause [VSp]] (where Sp is a subject pronoun) in a separate cell because it is a

.  In addition, it is sometimes difficult to identify a resumptive adverb with certainty. Some of
our OFr XV tokens may ultimately be found to fit better into the siV (resumptive) category.
 V2 loss in Old French and Old Occitan 

completely uncontroversial case of classic V2 behavior (that is, filling of one preverbal
position and incontrovertible V to C movement).

Table 1.╇ Categorization of word orders following fronted clause

Word order abbreviation example # grammar conservative/

progressive

Pronominal Inversion VSp 2a V2 conservative

Non-inversion SV 2b non-V2 progressive
Verb-initial with null or V (Sn) 5 (null) probably probably
non-pronominal inverted subject V2 conservative
Resumptive adverb Si V or si VS 3 V2 conservative
Preverbal XP constituent XV 6 V2 conservative

This variety of possibilities suggests a situation of competition (Kroch 1989) in

which two constructions, one generated by a conservative grammar (V2, with V-to-C
movement) and one generated by a progressive grammar (non-V2, without V-to-C),
compete over time. Throughout most of our data, the V2 and nonV2 word-order
choices are interchangeable in terms of meaning. For example, the subject of the
fronted clause is typically co-referential with the subject of the following main clause;
this can be expressed by any of the orders described above. (Like any such competi-
tion, however, there may be subtle pragmatic or semantic factors that increase the
probability of one of the alternates in a specific environment; see Footnote€16.)

3.  E
 mpirical study: The effect of fronted clauses on word order
in Old French

For the present study of OFr, we chose two texts from each of three different prose
genres, one from the first half of the 13th century and one from (roughly) the second
half, to track any potential diachronic developments. The genres are (a) literary texts
[La Queste del Saint Graal (1225) and Le Roman de Cassidorus (1267)], (b) historical
texts [Villehardouin’s Conquête de Constantinople (1206) and Joinville’s Vie de Saint
Louis (1306)], and (c) legal texts [charters (1231–1271)]. The dialects represented
are from the Northeastern area of French-speaking Europe, mainly Champagne. We
focused on the following fronted tensed subordinate clauses: se (“if ”), quant (“when”),
por ce que (“because” “so that”), en ce que (“so that”), endementiers que (“while”),
(aus)si tost com/que (“as soon as”), einsi com/que (“just as”), and avant que (“before”).
 Barbara Vance, Bryan Donaldson & B. Devan Steiner

Since we did not find clear differences between the literary and historical genres, we
have grouped them together in Section€3.1; legal texts are discussed in Section€3.2.9

3.1  Results and discussion: Old French literary and historical texts
In Tables 2–5, the data from literary and historical texts are organized by date and
type of fronted clause. Quant- and se-clauses were chosen because of their overall high
frequency in our texts, and por ce que for comparison with the legal texts. The other
subordinate clauses present lower token counts and are combined and reported as
“mixed clauses”.
Table 2 presents the most frequent fronted clause type – quant. In matrix clauses
following a quant clause, the preference in the early texts is for the Si V(S) word-
order (108 out of 128 tokens in Villehardouin and 170 of 202 tokens in the Queste).10
Cassidorus, from mid-century, provides evidence that a shift is beginning to take
place, with a rise in SV tokens (102/409) and slightly fewer SiV(S) tokens (281/409)
compared to the two early texts. Finally, in Joinville, the data point toward the comple-
tion of this shift, with a strong preference for the progressive SV word-order (118/128
tokens) and very few tokens of the conservative Si V(S) order.

Table 2.╇ Word order after fronted quant-clauses (‘When ….,___’)

Quant … V Sp V(Sn) Si V(S) XV(S) SV Total

Villehardouin (1206) 0 3 108 (84%) 1 16 (13%) 128

Queste (1225) 0 0 170 (84%) 0 32 (16%) 202
Cassidorus (1267) 0 8 281 (69%) 18 102 (25%) 409
Joinville (1306) 0 1 12 (9%) 0 115 (90%) 128
Sp = subject pronoun, Sn = nonpronominal subject, S = all subjects, () = optional constituent,
Si = resumptive adverb such as si, lors, (a)donc.

Table 3 demonstrates that se-clauses show a preference for the progressive SV

word-order across all time periods, from which we conclude that se-clauses are the

.  However, we do recognize a still somewhat puzzling syntactic difference between our two
earliest texts, Villehardouin and the Queste, when an adverb or prepositional phrase begins
the utterance (hence, in environments that do not directly concern us here). In the Queste, the
order SV after such constituents is vanishingly rare except after a restricted set of expressions
already identified by Foulet 1928, while in Villehardouin there is more variability. This differ-
ence has clearly contributed to different analyses proposed in Vance 1997 and e.g. Ferraresi€&
Goldbach 2002. The fact that the Villehardouin text also lacks robust statistics for subject
pronoun inversion (after fronted clauses, in the data presented here, and more generally), sug-
gests to us that it may share some archaic features with verse texts (see also Footnote€16).
.  Our data for quant- and se- clauses in the Queste come from pages 1–91 only.
 V2 loss in Old French and Old Occitan 

most progressive of the fronted clauses in question (despite the less robust than
expected preference in Cassidorus).

Table 3.╇ Word order after fronted se-clauses (‘If …, _____’)

Se… V Sp V(Sn) Si V(S) XV(S) SV Total

Villehardouin (1206) 1 2 0 1 29 (88%) 33

Queste (1225) 0 0 2 7 29 (76%) 38
Cassidorus (1267) 2 2 10 10 32 (57%) 56
Joinville (1306) 1 0 6 1 48 (86%) 56

Table 4 presents the por ce que data, where (even leaving Villehardouin aside due
to a low token count), one sees the same evolution from conservative to progressive
word order as reported for quant-clauses. However, we consider por ce que to be even
more conservative than quant, given that the early text Queste shows a significant
preference for the most clearly conservative order – VSp – and no SV orders at all, and
that VSp is still well-represented in the later two texts.

Table 4.╇ Word order after fronted por ce que clauses (‘Because …, _____’)

Por ce que V Sp V(Sn) Si V(S) XV(S) SV Total

Villehardouin (1206) 0 1 2 0 0 3
Queste (1225) 23 7 1 3 0 34
Cassidorus (1267) 3 2 2 2 10 (53%) 19
Joinville (1306) 4 6 2 0 22 (65%) 34

Finally, Table 5 presents the mixed group of other fronted clauses. These clauses
pattern similarly to both quant and por ce que, demonstrating an evolution from the
conservative to progressive word-orders over time. However, unlike quant, a relatively
strong presence of the conservative word orders is still found in the latest text, Join-
ville, and unlike por ce que, the rise of SV begins with the Queste. Thus, we consider
these to be more conservative than quant, for which the change to the progressive SV
order appears to be nearly complete at this same time period, and less conservative
than por ce que.

Table 5.╇ Word order after mixed other fronted clauses

Other fronted clauses V Sp V(Sn) Si V(S) XV(S) SV Total

Villehardouin (1206) 0 1 12 1 0 (0%) 14

Queste (1225) 9 20 53 1 24 (22%) 107
Cassidorus (1267) 12 30 6 14 34 (35%) 96
Joinville (1306) 8 6 0 2 48 (75%) 64
 Barbara Vance, Bryan Donaldson & B. Devan Steiner

Table 6 presents an overview of the emergent patterns discussed, focusing only on

the percentage of the progressive order SV in each text for each clause type.

Table 6.╇ Overall results for OFr: frequency of SV after fronted clause

Type of fronted clause Por ce que …, mixed clauses Quant …, Se …,

Villehardouin (1206) 0% SV* 0% SV 13% SV 88% SV

Queste (1225) 0% SV 22% SV 16% SV 76% SV
Cassidorus (1267) 53% SV 35% SV 25% SV 57% SV
Joinville (1306) 65% SV 75% SV 90% SV 86% SV

*There are only 4 por ce que tokens from this text.

The statistics indicating a preference (>50%) for the most progressive SV order
are concentrated in just two areas: the se-clauses (far right column) and the latest text
(bottom row). Furthermore, there is a general progression across the chart from least
progressive clause type (por ce que) to most progressive (se), with some anomalies
in the Queste and Cassidorus, perhaps due to the combination of clause types in our
mixed category. Finally, the progression over time from Villehardouin to Joinville is
quite clear, with the first three clause types preferring SV more and more over time
and se-clauses having already reached a somewhat stable state from the beginning of
the 13th century.
We can now identify a clear change taking place with respect to the participation
of fronted subordinate clauses in the OFr V2 grammar. Different fronted clause types
achieve a preference for the new grammar at different times, but by the beginning
of the 14th century, the conservative pattern is a minority in all contexts. Given the
documented spread of SV orders after other types of fronted constituents (adverbs,
direct objects, prepositional phrases, etc) over the next two centuries (cf. Vance 1997
and works cited there) we conclude that fronted clauses represent the first discernible
wave of V2 loss in OFr.11 The limited yet systematic use of SV after fronted clauses in
the early 13th century cannot, in our view, be used to argue that OFr is not a V2 lan-
guage, but rather indicates a very strong V2 constraint (contra Kaiser 2002) that is only
broken down slowly and systematically over several centuries.

.  Our data are not extensive enough at this point to distinguish between a situation in
which the change spreads from context to context and a scenario of “simultaneous unequal
actuation” across contexts, as promoted by Kroch 1989.
 V2 loss in Old French and Old Occitan 

3.2  Results and discussion: Old French legal texts (1251–1271)

The use of legal documents for syntactic research is controversial, because of the
rather formulaic language that could, in principle, have little in common with the
spoken language. However, the OFr Charters are interesting because they can usually be
attributed to a precise locality and because some regional variation in their syntax has
been found (cf. Dees 1980; Hinzelin 2007). Our charter data are from the northeastern
part of French-speaking Europe, including Champagne. Although we collected data
from both early and late 13th century charters, we excluded the earlier data because of
the limited token counts of the clause types in question.
The only type of fronted subordinate clause found in our charters is introduced by
por ce que “so that”. (8a) shows following SV order, while (8b) has an inverted subject.
(8) a. Et pour ce que ce soit ferme chose et estable, nous avons
and for this that this be firm thing and established we have
fait seeller ces lettres de nostre seel.
made to seal these letters of our seal
“And so that this be a firm and established thing, we have sealed these
letters with our seal.” (Aube Charter 87)
b. Et pour ce que ce soit ferme chose et estable ai
and for this that this be firm thing and established have
je seelees ces presentes lettres de mon propre seel en
I sealed these present letters of my own seal in
tesmoingnage de veritei.
witness of truth
“And so that this be a firm and established thing I have sealed the present
letters by my own seal in witness of the truth.” (Aube Charter 89)

Table 7 shows the distribution of word order types following the fronted clause. The
categories are the same as for the historical and literary texts, except for XV, which is
not found in the legal texts.

Table 7.╇ Old French Legal Documents by department of origin: word order after fronted
clauses beginning with por ce que ‘so that’

VSp V(Sn) SiV(S) SV Total

Vosges 39 6 1 23 (33%) 69
Hainaut 4 3 49 22 (28%) 78
Aube 2 0 0 40 (95%) 42
Flandres 2 0 2 22 (85%) 26
Luxembourg 28 2 3 36 (52%) 69
 Barbara Vance, Bryan Donaldson & B. Devan Steiner

The data from this 20-year period reveal significant regional variation. All the
departments except Aube are contiguous to Germanic-speaking areas, where we
might expect some contact-induced strengthening of V2, but the percentage of SV
orders varies for these four departments from roughly 30% (Vosges and Hainaut)
to€85% (Luxembourg). It is possible that variation is due to different writing practices
in different localities rather than to regional variation in the spoken language. How-
ever, supposing that our statistics represent true regional variation, let us focus on
the Aube data. The Aube department is located in Champagne, the probable region
of origin of three of our non-legal texts: Villehardouin, Cassidorus, and Joinville.12
Table 8 presents the Aube charter data with the parallel data from the literary and
historical texts.

Table 8.╇ Proportion of SV order after fronted clauses por ce que (all texts)

Text SV order

tokens %

Villehardouin, 1206 0/3 0%

Queste, 1225 0/34 0%
Cassidorus, 1267 10/19 53%
Joinville, 1306 22/34 65%
Aube Charters, 1251–1271 40/42 95%

Table 8 demonstrates that the legal charters are far more advanced in the use of
SV order (95%) after por ce que than even the latest literary/historical text (Joinville)
despite the fact that the legal texts cover the middle time period. Yet, referring back to
the raw numbers for the Aube department in Table 7, we see that variation still exists.
One interpretation is that the legal texts in the Aube follow the same general trajec-
tory as the literary/historical texts (from V2 to non-V2) but reflect a more innovative
grammar closer to the spoken language than other genres, despite the legal formulas.
Additional research is needed before other possible interpretations can be excluded,
but these results suggest systematic variation across both region and genre.

.  As far as we can determine, there are no suitable prose literary texts from Champagne
from the early 13th century. We have therefore retained the Queste in this slot, which is from
a nearby region (Bourgogne).
 V2 loss in Old French and Old Occitan 

4.  Fronted subordinate clauses in Old Occitan

We now examine the word order patters of Old Occitan (OOc), for which our study
is necessarily smaller than the OFr study because of lower token counts in the texts.
The relatively few published works on OOc syntax (compared to OFr) also provide
a less developed backdrop to the study than in OFr. Our detailed study of the effect
of fronted subordinate clauses on following word order is intended as a step toward
addressing the paucity of data on V2 in OOc and also as a counterpart to the OFr
study above.
Occitan is of interest because – like French – it has undergone a loss over time of
traits associated with a V2 constraint. In traditional treatments of OOc syntax, subject
inversion after a fronted non-subject constituent is taken to be a fundamental charac-
teristic, just as in OFr (cf. e.g. Jensen 1994:€362). As we noted above, Benincà (1995,
2006) includes OOc in her generative study of Medieval Romance and concludes that
all of these languages are V2, although OOc, like various Italian varieties, obeys a
descriptively less strict V2 constraint than was active in OFr. A rough idea of the dis-
tribution of word-order types in OOc comes from Sitaridou (2005). In a sample of 200
clauses from early OOc legal charters, 36 clauses (18%) are descriptively V3. None of
these 36 clauses appear to contain a fronted subordinate clause; our study thus adds a
new element to the consideration of V2 versus V3 in OOc.
As in the OFr study, we examine five prose texts: an earlier and a later work for
both literary and historical (or literary/historical) genres, as well as early charters. The
availability of charters in OOc from a century earlier than in OFr allow us to investi-
gate a greater time span. Looking at documents from central Occitania (Provence and
Languedoc), we investigate fronted subordinate clauses beginning with si “if ”,13 can
“when”, and a small group of mixed types (per so que “because”, “so that”, car “because”,
domentre “while”). The same kinds of word orders are found as in OFr:

(9) a. following SV order

E si tul nos demandas per te o per to mes, nos
and if you-it us ask for you or for your messenger we
lot rendrem sas logre.
it-to-you will-return without interest
“And if you ask it of us yourself or via your messenger, we will give it back
to you without interest.” (Charter 29)

.  OOc si ‘if ’ (= OFr se) should not be confused with the OFr resumptive adverb si “thus”.
 Barbara Vance, Bryan Donaldson & B. Devan Steiner

b. following V order (subject is null)

E can Karles ausi aysso, passec am los sieus
And when Charles heard this, passed-3pl with the his
entro Marelos.
as-far-as Marelos
“And when Charles heard this, he went with his men as far as Marelos…”
 (Gesta Karoli Magni, p. 55, lines 701–704)
c. following resumptive adverb
Cant illi retornava de sos autz raubimens, enaissi si
when she returned from her high raptures thus refl
mostrava paura e defallens e humana a totas.
showed-3sg poor and feeble and human to all
“When she came out of her holy rapture, then/thus she appeared weak and
feeble and human to all.” (Douceline, p. 115)
d. following VS order14
La qual cauza plus fizelmens a far e plus veraia, volc
the which thing more faithfully to do and more truly wanted
illi aver per lo dechat e.l conseill dell sant paire.
she to-have for the words and-the counsel of-the holy father
“To do this thing more faithfully and more truthfully, she wanted to have
the words and the counsel of the holy father.” (Douceline, p. 61)
In Table 9, we see that the distribution of these orders differs greatly from OFr. The raw
data is presented in the “details” column; the “global” column presents a generalized
view of the preponderant (S)V order. There is very little use of inversion after fronted
clauses, and the resumptive adverb strategy is so rare that we have not separated it here
from the V3 XV order. SV and V orders far outnumber the others in all texts and in
all time periods (except in some cells where the data are sparse). Here a clarification is
in order: whereas the V-initial orders are all consistent with a V2 grammar, as in OFr,
the distribution of null subjects in OOc complicates the interpretation of the V orders.
While OFr null subjects have typically been interpreted as necessarily postverbal
(since Foulet 1928), OOc null subjects enjoy a much freer distribution (cf. Lafont 1967;
Vanelli, Renzi & Benincà 1985; Sitaridou 2005; Vance 2009). Furthermore, Modern
Occitan still licenses null subjects. Consequently, the null subject orders in our chart are
equally compatible with a modern Occitan grammar. Although some considerations
that (for reasons of space) we cannot discuss here might lead us to re-classify a few

.  This is our clearest example of V2 inversion after a fronted subordinate clause. Because it
contains a fronted non-finite clause rather than a finite clause, we excluded it from our statis-
tical count. The VS examples in our data may represent V2, even if ambiguous.
 V2 loss in Old French and Old Occitan 

of our null subjects as postverbal, we have conflated V and SV orders as (S)V in our
global view in a way that should not compromise our conclusions.

Table 9.╇ Word order after fronted subordinate clauses in OOc

Text Cant … (‘when’) Si …, (‘if ’) Other clauses

details global details global details global

Chartes 13 SV 100% SV 20 V n/a

1034–1176 18 SV 79% (S)V n/a
early legal 9 XV
1 VS
Damiette 8 SV 94% (S)V n/a n/a n/a n/a
mid-13th 9V
historical 1 XV
Gesta KM 32 SV 96% (S)V 1 SV (50% SV) 30 SV 89% (S)V
mid-13th 20 V 1VS 10 V
literary 1 VS 2 VS
1 XVS 2 XV
1 XVS
Douceline 49 V 85% (S)V 6 XV (25% (S)V) 2V (86% (S)V)
late-13th- 31 SV 2V 4 SV
early-14th 4 VS 1 XV
literary 8 XV
2 XVS
Jerusalem 25 S(X)V 79% (S)V 1 SV (100% (S)V) n/a n/a
14th 17 V 1V
historical 5 V(X)S
1 XVS
5 XV

In stark contrast to OFr, the data from OOc do not show evidence of a change in
progress. No clear pattern emerges with respect to genre, clause type, or time period;
rather, there is a preponderance of orders consistent with the more modern grammar
(whenever token counts are robust). One possibility is that OOc is ahead of OFr in
losing V2, so that these data attest to a stage where the most advanced contexts have
already achieved a fairly stable new order. Another possibility, however, is that V2
loss in OOc cannot be tracked by looking at fronted clauses because they are (fairly)
consistently located outside the V2 core of the clause. Recall that Benincà’s (1995) view
of the Medieval Romance clause, which we now show in detail, posits two preverbal
positions (we abstract away from the more complex version of the left periphery pre-
sented in her later works):

(10) [TopP _______ [Top′ [CP _______ [C′ Vi [TP (subject) [T′ ti [VP …]]]]]]]

Under this expanded view, the OOc fronted clauses would consistently appear in the
Spec TopP position. Other constituents (XP or subject) could occupy Spec CP, and
 Barbara Vance, Bryan Donaldson & B. Devan Steiner

the verb might still move to C. To track loss of V-to-C in OOc, then, we would need
to focus on non-subject elements consistently situated in Spec CP, to see if there is a
progression over time from VS order (V-to-C movement) to SV order (V2 lost, verb
remains in T).
Fortunately, our data can be exploited in an additional way that helps confirm that
OFr and OOc fronted clauses behave differently syntactically. In our OOc data, but not
in our OFr data, there is variation in the position of the object clitic (OCL) in examples
where the matrix verb is accompanied by such a clitic. In our OFr examples, the object
clitic is always preverbal (see (6), for example), but in OOc it can be either preverbal
or postverbal. As Table 10 makes clear, there is a strong tendency in our data for the
OCL to be preverbal in OOc when there is some constituent – a subject or other XP –
between the fronted clause and the matrix verb. When there is no such constituent, the
OCL is nearly always postverbal (and always postverbal if the subject is null).15

Table 10.╇ Preverbal vs. Postverbal object clitics after fronted subordinate clauses (OOc)

Word order after Subj – initial XP-initial V initial with V initial with
fronted clause â•… inversion null subject

S Ocl V vs X Ocl V vs Ocl V S vs Ocl V vs

S V Ocl X V Ocl V Ocl S V Ocl

Text pre post pre post pre post pre post

Chartes, 11-12c 28 0 5 0 0 1 0 8
Damiette ~1250 0 0 1 0 0 0 0 2
Gesta ~1250 11 3 2 0 1 3 0 7
Douceline ~1400 12 0 11 0 1 2 0 14
Jérusalem ~1473 6 0 4 0 0 0 0 5
Total 57 3 23 0 2 6 0 36
% preverbal 90% 100% 25% 0%

We interpret this clear pattern as an indication that in OOc the fronted clause,
when it is the only constituent before the matrix verb, does not occupy the same posi-
tion (next to the verb) as the subject or XP. Rather, according to Benincà’s (1995, 2006)
and Skårup’s (1975) interpretation of the Tobler-Mussafia Law, object clitics can only
be postverbal if the position immediately to the left of the verb is left unfilled, so that
the verb behaves as if it were in absolute initial position. If no constituent appears in
Spec CP, then an object clitic following a fronted clause in Spec TopP will also follow
the verb.

.  Mériz 1978, Smith & Bergin (1984:114) and Skårup 1975 also note that object clitics are
predominately postverbal after a fronted subordinate clause in OOc.
 V2 loss in Old French and Old Occitan 

Returning to our OFr data, we are led to ask why we find only preverbal object
clitics after fronted clauses. It seems unlikely that the Tobler-Mussafia Law is simply
inactive in our data, since elsewhere in our texts we find postverbal OCLs in absolute
verb-initial clauses. Rather, we interpret the large number of examples with resumptive
si to indicate that, in contrast to OOc, OFr prefers to fill the Spec CP position. In exam-
ples like (6) it is the fronted clause that fills Spec CP, while in examples like (4a–b), the
resumptive adverb is in Spec CP.
This result suggests to us that the way in which Old French is to be interpreted
as a ‘stricter’ V2 language than OOc is that it requires the inner preverbal position to
be filled.16 Given that Modern German also permits resumptive adverb constructions
after fronted clauses, we find – contra Kaiser – that the presence of descriptively V3
clauses in our data does not force us to abandon the hypothesis that OFr was a V2
language of the same general type as the modern Germanic V2 languages. Indeed, it
is precisely in its observance of a strict adjacency between the verb in C0 and the con-
stituent of arbitrary category on its left that OFr departs from Medieval Romance and
patterns much more closely with Germanic V2. But if the V2 property were completely
stable during the 13th century, we would not expect to find the word order SV gaining
ground, first with se clauses, then with quant clauses, and then with other fronted
clauses, prefiguring the decline of V2 with other fronted constituents (adverbs, prepo-
sitional phrases, etc) that is to characterize the next two centuries. Rather, it seems
that the increasingly frequent occurrence of SV order after a fronted clause, as in (2b),
identifies this context as the first one in which the finite verb fails to move to C0.

.  A word about Early Old French (12th century and before) is in order here. Because we
have limited our study to prose texts within specific genres, we have been constrained to begin
our analysis of Old French with the 13th century. This restriction does not necessarily diminish
our view of the diachrony of fronted clauses, since it has long been recognized that the syntax
of early OFr texts is qualitatively different from that of later narrative prose (Hirschbühler
1990). However, in future work we intend to compare our results with those of Skårup 1975,
Grad 1956, and Franzen 1939, who have investigated mainly early verse texts. These authors
cite examples of fronted clauses with postposed object clitics similar to those we discovered
in our OOc data – evidence, in our view, of a more Romance-like earlier stage. They also suggest
that there is a tendency to pair “concessive” fronted clauses (those with meanings such as
“even if ”) with inversion more frequently than other types of fronted clauses; here, we
find possible confirmation in data such as our (2a). Thus it may be possible to uncover yet
another fine-grained distinction within our 13th century data, allowing us to track even
more closely the factors affecting the choice between conservative and innovative word
orders. Finally, the 12th century data already suggests a basic split between fronted se and
quant-clauses, on the one hand, and other types of fronted clauses – a situation that is not
unexpected given the advanced state at which we find the syntax of these clauses in the
13th€century, if it turns out to be possible to establish some continuity between these two
very different genres of Old French.
 Barbara Vance, Bryan Donaldson & B. Devan Steiner

5.  Summary and conclusion

In this study we have sought to bring some new data and statistics to bear on the
question of the V2 status of Old French and Old Occitan. Investigating the effect of
fronted subordinate clauses on following word order, we have found systematic varia-
tion in 13th-century OFr prose but very little variation in 11th–14th century Occitan
prose. In OFr, a competition (in the sense of Kroch 1989) appears to be in progress
between a Germanic-style V2 grammar – in which fronted clauses are found either
in a single preverbal position in the clause (Spec CP) or in a Spec TopP paired with
a resumptive adverb in Spec CP – and a progressive SVO grammar in which fronted
clauses co-occur with a preverbal subject because the finite verb no longer moves to€C.
Several factors are found to favor the more progressive grammar: the genre (legal
documents vs. literary/historical texts), the date (later vs. earlier), and the particular
type of fronted clause (se-clauses being the most likely to show nonV2 order). For
OOc, on the other hand, our study confirms that even in the earliest data, the outer
position Spec TopP is the usual landing site for fronted clauses, and Spec CP may be
empty. As a result, fronted clauses have little effect on following word order and cannot
help us track the evolution of V-to-C movement in OOc (at least not during the time
period investigated).
In future work, we hope to examine a much wider range of fronted constituents
in both OFr and OOc in order to begin filling out the story of V2 loss in these two
Romance varieties. Our results thus far make it clear, however, that the instances of V3
with fronted clauses in OFr that have been discussed here and in the previous litera-
ture cannot be taken as evidence that OFr was never a V2 language. Neither should we
hasten to adopt an analysis in which 13th-century OFr shares with general Medieval
Romance, including Occitan, a grammar that is considerably less strict than Modern
German in its adherence to descriptive V2 order. Rather, the data point to a systematic
evolution, over the course of Medieval French, from a strict V2 grammar to a non-V2
grammar. In the present paper we have documented what is perhaps the earliest avail-
able evidence of this change.

References

Primary Sources: Old French

Coq, Dominique. 1974. Chartes en langue française antérieures à 1271 conservées dans les dépar-
tements de l’Aube, de la Seine-et-Marne et de l’Yonne. Paris: Editions du Centre National de
la Recherche Scientifique.
Joinville, Jehan de. 1977. La Vie de Saint Louis. ed. by Noel Corbett. Sherbrooke: Editions
Naaman.
 V2 loss in Old French and Old Occitan 

Lanher, Jean. 1974. Chartes en langue française antérieures à 1271 conservées dans le département
des Vosges. Paris: Editions du Centre National de la Recherche Scientifique.
Mantou, Reine. 1987. Chartes en langue française antérieures à 1271 conservées dans les prov-
inces de Flandre orientale et de Flandre occidentale. Paris: Editions du Centre National de la
Recherche Scientifique.
Pauphilet, Albert, ed. 1984. La Queste del Saint Graal, roman du XIIIe siècle. Paris: Champion.
(Electronic version courtesy of Christiane Marchello-Nizia).
Palermo, Joseph, ed. 1963. Le Roman de Cassidorus. Paris: SATF. Volume 2.
Ruelle, Pierre. 1984. Chartes en langue française antérieures à 1271 conservées dans la province de
Hainaut. Paris: Editions du Centre National de la Recherche Scientifique.
Stein, Achim, et al. eds. 2006. New Amsterdam Corpus: Electronic corpus of Old French literary
texts (ca. 1150–1350), established by Anthonij Dees, reworked by Achim Stein, Pierre
Kunstmann & Martin-D. Gleßgen. Stuttgart: Institut für Linguistik/Romanistik.
Villehardouin, Geoffroi. 1978. La conquête de Constantinople. éd. Centre de Recherches sur les
Arts et le Langage (CRAL), Nancy. Paris, Bibliothèque Nationale Française, 2137. [Consulted
via New Amsterdam Corpus.]

Primary Sources: Old Occitan

Brunel, Clovis. 1926. Les plus anciennes chartes en langue provençale: recueil de pièces originales
antérieures au XIIIe siècle. Paris: Picard. (plus supplement, 1952).
Chabaneau, Camille, ed. 1888–1889. “La Prise de Jérusalem”. Revue des Langues Romanes
32:€581–608 and 33: 31–46 & 600–609.
Gout, Raoul. 1927. La Vie de Sainte Douceline. Paris: Ars et Fides.
Meyer, Paul, ed. 1877. “La Prise de Damiette en 1219”. Bibliothèque de l’Ecole des Chartes
38:€497–571.
Schneegans, Friedrich Edward. 1898. Gesta Karoli Magni ad Carcassonam et Narbonam. Halle:
Max Niemeyer.

Secondary Sources
Adams, Marianne. 1987. “From Old French to the Theory of Pro-Drop”. Natural Language and
Linguistic Theory 5, 1. 1–32.
Becker, Martin G. 2005. “Le Corpus d’Amsterdam face à une vieille question: L’ancien français
est-il une langue V2?”. Romance Corpus Linguistics II, ed. by Claus D. Pusch, Johannes
Kabatek & Wolfgang Raible. 345–358. Tübingen: Gunter Narr Verlag.
Benincà, Paola. 1995. “Complement Clitics in Medieval Romance: the Tobler-Mussafia law”.
Clause Structure and Language Change, ed. by Adrian Battye & Ian Roberts, 325–344.
Oxford: Oxford UP.
Benincà, Paola. 2006. “A Detailed Map of the Left Periphery of Medieval Romance”. Cross-
linguistic Research in Syntax and Semantics, ed. by Raffaella Zanuttini, Héctor Campos,
Elena Herburger & Paul H. Portner. 53–86. Washington, D.C.: Georgetown UP.
Besten, Hans den. 1983. “On the Interaction of Root Transformations and Lexical Deletive Rules”.
On the Formal Syntax of the Westgermania, ed. by Werner Abraham, 47–131. Amsterdam:
John Benjamins.
Dees, Anthonij. 1980. Atlas des formes et des constructions dans les chartes françaises du XIIIe
siècle. Tübingen: Niemeyer.
 Barbara Vance, Bryan Donaldson & B. Devan Steiner

Ferraresi, Gisella & Maria Goldbach. 2002. “V2 Syntax and Topicalization in Old French”.
Linguistische Berichte 189. 2–25.
Foulet, Lucien. 1928. Petite syntaxe de l’ancien français. Paris: Champion. Reprinted 1982.
Franzen, Torsten. 1939. Etude sur la syntaxe des pronoms personnels sujets en ancien français.
Ph.D. dissertation, Uppsala University.
Grad, Anton. 1956. “L’inversion du sujet dans la principale précédée d’une subordonnée en
ancien français”. Ljubljana: Razprave Dissertationes II. 1–28.
Hinzelin, Marc-Olivier. 2007. Die Stellung der klitischen Objektpronomina in den romanischen
Sprachen: Diachrone Perspektive une Korpusstudie um Okzitanischen sowie zum Katalanischen
und Französischen. Tübingen: Narr.
Hirschbühler, Paul. 1990. “La légitimation de la construction V1 à sujet nul en subordonnée
dans la prose et le vers en ancien français.” Revue Québécoise de Linguistique 19, 1. 33–55.
Jensen, Frede. 1994. Syntaxe de l’ancien occitan. Tübingen: Niemeyer.
Kaiser, Georg A. 2002. Verbstellung und Verbstellungswandel in den romansichen Sprachen.
Tübingen: Niemeyer.
Kroch, Anthony. 1989. “Reflexes of Grammar in Patterns of Language Change”. Language
Variation and Change 1: 3. 199–244.
Lafont, Robert. 1967. La phrase occitane. Montpellier: Presses Universitaires de France.
Ledgeway, Adam. 2008. “Satisfying V2 in Early Romance: Merge vs. Move”. Journal of Linguistics
44, 2. 437–470.
Meriz, Diana. 1978. “Observations on object pronoun collocation with finite verb-parts in
Medieval Occitan (to 1300).” Romania 99: 145–182, 289–310.
Poletto, Cecilia. 2005. “Si and e as CP Expletives in Old Italian”. Grammaticalization and
Parametric Variation, ed. by Montse Batllori, Maria-Lluïsa Hernanz, Carmen Picalle &
Francesc Roca. 206–235. Oxford: Oxford UP.
Rinke, Esther. 2003. “On the Licensing of Null Subjects in Old French”. Syntactic Structures
and Morphological Information, ed. by Uwe Junghanns & Luka Szucsich, 217–247. Berlin:
Mouton de Gruyter.
Rinke, Esther & Jürgen M. Meisel. 2009. “Subject Inversion in Old French: Syntax and infor-
mation structure”. Proceedings of the Workshop on Null subjects, Expletives, and Locatives
in Romance, ed. by Georg A. Kaiser & Eva-Maria Remberger, 93–130. Arbeitspapier 123.
Fachbereich Sprachwissenschaft, Universität Konstanz.
Roberts, Ian. 1993. Verbs and Diachronic Syntax. Dordrecht: Kluwer.
Sitaridou, Ioanna. 2005. “A corpus-based study of null subjects in Old French and Old Occitan”.
Corpora and Diachronic Linguistics, ed. by Claus D. Putsch, Johannes Kabatek & Wolfgang
Raible, 359–374. Tübingen: Narr.
Skårup, Povl. 1975. Les premières zones de la proposition en ancien français. Copenhagen:
Akademisk Forlag. [= Revue Romane Numéro Spécial 6.]
Smith, Nathaniel B. & Thomas G. Bergin. 1984. An Old Provençal Primer. New York: Garland.
Thurneysen, R. 1892. “Die Stellung des Verbums in Altfranzösischen”. Zeitschrift für Romanische
Philologie 16. 289–307.
Vance, Barbara. 1997. Syntactic Change in Medieval French. Dordrecht: Kluwer.
Vance, Barbara. 2009. “The evolution of subject pronoun systems in Medieval Occitan”. Ms,
Indiana University.
Vanelli, Laura, Lorenzo Renzi & Paola Benincà. 1985. “Typologie des pronoms sujets dans
les langues romanes”. Actes du XVIIe Congrès International de Linguistique et Philologie
Romanes, 1983, Vol 3. 163–176. Aix: Université de Provence.
part iii

Morphology and interfaces

 The loss and survival
of inflectional morphology
Contextual vs. inherent inflection in creoles

Ana R. Luís
University of Coimbra

Although recent evidence has shown that creoles are not exempt from overt
inflectional morphology, little is yet known about the exact range of inflectional
categories expressed by creoles. A detailed analysis of the verbal paradigm of
Korlai Creole Portuguese reveals that verbs encode conjugation classes but no
subject agreement endings. The same inflectional development has taken place in
two more Indo-Portuguese creoles, namely Daman and Diu Creole Portuguese.
Cross-linguistic evidence therefore casts serious doubts on the traditional claim
that creoles discard semantically dispensable units (McWhorter 2001).
In order to understand why creoles select purely formal inflectional
categories, this paper draws on the distinction, formulated by Booij (1994, 1996),
between inherent and contextual inflection. Based on this distinction, I argue that
the retention of conjugation class markers vs. the loss of agreement endings can
be subsumed under a wider generalisation about creole languages, namely their
preference for inherent rather than contextual inflection (Kihm 2003; Arends et€al.
2006; Plag 2008a).

Key words:╇ Korlai Creole Portuguese; creole inflection; theme vowels; inherent
inflection; contextual inflection

1.â•… Introduction

Despite recent attempts to demonstrate that creoles languages are not exempt from
overt inflection, the claim that creoles are inflectionally impoverished languages
remains largely correct.1 Indeed, creoles which do have overt inflection encode

.â•… An earlier version of this study was presented at the 2008 Meeting of the Associação de
Crioulos de Base Lexical Portuguesa e Espanhola (ACBLPE). I am deeply grateful to the audience
and also to Hugo Cardoso, Clancy Clements and John Holm for helpful suggestions. Thanks
also to two anonymous reviewers for comments on a previous version of this paper.
 Ana R. Luís

systematically fewer inflectional categories (e.g. tense, number, definiteness, etc.) and
thus have fewer inflectional affixes than their lexifier languages. Such inflectional
poverty appears to be determined, among other factors, both by psycholinguistic and
sociolinguistic forces.
For example, assuming that creole grammars are shaped by processes of Second
Language Acquisition (SLA), the fact that L2 learners acquire overt inflectional mor-
phology at a later stage of their acquisition process (DeGraff 2005) may explain why
creoles contain very little overt inflection. In addition, adults tend to impose on them-
selves attitudinal restrictions which may inhibit the acquisition of the target language
(Thomason€& Kaufman 1988; Klein & Perdue 1997; Siegel 2004). Research on adult
immigrant communities has shown, for example, that unmotivated adults, in an untu-
tored SLA context, tend not to go beyond a rudimentary (i.e. inflectionally poor) variety
of the target language (Klein & Perdue 1997; Siegel 2004).
While the scarcity of inflectional affixes is beyond dispute, very little is yet known
about the exact range of inflectional categories expressed by creoles. In particular, it
is not entirely clear why creoles seem to be selective with respect to the inflectional
categories they encode or why certain inflectional categories stand a higher chance
of surviving creolisation than others. While some studies have observed that subject-
agreement endings are among the inflectional endings that are most typically absent
from creoles (Kihm 2003; Veenstra 2003; Arends et al. 2006), none of these studies
have yet investigated the fate of entirely meaningless inflectional categories, such as
conjugation classes.
This paper attempts to account for the categorical selectivity of creoles by examining
the verbal paradigm of Korlai Creole Portuguese (KP). In KP, verbs have lost subject-
agreement endings, but have retained all three Portuguese conjugation classes (includ-
ing theme vowels and allomorphic alternations). The same inflectional development
has taken place in other Indo-Portuguese creoles, casting serious doubts on the tradi-
tional claim that creoles discard semantically dispensable units.
As part of my goal to understand why creoles select purely formal inflectional
markers and discard perfectly meaningful affixes, I will draw on the distinction between
inherent and contextual inflection, proposed by Booij (1994, 1996). It will be argued
that the retention of conjugation class markers vs. the loss of agreement endings can be
subsumed under a wider generalisation about creolization, namely that creoles favour
inherent inflection over contextual inflection (Kihm 2003; Arends et al. 2006).
This paper is organised as follows: Section€ 2 provides an outline of the inflec-
tional endings that are attested in the verbal paradigm of KP. Section 3 highlights both
empirical and psycholinguistic evidence in favour of Booij’s distinction between inher-
ent and contextual inflection. Section 4, then, argues that the inflectional categories
expressed by KP verb forms can be nicely accommodated under the view that creole
languages favour inherent over contextual inflectional (Kihm 2003). A summary of
our findings will be provided in Section€5.
 The loss and survival of inflectional morphology 

2.â•… The internal structure of inflected verbs in KP

Spoken on the northwest coast of India, KP has emerged from the contact between two Indo-
European languages, namely Portuguese (Romance) and Marathi (Indo-Aryan).2 Unlike
other Portuguese-based creoles, it has preserved a rare amount of overt inflectional
endings:3 verbs inflect for both tense and conjugation class and they also exhibit both
stem and suffix allomorphy. I will focus on tense suffixes and suffix allomorphs in
Section€2.1, on theme vowels and stem allomorphs in Section€2.2. Section 2.3, then,
offers contrastive evidence, from two Atlantic creoles.

