Kayne 1996

MICROPARAMETRIC SYNTAX AND DIALECT VARIATION
AMSTERDAM STUDIES IN THE THEORY AND

HISTORY OF LINGUISTIC SCIENCE
General Editor
E.F. KONRAD KOERNER
(University of Ottawa)
Series IV - CURRENT ISSUES IN LINGUISTIC THEORY
Advisory Editorial Board
Henning Andersen (Los Angeles); Raimo Anttila (Los Angeles)

Thomas V. Gamkrelidze (Tbilisi); John E. Joseph (Kong Kong)
Hans-Heinrich Lieb (Berlin); Ernst Pulgram (Ann Arbor, Mich.)
E. Wyn Roberts (Vancouver, B.C.); Danny Steinberg (Tokyo)
Volume 139
James R. Black and Virginia Motapanyane (eds)
Microparametric Syntax and Dialect Variation

MICROPARAMETRIC SYNTAX
AND DIALECT VARIATION
Edited by
JAMES R. BLACK
Memorial University of Newfoundland
VIRGINIA MOTAPANYANE
University of New Brunswick
JOHN BENJAMINS PUBLISHING COMPANY

AMSTERDAM/PHILADELPHIA
The paper used in this publication meets the minimum requirements of
American National Standard for Information Sciences — Permanence of
Paper for Printed Library Materials, ANSI Z39.48-1984.
Library of Congress Cataloging-in-Publication Data

Microparametric syntax and dialect variation / edited by James R. Black and Virginia
Motapanyane.
p. cm. - (Amsterdam studies in the theory and history of linguistic science. Series
IV, Current issues in linguistic theory, ISSN 0304-0763 ; v. 139)
Includes one contribution in French.
Includes bibliographical references and index.
Contents: Microparametric syntax : some introductory remarks / Richard S. Kayne - Une
analyse microparamétrique des moyens dans les langues romanes / J.-Marc Authier & Lisa
Reed - Treating that-trace variation / Philip Branigan - Negative particle questions / Lisa L.-
S. Cheng, C-T. James Huang & C.-C. Jane Tang - Imperative inversion in Belfast English /
Alison Henry - Scandinavian possessive constructions from a northern Swedish viewpoint /
Anders Holmberg & Görel Sandström - The occasional absence of anaphoric agreement in
Labrador Inuttut / Alana Johns - Hypothetical infinitives and crosslinguistic variations in
continental and Quebec French / France Martineau & Virginia Motapanyane - The second
person singular interrogative in the traditional vernacular of the Bolton metropolitan area /
Graham Shorrocks - Reflexives, pronouns, and subject/verb agreement in Icelandic and
Faroese / Knut Taraid Taraldsen - Adverbial quantifiers and dialectal variation in a
minimalist framework I Marie-Thérèse Vinet - Verb clusters in continental West Germanic
dialects / Jan-Wouter Zwart.
1. Grammar, Comparative and general—Syntax. 2. Language and languages-Variation. 3.
Linguistic geography. I. Black, James R. II. Motapanyane, Virginia. III. Series.
P291.M49 1996
415--dc20 96-38230
ISBN 90 272 3643 7 (Eur.) / 1-55619-594-X (US) (alk. paper) CIP
© Copyright 1996 - 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 75577 · 1070 AN Amsterdam · The Netherlands
John Benjamins North America · P.O.Box 27519 · Philadelphia PA 19118-0519 · USA
CONTENTS
Richard S. Kayne
Microparametric Syntax: Some Introductory Remarks ix
Jean-Marc Authier & Lisa Reed
Une analyse microparamétrique des moyens
dans les langues romanes 1
Philip Branigan
Tracing that-trace Variation 25
Lisa Cheng, James Huang & Jane Tang
Negative Particle Questions: A Dialectal Comparison 41
Alison Henry
Imperative Inversion in Belfast English 79
Anders Holmberg & Görel Sandström
Scandinavian Possessive Constructions from a
Northern Swedish Viewpoint 95
Alana Johns
The Anaphoric Agreement Morpheme in Labrador Inuttut 121
France Martineau & Virginia Motapanyane
Hypothetical Infinitives and Crosslinguistic Variation in
Continental and Québec French 145
Graham Shorrocks
The Second Person Singular Interrogative in the
Traditional Vernacular of the Bolton Metropolitan Area 169
Knut Taraid Taraldsen
Reflexives, Pronouns and Subject/V Agreement in
Icelandic and Faroese 189
Marie-Thérèse Vinet
Adverbial Quantifiers and Dialectal Variation in a
Minimalist Framework 213
Jan-Wouter Zwart
Verb Clusters in Continental West Germanic Dialects 229
General Index 259
PREFATORY NOTE
The papers in this volume originated as presentations at the 18th annual

meeting of the Atlantic Provinces Linguistic Association, held 28-29 October
1994 in Saint John, New Brunswick, Canada.
Selected papers on the conference theme of Microparametric Syntax and
Dialect Variation were subsequently refereed and revised in light of reviewers'
comments.
The editors gratefully acknowledge the generous support of the University
of New Brunswick - Saint John and of the Social Sciences and Humanities
Research Council of Canada for providing grants towards the costs of the con
ference and the publication of papers.
We also acknowledge with gratitude the help so freely and generously of
fered by linguists asked to comment on these papers: without their thorough and
conscientious assistance, this enterprise would have been impossible. Special
thanks as well to Janis Black for careful proofreading and many helpful sugges
tions.
JAMES R. BLACK
Department of Linguistics
St. John's, NF, Al 3X9, Canada
VIRGINIA MOTAPANYANE
Humanities
University of New Brunswick - Saint John
Saint John, NB, E2L 4L5, Canada
June 1996
MICROPARAMETRIC SYNTAX:
SOME INTRODUCTORY REMARKS
RICHARD S. KAYNE
Graduate Center, City University of New York
Comparative syntax can be thought of as that facet of syntactic theory di

rectly concerned with the question of how best to characterize the properties of
human languages that are not universal. Put another way, comparative syntax di
rectly addresses the question of how best to understand the notion of parameter
taken to underlie syntactic variation.
The study of differences among languages must obviously proceed in tan
dem with the study of what they have in common, i.e., with the study of the
principles of Universal Grammar (UG) that interact with language specific pa
rameters to yield observed variation. Similarly, there is every reason to believe
that the search for universal syntactic principles cannot proceed without close
attention being paid to syntactic variation.
At its most successful, comparative syntax simultaneously achieves two
primary kinds of results: it accounts for observed clusterings of syntactic pro
perties by showing that the several properties in question can all be traced back
to a single relatively more abstract parameter setting. And it shows that that op
timal parametric account depends on particular assumptions about the proper
formulation or understanding of the principles of universal grammar. In the latter
way, comparative syntax provides evidence bearing on questions which are not
themselves comparative in nature.
My own work in comparative syntax was at first limited to problems arising
from a comparison of French and English. I argued, for example, that the ab
sence in French of the so-called Exceptional Case Marking (ECM) construction
{John considers Bill to have been mistaken) need not, as Chomsky had thought,
be seen as an irreducible difference between the two languages.1 Rather, that
French-English difference should be related to others involving prepositions and
l
Cf. Chomsky (1980, 1981) and Kayne (1981).
 RICHARD S. KAYNE
prepositional complementizers,2 and ultimately be derived from an abstract dif

ference in the governing properties of prepositions in the two languages.
If correct or largely correct, this clustering of properties (involving ECM
constructions, for + lexical subject, preposition stranding and double object
constructions) provides evidence in favor of a certain approach to ECM con
structions (with an essential role played by an abstract prepositional head) and
against various others (e.g., the S-bar deletion approach), and thereby tells us
something about how UG is put together that we might not have discovered
without doing this kind of comparative work.
In the early to mid-eighties, it became apparent to me that a direct compari
son of French and English raised difficult problems to a greater extent than direct
comparison of French and Italian. In essence, in searching for clusters of pro
perties, one must make decisions about what syntactic differences can plausibly
be linked to what other syntactic differences. To a certain extent one is guided by
one's knowledge of syntax in general and by the theory within the framework of
which one is working. Such general considerations do place limits on the set of
hypotheses one takes seriously, but typically the set of plausible linkings re
mains larger than one would like. The size of that set will of course be affected
by the number of syntactic differences there are between the two languages in
question. The more there are to begin with, the harder it will be, all other things
being equal, to figure out the correct linkings.
A related point can be made by considering the question of how one goes
about testing one's comparative syntax hypotheses. Any proposal of the form
'These two (or three...) differences between the languages in question are re
lated to one another' can in principle be tested indirectly by examining the theo
retical consequences of the hypotheses about parameters and about universal
principles that flow from the original claim concerning the related differences.
But there is in addition a more direct way of testing such a claim, and that is to
examine other languages to see if the particular properties in question actually do
systematically cluster together. In pursuing this kind of examination, one finds
that hypotheses concerning French vs. English lead very directly to questions
concerning both other Romance languages and other Germanic languages.
(Languages further afield can sometimes be relevant, but they can also be too
different—for example, a hypothesis concerning past participles cannot be tested
in any direct manner in a language that has no past participles.) Hypotheses con
cerning French vs. Italian often do not immediately lead beyond the Romance
2
A related proposal concerning double object constructions is given in Kayne (1983). The ab
sence of double object constructions in French and Italian is not expected from the perspective
of Collins & Thráinsson (1993), whose analysis lacks the crucial abstract preposition.
MICROPARAMETRIC SYNTAX: SOME INTRODUCTORY REMARKS xi
languages: in this sense, the hypothesis testing task generated by French-Italian

comparison may be more manageable than the one generated by French-English
comparison.
This advantage of French-Italian over French-English comparative work led
me to concentrate on the former pair. I attempted, for example, to relate certain
differences between French and Italian past participle agreement to other aspects
of their syntax, in particular to a difference between them concerning passives of
causatives, and to another concerning inversion and null subjects.3
Although French and Italian are relatively similar syntactically, they still
show a rather large number of syntactic differences, so that the task of deciding
which pairs or sets of differences are significantly related to one another, and
therefore to be thought of as being traceable back to a single parametric differ
ence between the two languages is, even if somewhat easier than in the case of
French-English, still a difficult one.
It is consequently natural to take into comparative account languages that are
syntactically closer to French or Italian (or to each other) than French is to
Italian. And comparative syntax hypotheses, including those originating with
work on French and Italian, do turn out to be testable with great profit on the
less well-known dialects of France and Italy,4 the study of which permits the
discovery of clusterings of syntactic properties that otherwise would in all likeli
hood go unnoticed.
There is, for example, a strong correlation across these languages between
the possibility of having (some) null subjects and the possibility of having clitic
climbing out of an infinitival clause into the matrix (non-causative) clause; re
lated to this is movement of a clitic across a pre-infinitival adverbial).5 If correct
to a significant degree, this correlation supports decomposing clitic climbing into
several steps, in the spirit of successive cyclicity. A decompositional approach
receives additional support from the existence of 'clitic splitting', where two cli
tics originating as complements of the same infinitive can surface in two differ
ent positions. Clitic splitting in non-causatives had been thought not to exist in
Romance, but it turns out that there are a number of little known Romance lan
guages that do have it. It appears, however, to be limited to those Romance lan
guages with clitic-infinitive order.6
3 Cf. Kayne (1985).

4 Also, Belgium, Quebec and Switzerland.
5 Cf. Kayne (1989a).
6 At least one variety of Milanese seems to allow split clitics despite being an infinitive-clitic
language. This (yet to be understood) exception is probably correlated to the property of allow
ing a copy of the 'climbed' clitic to follow the embedded infinitive.
X11 RICHARD S. KAYNE
There is another correlation in Romance that involves the order between clitic
and infinitive: a Romance language/dialect will allow the counterpart of English
*John doesn 't know if to go to the movies tonight (with controlled PRO in the
presence of if), if and only if it has the order infinitive-clitic.7 This correlation,
particularly striking if one looks at little studied (from a syntactic point of view)
null subject languages that have clitic-infinitive order (Sardinian, Occitan,
Gardenese), appears to favor over other approaches an approach to PRO based
on a modified version of Chomsky's (1986) analysis, and to support (a certain
revision of) principles A and  of his binding theory.
Italian, although it almost invariably shows infinitive-clitic order, has an ap
parently anomalous pre-infinitival clitic in infinitival imperatives. This anomaly
can be made sense of by taking advantage of a correlation that holds quite
strongly across the dialects of Italy between pre-infinitival clitics in these im
peratives and clitic climbing of the familiar sort. The analysis called for involves
an abstract auxiliary to which to clitics can raise in the manner of clitic climb
ing. 8
The study of Italian dialects also radically changes one's perception of
Romance auxiliary selection. The picture and theory of auxiliary selection that
one arrives at by studying the most widely spoken Romance languages must be
considerably modified and enriched if one is to account for the remarkable di
versity found in the dialects.9
Comparative work on the syntax of a large number of closely related lan
guages can be thought of as a new research tool, one that is capable of providing
results of an unusually fine-grained and particularly solid character.10 If it were
possible to experiment on languages, a syntactician would construct an experi
ment of the following type: take a language, alter a single one of its observable
syntactic properties, examine the result to see what, if any, other property has
changed as a consequence of the original manipulation. If one has, interpret that
result as indicating that it and the original property that was altered are linked to
one another by some abstract parameter.
Although such experiments cannot be performed, I think that by examining
pairs (and larger sets) of ever more closely related languages, one can begin to
7
Cf. Kayne (1991).
8
Cf. Kayne (1992).
9
Cf. Kayne (1993).
10
As illustrated in particular by the various articles in this volume, most of which treat
Romance or Germanic languages. Johns' article on Inuttut and Cheng, Huang & Tang's article
on Chinese demonstrate the importance of microparametric syntax for other language families.
I make no attempt in these introductory remarks at a systematic survey of microparametric
work.
MICROPARAMETRIC SYNTAX: SOME INTRODUCTORY REMARKS xiii
approximate the results of such an experiment. To the extent that one can find
languages that are syntactically extremely similar to one another, yet clearly dis
tinguishable and readily examinable, one can hope to reach a point such that the
number of observable differences is so small that one can virtually see one pro
perty covarying with another.11
In addition to facilitating the accurate individuation of parameters and of the
principles of Universal Grammar required to interact with them, the technique of
examining a large number of very closely related languages promises to provide
a broad understanding of parameters at their finest-grained (microparameters),
i.e., to provide a handle on the question: what are the minimal units of syntactic
variation?
Consider, for example, the phenomenon of past participle agreement. Study
of the better known Romance languages shows past participle agreement with
the subject in SPEC-IP in passives to be general, perhaps exceptionless. In active
sentences, on the other hand, past participle agreement is found in French and
Italian, and to some extent in Catalan, but not in Spanish or Portuguese. On the
basis of these, one might postulate a parameter one setting of which allows past
participle agreement in actives.
Yet consideration of lesser known Romance languages indicates that there
must be parameters with finer-grained effects, to account for facts such as the
following: some of these languages allow past participle agreement in both WH-
and clitic constructions, some only in the latter. No Romance language/dialect,
as far as I can see, has past participle agreement with WH-phrases but not with
object clitics.12
Some allow past participle agreement with all direct object clitics, but some,
while having it with third person clitics, prohibit it with first and second person
clitics. Some allow past participle agreement with the partitive clitic, others do
not. Of those that allow clitics to follow past participles in the auxiliary-past par
ticiple construction, some allow past participle agreement only when the clitic
has moved up to the auxiliary, while others are freer. As far as I know, all
Romance languages with past participle agreement in actives allow such agree
ment in at least some reflexive clitic constructions when the auxiliary is the
equivalent of be. When there is a reflexive clitic in addition to an accusative
clitic, some require agreement with the accusative, others do not. When the
auxiliary is have, some prohibit agreement with the reflexive clitic (or its an
tecedent) completely, while others do not.
11
In the extreme case, one may find an isolated property distinguishing two very close di
alects—cf. Henry's article in this volume.
12
Cf. Kayne (1989b).
XIV RICHARD S. KAYNE
Our understanding of all of the above points will benefit from a closer look
at even more of these languages/dialects. In many of these cases, it is not clear
yet what the exact form of the relevant parameters will be, nor whether they will
concern agreement per se, or whether they will more centrally involve properties
of the pronouns/clitics, or of the auxiliaries, or of the participles, or some com
bination of these.
It seems reasonable to expect work in microparametric syntax to play a privi
leged role in the future in answering the more general question concerning the
form that syntactic parameters may take.13 Chomsky's recent work, for exam
ple, suggests the possibility that all syntactic variation might be expressible in
terms of strong/weak features on various functional heads; 14 microparametric
work will enable us to test this kind of hypothesis in a particularly interesting
way.
In the preceding discussion, I have assumed that the enormous amount of
syntactic variation that can be observed even within the set of Romance lan
guages/dialects lends itself to insightful characterization in terms of the notion
'parameter' as it has developed over the past fifteen or twenty years.
Alternatively put, we can take the study of microparametric variation to provide
an ideal testing ground for the very hypothesis that syntactic variation can be re
duced to a finite set of parameters (interacting with a set of universal principles).
Related to this is the question of how many irreducible syntactic parameters
there really are. Again, work in microparametric syntax should be invaluable,
and should begin to give us some sense of a lower bound for the number of pa
rameters (which in turn will bear on questions of learnability/acquisition). It is
also clear that the study of minimal syntactic variation is bound to provide crucial
evidence bearing on questions of diachronic syntax (which involves the study of
the minimally different stages in the evolution of the syntax of a language).
The question of the number of syntactic parameters leads in turn to the ques
tion of the number of syntactically distinct languages/dialects. To begin with, I
take it for granted that there is no syntactically significant distinction to be drawn
between 'language' and 'dialect' and no justification for neglecting the latter.
Now it is often estimated that the number of languages presently in existence is
4000-5000.15 Such estimates must evaluate the contribution of Italy as one.
Yet Renzi and Vanelli (1983) showed that in Northern Italy alone one can
13
The term 'micro-comparative' was used by Hellan & Christensen (1986). For discussion of
(microparameters vs.) macroparameters, cf. Baker (1996).
14
Cf. Chomsky (1995). In a general way, this seems compatible with the implications of the
approach of Kayne (1994) to word order variation.
15
For some discussion, cf. Comrie (1987:2-5) and Crystal (1987:284-285).
MICROPARAMETRIC SYNTAX: SOME INTRODUCTORY REMARKS XV
individuate at least 25 syntactically distinct languages/dialects solely by studying