2.1â•… Tense and aspect suffixes

In KP, verbs are overtly marked for progressive, completive and past:
(1) a. Progressive suffix: -n
b. Completive suffix: -d
c. Past suffixes: -[w]
The progressive marker -n is derived from the Portuguese gerund marker -ndo
(Ptg.: bebendo “drinking”); the completive marker -d from the Portuguese participle
marker -do (Ptg.: bebido “drunk”), and the past marker from the velar glide -[w],
which is orthographically represented as -u (Ptg.: bebeu “drank”).
While the progressive suffix and the completive suffix are invariable, the past tense
marker has two allomorphs which are sensitive to the conjugation class of the verb:
the -[w] suffix selects second and third conjugation stems, whereas the suffix -o selects
first conjugation stems. Both past allomorphs constitute robust evidence in favour of
our claim that Indo-Portuguese creoles have preserved genuine conjugation classes:
(2) a. kat-o (KP)
sing.cl1-pst
“sang”
b. kum-e-[w]
eat-cl2-pst
“ate”
c. sub-i-[w]
go.up-cl3-pst
“went up”

.╅ Throughout this paper, the data on Indo-Portuguese has been drawn from Clements
(1996), Clements & Koontz-Garboden (2001) and Cardoso (2006).
.╅ For reasons of space, I will restrict my analysis to KP alone. As alluded to above, however,
there are at least two other Indo-Portuguese creoles which exhibit a very similar (if not exactly
identical) verbal paradigm, namely Daman Creole Portuguese and Diu Creole Portuguese. For
evidence, see Clements & Koontz-Graboden (2002) and Cardoso (2006).
 Ana R. Luís

2.2â•… Conjugation classes, theme vowels and stem allomorphs

As noted by Aronoff (1994:46), conjugations tell us “which inflectional affixes will realise
the various morphosyntactic properties that the verb bears in a particular instance”.
Conjugation classes, therefore, have formal value, in so far as they help us to predict
the inflected forms of a given lexeme. Looking in more detail at Table 1, we see that the
verbal paradigm of KP is divided into the three conjugation classes.

Table 1.╇ Verbal paradigm of KP

class 1 class 2 class 3

unmarked forms kat-a kum-e sub-i

sing-cl1.unm eat-cl2.unm go.up-cl3.unm
“sing” “eat” “go up”
past forms kat-o kum-e-[w] sub-i-[w]
sing-cl1.pst eat-cl2-pst go.up-cl3-pst
“sang” “ate” “went up”
progressive forms kat-a-n kum-e-n sub-i-n
sing-cl1-prog eat-cl2-prog go up-cl3-prog
“singing” “eating” “going up”
completive forms kat-a-d kum-i-d sub-i-d
sing-cl1-compl eat-cl3-compl go.up-cl3-compl
“sang” “eaten” “gone up”

Each conjugation class carries a distinct theme vowel: class 1 takes the theme
vowel -a, class 2 takes the theme vowel -e and class 3 the theme vowel -i. All three
theme vowels are genuine stem-forming units which combine with the root to form
the base to which inflectional endings attach. Unlike most other Romance-based
creoles, inflected verbs in KP indeed make the distinction between roots, stems and
inflected verb forms. For example, verb forms such as kumeu comprise three distinct
morphological units: the root kum-, the stem kume- and the past allomorph -[w].
Further supporting our claim that Indo-Portuguese verb forms are organised
into conjugation classes is the fact that verb stems undergo allomorphic variation. As
shown in Table 2, verbs belonging to the first and second conjugation classes take two
stems: one primary and one secondary stem.

Table 2.╇ Primary and secondary stems in KP

Primary stem secondary stem

class 1 kata-n kat-o

class 2 kume-n kumi-d
class 3 subi-d –
 The loss and survival of inflectional morphology 

In the first conjugation (see Table 2), past forms take a theme-less stem (e.g. kant-o),
rather than an a-final stem (e.g. kant-a-n); in the second conjugation, verb lexemes
take a third conjugation stems in the participle form (i.e. kum-i-d not *kum-e-d),
rather than the default e-final stem (e.g. kum-e-n and kum-e-w).
Crucially, stem-alternations such as these are not triggered by semantic or
morphosyntactic factors, but merely by the formal distinctions underlying the
Indo-Portuguese verbal paradigm. Overall, then, conjugation classes in KP deter-
mine both the shape of stems (e.g. kume- vs kumi-) and the shape of tense suffixes
(i.e. -o vs -[w]).

2.3â•… Theme vowels in other Portuguese-based creoles

Although theme vowels have remained inflectionally distinct units in KP (and other
Indo-Portuguese creoles, see footnote 3), quite often the borrowing of morphologically
complex word forms is associated to the loss of morpheme boundaries or, in the case
at hand, to the loss of the inflectional boundary between the verb root and the theme
vowel. As I will show next, this is exactly what happened in a number of Portuguese-
based creoles, whose theme vowels have been reanalysed as phonemes and integrated
into the phonological form of the root.
In Guiné Bissau Kriyol, for example, each verb lexeme exhibits only one (unmarked)
verb form (Kihm 1994). Unlike in KP, then, Kriyol verb forms are invariable:

(3) a. seta (Kriyol)

“accept”
b. kume
“eat”
c. bibi
“drink”

While each one of the verb forms in (3) appears to contain a verb root which is
followed by a theme vowel, closer inspection reveals that each putative theme vowel
(-a (in 3a), -e (in 3b) and -i (in 3c)) has lost its inflectional status and has become
part of the phonological structure of the verb root. In particular, the lack of verbal
suffixes clearly shows that the form of Kriyol verbs does not depend on the conjuga-
tion class of the lexeme. So, despite their verb-final vowel, Kriyol verbs are effectively
monomorphemic. The same view is formulated by Kihm (1994) who points out that
“in no case could these [Kriyol] vowels be analysed as affixes or thematic vowels as
in Portuguese”.
In Sotavento Kabuverdianu, on the other hand, verbs are not entirely deprived
of internal morphological structure. In addition to the bare form, this creole also
 Ana R. Luís

contains an inflected past form which takes the past marker -ba, as shown in ((4)–(6))
(Baptista 2003; Baptista et al. 2007. Despite the suffixation of -ba, however, the inflected
past forms are not sensitive to the verb’s conjugation class. More concretely, not only
does the suffix fail to induce allomorphic alternation on the stem but it also fails to
undergo allomorphic variation (4a–6a). In other words, the past stem is identical to
the base form, and the past suffix is invariable. Based on this data, it seems reasonable
to conclude that stem-final vowels constitute indeed lexicalised theme vowels which
no longer operate as conjugation class markers in Sotavento Kabuverdianu.

(4) a. falta (Sotavento Kabuverdianu)

“miss”
b. falta-ba
“miss-ant”

(5) a. kore
“run”
b. kore-ba
“run-ant”

(6) a. parti
“break”
b. parti-ba
“break-ant”

2.4â•… Summary
This section has shown that KP is sensitive to conjugation class distinctions, while
other Portuguese-based creoles are not. This effectively means that only KP con-
tains genuine theme vowels. On the contary, both Kabuverdianu (spoken in the
Sotavento islands) and Kriyol exhibit lexicalised theme vowels, i.e. theme vow-
els which have lost their inflectional status. In the following sections, I will try
to understand why theme vowels may indeed survive creolization, despite their
entirely meaningless nature.

3.â•… Inflectional categories in creoles

In an attempt to explain why certain inflectional categories are systematically lost

whereas others stand a higher chance of being retained, Kihm (2003) has drawn on
 The loss and survival of inflectional morphology 

the �distinction between inherent and contextual inflection. This distinction, as will
be shown below, has allowed a number of scholars within creole linguistics, includ-
ing Arends et al. (2006) and Plag (2008a), to account for the categorical selectivity
of creoles.

3.1â•… Inherent vs. contextual inflectional

The distinction formulated by Booij (1994, 1996) emerged within a theory-internal
debate about the architecture of grammar. Contextual inflection comprises inflectional
categories which a word must have due to the syntactic context within which it occurs
(see Table 2). It is “dictated by syntax, such as person and number markers on the verbs
that agree with the subject and/or objects, agreement markers for adjectives, and struc-
tural case markers on nouns” (Booij 1996:€2). Inherent inflection, on the other hand, is
“not required by syntax but has syntactic relevance”. Examples are the category number
for nouns, comparative and superlative degree of the adjective, and tense and aspect for
verbs” (Booij 1996:€2). A summary of this distinction is provided in Table 3.

Table 3.╇ Distinction between inherent and contextual inflection (Booij 1994, 1996)

inherent inflection contextual inflection

nouns number case

(in)definiteness

adjectives degree number

(in)definiteness
verbs tense subject agreement
mood object agreement
aspect
voice
infinitive
conjugation
classes

Historically, the distinction between inherent and contextual inflection originated

as an argument against the view that morphology should be split into two separate
components of grammar. Known as the Split Morphology hypothesis, its proponents
claimed that derivation should be placed in the pre-syntactic component and apply
before inflection, while inflection should be located in the post-syntactic component,
after derivation (Anderson 1982; Perlmutter 1988). Booij, however, notes that there
are in effect two types of inflectional morphology and that certain types of inherent
inflection (e.g. number, degree, infinitive) may also apply before derivation, arguing
instead that both derivation and inflection should be kept together in one morphological
component (Booij 1996).
 Ana R. Luís

Interestingly, a number of creolists have shown that Booij’s bi-partite classification

proves useful in accounting for the nature of creole inflection. Kihm (2003), in particu-
lar, first noted that creoles “exhibit little or no contextual inflection in comparison with
the lexifier or substrate languages” (2003:€334).4 Illustrating with data from Louisiana
Creole, Kihm shows that the domain of the verb phrase encodes both imperfect and
perfect markers, but no subject-verb agreement:
(7) a. Shop-la frem a siz-er. (Louisiana Creole)
Shop-the close.impf at six hour
“The shop (always) closes at six o’clock”
b. Shop-la frem-e a siz-er.
Shop-the close-perf at six hour
“The shop closed at six o’clock” (see Neumann 1985:€196)

Similarly, within the domain of the noun phrase, Kriyol retains plural formation, as
shown in (8a), but no nominal agreement, as in (8b).
(8) a. gatu/gatu-s (Kriyol)
cat/cat-pl
“cat”/“cats”
b. gatu-s branku (*gatus brankus)
cat-pl white
“white cats” (Kihm 2003:€354)

So, while inherent categories, both nominal and verbal, tend to survive creolisation,
“what creole languages indeed appear to be lacking is contextual inflection, i.e. struc-
tural case marking of subjects and direct objects (…) and syntactic noun-adjective and
verb-argument agreement” (Kihm 2003:€358).
In line with Kihm (2003), Arends et al. (2006) provide further evidence in favour of
Khim’s generalisation, drawing specifically on four Romance-based creoles – Angolar,
Palenquero, Seychellois and Papiamentu. Within the domain of the VP, all four creoles
use overt inflection to express TMA categories (e.g. the gerund marker in Palenquero
in (9) and Papiamentu in (10), but none expresses subject verb agreement.
(9) sosega/soseg-ando (Palenquero)
rest/rest-ger
“to rest”/“resting” (see Kouwenberg & Murray 1994:€21)
(10) pogke suto ta necesit-ando ele (Papiamentu)
because 1.pl prog need-ger 3.pl
“because we are needing them” (see Lorenzino 2000:€39)

.╅ Kihm (2003) draws his evidence from (a) Louisiana Creole, Lesser Antillean Creoles and
Haitian (French-based), (b) Kriyol and Kabuverdianu (Portuguese-based) and (c) Juba Arabic
and Nubi (Arabic-based).
 The loss and survival of inflectional morphology 

At the level of the NP, there is overt marking of number on nouns, as shown for
Palenquero, in (11), and definiteness on articles, but no adjective-noun agreement or
case making.

(11) a. ndo muhé -re (Palenquero)

two woman-pl
“two women”
b. ndo papé -le
twp paper-pl
“two pieces of paper” (Schwegler 1996:€261)

Based on these findings, Arends et al. correctly conclude that “the distinguishing prop-
erty of prototypical creoles indeed lies in the absence of contextual inflectional mor-
phology” (2006:15).

3.2â•… Inherent inflection in SLA

The distinction between inherent and contextual inflection has also found psycho-
linguistic support in recent research on SLA. As argued by Plag (2008a), L2 language
learners acquire contextual inflection at a later stage of their acquisition process because
contextual inflection involves information exchange either between phrases or within
phrases which requires more advance processing procedures. Inherent inflection, on
the contrary, “does not presuppose information exchange between different constitu-
ents”. The absence of contextual categories in L2 varieties is thus primarily the result
of “the limited L2 processing capacities that are characteristic of the early stages of
SLA” (Plag 2008b).
The observation that L2 varieties and creoles exhibit a similar range of inflectional
categories has lent further support to the claim that creole grammars are shaped by
SLA processes (Siegel 2004; Plag 2008a). Under an SLA approach to creolisation, then,
the survival of inherent inflection and the loss of contextual inflection are assumed to
follow from the limited processing competence of the L2 language learners in the early
stages of creole formation.5

3.3â•… Summary
Contra McWhorter (2001), this section has surveyed evidence which shows that the sur-
vival of inflectional morphology in creoles is not dependent on the semantic relevance
of inflectional categories, but rather on whether inflectional affixes express inherent or

.â•… A similar view has been formulated in Veenstra (2003) and Luís (2008), who argue that
the inflectionally poverty of creoles may, in effect, result from the fact that most creoles have
derived from ‘basic’ (i.e. inflectionally rudimentary) L2 varieties (Klein & Perdue 1997).
 Ana R. Luís

contextual categories. Psycholinguistic evidence further reinforces this view, lending

support to the influence of SLA processes on the shaping of creole grammars.

4.â•… The survival of formal inflection in Indo-Portuguese

This final section returns to the empirical problem formulated at the outset, namely the
fact that KP and other Indo-Portuguese creoles have retained conjugation class markers.
Even though not explicitly included in Booij’s early distinction, conjugation classes
(and also declension classes) are indeed instances of inherent inflectional categories.
As observed by Spencer (1992:11), conjugation classes “serve no other purpose than
to help create a base to which inflectional desinences attach and to define the separate
morphological classes (conjugations)”. Crucially, conjugation classes provide indepen-
dent information about the verb and are not forced on the speaker by the syntactic
context (see also Katamba 1993:€224).6 On the contrary, person and number agree-
ment is assigned to the verb when it occurs in a larger syntactic context. In the case
of agreement, the features expressed on the verb are determined by the properties of
some other word or phrase in the same clause.
Based on this evidence, the survival of purely meaningless conjugation classes in
KP (and Indo-Portuguese) nicely shows that creoles target inherent inflection regard-
less of its semantic content (Luís 2008). Under an SLA approach to creole formation,
such retention follows from the fact that inherent inflection is acquired at an earlier
stage than contextual inflection (see section 3.2).
One caveat however is in order here. Having argued that conjugation classes are
not necessarily blocked by creolization, the question as to why these categories are so
extremely rare in Romance-based creoles remains open.7 The answer to this question,
I believe, can be found in the origins of Indo-Portuguese creoles. Sociolinguistically,
for example, Clements (1996) argues that the context within which KP and other
Indo-Portuguese creoles emerged differs in a number of ways from that of most other
Portuguese-based creoles. In particular, Clements claims that the isolation of substrate
speakers from their native community may have encouraged adult Marathi speakers
to learn more than just a rudimentary version of the L2.

.╅ Like other inherent categories, conjugation classes are also more likely to appear closer to
the root (Booij 1996) and more likely to induce allomorphic shape alternations on the base or
stem than contextual inflection (Booij 1996; Haspelmath 2002).
.╅ I am grateful to an anonymous reviewer for urging me to spell out this idea.
 The loss and survival of inflectional morphology 

This idea that L2 acquisition may benefit from the language learner’s receptive
attitude towards the target language has also been observed by Siegel (2004). As Siegel
points out, “limited second language acquisition” may result from “the limits learners
impose on themselves due to factors relating to identity or resistance”, rather than from
the limited degree of exposure to the Target Language (TL). So, adult language learners
who refuse to identify with the TL community tend to learn an L2 variety merely for
basic communicative purposes.
Assuming, then, that the development of Indo-Portuguese creoles developed
within a favourable learning context, it might seem plausible to suggest that these cre-
oles derived from a more elaborate (or advanced) L2 variety, which may have already
contained a wider range of inherent inflection (Luís 2008).8 As to the inflectionally
more impoverished creoles, I will assume that their grammar derived from ‘basic’ or
rudimentary L2 varieties containing hardly any overt inflection.
In addition to the sociolinguistic and attitudinal factors, other factors may have
played a role in shaping Indo-Portuguese grammar. Clements (1996), for example,
notes that Indo-Portuguese creoles differ from most of the other Portuguese-based in
two other important ways: (a) they emerged from the contact between Portuguese and
only one substrate language (i.e. Marathi for KP and Gujarati for both Daman and Diu
Portuguese Creoles), and (b) the substrate languages underlying Indo-Portuguese are
typologically inflecting and suffixing.
Overall, then, the emergence of a more elaborate kind of inherent inflection
in Indo-Portuguese creoles may have had its origins in a more advanced L2 variety
spoken by adult substrate speakers prior to creole formation. This L2 variety may
have contained a wider range of inherent inflectional categories and it may have
been shaped by sociolinguistic, attitudinal and, perhaps to some extent, also typo-
logical factors.9

.â•… Such advanced L2 variety is also known as ‘post-basic’ variety (Klein & Perdue 1997;
Venstra 2003).
.╅ However, the claim that the typological properties of the substrate languages may influence
the retention of overt inflection appears to be somewhat problematic. Among Asian creoles,
Sri Lanka Creole Portuguese has retained hardly any verbal inflection, even though one of
its substrate languages, Sinhala, contains both conjugation classes and tense suffixes. On the
contrary, Indo-Portuguese creoles, which have retained theme vowels, derive from substrate
languages without conjugation classes. Of course, since both creoles differ in many more ways,
it is difficult to exclude the hypothesis that other factors may have triggered the loss of verbal
endings in Sri Lanka Creole Portuguese.
 Ana R. Luís

5.â•… Conclusion

In this study, I have examined the verbal inflection of KP and investigated the causes
determining the loss and survival of its inflectional categories. Of crucial importance
has been the presence of purely formal conjugation class markers vs. the loss of seman-
tically meaningful person-number endings.
In previous studies, I have highlighted the importance of attitudinal and socio-
linguistic forces in triggering the overall emergence of inflectional morphology in
creoles (Luís 2008). Drawing on Clements’ (1996) exhaustive discussion of the socio-
cultural context within which KP took place, in combination with Siegel’s (2004)
insights about adult SLA processes, I took with Clements (1996) the view that KP
emerged from a favourable socio-linguistic situation which encouraged adult speakers
to learn the TL.
Such socio-cultural and attitudinal factors however were insufficient to explain
why creoles – despite their inflectional poverty – are selective with respect to their
inflectional features. Or, put more concretely, why creoles such as KP (and other Indo-
Portuguese creoles) preserve purely formal theme vowels and discard perfectly meaning-
ful agreement endings. Drawing on recent work by Kihm (2003), I have provided both
empirical and psycholinguistic evidence showing that the retention of conjugation
class markers follows from the tendency for creoles to favour inherent inflection.

References

Anderson, Stephen. 1982. “Where’s Morphology?” Linguistic Inquiry 13. 571–612.

Arends, Jacques, Josje Verhagen, Eva van Lier, Suzanne Dikker & Hugo Cardoso. 2006. “On the
Presence versus Absence of Morphological Marking in four Romance-based Creoles”.
The Structure of Creole Words: Segmental, syllabic and morphological aspects, ed. by Parth
Bhatt€& Ingo Plag, 223–241. Tübingen: Niemeyer.
Aronoff, Mark. 1994. Morphology by Itself. Mass.: MIT Press.
Baptista, Marlyse. 2003. “Inflectional Plural Marking in Pidgins and Creoles: A comparative study”.
Phonology and Morphology of Creole Languages, ed. by Ingo Plag, 315–332. Tübingen:
Max Niemeyer.
Baptista, Marlyse, Helena Mello & Miki Suzuki 2007. “Cape Verdean Creole and Guinea-Bissau
Creole”. Comparative Creole Syntax, ed. by John Holm & Peter Patrick, 53–82. London:
Battlebridge.
Booij, Geert. 1996. “Inherent versus Contextual Inflection and the Split Morphology Hypothesis’.
Yearbook of Morphology 1995, 1–16. Dordrecht: Kluwer.
Booij, Geert. 1994. “Against Split Morphology”. Yearbook of Morphology 1993, 27–49. Dordrecht:
Kluwer.
Cardoso, Hugo. 2006. “Diu Indo-Portuguese: Change and continuity since 1882”. Paper pre-
sented at the Associação de Crioulos de Base Lexical Portuguesa e Espanhola, University
of Coimbra.
 The loss and survival of inflectional morphology 

Clements, J. Clancy. 1996. The Genesis of a Language: The Formation and Development of Korlai
Portuguese. Amsterdam, Philadelphia: John Benjamins.
Clements, J. Clancy & Andrew Koontz-Garboden. 2002. “Two Indo-Portuguese Creoles in
Contrast”. Journal of Pidgin and Creole Languages 17: 2.191–236.
DeGraff, Michel. 2005. “Morphology and Word Order in ‘Creolization’ and Beyond”. The Oxford
Handbook of Comparative Syntax ed. by Guglielmo Cinque & Richard S. Kayne, 293–372.
Oxford: OUP.
Haspelmath, Martin. 2002. Understanding Morphology. Oxford: Arnold.
Katamba, Francis. 1993. Morphology. London: Macmillan.
Kihm, Alain. 2003. Inflectional Categories in Creole Languages. Phonology and Morphology of
Creole Languages ed. by Ingo Plag, 333–363. Tübingen: Max Niemeyer.
Kihm, Alain. 1994. Kriyol Syntax. The Portuguese-based Creole Language of Guiné-Bissau.
Amsterdam: John Benjamins.
Klein, Wolfgang & Clive Perdue. 1997. The Basic Variety (or: Couldn’t natural languages be
much simpler?). Second Language Research 13.4, 301–347.
Kouwenberg, Silvia & Eric Murray. 1994. Papiamentu. München/Newcastle: Lincom Europa.
Lorenzino, Gerardo. 2000. The Morphosyntax of Spanish-lexified Creoles. München/Newcastle:
Lincom Europa.
Luís, Ana R. 2008. “Tense Marking and Inflectional Morphology in Indo-Portuguese Creoles”.
Roots of Creole Structures: Weighing the contribution of substrates and superstrates ed. by
Susanne Michaelis, 83–121. Amsterdam: John Benjamins.
McWhorter, John H. 2001. The World’s Simplest Grammars are Creole Grammars. Linguistic
Typology 5. 125–166.
Neumann, Ingrid. 1985. Le créole de Beaux Bridge, Louisiane: morphosyntaxe, textes, vocabulaire.
Hamburg: Helmut Buske.
Perlmutter, David M. 1988. “The Split Morphology Hypothesis: Evidence from Yiddish”. Theo-
retical Morphology ed. by Michael Hammond & Michael Noonan, 79–100. San Diego:
Academic Press, Inc.
Plag, Ingo. 2008a. “Creoles as Interlanguages: Inflectional morphology”. Journal of Pidgin and
Creole Languages 23.1. 109–130.
Plag, Ingo. 2008b. “Creoles as Interlanguages: Syntactic structures”. Journal of Pidgin and Creole
Languages 23.2. 307–328.
Schwegler, Armin. 1996. Chi ma nkongo: Lengua y rito ancestrales en El Palenque de San Basilio
(Colombia). Madrid: Iberoamericana.
Siegel, Jeff. 2004. “Morphological Simplicity in Pidgins and Creoles”. Journal of Pidgin and Creole
Languages 19.1. 139–62.
Spencer, A. 1992. Morphological Theory. Oxford: Backwell.
Thomason, Sarah & Terrence Kaufman.1988.€ Language Contact, Creolization, and Genetic
Linguistics. Berkeley: University of California Press.
Veenstra, Tonjes. 2003. “What Verbal Morphology Can Tell us about Creole Genesis”. Phonology
and Morphology of Creole Languages, ed. by Ingo Plag, 293–313. Tübingen: Max Niemeyer.
 Allomorphy in pre-clitic imperatives
in Formenteran Catalan
An output-based analysis*

Eulàlia Bonet & Francesc Torres-Tamarit

Universitat Autònoma de Barcelona

In some Romance languages second person singular imperatives often surface

with an added extension, an accretion, when they are followed by enclitics. This
paper examines this phenomenon in Formenteran Catalan, where the accretion
has different shapes depending on the verb. We argue that the appearance of
the accretion is caused by a phonological constraint requiring a right-aligned
moraic trochee. The form of the imperative with the accretion coincides with the
inflectional stem that is found in other forms of the imperative paradigm. We
propose an output-based analysis in which Lexical Conservatism (Steriade 1999,
2008) and Uniform Exponence (Kenstowicz 1996, 1997, 2005) play a key role.

1.â•… Introduction

In Catalan, 2nd person singular (2sg) imperatives of conjugations II and III consist of a
bare stem, which surfaces as such in all contexts except under enclisis, where the form
of the imperative changes: additional material appears to the right of the stem; this
additional material, an accretion, has a different shape depending on the dialects and

*We wish to thank Florentina Mayans and Vicent Ferrer, our consultants for the Formenteran
Catalan data. Special thanks go to Adam Albright and Joan Mascaró for insightful sugges-
tions, and to Donca Steriade for lots of feedback and fruitful discussion. We are also indebted
to the audiences of LSRL 39 and the Phonology Circle (MIT) for helpful comments, and to
the suggestions and criticisms of two anonymous reviewers that have allowed us to improve
the written version of the paper. This research is sponsored by the Spanish Ministerio de
Ciencia e Innovación (project HUM2006-13295-CO2-01 and a grant, to Eulàlia Bonet, from
the Secretaría de Estado de Universidades y de Investigación), and also by the Generalitat de
Catalunya (research group 2009SGR1079).
 Eulàlia Bonet & Francesc Torres-Tamarit

on the verb. The examples in (1) and (2) illustrate this point with data from Central
Catalan, Majorcan and Formenteran (the accretion appears underlined).1
(1) prometre: “to promise” (conj.II)
in isolation with enclitics
a. Central Catalan [pnuÄmεt] [pnuÄmεtәli]
b. Majorcan [pnoÄmәt] [pnomәtәÄli]
c. Formenteran [pnuÄmәt] [pnumәÄtәli]
“promise!” “promise to him/her!”
Cf. [li pnuÄmεt] (Central Catalan)
to him/her promises (Indicative)
“(s/he) promises to him/her”
(2) bullir: “to boil” (conj.III)
in isolation with enclitics
a. Central Catalan [Äbuˆ] [Äbuˆәlә]
b. Majorcan [Äbuˆ] [buĈilә]
c. Formenteran [Äbuˆ] [buˆiÄ>әlә]
“boil!” “boil it(fem)!”
Cf. [lә Äβuˆ]
it(fem) boils (Indicative)
“(s/he) boils it”

In this paper we focus on Formenteran, which shows three different accretions

depending on the conjugation type. They appear illustrated in (3). Conjugation II has
two different patterns that depend on the presence (IIb) or absence (IIa) of a velar
segment, an extension, in certain forms of the verbal paradigm, and in the context of
enclisis. Conjugation III has two subtypes (IIIa and IIIb) that differ with respect to the
appearance, in IIIa, of a palatal extension /ә∫/ that avoids lexical stress on the root, a
phenomenon common also in other Romance languages. The extension /ә∫/ does not
appear when there is enclisis.
(3) Formenteran 2sg imperatives in enclisis
conj. in isolation with enclitics accretion
IIa Äpεnt pәnÄð-ә#lә ә “lose (it(fem))!”
IIb әÄpnәn әpnәŋ-Ä:ә#lә :ә “learn (it(fem))!”
IIIa әfәÄŠ-ә∫ әfәŠ-iÄ>ә#lә i>ә “add (it(fem))!”
IIIb Äbuˆ buˆ-iÄ>ә#lә i>ә “boil (it(fem))!”

.╅ Majorcan and Formenteran are varieties of Catalan spoken in the Balearic islands.
 Allomorphy in pre-clitic imperatives in Formenteran Catalan 

In the sections that follow we argue that these accretions appear for phonological
reasons but they are taken from other forms of the verb. This idea is best captured
resorting to the notions of Lexical Conservatism (Steriade 1999, 2008), and Uniform
Exponence (Kenstowicz 1996, 1997, 2005).

2.â•… Phonological motivation for the presence of an accretion

In Formenteran, as in some other Romance varieties, the presence of pronominal

enclitics triggers stress shift. The resulting pattern is a moraic trochee aligned at the
right edge of the V+Cl sequence.2 For space reasons the specific constraints respon-
sible for this pattern are grouped here under the cover constraint µTrochee], stated
in (4). For a more detailed analysis see Torres-Tamarit (2008).
(4) µTrochee]: assign one violation mark for any V+Cl sequence that lacks
a moraic trochee aligned at the right edge.

In (5) the stress shift caused by the constraint µtrochee] is illustrated with the impera-
tive of conjugation I verb cantar “to sing” followed by one or two clitics. From now on
we indicate the boundary between the verb and the clitic with the hash sign, #.
(5) a. Without enclisis: [Äkantә] “sing!”
b. With enclisis: [kәn(Ätә#lә)] “sing it(fem)!”
[kәn(Ätә#li)] “sing to him/her!”
[kәn(Ätә#m)] “sing to me!”
[kәntә#(Ämә#lә)] “sing it(fem) to me!”

As can be seen in (5), the 2sg imperative of conjugation I verbs ends in a vowel, an
inflectional vowel, and the addition of a clitic allows an unproblematic formation of a
right-aligned moraic trochee. However, 2sg imperatives of conjugations II and III end
in a consonant belonging to the stem; in these cases it is not possible to construct a
moraic trochee at the right edge of the V+Cl sequence, as shown in (6a). The presence
of the accretion solves this conflict, as can be seen in (6b).
(6) a. µTrochee] violated b. µTrochee] satisfied
*[(Äpεnt)#lә] [pәn(Äð-ә#lә)] “lose (it(fem))!”
*[ә(Äpnәn)#lә] [әpnәŋ-(Ä:ә#lә)] “learn (it(fem))!”
*[әfә(ÄŠәŠ)#lә] [әfәŠ-i(Ä>ә#lә)] “add (it(fem))!”
*[(Äbuˆ)#lә] [buˆ-i(Ä>ә#lә)] “boil (it(fem))!”

.â•… A well formed moraic trochee consists of two light syllables (ÄLL) or one heavy syllable
(ÄH). See Hayes (1995). A sequence HL cannot be parsed as a perfect moraic trochee because
it contains more than two moras.
 Eulàlia Bonet & Francesc Torres-Tamarit

One might wonder why an accretion is inserted instead of a plain epenthetic vowel:
e.g. *[bu(Ĉәlә)]. Epenthesis would equally avoid a violation of µTrochee]. As a matter
of fact, even though schwa epenthesis in Catalan is very common in certain contexts,
it is hardly ever used to break the contiguity relation between a verb and a clitic, as has
been argued in Bonet & Lloret (2005) for Central Catalan. Epenthesis is peripheral
whenever possible. For instance, in an underlying sequence like /pәndәm#n/ “let us lose
some!”, the consonant of the clitic cannot be syllabified and an epenthetic vowel has
to be inserted. There are two possible sites for the schwa: [pәnÄðәmnә], *[pәnÄðәmәn].
Both forms are equally bad with respect to syllabic wellformedness constraints, but
only the one which contains a peripheral epenthetic vowel allows the verb and the
clitic to be strictly adjacent: [pәnÄðәmnә]. This tight adjacency between the verb and
clitics is attributed by Bonet & Lloret (2005) to the very high ranking of the constraint
Align (V−Cl), stated in (7).
(7) Align (V−Cl): align the left edge of a pronominal clitic with the right edge
of the verb.

It is obvious that the presence of accretions like [:ә] or [i>ә] cannot be attributed to
phonological epenthesis, and we have just seen that epenthesis is peripheral to the
V+Cl sequence. These facts taken together indicate that even in cases like [pәnÄð-ә#lә],
illustrated in (3), the schwa that intervenes between the verb stem and the clitic has a
morphological affiliation; it is not an epenthetic vowel.

3.â•… Nature of the accretion

At first sight the choice of accretion that can be found, [ә], [:ә] or [i>ә], might seem
puzzling. However, once the pre-clitic 2sg imperatives are compared with the impera-
tive paradigm of each verb the puzzle disappears. The examples (8) to (11) are the
same ones that were shown in (3), but they appear now next to the whole imperative
paradigm.3 Notice that the form of the 2sg imperative in enclisis is the same that is
found in the plural forms of the imperative paradigm in isolation, excluding the per-
son/number markers [m] (for 1pl) and [w] (for 2pl), as shown by the underlining. The
relevant forms of the paradigm appear shaded; a hyphen isolates the verbal root and
person endings from the rest of the verbal morphology. Henceforth, verbal forms in
italics indicate the corresponding standard orthographic form.

.╅ Catalan, like other Romance languages, has a semantically (polite) 2nd person that is mor-
phologically 3rd person. In Formenteran imperatives this form is incompatible with enclisis,
and will be ignored in the rest of the paper.
 Allomorphy in pre-clitic imperatives in Formenteran Catalan 

(8) Perdre “to lose”

Imperative IIa 2sg in enclisis
1pl pәnÄð-ә-m pәnÄð-ә#lә
2sg Äpεnt 2pl pәnÄð-ә-w

(9) Aprendre “to learn”

Imperative IIb 2sg in enclisis
1sg әpnәŋ-Ä:ә-m әpnәŋ-Ä:ә#lә
2sg әÄpnәn 2pl әpnәŋ-Ä:ә-w

(10) Afegir “to add”

Imperative IIIa 2sg in enclisis
1pl әfәŠ-iÄ>ә-m әfәŠ-iÄ>ә#lә
2sg әfәÄŠ-ә∫ 2pl әfәŠ-iÄ>ә-w

(11) Bullir “to boil”

Imperative IIIb 2sg in enclisis
1pl buˆ-iÄ>ә-m buˆ-iÄ>ә#lә
2sg Äbuˆ 2pl buˆ-iÄ>ә-w

What we have called the accretion, [ә], [:ә] and [i>ә], which appears in 2sg impera-
tives under enclisis, corresponds to morphological material that appears between the
root and the person/number markers (φ-features) in 1pl and 2pl imperatives. There
is no consensus in the literature on the morphological status of this verbal material.
The schwa in examples like [pәnÄð-ә-m] or [pәnÄð-ә-w], present also in the sequence
with enclitics [pәnÄð-ә#lә], in (8), has usually been interpreted as a thematic vowel or
conjugation vowel (see, for instance, Mascaró 1986 or Oltra-Massuet 1999). Forms like
[әpnәŋ-Ä:ә-m] in (9), also present in [әpnәŋ-Ä:ә#lә], with a clitic, contain in addition
the velar extension [:] present in different forms of certain conjugation II verbs in
all dialects of Catalan (1sg Present Indicative, all persons of the Present Subjunctive,
Imperfect Subjunctive, Participle). This segment has spread to other verbs in Balearic
Catalan. The analysis of this velar extension is debated (see, for different interpreta-
tions, Wheeler 1985; Mascaró 1986; Pérez-Saldanya 1995; Viaplana 1995, among
others). Finally, forms like [әfәŠ-iÄ>ә-m] in (10) or [buˆ-iÄ>ә-m] in (11), also found in
[әfәŠ-iÄ>ә#lә] and [buˆ-iÄ>ә#lә], respectively, contain the segments [i>ә]: the last one
could be interpreted as the thematic vowel [ә], mentioned above; the segment [>] is the
spirantized allophone of the velar extension /:/ present also in conjugation IIb verbs;
finally, the first vowel, [i], is typical of conjugation III verbs; alternatively the sequence
[i>] can be interpreted as an allomorph of /:/. The exact morphological nature of all
 Eulàlia Bonet & Francesc Torres-Tamarit

these segments is not relevant for our purposes. What is important is (a) that in accre-
tions like [:ә] and [i>ә] at least two morphs are involved, and (b) that the accretions
do not contain morphosyntactic information (such as tense, mood, aspect, person or
number); they are some kind of thematic material.
As was shown in Section€ 2, with the examples in (6b), the appearance of the
accretion in 2sg imperatives under enclisis allows the construction of a moraic trochee
at the right edge of the V+Cl sequence. In 2sg imperatives this addition is possible
because the accretion is located after the bare root, where thematic material can appear
(compare (12b) with (12b′)). For 1pl or 2pl forms like [buˆ-iÄ>-ә-m] or [buˆ-iÄ>-ә-w]
the enclitized forms, like [buˆ-i(Ä>-ә-m)#lә] or [buˆ-i(Ä>-ә-w)#lә], do not surface
with a right-aligned moraic trochee. Here the problem cannot be repaired because no
accretion, no thematic material, is available to the right of φ-features (compare (12a)
with (12a′)).4
(12) Root – thematic material – φ-features Cl
a. 1pl: buˆ i(Ä>ә m)
a′. 1pl#Cl: buˆ i(Ä>ә m) #lә
b. 2sg: buˆ
b′. 2sg#Cl: buˆ i(Ä>ә #lә)

4.â•… Analysis

4.1â•… The role of corrLEX constraints

In previous sections we have argued that Formenteran avoids a phonological problem,
namely the impossibility of building a right-aligned moraic trochee, not by resorting to
phonological epenthesis but by using morphological material that is already present in
other forms of the same imperative paradigm. In other words, this phenomenon is an
instance of Lexical Conservatism, a notion developed in Steriade (1999, 2008), which
can be defined as “a class of grammatical conditions […] promoting the use of pre-
existing, familiar expressions, or parts or properties of such expressions. They penalize
the use of unprecedented, linguistically innovative expressions.” (Steriade 1999:€244).
CorrLEX constraints evaluate a correspondence relation established between candi-
dates and listed output forms, which constitute the base of an asymmetric relation.
The CorrLEX constraint in (13), evaluates a correspondence relation between listed

.╅ Similarly, infinitives and gerunds (the other verbal forms that allow enclisis), which end in
a consonant, violate the constraint µTrochee] in enclisis (cf. [kәn(Ätan)#lә] “singing it(fem)!”).
Here the accretion is also impossible because these verbal forms end in a tense marker and no
thematic material can be added afterwards.
 Allomorphy in pre-clitic imperatives in Formenteran Catalan 

output inflectional stems (root plus thematic material without φ-features) and impera-
tives under enclisis.

(13) CorrLEX InflstemImp (CorrLEXI): assign one violation mark for any
inflectional stem of a pre-clitic imperative that does not have a correspondent
in the inflectional stem of an imperative form (the base).

The constraint CorrLEXI is violated by any candidate whose inflectional stem does
not correspond to another inflectional stem of the imperative paradigm. This corre-
spondence relation is expressed through subscripts in the tableaux that illustrate the
analysis below. The constraint CorrLEXI is highly ranked in Formenteran.
A more specific version of this constraint penalizes any candidate whose inflec-
tional stem does not coincide with one that has the same φ-features. This constraint,
which is low ranked, is stated in (14).

(14) CorrLEX InflstemImp-φ (CorrLEXI-φ): assign one violation mark for any
inflectional stem of a pre-clitic imperative that does not have a correspondent in
the inflectional stem of an imperative form with the same φ-features (the base).

The fact that the surfacing inflectional stem in pre-clitic 2sg imperatives is always a
fully faithful copy of the inflectional stem of other imperative forms (namely 1pl and
2pl forms), and not a partial copy or an imperfect copy, means that all the OO faith-
fulness constraints Max, Dep and Ident are highly ranked. In (15) they appear as the
cover constraint Faith-OO.

(15) Faith-OO: cover constraint that includes Max-oo, Dep-oo and Ident-OO.

Summarizing so far, the two CorrLEX constraints we have proposed control what kinds
of correspondence relations are permitted, and OO faithfulness constraints control the
degree of similarity between two corresponding forms.5
The tableau in (16) illustrates the effects of all the constraints that have been intro-
duced so far. Under Listed output inflectional stems we have included all the inflec-
tional stems that can be found in the imperative paradigm in isolation, the only ones
that will be relevant with respect to the constraint CorrLEXI. The form [Äpεnti] is found
in the 2sg form; it is a free-standing form that is, at the same time, an inflectional stem.
The inflectional stem [pәnÄðәj] is found in 1pl and 2pl and is not a free-standing form;
in the listed output forms this inflectional stem appears with person/number markers
in parentheses (e.g. [pәnÄðәj (m,w)]).