the syntax of subject clitics. More recently, I have had the privilege of participat
ing in a Padua-based syntactic atlas/ (micro)comparative syntax project with
Paola Benincà,16 Cecilia Poletto, and Laura Vanelli, on the basis of which it is
evident that one can easily individuate at least 100 syntactically distinct lan
guages/dialects in Northern Italy.17 A very conservative estimate would be that
present-day Italy has at least 500 syntactically distinct languages/dialects.
500,000 would in consequence, I think, then be a very conservative extrapola
tion to the number of syntactically distinct languages/dialects in the world at pre
sent.
It is possible to arrive at a much more radical reevaluation based on the fol
lowing question: can anyone think of another person with whom they agree
100% of the time on syntactic judgments (even counting only sharp disagree
ments)? Or, more precisely, are there any two people who have exactly the same
syntactic judgments without exception?
By the nature of the question, it is not possible to give a positive answer,
since one could never be sure that two people who seemed to agree with each
other consistently would not disagree (sharply) on some judgment at some sub
sequent time. On the other hand, it is easy to think of many pairs of English
speakers, for example, who do differ sharply on some set of judgments. For
such pairs, the seemingly innocuous conclusion is that they do not speak exactly
the same variety of English.
Now we know that there are distinct varieties of English—many syntactic
differences have been discussed that distinguish American from British
English.18 And various regional syntactic differences within the United States or
within the United Kingdom are well known.19 But what if it turned out that for
every single pair of English speakers (and similarly for other languages) one
could find at least one clear syntactic difference?
My own experience in observing the syntax of English speakers, both lin
guists and non-linguists, makes me think that it is entirely likely that no two
speakers of English have exactly the same syntactic judgments. In which case
there must be many more varieties of English than is usually assumed. In fact, if
it is true that no two English speakers have the same (syntactic) grammar,20 then
16
Cf. Benincà (1994).
17
Cf. Poletto (1995).
18
Cf., for example, Zandwoort (1965:343), Merat (1974), Johansson (1979) and Trudgill &
Hannah (1994:56-82).
19
Cf., for example, Klima (1964), Trudgill & Chambers (1991); also Henry (1995).
20
Here and elsewhere, I gloss over the distinction between language/dialect and grammar. For
relevant discussion, cf. Chomsky (1995) on -language vs. I-language.
XVI RICHARD S. KAYNE
the number of varieties of English/ distinct grammars of English must be at least

as great as the number of native speakers of English. Extrapolating to the world
at large, one would reach the conclusion that the number of syntactically distinct
languages/dialects is at least as great as the number of individuals presently alive
(i.e., more than 5 billion).
Adding in those languages/dialects which have existed but no longer exist,
and those which will exist but do not yet exist, it becomes clear that the number
of syntactically distinct (potential) human languages is substantially greater than
5 billion.
One might object at this point that many of these languages will be distinct
from one another only to an insignificant degree. For example, two English
speakers might have identical judgments everywhere except in particle construc
tions, and even there, the differences might readily lend themselves to being
called 'tiny', especially if they had no effect on mutual comprehension. Yet such
tiny differences may (or may not) be of substantial theoretical importance.21
What if it is really true that the commonly cited number of 4000-5000
should, for the purposes of our attempt to understand the human language fac
ulty, be replaced by some number substantially greater than 5 billion?
It would of course be true that an exhaustive study of the syntax of the
world's languages would be rather more arduous than is sometimes thought.
(Nor will those linguists be comforted who admit that linguistic theorizing is im
portant but who think that it should wait until all languages have been studied.)
But if we set aside the unjustified and now entirely unrealistic idea that such an
exhaustive study (which would of course be arbitrarily limited to those lan
guages that happen to be spoken now and to a tiny number of extinct languages)
is a necessary component of the linguistic enterprise, we can focus instead on a
more important question: what is the significance of the number of possible hu
man languages for the acquisition of syntax?
Under the assumption that acquisition proceeds by parameter setting, the
child does not pick its language whole out of a set consisting of all possible lan
guages. Rather, it sets individual (syntactic) parameters, the end result of which
is (the syntactic component of) a grammar. If the number of possible languages
were so large that the number of parameters the child had to set was unmanage
able (i.e., not learnable in the amount of time available), there would indeed be a
problem.
However, the number of independent binary-valued syntactic parameters
needed to allow for 5 billion syntactically distinct grammars is only 33 (2 raised
21
For an example of theoretically important variation within English particle constructions,
cf. Emonds (1976:83-86); for recent discussion, cf. den Dikken (1995).
MICROPARAMETRIC SYNTAX: SOME INTRODUCTORY REMARKS XVii
to the 33rd power is about 8.5 billion). Although we do not yet have a clear idea
of the number of irreducible syntactic parameters, it seems likely that the number
will turn out to be greater than 33. At the same time, although again there is a lot
yet to be understood, it seems plausible that the child is capable of setting at least
that many syntactic parameters.
If the number of independent parameters is somewhat larger, say 50, then
the corresponding number of syntactically distinct grammars is somewhat more
than one thousand trillion. If the parameters are 100 in manageable number, then
the corresponding number of grammars is, innocuously, over one million trillion
trillion (i.e., greater than 10 raised to the 30th power).
Alongside these numbers, the increase in numerical coverage due to work in
microparametric syntax in recent years is quantitatively modest. Yet it may not
be premature to speak of the beginnings of a qualitative improvement in our un
derstanding of syntactic variation. Microparametric syntax is a powerful tool,
whose growth is perhaps to be compared with the development of the earliest
microscopes, that allows us to probe questions concerning the most primitive
units of syntactic variation. And since the invariant principles of UG can hardly
be understood in isolation from syntactic variation, this tool promises to provide
invaluable evidence that will shape our understanding of those principles them
selves.
REFERENCES
Baker, Mark C. 1996. The Polysynthesis Parameter. Oxford & New York: Oxford
Univ. Press.
Benincà, Paola 1994. La variazione sintattica. Studi di dialettologia romanza.
Bologna: Il Mulino.
Chomsky, Noam. 1980. 'On Binding". Linguistic Inquiry 11.1-46.
. 1981. Lectures on Government and Binding. Dordrecht: Foris.
. 1986. Knowledge of Language. New York: Praeger.
. 1995. The Minimalist Program. Cambridge: MIT Press.
Collins, Chris & Höskuldur Thráinsson. 1993. 'Object Shift in Double Object
Constructions and the Theory of Case". Papers on Case & Agreement II. MIT
Working Papers in Linguistics 19, ed. by C. Phillips, 131-174. Cambridge, Mass:
Dept. of Linguistics, MIT.
Comrie, Bernard, ed. 1987. The World's Major Languages. New York & Oxford:
Oxford University Press.
Crystal, David, ed. 1987. The Cambridge Encyclopedia of Language. Cambridge:
Cambridge University Press.
Dikken, Marcel den. 1995. Particles. On the Syntax of Verb-Particle, Triadic, and
Causative Constructions. New York & Oxford: Oxford University Press.
Emonds, Joseph E. 1976. A Transformational Approach to English Syntax. Root,
Structure-Preserving and Local Transformations. New York: Academic Press.
Hellan, Lars & Kirsti K. Christensen, eds. 1986. Topics in Scandinavian Syntax.
XVIII RICHARD S. KAYNE
Dordrecht: D. Reidel.
Henry, Alison. 1995. Belfast English and Standard English. Dialect Variation and
Parameter Setting. New York & Oxford: Oxford University Press.
Johansson, Stig. 1979. "American and British English Grammar: An Elicitation
Experiment". English Studies 60,2.195-215.
Kayne, Richard S. 1981. "On Certain Differences between French and English".
Linguistic Inquiry 12.349-371.
. 1983. "Le datif en français et en anglais". Analyses grammaticales du
français. Études publiées à ľ occasion du 50e anniversaire de Carl Vikner.
Revue Romane. [Numéro spécial 24], ed. by M. Herslund, O. Mordrup & F.
Sorensen, 86-98. English version in R. S. Kayne, 1984, Connectedness and
Binary Branching. Dordrecht: Foris.
. 1985. "L'accord du participe passé en français et en italien". Modèles
Linguistiques VIL 73-90.
. 1989a. "Null Subjects and Clitic Climbing". The Null Subject Parameter,
ed. by Osvaldo Jaeggli & Ken Safir, 239-261. Dordrecht: D. Reidel.
. 1989b. "Facets of Romance Past Participle Agreement". Dialect
Variation and the Theory of Grammar, ed. by Paola Benincà, 85-103.
Dordrecht: Foris.
. 1991. "Romance Clitics, Verb Movement and PRO". Linguistic Inquiry
22.647-686.
------. 1992. "Italian Negative Infinitival Imperatives and Clitic Climbing".
Hommages à Nicolas Ruwet, ed. by L. Tasmowski & Anne Zribi-Hertz, 300-312.
Ghent: Communication & Cognition.
. 1993. "Toward a Modular Theory of Auxiliary Selection". Studia
Linguistica 47.3-31.
. 1994. The Antisymmetry of Syntax. Cambridge: MIT Press.
Klima, Edward S. 1964. "Relatedness between Grammatical Systems". Language
40.1-20. Reprinted in Modern Studies in English. Readings in Transformational
Grammar, ed. by D.A. Reibel & S.A. Schane, 227-246. Englewood Cliffs, New
Jersey: Prentice-Hall.
Merat, Farokh 1974. Une comparaison grammaticale et lexicale de l'anglais bri
tannique et américain enseigné aux étrangers. Thèse de Doctorat de 3e Cycle,
Université de Paris VII.
Poletto, Cecilia 1995. "Split AGR and Subject Clitics in the Northern Italian
Dialects". Paper presented at the 18th GLOW [Generative Linguists of the Old
World] Colloquium, GLOW Newsletter 34.46-47.
Renzi, Lorenzo & Laura Vanelli. 1983. "I pronomi soggetto in alcune varietà ro
manze". Scritti linguistici in onore di Giovan Battista Pellegrini, 25-35. Pisa:
Pacini.
Trudgill, Peter & Jack K. Chambers, eds. 1991. Dialects of English. Studies in
Grammatical Variation. London: Longman.
& Jean Hannah. 1994. International English. A Guide to Varieties of
Standard English, 3rd edition. London: Edward Arnold.
Zandwoort, Reinard Willem. 1965. A Handbook of English Grammar, 3rd edition.
London: Longmans.
UNE ANALYSE MICROPARAMÉTRIQUE DES
MOYENS DANS LES LANGUES ROMANES*
J.-MARC AUTHIER & LISA REED
Université ď Ottawa & Pennsylvania State University
Abstract
This paper introduces novel micro-parametric data drawn from Canadian

French and Madrid Spanish middle constructions which indicate that two alleged
defining characteristics of this construction are not really defining characteristics
of it at all. In particular, it is presently assumed that Romance middles must be
generic and that they disallowfry-phrasesof the type found in passive sentences,
yet these dialects of Romance counterexemplify both claims. Based on these
data, as well as five additional considerations, it is argued that Romance middles
are associated with a structure similar, although not identical, to the one propo
sed in Roberts (1987), namely, one in which there is a PRO internal to VP which
bears the external theta-role. This account is shown to not only accommodate the
full range of micro-parametric variation but to also offer an explanation for the
generic constraint at work in most dialects of Romance.
* Nous tenons à exprimer notre gratitude aux personnes qui nous ont si généreusement aidé à
rassembler les nouvelles données dont notre analyse cherche à rendre compte: pour les données
du français canadien nous remercions Louise Deslauriers de Timmins en Ontario, Tanya
Goguen de Moncton au Nouveau-Brunswick, André Lapierre d'Ottawa, Sylvie Micheline
Mainville de Timmins en Ontario et Michelle Patry d'Ottawa; pour les données de l'italien
nous remercions Leonard Sbrocchi de Roseto Valfortore en Italie; et pour les données de l'espa
gnol nous remercions Maria-Luisa Rivero de Madrid. Nous tenons de plus à remercier de leurs
commentaires l'audience du colloque de 1994 de l'Association linguistique des provinces atlan-
tiques, les professeurs Julia Herschensohn et Paul Postal, ainsi que ľévaluateur anonyme sélec
tionné par les rédacteurs. Les recherches ayant mené à la rédaction de cet article ont été subven
tionnées en partie par une bourse post-doctorale (numéro 756-92-0036) attribuée à Lisa Reed
par le Conseil de la recherche en sciences humaines du Canada.
2 J.-MARC AUTHIER & LISA REED
0. De la problématique de deux propriétés dites

'caractéristiques' des moyens
La construction moyenne dans les langues romanes a fait l'objet d'un débat
soutenu au sein de la grammaire generative (voir par exemple Gross 1968,
Obenauer 1970, Ruwet 1972, Rizzi 1976, Chomsky 1981, Williams 1981,
Belletti 1982, Zubizarreta 1982, Keyser & Roeper 1984, Burzio 1986 [1981],
Wehrli 1986, Roberts 1987 [1985], Zubizarreta 1987 [1982], Cinque 1988,
Fellbaum & Zribi-Hertz 1989, Lyons 1989 et Fagan 1992). De ce débat a
émergé un certain accord parmi les chercheurs quant aux propriétés qui carac
térisent cette construction. Parmi ces propriétés se trouvent la présence obliga
toire du morphème SE/SI ainsi que l'apparition d'un objet thématique en position
de sujet de surface. Ces caractéristiques sont apparentes dans les exemples ci-
dessous tirés du français, de l'italien et de l'espagnol:
(1) a. Ces sakés japonais se boivent frais en été.

b. Questi sakè giapponesi si bevono freddi d'estate.
"Ces sakés japonais se boivent frais en été."
 Esos sakes japoneses se beben frescos.
"Ces sakés japonais se boivent frais."
Il est aussi tenu pour acquis qu'une troisième caractéristique des moyens ré
side dans le fait que ce genre de construction va de pair avec un contexte géné
rique, c'est à dire que les phrases contenant un verbe à la voix moyenne dénotent
un état de faits qui était attribuable, est attribuable ou sera attribuable à n'importe
quel individu ou ensemble d'individus. Les effets de cette contrainte générique
sont perceptibles dans les paires d'exemples comme (2a,b), deux exemples du
français continental qui se distinguent par la présence versus l'absence d'un ad
verbe de temps ponctuel et par la présence versus l'absence d'un temps géné
rique. 1
(2) a. Ces sakés japonais se boivent frais en été.

b. * Hier à huit heures, ces sakés japonais se sont bus frais.
1
Comme l'ont fait remarquer Belletti (1982) et Cinque (1988), le morphème SE/SI en
espagnol et en italien a au moins deux lectures possibles: la lecture moyenne et la lecture
impersonnelle. Le fait que les exemples de l'italien et de l'espagnol qui sont en apparence
parallèles à (2b) soient grammaticaux pourrait donc être lié à la présence de cette lecture
impersonnelle. Nous avons donc illustré la contrainte générique à l'aide d'exemples tirés du
français continental pour éviter ces complications.
LA CONSTRUCTION MOYENNE DES LANGUES ROMANES 3
Enfin, les verbes à la voix moyenne sont tenus d'avoir pour quatrième carac
téristique de ne pas tolérer la présence d'une phrase prépositionnelle en par
semblable à celle qui accompagnent souvent les verbes à la voix passive. Cette
caractéristique se base sur ľagrammaticalité de phrases telles que (3a,c), qui re
flètent les jugements de grammaticalité rapportés dans la plupart des études sur le
sujet (voir par exemple Ruwet 1972:110, Belletti 1982:7, et Aissen & Perlmutter
1983:368). Il est cependant à signaler que l'absence de phrases en par dans les
moyens est une contrainte qui n'est pas sans exceptions puisque des formes plus
anciennes du français utilisé en Europe permettaient la coexistence de la voix
moyenne et des phrases en par (cf. Martinon 1927, Stefanini 1962, Ruwet
1972, parmi d'autres). De plus, Cinque (1988:529, note 11), qui cite Lepschy
(1986, note 4), fait état d'une telle coexistence dans ce qu'il appelle les 'registres
rhétoriques' de l'italien moderne. Néanmoins, il est actuellement tenu pour vrai
que les registres parlés de la forme moderne des langues romanes ne permettent
pas l'utilisation des phraseserdans les constructions moyennes.
(3) a. * Ces sakés japonais se boivent frais par les gens,

b. * Questi sakè giapponesi si bevono freddi dalla gente.
"Ces sakés japonais se boivent frais par les gens."
 * Esos sakes japoneses se beben frescos por los estudiantes.
"Ces sakés japonais se boivent frais par les étudiants."
Personne ne semble cependant avoir remarqué l'existence de dialectes parmi

les langues romanes qui remettent en cause le bien fondé de ces deux dernières
caractéristiques des moyens. Par exemple, il existe des locuteurs francophones
de l'est du Canada pour qui la voix moyenne et les phrases en par peuvent coex
ister (voir les exemples en (4)) et pour qui la voix moyenne peut être exprimée
dans des contextes non-génériques à condition toutefois qu'une phraseenr
soit présente (voir les exemples en (5)).2
2
Certains chercheurs ont en fait remis en question bien avant nous le bien fondé de considérer
la contrainte générique comme étant une propriété caractéristique des moyens en se basant sur
deux types de données. Ces données ont cependant été jugées n'être que des contre-exemples ap
parents à cette contrainte, ce qui est aussi notre avis. Le premier type de données
'problématiques' se retrouve dans les travaux de Ruwet (1972:89,95), Zribi-Hertz (1982:349),
Zubizarreta (1987:150), Lyons (1989:177) et Fagan (1992:159). Ces données exhibent un
SE/SI moyen dans ce qui semble être une phrase non-générique (voir par exemple (ia), un exem
ple tiré de Zribi-Hertz (1982:349)). Or, Ruwet (1972:95), Fellbaum (1986:4), et Lyons
(1989:178) questionnent la valeur de contre-exemple de telles phrases du fait qu'elles sont à un
temps non-ponctuel et donc entièrement compatibles avec la définition de généricité que ľ on
suppose être requise dans les moyens (par exemple qu'à un moment du futur, un agent potentiel
(4) a. En général, ces débats s'enregistrent par Anne, qui est

notre technicienne la plus qualifiée.
b. Ce costume traditionnel se porte surtout par les femmes.
(entendu à la télévision canadienne en 1993)
(5) a. Hier, des rubans noirs se sont portés *(par les étudiantes).
b. Ces promesses se sont faites *(par au moins deux politiciens).
De plus, notre collègue, Madame María-Luisa Rivero, nous a informé que

dans son dialecte de l'espagnol, un dialecte parlé à Madrid, les exemples de
moyens avec une phrase en par sont communs (voir (6a,b)), et que les construc
tions moyennes non-génériques comme celle en (6b) sont aussi attestées.
(6) a. Eso se dice (por los militares).

"Ça se dit (par les soldats)."
b. Esas promesas se hicieron (por Ana).
ces promesses se sont faites (par Ana)
"Ces promesses ont été faites (par Ana)."
quelconque sera en mesure de commettre l'action en question). Pour étayer cette hypothèse on
peut avancer ľobservation de Ruwet qui note que si ľon ajoute à ce genre de phrase un adverbe
de temps ponctuel qui force la lecture événementielle de la phrase (voir (ib)), on obtient un ré
sultat agrammatical.
(i) a. Ce roman se lira bientôt à Moscou.
b. * Demain à huit heures et quart, ce roman se lira à Moscou.
Les phrases en (iia) et (iib) ci-dessous, tirées de Zubizarreta (1987:150) et de Cinque (1988:542)
respectivement, illustrent le deuxième type d'exemple problématique pour la contrainte géné
rique. Il est clair que ces exemples sont cette fois-ci non-génériques en cela qu'ils contiennent
un temps ponctuel. Cependant, l'hypothèse que ce sont de ce fait des contre-exemples à la con
trainte générique en vigueur dans les moyens est loin d'être acceptée par tous. Lyons
(1989:179), par exemple, a suggéré en se basant sur les recherches sur les verbes moyens et
neutres/ergatifs menées par Keyser & Roeper (1984:394), que les exemples comme (iia) pour
raient bien refléter une utilisation neutre de SE et Cinque (1988:542) a défendu l'hypothèse que
les exemples comme (iib) reflètent la lecture impersonnelle de SI. Le fait que les exemples par
allèles à (iib) soient agrammaticaux en français, une langue dans laquelle on ne trouve pas d'u
tilisation productive du SE impersonnel, suggère que l'hypothèse de Cinque est en effet correcte.
Quant à l'hypothèse de Lyons, le lecteur trouvera des données qui l'étayent dans la note 5.
(ii) a. La question est en train de se discuter dans la salle du conseil,
b. Oggi, a Beirut, si è ucciso un innocente,
aujourd'hui à Beirut SI est tué un innocent
"Aujourd'hui, un innocent a été tué à Beirut."
Dans cet article, nous montrerons que ces nouvelles données du français ca
nadien et de l'espagnol de Madrid permettent en fait de mieux comprendre la
contrainte aspectuelle (ou générique) qui se manifeste dans les constructions
moyennes de la plupart des langues romanes.3 De plus, nous mettrons au point
une analyse microparamétrique des moyens dans les langues romanes qui rendra
compte de leur propriétés quelque soit le dialecte considéré. En bref, nous pro
poserons que les constructions moyennes contiennent un argument externe qui
est actif du point de vue de la syntaxe et qui est projeté sous la forme de la caté
gorie vide PRO dans une position interne au VP (comme l'a suggéré Roberts
1987). Des différences mineures en termes de -sélection seront alors invoquées
pour rendre compte de la variation dialectale observée en ce qui concerne les
phrases en par et la contrainte générique dans les moyens.
1. De deux analyses divergentes de la voix moyenne

dans les langues romanes
On trouve dans la littérature deux grands types d'hypothèses qui tentent de

rendre compte du statut syntaxique de l'agent sous-entendu dans les construc
tions moyennes. Le premier type, représenté par les analyses de Williams
(1981:94, 106), Wehrli (1986:274), Cinque (1988:561, 565), Fellbaum &
Zribi-Hertz (1989:17) et Fagan (1992:142, 145), a pour principale caractéris
tique de ne pas traiter l'argument externe comme un agent actif du point de vue
syntaxique. D'après ces chercheurs, le SE/SI des moyens empêche le rôle théma
tique externe d'être représenté en syntaxe et ceci soit en l'effaçant (Cinque et
Wehrli) soit en bloquant sa projection à partir du lexique (Fellbaum & Zribi-
Hertz, ainsi que Williams et Fagan). Ce type d'approche traite donc le verbe à la
voix moyenne comme un prédicat monadique qui projette son argument interne
soit dans la position de sujet de surface (Williams et Fagan) soit en position
'normale' d'objet, la position de surface étant alors dérivée par une opération de
montée en position de sujet (Cinque, Fellbaum & Zribi-Hertz, ainsi que Wehrli).
Les deux versions de ce genre d'analyse sont illustrées en (7).
3 Nous n'avons considéré, pour les besoins de cet article, que le français, l'italien et l'espagnol.
Cependant, à en juger par ce que dit Cinque (1988:571-574) des autres langues romanes, nos
hypothèses semblent pouvoir s'étendre à ces langues, à l'exception sans doute du trentino, un
dialecte de l'italien qui, d'après Zubizarreta (1982:150, notes), n'a pas de SI moyen.
6 L-MARC AUTHIER & LISA REED
(7) a. L'analyse de Williams (1981:94, 106-voir aussi Fagan

1992:163, 164):
b. L'analyse de Wehrli (1986:274-voir aussi Cinque 1988:561,

565 et Fellbaum & Zribi-Hertz 1989:17):
D'un autre côté, Belletti (1982:5), Chomsky (1981:270-271), Keyser &

Roeper (1984:406), Roberts (1987:273-277), ainsi que Lyons (1989:172), se
basant sur les observations antérieures de Gross (1968), Ruwet (1972), et
Zubizarreta (1982), parmi d'autres, ont proposé une analyse de la voix moyenne
dans les langues romanes qui part de l'hypothèse que l'argument externe y est
syntaxiquement représenté soit comme un argument explicite (Belletti et Roberts)
soit comme un argument implicite de nature identique à celle du morphème de la
voix passive (Chomsky, Keyser & Roeper, ainsi que Lyons). Les deux versions
de ce type d'analyse sont données en (8). (Les raisons à l'origine du déplace
ment non-canonique du PRO sujet thématique en (8a) et celles à l'origine de l'ab
sence de représentation explicite du sujet thématique en (8b) seront explicitées
plus bas dans le texte,)
(8) a. L'analyse de Roberts (1987:273-277-qui s'inspire en partie de

celles de Gross 1968, Ruwet 1972, Belletti 1982:5 et Zubizarreta
1982):
b. L'analyse de Chomsky (1981:270-271-voir aussi Keyser &

Roeper 1984:406 ainsi que Lyons 1989:172):
Nous présenterons dans les sections 3 et 4 de cet article cinq arguments qui
étayent une analyse modifiée de la structure proposée par Roberts en (8a). Mais
pour bien saisir la portée de notre discussion, il est nécessaire de commencer par
passer en revue certains tests qui nous permettront de mieux cerner les propriétés
syntaxiques de l'argument externe dans les constructions moyennes.
2. De quatre tests qui distinguent les arguments externes qui

sont actifs en syntaxe de ceux qui ne le sont pas
Nous avons vu que les chercheurs en sont venus à deux conclusions diver
gentes en ce qui concerne le statut syntaxique des arguments externes dans la
construction moyenne des langues romanes. Certains sont d'avis que cet argu
ment est actif dans la syntaxe; d'autres qu'il ne l'est pas. Cependant, les re
cherches entreprises pour en arriver à une analyse de deux autres phénomènes en
anglais—la voix passive et les neutres/ergatifs—ont mené à la mise au point de
quatre tests qui ont pour but d'établir si un argument externe est actif ou inactif
du point de vue de la syntaxe. Par exemple, Chomsky (1981:143, note 60) attri
bue à Manzini (1980) l'observation que seuls les arguments externes qui sont
actifs en syntaxe peuvent contrôler un PRO dans les propositions de but à l'in
finitif, une observation qui a été utilisée depuis par Chomsky (1986:119),
Jaeggli (1986a:611) et Roberts (1987:70). De plus, Chomsky (1986:118),
Jaeggli (1986a:611) et Roberts (1987:70) ont montré que seuls les arguments
externes qui sont actifs en syntaxe peuvent être associés à des adverbes modifi
ant un agent et Chomsky (1986:118) et Jaeggli (1986a:611) ont remarqué que
seuls les arguments externes actifs en syntaxe peuvent être projetés sous la
forme de phrases en par. Enfin Stroik (1992:132, note 7) a découvert que seuls
les arguments externes actifs en syntaxe peuvent fonctionner comme des objets
de prédication.4 Pour bien comprendre la nature de ces quatre tests, examinons
tout d'abord les exemples français en (9a) et (9b), qui illustrent un contraste qui
distingue les passifs des neutres/ergatifs.
(9) a. Les porte-avions de l'ennemi ont été coulés pour PRO met
tre fin à une longue bataille navale.
4
Chomsky (1986:121), Jaeggli (1986a:614) et Roberts (1987:87) avaient en fait proposé que
le rôle thématique externe absorbé par le morphème du passif ne pouvait pas fonctionner
comme un objet de prédication afin de rendre compte de contrastes comme le suivant:
(i) A soldier visited the museum in uniform.
"Un soldat a visité le musée en uniforme."
(ii) The museum was visited *(by a soldier) in uniform.
"Le musée a été visité *(par un soldat) en uniforme."
Il existe cependant des exemples tels (iii)-(v), tirés de Stroik (1992:132, note 7), qui semblent
contredire cette hypothèse. Chomsky (1986:211, note 61) cite Roeper (1984) pour avoir noté
des exemples de ce type et il attribue à Luigi Rizzi l'hypothèse que le contraste qui oppose les
exemples en (i)-(ii) à ceux en (iii)-(v) est lié à la nature adverbiale plutôt qu'adjectivale des mo-
difieurs en (iii)-(v). Si cette hypothèse est correcte, les exemples en (iii)-(v) devraient être con
sidérés comme illustrant le test lié aux adverbes modifiant un agent,
(iii) This bank-job wasn't done alone.
"Ce holdup n'a pas été fait seul."
(iv) I can tell that this letter was written in a good mood.
"Je sais que cette lettre à été écrite de bonne humeur."
(v) That painting was painted blindfolded.
"Ce tableau a été peint les yeux fermés."
b. * Les porte-avions de l'ennemi ont coulé pour PRO mettre fin

à une longue bataille navale.
Le fait que le passif en (9a) mais pas le neutre/ergatif en (9b) tolère la pré
sence d'une proposition infinitive de but démontre que seules les phrases conte
nant un argument externe actif en syntaxe, par exemple les phrases passives,
permettent le contrôle du PRO de la proposition de but. Cependant, nous nous
devons de noter qu'il est impossible d'illustrer le fonctionnement de ce test ainsi
que celui des trois tests suivants au moyen de l'équivalent italien ou espagnol
des constructions de l'anglais et du français contenant des verbes comme to
melt/fondre, to sink/couler. Examinons par exemple les versions italiennes et es
pagnoles des exemples du français en (9a,b), qui apparaissent ci-dessous en
(10a,b) et (1 la,b) respectivement:
(10) a. Le portaerei nemiche sono state affondate per mettere fine

alla lunga battaglia navale,
b. Le portaerei nemiche sono affondate per mettere fine alla
lunga battaglia navale.
"Les porte-avions de l'ennemi ont été coulés pour mettre fin
à la longue bataille navale."
(11) a. Los portaviones del enemigo fueron hundidos para poner

fin a una larga batalla naval,
b. Los portaviones del enemigo se hundieron para poner fin a
una larga batalla naval.
"Les porte-avions de l'ennemi ont été coulés pour mettre fin
à une longue bataille navale."
Comme l'indique le contraste de grammaticalité qui oppose l'exemple du

français en (9b) à son 'équivalent' en italien et en espagnol en (10b) et (1 lb), les
verbes comme couler et fondre ne sont pas intégrés à des structures syntaxiques
équivalentes dans les langues romanes, contrairement à ce à quoi l'on pourrait
s'attendre. Plus précisément, ce type de verbe n'est associé à l'absence d'un ar
gument externe qu'en anglais et en français puisque c'est dans ces langues
seulement que de tels verbes sont incompatibles avec une proposition infinitive
de but. Il est possible que cette différence entre l'anglais et le français d'une part
et l'italien et l'espagnol d'autre part soit liée au fait que ces verbes de type neu
tre/ergatif sélectionnent l'auxiliaire avoir en anglais comme en français mais
prennent l'auxiliaire être en italien et ne peuvent apparaître qu'à la voix moyenne
en espagnol. Ceci ne veut pas dire qu'il n'existe pas de véritables verbes
'neutres' ou 'ergatifs' en italien et en espagnol. Nous nous contenterons donc de

noter ici que les membres appartenant à cette classe ne sont pas les mêmes à
travers les langues romanes. En effet, nous avons pu par exemple trouver en es
pagnol un verbe qui se comporte de la même manière que couler et fondre en
français. Il s'agit du verbe hervir "bouillir" qui, comme on s'y attendrait, donne
lieu à des contrastes comme celui en (9a,b), à savoir, El agua fue hervida para
PRO quitar olores "L'eau a été bouillie pour PRO se débarrasser des mauvaises
odeurs." versus *El agua ha hervido para PRO quitar olores. *"L'eau a bouilli
pour PRO se débarrasser des mauvaises odeurs."
Le test basé sur la présence possible d'un adverbe modifiant l'agent fonc
tionne de la même manière que le test lié à la présence d'une proposition de but.
Le contraste entre l'exemple (12a) et celui en (12b) démontre que seuls les ar
guments externes actifs en syntaxe peuvent être modifiés par de tels adverbes,
par exemple par intentionnellement ou délibérément. Comme le montre (12a), le
morphème du passif peut coexister avec des adverbes de ce type; par contre un
verbe neutre ou ergatif comme celui en (12b) ne le peut pas.
(12) a. Les policiers sont convaincus que ces pièces d'or ont été
fondues intentionnellement,
b. * Ces pièces d'or ont fondu intentionnellement.
De la même manière, le contraste en (13a,b) démontre que seuls les argu

ments externes actifs en syntaxe peuvent fonctionner comme des objets de prédi
cation.
(13) a. Cette machine à bouillir le linge est automatique.