.╅ The CorrLEX constraints and Faith-OO are asymmetrical in the sense that they take as
a Base a verb [V] (minus φ-features) and as the derived form the verb plus clitics [V+cl]. For
asymmetrical OO relations see Benua (1997).
 Eulàlia Bonet & Francesc Torres-Tamarit

(16) Perd-la “lose it(fem)!”: /pεnd#lә/ → [pәnÄðә#lә]

Listed output inflectional stems: [Äpεnti], [pәnÄðәj (m,w)]

/pεnd#lә/ µTrochee] Faith-OO CorrlexI CorrlexI-φ

a. (Äpεnti) lә *!
b. (Äpεni) lә * *! (M)
c. pәn(Äðәj lә) *
d. pәn(Äðij lә) *! (I) *
e. (Äpεn)ðik lә * *! *

Candidates (16a) and (16b) have a correspondent in the listed form [Äpεnti], as shown
by the subscript i. These candidates violate neither CorrLEXI (because the listed form
[Äpεnti] belongs to the imperative paradigm) nor CorrLEXI-φ, because their corresÂ�
pondent belongs to the 2sg and therefore has the same φ-features. Candidates (16c)
and (16d) have a correspondent in the listed form [pәnÄðәj (m,w)], and therefore satisfy
CorrLEXI; they violate CorrLEXI-φ because their correspondent inflectional stem
does not have the same φ-features. The candidate in (16e) violates the two CorrLEX
constraints because it has no correspondent in the imperative, as the use of a differ-
ent subscript (k) indicates. Faith-OO is violated by candidates (16b) and (16d) for
different reasons. In (16b) there is a Max-OO violation (M) because the listed form
[Äpεnti] contains a consonant that is not present in the inflectional stem of the can-
didate, [(Äpεni) lә]. In (16d) there is an Ident(f) violation (I) because the inflectional
stem of the listed form [pәnÄðәj (m,w)] ends in a schwa, which has a different feature
specification from the stem-final [i] of the candidate, [pәn(Äðij lә)]. The three top-
ranked constraints µTrochee], Faith-OO and CorrLEXI ensure that candidate (16c)
is the winner, even though it violates the low-ranked constraint CorrLEXI-φ. This
tableau also shows that µTrochee] dominates CorrLEXI-φ. In the Standard variety
of Catalan the winner would be (16b). This result would be achieved through the
promotion of CorrLEXI-φ, which would become a top-ranked constraint; the winner
would not be an exact copy of the inflectional stem of the listed output form [Äpεnti]
due to the higher ranking of the relevant Markedness constraints on syllable structure
over Max-OO.
As mentioned in Section€3, in 1pl or 2pl imperatives nothing is added or deleted to
avoid the stress requirement, the need for a right-aligned moraic trochee. For instance,
an input like 1pl /pεndәm#lә/ “let us lose it(fem)!” does not surface without the per-
son/number morph, /m/ (to become *[pәn(Äðәj lә)]), but preserves it, [pәn(Äðәjm) lә].
The constraint Max-φ (a specific version of Realize Morpheme, Walker 2000; Kurisu
2001) protects underlying person/number morphs from being deleted, as illustrated in
(17). Max-φ dominates µTrochee].
 Allomorphy in pre-clitic imperatives in Formenteran Catalan 

(17) Perdem-la “let us lose it(fem)!”: /pεndәm#lә/ → [pәnÄðәm#lә]

Listed output inflectional stems: [Äpεnti], [pәnÄðәj (m,w)]
/pεndәm#lә/ Max-φ µTrochee] Faith-OO CorrLEXi CorrLEXi-φ
a. pәn(Äðәjm) lә *
b. pәn(Äðәj lә) *!

The whole set of accretions can be obtained with the constraints given so far, as illus-
trated in tableaux (18) to (20).
(18) Aprèn-la “learn it(fem)!”: /әpnәn#lә/ → [әpnәŋÄ:ә#lә]
Listed output inflectional stems: [әÄpnәni], [әpnәŋÄ:әj (m, w)]
/әpnәn#lә/ Max-φ µTrochee] Faith-OO CorrLEXI CorrLEXI-φ
a. ә(Äpnәni)lә *!
b. әpnәŋ(Ä:әj lә) *
c. әpnә(Änәj lә) *! (M) *
d. әpnә(Änәi lә) *! (D)

The fully faithful candidate in (18a) is ruled out by µTrochee], and the winner is
(18b), the candidate that has an inflectional stem that is otherwise found in 1pl and 2pl
forms. The candidates in (18c) and (18d) are phonetically identical, and both violate
Faith-OO, but they correspond to different listed output inflectional stems. On the
one hand, the candidate in (18c), [әpnә(Änәj lә)], has a correspondent [әpnәŋÄ:әj (m, w)]
and therefore incurs a violation of Max-OO (M), because a consonant [:] has been
deleted. On the other hand, the candidate in (18d), [әpnә(Änәi lә)], has a correspondent
[әÄpnәni] and therefore violates Dep-OO (D), because a vowel [ә] has been inserted.
This candidate also shows that Faith-OO (Dep-OO to be more precise) dominates
CorrLEXI-φ.
The tableau in (19) contains a conjugation IIIb 2sg imperative. Here the form sur-
faces under enclisis with the accretion [i>ә], which contains the thematic vowel typical
of conjugation III verbs [i].
(19) Bull-la “boil it(fem)!”: /buˆ#lә/ → [buˆiÄ>ә#lә]
Listed output inflectional stems: [Äbuˆi], [buˆiÄ>әj (m,w)]
/buˆ#lә/ Max-φ µTrochee] Faith-OO CorrLEXI CorrLEXI-φ
a. (Äbuˆi) lә *!
b. buˆi(Ä>әj lә) *
c. bu(Ĉij lә) *!*(M) *
d. bu(Ĉik lә) *! *
 Eulàlia Bonet & Francesc Torres-Tamarit

In the tableau in (20), below, the verbal form is a conjugation IIIa 2sg imperative. As
mentioned earlier, this class has an extension /ә∫/ that is present only to avoid lexi-
cal stress on the root. This extension appears in the 2sg imperative in isolation (e.g.
[әfәÄŠә∫] “add!”), but not in enclisis (e.g. [әfәŠiÄ>ә#lә] “add it(fem)!”). As shown in (20),
the lack of the extension in enclisis is due to the fact that the inflectional stem that
is chosen is the one usually found in 1pl and 2pl imperatives, which never surface
with the extension. Leaving aside the fully faithful candidate (20a) and the winner in
(20b), which have already been accounted for, it is worth wondering why a candidate
like (20c), [әfә(ÄŠik€lә)], cannot be the optimal output. Like the winner, this candidate
satisfies the stress-related constraint µTrochee]. This candidate lacks the extension
but has the thematic vowel typical of conjugation III verbs, [i]. This form is in corres�
pondence with other forms of the verbal paradigm (Imperfect Indicative and Subjunc-
tive, Future, Conditional, Gerund, Infinitive), but not with the imperative. For this
reason this candidate violates not only CorrLEXI-φ, like the winner, but also the more
general CorrLEXI.

(20) Afegeix-la “add it(fem)!”: /әfәÄŠә∫#lә/ → [әfәŠiÄ>ә#lә]

Listed output inflectional stems: [әfәÄŠә∫i], [әfәŠiÄ>әj (m,w)]

/әfәŠә∫#lә/ Max-φ µTrochee] Faith-OO CorrLEXI CorrLEXI-φ

a. әfә(ÄŠәŠi)lә *! *(I)
b. әfәŠi(Ä>әj lә) *
c. әfә(ÄŠik lә) *! *

Contrary to conjugations II and III, in conjugation I verbs no apparent changes are

found in enclisis. A 2sg imperative in isolation like [Äkantә] “sing!” seems to be subject
only to predictable vowel reduction to schwa due to stress shift in enclisis, [kәnÄtә#lә]
“sing it(fem)!”; no additional phonological material appears. Under the present analy-
sis, the contraint ranking needed to account for conjugations II and III accounts also
for conjugation I. As illustrated in (21), the winner surfaces with the inflectional stem
corresponding to the plural forms, as we had seen for all the other cases.

(21) Canta-la “sing it(fem)!”: /kantә#lә/ → [kәnÄtә#lә]

Listed output inflectional stems: [Äkantәi], [kәnÄtәj (m, w)]

/kantә#lә/ Max-φ µTrochee] Faith-OO CorrLEXI CorrLEXI-φ

a. (Äkan)tәi lә *!
b. kәn(Ätәj lә) *
 Allomorphy in pre-clitic imperatives in Formenteran Catalan 

4.2â•… The role of Uniform Exponence

In previous sections the examples that were given to argue for the constraint ranking
that had proved adequate contained the clitic la (3sg fem accusative). Most Catalan
pronominal clitics are consonant-initial. Some other examples appear in (22).
(22) Bull-la “boil it(fem)!”: /buˆ#lә/ → [buˆi(Ä>ә#lә)]
Bull-lo “boil it(masc)!”: /buˆ#l/ → [buˆi(Ä>ә#l)]
Bull-me “boil for me!”: /buˆ#m/ → [buˆi(Ä>ә#m)]
Bull-te “boil for yourself!”: /buˆ#t/ → [buˆi(Ä>ә#t)]
Bull-li “boil for him/her!”: /buˆ#li/ → [buˆi(Ä>ә#li)]
Bull-ne “boil some!”: /buˆ#n/ → [buˆi(Ä>ә#n)]

However, there are two vocalic clitics in Catalan, hi /i/ (locative), and ho /u/ (neuter). In
these cases, as shown in (23), the accretion appears even though the fully faithful can-
didate would not violate µTrochee] or any other relevant phonological constraints.
(23) a. /buˆ#u/: [buˆi(Ä>әw)] “boil it(neut)!” Cf. fully faithful *[(Äbuˆu)]
b. /buˆ#i/: [buˆi(Ä>әj)] “boil here/there!” Cf. fully faithful *[(Äbuˆi)]

As shown in (24), the desired output (with a sad face ) cannot be selected by the
constraint hierarchy argued for in Section€4.1.

(24) Bull-ho “boil it(neut)!”: /buˆ#u/ → [buˆiÄ>ә#w]

Listed output inflectional stems: [Äbuˆi], [buˆiÄ>әj (m, w)]
/buˆ#u/ Max-φ µTrochee] Faith-OO CorrLEXI CorrLEXI-φ
a. (Äbuˆiu)
b. buˆi(Ä>әjw) *!

These cases can be accounted for by resorting to the constraint family Uniform
Exponence (Kenstowicz 1996, 1997, 2005). The fact that imperatives with the enclitics
hi and ho adopt the same inflectional stem as other pre-clitic imperatives is captured
through the more specific constraint Uniform Exponence (V+Cl), defined in (25).
(25) Uniform Exponence (V+Cl): the inflectional stem of a V+Cl sequence has
the same morphs in its various contexts of occurrence.

The table in (26) illustrates the pressure exerted by the contraint Uniform Exponence
(V+Cl) with one example of a consonant-initial clitic, /buˆ#lә/ “boil it(fem)!”, and
one example of a vocalic clitic, /buˆ#u/ “boil it(neut)!”. The constraint Uniform
Exponence (V+Cl) only demands that all forms surface with the same inflectional
stem, which makes the pairs (26a) and (26d) adequate. The constraint that decides
between these two pairs is µTrochee], violated by [(Äbuˆ)lә], which then favors (26d).
 Eulàlia Bonet & Francesc Torres-Tamarit

We have added the constraint CorrLEXI-φ to show that it must be dominated by

Uniform Exponence (V+Cl), since the grammatical forms (26d) fare worse than the
other candidates with respect to CorrLEXI-φ.
(26) [1] [2] UE (V+Cl), µTrochee] >> CorrLEXI-φ
a. [(Äbuˆ)#lә] – [(Äbuˆ#u)]  *[1] 
b. [(Äbuˆ)#lә] – [buˆi(Ä>ә#w)] * *[1] *
c. [buˆi(Ä>ә#lә)] – [(Äbuˆ#u)] *  *
 d. [buˆi(Ä>ә#lә)] – [buˆi(Ä>ә#w)]   **

All the cases we have seen up to this point contained only one clitic, but sequences of
two enclitics are also possible; these sequences also trigger stress shift, as was illus-
trated in (5b). For instance the clitic la, seen in previous examples, can be preceded by
a 1sg clitic, giving rise to the sequence me la, with stress on the first clitic [Ämәlә], due
to the constraint µTrochee]. Since the clitic cluster constitutes by itself a right-aligned
moraic trochee, one might expect the imperative to surface without the accretion, as a
bare root, for instance *[buˆ(Ämәlә)] “boil it(fem) for me!”. However, the grammatical
output surfaces with the accretion, [buˆi>ә(Ämәlә)]. Assuming Uniform Exponence,
this case can be accounted for in the same way we accounted for [buˆi(Ä>әw)]: the form
with two clitics is forced to surface with the accretion because the form with one clitic,
like [buˆiÄ>ә#lә], is bound to surface with it (without the accretion it would surface as
*[(Äbuˆ)lә]). The table in (27) is similar to the one in (26) but with the replacement of
the clitic ho with the sequence me la (and without CorrLEXI-φ).6,7
(27) [1] [2] UE (V+Cl) , µTrochee]
a. [(Äbuˆ)lә] – [buˆ(Ämәlә)]  *[1]
b. [(Äbuˆ)lә] – [buˆi>ә(Ämәlә)] * *[1]
c. [buˆi(Ä>әlә)] – [buˆ(Ämәlә)] * 
d. [buˆi(Ä>әlә)] – [buˆi>ә(Ämәlә)]  

.â•… Within Formenteran the clitics hi and ho are sometimes realized as [әj] and [әw], as men-
tioned by an anonymous reviewer. If one assumes an underlying form /j/ and /w/, the need
for UE constraints for this particular case would disappear; these clitics would behave like any
other consonantal clitic.
.╅ For ease of exposition, throughout the paper we have ignored plural clitics, which end
in a consonant, s. These clitics behave like singular clitics with respect to stress placement. For
instance the sequence [buˆiÄ>әlә] “boil it(fem)!” is realized, with the plural clitic, as [buˆiÄ>әlәs]
“boil them(fem)!”, even though the sequence [lәs] would constitute a right-aligned moraic
trochee, *[buˆi>ә(Älәs)]. Uniform Exponence with respect to stress location, ranked above
µTrochee], accounts for these facts. As noted by an anonymous reviewer, with only these two
constraints, however, a form like *[buˆiÄ>әmәlә] “boil it(fem) for me!” would win because it
shares the stress location with forms like [buˆiÄ>әlә]. This unwanted result can be ruled out by
a higher ranked constraint that requires two clitics to be parsed into a single foot, a constraint
argued for in Torres-Tamarit (2008).
 Allomorphy in pre-clitic imperatives in Formenteran Catalan 

5.â•… Problems with stratal approaches

The analysis proposed in this paper is based on the comparison between output
forms, through CorrLEX, Uniform Exponence and OO faithfulness constraints. In
stratal approaches (see the work by Paul Kiparsky, or Bermúdez-Otero in preparation)
this type of OO comparison is not possible, and therefore an alternative explanation
must be found. A possibility, suggested to us by Bermúdez-Otero (p.c.), is to posit
two “theme” allomorphs for 2sg imperatives. Moreover, following Mascaró (2007)
and Bonet, Lloret & Mascaró (2007), the allomorphs would be lexically ordered, as
indicated by the symbol “>”. The underlying representation (UR) of one example is
given in (28).

(28) perd-la “lose it(fem)!”: [pәnÄðә#lә] UR: /pεnd+{Ø > ә} #lә/

The constraint Priority, defined in (29), would penalize the choice of the dispreferred
allomorph.

(29) Priority: Respect lexical priority (ordering) of allomorphs.

Given an input containing allomorphs m1, m2, …, mn, and a candidate containing
mi in correspondence with mi, Priority assigns as many violation marks as
the depth of ordering between mi and the highest dominating morph(s).

For 2sg imperatives in isolation Priority would grant the choice of the Ø allomorph,
as shown in (30).

(30) perd “lose it(fem)!”: /pεnd/ → [Äpεnt]

/pεnd+{Ø > ә}/ µTrochee] Priority
a. (Äpεnt)
b. pәn(Äðә) *!

With enclisis, the higher ranked constraint µtrochee] would block the choice of the
preferred allomorph, and the candidate with the dispreferred allomorph would win,
as shown in (31).

(31) perd-la “lose it(fem)!”: /pεnd#lә/ → [pәnÄðә#lә]

/pεnd+{Ø > ә}#lә/ µTrochee] Priority

a. (Äpεn)lә *!
b. pәn(Äðәlә) *

A problem with this type of approach arises when trying to account for forms that have
the [:ә] or [i>ә] accretions. Recall that these accretions contain two different morphs,
one belonging to a velar extension and the other one being a theme vowel; none of
 Eulàlia Bonet & Francesc Torres-Tamarit

them can be considered an allomorph of a single morpheme. Therefore an analysis

based on allomorph ordering would not be possible for these cases.
Given that our analysis has to make reference to OO relations (an often criti-
cized mechanism), one might wonder whether the evidence for phonological condi-
tioning in the appearance of an accretion is strong enough. In an alternative analysis,
suggested to us by an anonymous reviewer, the accretion would be the result of a mor-
phological operation that would take place in the presence of enclitics. An immedi-
ate question arises: why should any additional morphology appear in the presence of
enclitics? A possible line of approach is suggested by the same anonymous reviewer:
since most forms of the imperative paradigm are taken from the Present Subjunctive,
the choice of accretion would render a second person imperative closer to the Present
Subjunctive. However, for many verbs belonging to conjugations II and III, the singu-
lar Present Subjunctive forms do not have the same morphology that is found in
enclisis. For instance, although the verb bullir “to boil” has the accretion [i>ә] with
enclitics, [buˆiÄ>ә#lә] “boil it(fem)!”, the singular forms of the Present Subjunctive
lack it ([Äbuˆis] or [Äbuˆ>әs] “you boil (PS)”, [Äbuˆi] or [Äbuˆ>ә] “s/he boils (PS)”,
cf. *[Äbuˆi>әs]). Moreover, in dialects like Central Catalan, the accretion is a schwa
(see examples like [Äbuˆәlә] “boil it(fem)!”, in (2a)), a vowel that is never found other-
wise in conjugation III verbs.

6.â•… Conclusions

In this paper we have argued that the appearance of an accretion in 2sg imperatives
followed by enclitics in Formenteran Catalan is forced by a phonological constraint
requiring a right-aligned moraic trochee. This accretion is morphological material that
is present in other forms of the imperative paradigm which is available for candidate
construction. The choice of the optimal candidate is regulated by CorrLEX constraints
and Uniform Exponence.

References

Benua, Laura. 1997. Transderivational Identity. Phonological Relations between Words. Ph.D.
dissertation, University of Massachusetts, Amherst, published as Phonological Relations
between Words. New York: Garland, 2000.
Bermúdez-Otero, Ricardo. In preparation. Stratal Optimality Theory. Oxford: Oxford University
Press.
Bonet, Eulàlia & Maria-Rosa Lloret. 2005. “More on Alignment as an Alternative to Domains:
The syllabification of Catalan clitics”. Probus 17.37–78.
Bonet, Eulàlia, Maria-Rosa Lloret & Joan Mascaró. 2007. “Allomorph Selection and Lexical
Preferences: Two case studies”. Lingua 117.903–927.
 Allomorphy in pre-clitic imperatives in Formenteran Catalan 

Hayes, Bruce. 1995. Metrical Stress Theory: Principles and Case Studies. Chicago: The University
of Chicago Press.
Kenstowicz, Michael. 1996. “Base Identity and Uniform Exponence: Alternatives to Cyclicity”.
Current Trends in Phonology: Models and Methods ed. by Jacques Durand & Bernard Laks,
vol. I, 363–393. Salford: University of Salford.
Kenstowicz, Michael. 1997. “Uniform Exponence: Exemplification and extension”. University
of Maryland Working Papers in Linguistics 5. Selected Phonology Papers from the Hopkins
Optimality Theory Workshop 1997/University of Maryland Mayfest 1997 ed. by Viola Miglio &
Bruce Morén, 139–155. University of Maryland.
Kenstowicz, Michael. 2005. “Paradigmatic Uniformity and Contrast”. Paradigms in Phonological
Theory ed. by Laura J. Downing, T. Alan Hall & Renate Raffelsiefen, 145–169. Oxford:
Oxford University Press.
Kurisu, Kazutaka. 2001. The Phonology of Morpheme Realization. Ph.D. dissertation, University
of California, Santa Cruz.
Mascaró, Joan. 1986. Morfologia. Barcelona: Enciclopèdia Catalana.
Mascaró, Joan. 2007. External Allomorphy and Lexical Representation. Linguistic Inquiry 38.
715–735.
Oltra-Massuet, Isabel. 1999. “On the Constituent Structure of Catalan Verbs”. MIT Working
Papers in Linguistics 33. Papers in Morphology and Syntax, Cycle One ed. by Karlos Arregi,
Benjamin Bruening, Cornelia Krause & Vivian Lin, 279–322. Cambridge, Mass.: MIT,
Department of Linguistics and Philosophy, MITWPL.
Pérez-Saldanya, Manuel. 1995. “Analogia i canvi morfològic: a propòsit de les formes verbals
velaritzades”. Caplletra 19. 279–305.
Steriade, Donca. 1999. “Lexical Conservatism in French Adjectival Liaison”. Formal Perspectives
in Romance Linguistics ed. by Barbara Bullock, Jean Marc Authier & Lisa Reed, 243–270.
Amsterdam & Philadelphia: John Benjamins.
Steriade, Donca. 2008. “A Pseudo-Cyclic Effect in Romanian Morphophonology”. Inflectional
Identity ed. by Asaf Bachrach & Andrew I. Nevins, 313–360. Oxford: Oxford University Press.
Torres-Tamarit, Francesc. 2008. Stress Shift in Formenteran Catalan Verb plus Enclitic(s) Sequences:
an OT Approach. MA thesis, Universitat Autònoma de Barcelona.
Viaplana, Joaquim. 1995. “Sobre la irregularitat verbal”. Caplletra 19.333–348.
Walker, Rachel. 2000. “Nasal Reduplication in Mbe Affixation”. Phonology 17.65–115.
Wheeler, Max W. 1985. “Analogia i psicologia: el desenvolupament de la morfologia verbal
balear”. Actes del XVI Congrés Internacional de Lingüística i Filologia Romàniques. Palma de
Mallorca 7–12 d’abril, 1980. Palma de Mallorca: Editorial Moll.
 Preverbal vowels in wh-questions
and declarative sentences
in Northern Italian Piacentine dialects*

Anna Cardinaletti & Lori Repetti

Università Ca’ Foscari Venezia/SUNY Stony Brook

In this paper, we discuss preverbal vocalic segments in interrogative and

declarative sentences. Their distribution depends on the type of subject (1sg, 1pl,
2pl vs 2sg, 3sg, 3pl), the presence of other preverbal clitics, and the verb tense
(present vs present perfect). We show that two different types of preverbal vocalic
segments should be differentiated: an “interrogative vowel” (in main wh-questions)
and a “subject field vowel” (in embedded wh-questions and in declarative
sentences). The two vowels realize different heads of the clausal skeleton (in the
CP and the IP layer, respectively), can be found in one and the same dialect,
and can co-occur in one and the same clause. Both types of vowels appear to be
incompatible with other preverbal clitics in some contexts, and the presence/
absence of the preverbal vowel in sentences with auxiliaries is predictable based on
the quality of the auxiliary-initial segment (consonant vs vowel).

1.â•… Introduction

In the Emilian dialect of Gazzoli (province of Piacenza), the preverbal vowel [ә] has a
very complicated distribution depending on, among other things, the type of sentence
(declarative vs interrogative), the subject (1sg, 1pl, 2pl vs 2sg, 3sg, 3pl), the presence
of other preverbal clitics, and the verb tense (present vs present perfect). For example,
the distribution of the preverbal vowel [ә] is different in wh-questions vs declarative
sentences. In wh-questions (with wh-phrases and wh-words), the preverbal vowel is
found in all six persons with the following distribution: the presence of the preverbal

*We would like to thank the audience at the LSRL and two reviewers of this volume for their
helpful comments, and our Piacentine informants for their time and patience. We use the fol-
lowing abbreviations: sg singular, pl plural, mas masculine, fem feminine, acc accusative, dat
dative, loc locative, neg negative, reflex reflexive, scl subject clitic, Z = Zörner (1989), and
M = Mandelli (1995).
 Anna Cardinaletti & Lori Repetti

vowel is preferred with a 1sg, 1pl, 2pl subject (1a), and the absence of the preverbal
vowel is preferred with a 2sg, 3sg, 3pl subject (1b). In declarative sentences, the prever-
bal schwa is optional with 1sg, 1pl, 2pl verbs (2a),1 and it is ungrammatical with 2sg,
3sg, 3pl verbs (2b).2 (We use the following symbols: ?(ә) = preference for the presence
of /ә/, (?ә) = preference for the absence of /ә/, (ә) = optional /ә/.)
(1) wh-questions
a. 1sg, 1pl, 2pl
kõ ki ?(ә) Ävo-jә via “with whom am I going away?”
kõ ki ? (ә) Änum-jә via “with whom are we going away?”
kõ ki ?(ә) Änε:-v via “with whom are you:pl going away?”
b. 2sg, 3sg, 3pl
kõ ki (?ә) Ävε-t via “with whom are you:sg going away?”
kõ ki (?ә) Äva-l via “with whom is he going away?”
kõ ki (?ә) Ävan-jә via “with whom are they going away?”
(2) declarative sentences
a. 1sg, 1pl, 2pl
(ә) Ävo via “I go away”
(ә) Änum via “we go away”
(ә) Änε via “you:pl go away”
b. 2sg, 3sg, 3pl
(*ә) t әskri:v “you:sg write”
(*ә) õ Äva via “he goes away”
(*ә) i Ävan via “they go away”

The 2sg form of the declarative sentence in (2b) displays a different verb with respect to
the rest of the paradigm: [t әskri:v] “you:sg write”.3 We use this verb to clearly show that
the preverbal vowel is impossible: *[ә t әskri:v]. In careful speech [әt әskri:v] is accept-
able with a slight pause between the subject clitic [әt] and the verb [әskri:v]; Â�however,

.╅ Although the 2pl is possible without the preverbal schwa, there is a strong preference for
the presence of the preverbal vowel.
.╅ A note on the semantics of the forms with and without the preverbal vowel: for other
northern Italian dialects, such as Paduan, it has been suggested that the presence of the pre-
verbal vowel denotes new information (Benincà 1983:€28). However, Zörner (1989: 150) notes
(and we concur) that this does not appear accurate for the Piacentine dialects.
.â•… In [t әskri:v], an epenthetic vowel (/ә/) is needed between the subject clitic /t/ and the
initial /s/ + consonant cluster of the verb in order to syllabify these consonants. This follows
the general rules of epenthesis in this dialect: an epenthetic vowel is inserted before an unsyl-
labified consonant or between two unsyllabified consonants (see Cardinaletti & Repetti 2004,
2008a for further discussion).
 Preverbal vowels in wh-questions and declarative sentences in Northern Italian 

the schwa before the /t/ subject clitic in forms such as [әt әskri:v] and [әt Ävε via]
“you:sg go away” is epenthetic and not the preverbal vowel discussed above. This Â�follows
the general rules of epenthesis in this dialect. See Cardinaletti and Repetti (2004),
(2008a) for further discussion.
For the 3sg and 3pl forms in (2b) [õ Äva via] and [i Ävan via], as well as the 3sg femi-
nine form [a Äva via] Äshe goes away”, the preverbal vowels /õ/, /i/, /a/ are true subject
clitic pronouns and do not enter the typology of functional vowels discussed in this
paper. These forms are impossible with the preverbal schwa: *[ә õ Äva via], *[ә i Ävan via],
*[ә a Äva via]. In nearby dialects, such as Donceto, the 3sg masculine form – [әl Äva via]
“he goes away” – contains an initial schwa that is epenthetic and is needed to syllabify
the 3sg masculine subject clitic /l/; it is not the functional vowel discussed in this paper.
See Cardinaletti and Repetti (2004), (2008a) for further discussion.
In this paper we will discuss the nature of the preverbal schwa in the above data,
and its distribution in various contexts. We will show that the preverbal vowels are
two different syntactic entities: an “interrogative vowel” (1) and a “subject field vowel”
(1a)–(2a). We will further show that the realization of the preverbal vowel(s) depends
on other considerations, such as phonological considerations and the presence of
other clitics.
This paper is organized as follows. In §2 we present our analysis of the preverbal
schwa as two different functional vowels, and in §3 we provide data from closely related
dialects to support our analysis of the Gazzoli facts. We then discuss the occurrence of
preverbal schwa with object clitics (§4) and with auxiliary verbs (§5) to illustrate how
various components of the grammatical system are involved in the realization of the
preverbal schwa. We conclude the paper in §6.

2.â•… Analysis of the preverbal schwa

In this section, we suggest that the preverbal vowel occurring in declarative sentences
and the preverbal vowel occurring in interrogative sentences are not one and the same
vowel, and that they realize different functional heads in the clausal skeleton. This
proposal is supported by the fact that the two vowels occur in different persons of
the paradigm in different sentence types (as seen in (1)–(2)), by their distribution in
embedded questions (§2.3), and by the fact that the two vowels can cooccur in main
wh-questions (§2.5).

2.1â•… Main wh-questions

In main wh-questions (1), the preverbal vowel is an “interrogative vowel” which is
merged in the Focus head of the CP layer (in Rizzi’s 1997 sense). This explains why it is
 Anna Cardinaletti & Lori Repetti

found in all six persons of the verbal paradigm. The preverbal vowel is the spell out of
the complex Q + Foc head found in main questions, following Rizzi’s (2006) analysis
of Italian. The interrogative head has an edge feature which attracts the wh-phrase. We
exemplify the derivation with the 2sg form.
(3) [FocP kõ ki ә [QP ә [YP vε-t [TP t vε … [VP t vε via kõ ki]]]]]

In interrogative structures, verb – subject clitic inversion is obtained by moving the verb
to Y across the subject clitic (see Cardinaletti & Repetti 2008a, 2010, for discussion).
We take the Y head to be located in the INFL layer. This is coherent with the wide-
spread proposal that no V-to-C movement takes place in Romance languages. See
Kayne (1994: 44, 139, n.15) and Sportiche (1999) for French, Suñer (1994) for Spanish,
Guasti (1996) and Cardinaletti (2007) for Italian, Munaro (1999) and Cardinaletti
and Repetti (2008a), (2010) for northern Italian dialects. V-to-Y movement is moti-
vated by the need to check the inflectional [wh] feature on the verb (Rizzi 1996, 2001)
against the Y head.
In §2.6, we will address the issue of preference/dispreference of the preverbal
vowel in the different forms of the verb paradigm.

2.2â•… Declarative sentences

In declarative sentences, the (optional) preverbal vowel is only found in some persons
of the paradigm (2a). We suggest that it is merged in a lower functional head (in (4) we
call the head Z) belonging to the IP layer (in Rizzi’s 1997 sense) that hosts the features
of 1sg, 1pl and 2pl, namely of the persons that do not have a subject clitic (there is no
evidence that the vowel has a different distribution in these three persons). We call this
vowel a “subject field vowel”, and in (4) we show the derivation for the 1sg.4
(4) [ZP (ә) [TP pro vo … [VP pro vo via]]]

In §2.6, we will address the issue of optionality vs impossibility of the preverbal vowel
in the different forms of the verb.

2.3â•… Embedded wh-questions

The patterning of schwa in embedded questions reflects the different behavior of the
preverbal vowel in main clauses. In both contexts, the preverbal vowel follows two pat-
terns: it is found in all persons of the paradigm, or it is found with 1sg, 1pl, 2pl forms. In
embedded questions, only the subject-field vowel is possible (5a), while the �interrogative

.╅ For some evidence that the subject field vowel is higher than subject clitics, see Cardinaletti
and Repetti (2004).
 Preverbal vowels in wh-questions and declarative sentences in Northern Italian 

vowel is ungrammatical (5b) (see Footnote€ 4). The vowel occurring in (5c) is an�
epenthetic vowel inserted to syllabify the 2sg subject clitic /t/ (see Section€1).
(5) a. õ lә sa mia õ:d (ә) vo “he doesn’t know where I am going”
b. õ lә sa mia ko:z (*ә) t әskri:v “he doesn’t know what she is writing”
c. õ lә sa mia õ:d әt vε “he doesn’t know where you:sg are going”

The occurrence of preverbal vowels in embedded wh-questions supports our proposal:

the CP head realized by the interrogative vowel is not available in embedded clauses,
while the IP-internal subject-field vowel found in main declarative clauses (2a) can appear
in interrogative embedded clauses, as well as in declarative embedded clauses (6):
(6) õ lә sa mia ke (ә) vo via “he doesn’t know that I am going away”

2.4â•… Functional vowels

The two types of vowels discussed above – both the “interrogative vowel” and the
“subject field vowel” – are referred to here as “functional vowels” since they realize
functional heads of the clausal skeleton. In the chart in (7), we show the contexts in
which each of the functional vowels can (optionally) appear. The “interrogative vowel”
is possible in main wh-questions in all persons, while the “subject field vowel” is pos-
sible in embedded wh-questions and declarative sentences, but only in the 1sg, 1pl, 2pl
forms. Nothing prevents the “subject-field vowel” from occurring in wh-questions as
well. Below we show that this is indeed the case.

(7) Distribution of 1sg, 1pl, 2pl 2sg, 3sg, 3pl

the preverbal vowels…
• interrog /ә/ • subj field /ә/ • interrog /ә/ • subj field /ә/
in main wh-questions: yes yes yes no
in embedded wh-Qs: no yes no no
in declarative sent.: no yes no no

Similar distributional patterns of what we call “functional vowels” are found in other
Northern Italian dialects and have been previously discussed, e.g. by Chinellato
(2004), Goria (2004) and Poletto (2000). However, many properties of the paradigms
in (1)–(2) have not been discussed. For example, previous analyses have not examined
cases in which a preverbal vowel found in all persons of the paradigm (what have been
called “invariable subject clitics” by Poletto 2000) is found in wh-questions. In fact,
cases like (1) have been claimed not to exist.5

.╅ For a thorough discussion of previous literature, see Cardinaletti and Repetti (2008b).
In that paper, we also analyze the distribution of functional vowels in main and embedded
yes-no questions.
 Anna Cardinaletti & Lori Repetti

Our data suggest that “functional vowels” can be found in both the CP layer
(“interrogative vowel”) and the IP layer (“subject field vowel”), in one and the same
dialect (for example, Gazzoli). Hence, there is no reason why they should not be able
to co-occur in the same phrase, and this is indeed what is found, as discussed in the
following section.

2.5â•… C
 o-occurrence of the “interrogative vowel”
and the “subject field vowel”
The preverbal schwa in the 1sg, 1pl, and 2pl can be pronounced as a long vowel in
main wh-questions, a fact which can be understood as follows: a long vowel is the
simultaneous realization of the “interrogative vowel” and the “subject field vowel”.6
As expected, no long preverbal vowel is found in the 2sg, 3sg, 3pl forms of main wh-
questions since there is only an “interrogative vowel” but not a “subject field vowel”
possible. Furthermore, no long preverbal vowel is found in embedded wh-questions
and declarative sentences since the “interrogative vowel” is not available.

(8) Distribution of the 1sg, 1pl, 2pl 2sg, 3sg, 3pl

preverbal vowels…
…in main wh-questions:
…in embedded wh-Qs and
declarative sentences:
• interrogative vowel • interrogative vowel
• subject field vowel
• subject field vowel • neither possible

2.6â•… On the distribution of the “interrogative vowel”

We have noted that in wh-questions (1), the preverbal vowel is preferred in the 1sg,
1pl, 2pl forms (and perhaps it is obligatory in the 2pl form, see Footnote€2), and it is
dispreferred in the 2sg, 3sg, 3pl forms. Why would this be? We can now propose an
answer to this question. In 1sg, 1pl, 2pl wh-questions, the preverbal schwa can be
either the interrogative vowel or the subject field vowel. Since both functional vowels
are optional, there is a higher possibility that (at least) one will be used. Alternatively,
in 2sg, 3sg, 3pl wh-questions the preverbal schwa can only be the (optional) interroga-
tive vowel. This may explain why the former are more commonly realized with the
preverbal vowel, than the latter.

.╅ Although no phonetic measurements have been done on vowel duration in Piacentine
dialects, vowel length distinctions are clearly audible.
 Preverbal vowels in wh-questions and declarative sentences in Northern Italian 

3.â•… Cross-linguistic data

The patterns observed above for the Gazzoli dialect are also found in other Piacentine
dialects.7 Wh-questions are reported both with and without the preverbal vowel: for
the dialect of Bobbio, Mandelli (1995) reports wh-questions with the interrogative
vowel [a] (underlined a in the following data), while for the dialect of Travo, Zörner
(1989) reports forms without the interrogative vowel ([a]).
(9) Bobbio: 2sg [ke dŠurnέl a ledŠ-at] “what newspaper did you:sg read?” (M 30)
what newspapers a read-you
3sg [kwant a kusta-l] “how much does it cost?” (M 30)
how much a cost-it
Travo: 2sg [kus fε-t] “what are you:sg doing?” (Z143)
what do-you:sg
[kus vø-t] “what do you want?” (Z 143)
what want-you:sg
[duv vε-t] “where do you:sg go?” (Z 144)
where go-you:sg
[parkέ ve\-at maj] “why don’t you:sg ever come?” (Z 145)
why come-you:sg never
3sg [kus vø-l] “what does he want?” (Z 143)
what want-he
[kwãd \irá-l] “when will he arrive?” (Z 144)
when will arrive-he
3pl [da duv ve\an-ja] “from where do they come?” (Z 144)
from where come-they
2pl [iÌ… kwãt si-v] ‘(in) how many are you:pl’ (Z 144)
in how many are-you:pl

In some cases, the status of a preverbal [a] in wh-questions is not clear. In the Piacen-
tine dialects, many wh-words can be realized as monosyllabic (for example, [koz]/
[dõd] “what/where”) or bisyllabic ([koza]/[dõda]). Given these latter forms, it is dif-
ficult to analyze the role of a preverbal [a] in the following sentences from Bobbio
(Mandelli 1995) and Groppallo (Zörner 1989).

.╅ In the Gazzoli dialect the preverbal functional vowel and the epenthetic vowel are both
[ә], while in the other Piacentine dialects discussed here (Bobbio, Groppallo, Travo) the func-
tional vowel and the epenthetic vowel are both [a].
 Anna Cardinaletti & Lori Repetti

(10) [koz]/[dõd] + /a/ [koza]/[dõda]

without /a/
Bobbio: [koz a vε\-at ke da fε] [koza vε\-at ke da fε] “what do you come
what a come-you here what come-you here here to do?’”(M 30)
to do to do
Groppallo: [dõd a vε-t] [dõda vε-t] “where are you:sg
where a go-you:sg where go-you:sg going?” (Z 171)
[dõd a vum-ja] [dõda vum-ja] “where are we
where a go-we where go-we going?” (Z 175)

Is the underlined /a/ in the data in (10) the preverbal vowel discussed above, or is it
part of the wh-word? There are three possible ways to analyze these data.
These forms could be interpreted as [koz]/[dõd] followed by a preverbal /a/. There is
some evidence in support of this approach. In the Groppallo cases in (10), Zörner (1989)
reports the wh-word as monosyllabic, and in the closely related dialect of Travo, [kus]
and [duv] can only be interpreted as monosyllabic (9). The problem with this analysis is
that in Gazzoli (1), the preverbal vowel is dispreferred with 2sg, 3sg, 3pl wh-questions;
however, the forms in (10) would have to be analyzed as including the preverbal vowel.
Alternatively, the forms in (10) could be interpreted as [koza]/[dõda] without a
preverbal /a/. In fact, the Bobbio form is written by Mandelli (1995) as a bisyllabic
word. The problem with this approach is that in Gazzoli (1), 1sg, 1pl, 2pl wh-questions
are preferred with a preverbal vowel, but the 1pl question from Groppallo would have
to be analyzed as not having the preverbal vowel.
A third possibility is that the forms are [koza a] and [dõda a], consisting of a bisyl-
labic wh-word [koza]/[dõda] plus an interrogative /a/; however, the two /a/’s are not
represented as long. Evidence in support of this analysis is that neither of these two
authors record long vowels, although these dialects clearly have vowel length distinc-
tions. So, if this vowel is indeed long, we would not expect to see it recorded as such.
The analysis of the forms in (10) is possibly a combination of all three options dis-
cussed above. However, given the limited amount of available data, a more definitive
analysis is not possible at this point.

4.â•… On the occurrence of functional preverbal vowels with object clitics

One interesting observation about the Piacentine dialects is that in wh-questions and
declarative sentences the presence of the preverbal functional vowel seems incompat-
ible with certain other preverbal clitics.
The functional vowels are fully compatible with post-verbal subject clitics (as
seen in (1) and (2)), but not with preverbal subject clitics. We would not expect the
 Preverbal vowels in wh-questions and declarative sentences in Northern Italian 

�
“interrogative vowel” with preverbal subject clitics since subject clitics are post-Â�verbal
in interrogatives. And the “subject field vowel” is in complementary distribution with
preverbal subject clitics: it is found with 1sg, 1pl, 2pl forms which do not have a prever-
bal subject clitic, and it is not found with 2sg, 3sg, 3pl forms which do have a preverbal
subject clitic. However, the functional vowels are absent with other types of prever-
bal clitics. Why is the realization of the functional vowel restricted in the presence of
another clitic? Is there a syntactic restriction against this combination or a phonologi-
cal constraint banning these clusters? In §4.1–4.2 we look into the cooccurrence of the
functional vowel with other consonant-initial and vowel-initial clitics. In §4.3 we inves-
tigate whether the restriction can be phonological, and we conclude that it is not.

4.1â•… Functional vowel + consonant-initial clitic

In phrases involving a (potential) functional vowel plus a consonant-initial clitic, the
status of the vowel preceding the consonantal clitic is not clear: is this vowel the func-
tional vowel, or is it an epenthetic vowel needed to syllabify the consonantal clitic? The
quality of the functional vowel and epenthetic vowel is identical: in Groppallo both are
/a/ and in Gazzoli /ә/.
In (11), the initial vowel can be analyzed as the subject-field vowel optionally pres-
ent in the 1sg in a declarative sentence, or it can be analyzed as an epenthetic vowel
needed to syllabify the following consonantal clitic.

(11) Groppallo: /a m rikfrd/ “I remember” (Z 274)

a reflex I-remember
/a g so sto/ “I was there” (Z 276)
a loc I-am been
Gazzoli: /ә t prumә́t/ “I promise you:sg”
ә dat I-promise

The problem was addressed by Vanelli (1984) in response to a series of articles on

epenthetic vowels and clitic pronouns in Romance languages. Analyzing the quality of
the vowel, Vanelli (1984) concludes that in some cases the vowel preceding the conso-
nantal clitic is a vocalic subject clitic pronoun, while in other cases it is an epenthetic
vowel. Since in (11), the quality of the two is the same (i.e. the quality of the functional
vowels and the epenthetic vowel is identical), we cannot determine what the nature of
the vowel is in these examples.

4.2â•… Functional vowel + vowel-initial clitic

We will now look into the possibility of the co-occurrence of the functional vowel(s)
and a vocalic clitic. As seen in the data below, the functional vowel is not realized in
the presence of another vowel-initial clitic (mas sg acc [õ] Gazzoli, [u] Groppallo, and
 Anna Cardinaletti & Lori Repetti

mas/fem pl acc [i] Gazzoli and Groppallo).8 Groppallo and Gazzoli both have the same
distribution of preverbal functional vowels illustrated in (1)–(2). In (12)–(15) we pro-
vide data showing the lack of a preverbal functional vowel when there is a preverbal
vowel-initial accusative clitic. The restriction holds for both the subject field vowel in
(12)–(13) and the interrogative vowel in (14)–(15):
(12) Gazzoli: /õ manŠ/ (*[ә õ manŠ])
acc I eat “I eat it”
/i manŠ/ (*[ә i ÄmanŠ])
acc I eat “I eat them”
(13) Groppallo: /u fávam nœŋ/ “we did it” (Z 276)
acc we did we
(14) Gazzoli: /kwand i vәd- jә/ “when do I see them?”
when acc I see I
/a ki õ do- jә/ “to whom do I give it?”
to whom acc I give I
/kwand i vәd- әt/ “when do you see them?”
when acc you see you
/a ki õ dε- t/ “to whom do you give it?”
to whom acc you give you
(15) Groppallo: /dõd i mә́tam- ja/ “where do we put them?” (Z 175)
where acc we put- we

In conclusion, we have not found evidence of the co-occurrence of the functional

vowel with other vowel-initial clitics.

4.3â•… 3sg feminine subject clitic + another clitic

Perhaps the restriction on the cooccurrence of the functional vowel with another clitic
is phonological. In order to test this hypothesis, we can see if there are restrictions on
the cooccurrence of the obligatory 3sg feminine subject clitic /a/ (Gazzoli and Grop-
pallo) with other clitics. We find that the obligatory 3sg feminine subject clitic /a/ can
co-occur with some consonantal clitics. In the examples in (16), the initial vowel can
only be analyzed as the mandatory 3sg fem subject clitic, and not as an epenthetic
vowel (which has the same quality in Groppallo).
(16) Groppallo: /a g fáva la sýpa/ “she made them the soup” (Z 272)
scl dat made the soup
Gazzoli: /a s láva/ “she gets washed”
scl reflex washes

.╅ All vowel-initial object clitics are accusative; there is no vowel-initial dative clitic.
 Preverbal vowels in wh-questions and declarative sentences in Northern Italian 

We can now see if there are restrictions on the cooccurrence of the scl /a/ with other
vocalic clitics (mas sg acc /õ/ (Gazzoli), /u/ (Groppallo), and gender-neutral pl acc
/i/ (Gazzoli and Groppallo)). Since the mas sg acc clitics are realized with a different
allomorph (/l/) in the presence of a third person subject (see Cardinaletti and Repetti
ms), the only vowel initial clitic we can investigate is 3pl acc /i/. The situation is further
complicated by the fact that there are restrictions on the syllabification of a third per-
son scl with a third person acc clitic when they cooccur: the two cannot form a single
syllable (see Cardinaletti and Repetti ms). Therefore, we expect that the 3sg feminine
subject clitic /a/ cannot syllabify with 3pl accusative clitic /i/. In fact, the 3pl acc /i/ is
syllabified by the insertion of a following epenthetic vowel, resulting in the syllable [ja]
(Groppallo) or [jә] (Gazzoli).9
(17) Groppallo: /a ja tája/ “she cuts them (fem)” (Z 174)
scl acc cuts
Gazzoli: /a jә fa le/ “she (emphatic) makes them (mas/fem)”
scl acc makes she

Since the mandatory 3sg feminine subject clitic /a/ can co-occur with another
vocalic clitic (although they cannot be syllabified together), there is no phonologi-
cal reason for the restriction on the occurrence of the functional vowels with other
vocalic clitics.
In conclusion, we have no clear evidence of the co-occurrence of the functional
vowels discussed in this paper with another clitic pronoun. There is no phonological
reason for this ban since obligatory subject clitic /a/, which is homophonous to the
functional vowels in some Piacentine dialects, can co-occur with other clitics. Nor
does there seem to be a syntactic reason for this ban: functional vowels and especially
the interrogative vowel realize functional heads which do not interfere with object
clitic placement. Instead, we propose that this ban is due to an economy principle. As
we have seen, the co-occurrence of vocalic subject and object clitics implies the use
of different allomorphs for object clitics, or the application of marked syllabification
strategies. These marked options can be avoided in the case of a functional vowel: since
the functional vowel is optional, it is not used in the presence of other clitics. Things
are different with a true subject clitic, which must be obligatorily realized. We hope to
develop this point more fully in future research.

.╅ If the 3pl acc clitic /i/ can syllabify with the following vowel-initial verb, it does.

Groppallo [a j a fat køz] “she cooked them (mas)” (Z 274)

scl acc has made cook

The acc clitic /j/ syllabifies as the onset of the following verb.
 Anna Cardinaletti & Lori Repetti

5.â•… On the occurrence of the preverbal vowel with auxiliaries

In some dialects, such as the Friulian dialect of San Michele al Tagliamento, preverbal
vowels (similar to the functional vowels investigated in this paper) and auxiliaries can
co-occur.

(18) Friulian: San Michele al Tagliamento (Poletto 2000)

a. Quantis caramelis *(i) a-tu mangiat? (p. 25, 60)
how many sweets i have-you:sg eaten?
b. Coma (i) a-tu fat il compit? (p. 60)
how (i) have-you:sg done the task?

Renzi and Vanelli (1983: 129) and Poletto (2000: 183,n.19) have claimed that in Emilian
dialects, preverbal vowels are not possible with auxiliaries. This claim is correct for the
Gazzoli cases like (19) and (20b), but it is not generally true, as shown by the gram-
maticality of (20a). The data are organized so that the forms in (19) contain the “have”
auxiliary, and the forms in (20) contain the “be” auxiliary, and the forms in (a) are 1sg,
1pl, 2pl, and the forms in (b) are 2sg, 3sg, 3pl.10

(19) “have” auxiliary

a. (*ә) Äo buÄvi:d kwã:d (*ә) Äo-jә buÄvi:d? “I have drunk”/“when …?”
(*ә) Äum buÄvi:d kwã:d (*ә) Äum-jә buÄvi:d? “we have drunk”/“when …?”
(*ә) Äi buÄvi:d kwã:d (*ә) Äi:-v buÄvi:d? “you:pl have drunk”/
“when …?’
b. (*ә) t Äe buÄvi:d kwã:d (*ә) Äe–t buÄvi:d? “you:sg have drunk”/
“when …?”
(*ә) l Äa buÄvi:d kwã:d (*ә) Äa–l buÄvi:d? “he has drunk”/’when …?”
(*ә) j Äan buÄvi:d kwã:d (*ә) Äan-jә buÄvi:d? “they have drunk’/’when …?’

(20) “be” auxiliary

a. (ә) Äso na via kwã:d (ә) Äso-jә na via? “I am gone away”/“when …?”
(ә) Äsum ana via kwã:d (ә) Äsum-jә na via? “we are gone away”/
“when …?”
(ә) Äsi na via kwã:d (ә) Äsi:-v na via? “you:pl are gone away”/
“when …?”

.╅ In the declarative forms in (19b) and (20b), which are included for completeness, the
preverbal vowel is ungrammatical, as with simple verbs (see Section€1), and confirms the
observation that no subject-field vowel is found in the 2sg, 3sg, 3pl. Notice also that the
epenthetic vowel is not found; this is because the subject clitic syllabifies with the vowel-
initial auxiliary.
 Preverbal vowels in wh-questions and declarative sentences in Northern Italian 

b. (*ә) t Äe na via kwã:d (*ә) Äe-t na via? “you:sg are gone away”/
“when …?”
(*ә) l Äε na via kwã:d (*ә) Ä ε-l na via? “he is gone away”/“when …?”
(*ә) j Äen na via kwã:d (*ә) Äen-jә na via? “they are gone away”/
“when …?”

Two facts are surprising in these paradigms. First, consider the contrast between (19a)
and (20a) in both declarative and interrogative sentences: in the 1sg, 1pl, 2pl, the func-
tional vowels are possible with the “be” auxiliary but are not possible with the ‘have’
auxiliary. Second, the interrogative forms in (19b) and (20b) are, surprisingly, ungram-
matical if compared with the optionality of [ә] in (1b).
We cannot provide a syntactic account of these seemingly complicated patterns.
As can be seen in the above data, the choice of the auxiliary (“have” or “be”) does not
affect whether or not the preverbal vowel is used: it is never found with the “have”
auxiliary, and it is only found with the 1sg, 1pl, 2pl “be” auxiliary forms. In addition,
the distribution of the preverbal vowel is identical in declarative sentences and wh-
questions in (19)–(20), where we claim two different vowels are involved (“subject-
field vowel” and “interrogative vowel”).
We propose that the insight behind this seemingly complicated pattern is not syn-
tactic, but phonological in nature. The ungrammaticality of (19) and (20b), is due to a
phonological restriction, and specifically a constraint against schwa + stressed vowel:
*[ә + ÄV]. Notice that the “be” forms in (20a), which allow the preverbal schwa, are
consonant-initial ([so], [sum], [si]), while the “have” forms in (19) and the “be” forms
in (20b), which do not allow the preverbal schwa, are vowel-initial.11
Independent evidence of the constraint banning schwa + stressed vowel: *[ә + ÄV]
is found. In normal speech an unstressed word-final schwa is deleted when followed
by a word-initial stressed vowel.12
(21) Gazzoli: /õ vedә ána/ > [õ ved- ána] “he sees Anna”
/әm pja:zә aj/ > [әm pja:z- aj] “I like Ai (town name)”

We predict that the same pattern would hold for lexical verbs beginning with a stressed
vowel, namely that the preverbal vowel is not possible with vowel-initial verbs. We
cannot test this pattern with vowel-initial verbs (other than auxiliaries or the copula)

.╅ The consonant-initial and the vowel-initial forms coincide with the person split which
pervasively shows up in these dialects: 1sg, 1pl, 2pl vs. 2sg, 3sg, 3pl. This seems to be a lexical
accident. The Gazzoli paradigm of “be” should be compared with the Italian paradigm of the
same verb, which only has one vowel-initial form in the 3sg (è “he/she is”).
.╅ Although the schwa in (21) is in a different prosodic context with respect to the func-
tional vowel (word-final vs proclitic), certain important aspects of the two contexts are the
same, including the fact that the schwa is before a stressed vowel.
 Anna Cardinaletti & Lori Repetti

because, to the best of our knowledge, no such verbs exist in the Gazzoli dialect. Com-
mon Romance vowel-initial verbs are consonant-initial in the dialect of Gazzoli: It.
amare = /vu’le bεŋ/ “to love”, It. entrare = /vni dεntr/ “to enter”, It. uscire = /na føra/ “to
exit”, It. incontrare = /tru’va/ “to meet”, It. odorare, annusare = /na’za/ “to smell”, etc.
Other Piacentine dialects appear to have the same restriction.13 In Groppallo and
Travo (Zörner 1989), none of the sentences with vowel-initial auxiliaries are recorded
with the preverbal vowel (22); however, sentences with consonant-initial auxiliaries do
allow for the optional subject-field vowel (23).
(22) Groppallo: [m e–l estó u ta vjaj] “how was your trip?” (Z 171)
[m e-la stá a partída] “how was the game?” (Z 171)
[kwãd ε-t fat kula vjaj ke] “when did you take this trip?” (Z 171)
[kwãd e–l nasýd] “when was he born?” (Z 171)
Travo: [kwãd e–l nasí] “when was he born?” (Z 144)
[m e–l andέ la partída] “how did the game go?” (Z 145)
(23) Groppallo: [a sum \yd da naskús] “we came hidden” (Z 172)
[a sum astέ] “we were” (Z 276)
[so sto] “I was” (Z 276)

6.â•… Conclusions

In sum, we have shown that the preverbal vocalic segments in the data in (1)–(2) (what
we call here “functional vowels”) are of two different types: “interrogative vowels” (1) and
“subject field vowels” (1a)–(2a). Our data suggest that “functional vowels” can be found:

–â•fi in both the CP layer (“interrogative vowel”) and the IP layer (“subject field vowel”),
–â•fi in one and the same dialect (for example, Gazzoli),
–â•fi in one and the same phrase (as shown by the occurrence of long functional vowels
in the 1sg, 1pl, 2pl in main wh-questions).

Both types of vowels appear to be incompatible with other preverbal clitics in

some contexts, and the presence/absence of the preverbal vowel in sentences with
auxiliaries is predictable based on the quality of the auxiliary-initial segment
�(consonant vs vowel).

.â•… Zörner (1989) provides data involving subject-field /a/ + an unstressed vowel (although,
like Gazzoli, there are no data with a verb beginning with a stressed vowel): Groppallo: [(a) ãdúm]
“we go”, [(a) ãdέma] “we were going”, etc. (Z 298); Travo: [(a) ãdr↜f́] “I will go”, [(a) ãdrís]
“I would go”, etc. (Z 299).
 Preverbal vowels in wh-questions and declarative sentences in Northern Italian 

References

Benincà, Paola. 1983. “Il clitico a nel dialetto padovano”. Scritti linguistici in onore di Giovan
Battista Pellegrini ed. by Paola Benincà et al. 25–35. Pisa: Pacini.
Cardinaletti, Anna. 2007. “Subjects and Wh-questions. Some new generalizations”. Romance
Linguistics 2006: Selected Papers from the 36th Linguistic Symposium on Romance Languages
(LSRL), ed. by Juan Camacho, Nydia Flores Ferrán, Liliana Sánchez, Viviane Déprez, María
José Cabrera, 57–79. Amsterdam: Benjamins.
Cardinaletti, Anna & Lori Repetti. 2004. “Clitics in Northern Italian Dialects: Phonology, Syntax
and Microvariation”. University of Venice Working Papers in Linguistics 14.7–106.
Cardinaletti, Anna & Lori Repetti. 2008a. “The Phonology and Syntax of Preverbal and Postver-
bal Subject Clitics in Northern Italian Dialects”. Linguistic Inquiry 39.523–563.
Cardinaletti, Anna & Lori Repetti. 2008b. “Functional Vowels in Main Questions in Northern
Italian Dialects”. Paper presented at Going Romance, Groningen, December 2008.
Cardinaletti, Anna & Lori Repetti. 2010. “Proclitic vs Enclitic Pronouns in Northern Italian
Dialects and the Null-Subject Parameter”. Syntactic Variation. The Dialects of Italy, ed. by
Roberta D’Alessandro, Adam Ledgeway & Ian Roberts, 119–134. Cambridge: Cambridge
University Press.
Cardinaletti, Anna & Lori Repetti. Ms. “Clitic Clusters in Northern Italian Dialects”.
Chinellato, Paolo. 2004. “La microvariazione del clitico A in alcune varietà venete”. Italia Dial-
ettale LXV. 43–72.
Goria, Cecilia. 2004. Subject Clitics in the Northern Italian Dialects. A comparative study based on
the Minimalist Program and Optimality Theory. Dordrecht: Kluwer.
Guasti, Maria Teresa. 1996. “On the Controversial Status of Romance Interrogatives”. Probus
8.161–180.
Kayne, Richard S. 1994. The Antisymmetry of Syntax. Cambridge, Mass: MIT Press.
Mandelli, Enrico. 1995. Il dialetto bobbiese. Bobbio: Tipografia Columba.
Munaro, Nicola 1999. Sintagmi interrogativi nei dialetti italiani settentrionali. Padova: Unipress.
Poletto, Cecilia. 2000. The Higher Functional Field: Evidence from Northern Italian dialects.
Oxford: Oxford University Press.
Renzi, Lorenzo & Laura Vanelli. 1983. “I pronomi soggetto in alcune varietà romanze”. Scritti lin-
guistici in onore di Giovan Battista Pellegrini ed. by Paola Benincà et al., 121–145. Pisa: Pacini.
Rizzi, Luigi. 1996. “Residual Verb-second and the Wh-criterion”. Parameters and Functional
Heads. Essays in Comparative Syntax, ed. by Adriana Belletti & Luigi Rizzi, 63–90. New
York & Oxford: Oxford University Press.
Rizzi, Luigi. 1997. “The Fine Structure of the Left Periphery”. Elements of Grammar, ed. by
Â�Liliane Haegeman, 281–337. Dordrecht: Kluwer.
Rizzi, Luigi. 2001. “On the Position “Int(errogative)” in the Left Periphery of the Clause”. Current
Studies in Italian Syntax, ed. by Guglielmo Cinque & Giampaolo Salvi, 287–296. Amster-
dam: Elsevier.
Rizzi, Luigi. 2006. “On the Form of Chains: Criterial positions and ECP effects”. Wh-Movement.
Moving On, ed. by Lisa Lai-Shen Cheng & Norbert Corver, 97–133. Cambridge, Mass: MIT
Press.
Sportiche, Dominique. 1999. “Subject Clitics in French and Romance: Complex inversion
and clitic doubling”. Studies in Comparative Syntax, ed. by Kyle Johnson & Ian Roberts,
189–221. Dordrecht: Kluwer.
 Anna Cardinaletti & Lori Repetti

Suñer, Margarita. 1994. “Verb-Movement and the Licensing of Argumental Wh-phrases in

Spanish”. Natural Language and Linguistic Theory 12. 335–372.
Vanelli, Laura. 1984. “Pronomi e fenomeni di prostesi vocalica nei dialetti italiani settentrionali”.
Revue de linguistique romane 48. 281–295.
Zörner, Lotte. 1989. Die Dialekte von Travo und Groppallo.Wien: Der Österreichischen Akad-
emie der Wissenschaften.
 Pitch accent, focus, and the interpretation
of non-wh exclamatives in French*

Heather Burnett
University of California Los Angeles

This paper examines the cross-linguistic realization of the class of exclamatives in

the Romance languages. I argue that, while the syntactic and semantic properties
of exclamative sentences are usually viewed as being licensed by wh-morphology,
other grammatical features such as f(ocus) marking may serve a similar purpose
in the construction of exclamative meaning. In particular, I argue that
exclamations with focused gradation quantifiers, such as the Québec French
sentence J’ai vu un film ASSEZ bon! “I saw SUCH a good movie!” display many
of the same semantic properties that have been observed with wh-exclamatives.
I propose that the semantic content of this new type of exclamative is a gradation
construction with an implicit threshold clause and focus on the quantifier.
I propose that the exclamative operator binds the threshold yielding an extreme
degree reading, and that the presence of this operator is licensed by the focus
structure of the sentence.

1.â•… Introduction

This paper examines the cross-linguistic realization of the class of exclamative sentences,
with a special focus on French and other Romance languages. It is well known that,
although almost all utterances can be used in the act of exclaiming, certain utterances are
syntactically and semantically tied to this use. For example, if I am surprised at the fact
that you got your hair cut, I can express this by using the relevant assertion, the French
sentence in (1a), with a particular exclamatory intonation (1b).

*I would like to thank David-Étienne Bouchard, Daniel Büring, Jessica Rett, Dominique
Sportiche, Daniel Valois, and the audience at LSRL39 for their comments and suggestions.
I would also like to thank David-Étienne Bouchard, Caroline Chevalier, Philippe Gagnon,
Mireille Tremblay, and Daniel Valois for their judgements on Québec French; Emanuela Sanfelici
for her judgements on Italian, and Calixto Agüero-Bautista for his judgements on Spanish.
This research was supported in part by a SSHRC doctoral fellowship (#752-2007-2382), and
the SSHRC MCRI Les voies du français (PI: France Martineau).
 Heather Burnett

(1) a. Tu t’es fait couper les cheveux.

You CL-have made cut the hair
“You got your hair cut.”
b. Tu t’es fait couper les cheveux!
“You got your hair cut!”

However, some utterances are infelicitous in non-exclamative contexts. These are

known as exclamative sentences, and, as discussed in many works such as Michaelis &
Lambrecht (1996), Zanuttini & Portner (2003), and Rett (2009), these utterances have
particular semantic and pragmatic properties, which will be discussed over the course
of this paper. Some examples of sentences that have previously been argued to belong
to the exclamative class in French are shown in (2).
(2) a. Qu’est-ce qu’ils sont beaux, tes cheveux!
What-is-it that-they are nice, your hair
b. Qu’ils sont beaux, tes cheveux!
What-they are nice, your hair
c. Ostie qu’ils sont beaux, tes cheveux! (Québec French)
Ostie that-they are nice, your hair
“How nice your hair is!”

In most current theories of the syntax and semantics of exclamative sentences, the
force of an utterance used in the act of exclaiming is given by an exclamative illo-
cutionary force operator (ex. Zanuttini & Portner’s F(active) morpheme, and Rett’s
Degree E-FORCE). In order to apply, these operators require a specific semantic and
discourse structure, and this is licensed by particular syntactic configurations. By far,
the most studied of these configurations is the wh-morphology in wh-exclamatives
such as those in (2a,b). In fact, some authors have even argued that exclamatives, by
definition, contain wh morphology (cf. Obenauer (1994) for French).
In this paper, I argue that the exclamative operator can be licensed by other
means. In particular, I argue that it can be licensed by prosody, more specifically,
by contrastive focus on a certain class of non-wh items with the proper semantics.
These elements are gradation quantifiers. Following Heim (2000), I call gradation
quantifiers those quantifiers that compare the degree to which a property holds of
an individual with a “standard” degree or “threshold” specified by a consecutive
clause. The set of English gradation quantifiers includes too…for, so…that, and
enough…for.
Furthermore, I argue that, while gradation exclamatives are not unheard of in
languages like English, they constitute a robust class of exclamatives in the Romance
languages. In most of what follows, my data will come from Québec French (QF),
where the construction in question is widespread; however, as will be discussed below,
we find similar patterns in European French, Italian, and Spanish.
 Pitch accent, focus, and the interpretation of non-wh exclamatives in French 

The main goal of my paper is the analysis of the following paradigm of utterances
containing gradation quantifiers, exemplified here by assez “enough”:

(3) a. J’ai lu un livre assez difficle (pour moi). Assertion

“I read a book difficult enough (for me)”
b. J’ai lu un livre ASSEZ difficile (pour moi)…pas TROP difficile.
 Contrastive Focus
“I read a book difficult ENOUGH (for me)…not TOO difficult”
c. J’ai lu un livre assez difficile (pour moi)! Exclamation
“I read a book that difficult enough for me!”
d. J’ai lu un livre ASSEZ difficile! Exclamative
“I read SUCH a difficult book!/What a difficult book I read!”

(3a) is a simple assertion: it communicates the fact that I read a book that was difficult
enough for me. (3b) is another assertion, but one with contrastive focus on assez. It
states that I read a book difficult enough, and contrasts the degree of the book’s dif-
ficulty with another degree already in the common ground: in this case, the degree of
being “too” difficult. (3c) is an exclamation similar to (1c): it expresses the speaker’s
surprise at the fact that I read a book that was difficult enough for me (perhaps books
are usually too easy). Finally, (3d), I will argue, is an exclamative: it expresses surprise
at the degree of the book’s difficulty, namely, that it is extreme.
The heart of my proposal concerns the relation between (3b) and (3d): I argue
that what assertions with f(ocus)-marked gradation quantifiers have in common with
gradation exclamatives, and exclamatives more generally, is their presuppositions: In
order to be felicitous, both contrastively focused gradation constructions and exclama-
tives require that a degree property be salient in the common ground. Following Rett
(2009), I assume that this is a precondition on the application of exclamative operator,
one that is met in the case of exclamatives and focused gradation constructions, but
not in many other constructions.
The paper is organized as follows: In Section€1, I present the basic data on the use
of gradation constructions in assertions with and without focus. I present the main
proposals from the literature that I adopt for the analysis of the semantics of gradation
quantifiers and contrastive focus. In Section€2, I examine gradation constructions used
in the act of exclaiming. Following Rett, I argue that unfocused gradation construc-
tions used in these contexts are simple exclamations: they express surprise on behalf
of the speaker at a proposition in the common ground. I then examine sentences used
in the act of exclaiming with focused gradation quantifiers. I propose that these utter-
ances form part of a special subclass of exclamations: exclamative sentences. I apply
both Portner & Zanuttini (2003)’s and Rett (2009)’s semantic critera for identifying
exclamatives, and show that, despite their lack of wh-morphology, exclamations with
focused quantifiers satisfy them. Finally, in Section€ 3, I present a semantic analysis
 Heather Burnett

of these gradation exclamatives, and show how it explains two puzzling properties of
the construction.

2.â•… Gradation quantifiers in assertions

In theories that take into account the communicative function of an utterance (for
example Searle (1969)), its meaning is often divided into two components: the semantic
content of an utterance, and the illocutionary force with which it is uttered. The semantic
content of an utterance is, in simplest terms, its denotation: it is the value that is obtained
by combining the various lexical items that constitute the expression together by means
of compositional semantic rules like function application etc. The illocutionary force of
an utterance has to do with the intention of the speaker: an expression that is uttered
with the speaker wishing to communicate its truth is known as an assertion. Assertions
can be descriptively correct or incorrect, i.e. true or false.
In this section, I present the basic data on the use of gradation quantifiers in asser-
tions in Québec French and other Romance languages, and the analysis of their semantic
content that I am adopting. I then consider the sentences in which they are focused, and
present the semantic analysis that I am adopting for contrastive focus (CF).

2.1â•… Assertions with no focus

As mentioned in the introduction, gradation quantifiers compare the degree to which
a property, introduced by an adjective, holds of an individual with a threshold degree
given by a consecutive clause. Assertions containing non-focused gradation quantifiers
in Québec French are shown below with the quantifiers tellement…que “so much…
that”, assez…pour “enough…for”, and trop…pour “too much…for”.
(4) a. J’ai lu un assez bon livre pour l’acheter.
“I read a book on this subject good enough to buy it”
b. J’ai lu un livre tellement difficile que j’ai commence à pleurer.
“I read a book so difficult that I started crying”
c. J’ai lu un livre trop difficile pour l’apprécier.
“I read a book too difficult to appreciate it”

Formally, following von Stechow (1984), I assume that gradeable adjectives like difficile
“difficult” denote relations between individuals and degrees. Thus, we can represent the
denotation of difficile as the function in (5).
(5) [[difficile]] = λdd λxe. x is d difficult

difficile takes a degree and an individual and yield “true” just in case the individual
is difficult to that degree. Gradation quantifiers take a threshold degree to form a
 Pitch accent, focus, and the interpretation of non-wh exclamatives in French 

generalized quantifier over sets of degrees. Denotations for assez “enough”, trop “too”
and tellement “so” in a model M under an assignment function α, are shown in (6).
(6) a. [[assez…pour]]M,α = λd’dλP<d,t>. max1(P) ≥ d’ and d’ = min({d: it is
�required for2 P(d) = 1})
b. [[trop…pour]]M,α = λd’dλP<d,t>. max(P) ≥ d’ and d’ = max({d: it is allowed
for P(d) = 1})
c. [[tellement…que]]M,α = λd’dλP<d,t>. max(P) ≥ d’ and d’ = min({d: it is
required for P(d) = 1})

The quantifier assez…pour, for example, states that the maximum degree in a degree
property P is greater than (or equal to) another degree, and this degree is the mini-
mum that is required for the property to hold. So, under this account, the truth
conditions for a sentence like (7a) are as shown in (7b), i.e. the sentence is true just
in case John’s handsomeness is greater than or equal to the handsomeness that is
required to date Marie.
(7) a. Jean est assez beau pour sortir avec Marie.
“Jean is handsome enough to date Marie.”
b. 1 iff max({d: Jean is d handsome})≥ ddate Marie

Although gradation quantifiers canonically select for an overt consecutive clause, if

the threshold of comparison can be recovered from context, the that/for clause may be
left unpronounced.
(8) Le prof m’a demandé de lire un livre qui était assez difficile.
“The teacher asked me to read a book that was difficult enough.”

.╅ As discussed in Rullman (1995), the proper truth conditions for sentences with compara-
tives and gradation quantifiers are given through comparing the maximum degree to which
a property holds with another degree. Since assez, trop etc. apply to sets of degrees, we must
make use of a maximality operator such as in (i).
(i) Let DEG be a set of degrees ordered by the relation ≥,
the max(DEG) = ιd[d є DEG and for all d’ є DEG [d’ ≤ d]] (Rullman (1995: 68))
.╅ It is well known that the quantifiers so, too, and enough and their French and Italian
translations involve some type of modality inside the consecutive clause (cf. the discussion
in Meier (2003)). In fact, it is in the formalization of the modal in the subordinate clause
that the majority of the proposals for the semantics of gradation quantifiers differ. Since, as
we will see, the semantic content of the consecutive clause is not going to be of particular
importance in the phenomenon that we are analyzing here, in the definitions provided in (6),
I have simply stated this modality in English words. For alternative proposals for the semantics
of the consecutive clause, see, among others, von Stechow (1984), Heim (2000), Meier (2003),
and Harquard (2006).
 Heather Burnett

(8) is good when a phrase headed by pour, for example pour moi “for me”, is straight-
forwardly supplied by the discourse context. Formally, I represent the filling in of the
threshold from context as the insertion of a free variable, say d1, as the first argument
of the quantifier.

(9) [[J’ai lu un livre assez difficile]] M,α = 1 iff max({d: I read d difficult book}) ≥ α(d1)

The truth of (9), where the value of the threshold is implicit, is dependent on the value
assigned to d1 by the assignment function α.
Québec French also contains a number of gradation quantifiers that obligatorily
take an implicit threshold, i.e, the value to which the degree of the main predicate is com-
pared is always supplied by context. An example of such a quantifier is pas mal “fairly”.
(10) J’ai lu un livre qui était pas mal difficile.
“I read a book that was fairly difficult.”

Like assez, quantifiers like pas mal relate degree properties with threshold degrees.
However, the latter are lexically specified to take implicit variables as threshold
arguments. Thus, the lexical entry for pas mal can be directly given as (11).
(11) [[pas mal]]M,α = λddλP〈d,t〉. max(P) ≥ α(d)

A sentence like J’ai lu un livre qui était pas mal difficile is true just in case α maps d to
the degree to which a book needs to be difficult in order to count as “fairly” difficult,
and the degree to which the book that I read is difficult exceeds the value of d.
In summary, I have followed the majority of the literature on comparatives and
gradation quantifiers in supposing that they combine with a threshold supplied by a con-
secutive clause to form a generalized quantifier over degrees: a function from a degree
property to a truth value. When the threshold is recoverable from context, or when the
threshold is lexically specified to be implicit, I suppose that the quantifier takes a free
variable as its first argument, and its value is given by an assignment function.

2.2â•… Assertions with focused gradation quantifiers

I now turn to assertions in which these quantifiers are focused, or f-marked. By focus/
f-marking, I simply mean pitch accent (stress) on the quantifier. As shown in (12),
asserting a proposition with stress on the degree quantifier results in contrastive focus
(CF) on the quantifier.
(12) Le prof m’a demandé de lire un livre qui était ASSEZ difficile pour moi…pas
TROP difficile.
“The teacher asked me to read a book that was difficult ENOUGH for me…not
TOO difficult”

Again, the consecutive clause may be dropped if it is clear from context.

 Pitch accent, focus, and the interpretation of non-wh exclamatives in French 

(13) Le prof m’a demandé de lire un livre qui était ASSEZ difficile…pas TROP difficile.
“The teacher asked me to read a book that difficult ENOUGH…not TOO difficult”

In the analysis of contrastive focus, I follow much recent work that proposes that, in
a sentence with focus marking, it is the destressed material, not the stressed material,
that is semantically marked. In particular, I assume that this unstressed material must
be given (Selkirk (1996), Schwarzchild (1999)) or anaphoric to the context (Büring
(2008)). This is stated as the generalization in (14).
(14) The absence of f-marking indicates giveness in the discourse. (Schwarzchild
(1999; his (10b)))

Schwarzchild gives the following definition of giveness:

(15) An utterance U counts as GIVEN iff it has a salient antecedent A and A entails
U modulo ∃-type shifting. (Schwarzchild (1999; his (18))

In other words, for sentences with f-marked constituents to be well formed, they must
have antecedents in the discourse that are composed of the non-f-marked constituents
of the sentence. Therefore, in a sentence with a focused gradation quantifier, for the
truth of the entire sentence to be evaluated, the degree property that forms the argu-
ment of the focused quantifier must be salient in the discourse. For example, focus
on assez in (16a) requires that the book’s difficulty (i.e. the property: λd. I read a book
d-difficult) be salient in the discourse. The meaning of (16a) would then be a function
from worlds in which we have been talking about the difficulty of a book that I read
(the presupposition is indicated by square brackets ([])) to truth values (16b).
(16) a. J’ai lu un livre ASSEZ difficile.
“I read a book difficult ENOUGH”
b. λw. [λd λw’. I read a book d-difficult in w’ is salient in the discourse]
assez(α(d’))(λd. I read a book d-difficult) = 1 in w.

2.3â•… Summary
In summary, in the spirit of von Stechow (1984) inter alia, I assume that gradation
quantifiers denote relations between degrees: they relate the maximum degree to
which the main predicate holds to a threshold degree that can be overtly specified (in
the case of assez, tellement, and trop). Thus, when the threshold is overtly specified,
the gradation construction denotes a closed proposition, and when the threshold
is given by context, it denotes an open proposition, i.e. a proposition with one free
variable of type d.
Following Schwarzchild (1999) inter alia, I assume that the role of focus is to mark
“new” information. In particular, I assume that, when a constituent smaller than the
entire sentence is f-marked, the remaining part of the sentence must be given, i.e. salient
 Heather Burnett

in the previous discourse. In the case where the gradation quantifier is f-marked, what
must be salient in the discourse is a degree property.

3.â•… Gradation quantifiers in exclamations

In this section, I present the data on unfocused and focused gradation quantifiers in
utterances used in the act of exclaiming. In other words, I consider what happens when
we utter an expression denoting a proposition containing a gradation quantifier with
the illocutionary force as an exclamation.

3.1â•… Exclamations with no focus

In this section, I examine utterances used in the act of exclaiming containing unfo-
cused gradation quantifiers. As is standard, I take the act of “exclaiming” to be the
expression of surprise on behalf of the speaker.3 The first type of exclamation is the
simple propositional exclamation: when the speaker expresses surprise at the truth
of a particular proposition. In English and French etc. these exclamations are simply
indicated by uttering the entire expression with exclamational intonation. An example
of such an exclamation formed from a gradation construction is shown in (17).
(17) J’ai lu un livre qui était assez difficile pour moi!
“I read a book that was difficult enough for me!”
Following Kaplan (1999) and Rett (2009), I assume that, while assertions are descrip-
tively (in)correct (true/false), exclamations are expressively (in)correct. Rett provides
the following “expressive conditions” for exclamations:
(18) “We now have two requirements for what makes the utterance of an
exclamation expressively correct: its content must be salient, and the
speaker must find this content surprising in some way” (Rett (2009: 3))
Thus, the expressive correctness of an utterance is given by the illocutionary operator
PROPOSITION E-FORCE, which is a function from a proposition in the conversational
background to a speaker’s expression of surprise.
(19) PROPOSITION E-FORCE(p) is expressively correct in context C iff p is salient
in C, and the speaker in C is surprised that p. (Rett (2009; her (15)))
So PROPOSITION E-FORCE applied to the proposition J’ai lu un livre assez difficile
pour moi yields the exclamation in (17), which is expressively correct just in case the

.╅ There are many apparent counter examples to this claim; however, I refer the reader to
the discussion in Zanuttini & Portner (2003) and Rett (2009) for arguments that surprise is,
truly, the right notion.
 Pitch accent, focus, and the interpretation of non-wh exclamatives in French 

proposition that I read a book difficult enough for me is salient in the context, and the
speaker is surprised at this fact.
In the next sections, I examine the use of gradation constructions with focused
quantifiers to express speaker surprise. I argue that these utterances are more than sim-
ple exclamations; rather, they form part of the class of exclamatives in French. However,
I first review the literature on the criteria for classifying utterances as exclamatives.

3.2â•… Identifying exclamatives

Propositional exclamations are completely general: PROPOSITION E-FORCE may
be applied to any proposition to create one, provided that this proposition is associ-
ated with surprise on behalf of the speaker; however, different types of exclamations
are more restricted. In particular, exclamatives, utterances of the type shown in
(20), have a number of syntactic and semantic restrictions that do not apply to
propositional exclamations.
(20) Comme t’as de beaux cheveux!
“What nice hair you have!”

For Rett, what distinguishes exclamatives from other exclamations is that, as noticed
by Milner (1978) and Gérard (1980), the former are subject to the following semantic
restriction that the latter are not:
(21) The (Extreme) Degree Restriction: Exclamatives may only express the surprise
of the speaker at the extreme degree to which a property holds.

While an exclamative like What languages Mimi speaks! (in French: Quelles langues
qu’elle parle, Mimi!) is appropriate to express surprise at the degree of difficulty or
“exotic-ness” of the languages that Mimi speaks, it is inappropriate to express surprise
at what those individual languages are. For example, consider the following scenario
described by Rett: You’ve heard that Mimi speaks two Romance languages in addition
to speaking English. You know that Mimi’s mother is Swiss, and so you assume that
these two languagse are French and Italian. However, you learn that Mimi instead
speaks Portugese and Romanian. In this case, the propositional exclamation in (28a) is
expressively correct; however, the exclamative is not (22b).
(22) a. (Wow), Mimi speaks Portugese and Romanian!
((Wow), Mimi parle le portugais et le roumain!)
b. #(My), What languages Mimi speaks! (Rett (2009:€5),
(Quelles langues qu’elle parle, Mimi!)