? Maintenant le linge peut être bouilli tout en lisant,
b. Cette machine à bouillir le linge est automatique.
* Maintenant le linge bout tout en Usant.
Enfin, le contraste en (14a,b) illustre le fait que seules les constructions

contenant un argument externe actif du point de vue syntaxique tolèrent la pré
sence d'une phrase en par.
(14) a. Ces pièces d'or ont été fondues par les voleurs,
b. * Ces pièces d'or ont fondu par les voleurs.
Ces quatre tests sont bien entendu applicable à la construction moyenne et

devraient nous permettre de choisir entre les deux structures attribuées aux
moyens qui apparaissent en (7) et (8).
3. Application des tests aux moyens dans les langues romanes
Afin d'étayer son hypothèse que les moyens dans les langues romanes conti
ennent un argument externe actif en syntaxe, Roberts (1987:275) a appliqué le
test lié aux propositions infinitives de but ainsi que le test des adverbes modifiant
un agent à des phrases contenant un verbe conjuge à la voix moyenne. Les résul
tats de ces tests indiquent qu'au prime abord, il y a en effet dans ces construc
tions un argument externe présent dans la composante syntaxique de la gram
maire. Les phrases qui le montrent apparaissent ci-dessous. De plus, nous avons
ajouté aux données de Roberts les exemples en (15c), (16c) et (17c) qui illus
trent l'application du test lié au fonctionnement de l'argument exteme comme
objet de prédication et les résultats de ce test concordent avec ceux des deux
autres.
(15) a. Cette racine se mange pour maigrir. (Zubizarreta 1987:150)

b. Les contrats de location, ça se lit attentivement.
 Ce musée militaire ne se visite qu'en uniforme.
(16) a. Le mele si mangiano per perdere peso.

"Les pommes, ça se mange pour maigrir/'
b. í bambini si lavano volentieri. (Roberts 1987:275)
"Les enfants, ça se lave de bonne volonté."
e. Questi musei militari non si visitano che in divisa.
"Ces musées militaires ne se visitent qu'en uniforme."
(17) a. Una fabrica, eso se incendia para cobrar el seguro.

"Une usine, ça se brûle pour toucher l'assurance."
b. Los contratos, eso se lee con cuidado.
"Les contrats, ça se Ht avec attention."
 Estos museos militares no se visitan más que de uniforme.
"Ces musées militaires ne se visitent qu'en uniforme."
Cependant Cinque (1988, § 4.1 et 4.2) a fait valoir qu'on ne peut pas se fier
à de tels résultats puisqu'on pourrait attribuer la grammaticalité de ces phrases
(comme il le fait d'ailleurs) à une lecture impersonnelle arguméntale du mor
phème SE/SI. Pour éliminer ce problème, Cinque propose l'utilisation de verbes
à l'infinitif avec le morphème SE/SI puisque de tels verbes apparaissent dans des
phrases où le Cas nominatif ne peut pas être assigné et que le SE/SI impersonnel
ajustement besoin d'un tel Cas. Bien que Cinque (1988:562) ait suggéré que les
phrases infinitives avec un SE/SI moyen ne se comportent pas vis à vis des tests
comme des phrases contenant un argument exteme actif en syntaxe, nous avons
pu construire des phrases en français, en italien et en espagnol qui semblent
prouver que les conclusions de Cinque étaient un peu hâtives.
(18) a. Vu que ces manifestants sont aussi méprisés qu'ils sont

mal armés, ils auront l'avantage de pouvoir se disperser
pour PRO satisfaire un electorat avide d'ordre et de sécurité,
b. Les contrats rédigés par M. Leblanc ont la réputation de
devoir se lire avec précaution,
 Ces musées militaires ont le désavantage de ne pouvoir se
visiter qu'en uniforme.
(19) a. Visto che questi dimostranti sono così odiati come male ar
mati, avranno il vantaggio di potersi disperdere per PRO
soddisfare un elettorato avido di ordine e di sicurezza.
"Vu que ces manifestants sont aussi méprisés qu'ils sont
pour satisfaire un electorat avide d'ordre et de sécurité."
b. I contratti preparati dal signor Leblanc hanno la reputazione
di doversi leggere attentamente.
"Les contrats rédigés par M. Leblanc ont la réputation de
devoir se lire avec précaution."
e. Questi musei militari hanno lo svantaggio di non potersi
visitare che in divisa.
"Ces musées militaires ont le désavantage de ne pouvoir se
visiter qu'en uniforme."
(20) a. Dado que los manifestantes son tan despreciados como están
mal armados, tendrán la ventaja de poder dispersarse para
PRO satisfacer a un electorado ávido de orden y seguridad.
pour satisfaire un electorat avide d'ordre et de sécurité."
b. Dado que estos manifestantes son tan despreciados como
están mal armados, tendrán la ventaja de poder dispersarse
con entusiasmo.
avec enthousiasme."
c. Estos museos militares tienen la desventaja de no poderse

visitar más que de uniforme.
"Ces musées militaires ont le désavantage de ne pouvoir se
visiter qu'en uniforme."
En effet, la grammaticalité de phrases comme celles en (18a,c), (19a,c) et

(20a,c) indique que d'après nos trois premiers tests, les moyens dans les
langues romanes contiennent un argument externe actif en syntaxe (et suggère de
plus que Roberts avait raison de supposer que la lecture moyenne de SE/SI est
disponible même lorsque le verbe auquel se rattache ce morphème est conjuge
comme dans les phrases en (15)-(17).)5 Enfin, les données issues du français
canadien et de l'espagnol de Madrid en (4)-(6) montrent que le dernier test, celui
lié à la présence d'une phrase en par, indique lui aussi la présence d'un argument
externe actif du point de vue de la syntaxe. Les résultats des quatre tests tendent
donc à prouver que l'analyse des moyens défendue par Williams (1981), Wehrli
(1986), Cinque (1988), Fellbaum & Zribi-Hertz (1989) et Fagan (1992) est in
correcte.
4. De ľhypothese d'un argument externe explicite dans

les moyens des langues romanes
Dans la section précédente, nous avons examiné les résultats de quatre tests
qui militent contre toute analyse des moyens qui se base sur des structures du
type en (7a,b). Cependant, il nous reste encore à choisir entre deux hypothèses
possibles: l'hypothèse de l'argument explicite défendue par Belletti (1982) et
Roberts (1987), une hypothèse qui se base sur la structure en (8a), et l'hypo
thèse de l'argument implicite proposée par Chomsky (1981) et Keyser & Roeper
(1984), qui se base sur la structure en (8b). Ces deux types d'analyse sont com
patibles avec les données présentées à la section 3 puisque les quatre tests utili
sés ne peuvent distinguer qu'entre les arguments externes qui sont présents en
syntaxe et ceux qui ne le sont pas et qu'en conséquence les arguments externes
implicites et explicites se comportent de manière identique par rapport à ces tests.
5
L'application de ces tests à la classe de contre-exemples apparents mentionnée dans la note
numéro 2 indique que Lyons (1989:178) avait raison d'analyser de telles phrases comme reflé
tant l'utilisation neutre du morphème SE.
(i) a. ?* La question est en train de se discuter dans la salle du conseil pour
faire peur au gouvernement fédéral.
b. * La question est en train de se discuter dans la salle du conseil en hurlant.
c. * La question est en train de se discuter en uniforme au Q.G. du bataillon.
Il nous reste donc à apporter une réponse à la question suivante: des deux struc
tures en (8a) et (8b), laquelle est la bonne?
Nous nous proposons de résoudre cette question à ľaide d'un test mis en
place par Jaeggli (1986a:616-617), un test lié à l'observation qu'il existe deux
types de Contrôle: le Contrôle argumentai et le Contrôle thématique. Alors que le
Contrôle thématique se fait à partir d'arguments implicites, le Contrôle argumen
tai ne peut se faire qu'à partir d'arguments explicites qui occupent une position
arguméntale dans la structure. En se basant sur des contrastes tels celui de
(21a,b), Jaeggli (1986a) a proposé que le Contrôle d'un PRO sujet d'une pro
position infinitive au passif est un exemple de Contrôle argumentai. C'est à dire
que seul (21a) contient un argument explicite (John) qui peut contrôler le PRO
alors que le morphème du passif en anglais ne peut ce faire, étant donné sa na
ture d'argument implicite.6
(21) a. John wants PRO to be loved by everyone.

"Jean veut être aimé de tous."
b. * The bridge was blown up PRO to be awarded a medal.
* "Le pont a été dynamité pour être décoré."
Donc, étant donné que le Contrôle d'un PRO sujet d'une infinitive à la voix
passive est sensible au statut implicite ou explicite de l'argument externe de la
proposition principale, il devrait être possible d'utiliser ce test pour déterminer le
statut (implicite ou explicite) de l'argument externe dans les constructions
moyennes. Or, pour autant que nous le sachions, personne n'a remarqué que les
constructions moyennes dans les langues romanes permettent le contrôle d'un
6
Les jugements de grammaticalité rapportés par nos locuteurs natifs de l'italien et de l'espag
nol indiquent que les constructions passives dans ces deux langues diffèrent de celles de l'an
glais et du français en cela qu'elles tolèrent le Contrôle à l'intérieur de propositions infinitives
au passif. Ce contraste est illustré ici par (i) versus (ii):
(i) a. * These lies were told PRO to be admired by others.
b. * Ces mensonges ont été racontés pour PRO être admiré par les gens,
(ii) a. Ieri, alle due, queste mensogne sono state raccontate per essere ammirati dalla gente,
hier à 2 h. ces mensonges sont été racontés pour être admiré des gens
"Hier à 2 h., quelqu'un a raconté ces mensonges pour se faire admirer par les gens."
b. Ayer a las ocho esas mentiras fueron contadas para ser admirado,
hier à huit heures ces mensonges ont été racontés pour être admiré
"Hier à huit heures, quelqu'un a raconté ces mensonges pour se faire admirer."
Il semble donc que, contrairement à ce que Jaeggli (1986a) supposait, les constructions à la
voix passive dans les langues romanes n'ont pas exactement les mêmes propriétés. En particu
lier, le statut grammatical des phrases en (ii) indique que le passif en italien et en espagnol con
tient un argument externe explicite.
PRO contenu dans une proposition infinitive à la voix passive, ce qui indique
qu'elles contiennent un argument nul explicite.
(22) De tels mensonges se racontent parfois pour PRO être ad

miré par les gens.
(23) Tali mensogne si raccontano spesso per PRO essere ammi

rati dalla gente.
"De tels mensonges se racontent souvent pour être admiré
par les gens."
(24) Esas mentiras se cuentan á veces para PRO ser admirado.

"De tels mensonges se racontent parfois pour être admiré."
En résumé, nous avons défendu l'hypothèse que toute construction moyenne

dans les langues romanes contient un argument explicite interne au VP; c'est à
dire que ces constructions ont en gros la structure en (8a) proposée par Roberts
(1987). Nous devons maintenant rendre compte des différences microparamé
triques qui existent entre les constructions moyennes issues de différentes
langues et dialectes romans. Ces différences nous amèneront en fin de compte à
modifier deux aspects de l'analyse de Roberts.
5. Une analyse microparamétrique des moyens dans les

langues romanes
Comme nous l'avons déjà mentionné au début de cet article, les chercheurs
qui se sont penchés sur le problème des moyens ont pris pour acquis que ce type
de construction est incompatible avec un argument externe exprimé au moyen
d'une phrase en par. Cela a donné lieu à des analyses de la voix moyenne qui
comportent des mécanismes qui ont pour effet de bloquer l'apparition d'une
phrase en par. Or, nous avons vu que les phrases en par sont en fait licites dans
certaines variétés du français canadien et de l'espagnol continental, ce qui a pour
conséquence de nous amener à modifier de tels mécanismes.
Belletti et Roberts ont proposé des analyses qui empêchent l'apparition d'une
phrase en par dans les moyens de deux manières différentes, comme l'indiquent
les entrées lexicales ci-dessous.
(25) Le SE/SI moyen dans les langues romanes: Belletti (1982:5)

Assignation: Ө-externe; pas de PRO
Absorption: ACC
C-sélection: -
(26) Le SE/SI moyen dans les langues romanes: Roberts (1987:273)

Assignation: Ө-externe; chaîne-Ө avec PRO
Absorption: ACC
C-sélection: -
Comme on le voit en (25), l'analyse de Belletti suppose que le morphème

SE/SI des moyens n'absorbe pas le rôle thématique externe mais qu'en fait ce
rôle lui est assigné directement en syntaxe. De plus, étant donné que le mor
phème des moyens est de ce fait un argument, il ne peut pas C-sélecüonner une
autre catégorie comme par exemple une phrase en par parce que le DP contenu
dans cette phrase en par ne serait pas Ө-marqué, ce qui donnerait lieu à une vio
lation du Critère-Ө et de la Condition de visibilité. Roberts, par contre, adopte
une approche différente. Comme le montre son entrée lexicale en (26), il sup
pose que le morphème SE/SI des moyens reçoit le rôle thématique externe et que
ce morphème forme une chaîne thématique avec un PRO en position de sujet.
Cependant, ce PRO doit se déplacer vers une position post-verbale de manière à
libérer la position de sujet dans laquelle l'objet thématique doit monter pour être
en mesure de recevoir le Cas nominatif. (Ces déplacements sont illustrés plus
haut dans la structure en (8a).) Cette analyse a pour effet d'empêcher le mor
phème SE/SI de -sélectionner toute autre phrase puisque tout argument réalisé
dans cette phrase serait privé de rôle thématique, ce qui résulterait en une viola
tion du Critère-Ө et de la Condition de visibilité dans tout moyen contenant une
phrase en par. Ce qui se dégage donc de ces deux analyses, c'est que pour per
mettre ou empêcher l'apparition d'une phrase en par dans les moyens, il suffit de
paramétriser les propriétés thématiques et les propriétés de C-sélecüon du mor
phème SE/SI. Nous suggérerons donc ici de substituer aux entrées lexicales en
(25) et (26) celles en (27a) et (27b) ci-dessous. L'entrée lexicale en (27a) corre
spond bien sûr aux dialectes du français canadien et à l'espagnol madrilène qui
permettent les phrases en (4)-(6). Nous supposons avec Belletti et Roberts que
le morphème SE/SI absorbe le Cas accusatif mais notre analyse diffère des leurs
en cela que nous proposons aussi que SE/SI absorbe le rôle thématique externe et
-sélectionne soit une phrase en par, soit un DP directement. L'entrée lexicale en
(27b) correspond aux autres dialectes des langues romanes. Pour cette entrée,
nous supposons à nouveau, tout comme Belletti et Roberts, que SE/SI est un ab-
sorbeur de Cas accusatif, mais nous nous dissocions de leurs hypothèses res
pectives en proposant que ce morphème joue aussi le rôle d'absorbeur de rôle
thématique externe et qu'il -sélectionne exclusivement un DP.
(27) a. Le SE moyen du français et de l'espagnol madrilène