To account for this defining characteristic of exclamative clauses, Rett proposes that
the illocutionary force operator applies not to propositions in exclamatives, but rather
to degree properties that are salient in the discourse. Thus, for Rett, the semantic
 Heather Burnett

content of an exclamative is a function from degrees to propositions, and the exclamative

operator, DEGREE E-FORCE, applies to these degree properties.

(23) DEGREE E-FORCE(D〈d,〈s,t〉〉) is expressively correct in context C iff D is salient

in C and ∃d, d > s [the speaker in C is surprised that λw. D(d) in w]

Thus, in Rett’s theory, “the utterance of an exclamative is expressively correct if its con-
tent is a degree property which is salient in the discourse, the speaker is surprised that
a specific degree holds of that degree property, and that degree exceeds a contextually
provided standard s” (Rett (2009:€11)).
Zanuttini & Portner (2003) provide two4 other criteria for distinguishing excla-
matives from simple exclamations. Firstly, they propose that, since like Rett, they
propose that the semantic content of an exclamative is presupposed, i.e. already in
the common ground, they may only be appear in factive contexts.5 For example,
while English exclamatives can be embedded under know, they are ungrammatical
under non-factive verbs like think and wonder.

(24) Mary knows/*thinks/*wonders how very cute he is.

Zanuttini & Portner’s second criterium is that exclamative sentences are illicit in both
parts of question/answer pairs. As shown below (25a), for an utterance with a wh word
to be a question, inversion must take place. If there is no inversion, the utterance must
be an exclamative (25b).

(25) a. Quel homme a-t-elle épousé?

Which man has-she married
“Which man did she marry?”
b. Quel homme (qu’) elle a épousé!
Which man (that) she has married
“What a man she married!”

Additionally, using an exclamative to answer a question is bizarre.

.â•… They actually provide three criteria, but their third one is sufficiently similar to Rett’s
extreme degree restriction.
.╅ This observation needs to be refined somewhat, in particular to upward monotone
factive contexts. Note that a downward monotone operator like regrets does not license a
wh-exclamative, nor does it license a degree quantifier exclamative either.

(i) a. *Mary regrets how very cute he is!

b. *Mary regrets that he did SO much work!
 Pitch accent, focus, and the interpretation of non-wh exclamatives in French 

(26) A: Est-ce qu’elle a épousé un homme impressionnant?

“Did she marry an impressive man?”
B: #Quel homme (qu’)elle a épousé!
“# What a man she married!”

3.2.1â•… Gradation exclamatives

With these properties in mind, I consider exclamations with a focused gradation
quantifiers such as (27).
(27) J’ai lu un livre ASSEZ difficile!
I-have read a book ENOUGH difficult
“What a difficult book I read!”
I argue that these constructions display the same properties that Rett and Zanuttini &
Portner have noticed with wh-exclamatives. Firstly, exclamations with focused grada-
tion quantifiers display Rett’s extreme degree restriction: Clearly, the speaker’s surprise
may only be about the degree to which the adjectival property holds. Additionally,
this degree must be “extreme”. For example, suppose Jean is very ugly and, as such, we
expect him to marry an ugly girl. If it turns out that he marries someone moderately
beautiful, we cannot say (28), even though this fact is surprising.
(28) #Jean a épousé une fille ASSEZ belle! …mais elle était pas
Jean has married a girl ENOUGH beautiful! …but she was not
extrêmement belle.
extremely beautiful
“# Jean married SUCH a beautiful girl!…but she wasn’t extremely beautiful”
Secondly, exclamations with focused quantifiers may only be embedded in contexts
where the truth of the proposition containing the quantifier is presupposed or asserted.
For example, ASSEZ is only possible with the factive savoir “to know” (29a), or c’est vrai
“it’s true”.
(29) a. Marie sait que Jean a épousé une fille ASSEZ belle!
Marie knows that Jean has married a girl ENOUGH beautiful
“Marie knows that Jean married SUCH a beautiful girl!”
b. C’est vrai que Jean a épousé une fille ASSEZ belle!
It’s true that Jean has married a girl ENOUGH beautiful
“It’s true that Jean married SUCH a beautiful girl!”
c. *?Marie se demande si Jean a épousé une fille ASSEZ belle!
Marie CL wonders if Jean has married a girl ENOUGH beautiful
“*? Marie wonders if Jean married SUCH a beautiful girl!”
d. *Marie pense que Jean a épousé une fille ASSEZ belle!
Marie thinks that Jean has married a girl ENOUGH beautiful
“*Marie thinks that Jean married SUCH a beautiful girl!”
 Heather Burnett

Finally, focused gradation constructions are impossible in questions (30), and are
somewhat bizarre when used as answers (31).
(30) a. *Jean, il a-tu épousé une fille ASSEZ belle? (Yes/No Question)
Jean, he has-Q married a girl ENOUGH beautiful
“*Did Jean marry SUCH a beautiful girl?”
b. *Qui a épousé une fille ASSEZ belle? (Wh-Question)
Who has married a girl ENOUGH beautiful
“*Who married SUCH a beautiful girl?”

(31) A: Jean, il a-tu épousé une belle fille?

“Did Jean marry a beautiful girl?”
B: ?Jean, il a épousé une fille ASSEZ belle!
“Jean married SUCH a beautiful girl!”

In conclusion, I have argued that sentences with focused gradation quantifiers con-
form to the semantic criteria for identifying exclamatives proposed by both Zanuttini &
Portner (2003) and Rett (2009). I therefore propose that this construction, that I will
henceforth refer to as the gradation exclamative construction, constitutes a new class
of exclamatives, one that does not involve wh-words.
In the next section, I show that gradation exclamatives have two rather puzzling
semantic and syntactic properties that distinguish them from assertions with con-
trastively focused gradation quantifiers. Thus, not only do gradation exclamatives
have many similarities with wh-exclamatives, they also display many differences
from regular and contrastively focused assertions.

3.3â•… Additional syntactic and semantic properties of gradation exclamatives

In addition to displaying similar semantic properties to wh-exclamatives, gradation
exclamatives further distinguish themselves from assertive gradation constructions in
both a semantic way and a syntactic way.

3.3.1â•… Semantic neutralization

The most striking property of gradation exclamatives concerns the semantic behaviour
of the gradation quantifiers that participate in this construction. In the previous parts of
the paper, I have illustrated the construction using the quantifier assez “enough”, which,
in its normal assertive use means “sufficiently…for”. In an assertion with contrastive
focus, it keeps this meaning. However, as shown in all the examples in the previous
section, when focused in an exclamation, it means “extremely”.
The change in the meaning of assez is not limited to this lexical item. In fact, in
Québec French, all of the gradation quantifiers assez, trop ‘too”, tellement “so”, and pas
mal “fairly”, mean “extremely” when used in a gradation exclamative.
 Pitch accent, focus, and the interpretation of non-wh exclamatives in French 

(32) a. Marie est ASSEZ belle!

Marie is ENOUGH beautiful
b. Marie est TROP belle!
Marie is TOO beautiful
c. Marie est TELLEMENT belle!
Marie is SO beautiful
d. Marie est PAS MAL belle!
Marie is FAIRLY beautiful
“Mary is SO beautiful!/How beautiful Mary is!”

We find a similar pattern of semantic neutralization in Italian: my speaker tells me that

all four quantifiers talmente, tanto “so”, troppo “too”, and abbastanza “enough” all mean
“extremely” when focused in an exclamation.
(33) a. Mario è TALMENTE bello!
Mario is SO handsome
b. Mario è TANTO bello!
Mario is SO handsome
c. Mario è TROPPO bello!
Mario is TOO handsome
d. Mario è ABBASTANZA bello!
Mario is ENOUGH handsome

However, there seems to be significant dialectal and cross-linguistic variation in which

quantifiers can participate in the construction:6 while almost all gradation quantifiers
in Québec French and Italian are neutralized into meaning “extremely” when focused
in an exclamation, the only quantifiers that can appear in gradation exclamatives in
European French (34), Spanish (35) and English (36) are those that mean, roughly, so:7
(34) a. Elle fut SI serviable! (European French (Gérard (1980:3)))
She was SO obliging!
b. C’est TELLEMENT extraordinaire!
It’s SO extraordinary!
c. Il nous a fait TANT de peine!
They gave us SO MUCH sadness!

.╅ And indeed, within a single language, why some quantifiers like assez can participate in
the construction, but other similar ones like the comparative plus ‘more’, at this point, remains
mysterious.
.╅ However, for some speakers of English, rather also fits this pattern: Hey, that turned out
to be a RATHER good film.
 Heather Burnett

(35) Ayer leí un libro TAN bueno! (Spanish)

Yesterday (I) read a book SO good!
(36) a. John is SO tall!
b. John is SUCH a tall man!

However, some speakers of European French and English also accept the following
examples with too/trop:

(37) a. C’est TROP bien!8

b. That’s TOO good!

In summary, gradation exclamatives have different semantic properties from contras-

tively focused gradation constructions. In particular, with quantifiers that do not have
an “extreme” meaning, their meaning becomes extreme.
In principle, the quantifiers presented in this section are not the only ones that can
form gradation exclamatives. Other elements in QF like the swear word crissement “lit.
christly” and the adverb énormément “enormously” work very well in exclamatives.

(38) a. Ta robe est CRISSEMENT belle!

Your dress is CRISSEMENT beautiful
“Your dress is SO beautiful!”
b. Ta robe est ÉNORMÉMENT belle!
“Your dress is ENORMOUSLY beautiful!”

However, since these quantifiers already have an extreme meaning, we do not

immediately see their semantic neutralization.

3.3.2â•… Syntactic neutralization

The second property that distinguishes gradation exclamatives from contrastively
focused assertions was first noticed by Gérard (1980) for European French. Recall that,
in CF assertions, the consecutive clause selected for by assez, tellement, and trop may be
either overt (Jean est ASSEZ beau pour sortir avec Marie ‘Jean is handsome ENOUGH
to date Mary’), or left implicit if it is recoverable from context. Gérard observes that
when a gradation construction is used in an exclamation, it becomes an énoncé
tronqué “truncated utterance” (Gérard (1980:€3)): that/for must be implicit.
(39) a. J’ai lu un ASSEZ bon livre! *pour…
I-have read a ENOUGH good book ↜for…

.â•… cf. Rapper Fatal Bazooka’s 2007 hit J’aime trop ton boule!
 Pitch accent, focus, and the interpretation of non-wh exclamatives in French 

b. J’ai lu un TELLEMENT bon livre! *?que…

I-have read a SO good book ╇ that…
c. J’ai lu un TROP bon livre! *pour…
I-have read a TOO good book ↜for

In other words, despite the fact that tellement, assez, and trop all select for their dif-
ferent consecutive clauses, when used in a gradation exclamative, they all become
like pas mal, that obligatorily takes an free variable as a threshold. This syntactic
neutralization of gradation quantifiers is also observed in Italian (40a), Spanish
(40b), and English (40c).
(40) a. Mario è TROPPO bello *per…
Mario is TOO handsome *to…
b. Ayer, leí un libro TAN bueno *que decidí comprarlo!
Yesterday, (I) read a book SO good *that (I) decided to buy it!
c. John is SUCH a tall man *that he can dunk a basketball!

In summary, unlike gradation constructions with contrastive focus, where the con-
secutive clause that specifies the threshold may only be left implicit if it is straightfor-
wardly recoverable from the context, gradation exclamatives may not take an overtly
specified threshold.9 Intuitively, this is because, in the exclamative, threshold degree
must always be surprisingly extreme, and there is no consecutive clause that can
properly denote this degree. I formalize this intuition when I present my semantic
analysis of gradation exclamatives in the next section.

4.â•… A semantic analysis of gradation exclamatives

In Section€3.2, I argued that gradation exclamatives display many of the same semantic
properties as wh-exclamatives, properties that have been attributed to the presence
of an exclamative illocutionary operator, such as Rett’s DEGREE E-FORCE, the
definition of which is repeated below.

.╅ The judgements on the possibility of having a consecutive clause with the focused quanti-
fier so are a little tricky: this is because, unlike too or enough, in order for an assertion of the
form so P that Q to be true, P has to independently be true. For example, John is so tall that
Q entails that John is tall. Therefore, it is easy to add an overt consecutive clause like that it’s
surprising to construct an assertion that has a similar content to the exclamative. The point
here is not whether, if we change the intonation of the utterance, it is possible to include an
overt standard that is, itself, extreme; it’s that, in the basic uses of these quantifiers in excla-
matives, the threshold is not coming from the previous discourse, but rather from the act of
exclaiming itself.
 Heather Burnett

(41) DEGREE E-FORCE(D〈d,〈s,t〉〉) is expressively correct in context C iff D is salient

in C and ∃d, d > s [the speaker in C is surprised that λw. D(d) in w]

I therefore propose that it is this operator that appears in gradation exclamatives to

give them the extreme degree interpretation that we have seen in the previous sections
of the paper.
Unlike PROPOSITION E-FORCE, that is allowed to apply to any structure whose
semantic content is a proposition, and therefore has a very wide distribution, the dis-
tribution of DEGREE E-FORCE is more restricted: it only applies to degree properties,
and then, only when there is an appropriate degree property salient in the discourse.
There are very few syntactic/discourse configurations that meet this description.
According to Rett, wh-exclamatives are one of them. I propose that sentences with
contrastively focused gradation quantifiers with implicit thresholds are another.
Firstly, as shown in (41), the exclamative operator must apply in contexts in which
there a degree property is already present in the common ground. As discussed in
Section€ 3.2, focus on the gradation quantifier signals the presence of such a salient
property. For example, consider the exclamative in (42).
(42) J’ai vu un film ASSEZ bon!
I-have seen a film ENOUGH good
“I saw SUCH a good movie!”

Pitch accent on the quantifier, requires that the film’s “goodness” be salient in the
discourse.
(43) λd1. I saw a d1 good movie

Thus, the first condition for expressive correctness given by (41) is met.
Secondly, the exclamative operator must apply to a degree property. It then, itself,
introduces an objectively high threshold degree and asserts that the values for which
the property is true are greater than this threshold. The semantic content of (42), which
can be represented as the function in (43), is not directly a degree property. However,
it is, as discussed in Section€4.1., an open proposition: a proposition with a free degree
variable, d’.
(44) λw. max({d: I saw a d good movie}) ≥ α(d’) in w

In assertions, the truth of (44) is given by looking at the value that α assigns to d’. Thus,
the entire expression is equivalent to the degree property in (45).
(45) λd’λw. max({d: I saw a d good movie}) ≥ d’ in w

I propose that, in a gradation exclamative, DEGREE E-FORCE applies to the property

in (45); in other words, it binds the free threshold variable. Therefore, in my proposal,
the expressive correctness conditions for (42) are given in (46).
 Pitch accent, focus, and the interpretation of non-wh exclamatives in French 

(46) J’ai vu un film ASSEZ bon! is expressively correct in context C iff λd1.
I saw a d1 good movie is salient in C and ∃d2, d2 > s [ I am surprised that
max({d: I saw a film d good}) > d2]

In other words, J’ai vu un film ASSEZ bon! is predicted to be expressively correct just in
case I am surprised that the film’s “goodness” is extremely high.

4.1â•… Consequences
This proposal has a number of consequences, in particular with respect to the
properties discussed in Section€3.3.
Firstly, it explains the semantic neutralization facts: In their canonical use in
assertions, gradation quantifiers differ in the size of the threshold that the content of
the main clause is being compared to. Since, with every quantifier, the exclamative
operator binds the threshold and asserts that it is surprisingly high, sentences with
assez, tellement, trop, pas mal etc. are all predicted to be expressively correct in the
same situations.
Secondly, it explains the syntactic neutralization facts: Since the exclamative
operator binds the threshold value, it is impossible to add a subordinate clause that
would overtly specify it. This analysis also predicts that, when a for-clause is present,
the only type of exclamation that is possible is propositional:
(47) J’ai bu trop de café pour m’endormir!
“I drank too much coffee to get to sleep!”

Unlike an exclamative, (47) is not required to have an “extreme degree” reading: it

is felicitous even if drinking only 2 cups of coffee is sufficient to inhibit the speaker’s
sleeping, and the fact that she drank this amount is surprising.
In summary, I proposed that, in gradation exclamatives, the exclamative
DEGREE E-FORCE operator binds the free variable inside a gradation construction
with an implicit threshold clause. I proposed that this is possible because focus on the
gradation quantifier sets up the proper discourse context for the illocutionary force
operator to apply.

5.â•… Conclusion

In conlusion, I have presented data and a semantic analysis of a previously undiscussed

class of exclamative sentences: gradation exclamatives. I proposed that the semantic
content of a gradation exclamative is a gradation construction with an implicit thresh-
old clause and focus on the quantifier. I proposed that DEGREE E-FORCE, argued for
in Rett (2009), binds the threshold yielding an extreme degree reading, and that the
presence of this operator is licensed by the focus structure of the sentence.
 Heather Burnett

I argued that these utterances constitute a robust class of exclamatives within the
Romance languages, particularly in Québec French and Italian. I showed that, despite
their lack of wh-morphology, gradation exclamations display many of the same seman-
tic properties observed with wh-exclamatives. This result is significant because it shows
that the semantic properties that characterize wh-exclamatives are not uniquely due to
the presence of wh-morphology in these utterances. This paper thus supports the work
of Zanuttini & Portner (2003) and Rett (2009) that define the exclamative clause type
on the basis of semantic, rather than syntactic or morphological properties.

References

Büring, D. 2008. “What’s New (and What’s Given) in the Theory of Focus.” in the Proceedings of
the 2008 meeting of the Berkely Linguistics Society.
Gérard, J. 1980. L’ exclamation en français. Niemeyer, Tübingen.
Harquard, V. 2006.“Aspects of too and enough Constructions.” Proceedings of Semantics and
Linguistic Theory 15, ed. by E. Georgala & J. Howell. Ithaca, NY: CLC Publications.
Heim, I. 2000. “Degree Operators and Scope.” Proceedings of Semantics and Linguistic Theory 10,
ed. by B. Jackson & T. Matthews. Ithaca, NY: CLC Publications.
Kaplan, D. 1999. “The Meaning of “ouch” and “oups””. Ms. University of California, Los Angeles.
Meier, C. 2003. “The Meaning of too, enough, and so that”. Natural Language Semantics, 11: 69–107.
Michaelis, L. & K. Lambrecht, 1996. “The Exclamative Sentence Type in English,” Conceptual
Structure, Discourse and Language, ed. by A. Goldberg. 375–389. CSLI.
Milner, J-C. 1978. De la syntaxe à l’interprétation: Quantités, insultes, exclamations. Paris: Édi-
tions de Seuil.
Obenauer, H-G. 1994. Aspects de la syntaxe A-barre: Effets d’intervention et mouvements
desquantifieurs. Thèse d’État. Université de Paris VIII.
Rett, J. 2009. “A Degree Account of Exclamatives”. Proceedings of Semantics and Linguistic Theory
18. Ithaca, NY: CLC Publications.
Rullman, H. 1995. Maximality in the Semantics of Wh-constructions. Ph.D. Dissertation, University
of Massachusetts, Amherst.
Schawarzchild, R. 1999. “GIVENness, AvoidF, and other Constraints on the Placement of
Accent.” Natural Language Semantics, 7: 141–177.
Searle, J. 1969. Speech Acts. Cambridge: Cambridge University Press.
Selkirk, E. 1996. “Sentence Prosody: Intonation, Stress and Phrasing,” The Handbook of
Phonological Theory, ed. By J. Goldsmith. Basil Blackwell: London.
Von Stechow, A. 1984. “Comparing Semantic Theories of Comparison.” Journal of Semantics.
3: 1–77.
Zanuttini, R. & Portner, P. 2003. “Exclamative Clauses: at the Syntax-Semantics Interface.”
Language. 79: 39–81.
 Detours along the perfect path

Patrícia Amaral & Chad Howe

University of Liverpool & The University of Georgia

The development of periphrastic past constructions in Romance, including

those that do not take a reflex of Latin habere as an auxiliary, has been analyzed
along a continuum from a resultative construction to a perfect, and in some
cases to a perfective (see Harris 1982; Fleischman 1983). This paper argues
that the development of the Pretérito Perfeito Composto (PPC) in Portuguese
does not adhere to the proposed typologies of periphrastic past evolution in
Romance. Using diachronic corpus data, we revisit the proposed resultative >
perfect grammaticalization path (see Bybee et al. 1994) and contend that the
developmental trajectory of the Portuguese PPC is distinct from other cases
of periphrastic past evolution in Romance languages, specifically Spanish. We
demonstrate that the iterative meaning unique to the PPC in contemporary
Portuguese arises in morphosyntactically ambiguous contexts in which the
ter + Past Participle construction co-occurs with semantically plural complements.

1.â•… Introduction1

The functional distribution of periphrastic past (i.e. Perfect) forms in Romance has
been treated extensively in the literature on language change. The dominant view
(e.g. Harris 1982; Fleishman 1983; Squartini & Bertinetto 2000; Schwenter & Torres
Cacoullos 2008) is that the verbal periphrases formed by ‘have/be’ + Past Participle
across varieties of Romance form a continuum whose extremes are measured by the
degree to which the form has developed uses beyond its source as a small clause con-
struction. It is argued, for example, that the periphrastic past in Sicilian is at one end of
the continuum: it retains the original meaning of the resultative construction, refer-
ring exclusively to present states resulting from past actions (see Harris 1982). At
the other end of the proposed spectrum, the passé composé in Modern French has
developed into a perfective past, exhibiting the characteristics typical of perfective

.╅ We would like to thank two anonymous reviewers and the audience at the 39th Linguistic
Symposium on Romance Languages in Tucson, Arizona, for their helpful comments on this
paper. All remaining errors are our own.
 Patrícia Amaral & Chad Howe

forms cross-linguistically – e.g. compatibility with definite past adverbials and use in
sequenced narratives (see Bybee et al. 1994). In the intervening stages of this historical
path, the periphrastic past acquires additional meanings characteristic of perfect, or
anterior, constructions (see Comrie 1976).2 In this model, the Portuguese Pretérito
Perfeito Composto (henceforth PPC) and the pretérito perfecto compuesto in Mexican
Spanish (henceforth Perfect) are situated at a stage that precedes the emergence of
the prototypical value of the Perfect, “past reference with current relevance”. Both
forms are described as referring to “[p]ast situations still ongoing in the present”
(Schwenter€ & Torres Cacoullos 2008:€ 7). The current analysis focuses primarily on
these intermediary stages.
This paper revisits the widely cited typology proposed by Harris (1982) and
Fleishman (1983) and makes two principle claims. First, the Harris-style approach
to categorizing periphrastic past constructions in Romance, though identifying the
extreme ends of a robust diachronic pattern, is too coarse-grained a generalization and
fails to capture the full range of interpretations found across the Romance varieties. In
particular, we show that the characterization of the Portuguese PPC as being function-
ally parallel to the Perfect in Mexican Spanish overlooks relevant semantic distinc-
tions between the two forms, both in synchrony and diachrony. Second, we argue that
the assumption of discrete stages of development obscures semantic details that are
important to understand the relation between the aspectual categories included in the
diachronic trajectory of periphrastic pasts in Romance.
Unlike what happens in other Romance languages, the PPC requires an itera-
tive or durative interpretation, and hence is incompatible with ‘once’ adverbials, as
shown in (1) (see Campos 1986; Giorgi & Pianesi 1997; Schmitt 2001).
(1) A Ana tem chegado atrasada *uma vez.
“Ana has been arriving late *once.”

It is unclear why the aspectual properties of the PPC should represent an archaic
stage in the development of the periphrastic past in Romance, as argued by Harris.
In fact, there is both synchronic and diachronic evidence for the availability of the
“past reference with current relevance” interpretation of the PPC. For Harris and oth-
ers (e.g. Squartini & Bertinetto 2000), this type of interpretation, characteristic of the
periphrastic past constructions found in Catalan and Peninsular Spanish, represents
a later stage in the diachronic path of the perfect. Hence, the evidence provided by the
Portuguese data runs counter to the predictions of a unidirectional model like the one

.╅ For the purposes of the current analysis, we distinguish between the aspectual category of
“perfective” and the functional category “perfect”, which does indeed display functions akin to
perfective forms in some varieties of Romance.
 Detours along the perfect path 

proposed by Harris. The meaning of the PPC offers compelling evidence against the
claim that the development of periphrastic past forms is a unified phenomenon across
Romance. Rather, it seems that the Portuguese PPC exemplifies a distinct development
from the one observed in other Romance languages.
The structure of the paper is as follows. In Section€2 we provide an overview of
the PPC in contemporary Portuguese. In Section€3 we present our proposal regarding
the diachronic development of the PPC giving rise to the iterative interpretation found
in synchrony, offering quantitative evidence from corpus data to support our claims.
Section 4 presents a comparison of the PPC and the Perfect in Mexican Spanish and
develops our argument concerning divergent paths of semantic change. We conclude
in Section€5 with a summary of the central claims.

2.â•… The Portuguese PPC in synchrony

The Portuguese PPC differs from other Perfect forms in Romance in that it denotes a
plurality of events or a continuation of a state.3 Thus, in (2), the stative predicate estar
doente “to be sick” allows for (i) an iterative reading in which Pedro has been experi-
encing periods of illness recently (with gaps during which he was not sick), and may
or may not be sick at the present, and (ii) a durative reading in which Pedro’s state of
illness started in the past and continued (uninterrupted) into the present, possibly
holding at utterance time.

(2) O Pedro tem estado doente.

“Pedro has been sick.”

While the durative reading of example (2) is also available in other varieties of
Romance, as well as for other perfects cross-linguistically (see Comrie 1976; Bybee et€al.
1994, among others), the iterative reading is not as widely attested. This reading of
the PPC is the only possible meaning with non-stative verbs; in (1), with the achieve-
ment predicate chegar “to arrive”, the PPC requires multiple instances of Ana’s arriving,
hence the incompatibility with the adverbial uma vez “once”. Crucially, in synchrony,

.╅ The multiple event reading of the PPC in contemporary Portuguese is required only in
the Indicative. Non-iterative meanings, such as an existential interpretation, are available in
embedded clauses with the verb ter in the Subjunctive, as in Espero que a Ana tenha conseguido
um bilhete para o concerto “I hope that Ana has managed to buy a ticket for the concert”. The
iterative reading of the PPC is also not required when the reference time is not the same as
utterance time (see Peres 1996; Mateus et al. 2003:142).
 Patrícia Amaral & Chad Howe

the Portuguese PPC cannot be used to refer to a relevant present state resulting from a
recent event, as shown in (3):

(3) Onde está a Ana?

“Where is Ana?”
a. Está aqui: acabou de chegar. (aspectual periphrasis)
“She’s here: she has just arrived.”
b. Está aqui: chegou agora. (simple past perfective)
“She’s here: she has arrived just now.”
c. ??Está aqui: a Ana tem chegado. (PPC)
“??She’s here: she has been arriving.”

In (3c), the use of the PPC does not provide a felicitous response to the question Where
is Ana? – i.e. it is not compatible with a resultative interpretation, as in (3a,b). This
observation stands in direct contrast with other periphrastic past forms in Romance
that do allow for non-iterative/durative meanings.
This property of the PPC is essentially aspectual and concerns the distribution of
instances of an event (or state) throughout an interval of evaluation.4 Incompatibility
with ‘once-only’ events (e.g. “to die”), as in (4a), offers a particularly striking demon-
stration of this property. However, with a plural subject, as in (4b), the PPC can be
used because the possibility of multiple dying events is made available.

(4) a. *O animal tem morrido nesta rua.

“??The animal has died repeatedly on this street.”
b. Muitos animais têm morrido nesta rua.
“Many animals have been dying on this street.”

The aspectual requirement of event iteration pertains to the temporal distribution of

the described eventuality; the events denoted by the verb must be distributed evenly
throughout the interval of evaluation. In other words, (4b) means that the dying events
of many animals did not occur simultaneously. The sentences in (5) further illustrate
this observation.

(5) a. A árvore tem florido.

“The tree has been blooming.”
b. As árvores têm florido.
“The trees have been blooming.”

.╅ Past and future perfect forms in Portuguese do not share the property of required iterative
or durative meaning with the PPC.
 Detours along the perfect path 

In both (5a) and (5b), multiple blooming events are required. With (5a), there must
be sequential occasions of blooming in which the same tree participates, precluding
the possibility of all of the events occurring at the same time. By the same token, (5b)
requires that multiple blooming events (involving multiple trees) are distributed along
the interval of evaluation, and hence the sentence cannot be used to describe a scenario
in which all the trees bloomed simultaneously.
As a result of this requirement of regular temporal distribution, the PPC cannot
be used with adverbials denoting a specific cardinality of times. We can observe this
effect with the adverbial três vezes “three times”.
(6) A Ana tem chegado atrasada *três vezes.
“Ana has been arriving late (repeatedly) *three times.”

We will return to this restriction on cardinal adverbials in Section€4, as it turns out to

be critical to our argument against the classification of the Portuguese PPC as func-
tionally parallel to the Perfect in Mexican Spanish.
Concerning the temporal properties of the PPC, the interval of evaluation must
include utterance time as its right boundary (Campos 1986; Cabredo-Hofherr et al.
2007). Hence, the PPC is not compatible with modifiers that do not include utterance
time, as in example (7), from Cabredo-Hofherr et al. (2007:4, example€26).
(7) Até agora/#Até ontem, tenho andado no mundo de cabeça levantada.
“Until now/#Until yesterday, I have been walking around the world with my
head held high.”

In what follows, we propose an account of the development of the iterative/durative

meaning of the PPC. We show that the evolution of the PPC as a marker of event
plurality does not comply with the unidirectional model of the development of peri-
phrastic past forms in Romance broadly accepted in the literature.5

3.â•… The PPC in diachrony

Corpus data provide evidence for the availability of a “past reference with current
relevance” interpretation of the PPC in the diachrony of Portuguese. In Harris’ pro-
posal, this type of interpretation is characteristic of Peninsular Spanish and Catalan
and may be followed by the last stage of the development of periphrastic past construc-
tions, the perfective interpretation (as in the passé composé in French).

.╅ For a description of the PPC as a marker of event plurality, or pluractional operator, see
Cabredo-Hofherr et al. (2007).
 Patrícia Amaral & Chad Howe

In the examples below, the verb in the PPC denotes a single event that occurred
in the recent past and whose resultant state is pragmatically relevant at the utterance
time.6 In (8), taken from the Corpus do Português (CdP, Davies & Ferreira 2006), both
the number of the NP a capitania de um dos navios and contextual information provide
evidence for a one-time event interpretation: under normal circumstances, someone is
chosen as the captain of a ship going on a specific expedition only once.
(8) Eu ey por bem que Nicolaao Jusarte, fidalgo de minha casa, a que tenho ffeito
merce da capitania de um dos navios que vao pera a India nesta armada
d’outubro, vaa no navio do Porto
“I order that Nicolau Jusarte, nobleman of my house, to whom I have awarded
the honor of being the captain of one of the ships that will go to India in the
October fleet, shall go on the ship from Oporto” (CdP, 16th century)

In example (9), the context of utterance makes it apparent that the PPC denotes a
single event whose resultant state (the food having been chosen) is under discussion.
(9) [Context: at the Nobleman’s request, the Waiter enumerates what dishes are
available and is forced to exhaustively provide the whole list of foods]
MORGADO: Está bem,€tenho€escolhido. Mande-me um pratinho de salada e dez
réis de pão. CAREIRO: Sim, senhor (.â•›.â•›.) –Abalaram-se os montes, pariram um
rato. Tanta coisa para dez réis de pão e um pratinho de salada.
“NOBLEMAN: Alright, I have chosen. Give me a little salad plate and ten réis
[old Portuguese currency] of bread.
CAREIRO: Oh well! (.╛.╛.) –The mountain has€labored and€brought forth a mouse.
All this just for ten réis of bread and a little salad plate” (CdP, 19th century)
In (8) and (9), the iterative interpretation that would be obtained for the PPC with the
same non-stative predicates in synchrony is ruled out by the context. In synchrony,
given these contextual premises, examples (8) and (9) would be semantically anoma-
lous. We believe that the non-iterative meaning in example (9), from the 19th century,
is in fact a relic of a previous diachronic stage of the ter + Past Participle construction in
Portuguese, which is retained with verbs like escolher “to choose”, dizer “to say”, and
ouvir “to hear”. It has been claimed that such forms are “frozen” due to high frequency
(Squartini 1998:€156; Detges 2000).
We take the existence of these types of examples as evidence of a stage in the
development of the ter + Past Participle construction at which the iterative reading
was not required. Such examples are left unaccounted for by Harris’ proposal, since
they are instances of the ‘prototypical’ perfects argued to emerge at a later stage than

.╅ Harre (1991:144) also discusses a possible intermediary stage of the development of the
PPC in which the form is interpreted as referring to a prior event. We believe that her under-
lying intuition is consistent with our claim about the resultative perfect interpretation.
 Detours along the perfect path 

the stage attributed to Portuguese. Hence, we argue that the development of the PPC
reveals a diachronic path that has been overlooked in the literature.
For Harris, the Portuguese PPC and the Mexican Spanish Perfect constitute
stalled stages of the development of periphrastic past forms. Instead, we maintain that
the Portuguese case represents a distinct semantic trajectory. We claim that for the
PPC the iterative interpretation arises out of the resultative interpretation (henceforth
resultative perfect),7 instead of preceding it. Perhaps more importantly, assuming dis-
cretely ordered diachronic stages obscures the semantic relations between possible
interpretations of the periphrastic past form. These interpretations may co-exist and
partially overlap during long periods of time. The emergence of the iterative meaning
presupposes that the periphrastic form entails the existence of a prior event. Crucially,
an ambiguity between a single-event and a multiple-event interpretation may arise
due to the presence of pluralizing elements in the VP. Hence, the Portuguese data are
problematic for an analysis that assumes a relation of precedence and non-overlap
between the iterative and the resultative perfect interpretations. Before presenting our
account, we briefly review the properties of the periphrastic construction that is at the
origin of the PPC.

3.1â•… The source construction of the PPC

The diachronic change undergone by the verb ter (from Latin tene re) is analogous
to the development of haver (from Latin habe re), both verbs meaning “to have” (see
Viotti 1998). At different points in their respective diachronies, both verbs expressed
meanings akin to possession, and selected for the small clause construction that is con-
sidered the source of the periphrastic past forms in Romance (Salvi 1987; de Acosta
2006). An example of the small clause construction with habe re is given in (10), from
Salvi (1987:226):
(10) habeo epistulam scriptam
have-1sg letter-fem-sg-acc written-fem-sg-acc

In (10), the past participle has a predicative function and is a complement of the direct
object of the verb habēre (i.e. letter), hence displaying morphological agreement with
this argument of the verb. There is no obligatory co-indexation between the subject of
habeo (i.e. the speaker in (10)) and the logical subject of the participle. This construc-
tion was restricted to past participles of telic verbs, in particular change of state verbs.

.â•… We adopt the definition of “resultative perfect” from Condoravdi & Deo (2008): “…the
relevant state holds at the reference time as a result of an event of the type denoted by the verb
having occurred. For instance, John has put the cake in the oven implies that the cake is now in
the oven as a result of John’s putting it there” (Condoravdi & Deo 2008:€3).
 Patrícia Amaral & Chad Howe

At least until the 16th century, both verbs haver and ter occurred in the periphrastic
construction with the past participle in Portuguese, with ter eventually becoming the
auxiliary for the PPC. Both haver and ter underwent desemanticization, although this
process is attested earlier in the distribution of Portuguese haver, reflecting the evolu-
tion of Latin habe re (cf. Mattos & Silva 1991; Viotti 1998; Cardoso & Pereira 2003).
Wigger (2004) observes that while haver occurs almost categorically with past parti-
ciples in the 15th century, this distribution is reversed by the 17th and 18th€centuries,
where ter is clearly the dominant choice in these constructions (2004:178). In our
subsequent analyses of corpus data, we focus on tokens of the ter + Past Participle
construction from the 16th century as a locus for the emergent multiple event meaning
of the PPC. These data will then be compared to tokens from the 18th century in which
the multiple event meaning of the PPC had already generalized.

3.2â•… ter across constructions

In contemporary Portuguese (as well as in Spanish and Galician), the resultative con-
struction described above is retained with the verb ter. In (11) and (12) we consider
some of its semantic properties, in comparison with the PPC. Sentence (11) denotes a
state that holds at speech time (the state of a certain door being closed). The same predi-
cate with the PPC in (12), on the other hand, denotes multiple door-closing events.

(11) Tenho a porta fechada.

have-1sg the door-fem-sg closed-fem-sg
“I have the door closed.”

(12) Tenho fechado a porta.

have-1sg close-ppart the door-fem-sg
“I have been closing the door.”

Accordingly, the two constructions differ with respect to their entailments. Compare
(13a) and (13b).

(13) a. ??Tenho a porta fechada, mas a porta não está fechada.

“??I have the door closed, but the door is not closed.”
b. Tenho fechado a porta, mas a porta não está fechada.
“I have been closing the door, but the door is not closed (now).”

In (13a) the coordination of the resultative construction with a clause that negates
the asserted state (in this case, the door being closed) yields a contradiction. This is
not the case in (13b), since the PPC denotes an eventuality that is distributed over the
interval of evaluation and is right-bounded by the speech time, but may or may not be
true at speech time. The resultative construction contributes no entailment of a prior
event denoted by the verb (although one may infer the existence of an event yielding
the resultant state), and hence is not compatible with instrumental phrases, as well as
 Detours along the perfect path 

with adverbial modifiers of the eventive predication. The PPC, on the other hand, may
co-occur with such modifiers.8 This is demonstrated in (14a) and (14b):
(14) a. ??Tenho a porta fechada com cuidado/com ajuda/rapidamente/frequentemente.
“I have the door closed carefully/with help/quickly/frequently.”
b. Tenho fechado a porta com cuidado/com ajuda/rapidamente/frequentemente.
“I have closed the door carefully/with help/quickly/frequently.”

For the purposes of explaining the diachronic development of the PPC, this distinction
is relevant for the following reason. We claim that the resultative construction is the
historical precursor of the PPC in Portuguese. The emergence of the iterative interpre-
tation of the PPC requires that the construction ter + Past Participle entails the prior
occurrence of the event denoted by the verb, as the PPC in synchrony denotes regular
event iteration over a time interval. As the prior occurrence of the event denoted by the
verb becomes part of the encoded meaning of the form, we expect to see an expansion
in the aspectual verb classes that may occur in the construction: the construction is no
longer restricted to telic verbs, as was the case in the purely stative resultative construc-
tion. We argue that the aspectual shift from the entailment of existence of a single prior
event to the entailment of existence of multiple prior events was induced by contexts
in which the interpretation of the VP was compatible with both a single and a multiple
event interpretation. In what follows, we will offer corpus-based evidence for these
claims regarding the development of the ter + Past Participle construction.