Absorption: Ө-externe
ACC
C-sélection: (par) DP
b. Le SE/SI moyen des autres langues romanes

Absorption: Ө-externe
ACC
C-sélection: DP
Examinons maintenant quelques-unes des conséquences qui découlent de ces

entrées lexicales. L'option 'phrase en par présente dans les caractéristiques de
-sélection en (27a) permet de traiter les données en (4)-(6) d'une manière paral
lèle au morphème du passif si nous adoptons l'analyse de la voix passive
défendue par Jaeggli (1986a). Jaeggli propose que le morphème du passif, sym
bolisé par -EN, absorbe le rôle thématique externe du prédicat ainsi que le trait de
Cas accusatif associé au verbe. Ces propriétés correspondent à celles que nous
avons attribuées au SE/SI moyen dans les langues romanes. De plus, lorsque
l'option 'phrase en par' est choisie, le DP contenu dans cette phrase reçoit un
Cas de la préposition et un rôle thématique transmis par le morphème du passif
ou par le morphème SE/SI des moyens, selon la construction considérée.
Cependant, notre entrée lexicale pour le morphème des moyens dans les langues
romanes n'est pas identique à celle donnée par Jaeggli pour le morphème du
passif. Nous proposons en particulier qu'à l'encontre du morphème du passif, le
morphème SE/SI des moyens peut aussi -sélectionner directement un DP. De
plus, lorsque cette option est choisie, le DP sélectionné est privé de Cas puisque
le SE/SI moyen absorbe le Cas accusatif du verbe. Ce DP est donc sans matrice
phonétique et doit apparaître en syntaxe sous la forme de la catégorie vide PRO.
(Nous supposons avec Jaeggli 1986b, Roberge 1986, 1990 et Authier 1988,
1992a que PRO est une catégorie vide sans Cas.) En conséquence, l'entrée lexi
cale en (27a) non seulement rend compte du fait que les phrases en par sont
possibles en français canadien et en espagnol madrilène mais prédit aussi le
statut explicite de l'argument externe et sa réalisation en tant que catégorie vide
lorsqu'il n'y a pas de préposition par pour lui assigner un Cas.
Passons maintenant à l'entrée lexicale du SE/SI des moyens correspondant

aux autres dialectes des langues romanes en (27b). Dans ces dialectes, le mor
phème des moyensC-sélectionne uniquement un DP qui, pour les raisons liées à
la Théorie du Cas que nous venons de mentionner, a toujours le statut de la caté
gorie vide PRO. En conséquence, dans ces dialectes, les contructions moyennes
ne peuvent pas contenir de phrase en par.
Il ne nous reste plus qu'à rendre compte du comportement non homogène
des divers dialectes quant à la contrainte aspectuelle/générique. Nous nous pro
posons ici de suggérer que les entrées lexicales en (27a) et (27b) peuvent aussi
être utilisées pour prédire la présence de cette contrainte dans certains dialectes et
son absence dans d'autres. Nous aimerions en particulier proposer de dériver la
contrainte aspectuelle de l'hypothèse que le morphème des moyens est un absor-
beur de Cas. En effet, comme nous l'avons déjà mentionné, le fait que ce mor
phème absorbe le Cas accusatif a pour effet de forcer l'argument externe dans les
moyens qui ne contiennent pas de phrase en par à apparaître sous la forme d'un
PRO. Or, Authier (1988, 1992a,b) a proposé que PRO doit être identifié et peut
l'être de deux manières différentes. Plus précisément, PRO peut être identifié soit
par des marques d'accord faisant partie de la catégorie Xo la plus proche qui
constitue la tête d'un XP qui domine ce PRO (dans le cas qui nous occupe, cet
identificateur potentiel serait le morphème SE/SI en AGRS), soit en vertu d'une
relation de liage à un opérateur non-sélectif (au sens de Lewis 1975) qui lui con
fère une interprétation quasi-quantificationnelle.
Dans le cas des moyens dans les langues romanes, la première façon d'iden
tifier le PRO correspondant à l'argument externe n'est pas disponible puisqu'il
est évident que le morphème SE/SI s'accorde non pas avec l'argument externe,
mais reflète plutôt les marques d'accord de l'objet thématique monté en position
de sujet dans la composante syntaxique. (Ceci se voit par exemple dans les
phrases en (la,c) qui montrent bien l'accord entre le SE/SI moyen, le verbe, et
l'objet thématique en position de sujet.) Ce phénomène d'accord devient particu
lièrement clair en français, une langue dans laquelle on trouve parfois des
phrases à la voix moyenne où l'objet thématique est à la première ou à la deux
ième personne. Dans ces cas, le morphème des moyens s'accorde en personne et
en nombre avec le verbe et l'objet thématique.7
7
Pour des raisons qui nous échappent, les locuteurs de l'italien et de l'espagnol que nous
avons consultés jugent inacceptables les moyens contenant un sujet dérivé à la première ou à la
deuxième personne. Ces jugements, illustrés en (i) et (ii) ci-dessous contrastent avec ceux se
rapportant au français en (28a,b) dans le texte,
(i) a. ?* Dicono che mi leggo bene.
"On dit que je me lis bien."
(28) a. "Nous ne nous vendrons jamais, nous autres," disait Zola

sans envie, mais avec un peu de tristesse. (Exemple tiré
d'Alphonse Daudet, Trente ans de Paris, cité par Sandfeld
1928:133.)
b. Si tu étais une chemise, tu te vendrais mal. (exemple attri
bué à Yves-Charles Morin par Postal 1989:105)
c. On dit que je me lis bien. (Exemple attribute à Richard
Kayne par Osvaldo Jaeggli lors d'un séminaire à USC en
1987.)
d. (En parlant à sa voiture) Tu te pilotes bien, ma vieille.
Donc, étant donné que PRO ne peut pas être identifié par le morphème des
moyens, son identification doit se faire par le truchement d'un opérateur non-
sélectif. C'est en fait ce processus d'identification par un opérateur non-sélectif
qui explique les faits se rapportant à la contrainte générique. En effet, comme l'a
montré Lewis (1975:7), la sémantique des opérateurs non-sélectifs (une classe
d'opérateurs qui comprend les adverbes de quantification en général, parfois et
jamais ainsi que l'adverbe de quantification nul correspondant à en général) est
telle que le domaine de quantification de ces opérateurs est un ensemble de cas,
c'est à dire un ensemble d'événements ou d'intervalles de temps. Par exemple,
dire qu'un état de choses a cours en général, parfois, ou n'a jamais cours, c'est
vérifier si cet état de choses a cours ou n'a pas cours à travers un grand nombre
d'événements. C'est cette particularité sémantique des opérateurs non-sélectifs
que nous croyons être responsable de la contrainte générique en vigueur dans les
moyens qui ne contiennent pas de phrase en par. Dans ces cas-là le SE/SI moyen
-sélectionne un DP qui, en raison de la Théorie des Cas et de la propriété du
SE/SI moyen d'absorber le Cas accusatif, apparaît sous la forme d'un PRO. Or,
comme ce PRO ne peut être identifié que par un opérateur non-sélectif et que la
sémantique de ce genre d'opérateur en restreint la distribution à des phrases qui
permettent la quantification sur les cas, à savoir les phrases génériques, il s'en
suit que toute phrase moyenne sans phraseerdoit être générique.
Enfin, il est important de noter que notre explication de la contrainte géné
rique dans les moyens en termes de -sélection et des conditions d'identification
imposées à PRO prédit qu'en théorie, une langue ou un dialecte qui possède un
b. ?* Tu ti guidi bene, vecchia mia.

"Tu te conduis bien, ma vieille.'
(ii) a. * Dicen que me leo bien.
"On dit que je me lis bien."
b. * Tu te conduces bien.
"Tu te conduis bien."
morphème SE/SI moyen pourrait choisir pour ce morphème les mêmes propriétés
de -sélection que celles données par Jaeggli (1986a) pour le morphème du
passif (c'est à dire que le SE/SI moyen absorberait le Cas accusatif et le rôle thé
matique externe mais ne -sélectionnerait qu'une phrase en par et ceci faculta
tivement.)8 A première vue, le SE moyen de l'espagnol madrilène exemplifié en
(6b) paraît justement être un tel cas puisque les exemples du type Ayer a las ocho
esas promesas se hicieron "Hier à huit heures, ces promesses se sont faites" sont
grammaticaux. Cependant, ces données ne sont pas concluantes puisque la
grammaticalité de telles phrases pourrait, si l'on suit l'analyse de Cinque (1988),
être attribuée à la lecture impersonnelle arguméntale du morphème SE.
En conclusion, nous avons, dans cet article, introduit des différences micro
paramétriques liées à la voix moyenne dans les langues romanes et nous avons
défendu trois grandes hypothèses. Tout d'abord, nous avons proposé que les
constructions moyennes dans ces langues ont une structure très proche de celle
en (8a), la seule différence étant que le PRO interne au VP n'est pas déplacé dans
cette position comme le suggère Roberts (1987), mais y est inséré dans la base.
De plus, nous avons suggéré que le morphème des moyens dans les langues
romanes est associé aux entrées lexicales en (27a,b) en nous basant sur le fait
que seules ces entrées lexicales permettent de rendre compte de la variation mi
croparamétrique qui caractérise la voix moyenne dans ce groupe de langues.
Enfin, les données nouvelles que nous avons examinées ont montré que deux
des propriétés qui avaient été qualifiées de caractéristiques des moyens par le
passé en fait ne méritent pas cette qualification. Nous avons donc défendu l'hy
pothèse que ces deux propriétés (l'absence d'une phrase en par et la contrainte
générique), que l'on retrouve dans la majorité des langues romanes, sont tout
simplement les conséquences qui découlent du choix des propriétés de C-sélec-
tion attribuées au morphème des moyens ainsi que celles qui découlent des
conditions d'identification imposées par la grammaire universelle à la catégorie
vide PRO. 9
8
On doit tout de même garder à l'esprit les données de la note numéro 6 concernant l'italien et
l'espagnol qui semblent indiquer que l'analyse du passif de Jaeggli (1986a) pour les langues
romanes est incomplète.
9
Le lecteur attentif ne manquera pas de se poser la question de savoir si oui ou non nos hypo
thèses concernant les moyens dans les langues romanes s'étendent à d'autres langues comme
par exemple l'anglais. Etendre nos hypothèses à l'anglais est certainement du domaine du pos
sible et le fait que les moyens de l'anglais soient soumis à une contrainte générique et ne soient
pas compatibles avec une phrase en by suggère à première vue qu'une telle extension soit en
fait désirable. Par exemple, on pourrait émettre l'hypothèse que les moyens de l'anglais conti
ennent un morphème SE/SI nul (cette idée ayant été proposée par Keyser & Roeper) et associer
ce morphème aux propriétés de -sélection en (27b). On pourrait aussi adopter la récente hypo-
BIBLIOGRAPHIE
, Peter & Maaike Schoorlemmer. 1995. "Middles and Nonmovement".
Aissen, Judith & David Perlmutter. 1983. "Clause Reduction in Spanish". Studies
in Relational Grammar 1, red. par David Perlmutter, 360-403. Chicago:
University of Chicago Press.
Authier, J.-Marc. 1988. The Syntax of Uns eleetive Binding. Thèse de doctorat,
Univ. of Southern California, Los Angeles, Cal.
. 1992a. "A Parametric Account of V-Governed Arbitrary Null
Arguments". Natural Language and Linguistic Theory 10.345-374.
. 1992b. "Is French a Null Subject Language in the DP"? Probus 4.1-16.
Belletti, Adriana. 1982. "Morphological Passive and PRO-Drop: The Impersonal
Construction in Italian". Journal of Linguistic Research 2.1-34.
Burzio, Luigi. 1986. Italian Syntax: A Government-Binding Approach. Dordrecht:
D. Reidel. (Version revue et corrigé d'une thèse de doctorat, Intransitive Verbs
and Italian Auxiliaries, MIT, 1981.)
Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris.
. 1986. Knowledge of Language: Its Nature, Origin, and Use. New York:
Praeger.
Cinque, Guglielmo. 1988. "On si Constructions and the Theory of crb".
thèse de Stroik (1992) qui suggère que les moyens de l'anglais comportent un PRO qui reçoit le
rôle thématique externe puisque comme ľa montré Authier (1988), ce PRO peut être lié à un
opérateur non-sélectif, tout comme PRO. L'une ou l'autre de ces deux hypothèses rendrait
compte des deux caractéristiques que nous venons de mentionner. Cependant, il existe des faits
qui incitent à la prudence lorsque l'on songe à uniformiser le traitement des moyens des langues
romanes et de l'anglais. Par exemple, il est bien connu que la voix moyenne a un caractère
beaucoup moins productif en anglais que dans les langues romanes. (Il suffit pour s'en conva
incre d'essayer de traduire en anglais les exemples au moyen présentés dans cet article.) De
plus, Jaeggli (1986a:611) et Roberts (1987:189,190) ont montré que les tests classiques qui
permettent de dépister les arguments externes qui sont actifs en syntaxe ont des résultats posi
tifs dans les langues romanes mais des résultats négatifs en anglais. Enfin notre nouveau test
lié au Contrôle à l'intérieur d'une proposition infinitive au passif appliqué aux moyens semble
indiquer la présence d'un argument externe explicite dans les langues romanes mais pas en an
glais. (Cf. le statut grammatical des phrases en (22)-(24) versus le statut nongrammatical de
phrases comme *Bridges blow up easily to be awarded medals) Hâtons-nous d'ajouter, cepen
dant, que ces faits ne devraient pas forcément nous amener à rejeter l'hypothèse que les moyens
de l'anglais sont entièrement différents de ceux des langues romanes. En effet, Stroik (1992) a
attribué les résultats négatifs pour l'anglais de certains des tests que nous avons appliqués aux
moyens à des facteurs indépendants (mais voir la réplique de  & Schoorlemmer 1995).
Nous concluons donc que l'étude des moyens de l'anglais mérite des recherches plus approfon
dies. (Le lecteur intéressé peut se référer entre autres à Fellbaum 1986, à Roberts 1987, à Fagan
1988, 1992, à Massam 1992, à Stroik 1992, et à  & Schoorlemmer 1995 qui ont étu
dié en détail les propriétés particulières des moyens de l'anglais.)
Fagan, Sarah, 1988, "The English Middle". Linguistic Inquiry 19.181-203.