3.3â•… The diachronic path to the iterative interpretation

The two constructions introduced in the previous section are exemplified in (15) and
(16): (15) is an instance of the resultative construction with a stative interpretation,
while (16), from the Tycho Brahe Parsed Corpus of Historical Portuguese (TB), exempli-
fies the PPC:
(15) enfreão potentissimos Reys, tem sogeitas varias nações & resistem ha muitos
annos a fortes inimigos
“…[they] face very powerful kings, [they] have subjected several nations,
and have resisted [lit. resist-pres] for many years to strong enemies”
 (CdP, 17th€century)

.╅ By the same token, an agentive phrase that denotes the cause of the event is only compat-
ible with the PPC and not with the resultative construction:
(i) ??Tenho a porta fechada com o vento. (Resultative)
“??I have the door closed with the wind.”
(ii) A porta tem-se fechado com o vento. (PPC)
The door has-se close-ppart with the wind
“The wind has been closing the door.”
 Patrícia Amaral & Chad Howe

These constructions are distinct on both semantic and morphosyntactic grounds. In

(15), the participial adjective sogeitas agrees in gender and number with the object NP
várias nações. Note that the construction ter + Past Participle is coordinated with verbs
in the Present tense (i.e. enfrentam and resistem) that denote present states. In (16),
on the other hand, there is overt non-agreement between the participle feito and the
object NP abomináveis estragos e homicídios:
(16) Por ouro tem a ira feito abomináveis estragos e homicídios no mundo
“For the sake of gold, rage has inflicted atrocious damages and murder
on the world” (TB, 16th century)

The diachronic corpora provide evidence for yet another type of construction with
ter + Past Participle:
(17) Eu tenho recebido tanto contentamento com vossas cartas, pelas quais tenho visto,
e pelas obras sabido, como me tendes bem servido
“I have received so much satisfaction with your letters, by which I have seen,
and by the works known, how you have served me well.” (TB, 16th century)

(18) Com isto tenho dito do sal o que me preguntastes

“With this I have said about the salt what you have asked me”
 (TB, 16th century)

On the basis of morphosyntactic criteria, it is not possible to classify (17) and (18)
as instances of either the resultative construction (as in (15)) or the PPC (as in (16)).
In (17) and (18), the participle is [masc.sg.] and hence could agree with the comple-
ment of the verb, as is the case in the resultative construction. However, (17) and (18)
contain atelic verbs, which are not found in the resultative construction. Note also the
instrumental phrase com vossas cartas “with the letters from you”, not compatible with
the stative resultative construction, hence confirming the emergent eventive interpre-
tation of the ter + Past Participle form.
Table 1 below provides an overview of the different morphosyntactic patterns
attested in the 16th and 18th centuries data from the TB corpus. All tokens of ter€+ Past
Participle constructions were extracted from the corpus and coded according to one of
three categories based on morphosyntactic features:9 (a) Structurally PPC€– i.e. overt
non-agreement between the participle and the complement (as in (16)); (b) Structur-
ally Resultative – i.e. overt agreement between the participle and the complement (as

.╅ The categories used in classifying the tokens for quantitative analysis were chosen as an
measurable reflex of the distribution of ter + Past Participle constructions. They do not pre-
clude the possibility, for example, of a structurally PPC token expressing a non-iterative or
durative meaning (see Cardoso & Pereira 2003). Cases in which a Past Participle preceded ter
were also extracted.
 Detours along the perfect path 

in example€15); and (c) Ambiguous (see examples€17 and 18).10 The results shown in
Table€1 suggest considerable expansion of the PPC tokens, 24.1% in the 16th century to
78.4% in the 18th. In the subsequent discussion of corpus results, we will concentrate
specifically on the distributional patterns observed with the tokens labelled as PPC.

Table 1.╇ Overall Distribution of Construction Types by Century (from TB Corpus)

Construction Type 16th Century 18th Century

PPC 24.1% (N=68) 78.4% (N=410)

Ambiguous 61.7% (N=174) 19.9% (N=104)
Resultative 16.3% (N=46) 1.7% (N=9)
Total 288 523

Our claim regarding the proposed path of development for the PPC makes two
important predictions. First, as the iterative meaning of the PPC becomes more general,
we should observe patterns of expansion across verb types such that predicates ame-
nable to event duration or repetition are favored. Specifically, we predict that atelic
predicates will represent a significant indicator for expansion of the PPC tokens. Further-
more, we argue that the iterative interpretation was induced in contexts in which the
resultative construction had acquired an eventive entailment and this entailment was
compatible with both a single and a multiple event interpretation. Qualitative analysis
of the corpus data suggests that certain features of the nominal arguments of the verb
may have given rise to the aspectual profile of the PPC described above, namely the
temporal distribution of the events of the type denoted by the verb through the inter-
val of evaluation. We expect that the analysis of the corpus data will reveal specific
patterns of interaction between the PPC constructions and the semantic plurality of
the verbal complements.
Both predictions are borne out. Let us start with the first prediction. Table 2 rep-
resents the coding for each PPC token in the 16th and 18th century samples according
to its Aktionsart, following the lexical verb classes proposed by Vendler (1967).

Table 2.╇ Distribution of PPC Tokens by Aktionsart (from TB Corpus)

Aktionsart 16th Century 18th Century

Stative 0%â•… (N=0) 24.6% (N=101)

Activity 27.9% (N=19) 32.4% (N=133)
Accomplishment 23.5% (N=16) 9.8% (N=40)
Achievement 48.5% (N=33) 33.2% (N=136)

.╅ Tokens with intransitive verbs were labelled as PPCs given that these predicates are not
compatible with the resultative construction.
 Patrícia Amaral & Chad Howe

Table 2 demonstrates the expansion of the PPC tokens among the group of atelic
predicates,11 with statives representing almost 25% of all tokens in the 18th century data
despite being non-existent in the 16th century sample.12 The change in the percentage
of stative and activity predicates from the 16th century (28%) to the 18th€century (57.1%)
and the concomitant decline in the distribution of telic predicates (i.e. from 72% to 43%),
taken together, are significant trends in the development of the PPC (χ2 =€19.868;
p€< 0.00001). The patterns observed in Table 2 are consistent with our claim about the
emergence of iterative/durative meaning with the PPC, which would, we argue, result
in increased compatibility with atelic predicates.
With respect to the second of our predictions, we observe object plurality as a
precipitating factor in the emergence of the multiple event meaning, specifically the
iterative interpretation that is obtained with telic predicates. Our claim concerns the
interaction between the ter + Past Participle construction and potentially pluraliz-
ing elements in the context. For example, in (19), the multiple event interpretation is
favored both by the meaning of the direct object of the verb and by the plural recipient
argument. The clitic pronoun o “it” refers to a proposition expressed in the previous
discourse: the assertion that the writer has a gift for drawing. This propositional object
is compatible both with a single and a plural interpretation of the eventuality of show-
ing. By the same token, one may infer either a single event of showing to all the men-
tioned individuals or (more plausibly) multiple events of showing, each corresponding
to a different individual to whom the artist’s talent was shown.
(19) porque aos serenissimos ifantes e aos môres senhores d’esta côrte o tenho eu, sendo
moço, mostrado
“since to the most serene princes and to the most important lords of this court
I have shown this, while I was a young man” (TB, 16th century)

Similarly, the mass noun contentamento in the NP complement of the verb in (17)
above, though morphologically singular, allows for both an interpretation according
to which there was a single satisfaction-receiving event or multiple events of this type.
Note that, pragmatically, the most natural interpretation would involve a scenario in
which the author’s incremental increase in satisfaction was the result of a series of letter-
receiving events. This interpretation is strengthened by the adjunct com vossas cartas
“with your letters”, which distributes the “portions” of satisfaction over non-overlapping

.╅ Note that we have not quantified other structural elements that may have had an ateli-
cizing effect on these tokens – e.g. negation and adverbials. Nevertheless, the omission of these
elements should not detract from the effects observed in Table 2.
.â•… To account for the effects of high frequency verb tokens (e.g. ser ‘to be’), we further cal-
culated the type frequencies for the PPC tokens. The results were parallel to those presented
for the raw tokens provided in Table 2.
 Detours along the perfect path 

occasions of receiving individual letters. While it is possible that the author received
multiple letters at the same time, it seems that a more plausible scenario is the occur-
rence of a series of letter-receiving events distributed over some interval of time. The
ambiguity arising in the verbal complex stems from the semantic properties of the
noun in the object NP. The homogeneity of mass nouns allows for the type of event
individuation that could be at the origin of the iterative interpretation: since “por-
tions” of satisfaction are no different in nature from “a sum” of satisfaction, a plural
interpretation of this eventuality of satisfaction-receiving amounts to a distribution of
sub-events of the same type throughout the interval of evaluation.
In order to demonstrate this effect in the data, we coded all of the extracted
tokens from the 16th and 18th centuries according to the plurality of the complements
(excluding adjuncts; see also Schwenter & Torres Cacoullos 2008). Table 3 presents
the distribution of complement types in the PPC tokens. The specific categories are
defined as follows: Singular: tokens with singular complements; Plural: complements
with overt plural morphology (like (16)); Quantifier: complements that include a quan-
tifier (example€20 below, quantifier underlined); Mass: complements that include mass
nouns (as in example€17); Abstract: tokens with propositional or anaphoric comple-
ments (like Example€19); and Intransitive: tokens that do not take a complement.
(20) E de tudo o que me tendes sprito que la fizestes…
“and from everything which you have written to me that you did there…”
 (TB, 16th century)

In the PPC tokens from the 16th century, there is a clear preponderance of comple-
ment types that, according to our claim, would precipitate the type of ambiguity that
gives rise to the multiple event reading.

Table 3.╇ Distribution of PPC Tokens by Complement Type (from TB Corpus)

Complement Type 16th Century 18th Century

Singular 13.2% (N=9) 10.8% (N=44)

Plural 45.6% (N=31) 21.2% (N=86)
Quantifier 20.6% (N=14) 13.8% (N=56)
Mass 10.3% (N=7) 4.9% (N=20)
Abstract 4.4% (N=3) 5.4% (N=22)
Intransitive 5.9% (N=4) 43.8% (N=178)
Total 68 406

Over 75% (N=52) of the PPC tokens from the 16th century occurred with either
plural complements, quantifiers, or mass nouns, all of which, we argue, play a role in
the transfer of plurality from the nominal domain to the verbal one. The 16th century
data also reveal a clear disfavoring of intransitive verbs; at this stage in the diachrony
 Patrícia Amaral & Chad Howe

of the PPC we would still expect to observe the erstwhile structural conditions of the
resultative source construction.
The distribution of complement types changes considerably in the 18th century
data with decreased proportions of pluralizing complements and a significant increase
in the compatibility with intransitives (6% to 44%). These results suggest that by the
18th century the iterative meaning of the PPC had generalized and consequently no
longer necessitated an overtly plural complement. Since this meaning was merely
emergent in the 16th century, it depended on collocational effects, i.e. on the interaction
with pluralizing elements within the VP (for semantic accounts of interaction between
plurality in the nominal and the verbal domains see Cusic 1981; Van Geenhoven 2004).
In fact, we find a higher percentage of such complements in this century. In other
words, potential event plurality in the 16th century PPC was still largely a function of
the plurality of the complements; by the 18th century, it had become semanticized as
part of the conventional meaning of the PPC.
We now turn to a brief discussion of the PPC and its purported similarity with the
Mexican Spanish Perfect.

4.â•… The perfect path revisited

As argued in Sections€1 and 2, the claimed functional parallels between the Portuguese
PPC and the Perfect in Mexican Spanish are limited to the observation that the two
structures encode some notion of continuation of a state in the present. This observa-
tion, however, fails to distinguish the two forms in terms of their aspectual properties.
If we consider modification with specific cardinal adverbials, for example, the Mexican
Perfect does not exhibit the same behavior as the PPC. Compare the example from
Mexican Spanish in (21a) (from Lope Blanch 1976) with its Portuguese ‘counterpart’
in (22).

(21) a. Sí; he ido dos ocasiones [a su tierra].

“Yes. I have gone on two occasions [to his hometown].”
b. Sí; he ido una vez a su tierra.
“Yes. I have gone once to his hometown.”

(22) Eu tenho ido *duas vezes [à terra dele].

“I have been going *twice [to his hometown].”

The comparison between (21a) and (22) shows that, unlike the PPC, the Perfect in
Mexican Spanish may co-occur with specific cardinal adverbials. The comparison
between (21) and (22) suggests that event plurality is in fact not part of the core mean-
ing of the Mexican Perfect, as demonstrated by the acceptability of una vez in (21b).
 Detours along the perfect path 

For Schwenter & Torres Cacoullos, the Perfect in Mexican Spanish is a “continuative
perfect or a perfect of persistent situation” (2008:6, emphasis added; see also Comrie
1976; Dahl 1985). Similarly, according to Thibault (2000:98), the Spanish haber Perfect
in some varieties has a plural meaning that developed as function of frequent col-
location with indicadores iterativos “iterative indicators” such as algunas veces “some-
times” in (23) (see also Moreno de Alba 1978 for comments on the Perfect in Mexican
Spanish).

(23) he visto algunas veces que soldados pláticos se burlan de los nuevamente venidos
de Spaña
“I have seen sometimes that career soldiers make fun of the newly-arrived guys
from Spain” (Diálogo de la lengua, cited in Thibault 2000:€98, 16th century)

Thibault further comments that “[e]l uso de los indicadores aspectuales … resultaría
imposible en portugués … el PC [haber perfect in Spanish] acompaña una iteración,
no la determina” (2000:99). This mention of indicadores aspectuales refers to the type
of adverbial modification demonstrated in (21a) and (22). In terms of their respective
developments, it is only with the PPC that we observe semanticization of the multiple
event meaning; with the Perfect in Mexican Spanish, event iteration continues to be
a function of the interaction of the haber + Past Participle construction with plural
arguments and certain adverbials. Thus, Thibault’s comments are appropriate in that
the iterative or durative interpretation with the Perfect in Mexican Spanish is not part
of its core meaning; it is made available contextually.
In their analysis of the distribution of the Perfect vis-à-vis the simple perfective
past (pretérito) in Mexican Spanish, Schwenter & Torres Cacoullos provide quantita-
tive evidence for a number of factors that favor the use of the Perfect, one of which is
Aktionsart. Perfects are favored with durative predicates (probability =.52) and disfa-
vored with punctual ones (.39) (2008:20). Of the remaining significant factors, both
interaction with frequency adverbials and co-occurrence with plural complements
have a greater effect on Perfect usage than Aktionsart, suggesting that the claimed
‘plural’ meaning of the Perfect in Mexican Spanish (see Moreno de Alba 1978; Thibault
2000) is still largely a function of the plurality of co-occurring elements. This observa-
tion, coupled with the fact that in Portuguese the functional overlap between the PPC
and the simple past (Pretérito Perfeito Simples) is, at the very least, qualitatively distinct
from the simple vs. periphrastic past distinction in Spanish, provides further evidence
for our claim that the PPC and the Perfect in Mexican Spanish represent divergent
developments.
In sum, a multiple event meaning is available with two structurally similar forms
in two different Romance varieties. With the Perfect in Mexican Spanish, this mean-
ing is part of an inventory of contextually motivated interpretations associated with a
perfect (see Comrie 1976) but is not part of the core semantics of the form. Conversely,
 Patrícia Amaral & Chad Howe

event plurality, as described in Section€2, is an indelible component of the meaning of

the PPC, one that is not subject to the type of contextual factors that license plural or
continuative meanings with the haber perfect in Spanish. We maintain that the dis-
tinction between the PPC and the Mexican Spanish pretérito perfecto compuesto sug-
gests divergent paths of semantic development. The consequence of this claim in the
case of the PPC is that this construction seems to fall outside the scope of the typology
of periphrastic pasts as described by Harris (1982).

5.â•… Conclusions

We have demonstrated that the emergence of the iterative meaning of the Portuguese
PPC was favored by ambiguous contexts in which the semantic properties of the nomi-
nal arguments of the verb license the interpretation of event iteration. Using different
patterns of distribution as a diagnostic, we have offered quantitative corpus evidence
of the development of the multiple event meaning from the resultative source con-
struction with an eventive entailment. The subsequent semanticization of the multiple
event interpretation in these cases has given rise to the aspectual properties of the PPC
observed in synchrony.
Further comparison between the PPC and the Perfect in Mexican Spanish suggests
that the plural meaning typically ascribed to these structures is too coarse-grained.
Our synchronic and diachronic analyses offer compelling evidence of the potential for
strictly categorical treatments of semantic change to obscure deviations among indi-
vidual languages and varieties. Indeed, these deviations may turn out to reveal some-
thing more significant about the nature of pathways of grammaticalization, offering a
more nuanced account of the multi-faceted process of semantic change.

References

Bybee, Joan, Revere Perkins & William Pagliuca. 1994. The Evolution of Grammar: Tense, aspect,
and modality in the languages of the world. Chicago: University of Chicago Press.
Cabredo-Hofherr, Patrícia, Brenda Laca & Sandra de Carvalho. 2007. “When “Perfect” means
“Plural”: the Present Perfect in NE Brazilian Portuguese.” (Available at http://web.uni-frankfurt.
de/fb10/rathert/forschung/pdfstense/laca.pdf).
Campos, Maria Henriqueta C. 1986. “L’opposition du portugais Pretérito Perfeito Simples-Pretérito
Perfeito Composto: un cas singulier dans l’ensemble des langues romanes”. Morphosyntaxe
des Langues Romanes. Actes et du XVIIe Congrès International de Linguistique et Philologie
Romanes (Aix-en-Provence), 411–22.
Cardoso, Adriana & Susana Pereira. 2003. “Contributos para o estudo da emergência do tempo
composto em Português”. Revista da ABRALIN 2: 159–181.
 Detours along the perfect path 

Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press.

Condoravdi, Cleo & Ashwini Deo. 2008. “Aspect Shifts in Indo-Aryan”. Paper presented at the
Workshop on Formal Approaches to the Relation of Tense, Aspect and Modality, Seoul,
Korea, July 2008.
Cusic, David. 1981. Plurality and Aspect. Ph.D. Dissertation, Stanford University.
Dahl, Östen. 1985. Tense and Aspect Systems. Oxford: Blackwell.
Davies, Mark & Michael Ferreira. 2006. Corpus do Português (45 million words, 1300s–1900s).
Available at http://www.corpusdoportugues.org.
de Acosta, Diego. 2006. HAVE + PERFECT PARTICIPLE in Romance and English: Synchrony
and diachrony. Ph.D. dissertation, Cornell University.
Detges, Ulrich. 2000. “Time and Truth: The Grammaticalization of Resultatives and Perfects
within a Theory of Subjectification”. Studies in Language 24.345–377.
Fleischman, Suzanne. 1983. “From Pragmatics to Grammar: Diachronic reflections on complex
pasts and futures in Romance”. Lingua 60.183–214.
Giorgi, Alessandra & Fabio Pianesi. 1997. Tense and Aspect: From semantics to morphosyntax.
New York: Oxford University Press.
Harre, Catherine. 1991. Tener + Past Participle: A case study in linguistic description. London:
Routledge.
Harris, Martin. 1982. “The ‘Past Simple’ and ‘Present Perfect’ in Romance”. Studies in the Romance
Verb ed. by Martin Harris & Nigel Vincent, 42–70. London: Croom Helm.
Lope Blanch, Juan M. 1976. El habla popular de México: materiales para su estudio. México:
UNAM.
Mateus, Maria H. M., Ana Maria Brito, Inês Duarte & Isabel Hub Faria. 2003. Gramática da
Língua Portuguesa. Lisboa: Caminho.
Mattos e Silva, Rosa V. 1991. “Caminhos de mudanças sintático-semânticas no português antigo”.
Cadernos de Estudos Linguísticos 20.59–74.
Moreno de Alba, José. 1978. Valores de formas verbales en el español de México. México: UNAM.
Peres, João. 1996. “Sobre a semântica das construções perfectivas do Português”. Congresso
Internacional sobre o Português V. 2.33–58.
Salvi, Giampaolo. 1987. “Syntactic Restructuring in the Evolution of Romance Auxiliaries”.
Historical Development of Auxiliaries ed. by Martin Harris & Paolo Ramat, 225–36.
Amsterdam: Benjamins.
Schmitt, Cristina. 2001. “Cross-linguistic Variation and the Present Perfect: The case of Portuguese”.
Natural Language and Linguistic Theory 19.403–453.
Schwenter, Scott. A. & Rena Torres Cacoullos. 2008. “Defaults and Indeterminacy in Temporal
Grammaticalization: The ‘perfect’ road to perfective”. Language Variation and Change
20.1–39.
Squartini, Mario. 1998. Verbal periphrasis in Romance. Aspect, actionality, and grammaticalization.
Berlin, New York: Mouton de Gruyter.
Squartini, Mario & Pier Marco Bertinetto. 2000. “The simple and compound past in Romance
languages”. Tense and aspect in the languages of Europe ed. by Östen Dahl, 403–439. Berlin/
New York: Mouton de Gruyter.
Thibault, André. 2000. Perfecto simple y perfecto compuesto en español preclásico. Tübingen:
M.N. Verlag.
Tycho Brahe Parsed Corpus of Historical Portuguese. (approx. 2.3 million words, 1400s–1800s).
Available at http://www.tycho.iel.unicamp.br.
 Patrícia Amaral & Chad Howe

Van Geenhoven, Veerle. 2004. “For-adverbials, Frequentive Aspect, and Pluractionality”. Natural
Language Semantics 12.135–90.
Vendler, Zeno. 1967. Linguistics in Philosophy. New York: Cornell University Press.
Viotti, Evani. 1998. “Uma história sobre “ter” e “haver””. Cadernos de Estudos Linguísticos 34.41–50.
Wigger, Lars-Georg. 2004. Die Entwicklungsgeschichte der romanischen Vergangenheitstempora
am Beispiel des Pretérito Perfeito Composto im Portugiesischen. Ph.D. Dissertation, Universität
Tübingen.
 Grammaticalization of commencer/cominciare
“to begin” in French and Italian*

Jason Brazeal
University of Texas at Austin

This study presents a synchronic and diachronic analysis of the French and
Italian aspectualizers commencer and cominciare from the perspective of
grammaticalization theory. They show properties of both full verbs and
auxiliaries, and thus have traditionally proven difficult to analyze. It is argued
that these aspectualizers are best explained by taking into account their historical
development from Latin initiare “to initiate”. Synchronic and diachronic data
are presented in order to locate them on the Verb-to-TAM chain, a continuum
ranging from full lexical verbs on one end to purely grammatical tense, aspect,
and mood affixes on the other. Their mixed semantic and syntactic properties are
due to their position on the continuum between lexical verbs and auxiliaries.
The differences between commencer and cominciare with respect to their degree
of grammaticalization are also discussed.

Key words:╇ grammaticalization; aspectualizers; auxiliary; commencer; cominciare

1.â•… Introduction

Aspectualizers are predicates like begin, continue, and stop, which convey informa-
tion about aspect, the internal temporal organization of a situation (Comrie 1976).
They add nuances about an event or state such as whether it is just starting, already in
progress, or coming to a close. These predicates have traditionally proven difficult to
analyze, given that they exhibit properties of both full verbs and auxiliaries. Work in

*I would like to thank Cinzia Russi, Catherine Léger, Antonia Greb, the audience of the 39th
Linguistic Symposium on Romance Languages, the Department of French and Italian and the
Graduate School of the University of Texas at Austin, and three anonymous reviewers for their
assistance, comments, and suggestions. Any errors that remain are my own.
 Jason Brazeal

grammaticalization theory has suggested that their syntactic and semantic properties
are a result of their evolution from full lexical verbs (Heine 1993). The grammaticaliza-
tion of aspectualizers and other auxiliary-like structures in Romance languages has
been examined, but detailed studies on particular aspectualizers are lacking (Lamiroy
1999). The current study presents an in-depth analysis of the French and Italian
aspectualizers commencer/cominciare “to begin” from the perspective of grammati-
calization theory. Grammaticalization is a process of language change whereby lexical
items take on grammatical functions or grammatical items become more grammati-
cal (Meillet 1912; Kuryłowicz 1965). It is argued that the properties of commencer/
cominciare are best explained with recourse to diachronic data; by examining their
historical development, the confusion regarding their status as full verbs or auxiliaries
is resolved. Grammaticalization theory allows us to conceive of these two categories
as sets of points on a continuum ranging from full lexical verbs on one end to purely
grammatical affixes on the other end. It will be shown that the mixed behavior of
�commencer/cominciare is due to their intermediate position on the continuum between
full verbs and auxiliaries.
The paper will proceed as follows: section one introduces data on the mixed
properties of commencer/cominciare, section two summarizes the previous research,
section three gives a brief overview of grammaticalization theory and the development
of auxiliary-type structures, section four presents a synchronic analysis of commencer/
cominciare, section five examines their diachronic development from Latin, section
six discusses the differences between commencer and cominciare, and section seven
concludes and offers suggestions for future research.

2.â•… Data

The aspectualizers commencer/cominciare “to begin” have syntactic and semantic

properties that make them difficult to classify as full verbs or auxiliaries. Syntactically,
they often behave like full verbs, constituting the main predicate of the sentence with
or without a direct object NP. Their transitive and intransitive uses are exemplified
in€(1) and (2).
(1) a. Jean commence le livre.
b. Gianni comincia il libro.
“John is beginning the book.”

(2) a. La réunion commence à 8 heures.

b. La riunione comincia alle 8.
“The meeting begins at 8.”
 Grammaticalization of commencer/cominciare 

Aspectualizers may also take infinitival complements introduced by a preposition, as

in (3).
(3) a. Marie commence à écrire le livre.
b. Maria comincia a scrivere il libro.
“Mary is beginning to write the book.”

Semantically, the aspectualizer and the infinitival complement in (3) constitute a single
event, ‘begin to X’ (Rochette 1999). A single event interpretation is not possible with
other verbs which take infinitival complements. Consider the following examples:
(4) a. Hier, Jean a promis d’écrire une lettre aujourd’hui.
b. Ieri, Gianni ha promesso di scrivere una lettera oggi.
“Yesterday, John promised to write a letter today.”
(5) a. *Hier, Jean a commencé à écrire une lettre aujourd’hui.
b. *Ieri, Gianni ha cominciato a scrivere una lettera oggi.
“Yesterday, John began to write a letter today.”

In (4) the presence of distinct time adverbials in the matrix and infinitival clauses is
acceptable, while in (5) it renders the sentence ungrammatical. This is evidence that
constructions with aspectualizers like (5) consist of a single proposition at the level
of semantic representation (Rochette 1999; Newmeyer 1975). With regards to this
property, aspectualizers resemble the auxiliaries avoir/avere in (6).
(6) a. *Hier, Jean a écrit une lettre aujourd’hui.
b. *Ieri, Gianni ha scritto una lettera oggi.
“Yesterday, John wrote a letter today.”

These auxiliaries also occur with a nonfinite verb form (past participle) and express a
single event, ‘have X-ed’.
From these data it is clear that commencer/cominciare share some properties with
full verbs (the transitive and intransitive uses in (1) and (2)), and some properties
with auxiliaries (the single event interpretation in combination with a nonfinite form
in (3)). We will now review some previous analyses of aspectualizers that attempt to
account for these mixed properties.

3.â•… Previous accounts

There is little agreement in the literature on the analysis of commencer/cominciare.

Researchers often focus on the underlying syntax of aspectualizers and attempt to clas-
sify them as raising or control verbs based on their semantic and syntactic properties.
 Jason Brazeal

An example of this type of analysis is Perlmutter (1970), who posits two different sub-
categorization frames for English begin. He considers the cases with infinitival com-
plements to be raising structures, whose surface subject originates in the embedded
infinitival clause and is then raised to the main clause subject position. The transitive
and intransitive uses of begin are control structures since agentive nominalization,
embedding under other control verbs, and imperatives with begin are all permitted.
Perlmutter (1970) maintains that begin enters into two different deep structures, one
control and one raising, but leaves open the question of whether there are actually
two homophonous verbs begin, although the title of his paper suggests that there are.
Ruwet (1972) provides a similar analysis for commencer, presenting much of the same
evidence in French as Perlmutter (1970) does in English.
Examples of analyses which posit a single deep structure for begin are Newmeyer
(1975) and Rochette (1999). They consider all instances of begin and commencer to be
raising structures. Newmeyer (1975) accounts for the different surface constructions
in which begin occurs by deriving them all from a deep raising structure through several
transformational operations, differing from Perlmutter’s (1970) analysis in both the
type and the number of transformations in the derivation. Rochette (1999) highlights
the mechanism of semantic selection to explain the various surface structures in which
begin and commencer occur. She bases her analysis on Pustejovsky’s (1991) model of
event semantics, claiming that aspectualizers select the natural class of event-types
whose head constitutes a process, corresponding to Vendler’s (1957) activities and
accomplishments.1 Rochette (1999) proposes that aspectualizers are raising verbs that
select an activity or accomplishment, which may be realized as a nominal or verbal
projection and may be ‘concealed’ as in (1). The interpretation of this concealed argu-
ment is determined by the context, so (1) may be interpreted, for example, as (7). An
accomplishment interpretation like read a book is possible, but not a state or achieve-
ment interpretation like have a book or find a book.

(7) a. Jean commence {à lire/*à avoir/*à trouver} le livre.

b. Gianni comincia {a leggere/*ad avere/*a trovare} il libro.
“John is beginning {to read/*to have/*to find} the book.”

.â•… Rochette (1999:159) states that “aspectual verbs are characterized as verbs that semanti-
cally select a process”, but Pustejovsky’s (1991) ‘process’ corresponds only to Vendler’s (1957)
activities. When decomposed into their subeventual structure, however, accomplishments
consist of an activity that brings about a change (see Pustejovsky 1991 for a detailed analysis).
In this quote and throughout her paper, Rochette’s (1999) use of the terminology ‘process’ is
misleading; she uses ‘process’ to refer to event types that contain an activity at any level of their
semantic representation, thus encompassing Vendler’s (1957) activities and accomplishments.
 Grammaticalization of commencer/cominciare 

As with prototypical raising verbs like seem, the deep subject of the infinitival comple-
ment raises to the subject position of the aspectualizer in surface structure. In addi-
tion, example (2) illustrates that when the process is a nominal, it must be realized
in subject position. While Rochette does not go into great detail about this specific
example, she does suggest that an “implicit predicate such as to take place” may be
present (1999:162). It is not clear from her analysis exactly how the intransitive uses of
these aspectualizers are derived. She also leaves open the question of representing the
concealed process in the syntax.
Another important account of aspectualizers is Lamiroy (1987); she proposes a
semantic hierarchy (8a) with a corresponding syntactic hierarchy (8b) to explain the
facts of spatio-temporal expressions in French.

(8) a. space < aspect < serialization < tense (Lamiroy 1987:289)
b. control < raising

A restricted class of motion verbs (monter, aller, sortir, etc.) is located on the left side of
the hierarchy; they indicate motion in space and their infinitival complements describe
“an action that will take place where and when the movement has ended” (Lamiroy
1987:289). The future proche usage of aller is located on the right side of the hierarchy;
in this usage, aller indicates that its infinitival complement will be realized in the future.
The link between the semantic and syntactic hierarchies is explained by the selectional
restrictions on the subject. For example, when used as a motion verb, the subject of
aller is restricted to an animate NP, a property typical of control verbs. As a marker of
future tense, there are no semantic restrictions on the subject of aller, which is
�chara�c�teristic of raising structures. Lamiroy (1987) accounts for the mixed behav-
ior of aspectualizers by locating them at a intermediate position on the semantic and
syntactic hierarchies in (8). Her analysis has the advantage of embracing a continuum
of structures; as she points out, it explains the fact that aspectualizers show contradic-
tory properties of both raising and control structures. The idea that the constructions
in which aspectualizers occur represent a continuum is, as we shall soon see, one of
the key points of the grammaticalization analysis. Before presenting the analysis, we
briefly examine the basic tenets of grammaticalization theory.

4.â•… Grammaticalization theory

From a diachronic perspective, grammaticalization is a process by which new grammati�

cal elements arise as they evolve from full lexical items or existing grammatical elements
take on more grammatical functions (Meillet 1912; Kuryłowicz 1965). Synchronically,
 Jason Brazeal

grammaticalization theory provides a useful framework for evaluating membership

in grammatical categories and ordering the different subcategories thereof (Lehmann
1995). Although this panchronic view of grammaticalization is rejected by some
researchers (e.g. Newmeyer 1998) on the grounds that diachronic statements have no
place in synchronic descriptions, I claim that a panchronic approach which takes into
account the diachronic development of aspectualizers leads to a better understanding
of their mixed properties.
Heine (1993) discusses the type of grammaticalization that is most relevant for
aspectualizers, the evolution of lexical verbs into markers of tense, aspect, and mood
(TAM). These grammatical notions are complex and abstract, and evidence from many
typologically different languages has shown that they almost always develop from sim-
pler lexical structures called ‘event schemas’ (Heine 1993). An example of an event
schema is the motion schema, as represented in (9).
(9) Motion schema (Heine 1993:31)
‘X moves to/from Y’

The motion schema in (9) contains an agent X, a locative element Y, and a verb of motion
(e.g. come, go, walk, pass, etc.). The lexical source verb in the event schema (moves
in€(9)) gradually loses properties common to full verbs and develops into an auxiliary,
or in more advanced stages of grammaticalization, an affix, while the complement Y
comes to be reanalyzed as the main verb. Event schemas provide the source for the
development of TAM markers; we now turn to the various stages through which lexi-
cal verbs pass on their way to becoming TAM markers.
The development of a TAM marker from a lexical verb is a gradual process
whereby the verb loses many of the properties typical of a full lexical verb and takes on
characteristics of a grammatical marker. The changes are not abrupt; a form undergoes
many gradual transitions during the process of grammaticalization. These transitions
in each linguistic domain are similar across many unrelated languages, which has led
researchers to posit a common pathway of evolution for these grammatical markers,
the Verb-to-TAM chain (Heine 1993). Table 1 outlines six stages in the grammaticaliza-
tion of TAM markers from lexical verbs.

Table 1.╇ Verb-to-TAM chain (Heine 1993:58)

lexical grammatical

Overall Stage: A B C D E F G

Desemanticization I II III
Decategorialization I II III IV V
Cliticization I II III
Erosion I II III
 Grammaticalization of commencer/cominciare 

The overall stages of grammaticalization in the Verb-to-TAM chain (capital letters

A–G) are not meant to be discrete, but rather represent sets of points on a continuum.
Each stage in the overall grammaticalization of an item is characterized by changes
in linguistic parameters at the semantic, morphosyntactic, morphophonological,
and/or phonetic levels (Roman numerals I, II, etc.). The changes in each domain are
manifested as dynamic processes of ‘desemanticization’, ‘decategorialization’, ‘cliticiza-
tion’, and ‘phonetic erosion’, respectively, all of which work in tandem to reduce the
autonomy of the item (Heine 1993; Lehmann 1995). Desemanticization, or semantic
‘bleaching’, is a reduction of the lexical content of a linguistic unit accompanied by a
gain in grammatical function. Decategorialization refers to the change from a primary
lexical category such as noun or verb to a secondary category such as pronoun or adpo-
sition. This process entails a loss of the morphosyntactic privileges typical of primary
categories, such as the ability to inflect in all cases or tenses. Cliticization represents a
further reduction of syntactic autonomy and eventual fusion to a host. Cliticization is
often accompanied by phonetic reduction, but these two processes usually apply only
at the later stages of grammaticalization. There is no evidence that the aspectualizers
commencer/cominciare have undergone cliticization or erosion. They are not fused to
other elements, and their phonetic forms have changed very little since the Old French
and Old Italian periods. Given the lack of evidence for these processes, our analysis
will concentrate on the earlier stages, A–D, examining in detail the transitions between
each stage.

5.â•… Grammaticalization of commencer/cominciare: Synchronic analysis

As shown in Table 1, the first changes to the lexical verb involve desemanticization
(Lehmann’s (1995) semantic ‘attrition’). The event schemas upon which grammaticaliza-
tion operates lose their connection to the concrete, human world and become more
abstract, eventually coming to express the grammatical notions of TAM. Desemantici-
zation may affect the entire construction undergoing grammaticalization, not just the
lexical verb. In the source construction, the subject X is typically human, the Y element
is a concrete object or location, and the verb expresses a lexical concept. During the
passage from stage A to stage B, the Y element comes to express a dynamic situation,
often appearing as a nonfinite verb form (Heine 1993). Thus, in addition to the loss
of lexical meaning in the verb at the expense of grammatical function, the semantic
properties of its arguments also tend to change. Their ability to take non-human sub-
jects and infinitival complements suggests that commencer/cominciare have advanced
beyond the lexical verb stage A (see (2) and (3)).
The transition from stage B to stage C is characterized by more semantic loss.
�Consequently, the selectional restrictions on the subject tend to be eliminated, and
 Jason Brazeal

the lexical verb comes to express some notion of TAM (Heine 1993). This is certainly
the case with commencer/cominciare. In fact, the lack of selectional restrictions on the
subject served as an argument for earlier raising analyses of aspectualizers mentioned
in section two.2 In stage C, the verb is associated strongly with the infinitival comple-
ment (Y), and they must refer to the same time (Heine 1993). This was illustrated for
�commencer/cominciare in Section€1 (see (5) and discussion thereof). Another prop-
erty of stage C items is that finite clausal complements are typically disallowed (Heine
1993). This is also true for commencer/cominciare, as shown in (10).
(10) a. Jean {espère/*commence} que Marie écrira le livre.
b. Gianni {spera/*comincia} che Maria scriva il libro.
“John hopes/*begins that Mary will write the book.”

It is likely that commencer/cominciare have not yet arrived at stage D on the Verb-
to-TAM chain since NP complements are usually not possible at this stage (Heine
1993). Due to their more decategorialized status, stage D items are also much less verb-
like than items at earlier stages, often appearing in a very restricted set of possible
tenses and moods; this phenomenon is called ‘paradigmaticity’ (Lehmann 1995). This
is not the case with commencer/cominciare, which may occur in the entire range of
tenses and moods.
This analysis explains the synchronic properties of commencer/cominciare and
resolves the issue of their categorization as full verbs or auxiliaries. As stage C items
on the Verb-to-TAM chain, they are between full verbs (stage A) and those which
are most commonly classified as auxiliaries like have and be (stage D/E). To complete
the analysis, it is ideal to have historical data in order to support the claims made in
this section. We will now present the diachronic data relevant for the development of
commencer/cominciare.

6.â•… Grammaticalization of commencer/cominciare: Diachronic analysis

Both commencer and cominciare can be traced back to a reconstructed Vulgar Latin
form *cuminitiare or *cominitiare, a compound of the Latin cum “with” and initiare “to
initiate” (Rey 1992; Cortelazzo & Zolli 1999). Thus, the verb initiare served as the lexical

.╅ Lamiroy shows that there do exist some selectional restrictions on the subject, mentioning
the ungrammaticality of expletive il when the infinitival complement is s’agir ‘to be important
to’ (1987:280). The reader is reminded that the stages are not discrete entities and merely
represent sets of points on the Verb-to-TAM chain. Therefore, stage C items may share some
properties with both stage B and stage D items with respect to some linguistic parameter(s).
See Heine’s discussion on the Overlap Model (1993:48) for details.
 Grammaticalization of commencer/cominciare 

source concept for commencer/cominciare. The constructions into which it entered are
instantiations of Heine’s action schema, shown below in (11).

(11) Action schema (Heine 1993:31)

‘X does Y’

X refers to the agent, Y is the complement NP, and does represents an action verb. In
this case, the action schema may have an additional element Z, given the meaning of
initiate. The proposed source schema for commencer/cominciare is given in (12).

(12) Source schema for commencer/cominciare

‘X initiates Y (into Z)’

This source event schema is concrete and grounded in the human experiences of that
epoch. The subject and the object are necessarily human, and in Classical Latin, the Z
element originally referred to a secret religious organization or a cult (Lewis & Short
1879; Glare 1983).

(13) se pro aegro eo vovisse ubi primum convaluisset, Bacchis eum se initiaturam
 (Livy, Ab Urbe Condita 39.9, 1st century A.D.)
“while he was sick, she had vowed to him that as soon as he had recovered,
she would initiate him into the Bacchic rites”

In later Christian writings, initiare was adopted as a general term for baptize (a person),
ordain (a priest), consecrate (a church), etc. (Jaberg 1925). The meaning of baptize is
exemplified in (14), which refers to John the Baptist.

(14) Quis enim corpus domini dignius initiaret…?

 (Tertullian, De Monogamia 8, 3rd century A.D.)
“For who was more worthy to baptize the body of the Lord…?” ╛╛╛

Here initiare may be interpreted as an initiation into the Christian religion (the unex-
pressed Z element), therefore this meaning is quite similar to that in (13).
This meaning was extended to refer to the introduction of someone to a practice
or skill (Glare 1983).
(15) rationes legebam…cursim (aliis enim chartis, aliis sum litteris initiatus)
 (Pliny the Younger, Epistulae 5.14, 1st century A.D.)
“I was reading their accounts…quickly (for I have been introduced to papers
and letters of another sort)”

Although the change is subtle, this meaning is more general than those in (13) and
(14). In (13) and (14) initiare describes a specific type of rite, the Z element referring
to a religious organization. In (15) the semantic scope of initiare and the Z element
is broader; it may refer to any practice or skill. This is the beginning of the deseman-
ticization process, which affects both the verb and the Z element in this case.
 Jason Brazeal

As early as the 3rd or 4th century we find the additional meanings of “to originate,
begin”, exemplified in (16) and (17).