_ _ _ . 1992. The Syntax and Semantics of Middle Constructions: A Study with
Speciai Reference to German. Cambridge: Cambridge University Press.
Felibaum, Christiane, I986. On the Middle Construction in English. Bloomington,
Indiana: éditions IULC.
_ — _ & Anne Zribi-Hertz, 1989. The Middle Construction in French and
English: A Comparative Study of its Syntax and Semantics. Bloomington,
Indiana: éditions IULC.
Gross, Maurice. 1968. Grammaire transformationnelle du français: Syntaxe du
verbe. Paris: Larousse.
Jaeggli, Osvaldo. 1986a. "Passive". Linguistic inquiry 17.587-622.
. 1986b. 'Three Issues in the Theory of Clitics: Case, Doubled NPs, and
Extraction". Syntax and Semantics 19: The Syntax of Pronominal Clitics, red.
par Kagit Borer, 15-42. Orlando: Academic Press.
Keyser, Samuel Jay & Thomas Roeper. 1984. "On the Middle and Ergative
Constructions in English". Linguistic Inquiry 15.381-416.
Lepschy, Giulio. 1986. "Aspects of Italian Constructions with si". The Italianist
6.139-151.
Lewis, David. 1975. "Adverbs of Quantification". Formal Semantics of Natural
Language, red. par Edward Keenan, 3-15. Cambridge: Cambridge University
Press.
Lyons, Christopher. 1989. "L'aspect générique et la voix moyenne". Travaux de
linguistique 19.171-186.
Manzini, Maria Rita. 1980. "On Control", ms., MIT.
Martinon, Philippe. 1927, Comment on parle en français. Paris: Larousse.
Massam, Diane. 1992. "Null Objects and Non-Thernatic Subjects". Journal of
Linguistics 28.115-137.
Obenauer, Hans-Georg. 1970. La construction pronominale passive en français
moderne. Mémoire de maîtrise inédit, Université Paris VIII.
Postal, Paul M. 1989, Masked Inversion in French. Chicago: University of Chicago
Press.
Rizzi, Luigi. 1976. "La montée du sujet, le SI impersonnel et une règle de restruc
turation dans la syntaxe italienne". Recherches linguistiques 4.158-185.
Roberge, Yves. 1990. The Syntactic Recoverability of Null Arguments. Montréal:
McGill Queen's University Press. (Version revue et corrigée d'une thèse de
doctorat, Univ, of British Columbia, 1986.)
Roberts, lan. 1987, The Representation of Implicit and Dethematized Subjects.
Dordrecht: Foris. (Version revue et corrigé d'une thèse de doctorat, Univ. of
Southern California, Los Angeles, 1985.)
Roeper, Thomas, 1984. "Implicit Arguments and the Projection Principle", ms.,
Univ. of Massachusetts.
Ruwet, Nicolas. 1972. Théorie syntaxique et syntaxe du français. Paris: Éditions du
Seuil.
Sandfeld, Kristian. 1928. Syntaxe du français contemporain I: Les pronoms, Paris:
Librairie ancienne Honoré Champion.
Stéfanini, Jean. 1962. La voix pronominale en Ancien et en Moyen Français. Aix-

en-Provence: Ophrys.
Stroik, Thomas. 1992. "Middles and Movement". Linguistic Inquiry 23.127-137.
Wehrli, Eric. 1986. 'On Some Properties of French Clitic SE". Syntax and
Semantics 19: The Syntax of Pronominal Clitics, red. par Hagit Borer, 263-283.
Orlando: Academic Press.
Williams, Edwin. 1981. "Argument Structure and Morphology". The Linguistic
Review 1.81-114.
Zribi-Hertz, Anne. 1982. "La construction SE-moyen du français et son statut
dans le triangle: moyen-passif-réfléchi". Linguisticae Investigationes 6.345-
401.
Zubizarreta, María-Luisa. 1987. Levels of Representation in the Lexicon and in the
Syntax. Dordrecht: Foris. (Version revue et corrigée d'une thèse de doctorat, On
the Relation of the Lexicon to Syntax, MIT, 1982.)
TREATING THAT-TRACE VARIATION
PHILIP BRANIGAN
Introduction
In principle, describing the grammars of two dialects of a language is no dif

ferent than describing the grammars of two quite different languages. In both
cases, we expect universal grammar to provide the basic template to which each
dialect or language will conform. Beyond obvious lexical variation, differences
must be relatively superficial, perhaps even limited in the syntax to variation
among the strength of features of functional heads (Borer 1984; Chomsky
1993). There is, however, one significant difference between two dialects and
two languages. The data available to a child acquiring one dialect will be quite
similar to that available to a child learning a second dialect. We should expect, as
a consequence, that much of the grammars of two dialects will be the same,
since there should be similar trigger experiences involved in acquisition of the
two grammars.
One consequence of this uncontroversial thesis is the following. All things
being equal, we should prefer analyses in which the grammars of two dialects of
the same language are very similar. In particular, if there are seemingly deep
principles of UG underlying a particular phenomenon, we should not be eager to
suppose that two dialects relate to the deep principles in any radically different
fashion. The null hypothesis should always be that dialectal variation is lexical,
rather than the result of fundamental differences in grammars, or different set
tings of parameters.
In this paper, I apply this heuristic strategy to the analysis of variation in the
'that-trace effect'. The core data (la)-(ld) are well-known, thanks to Sobin
(1987).1
1
I make use of the following labelling conventions. Coindexation indicates membership in
the same chain. The category S is an abbrevation for AGRSP.
26 PHILIP BRANIGAN
(1) a. * WhOi did Jay wonder if ti had arrived yet?

b. the guy Oi that ti called us
c. (*) Which applei do the experts agree that ti had been the tastiest?
d. Which appk i do the experts agree ti had been the tastiest?
All native speakers of English reject sentences in which a subject is extracted

past an adjacent if complementiser, as in (la). Speakers uniformly accept relative
clauses in which an invisible relative pronoun is extracted to the specifier posi
tion of the first dominating CP, with a that complementiser heading CP, as in
(lb).
Judgments vary with respect to sentences of the (lc) type: some speakers
reject this class of sentences; others find it well-formed. Moreover, judgments
are quite sharp, contrasting with the relatively weak judgments associated with
subjacency violations. Sentences of this type involve long WH-movement of a
subject past an overt that complementiser in its own clause. Let us refer to the
conservative group which finds (lc) ungrammatical as dialect A, and the more
permissive group as dialect B. There are no clear boundaries to these 'dialects',
but the terminology will aid the exposition. Grammaticality judgments in dialects
A and  again converge with sentences like (Id), in which the subject is ex
tracted via WH-movement, but without a visible that complementiser in its origi
nal clause. All speakers find such sentences grammatical.
The variability in judgments is unexpected under the conventional ECP ac
count of (la), originally put forward in Chomsky (1981). If a subject trace fails
to be properly governed in (la), it should equally fail to be properly governed in
(lc) in any dialect of English, as the structure of the sentences should be identi
cal in both cases. This suggests that one of the following two conclusions
should be drawn: either the ECP does not provide the right account of the limits
on WH-extraction of the subject of finite clauses, or the structures of (la) and
(lc) are not parallel for those speakers who accept the latter as well-formed, or
both.
My claim here will be that both conclusions should be drawn. The ECP—a
principle requiring proper government of traces—has no role to play in explain
ing the limits on subject extraction. Instead, a uniformity constraint on chains
lies behind the ungrammaticality of the relevant cases. And the structure of (lc)
is quite different from the structure of (la) in the dialects which accept the for
mer. The different structures come about because of a difference in the phono
logical form of a single functional head: a complementiser which checks subjects
in SPECCP. Thus, variation in these two dialects is ultimately a lexical difference
TREATING -TRACE VARIATION 27
and nothing more, although deep principles of UG can be observed in studying

the contrasts.
The form of the paper is as follows. I begin with a discussion of the that-
trace effect in other Germanic languages. The effect is shown to follow from a
condition on uniform chains, which requires -bar chains originating in an A-bar
position to be in a certain configuration. It is then shown that the same treatment
can be used for the English dialect A data. Finally, the permissive dialect  data
is shown to follow from the supposition that an overt that complemenüser can be
situated in   position from which it is excluded in dialect A.
I assume throughout the Minimalist model of Chomsky (1993, 1994
[1995]).
1. The thai-trace effect in verb-second complements
The contrast between (1c) and (1d) is possible in English because the com
plemenüser can be omitted or included freely in complements to the appropriate
class of verbs. The optional presence of a complemenüser is not a necessary-
feature of language. Other languages may require or prohibit overt complemen-
tisers in parallel contexts. Within Germanic, Danish (2) and German (3) provide
a cross-linguistic minimal pair in this respect.
(2) De sagkyndige siger *(at) dette aeble smager ikke bedst.

the experts say that this apple tastes not best
(3) Die Experten sagen (*dass) dieser Apfel schmeckt nicht am besten.
the experts say that this apple tastes not the best
(The V/2 structure of (4) is shown by the position of the finite verb smager
in front of the negator ikke.) As the data shows, Danish and German differ in
the restrictions they impose on complementisers in V/2 complements. Danish re
quires the at complemenüser in such clauses. German does not allow a comple
menüser to appear (den Besten 1983).
Significantiy, this difference between the two languages appears to go hand-
in-hand with the possibility of a phrase undergoing WH-movement out of the
V/2 complement (Vikner 1991).
(4) * Hvilket æble siger de sagkyndige at smager ikke bedst?

which apple say the experts that tastes not best
(5) Welcher Apfel sagen die Experten schmeckt am besten?
which apple say the experts tastes the best
28 PHILIP BRANIGAN
In Danish, this type of extraction is impossible, as example (4) demon

strates. In the German (5), in contrast, extraction is possible, presumably be
cause the complementiser is missing. The fact that the presence of an overt
complementiser blocks extraction suggests a clear parallel with the English con
trast between (1c) and (1d). The impossible Danish (4) appears to pattern with
English (1c), while the grammatical (5) acts like English (1d).
The Danish/German contrast involves a structure which seems initially to be
quite different from that of the English sentences. The extraction site of the WH-
phrase in the German/Danish case is SPEC.C.2 It is only the presence of a *topic'
category in SPECC which is responsible for the verb movement to  which pro
duces the V/2 order. When no category appears in SPEC-, verb raising is im
possible, as shown by the (6) examples.
(6) a. * De sagkyndige siger at smager dette æble bedst.

the experts say that tastes this apple best
b. * Die Experten sagen schmeckt dieser Apfel am besten,
the experts say tastes this apple the best
The 'topic5 category in a V/2 clause need not be the subject. Complements
and adjuncts may also raise to SPECC and legitimate verb raising in the same
manner. And the same contrast between Danish and German is found when ob
jects or adjuncts are extracted from an embedded V/2 SPECC: (7). (Example
(7d) is taken from Vikner (1991).)
(7) a. De sagkyndige siger at dette æble foretrækker de ikke.

the experts say that this apple prefer they not
b. * Hvilket æble siger de sagkyndige at foretrækker de ikke.
which apple say the experts that prefer they not
c. Sie hat gesagt Geschichte haben die Kinder gelernt
she has said history have the children learned
d. Wiei hat sie gesagt ti haben die Kinder Geschichte ti gelernt?
how has she said have the children history learned
Again, the fronted position of the verb shows that WH-movement is preceded by
movement of the complement or adjunct phrase to SPEC.C.
I assume that the derivation is fixed in this respect (in both languages): there
are two distinct -bar movement operations involved in WH-extraction from a
41 assume this to be the case even when subjects are extracted, contra Travis (1984); Zwart
(1993). For arguments against the dual position analysis of verb-second, see Schwartz &
Vikner (1989); Branigan (1995).
TREATING THAT-TRACE VARIATION 29
V/2 complement. The first is topicalisation, which takes the WH-phrase from its
position inside the clause to the SPECC topic position. This movement results—
for whatever reason—in movement of the finite verb to  Topicalisation pro
duces a chain with its head in an -bar position and its tail in an A-position (or
possibly an adjunct, 'broadly L-related' position). This chain is not uniformly A-
bar. The second movement then raises the WH-phrase from SPEC  to a higher
SPEC  position, possibly through a sequence of intermediate SPECC positions.
This produces a second chain, all of the members of which occupy SPEC.C. This
second chain is uniformly A-bar.
I assume as well that the overt complementiser found in embedded V/2
complements in Danish heads a second CP category, which takes the V/2 clause
as its complement (Vikner 1991; Iatridou & Kroch 1992). Thus the structures of
the complement clauses in (4) and (5) are as shown in (8a) and (8b), respec
tively.
(8) a. hvilket æbk i [CP at [CP ti smager [s ti ikke t v bedst ]]

b. welcher Apfeli [Cp ti schmeckt [s ti am besten t v ]]
One further issue remains to be clarified before the extraction constraints can
be addressed. The CP-recursion structure found in Danish embedded V/2 is
somewhat problematic from the point of view of a restrictive theory of selection.
A V/2 clause is found only in a selected position—such clauses cannot
serve as subjects or adjuncts, nor are they allowed to be complements to other
than a restricted class of predicates (bridge verbs, loosely put). In short, V/2
complements must be selected by an appropriate predicate. Under conventional
assumptions, a head can select properties only on the head of its complements.
This means that the matrix verb should be unable to select a V/2 clause in the
structure (8a). As the verb must in fact do so, we may suppose that the structure
of (8a) is modified in the LF component. Selection will be possible if covert
movement raises the V- complex to the next head, where it can be selected di
rectly at LF.3 Then the LF representation of (8a) will be (9). This -raising op
eration makes sense if we suppose, with Law (1991), that some complementis-
ers are 'expletive' elements which must be eliminated from the LF representa-
3
An anonymous reviewer points out that the matrix verb might select the V/2 complement in
directly, by selecting a complementiser which has the particular property of selecting a V/2
complement. In the absence of any reason to suppose such a complementiser, I prefer the text
analysis. Although indirect selection is certainly a coherent notion, by resorting to such selec
tion 'magic', we mask the relationship between the matrix verb and its complement type,
rather than explaining it.
 PHILIP BRANIGAN
tion. Then movement of the lower V- complex is necessary in order to satisfy
Full Interpretation, because only such movement will eliminate the expletive 
from the LF representation.
(9) V[CpV-C-C[CPXPtc[s...t...]]]
Now the parameters of the problem are clear. While extraction from SPEC- in a
V/2 complement is possible, as occurs in German, extraction from a lower
SPEC- is impossible in CP-recursion constructions, such as those found in
Danish.
The obvious place to look for a solution is the bounding theory. Let us there
fore quickly dispose of this possibility. Suppose a mechanism similar to those
employed in Chomsky's (1986) theory of barriers. In such a theory, a comple
ment is a barrier unless it is L-marked. In the case of the CP-recursion structures
which concern us here, the consequence seems to be that the second CP should
be a barrier. If the upper CP were L-marked (by the matrix verb) and the lower
CP were not, then movement from the lower SPEC- would cross a CP barrier.
However, any theory of this sort must stipulate, as Chomsky did with the IP cat
egory, that adjunction to the appropriate barrier is not allowed. In the absence of
any justification for this stipulation, a barriers-based approach does not hold
much promise.
The alternative is a constraint on representations, like the ECP. In the frame
work I have adopted, there is no notion of government, so there can be no no
tion of proper government. But it is possible to formulate a Minimalist counter
part to the ECP, by relying on the notion of a checking domain. We require some
principle which constrains the formation of a uniform -bar chain, that is, of a
chain the head and tail of which occupy distinct SPEC- positions. We cannot
exclude such chains entirely, as they are well-formed in the German structures,
where there is no CP-recursion in embedded V/2. Let us suppose then that a uni
form -bar chain must not terminate in the specifier of a CP, the head of which
has been raised. This can be captured in a general LF constraint on chains, which
I refer to as RES(ECP) (the 'residue' of the ECP).
(10) RES(ECP) (initial formulation)

The members of a uniform -bar chain must all be contained in the
checking domain of a complementiser.
Let us take this formulation as a first approximation only, which will be re
fined as we proceed. Conceptually, the RES(ECP) is quite closely related to the
ECP. Both impose an extra licensing condition on traces. The ECP does so di
rectly, requiring proper government in place of simple government. The
RES(ECP) also imposes an extra constraint on traces, but it does so indirectly, by
referring to members of uniform chains.
Now let us compare the effects of the RES (ECP) on the grammatical German
example (5) and the ungrammatical Danish example (4). In the German case,
there is a single CP projection in the complement clause. The tail of the chain
formed by WH-movement is in SPEC.C, and this is part of the checking domain
of the complementiser. So RES(ECP) is satisfied for the tail of the uniform A-bar
chain. As for the head of the -bar chain, it too occupies a SPECC with no CP
recursion, so we may assume that RES(ECP) is satisfied here, too. In the Danish
case, the effect of RES (ECP) is quite different. While the tail of the uniform A-bar
chain again occupies a SPECC position, it is not a part of the checking domain of
any complementiser at LF. The checking domain of any head is determined
derivationally (Chomsky 1993), so that it may change as movement creates new
X°-chains. While the original checking domain of the lower  includes its speci
fier, movement of that head to a higher position produces a new checking do
main for the new binary Xo chain, which does not include its original specifier.
As noted above, the V- complex must raise to the higher  in the LF component
in order for selection of the V/2 complement to be possible. So LF movement of
the V- complex to the higher  will leave the tail of the uniform -chain in vio
lation of RES(ECP). We must also be sure that normal A-bar chains do not run
afoul of the RES (ECP). This is quite straightforward. In A-bar chains which do
not originate in SPECC, the tail normally does not occupy an A-bar position.
Complements to categories other than  are either L-related or immobile. L-re-
lated categories can never serve as the tails of uniform A-bar chains. The same is
true of subjects of small clauses or ECM complements, which occupy a
SPEC-AGR position.
So the RES(ECP) has the right empirical effects for distinguishing the gram
matical German examples from the ungrammatical Danish ones. The next step
must be to see how it can best be formulated in principled, minimalist terms. The
definition (10) requires that chain members be in the checking domain of some
complementiser. We might seek to eliminate the reference to a single syntactic
category  in favor of a more general constraint. Let us suppose that the prop
erty which distinguishes  from other categories is the fact that the domain of 
is not L-related. In other words, everything in its domain is in an A-bar position.
Then we can relate the A-bar properties of the chain constrained by the RES (ECP)
to this property of  The revised RES(ECP) in (11) expresses this idea.
32 PHILIP BRANIGAN
(11) RES(ECP)
Any category α belonging to a uniform chain Σ must be in the
checking domain of a head γ, such that the features which render Σ
uniform are determined for α by the corresponding properties of γ.
More verbosely, in the case of WH-movement, the RES(ECP) now requires

that a chain which is uniformly -bar have all its members in the checking do
mains of heads which are not lexical (or L-related). (One immediate consequence
of the RES(ECP) will be that WH-movement from adjoined positions should be
impossible, under the assumption that adjoined positions are -bar, rather than
L-related. This then requires us to adopt a theory in which at least those
'adjuncts' which may undergo -bar movement must originate in an L-related
position, as proposed by Larson (1988) and Chomsky (1994), among others.)
In formulating the RES(ECP) in terms of chain uniformity with respect to
certain features, we are led to ask whether features other than L-relatedness are
relevant to this constraint. It appears that Ө-marking is also relevant to the
RES(ECP), so that a chain which is uniformly Ө-bar will be rejected if it contains
members which are not in the checking domain of an appropriate head. The ef
fect of this more general formulation of the RES(ECP) can be seen in the un-
grammatical passives in (12).
(12) a. The car was repaired by my sister.