(16) ex his initiata sunt cetera

 (Tertullian, Adversus Valentinianos XV, 3rd century A.D.)
“the others originated from these”

(17) ver enim tunc initiatur

 (Firmicus Maternus, Matheseos Libri VIII, 4th century A.D.)
“then spring certainly began”

In (16) and (17), initiare lacks the majority of its original lexical content, that of admis-
sion into a cult. This very concrete notion was bleached out in the desemanticization
process, and by the 3rd or 4th century initiare had acquired the more general meaning
of “begin”. Jaberg (1925) notes that the older meanings of initiare did not disappear as
the verb evolved. The meanings of “admit into a cult”, “introduce to a practice or skill”,
and “originate, begin” all coexisted at the same time.
Examples (16) and (17) suggest that initiare was decategorialized to some extent
as well. As a consequence of the desemanticization process, we would expect initiare
to undergo paradigmaticization, becoming less verb-like and appearing in a more
restricted set of morphological forms (Lehmann 1995). Examples in the passive voice
such as (16) and (17) are frequent in Late Latin. Furthermore, due to the bleaching of
its lexical meaning, initiare started to appear without a direct object (Y in the source
schema). Eventually, the Z element in the source schema, “into a cult” was left unex-
pressed, reduced to the semantically empty “in space or time”. A sketch of the develop-
ment of initiare is given in (18).

(18) Development of Latin initiare3

X initiates Y (into Z) → X introduces Y to Z →
Y is introduced in space or time → Y begins

The first stage is exemplified in (13), the second stage in (15), and the third stage
in (16) and (17). Note that the change from the third to the fourth stage does not
constitute further semantic change. In fact, this last development probably occurred
not as a result of further grammaticalization, but rather the loss of the Latin passive
morphology. Grammaticalization is not a process that occurs in isolation; it may

.â•… Initiare presents a case of ‘divergence’ (Hopper 1991). In addition to the developments
mentioned here, the lexical verb initiare survives in French and Italian as initier and iniziare,
respectively.
 Grammaticalization of commencer/cominciare 

interact with other changes occurring in the language at the same time. Further research,
specifically quantitative analyses of Latin corpora, will be needed to verify the claims
made in (18).
The pattern of development sketched above is not exactly the same as the typi-
cal one discussed in Heine (1993). Regarding the types of lexical verbs which tend
to grammaticalize into auxiliary-type structures, he cites Bybee, Perkins, & Pagliuca’s
(1994) idea that the verbs that tend to be used in source schemas are usually ones
with general meanings like go or come, appropriate in a wide range of contexts, rather
than ones which denote specific actions like stroll or swim. The case of initiare is one
in which the source concept is quite specific, and therefore it is a counterexample to
this generalization. Moreover, in most of the source schemas the main verb develops
into the auxiliary and the complement Y is reanalyzed as the new main verb. This is
clearly not the case with initiare. As previously noted, it seems that grammaticalization
interacted with the morphological changes going on at the same time (i.e. loss of syn-
thetic passive), which resulted in the complement Y becoming the new subject. This
suggests that our conception of the typical pattern of auxiliary development discussed
by Heine (1993) must remain flexible so that it may be adapted to take into account
other language-specific changes that occur parallel to grammaticalization. The last
change in (18) is, nevertheless, an important development since it was the one which
eventually permitted *cominitiare to start taking direct objects and infinitival comple-
ments. My claim is that this morphosyntactic change essentially reset the decategorial-
ization process and constituted a return to the original action schema, this time with
the form ‘X begins (Y)’. Ideally this claim would be substantiated with data, examples
of �*cominitiare used intransitively, transitively, and eventually with infinitival comple-
ments. As discussed below, though, this data is unfortunately not available.
The precise origin of *cominitiare is not known; it is first attested in Northern
France in the 10th century as commencier (Jaberg 1925). Because it doesn’t appear in
the works of Late Latin authors, Jaberg (1925) posits that *cominitiare arose between
the 5th and 10th centuries A.D. He cites several pieces of indirect evidence (see Jaberg
1925:123) but maintains that it is impossible to know the date of origin with absolute
certainty. Having already discussed initiare, we now turn to the prefix cum-.
The Latin preposition cum “with” functioned as a productive verbal prefix which
added nuances such as accompaniment (loqui “to speak”/colloqui “to converse”) and
completion (facere “to do, make”/conficere “to finish, accomplish”) in Latin. These are
often called expressive or intensive nuances in dictionaries and etymologies (cf. cum
intensif in the DAF entry for commencer). They tended to fade in Late Latin, though,
with new verbs derived with cum- having no real difference in meaning from the unpre-
fixed form (Ernout & Meillet 1967). Without data on the precise origin of *cominitiare,
we are not able to say exactly what the semantic contribution of cum- was, however we
 Jason Brazeal

can say with some degree of certainty that both before and after the addition of cum-,
this verb had the meaning of “to begin”.4
Fortunately, data on commencer/cominciare in early French and Italian are avail-
able. All three constructions that we find in the modern languages – transitive (19, 22),
intransitive (20, 23), and infinitival complement (21, 24) – were already present in Old
French and Old Italian, as the following examples show.5
(19) Henri, roi d’Engletière, ki voloit commencier la gierre
“Henry, king of England, who wanted to begin the war”
 (Philippe Mousket, Chronique, 1243)
(20) quant ce fu fait, un grant murmure commence
“when it was done, a great murmur began”
 (Christine de Pisan, Chemin de long estude, 1402)
(21) quant li hons commence a nestre
“when the man begins to be born” (Rutebeuf, Oeuvres complètes, 1260)
(22) ne l’ora che comincia i tristi lai la rondinella
“at the hour when the swallow begins its sorrowful song”
 (Dante, Commedia, 14th century)
(23) ma però l’adolescenza non comincia dal principio della vita
“but adolescence does not start from the beginning of life”
 (Dante, Convivio, 14th century)
(24) amor m’ha coninzato a meritare
“love has begun to be deserving to me”
 (Rinaldo d’Aquino, Per fino amore vao sì letamente, 13th century)

.╅ Jaberg (1925) has an interesting hypothesis regarding the formation of *cominitiare. He
suggests that it may have originated in the language of the clergy who tended to speak both
Latin and a Romance dialect during this period. He gives other examples such as consecrare
“to consecrate” and communicare “to partake, participate in (e.g. communion)”, noting that
they express collective actions in the church. These use of cum- in these words incorporated
the accompaniment nuance, perhaps paving the way for *cominitiare. Furthermore, given the
different meanings of initiare which all coexisted, there was an ambiguity between the reli-
gious meanings (baptize, ordain, etc.) and non-religious meanings (introduce, begin, etc.)
The church wanted to eliminate this ambiguity, and thus new, more specific words were
employed for the religious meanings (baptizare “to baptize”, ordinare “to ordain”, etc.). This
caused *cominitiare to specialize with the meaning “to begin”, which spread quickly since it
eliminated the ambiguity still present in initiare.
.â•… Examples (19)–(21) are taken from the ARTFL-FRANTEXT corpus of French, and
examples (22)–(24) are from the OVI corpus of Italian.
 Grammaticalization of commencer/cominciare 

These examples show that there were no restrictions on the subject in Old French
and Old Italian; both animate (human and non-human) and inanimate subjects were
possible. As for the object, recall that initiare permitted only animate objects in clas-
sical Latin. In Old French and Old Italian, however, commencer/cominciare could also
take inanimate objects and infinitival complements. This suggests that decategorializa-
tion has taken place since the further along the Verb-to-TAM chain a verb goes, the
more likely it is to take a nonfinite verb as its complement (Heine 1993). Furthermore,
when items at stage C on the Verb-to-TAM chain do take NP complements, it is likely
that they refer to an activity of some sort (Heine 1993:61). This was pointed out by
Rochette (1999) and Lamiroy (1987) for commencer in Modern French, and it applied
to �commencer/cominciare in Old French and Old Italian as well. The direct objects in
(19) and (22), “the war” and “its sorrowful song”, show that NP complements were not
concrete objects, but rather some kind of event that occurs or takes place. Finally, in
(21) and (24), the aspectualizer and the infinitival complement constitute a single event
at the level of semantic representation. This interpretation is similar to that of (7).
These semantic and syntactic properties are characteristic of stage C items, sug-
gesting that commencer/cominciare were already at stage C on the Verb-to-TAM chain
at the time of these early Romance attestations. Although these aspectualizers do not
appear to have progressed to stage D yet, there are some important developments to
consider in their evolution from Old French and Old Italian, as well as some interesting
differences between modern commencer and cominciare.

7.â•… Differences between commencer and cominciare

It is often claimed that grammaticalization in French is, in general, more advanced

than in Italian, although auxiliary-type structures sometimes show contradictory
results (Lamiroy 1999). When examining the data on paradigmatic variability, auxi�
liary selection, and clitic placement for commencer and cominciare, this is exactly
what we find.
First of all, these verbs show varying degrees of ‘paradigmatic variability’, the
possibility of using other lexical items in their place. As an element becomes more
grammaticalized, it often becomes more obligatory, meaning that there is less lexical
choice to express the same concept (Lehmann 1995). While both French and Italian
have inchoative expressions other than commencer/cominciare, they are much more
numerous in Italian (Lamiroy 1999). These alternate forms are often less frequent
that commencer/cominciare
� (e.g. there are 36 instances of incominciare in the LIP and
205 of cominciare), but the presence of more lexical choice in Italian suggests that
cominciare is less grammaticalized than commencer.
 Jason Brazeal

Secondly, cominciare occurs with both past tense auxiliaries, but commencer is
found only with avoir, as shown in (25).

(25) a. È cominciato a cadere la pioggia. (Lamiroy 1999:36)

b. La pluie a commencé à tomber.
“The rain began to fall.”

Infinitival complements in Italian impose their own auxiliary on cominciare, indi�

cating a greater degree of syntagmatic cohesion, or ‘bondedness’, than the French con-
struction (Lehmann 1995). This points to a greater degree of grammaticalization,
thus the data on auxiliary selection suggest that cominciare is more grammaticalized
than commencer.
With respect to the clitic placement in (26), the Italian construction again shows
a greater degree of bondedness.

(26) a. J’ai commencé à le faire./*Je l’ai commencé à faire.

b. Ho cominciato a farlo./L’ho cominciato a fare.
“I began to do it.”

In Italian, when the infinitive has an accompanying clitic, it may be attached to the
end of the infinitive or climb to a position before the auxiliary. In French, this clitic
must intervene between the aspectualizer and the infinitival complement.6 That the
Italian construction exhibits greater syntagmatic cohesion also seems to suggest that
�cominciare is more grammaticalized than commencer.
Why is it that commencer/cominciare show contradictory results with regards to
the degree of grammaticalization? It is important to realize that grammaticalization is
a gradual, dynamic process, and that each of its components may proceed at different
speeds. If we look only at paradigmatic variability, commencer is the more grammati-
calized form; if we consider only the parameter of bondedness, it is cominciare. It is
not always possible to situate an item at one exact point on the continuum due to the
multi-faceted and dynamic nature of the process of grammaticalization. Nevertheless,
it is still quite informative to state that commencer/cominciare are both located at an
intermediate position on a continuum of structures ranging from full lexical verbs to
purely grammatical affixes.

.╅ Clitic climbing, as in the second sentence of (26a), was possible in Old French, but in Modern
French it is grammatical only with causative and perception verbs (van Riemsdijk 1999).
 Grammaticalization of commencer/cominciare 

8.â•… Conclusion

This study has highlighted the usefulness of grammaticalization theory as an explanatory

device for areas of the grammar which prove troublesome for traditional approaches.
The aspectualizers commencer/cominciare have properties of both full verbs and auxi�
liaries. The analysis from the perspective of grammaticalization theory explains these
mixed properties by locating them at an intermediate position on the Verb-to-TAM
chain between full verbs and auxiliaries. They show properties of both because they
developed from a full verb in Latin and have become more grammatical over time.
They have undergone desemanticization and decategorialization and have progressed
to stage C on the Verb-to-TAM chain. Thus, they have properties of both lexical and
functional categories.
On the other hand, the current study leaves several questions unanswered. One
is the nature of the element preceding the infinitival complement of commencer/
cominciare, which I have called a preposition. Although I use the traditional term ‘prep-
osition’, this element has been analyzed as a complementizer by many researchers (e.g.
Spang-Hanssen 1963; Van Goethem 2005). Regardless of the syntactic analysis adopted,
the synchronic and diachronic variation in these elements is in need of a more detailed
treatment. Quantitative data from diachronic corpora will be essential for a comprehen-
sive study of these elements, and would also strengthen the grammaticalization analysis
presented here. Finally, all the modern examples of commencer/cominciare here were
from standard French and Italian. An in-depth study of these aspectualizers in non-
standard dialects as well as in other Romance languages would surely be informative, as
they may be more or less grammaticalized in other dialects and languages.
The present study has implications for linguistic theory with regards to the clas-
sification of linguistic items and the inclusion of diachronic data in linguistic analyses.
Many studies have shown that linguistic categories, even well-established ‘universal’
categories like noun and verb, are gradient (see Bybee & McClelland 2005 and refer-
ences therein). We have seen that by viewing aspectualizers as occupying intermediate
points on a continuum, their mixed properties can be more easily accounted for.
In addition, diachronic data have often been excluded from linguistic analyses on
the grounds that they don’t belong in any synchronic analysis. It is clear that the classi-
fication of commencer/cominciare as full verbs or auxiliaries is not at all straightforward
based on their synchronic properties alone. I would contend that diachronic data should
be included in linguistic analyses, when it is available, as it sometimes provides addi-
tional insights not available from synchronic data alone. Grammaticalization provides
a framework for integrating the diachronic and synchronic dimensions of linguistic
analysis. While obviously not the appropriate framework in every case, this panchronic
approach has certainly proven useful in the analysis of commencer/cominciare.
 Jason Brazeal

References

ARTFL-FRANTEXT (American and French Research on the Treasury of the French Language) –
http://humanities.uchicago.edu/orgs/ARTFL/
Bybee, Joan & James L. McClelland. 2005. “Alternatives to the Combinatorial Paradigm of
Linguistic Theory Based on Domain General Principles of Human Cognition”. The Linguistic
Review 22.381–410.
Bybee, Joan, Revere Perkins, & William Pagliuca. 1994. The Evolution of Grammar: Tense, Aspect,
and Modality in the Languages of the World. Chicago: Chicago University Press.
Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press.
Cortelazzo, Manlio & Paolo Zolli, eds. 1999. Il nuovo etimologico: DELI – dizionario etimologico
della lingua italiana. Bologna: Zanichelli.
DAF (Dictionnaire de l’Académie française, 9ème édition) – http://atilf.atilf.fr/academie9.htm
Ernout, Alfred & Antoine Meillet. 1967. Dictionnaire étymologique de la langue latine: histoire
des mots, 4ème édition. Paris: Klincksieck.
Glare, P.G.W., ed. 1983. Oxford Latin Dictionary. Oxford: Oxford University Press.
Heine, Bernd. 1993. Auxiliaries: Cognitive Forces and Grammaticalization. New York: Oxford
University Press.
Hopper, Paul. 1991. “On Some Principles of Grammaticization”. Approaches to Grammati-
calization, ed. by Elizabeth Closs Traugott & Bernd Heine, 17–36. Amsterdam: John
Benjamins.
Jaberg, Karl. 1925. “Der Sprach- und Sachatlas Italiens und der Südschweiz und die Bezeich-
nungsgeschichte des Begriffes ‘anfangen’↜”. Revue de linguistique romane 1.118–145.
Kuryłowicz, Jerzy. 1965 [1975]. “The Evolution of Grammatical Categories”. Esquisses linguis-
tiques, vol. 2, 38–54. Munich: W. Fink.
Lamiroy, Béatrice. 1999. “Auxiliaires, langues romanes et grammaticalisation”. Langages 135.33–45.
Lamiroy, Béatrice. 1987. “The Complementation of Aspectual Verbs in French”. Language
63.278–298.
Lehmann, Christian. 1995. Thoughts on Grammaticalization. Munich: Lincom Europa.
Lewis, Charlton T. & Charles Short, eds. 1879. A Latin Dictionary. Oxford: Clarendon.
LIP (Lessico di frequenza dell’italiano parlato) – http://badip.uni-graz.at/index.php
Meillet, Antoine. 1912 [1948]. “L’évolution des formes grammaticales”. Linguistique historique et
linguistique générale, 130–148. Paris: Champion.
Newmeyer, Frederick. 1975. English Aspectual Verbs. The Hague: Mouton.
Newmeyer, Frederick. 1998. Language Form and Language Function. Cambridge, Mass.:
MIT Press.
OVI (Opera del vocabolario italiano) – http://www.lib.uchicago.edu/efts/ARTFL/projects/OVI/
Perlmutter, David. M. 1970. “The Two Verbs Begin”. Readings in English Transformational
Grammar, ed. by Roderick A. Jacobs & Peter S. Rosenbaum, 107–120. Waltham, Mass.: Ginn.
Pustejovsky, James. 1991. “The Syntax of Event Structure”. Cognition 41. 47–81.
Rey, Alain, ed. 1992. Dictionnaire historique de la langue française. Paris: Le Robert.
Rochette, Anne. 1999. “The Selectional Properties of Aspectual Verbs”. Beyond Principles and
Parameters: Essays in Memory of Osvaldo Jaeggli, ed. by Kyle Johnson & Ian Roberts,
145–165. Dordrecht: Kluwer.
Ruwet, Nicholas. 1972. Théorie syntaxique et syntaxe du français. Paris: Seuil.
Spang-Hanssen, Ebbe. 1963. Les prépositions incolores du français. Copenhagen: GEC Gads Forlag.
 Grammaticalization of commencer/cominciare 

Van Goethem, Kristel 2005. “La préposition française à et ses correspondants romans (espagnol,
italien) et germaniques (néerlandais, allemand)”. Beiträge zum romanisch-deutschen und
innerromanischen Sprachvergleich, ed. by Christian Schmitt & Barbara Wotjak, 305–316.
Bonn: Romanistischer Verlag.
Van Riemsdijk, Henk, ed. 1999. Clitics in the Languages of Europe. Berlin: Mouton de Gruyter.
Vendler, Zeno. 1957. “Verbs and Times”. The Philosophical Review 66.143–60.
 Index of subjects, terms and languages
A regressive assimilation╇
accretion╇ 337–342, 345,
347–350
action schema╇ 413, 415
affectedness╇ 224, 228
Agree╇ 178–179, 181, 205–206,
212, 218, 222, 235, 249–251,
253, 257–263, 291, 294, 396
agreement╇ 174, 185, 204–205,
212–213, 222, 232, 241,
244, 246–247, 291, 293,
323–324, 329–332, 393, 396
(dis)agreement╇ 179,
249–251, 254, 256–257, 260
extra-grammatical
agreement╇ 249
number agreement╇ 171,
179–181, 185, 203–214,
249–250, 254–264, 286,
291–298, 324, 329–332, 334,
340–344, 377, 385
person agreement╇ 195–196,
203–208, 210, 227, 232,
235, 239, 249–263, 291, 334,
337, 340–344, 350, 363, 365
strong agreement╇ 268
Aktionsart╇ 397, 401
anaphor╇ 187, 192–195, 198,
200–201 , 231–235, 237,
246, 375, 399
nominal anaphora╇ 237–238
animacy╇ 196–197, 213, 269, 271,
277, 279
anominal pronouns╇ 244, 246
apocope╇ 130, 133, 138,
140–145, 149
aspectual categories╇ 388
aspectual properties╇ 400, 402
aspiration╇ 15–16
assimilation╇ 26–27, 39, 46
anticipatory assimilation╇ 25
local assimilation╇ 27, 39
long-distance
assimilation╇ 39
26–27, 37, 39
vowel-to-vowel assimilation
(V-to-V assimilation,
VVA)╇ 25–28, 35, 39–40
atelic predicates╇ 397, 398
B 301–302, 308, 310, 315, 386
bare indices╇ 235, 244
binding╇ 187–189, 192–195,
199–201, 221, 292
bleaching╇ 411, 414
bondedness╇ 418
C cliticization╇ 410–411
cardinal adverbials╇ 391, 400
case╇ 190–195, 198–199,
203–214, 224–225,
228–229, 236–245,
257–259, 262–263,
285–286, 288–298,
329–332, 371–377, 393–396.
Catalan╇ 44, 55, 63–70,
72–77, 81–85, 88–93, 113,
156, 203–204, 211–214,
239–240, 254, 260, 267,
269, 271, 274, 278–281,
337–338, 340–341, 344, 347,
350, 388, 391
Formenteran Catalan╇
337–340, 342–343, 348
Majorcan Catalan╇ 63–67,
81–83, 92, 338
Minorcan Catalan╇ 274,
278–281
central coincidence╇ 223, 228
Cibao╇ 18
change╇ 305, 310, 318, 387, 393,
406, 410, 411
change in progress╇ 92,
302, 315
linguistic change╇ 143, 273
morphosyntactic
change╇ 415
semantic change╇ 389,
402, 414
sound change╇ 129–130,
133, 135–136, 138, 140, 145,
148–149
syntactic change╇ 301
clause type╇ 269–270, 276,
clitic╇ 97–98, 111–115, 119–125,
203–214, 220–223,
228–229, 232–234, 239–241,
243–245, 339–341, 343,
353–357, 360–364, 398
climbing╇ 172–173, 417–418
cluster╇ 203, 207–209,
212–213, 348
linear ordering of clitics╇
209–210
cluster
clitic cluster see Clitic
consonant cluster╇ 23,
159, 354
coarticulation╇ 27
coarticulatory╇ 26, 28–29,
31, 39
Composite Group╇ 95–96,
105–108
compounds╇ 98–107, 293
contact╇ 73, 81 83, 92, 273,
281, 325
dialect contact╇ 139
language contact╇ 268, 272
contextual inflection╇ 323–324,
329–332
continuative perfect╇ 401
contrast╇ 48–49, 63–64,
72–77, 143
contrast preservation╇
63–64, 73
phonemic contrast╇ 47, 63,
73, 77, 143
control╇ 407–409
counterbleeding╇ 158–159
 Index of subjects, terms and languages
CREA╇ 220
creole╇ 111, 124–125, 323–334
creole languages╇ 323–324, 330
Indo-Portuguese
creoles╇ 323–325, 327,
332–334
current relevance╇ 388, 391
D 180–185, 190, 258, 355,
decategorialization╇ 410–411,
415, 417, 419
defective phase╇ 295, 297–298
definiteness╇ 286, 292, 324,
329, 331
deletion╇ 15–17, 129–131,
134–135, 137–138, 144, 149
desemanticization╇ 394,
410–411, 413–414, 419
devoicing╇ 75–77, 153–154,
156–159, 163–164
diachronic╇ 2–4, 8–11, 15, 43–44,
387–389, 392–393, 395–396
See also Synchronic and/or
diachronic
Differential Object
Marking╇ 181, 185
discourse╇ 178, 184, 190, 198,
236–238, 250, 261, 269, 271,
303, 370, 374–378, 383–385
distance╇ 187, 193–198, 200,
205–206, 212–213,
269–270, 276, 279–281
divergence╇ 414
DP╇ 176, 189, 191, 200, 217, 219,
221–222, 228–229, 243,
260–261, 263, 286, 288,
290–298
Dutch╇ 95–96, 99–100,
103–107, 156
E genitive╇ 232, 285–295, 297–298
enclisis╇ 337–343, 345–346,
349–350
epenthesis╇ 133, 153–154,
158–159, 163–164, 340,
354–355
events
iteration╇ 390, 395, 401–402
schemas╇ 410, 413
exclamatives╇ 369–372, 377–386
F hierarchy╇ 409
feature╇ 203–210, 212–214,
220, 234–235, 249–250,
256–264, 281, 285–286,
291–292, 294–298, 334,
341–344, 356, 369, 396–397
interpretable features╇ 234,
257–259, 291–292, 295–298
feature spreading rules╇
27, 39
focus╇ 169–174, 177–178,
369–372, 374–376, 380,
383–385, 394, 407
French╇ 18, 22, 25–29, 39–40,
74–75, 153–154, 156, 158,
187, 189–190, 192, 201,
204–205, 222, 232, 239,
285–286, 301–302, 307–308,
311, 313, 317–318, 330, 356,
369–370, 372–374, 376–377,
380–382, 386–387, 391,
405–406, 408–409, 411,
414, 416–419
fronted subordinate clause╇
301–302, 305, 310– 311,
313–31, 318
functional vowel╇ 355,
359–363, 365
G 199–201
Galician╇ 45–47, 134, 136–138,
144, 146–149, 394
See also Galician-Portuguese
gender╇ 81, 84, 87–92, 231–247,
249–250, 256–258,
260–264, 268–269,
279, 289, 291–292,
294–298, 396
in pronouns╇ 238
in Romanian╇ 234
genderless pronouns╇
231–232, 236, 244–245, 247
Germanic╇ 95, 99, 104–105, 107,
153, 156, 303, 312, 317–318
gradation quantifiers╇ 369–376,
379–381, 383–385
gradience╇ 25
grammaticalization╇ 387,
402, 404, 406, 409–412,
414–415, 417–419
H
constraint hierarchy╇
280, 347
implicational thematic
hierarchy╇ 223–224
prosodic hierarchy╇ 96,
105–106, 108, 131, 133
variable hierarchy╇ 280
hypercorrection╇ 15–19, 22
I
identical vowel
simplification╇ 129–131,
135–137, 139, 149
identification╇ 250, 268
iterative/durative
meaning╇ 390–391, 398
imperatives╇ 337–344, 346–347,
349–350, 408
Indirect Object
Construction╇ 217
inflectional morphology╇ 258,
323–324, 329, 331, 334
inherent inflection╇ 323–324,
329, 331–334
initial accent in French╇ 39–40
insertion╇ 139, 293–294, 363
[s] insertion╇ 15–16, 18, 20–22
coda insertion╇ 18, 22–23
intensification╇ 187–190, 192,
interrogative vowel╇ 353, 355,
357–359, 361–363, 365–366
IP-adverbs╇ 175
Italian╇ 74, 95–96, 99–101,
104–107, 196, 204, 212–213,
222, 239, 246, 252, 272, 291,
293, 313, 353–354, 356–357,
365, 369–370, 373, 377, 381,
383, 386, 405–406, 411, 414,
416–419
Bobbio dialect╇ 359–360
Donceto dialect╇ 355
Gazzoli dialect╇ 353, 355,
358–366
Groppallo dialect╇
359–363, 366
Piacentine dialects╇ 353–354,
358–360, 363, 366
Travo dialect╇ 359–360, 366
J
Juxtaposition Genitive╇ 285, 288
L
Latin╇ 23, 44–46, 73–75, 77,
129–131, 133–135, 137–138,

140–143, 145, 148–149,
238, 239, 246, 285, 387,
393–394, 405–406,
412–417, 419
left periphery╇ 315
legal texts╇ 307–308, 311–312
lengthening╇ 147–149, 153–158,
162–163
moraic lengthening╇ 137,
147, 149
lenition╇ 47, 55, 64–65, 70,
73–77
Leonese╇ 129, 134–136
lexical conservatism╇ 337,
339, 342
light vs. heavy syllable╇ 132, 140
literacy╇ 15, 17, 19, 21
Logophoricity╇ 187, 193–200,
198, 200–201
Luiseño╇ 222
M opacity╇ 153–154, 156–160,
MAP╇ 218, 224–229
markedness╇ 16, 22, 154–156,
158–159, 162, 164, 203–204,
206, 210, 213–214, 228
positional markedness╇
155–156
minimal word╇ 130–131, 133,
138–140, 149
word minimality╇ 129–131,
133, 137, 140, 143, 145,
147, 149
minimalism╇ 290, 294
mora-sharing╇ 154–155, 157
morphological uniformity
principle╇ 268
Morphosyntactic Alignment
Principle╇ 217–218,
223–224, 229
motion schema╇ 410
motion verb╇ 409
N Case Constraint╇ 203–204
nasal╇ 43–50, 55–58, 130, 134, 148
palatal nasal╇ 43–50, 52–58
natural gender╇ 234,
237–239, 258
negative polarity items╇ 173
neuter╇ 246
neuter gender╇ 234, 238,
246–247
neuter pronouns╇ 231, 233–234,
236, 242, 245–246
Index of subjects, terms and languages 
null objects╇ 231–232, 234–236
null subject╇ 240, 254, 257,
260–261, 267–271,
274–275, 278–280, 303,
306, 314–316
null-subject language╇
249–250, 252, 263
number See Agreement
O 48, 50–58
object clitic╇ 212, 240, 316, 363
Obligatory Contour
Principle╇ 121–123
oblique╇ 217–219, 223–225,
285–286, 288–290,
295–296, 298
Old French╇ 156, 285, 301–302,
307–308, 311, 317–318, 411,
416–418
Old Occitan╇ 301–302, 313,
318–319
162–164
OT-CC╇ 153–154, 160, 162–164
Overlap Model╇ 412
overt╇ 253, 260–261, 263,
267–275, 277–281, 285, 292,
294–295, 306, 323–325,
330–331, 333, 373, 382–383
P 242, 244, 252–253, 257, 261,
Palenquero╇ 111–126
paradigmatic variability╇
417–418
parametric variation╇ 268
parasitic gaps╇ 231–233, 235
PCC╇ 204–207, 212–214
perception╇ 23, 58
perfect of persistent
situation╇ 401
periphrastic past╇ 387–391, 393,
401–402
person See Agreement Person
phase╇ 285, 291, 294–295,
297–298
phonetic erosion╇ 411
phonological phrase╇ 95–96,
99, 105–106, 108
phonological word╇ 95–96,
104–106
pluractional operator╇ 391
polarity╇ 113, 117, 120–122,
125, 173
popular speech╇ 18
Portuguese╇ 43–50, 53, 55–58,
60, 74–75, 124, 129–130,
134–138, 144, 146–149, 252,
271, 323–328, 330, 332–334,
387–395, 400–402
Brazilian Portuguese╇ 43,
170, 271
See also BP╇ 43, 44, 47,
European Portuguese,╇ 44,
48, 58
See also EP╇ 44, 47, 48
Galician-Portuguese╇ 43, 45,
129–130, 134–135, 137–138,
144, 146, 148
See also Galician
Portuguese-based creoles╇ 325,
327–328, 332
possessive╇ 111–113, 115, 118,
124–125, 187, 189, 191–192,
194, 222, 232, 289, 293–294
precedence╇ 153–154,
162–164, 393
preposition╇ 285–286, 289–290,
292–293, 407, 415, 419
Prepositional Dative
Construction╇ 217
present perfect╇ 353
pro╇ 172, 222, 231–235, 238, 240,
263, 267, 269, 289, 356, 413
Probe-Goal╇ 249, 263, 292, 294
pronoun╇ 171–172, 193, 204,
208, 222, 231–235, 237, 239,
244–245, 253, 257, 294,
304, 306, 308, 398, 411
prosodic╇ 39–40, 85, 95, 99, 102,
104–107, 130–133, 138–139,
142–147, 149
prosodic hierarchy see
Hierarchy
prosodic word╇ 130–131,
133–134, 139–140, 144–146,
148–149
Prosodic Minimality
Condition╇ 133
pseudo-cleft╇ 169, 171–174,
182, 185
R
raising╇ 292–294, 407–409, 412
vowel raising╇ 39, 100
reconstruction╇ 220
 Index of subjects, terms and languages

resultative╇ 387, 390, 392–397, Dominican Spanish╇ 15–18, trochee

400, 402
resyllabification╇ 158–160
Romance╇ 1–8, 10–11, 22, 74,
77, 95, 99, 104–105, 107,
134, 143, 156, 159, 210, 214,
231, 236–237, 239–241,
246–247, 252, 254, 274,
293–294, 303–304, 317–318,
325–326, 330, 332, 337–340,
356, 361, 366, 369–370, 372,
377, 386–391, 393, 401, 406,
416–417, 419
Hispano-Romance╇ 129, 144
Ibero-Romance╇ 47, 73, 74,
76, 129
Medieval Romance╇
303–304, 313, 315, 317–318
Western Romance╇ 73, 74,
240, 242, 267, 274
Romanian╇ 203, 204, 210, 214,
231–234, 236, 240–247, 377
S variationist╇ 81, 83, 267–269,
San Basilio de Palenque╇ 111, 114
sandhi╇ 22, 114, 126
Second Language
Acquisition╇ 324, 333
semantic attrition╇ 411
semantic verb type╇ 271, 277,
279–280
Sicilian╇ 387
small clause╇ 218–219, 387, 393
sociolinguistic╇ 17–19, 23, 81–82,
84, 91–92, 139, 267–268,
273–275, 280, 324, 333
Spanish╇ 15–18, 23, 43–50, 53,
55–57, 61, 63–64, 73–77,
81–85, 87–93, 111–112, 114,
129–130, 132–140, 142–149,
169–170, 172, 174–176, 178,
183, 203–214, 217–222,
224–226, 228–229, 240,
245, 249–257, 260–264,
267–269, 271, 273–274,
278–281, 356, 369–370,
381–383, 387–389, 391,
393–394, 400–402
Caribbean Spanish╇ 170,
249–250, 256, 263, 280
20, 22, 169–170, 175,
268, 273
Puerto Rican Spanish╇
48–49, 273
Valladolid Spanish╇ 274,
278–280
Spanish Royal Academy
corpus╇ 220
speech connectivity╇ 269,
276–277
speech encoding╇ 95–96,
99–100, 104–105, 107–108
speech production╇ 96, 104
spirantization╇ 64, 76
spurious se╇ 208
stress shift╇ 339, 346, 348
subject inversion╇ 250,
302–303, 313
subjects╇ 175–176, 222, 242,
249–252, 256–258,
261–264, 286, 308, 314–315,
330, 411, 417
subject clitic╇ 353–357,
361–364
subject expression╇ 267–269,
272–273, 278–279
subject form╇ 275, 277–278
thematic subjects╇ 268–269
switch reference╇ 269, 276
syllable weight╇ 131, 140, 143, 145
synchronic╇ 130, 143, 154, 290,
402, 406, 410–412, 419
synchronic and/or
diachronic╇ 2, 11, 294,
388, 402
syntactic verb type╇ 269,
271–272, 277, 279
T Weak Crossover╇ 222
telic predicates╇ 398
tense╇ 178, 181, 185, 256,
269–270, 291, 324–325,
327, 329, 333, 342, 353,
396, 405, 409–410, 418
T-feature╇ 286, 291, 295, 298
Tobler-Mussafia Law╇ 316–317
topic continuity╇ 270, 279
TP-internal focus╇ 169,
171, 183
moraic trochee╇ 131–132, 138,
140, 142–143, 145, 149, 339,
342, 344, 348
syllabic trochee╇ 132, 140,
142, 144–145, 149
Turbidity╇ 159
U
Uniform Exponence╇ 339,
347–350
uninterpretable╇ 220, 257–259,
291–292, 295–298
units╇ 323–324, 326–327
V
variable╇ 229, 231, 268–272,
275–281
variation╇ 2, 26, 44, , 64, 82,
153–154, 180, 267–268,
272–274, 303, 311–312, 316,
318, 326, 328
synchronic variation╇ 4,
43, 419
271–272, 274, 280–281
verb form ambiguity╇ 269,
272, 277
verb-second╇ 301
Verb-to-TAM chain╇ 405,
410–412, 417, 419
Vowel Harmony╇ 25–26
Vowel-to-vowel assimilation
See Assimilation
VVA
See Vowel-to-vowel
assimilation
VP-adverbs╇ 175
W
weakening╇ 64, 73, 75–77
weight-bearing codas╇ 154
Western Romance
languages╇ 267, 274
word order╇ 220, 228, 292, 302,
305, 307–309, 311, 313,
315–318
Y
yod augmentation╇ 139–140
 CURRENT ISSUES IN LINGUISTIC THEORY

E. F. K. Koerner, Editor
Zentrum für Allgemeine Sprachwissenschaft, Typologie
und Universalienforschung, Berlin
efk.koerner@rz.hu-berlin.de

Current Issues in Linguistic Theory (CILT) is a theory-oriented series which welcomes contributions from
scholars who have significant proposals to make towards the advancement of our understanding of lan-
guage, its structure, functioning and development. CILT has been established in order to provide a forum
for the presentation and discussion of linguistic opinions of scholars who do not necessarily accept the
prevailing mode of thought in linguistic science. It offers an outlet for meaningful contributions to the
current linguistic debate, and furnishes the diversity of opinion which a healthy discipline must have.
A complete list of titles in this series can be found on the publishers’ website, www.benjamins.com