b. * Ten pounds are weighed by this book.
 * The dust was bit(ten) by my car a week ago.
The objects of weigh and bite cannot be raised to subject position, even
though they are regular NPs and NP objects are usually raised in passives. This
follows from RES(ECP), because the objects do not occupy a Ө-position in either
case. Movement of the object to SPEC-AGRS would create a uniform -chain, all
of whose elements are not Ө-marked. The tail of this chain is not in the checking
domain of any head, though, so it does not satisfy RES(ECP).
Another effect of the RES(ECP) is that topicalisation of S will be blocked, ac
counting for the ungrammaticality of (13b).
(13) a. ? Peter promised to tell Mark that it would snow and snow¡, he
told Mark that it would ti.
b. * Peter promised to tell Mark that it would snow and [s it would
snow ] i , [s he told Mark that ti.]
In this example, the S complement to that is topicalised to SPECC. If we as

sume that the complement to  occupies an -bar position, just as the comple
ment to V occupies an A position, then the topicalisation forms a uniform A-bar
chain, the tail of which is not in the checking domain of  RES(ECP) then ex
cludes this structure at LF. (Even speakers of the less restrictive dialect  reject
structures like (13b), which shows that the that complementiser itself is never
able to license an empty category as its complement.)
The notion of a uniform chain must be sharpened. Developing a suggestion
by Browning (1987), Chomsky (1993) and Chomsky & Lasnik (1993) propose
that the intermediate traces of non-uniform chains are deleted, resulting in weak
subjacency effects if the affected chain had contained ηοη-γ-marked traces. For
their purposes, a uniform chain is uniform with respect to either the -bar status
of its members or the marked status of its members. A chain with at least one
uniform property is immune to deletion of intermediate traces. Let us refer to this
property of chains as 'weak' uniformity. The notion of uniformity I require is a
more restricted notion, which I will call 'strong' uniformity. A chain is strongly
uniform if it is uniform with respect to both the -bar status of its members and
their marked status. Thus, a chain ending in a Ө-position will never be strongly
uniform, while a chain ending in a ոօո-Ө position may be, as long as it is made
up entirely of elements either in A-positions or in -bar positions.
Let us fine-tune the RES(ECP) a final time to reflect ¿lis distinction:
(14) RES(ECP) (final version)

Any category α belonging to a strongly uniform chain Σ must be
within the checking domain of a head γ, such that the features
which render Σ uniform are determined for α by the corresponding
properties of γ.
2. The variable thai-trace effect in English
The literature on this phenomenon, notably Sobin (1987), Rizzi (1990) and
Sobin (1991), proposes certain alterations to the formulation of the ECP which
are designed to provide a means for proper government to take place in (lc), for
those speakers who find this sentence grammatical. In both cases, the mecha
nism involves strengthening the that complementiser by a procedure which Com
dexes it with the subject, so that it counts as a proper governor. The same pro
cedure is said to take place with the that complementiser found in subject relative
clauses like (lb).
34 PHILIP BRANIGAN
In essence, the strategy adopted is to patch the ECP to deal with the awkward
data. The strategy is not in itself unreasonable—we should hardly expect to
achieve exactly the right formulation of a principle of UG on the first try. But re
cent developments in the form of the theory as a whole—in particular,
Chomsky's (1993) minimalist program—make this strategy for characterising
dialectal variation less attractive, for the simple reason that the ECP is not a for-
mulable notion in the minimalist model. The ECP requires that a restricted form
of government obtain for all traces. In the minimalist model, the notion of gov
ernment is not defined, as all relations between head and phrasal category can be
defined in the more fundamental local -bar relations. Without government,
there is no proper government. Without proper government, there can be no
ECP. Then there is little point in reconciling the properties of dialect  with the
ECP. What is needed is a new account of the ungrammaticality of (lc) in dialect
A.
It remains to be seen how RES (ECP) can account for that-trace effects in
English, where the relevant contexts are not V/2 complements. Conventional
wisdom has it that subjects occupy an A-posiüon within the clause—SPEC-AGRS,
in the minimalist model. If this is right, then there is in fact no way to derive the
English facts from RES(ECP). But I maintain that there is sufficient evidence that
conventional wisdom is wrong in this respect, as argued in Branigan (1992,
1996) and Harley (1995). The most direct evidence comes from the locative in
version construction, in which a PP is preposed to the position normally occu
pied by subjects, leaving the subject elsewhere. Example (15) illustrates.
(15) Behind this door sleep three very large dogs.
If the locative PP occupied SPEC-AGRS, then the fact that the subject agrees
with the finite verb is inexplicable.4 If the locative PP occupies a different posi
tion, though, then covert movement of the subject to SPEC-AGRS allows the usual
nominative Case and agreement checking to take place. I conclude that the pre-
verbal position usually occupied by subjects in finite clauses is not SPEC-AGRS,
but rather a higher position. I now propose that this position is SPEC-, so that
the structure of a finite complement clause may involve 4CP recursion', as in
(16). 5
4
But cf. Hoekstra & Mulder (1990) and Watanabe (1993) for alternative proposals concerning
agreement in locative inversion.
5
Other literature which points in the same direction includes Culicover (1991) and Shlonsky
(1992). Culicover's Topic projection and Shonsky's AGRC projection can be taken as roughly
equivalent to my  projection, for the purposes of this paper. The RES (ECP) is indifferent to
(16) John said [CP that [Œ Peter  [s t loves green tea ]]]
Some terminology will prove convenient to allow us to distinguish between

the various Cs. Let us refer to the  with the subject as its specifier as the
'primary' C. I use the notationCto single out a primary  in a phrase marker.
(The symbol π has no further significance.) The head of the higher CP(s) asso
ciated with the same clause can be referred to as secondary C(s). The primary 
in (16) is phonetically null; the secondary  is that, I suppose as well that the
primary  must adjoin to the secondary  at some point in the derivation, just as
the V- complex must adjoin to at in Danish embedded V/2 complements.
Without argument, I will assume that this -movement takes place in the overt
syntax, so that the S-structure of the complement clause in (16) is (17).6
(17) [CP C -that [CP Peteri tc [s ti loves green tea ]]]
The that-trace effect in the -dialects now follows from RES(ECP). Consider ex
ample (lc), once more.
(lc) (*) Which appk i do the experts agree that ti had been the tastiest?
Given the SPEC.C position occupied by subjects prior to any WH-movement,

the WH-phrase can only reach its surface position via two distinct -bar move
ments. The first raises the subject from its Case position within the clause to the
specifier of the primary C. The next takes the subject from SPECC to a higher
SPEC.C, to have its WH-features checked. Alongside the movement of the sub
ject, the primary  undergoes head-movement, so that the LF-representation will
be (18).
(18) [CP whOi did [s John say [CP C -that [ ti tc [s ti had ...] ]]]]
The uniform -bar chain in (18) violates RES(ECP) because its tail is not in
the checking domain of any complementiser. The same explanation holds for
(19), where the locative PP occupies the position normally used by subjects, be
fore WH-movement raises it higher.
the label used on the category for which the subject is the specifier. What matters is that the
subject must occupy an -bar specifier position.
6
See Branigan (1995) for arguments that C-to-C movement takes place only in the overt syn
tax.
36 PHILIP BRANIGAN
(19) * Behind which doori did Tom say [cp C-that [CP ti t c [s sleep three
large dogs ti ]]]
When the that complementiser is not present, the primary complementiser re
mains in place, so that RES(ECP) is satisfied in (20).
(20) a. Which eli do the experts agree [CPti[s ti had been the
tastiest ]]
b. Behind which doori did Tom say [ cp ti C [s sleep three
large dogs ti ]]
As C71 is not raised, the tail of the uniform -bar chain is in its checking domain
at LF, thereby satisfying the RES(ECP).
The remainder of the (1) sentences (in dialect A) fall readily into place.
Consider first example (la).
(la) * WhOi did Jay wonder if ti had arrived yet?
The derivation of (la) matches that of (lc). The subject is extracted from
SPECC by WH-movement, giving rise to a strongly uniform chain. And the pri
mary complementiser must again adjoin to the secondary —in this case, the if
complementiser. The RES(ECP) is then violated by the LF representation.
The subject relative in (lb) is unproblematic, as long as the structure of the
relative clause is that shown in (21).7
(21 ) the guy [cp Oi that" [s ti called us ]]
In this case, the primary complementiser takes the form that. The primary 
has phonetic content in a relative clause of this type, because its specifier has
none. Examples like (22) are excluded as a 'doubly-filled Comp' (DFC) effect
(Chomsky & Lasnik 1977).
(22) the guy whOithatcalled us
As DFC effects are common elsewhere in the grammar, this need not be a
matter for concern. What is significant in the structure of (21) is the absence of
any uniform -bar chains in the relative clause. The sole -bar chain terminates
7
Branigan (1992) shows that 'vacuous movement effects' can be explained on the basis of a
structure for subject relatives similar to that proposed in the text.
TREATING -TRACE VARIATION 37
in an A-position: SPEC-AGRS. This renders the chain immune to the RES(ECP), so

the grammatical status of the expression is expected.
All that now remains is to explain why the overt complementiser does not
render the (lc) example ungrammatical in the permissive dialect B. The simple
answer is that speakers of dialect  may treat non-relative that as a primary  In
dialect B, then, (lc) may be represented as (23).
(23) which appleSj do the experts agree [CP ti that [s ti are the tastiest ]]
Here, no movement of the primary complementiser takes place, so the uniform

-bar chain satisfies RES(ECP).
The DFC effect—whatever its actual explanation—ensures that sentences
with the form of (24) will remain impossible even in dialect B.
(24) V [CP (WH-phrase)  [CP NP that [ s . . . ]]]
3. Conclusion
The data in (1) pose problems for an ECP account of constraints on subject
extraction. The ECP is too coarsely grained to deal with the actual variability of
the data, especially if the structures assigned to identical strings are invariant
across dialects. By relaxing the latter assumption, and by reformulating the ECP
as RES(ECP), a minimalist constraint on chains, the theory is made more flexi
ble, and the necessary fine distinctions in the data can then be drawn. The pic
ture of English that-trace variation which results from this analysis includes a
limited, and familiar, brand of variation. Dialect A and Dialect  differ only in
the form of the primary complementiser in finite complement clauses, as sum
marised in (25).
(25) Forms of the complementisers in finite clauses
Dialect A Dialect 
non-relative relative non-relative relative
primary : ø that or ø that or ø that or ø
secondary C: that or if that or ø that or if that or ø
No variability in agreement or indexing procedures is necessary. And the

grammars of English fall into place as reflexes of Universal Grammar, with
syntactic variation only in the system of functional heads.
38 PHILIP BRANIGAN
REFERENCES
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 & Philadelphia: John Benjamins.
Borer, Hagit. 1984. Parametric Syntax: Case Studies in Semitic and Romance
Languages. Dordrecht: Foris.
Branigan, Philip. 1992. Subjects and Complementizers. MIT dissertation.
. 1996. "Verb-second and the -bar Syntax of Subjects". Studia
Linguistica 50,1.51-79.
Browning, Marguerite. 1987. Null Operator Constructions. MIT dissertation.
Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris.
———. 1986. Barriers. Cambridge, Mass.: MIT Press.
. 1993. "A Minimalist Program for Linguistic Theory". The View from
Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, ed. by
Kenneth Hale & Samuel Jay Keyser, 1-52. Cambridge, Mass.: MIT Press.
———. 1994. "Bare Phrase Structure". MIT Occasional Papers in Linguistics 5.
[Published as Chomsky, Noam. 1995. "Bare Phrase Structure". Government
and Binding Theory and the Minimalist Program, ed. by Gert Webelhuth, 381-
439. Oxford and Cambridge, Mass.: Blackwell]
— — . & Howard Lasnik. 1977. "Filters and Control". Linguistic Inquiry
8.425-504.
. 1993. "Principles and Parameters Theory". Syntax: Ein internationales
Handbuch zeitgenössicher Forschung = An International Handbook of
Contemporary Research, ed. by Joachim Jacobs, Arnim von Stechow, Wolfgang
Sternefeld & Theo Vennemann, 506-569. Berlin: Walter de Gruyter.
Culicover, Peter W. 1991. "Topicalization, Inversion, and Complementizers in
English", ms., Ohio State University.
Harley, Heidi. 1995. Subjects, Events and Licensing. MIT dissertation.
Hoekstra, Teun, & Rene Mulder. 1990. "Unergatives as Copular Verbs:
Locational and Existential Predication". The Linguistic Review 7.1-79.
Iatridou, Sabine, & Anthony Kroch. 1992. "The Licensing of CP Recursion and
its Relevance to the Germanic Verb-Second Phenomenon". Working Papers in
Scandinavian Syntax 50.1-25.
Larson, Richard. 1988. "On the Double Object Construction". Linguistic Inquiry
19.335-391.
Law, Paul, 1991. Effects of Head-Movement on Theories of Subjacency and Proper
Government. MIT dissertation.
Rizzi, Luigi. 1990. Relativized Minimality. Cambridge: MIT Press.
Schwartz, Bonnie & Sten Vikner. 1989. "All Verb-second Clauses are CPs".
Working Papers in Scandinavian Syntax 43.27-49
Shlonsky, Ur. 1992. "The Representation of Agreement in Comp and Subject
Clitics in West Flemish". Geneva Generative Papers 0,0.27-38.
Sobin, Nicholas. 1987. "The Variable Status of Comp-trace Phenomena".
Natural Language and Linguistic Theory 5.33-60.
. 1991. "Agreement in CP". Lingua 84.43-54.
Travis, Lisa. 1984. Parameters and Effects of Word Order Variation. MIT disserta
tion.
Vikner, Sten. 1991. Verb Movement and the Licensing of NP-Positions in the
Germanic Languages. Université de Genève dissertation.
Watanabe, Akira. 1993. AGR-based Case theory. MIT dissertation.
Zwart, . Jan Wouter. 1993. Dutch Syntax: A Minimalist Approach. Univ. of
Groningen dissertation.
NEGATIVE PARTICLE QUESTIONS:
A DIALECTAL COMPARISON*
LISA L.-S. CHENG, C.-T. JAMES HUANG & C.-C. JANE TANG
University of California, Irvine & Academia Sinica, Nanking, Taipei
1. Introduction
There are a number of ways to form yes-no questions in Chinese, though
not every dialect employs all the choices. In this paper, we discuss a particular
yes-no construction which uses negation markers to form yes-no questions, as
in (1)-(3).
(1) Mandarin
hufei kan-wan-le nei-ben shu meiyou
Hufei read-finish-PERF that-cL book not-have
"Has Hufei finished reading that book?"
(2) Cantonese
wufei lei-zo mei
Wufei come-PERF not-yet
"Has Wufei come yet?"
(3) Taiwanese
i u tsiak beng bo
he have eat rice not-have
"Did he eat?"
* We thank the audiences of the APL A meeting on Mircroparametric Syntax and Dialect
Variation, the Workshop on Theoretical East Asian Linguistics at UC Irvine, the Stanford
Workshop on Historical Chinese Linguistics, and audiences at the universities of Connecticut,
Toronto, Illinois at Urbana-Champaign and Arizona. In particular, we would like to thank
Audrey Li, Richard Kayne, Chao-Fen Sun, Rint Sybesma, and Pei-Chuan Wei for their com
ments and suggestions. We also thank Cheng-Sheng Liu, Xiaoguang Li, Ruo-Mei Hsieh, Sze-
Wing Tang, Teresa Griffith, Pauline Huynh and Deng Wei for their native speaker judgements.
42 LISA L.-S. CHENG, C.-T. JAMES HUANG & C.-C. JANE TANG
In (1)-(3), the question is marked by a negation marker at the end of the

sentence. We call yes-no questions such as these Negative Particle Questions
(henceforth NPQs). As shown in (1)-(3), the negation marker appears at the end
of the sentence in NPQs, in contrast with the typical preverbal position of nega
tion markers, as shown in (4)-(6).
(4) Mandarin
hufei meiyou -wan nei-ben shu
Hufei not-have read-finish that-cL book
"Hufei did not finish reading that book."
(5) Cantonese
wufei mei lei
Wufei not-yet come
"Wufei has not come yet."
(6) Taiwanese
i bo tsiak beng
he not-have eat rice
"He did not eat."
In this paper, we will first discuss the negation markers in Mandarin,