315 COLINA, Sonia, Antxon OLARREA and Ana Maria CARVALHO (eds.): Romance Linguistics 2009.
Selected papers from the 39th Linguistic Symposium on Romance Languages (LSRL), Tucson, Arizona,
March 2009. 2010. xiv, 426 pp.
314 LENKER, Ursula, Judith HUBER and Robert MAILHAMMER (eds.): English Historical Linguistics
2008. Selected papers from the fifteenth International Conference on English Historical Linguistics (ICEHL
15), Munich, 24-30 August 2008.. Volume I: The history of English verbal and nominal constructions. 2010.
vii, 281 pp.
313 ARREGI, Karlos, Zsuzsanna FAGYAL, Silvina A. MONTRUL and Annie TREMBLAY (eds.): Romance
Linguistics 2008. Interactions in Romance. Selected papers from the 38th Linguistic Symposium on
Romance Languages (LSRL), Urbana-Champaign, April 2008. 2010. vii, 266 pp.
312 OPERSTEIN, Natalie: Consonant Structure and Prevocalization. 2010. x, 234 pp.
311 SCALISE, Sergio and Irene VOGEL (eds.): Cross-Disciplinary Issues in Compounding. 2010. viii, 382 pp.
310 RAINER, Franz, Wolfgang U. DRESSLER, Dieter KASTOVSKY and Hans Christian LUSCHÜTZKY
(eds.): Variation and Change in Morphology. Selected papers from the 13th International Morphology
Meeting, Vienna, February 2008. With the assistance of Elisabeth Peters. 2010. vii, 249 pp.
309 NICOLOV, Nicolas, Galia ANGELOVA and Ruslan MITKOV (eds.): Recent Advances in Natural
Language Processing V. Selected papers from RANLP 2007. 2009. x, 338 pp.
308 DUFRESNE, Monique, Fernande DUPUIS and Etleva VOCAJ (eds.): Historical Linguistics 2007. Selected
papers from the 18th International Conference on Historical Linguistics, Montreal, 6–11 August 2007. 2009.
x, 311 pp.
307 CALABRESE, Andrea and W. Leo WETZELS (eds.): Loan Phonology. 2009. vii, 273 pp.
306 VIGÁRIO, Marina, Sónia FROTA and M. João FREITAS (eds.): Phonetics and Phonology. Interactions
and interrelations. 2009. vi, 290 pp.
305 BUBENIK, Vit, John HEWSON and Sarah ROSE (eds.): Grammatical Change in Indo-European
Languages. Papers presented at the workshop on Indo-European Linguistics at the XVIIIth International
Conference on Historical Linguistics, Montreal, 2007. 2009. xx, 262 pp.
304 MASULLO, Pascual José, Erin O'ROURKE and Chia-Hui HUANG (eds.): Romance Linguistics 2007.
Selected papers from the 37th Linguistic Symposium on Romance Languages (LSRL), Pittsburgh, 15–18
March 2007. 2009. vii, 361 pp.
303 TORCK, Danièle and W. Leo WETZELS (eds.): Romance Languages and Linguistic Theory 2006. Selected
papers from ‘Going Romance’, Amsterdam, 7–9 December 2006. 2009. viii, 262 pp.
302 FERRARESI, Gisella and Maria GOLDBACH (eds.): Principles of Syntactic Reconstruction. 2008.
xvii, 219 pp.
301 PARKINSON, Dilworth B. (ed.): Perspectives on Arabic Linguistics. Papers from the annual symposium
on Arabic linguistics. Volume XXI: Provo, Utah, March 2007. 2008. x, 206 pp.
300 VAJDA, Edward J. (ed.): Subordination and Coordination Strategies in North Asian Languages. 2008.
xii, 218 pp.
299 GONZÁLEZ-DÍAZ, Victorina: English Adjective Comparison. A historical perspective. 2008. xix, 252 pp.
298 BOWERN, Claire, Bethwyn EVANS and Luisa MICELI (eds.): Morphology and Language History. In
honour of Harold Koch. 2008. x, 364 pp.
297 DOSSENA, Marina, Richard DURY and Maurizio GOTTI (eds.): English Historical Linguistics 2006.
Selected papers from the fourteenth International Conference on English Historical Linguistics (ICEHL 14),
Bergamo, 21–25 August 2006. Volume III: Geo-Historical Variation in English. 2008. xiii, 197 pp.
296 DURY, Richard, Maurizio GOTTI and Marina DOSSENA (eds.): English Historical Linguistics 2006.
Selected papers from the fourteenth International Conference on English Historical Linguistics (ICEHL 14),
Bergamo, 21–25 August 2006. Volume II: Lexical and Semantic Change. 2008. xiii, 264 pp.
295 GOTTI, Maurizio, Marina DOSSENA and Richard DURY (eds.): English Historical Linguistics 2006.
Selected papers from the fourteenth International Conference on English Historical Linguistics (ICEHL 14),
Bergamo, 21–25 August 2006. Volume I: Syntax and Morphology. 2008. xiv, 259 pp.
294 FRELLESVIG, Bjarke and John WHITMAN (eds.): Proto-Japanese. Issues and Prospects. 2008.
vii, 229 pp.
293 DETGES, Ulrich and Richard WALTEREIT (eds.): The Paradox of Grammatical Change. Perspectives
from Romance. 2008. vi, 252 pp.
292 NICOLOV, Nicolas, Kalina BONTCHEVA, Galia ANGELOVA and Ruslan MITKOV (eds.): Recent
Advances in Natural Language Processing IV. Selected papers from RANLP 2005. 2007. xii, 307 pp.
291 BAAUW, Sergio, Frank DRIJKONINGEN and Manuela PINTO (eds.): Romance Languages and
Linguistic Theory 2005. Selected papers from ‘Going Romance’, Utrecht, 8–10 December 2005. 2007.
viii, 338 pp.
290 MUGHAZY, Mustafa A. (ed.): Perspectives on Arabic Linguistics. Papers from the annual symposium on
Arabic linguistics. Volume XX: Kalamazoo, Michigan, March 2006. 2007. xii, 247 pp.
289 BENMAMOUN, Elabbas (ed.): Perspectives on Arabic Linguistics. Papers from the annual symposium on
Arabic Linguistics. Volume XIX: Urbana, Illinois, April 2005. 2007. xiv, 304 pp.
288 TOIVONEN, Ida and Diane NELSON (eds.): Saami Linguistics. 2007. viii, 321 pp.
287 CAMACHO, José, Nydia FLORES-FERRÁN, Liliana SÁNCHEZ, Viviane DÉPREZ and María José
CABRERA (eds.): Romance Linguistics 2006. Selected papers from the 36th Linguistic Symposium on
Romance Languages (LSRL), New Brunswick, March-April 2006. 2007. viii, 340 pp.
286 WEIJER, Jeroen van de and Erik Jan van der TORRE (eds.): Voicing in Dutch. (De)voicing – phonology,
phonetics, and psycholinguistics. 2007. x, 186 pp.
285 SACKMANN, Robin (ed.): Explorations in Integrational Linguistics. Four essays on German, French, and
Guaraní. 2008. ix, 239 pp.
284 SALMONS, Joseph C. and Shannon DUBENION-SMITH (eds.): Historical Linguistics 2005. Selected
papers from the 17th International Conference on Historical Linguistics, Madison, Wisconsin, 31 July - 5
August 2005. 2007. viii, 413 pp.
283 LENKER, Ursula and Anneli MEURMAN-SOLIN (eds.): Connectives in the History of English. 2007.
viii, 318 pp.
282 PRIETO, Pilar, Joan MASCARÓ and Maria-Josep SOLÉ (eds.): Segmental and prosodic issues in
Romance phonology. 2007. xvi, 262 pp.
281 VERMEERBERGEN, Myriam, Lorraine LEESON and Onno CRASBORN (eds.): Simultaneity in Signed
Languages. Form and function. 2007. viii, 360 pp. (incl. CD-Rom).
280 HEWSON, John and Vit BUBENIK: From Case to Adposition. The development of configurational syntax
in Indo-European languages. 2006. xxx, 420 pp.
279 NEDERGAARD THOMSEN, Ole (ed.): Competing Models of Linguistic Change. Evolution and beyond.
2006. vi, 344 pp.
278 DOETJES, Jenny and Paz GONZÁLEZ (eds.): Romance Languages and Linguistic Theory 2004. Selected
papers from ‘Going Romance’, Leiden, 9–11 December 2004. 2006. viii, 320 pp.
277 HELASVUO, Marja-Liisa and Lyle CAMPBELL (eds.): Grammar from the Human Perspective. Case,
space and person in Finnish. 2006. x, 280 pp.
276 MONTREUIL, Jean-Pierre Y. (ed.): New Perspectives on Romance Linguistics. Vol. II: Phonetics,
Phonology and Dialectology. Selected papers from the 35th Linguistic Symposium on Romance Languages
(LSRL), Austin, Texas, February 2005. 2006. x, 213 pp.
275 NISHIDA, Chiyo and Jean-Pierre Y. MONTREUIL (eds.): New Perspectives on Romance Linguistics. Vol.
I: Morphology, Syntax, Semantics, and Pragmatics. Selected papers from the 35th Linguistic Symposium on
Romance Languages (LSRL), Austin, Texas, February 2005. 2006. xiv, 288 pp.
274 GESS, Randall S. and Deborah ARTEAGA (eds.): Historical Romance Linguistics. Retrospective and
perspectives. 2006. viii, 393 pp.
273 FILPPULA, Markku, Juhani KLEMOLA, Marjatta PALANDER and Esa PENTTILÄ (eds.): Dialects
Across Borders. Selected papers from the 11th International Conference on Methods in Dialectology
(Methods XI), Joensuu, August 2002. 2005. xii, 291 pp.
272 GESS, Randall S. and Edward J. RUBIN (eds.): Theoretical and Experimental Approaches to Romance
Linguistics. Selected papers from the 34th Linguistic Symposium on Romance Languages (LSRL), Salt Lake
City, March 2004. 2005. viii, 367 pp.
271 BRANNER, David Prager (ed.): The Chinese Rime Tables. Linguistic philosophy and historical-
comparative phonology. 2006. viii, 358 pp.
270 GEERTS, Twan, Ivo van GINNEKEN and Haike JACOBS (eds.): Romance Languages and Linguistic
Theory 2003. Selected papers from ‘Going Romance’ 2003, Nijmegen, 20–22 November. 2005. viii, 369 pp.
269 HARGUS, Sharon and Keren RICE (eds.): Athabaskan Prosody. 2005. xii, 432 pp.
268 CRAVENS, Thomas D. (ed.): Variation and Reconstruction. 2006. viii, 223 pp.
267 ALHAWARY, Mohammad T. and Elabbas BENMAMOUN (eds.): Perspectives on Arabic Linguistics.
Papers from the annual symposium on Arabic linguistics. Volume XVII–XVIII: Alexandria, 2003 and
Norman, Oklahoma 2004. 2005. xvi, 315 pp.
266 BOUDELAA, Sami (ed.): Perspectives on Arabic Linguistics. Papers from the annual symposium on Arabic
linguistics. Volume XVI: , Cambridge, March 2002. 2006. xii, 181 pp.
265 CORNIPS, Leonie and Karen P. CORRIGAN (eds.): Syntax and Variation. Reconciling the Biological and
the Social. 2005. vi, 312 pp.
264 DRESSLER, Wolfgang U., Dieter KASTOVSKY, Oskar E. PFEIFFER and Franz RAINER (eds.):
Morphology and its demarcations. Selected papers from the 11th Morphology meeting, Vienna, February
2004. With the assistance of Francesco Gardani and Markus A. Pöchtrager. 2005. xiv, 320 pp.
263 BRANCO, António, Tony McENERY and Ruslan MITKOV (eds.): Anaphora Processing. Linguistic,
cognitive and computational modelling. 2005. x, 449 pp.
262 VAJDA, Edward J. (ed.): Languages and Prehistory of Central Siberia. 2004. x, 275 pp.
261 KAY, Christian J. and Jeremy J. SMITH (eds.): Categorization in the History of English. 2004. viii, 268 pp.
260 NICOLOV, Nicolas, Kalina BONTCHEVA, Galia ANGELOVA and Ruslan MITKOV (eds.): Recent
Advances in Natural Language Processing III. Selected papers from RANLP 2003. 2004. xii, 402 pp.
259 CARR, Philip, Jacques DURAND and Colin J. EWEN (eds.): Headhood, Elements, Specification and
Contrastivity. Phonological papers in honour of John Anderson. 2005. xxviii, 405 pp.
258 AUGER, Julie, J. Clancy CLEMENTS and Barbara VANCE (eds.): Contemporary Approaches to
Romance Linguistics. Selected Papers from the 33rd Linguistic Symposium on Romance Languages (LSRL),
Bloomington, Indiana, April 2003. With the assistance of Rachel T. Anderson. 2004. viii, 404 pp.
257 FORTESCUE, Michael, Eva Skafte JENSEN, Jens Erik MOGENSEN and Lene SCHØSLER (eds.):
Historical Linguistics 2003. Selected papers from the 16th International Conference on Historical
Linguistics, Copenhagen, 11–15 August 2003. 2005. x, 312 pp.
256 BOK-BENNEMA, Reineke, Bart HOLLEBRANDSE, Brigitte KAMPERS-MANHE and Petra
SLEEMAN (eds.): Romance Languages and Linguistic Theory 2002. Selected papers from ‘Going Romance’,
Groningen, 28–30 November 2002. 2004. viii, 273 pp.
255 MEULEN, Alice ter and Werner ABRAHAM (eds.): The Composition of Meaning. From lexeme to
discourse. 2004. vi, 232 pp.
254 BALDI, Philip and Pietro U. DINI (eds.): Studies in Baltic and Indo-European Linguistics. In honor of
William R. Schmalstieg. 2004. xlvi, 302 pp.
253 CAFFAREL, Alice, J.R. MARTIN and Christian M.I.M. MATTHIESSEN (eds.): Language Typology. A
functional perspective. 2004. xiv, 702 pp.
252 KAY, Christian J., Carole HOUGH and Irené WOTHERSPOON (eds.): New Perspectives on English
Historical Linguistics. Selected papers from 12 ICEHL, Glasgow, 21–26 August 2002. Volume II: Lexis and
Transmission. 2004. xii, 273 pp.
251 KAY, Christian J., Simon HOROBIN and Jeremy J. SMITH (eds.): New Perspectives on English
Historical Linguistics. Selected papers from 12 ICEHL, Glasgow, 21–26 August 2002. Volume I: Syntax and
Morphology. 2004. x, 264 pp.
250 JENSEN, John T.: Principles of Generative Phonology. An introduction. 2004. xii, 324 pp.
249 BOWERN, Claire and Harold KOCH (eds.): Australian Languages. Classification and the comparative
method. 2004. xii, 377 pp. (incl. CD-Rom).
248 WEIGAND, Edda (ed.): Emotion in Dialogic Interaction. Advances in the complex. 2004. xii, 284 pp.
247 PARKINSON, Dilworth B. and Samira FARWANEH (eds.): Perspectives on Arabic Linguistics. Papers
from the Annual Symposium on Arabic Linguistics. Volume XV: Salt Lake City 2001. 2003. x, 214 pp.
246 HOLISKY, Dee Ann and Kevin TUITE (eds.): Current Trends in Caucasian, East European and Inner
Asian Linguistics. Papers in honor of Howard I. Aronson. 2003. xxviii, 426 pp.
245 QUER, Josep, Jan SCHROTEN, Mauro SCORRETTI, Petra SLEEMAN and Els VERHEUGD (eds.):
Romance Languages and Linguistic Theory 2001. Selected papers from 'Going Romance', Amsterdam, 6–8
December 2001. 2003. viii, 355 pp.
244 PÉREZ-LEROUX, Ana Teresa and Yves ROBERGE (eds.): Romance Linguistics. Theory and Acquisition.
Selected papers from the 32nd Linguistic Symposium on Romance Languages (LSRL), Toronto, April 2002.
2003. viii, 388 pp.
243 CUYCKENS, Hubert, Thomas BERG, René DIRVEN and Klaus-Uwe PANTHER (eds.): Motivation in
Language. Studies in honor of Günter Radden. 2003. xxvi, 403 pp.
242 SEUREN, Pieter A.M. and Gerard KEMPEN (eds.): Verb Constructions in German and Dutch. 2003.
vi, 316 pp.
241 LECARME, Jacqueline (ed.): Research in Afroasiatic Grammar II. Selected papers from the Fifth
Conference on Afroasiatic Languages, Paris, 2000. 2003. viii, 550 pp.
240 JANSE, Mark and Sijmen TOL (eds.): Language Death and Language Maintenance. Theoretical, practical
and descriptive approaches. With the assistance of Vincent Hendriks. 2003. xviii, 244 pp.
239 ANDERSEN, Henning (ed.): Language Contacts in Prehistory. Studies in Stratigraphy. Papers from the
Workshop on Linguistic Stratigraphy and Prehistory at the Fifteenth International Conference on Historical
Linguistics, Melbourne, 17 August 2001. 2003. viii, 292 pp.
238 NÚÑEZ-CEDEÑO, Rafael, Luis LÓPEZ and Richard CAMERON (eds.): A Romance Perspective on
Language Knowledge and Use. Selected papers from the 31st Linguistic Symposium on Romance Languages
(LSRL), Chicago, 19–22 April 2001. 2003. xvi, 386 pp.
237 BLAKE, Barry J. and Kate BURRIDGE (eds.): Historical Linguistics 2001. Selected papers from the 15th
International Conference on Historical Linguistics, Melbourne, 13–17 August 2001. Editorial assistance Jo
Taylor. 2003. x, 444 pp.
236 SIMON-VANDENBERGEN, Anne-Marie, Miriam TAVERNIERS and Louise J. RAVELLI (eds.):
Grammatical Metaphor. Views from systemic functional linguistics. 2003. vi, 453 pp.
235 LINN, Andrew R. and Nicola McLELLAND (eds.): Standardization. Studies from the Germanic languages.
2002. xii, 258 pp.
234 WEIJER, Jeroen van de, Vincent J. van HEUVEN and Harry van der HULST (eds.): The Phonological
Spectrum. Volume II: Suprasegmental structure. 2003. x, 264 pp.
233 WEIJER, Jeroen van de, Vincent J. van HEUVEN and Harry van der HULST (eds.): The Phonological
Spectrum. Volume I: Segmental structure. 2003. x, 308 pp.
232 BEYSSADE, Claire, Reineke BOK-BENNEMA, Frank DRIJKONINGEN and Paola MONACHESI
(eds.): Romance Languages and Linguistic Theory 2000. Selected papers from ‘Going Romance’ 2000,
Utrecht, 30 November–2 December. 2002. viii, 354 pp.
231 CRAVENS, Thomas D.: Comparative Historical Dialectology. Italo-Romance clues to Ibero-Romance
sound change. 2002. xii, 163 pp.
230 PARKINSON, Dilworth B. and Elabbas BENMAMOUN (eds.): Perspectives on Arabic Linguistics.
Papers from the Annual Symposium on Arabic Linguistics. Volume XIII-XIV: Stanford, 1999 and Berkeley,
California 2000. 2002. xiv, 250 pp.
229 NEVIN, Bruce E. and Stephen B. JOHNSON (eds.): The Legacy of Zellig Harris. Language and
information into the 21st century. Volume 2: Mathematics and computability of language. 2002. xx, 312 pp.
228 NEVIN, Bruce E. (ed.): The Legacy of Zellig Harris. Language and information into the 21st century.
Volume 1: Philosophy of science, syntax and semantics. 2002. xxxvi, 323 pp.
227 FAVA, Elisabetta (ed.): Clinical Linguistics. Theory and applications in speech pathology and therapy. 2002.
xxiv, 353 pp.
226 LEVIN, Saul: Semitic and Indo-European. Volume II: Comparative morphology, syntax and phonetics.
2002. xviii, 592 pp.
225 SHAHIN, Kimary N.: Postvelar Harmony. 2003. viii, 344 pp.
224 FANEGO, Teresa, Belén MÉNDEZ-NAYA and Elena SEOANE (eds.): Sounds, Words, Texts and Change.
Selected papers from 11 ICEHL, Santiago de Compostela, 7–11 September 2000. Volume 2. 2002. x, 310 pp.
223 FANEGO, Teresa, Javier PÉREZ-GUERRA and María José LÓPEZ-COUSO (eds.): English Historical
Syntax and Morphology. Selected papers from 11 ICEHL, Santiago de Compostela, 7–11 September 2000.
Volume 1. 2002. x, 306 pp.
222 HERSCHENSOHN, Julia, Enrique MALLÉN and Karen ZAGONA (eds.): Features and Interfaces in
Romance. Essays in honor of Heles Contreras. 2001. xiv, 302 pp.
221 D’HULST, Yves, Johan ROORYCK and Jan SCHROTEN (eds.): Romance Languages and Linguistic
Theory 1999. Selected papers from ‘Going Romance’ 1999, Leiden, 9–11 December 1999. 2001. viii, 406 pp.
220 SATTERFIELD, Teresa, Christina TORTORA and Diana CRESTI (eds.): Current Issues in Romance
Languages. Selected papers from the 29th Linguistic Symposium on Romance Languages (LSRL), Ann
Arbor, 8–11 April 1999. 2002. viii, 412  pp.
219 ANDERSEN, Henning (ed.): Actualization. Linguistic Change in Progress. Papers from a workshop held
at the 14th International Conference on Historical Linguistics, Vancouver, B.C., 14 August 1999. 2001.
vii, 250 pp.
218 BENDJABALLAH, Sabrina, Wolfgang U. DRESSLER, Oskar E. PFEIFFER and Maria D. VOEIKOVA
(eds.): Morphology 2000. Selected papers from the 9th Morphology Meeting, Vienna, 24–28 February 2000.
2002. viii, 317 pp.
217 WILTSHIRE, Caroline R. and Joaquim CAMPS (eds.): Romance Phonology and Variation. Selected
papers from the 30th Linguistic Symposium on Romance Languages, Gainesville, Florida, February 2000.
2002. xii, 238 pp.
216 CAMPS, Joaquim and Caroline R. WILTSHIRE (eds.): Romance Syntax, Semantics and L2 Acquisition.
Selected papers from the 30th Linguistic Symposium on Romance Languages, Gainesville, Florida, February
2000. 2001. xii, 246 pp.
215 BRINTON, Laurel J. (ed.): Historical Linguistics 1999. Selected papers from the 14th International
Conference on Historical Linguistics, Vancouver, 9–13 August 1999. 2001. xii, 398 pp.
214 WEIGAND, Edda and Marcelo DASCAL (eds.): Negotiation and Power in Dialogic Interaction. 2001.
viii, 303 pp.
213 SORNICOLA, Rosanna, Erich POPPE and Ariel SHISHA-HALEVY (eds.): Stability, Variation and
Change of Word-Order Patterns over Time. With the assistance of Paola Como. 2000. xxxii, 323 pp.
212 REPETTI, Lori (ed.): Phonological Theory and the Dialects of Italy. 2000. x, 301 pp.
211 ELŠÍK, Viktor and Yaron MATRAS (eds.): Grammatical Relations in Romani. The Noun Phrase. with a
Foreword by Frans Plank (Universität Konstanz). 2000. x, 244 pp.
210 DWORKIN, Steven N. and Dieter WANNER (eds.): New Approaches to Old Problems. Issues in Romance
historical linguistics. 2000. xiv, 235 pp.
209 KING, Ruth: The Lexical Basis of Grammatical Borrowing. A Prince Edward Island French case study.
2000. xvi, 241 pp.
208 ROBINSON, Orrin W.: Whose German? The ach/ich alternation and related phenomena in ‘standard’ and
‘colloquial’. 2001. xii, 178 pp.
207 SANZ, Montserrat: Events and Predication. A new approach to syntactic processing in English and
Spanish. 2000. xiv, 219 pp.
206 FAWCETT, Robin P.: A Theory of Syntax for Systemic Functional Linguistics. 2000. xxviii, 360 pp.
205 DIRVEN, René, Roslyn M. FRANK and Cornelia ILIE (eds.): Language and Ideology. Volume 2:
descriptive cognitive approaches. 2001. vi, 264 pp.
204 DIRVEN, René, Bruce HAWKINS and Esra SANDIKCIOGLU (eds.): Language and Ideology. Volume 1:
theoretical cognitive approaches. 2001. vi, 301 pp.
203 NORRICK, Neal R.: Conversational Narrative. Storytelling in everyday talk. 2000. xiv, 233 pp.
202 LECARME, Jacqueline, Jean LOWENSTAMM and Ur SHLONSKY (eds.): Research in Afroasiatic
Grammar. Papers from the Third conference on Afroasiatic Languages, Sophia Antipolis, 1996. 2000.
vi, 386 pp.
201 DRESSLER, Wolfgang U., Oskar E. PFEIFFER, Markus A. PÖCHTRAGER and John R. RENNISON
(eds.): Morphological Analysis in Comparison. 2000. x, 261 pp.
200 ANTTILA, Raimo: Greek and Indo-European Etymology in Action. Proto-Indo-European *aǵ-. 2000.
xii, 314 pp.
199 PÜTZ, Martin and Marjolijn H. VERSPOOR (eds.): Explorations in Linguistic Relativity. 2000.
xvi, 369 pp.
198 NIEMEIER, Susanne and René DIRVEN (eds.): Evidence for Linguistic Relativity. 2000. xxii, 240 pp.
197 COOPMANS, Peter, Martin EVERAERT and Jane GRIMSHAW (eds.): Lexical Specification and
Insertion. 2000. xviii, 476 pp.
196 HANNAHS, S.J. and Mike DAVENPORT (eds.): Issues in Phonological Structure. Papers from an
International Workshop. 1999. xii, 268 pp.
195 HERRING, Susan C., Pieter van REENEN and Lene SCHØSLER (eds.): Textual Parameters in Older
Languages. 2001. x, 448 pp.
194 COLEMAN, Julie and Christian J. KAY (eds.): Lexicology, Semantics and Lexicography. Selected papers
from the Fourth G. L. Brook Symposium, Manchester, August 1998. 2000. xiv, 257 pp.
193 KLAUSENBURGER, Jurgen: Grammaticalization. Studies in Latin and Romance morphosyntax. 2000.
xiv, 184 pp.
192 ALEXANDROVA, Galina M. and Olga ARNAUDOVA (eds.): The Minimalist Parameter. Selected papers
from the Open Linguistics Forum, Ottawa, 21–23 March 1997. 2001. x, 360 pp.
191 SIHLER, Andrew L.: Language History. An introduction. 2000. xvi, 298 pp.
190 BENMAMOUN, Elabbas (ed.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on
Arabic Linguistics. Volume XII: Urbana-Champaign, Illinois, 1998. 1999. viii, 204 pp.
189 NICOLOV, Nicolas and Ruslan MITKOV (eds.): Recent Advances in Natural Language Processing II.
Selected papers from RANLP ’97. 2000. xi, 422 pp.
188 SIMMONS, Richard VanNess: Chinese Dialect Classification. A comparative approach to Harngjou, Old
Jintarn, and Common Northern Wu. 1999. xviii, 317 pp.
187 FRANCO, Jon A., Alazne LANDA and Juan MARTÍN (eds.): Grammatical Analyses in Basque and
Romance Linguistics. Papers in honor of Mario Saltarelli. 1999. viii, 306 pp.
186 MIŠESKA TOMIĆ, Olga and Milorad RADOVANOVIĆ (eds.): History and Perspectives of Language
Study. Papers in honor of Ranko Bugarski. 2000. xxii, 314 pp.
185 AUTHIER, Jean-Marc, Barbara E. BULLOCK and Lisa A. REED (eds.): Formal Perspectives on Romance
Linguistics. Selected papers from the 28th Linguistic Symposium on Romance Languages (LSRL XXVIII),
University Park, 16–19 April 1998. 1999. xii, 334 pp.
184 SAGART, Laurent: The Roots of Old Chinese. 1999. xii, 272 pp.
183 CONTINI-MORAVA, Ellen and Yishai TOBIN (eds.): Between Grammar and Lexicon. 2000.
xxxii, 365 pp.
182 KENESEI, István (ed.): Crossing Boundaries. Advances in the theory of Central and Eastern European
languages. 1999. viii, 302 pp.
181 MOHAMMAD, Mohammad A.: Word Order, Agreement and Pronominalization in Standard and
Palestinian Arabic. 2000. xvi, 197 pp.
180 MEREU, Lunella (ed.): Boundaries of Morphology and Syntax. 1999. viii, 314 pp.
179 RINI, Joel: Exploring the Role of Morphology in the Evolution of Spanish. 1999. xvi, 187 pp.
178 FOOLEN, Ad and Frederike van der LEEK (eds.): Constructions in Cognitive Linguistics. Selected papers
from the Fifth International Cognitive Linguistics Conference, Amsterdam, 1997. 2000. xvi, 338 pp.
177 CUYCKENS, Hubert and Britta E. ZAWADA (eds.): Polysemy in Cognitive Linguistics. Selected papers
from the International Cognitive Linguistics Conference, Amsterdam, 1997. 2001. xxviii, 296 pp.
176 VAN HOEK, Karen, Andrej A. KIBRIK and Leo NOORDMAN (eds.): Discourse Studies in Cognitive
Linguistics. Selected papers from the 5th International Cognitive Linguistics Conference, Amsterdam, July
1997. 1999. vi, 187 pp.
175 GIBBS, JR., Raymond W. and Gerard J. STEEN (eds.): Metaphor in Cognitive Linguistics. Selected papers
from the 5th International Cognitive Linguistics Conference, Amsterdam, 1997. 1999. viii, 226 pp.
174 HALL, T. Alan and Ursula KLEINHENZ (eds.): Studies on the Phonological Word. 1999. viii, 298 pp.
173 TREVIÑO, Esthela and José LEMA (eds.): Semantic Issues in Romance Syntax. 1999. viii, 309 pp.
172 DIMITROVA-VULCHANOVA, Mila and Lars HELLAN (eds.): Topics in South Slavic Syntax and
Semantics. 1999. xxviii, 263 pp.
171 WEIGAND, Edda (ed.): Contrastive Lexical Semantics. 1998. x, 270 pp.
170 LAMB, Sydney M.: Pathways of the Brain. The neurocognitive basis of language. 1999. xii, 418 pp.
169 GHADESSY, Mohsen (ed.): Text and Context in Functional Linguistics. 1999. xviii, 340 pp.
168 RATCLIFFE, Robert R.: The “Broken” Plural Problem in Arabic and Comparative Semitic. Allomorphy and
analogy in non-concatenative morphology. 1998. xii, 261 pp.
167 BENMAMOUN, Elabbas, Mushira EID and Niloofar HAERI (eds.): Perspectives on Arabic Linguistics.
Papers from the Annual Symposium on Arabic Linguistics. Volume XI: Atlanta, Georgia, 1997. 1998.
viii, 231 pp.
166 LEMMENS, Maarten: Lexical Perspectives on Transitivity and Ergativity. Causative constructions in
English. 1998. xii, 268 pp.
165 BUBENIK, Vit: A Historical Syntax of Late Middle Indo-Aryan (Apabhraṃśa). 1998. xxiv, 265 pp.
164 SCHMID, Monika S., Jennifer R. AUSTIN and Dieter STEIN (eds.): Historical Linguistics 1997. Selected
papers from the 13th International Conference on Historical Linguistics, Düsseldorf, 10–17 August 1997.
1998. x, 409 pp.
163 LOCKWOOD, David G., Peter H. FRIES and James E. COPELAND (eds.): Functional Approaches to
Language, Culture and Cognition. Papers in honor of Sydney M. Lamb. 2000. xxxiv, 656 pp.
162 HOGG, Richard M. and Linda van BERGEN (eds.): Historical Linguistics 1995. Volume 2: Germanic
linguistics.. Selected papers from the 12th International Conference on Historical Linguistics, Manchester,
August 1995. 1998. x, 365 pp.
161 SMITH, John Charles and Delia BENTLEY (eds.): Historical Linguistics 1995. Volume 1: General issues
and non-Germanic Languages.. Selected papers from the 12th International Conference on Historical
Linguistics, Manchester, August 1995. 2000. xii, 438 pp.
160 SCHWEGLER, Armin, Bernard TRANEL and Myriam URIBE-ETXEBARRIA (eds.): Romance
Linguistics: Theoretical Perspectives. Selected papers from the 27th Linguistic Symposium on Romance
Languages (LSRL XXVII), Irvine, 20–22 February, 1997. 1998. vi, 349 pp. + index.
159 JOSEPH, Brian D., Geoffrey C. HORROCKS and Irene PHILIPPAKI-WARBURTON (eds.): Themes in
Greek Linguistics II. 1998. x, 335 pp.
158 SÁNCHEZ-MACARRO, Antonia and Ronald CARTER (eds.): Linguistic Choice across Genres. Variation
in spoken and written English. 1998. viii, 338 pp.
157 LEMA, José and Esthela TREVIÑO (eds.): Theoretical Analyses on Romance Languages. Selected papers
from the 26th Linguistic Symposium on Romance Languages (LSRL XXVI), Mexico City, 28–30 March,
1996. 1998. viii, 380 pp.
156 MATRAS, Yaron, Peter BAKKER and Hristo KYUCHUKOV (eds.): The Typology and Dialectology of
Romani. 1997. xxxii, 223 pp.
155 FORGET, Danielle, Paul HIRSCHBÜHLER, France MARTINEAU and María Luisa RIVERO (eds.):
Negation and Polarity. Syntax and semantics. Selected papers from the colloquium Negation: Syntax and
Semantics. Ottawa, 11–13 May 1995. 1997. viii, 367 pp.
154 SIMON-VANDENBERGEN, Anne-Marie, Kristin DAVIDSE and Dirk NOËL (eds.): Reconnecting
Language. Morphology and Syntax in Functional Perspectives. 1997. xiii, 339 pp.
153 EID, Mushira and Robert R. RATCLIFFE (eds.): Perspectives on Arabic Linguistics. Papers from the
Annual Symposium on Arabic Linguistics. Volume X: Salt Lake City, 1996. 1997. vii, 296 pp.
152 HIRAGA, Masako K., Chris SINHA and Sherman WILCOX (eds.): Cultural, Psychological and
Typological Issues in Cognitive Linguistics. Selected papers of the bi-annual ICLA meeting in Albuquerque,
July 1995. 1999. viii, 338 pp.
151 LIEBERT, Wolf-Andreas, Gisela REDEKER and Linda R. WAUGH (eds.): Discourse and Perspective in
Cognitive Linguistics. 1997. xiv, 270 pp.
150 VERSPOOR, Marjolijn H., Kee Dong LEE and Eve SWEETSER (eds.): Lexical and Syntactical
Constructions and the Construction of Meaning. Proceedings of the bi-annual ICLA meeting in
Albuquerque, July 1995. 1997. xii, 454 pp.
149 HALL, T. Alan: The Phonology of Coronals. 1997. x, 176 pp.
148 WOLF, George and Nigel LOVE (eds.): Linguistics Inside Out. Roy Harris and his critics. 1997.
xxviii, 344 pp.
147 HEWSON, John: The Cognitive System of the French Verb. 1997. xii, 187 pp.
146 HINSKENS, Frans, Roeland van HOUT and W. Leo WETZELS (eds.): Variation, Change, and
Phonological Theory. 1997. x, 314 pp.
145 HEWSON, John and Vit BUBENIK: Tense and Aspect in Indo-European Languages. Theory, typology,
diachrony. 1997. xii, 403 pp.
144 SINGH, Rajendra (ed.): Trubetzkoy's Orphan. Proceedings of the Montréal Roundtable on
“Morphonology: contemporary responses” (Montréal, October 1994). In collaboration with Richard
Desrochers. 1996. xiv, 363 pp.
143 ATHANASIADOU, Angeliki and René DIRVEN (eds.): On Conditionals Again. 1997. viii, 418 pp.
142 SALMONS, Joseph C. and Brian D. JOSEPH (eds.): Nostratic. Sifting the Evidence. 1998. vi, 293 pp.
141 EID, Mushira and Dilworth B. PARKINSON (eds.): Perspectives on Arabic Linguistics. Papers from the
Annual Symposium on Arabic Linguistics. Volume IX: Washington D.C., 1995. 1996. xiii, 249 pp.
140 BLACK, James R. and Virginia MOTAPANYANE (eds.): Clitics, Pronouns and Movement. 1997. 375 pp.
139 BLACK, James R. and Virginia MOTAPANYANE (eds.): Microparametric Syntax and Dialect Variation.
1996. xviii, 269 pp.
138 SACKMANN, Robin and Monika BUDDE (eds.): Theoretical Linguistics and Grammatical Description.
Papers in honour of Hans-Heinrich Lieb. 1996. x, 375 pp.
137 LIPPI-GREEN, Rosina L. and Joseph C. SALMONS (eds.): Germanic Linguistics. Syntactic and
diachronic. 1996. viii, 192 pp.
136 MITKOV, Ruslan and Nicolas NICOLOV (eds.): Recent Advances in Natural Language Processing.
Selected Papers from RANLP ’95. 1997. xii, 472 pp.
135 BRITTON, Derek (ed.): English Historical Linguistics 1994. Papers from the 8th International Conference
on English Historical Linguistics (8 ICEHL, Edinburgh, 19–23 September 1994). 1996. viii, 403 pp.
134 EID, Mushira (ed.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic
Linguistics. Volume VIII: Amherst, Massachusetts 1994. 1996. vii, 261 pp.
133 ZAGONA, Karen (ed.): Grammatical Theory and Romance Languages. Selected papers from the 25th
Linguistic Symposium on Romance Languages (LSRL XXV) Seattle, 2–4 March 1995. 1996. vi, 330 pp.
132 HERSCHENSOHN, Julia: Case Suspension and Binary Complement Structure in French. 1996. xi, 200 pp.
131 HUALDE, José Ignacio, Joseba A. LAKARRA and R.L. TRASK (eds.): Towards a History of the Basque
Language. 1996. 365 pp.
130 EID, Mushira (ed.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic
Linguistics. Volume VII: Austin, Texas 1993. 1995. vii, 192 pp.
129 LEVIN, Saul: Semitic and Indo-European. Volume I: The Principal Etymologies. With observations on
Afro-Asiatic. 1995. xxii, 514 pp.
128 GUY, Gregory R., Crawford FEAGIN, Deborah SCHIFFRIN and John BAUGH (eds.): Towards a
Social Science of Language. Papers in honor of William Labov. Volume 2: Social interaction and discourse
structures. 1997. xviii, 358 pp.
127 GUY, Gregory R., Crawford FEAGIN, Deborah SCHIFFRIN and John BAUGH (eds.): Towards a Social
Science of Language. Papers in honor of William Labov. Volume 1: Variation and change in language and
society. 1996. xviii, 436 pp.
126 MATRAS, Yaron (ed.): Romani in Contact. The history, structure and sociology of a language. 1995.
xvii, 208 pp.
125 SINGH, Rajendra (ed.): Towards a Critical Sociolinguistics. 1996. xiii, 342 pp.
124 ANDERSEN, Henning (ed.): Historical Linguistics 1993. Selected papers from the 11th International
Conference on Historical Linguistics, Los Angeles, 16–20 August 1993. 1995. x, 460 pp.
123 AMASTAE, Jon, Grant GOODALL, M. MONTALBETTI and M. PHINNEY (eds.): Contemporary
Research in Romance Linguistics. Papers from the XXII Linguistic Symposium on Romance Languages, El
Paso/Juárez, February 22–24, 1992. 1995. viii, 381 pp.
122 SMITH, John Charles and Martin MAIDEN (eds.): Linguistic Theory and the Romance Languages. 1995.
xiii, 240 pp.
121 HASAN, Ruqaiya, Carmel CLORAN and David G. BUTT (eds.): Functional Descriptions. Theory in
practice. 1996. xxxvi, 381 pp.
120 STONHAM, John T.: Combinatorial Morphology. 1994. xii, 206 pp.
119 LIPPI-GREEN, Rosina L.: Language Ideology and Language Change in Early Modern German. A
sociolinguistic study of the consonantal system of Nuremberg. 1994. xiv, 150 pp.
118 HASAN, Ruqaiya and Peter H. FRIES (eds.): On Subject and Theme. A discourse functional perspective.
1995. xii, 414 pp.
117 PHILIPPAKI-WARBURTON, Irene, Katerina NICOLAIDIS and Maria SIFIANOU (eds.): Themes in
Greek Linguistics. Papers from the First International Conference on Greek Linguistics, Reading, September
1993. 1994. xviii, 534 pp.
116 MILLER, D. Gary: Ancient Scripts and Phonological Knowledge. 1994. xvi, 139 pp.
115 EID, Mushira, Vicente CANTARINO and Keith WALTERS (eds.): Perspectives on Arabic Linguistics.
Papers from the Annual Symposium on Arabic Linguistics. Volume VI: Columbus, Ohio 1992. 1994.
viii, 238 pp.
114 EGLI, Urs, Peter E. PAUSE, Christoph SCHWARZE, Arnim von STECHOW and Götz WIENOLD
(eds.): Lexical Knowledge in the Organization of Language. 1995. xiv, 367 pp.
113 MORENO FERNÁNDEZ, Francisco, Miguel FUSTER and Juan Jose CALVO (eds.): English Historical
Linguistics 1992. Papers from the 7th International Conference on English Historical Linguistics, Valencia,
22–26 September 1992. 1994. viii, 388 pp.
112 CULIOLI, Antoine: Cognition and Representation in Linguistic Theory. Texts selected, edited and
introduced by Michel Liddle. Translated with the assistance of John T. Stonham. 1995. x, 161 pp.
111 TOBIN, Yishai: Invariance, Markedness and Distinctive Feature Analysis. A contrastive study of sign
systems in English and Hebrew. 1994. xxii, 406 pp.
110 SIMONE, Raffaele (ed.): Iconicity in Language. 1995. xii, 315 pp.
109 PAGLIUCA, William (ed.): Perspectives on Grammaticalization. 1994. xx, 306 pp.
108 LIEB, Hans-Heinrich: Linguistic Variables. Towards a unified theory of linguistic variation. 1993.
xiv, 261 pp.
107 MARLE, Jaap van (ed.): Historical Linguistics 1991. Papers from the 10th International Conference on
Historical Linguistics, Amsterdam, August 12–16, 1991. 1993. xviii, 395 pp.
106 AERTSEN, Henk and Robert J. JEFFERS (eds.): Historical Linguistics 1989. Papers from the 9th
International Conference on Historical Linguistics, New Brunswick, 14–18 August 1989. 1993. xviii, 538 pp.
105 HUALDE, José Ignacio and Jon Ortiz de URBINA (eds.): Generative Studies in Basque Linguistics. 1993.
vi, 334 pp.
104 KURZOVÁ, Helena: From Indo-European to Latin. The evolution of a morphosyntactic type. 1993.
xiv, 259 pp.
103 ASHBY, William J., Marianne MITHUN and Giorgio PERISSINOTTO (eds.): Linguistic Perspectives on
Romance Languages. Selected Papers from the XXI Linguistic Symposium on Romance Languages, Santa
Barbara, February 21–24, 1991. 1993. xxii, 404 pp.
102 DAVIS, Philip W. (ed.): Alternative Linguistics. Descriptive and theoretical modes. 1996. vii, 325 pp.
101 EID, Mushira and Clive HOLES (eds.): Perspectives on Arabic Linguistics. Papers from the Annual
Symposium on Arabic Linguistics. Volume V: Ann Arbor, Michigan 1991. 1993. viii, 347 pp.
100 MUFWENE, Salikoko S. and Lioba MOSHI (eds.): Topics in African Linguistics. Papers from the XXI
Annual Conference on African Linguistics, University of Georgia, April 1990. 1993. x, 304 pp.
99 JENSEN, John T.: English Phonology. 1993. x, 251 pp.
98 EID, Mushira and Gregory K. IVERSON (eds.): Principles and Prediction. The analysis of natural
language. Papers in honor of Gerald Sanders. 1993. xix, 382 pp.