Cantonese and Taiwanese as they are significant for the formation of NPQs. We
show that negation markers vary depending on the aspect or verb type. In sec
tion 3, we briefly consider NPQs in Classical Chinese, which sheds light on the
historical development of negation markers as question particles. We then com
pare the formation of NPQs in these three dialects in Chinese. It is shown that
Mandarin NPQs observe the typical agreement requirement between negation and
aspect/verb while Cantonese and Taiwanese do not maintain such a requirement
in NPQs. We argue that the contrast between Mandarin NPQs and
Cantonese/Taiwanese NPQs results from a difference in the derivation of NPQs:
NPQs in Mandarin Chinese involve the movement of a negation marker to the
sentence final position while no such movement is involved in the formation of
NPQs in Cantonese and Taiwanese.
CHINESE NEGATIVE PARTICLE QUESTIONS: A DIALECTAL COMPARISON 43
1.1 Are NPQs reduced A-not-A or VP-not-V questions?1
Before we proceed to the discussion on the formation of NPQs in different

dialects, we need first to address the question of whether or not they are derived
from other types of yes-no questions in Chinese. There are numerous ways of
forming yes-no questions in Chinese (see Appendix). Among them, two might
appear to resemble NPQs, namely A-not-A and VP-not-V questions. In particular,
one may question the status of NPQs as a different type from A-not-A and VP-
not-V questions. In this section, we examine data from Mandarin (with preverbal
adjuncts and sentence-final question particles) to show that in Mandarin, NPQs
cannot be reduced forms of either A-not-A or VP-not-V questions. Moreover, as
we will see in the discussion in Section 2, there is an asymmetry in the use of
various negation markers in A-not-A questions and NPQs in Cantonese and
Taiwanese, which offers further evidence for NPQs being separate from A-not-A
questions. See § 7 & 8 for further comparisons between NPQs and other types of
yes-no questions (see also Yue-Hashimoto 1988, 1992 and 1993).
1.1.1 Preverbal adjuncts

Non-temporal and locative preverbal adjuncts can appear in NPQs (7) but not
in A-not-A (8) and VP-not-V (9) questions.
(7) a. ta chang qu bu
he often go not
"Does he go often?"
b. ta yijing kan-wan shu meiyou
he already read-finish book not-have
"Did he already finish reading the book?"
(8) a. * ta һ qu-bu-qu

he often go-not-go
"Does he go often?"
b. * ta yijing you-meiyou kan-wan shu
he already have-not-have read-finish book
1
There are other types of questions on a par with VP-not-V, such as VP-not-VP and V-not-VP.
We will only discuss VP-not-V questions. The arguments can be easily extended to the other
types.
(9) a. * ta chang pian-ni-bu-pian

he often cheat-you-not-cheat
"Does he often cheat you?"
b. * ta yijing kan-wan shu mei-kan-wan
he already read-finish book not-read-finish
If NPQs are derived from A-not-A questions or VP-not-V questions by deleting

the post-negation part, the contrast between (7) and (8)-(9) cannot be explained.
1.1.2 Co-occurrence with ma/ne
In Mandarin Chinese, question particles such as ma and ne can occur in

sentence final position in questions, as shown in (10). Ma is a yes-no question
particle while ne is the optional WH-question particle.
(10) a. ta lai-le *ne/ma

h COme-PERF WH/Y-N
"Did he come?"
b. ta mai-le shenme (ne)/*ma
Һ buy-PERF What WH/Y-N
"What did he buy?"
As we can see in (11), A-not-A and VP-not-V questions can co-occur with the
question particle ne, though they cannot appear with ma.2 However, NPQs can
not co-occur with either ma or ne, as in (12).
(11) a. ta lai-bu-lai ne/*ma

he come-not-come WH/Y-N
"Is he coming?"
b. ta you-meiyou lai ne/*ma
he have-not-have come WH/Y-N
"Did he come?"
2
Though A-not-A questions take ne as a question particle, they are still interpreted as yes-no
questions. This may seem strange at first glance. However, given Huang's (1991) proposal
which treats the formation of A-not-A questions on a par with typical WH-questions, the fact
that ne is used is not unexpected.
c. ta xihuan ni-bu-xihuan ne/*ma

he like you-not-like WH/Y-N
"Does he like you?"
(12) a * ta qu bu ne/ma
he go not WH/Y-N
"Is he going?"
b. * ta you qian meiyou ne/ma
he have money not-have WH/Y-N
"Did he have money?"
Again, if NPQs are derived from A-not-A or VP-not-V questions by deleting

the post-negation elements, we would expect ne to be able to appear in NPQs.
However, as shown above, neither ma or ne can appear in this type of question.
The two arguments presented above show that NPQs are yes-no questions of
a different type from A-not-A and VP-not-V questions. They cannot be derived
from the latter types.
2. Negation Forms
The crucial element in NPQs is the negation marker. To understand the for
mation of NPQs, we must first consider the properties of negation in Chinese.
Every dialect in Chinese has more than one negation form. The negation form
varies depending on the aspectual markings on the verb or the verb type itself. In
other words, there is a matching or agreement requirement which holds between
the negation marker and the aspect/verb form. (In the following discussion, we
will not consider the negation marker in imperatives.)
2.1 Mandarín
Mandarin has two negation markers: bu and meiyou (see Wang 1965, Chao
1968 and Li & Thompson 1981).  is used with bare verbs and modals.
Meiyou is used with various aspects and with accomplishment verbs.3 In the ex-
3
There are some apparent counterexamples to this description of the usage ofbuw and meiyou.
As we can see in (i), bu seems able to appear with the aspectual marker -zhe "PROG":
(i) ta bu bao-zhe zhen-tou shui-jiao
he not hold-PROG pillow sleep
"He does not sleep by holding a pillow."
amples below, we see that meiyou cannot appear with the modal hui (13b),
while bu cannot appear with the perfective aspect -le (14b) or the experiential as
pect -guo (14e). (14a) shows that meiyou is interpreted as perfective without the
presence of the perfective aspect -le (and in fact, meiyou cannot co-occur with
-le (14c), see Wang (1965) for an account of this restriction).
(13) a. ta bu lai
he not come
"He is not coming."
b. hufei bu/*meiyou hui qu
Hufei not/not-have will go
"Hufei will not go."
(14) a. hufei meiyou qu xuexiao

Hufei not-have go school
"Hufei did not go to school."
b. * hufei bu qu-le xuexiao
Hufei not go-PERF school
 * hufei meiyou qu-le xuexiao
Hufei not-have go-PERF school
d. hufei meiyou qu-guo
Hufei not-have go-EXP
"Hufei has not been (there)."
e. * hufei bu qu-guo
Hufei not go-EXP
"Hufei has not been (there)."
However, it should be noted that (ii) is ungrammatical.

(ii) * ta bu bao-zhe zhen-tou
he not hold-PROG pillow
"He is not holding a pillow/'
Example (ii) shows that b u cannot appear with the progressive marker -zhe. The contrast be
tween (i) and (ii) is due to the fact that -zhe does not really have an aspectual reading in (i) but
rather an instrumental reading.
It should be noted that meiyou appears able to co-occur with neng "can":
(iii) ta mei(you) neng qu
he not-have can go
"He could not go."
Both of these negation markers can appear in NPQs as question particles.
(15) hufei hui qu bu

Hufei will go not
"Will Hufei go?"
(16) hufei qu-le meiyou

Hufei go-PERF not-have
"Did Hufei go?"
Note that both bu and meiyou can be used in A-not-A questions.4 As we will
see in subsequent sections, in Cantonese and Taiwanese, not all negation mark
ers can appear in A-not-A questions, further supporting our claim that NPQs can
not be derived from A-not-A questions.
(17) a. qiaofeng qu-bu-qu

Qiaofong go-not-go
"Is Qiaofeng going?"
b. qiaofeng you-meiyou qu
Qiaofeng have-not-have go
"Did Qiaofeng go?"
2.2 Cantonese
Cantonese has three negation forms: m, mou, and mei (see Cheung 1972
and Yue-Hashimoto 1993).5 M is used with bare verbs and modals (on a par
with bu in Mandarin) and cannot be used with aspectual markers (18a-c).  is
4
In Beijing Mandarin, instead of (17b), it is possible to say (i):
(i) qiaofeng qu-mei-qu
Qiaofeng go-not-go
"Did Qiaofeng go?"
5
It should be noted that Mandarin also has a negation marker, wei "not yet" which corresponds
to mei "not-yet" in Cantonese. However, the negation form wei "not-yet" has to co-occur with
the adverbials shang "yet" or hai "yet".
(i) a. ta shang/hai wei lai.
he yet not come
"He has not come yet."
b. * ta wei lai
he not come
The contrast in (i) shows that wei is no longer a free form in Mandarin.
used with various aspects and accomplishment verbs and, like meiyou in
Mandarin, it cannot co-occur with the perfective aspect marker and its mere
presence is interpreted as perfective. Mei is similar to mou except that the former
has an added meaning of "not yet".
(18) a. keoi m lei

he not come
"He is not coming."
b. keoi m/*mou/*mei hoyi lei
he not/not-have/not-yet can come
"He will not come."
 * keoi m lei-zo
he not come-PERF
"He didn't come."
(19) a. keoi mou lei

he not-have come
"He didn't come."
b. * keoi mou lei-zo
he not-have come-PERF
"He didn't come."
(20) a. keoi mei lei

he not-yet come
b. * keoi mei lei-zo
he not-yet come-PERF
In contrast with Mandarin, which allows both bu and meiyou to be used in

NPQs, in Cantonese, only mei can appear in NPQs:6
6
It should be noted that mou "not-have" can be used in NPQs only when the verb is the pos
sessive verb jau "to have", as shown in (i):
(i) keoi jau tsin mou
he have money not-have
"Does he have money?"
This indicates that there is a contrast between mou being the suppletive form of NEG plus the
aspectual jau "perfective" and mou being the suppletive form of NEG plus the verb jau "to
have". We will leave this issue open.
(21) a. keoi lei-zo mei

he come-PERF not
"Has he come yet?"
b. * keoi lei m
he come not
"Is he coming?"
c. * keoi lei-zo mou
he come-PERF not
"Did he come?"
However, both m and mou can appear in A-not-A questions while mei cannot.
(22) a. keoi lei-m-lei

he come-not-come
"Is he coming?"
b. keoi you-mou lei
he have-not-have come
"Did he come?"
 * keoi lei-mei-lei
he come-not-yet-come
"Has he come?"
This shows a complementary distribution of negation markers in these two types

of questions: the ones that appear in A-not-A questions cannot appear in NPQs.
We will come back to the complementary distribution noted here in section 5.4.
2.3 Taiwanese
Taiwanese has four monosyllabic negation markers, m, bo, be, and buei
(see P. Li 1971, Teng 1992 and T.-C. Tang 1993). M is the neutral negation, bo
the perfective negation, be the future negation and buei is the negation marker
indicating "not-yet".
(23) a. i m lai
he not come
"He is not coming."
b. і be lai
he not-FUT come
"He will not come."
 * i m/bo/buei e lai
he not/not-have/not-yet will come
"He will not come."
d. i m/*bo/*buei/be gaN chu-ki
he not/not-have/not-yet/not-FUT dare out-go
"He dare not/will not dare go out."
The examples in (23a-d) show that bo "not-have" and buei "not-yet" cannot
appear with modals (such as e "will" and gaN "dare"). M can appear with typical
modals except e "will" (probably due to the fact that be "not-future" is the sup
pletive form of NEG and e "will").
(24) a. i bo ki hak-hao
he not-have go school
"He did not go to school."
b. * i m u ki hak-hao
he not have go school
"He didn't go to school."
(25) . і a buei lai

he yet not-yet come
b. * і a buei lai a
he yet not-yet come-PERF
All four negation markers in Taiwanese can appear in NPQs.
(26) a. li ki m
you go not
"Are you going?"
b. i u tsiak beng bo
he have eat rice not-have
"Did he eat?"
c. i e ki be
he will go not-FUT
"Will he go?"
d. і ki buei
he go not-yet
"Has he gone?"
A-not-A questions in Taiwanese are restricted to certain verbs such as si "to be"
and only the negation marker m.
(27) . і si-m-si hakseng

he be-not-be student
"Is he a student?"
b. * і lai-m-lai
he come-not-come
"Is he coming?"
(28) a. * і lai-bo-lai
he come-not-have-come
"Did he come?"
b. * і lai-be-lai
he come-not-FUT-come
"Will he come?"
c. * і lai-buei-lai
he come-not-yet-come
"Has he come yet?"
Hence, Taiwanese offers further evidence for our claim that NPQs are not de
rived from A-not-A questions.
3. Classical Chinese
NPQs can be traced back to Classical Chinese—Pre-Qin Dynasty to Han

Dynasty. Zhang (1990) notes that the appearance of NPQs in Classical Chinese
predates the appearance of other types of yes-no questions. This further sup
ports our analysis that NPQs are not derived from other types of yes-no ques
tions. Furthermore, Zhang (1990) proposes that some NPQs are from [VP-NEG +
Q-PARTICLE]. He shows that there were sentences of the VP-NEG form followed
by a question particle, as in (29), where the negation is fou followed by the
question particle hu. It should be noted that Classical Chinese has more than a
dozen negation markers. However, only bu,fou, wei, fei, and wu can occur in
NPQs.
(29) mci ze dongxin fou-hu (Gongsunchou, Shang)

if-so then move-heart not-QPAimcLE
"If this is so, will you be inclined (to do it)?"
According to Zhang (1990), the sentence-final question particle later disap

peared in these cases (i.e., in sentences with negation following the VP), as in
(30). This development can be interpreted as either the incorporation of the
question particle into the negation (and thus fou at this stage was no longer a
mere negation marker) or the deletion of the question particle with the negation
taking over the function of the question particle.
(30) zhi ke fou (Zhuangzi, 10)

know possible not
"(Someone) knows whether it's possible."
(31) jie wei (Zhongbenqijing, 148)

understand not-yet
"Do (you) understand it yet?"
We also see [NEG+ QPARTICLE] as well as NEG as a question particle occurring

in the same text:
(32) jia-zhong suo you mi dang yu bu~ye (Zapiyujing, 509)

house-in have rice should give not QPARTICLE
"Should we give the rice in the house to (someone)?"
(33) you gui mai bu (Zapiyujing, 507)

have ghost sell not
"Do you have ghost for sale?"
The data in Classical Chinese above show the historical development of negation
markers as question particles. We will see that this sheds light on the formation
of NPQs in various dialects of Chinese.
4. Dialectal differences in NPQs
We have so far presented data showing that the three dialects under discus
sion differ with respect to the number of negation forms they have as well as
which one(s) can be used in NPQs. It is also clear that there is agreement be
tween the negation form and the aspect/verb. We now present data showing that
such agreement is maintained in NPQs in Mandarin while in Cantonese and
Taiwanese, the agreement collapses in NPQs.
Consider first examples of NPQs in Mandarin.
(34) a. * ta qu-le bu
Һ gO-PERF not
"Did he go?"
b. * ta qu-guo bu
Һ gO-EXP not
"Has he gone?"
c. ta qu bu
he go not
"Is he going?"
(35) a. ta qu-le meiyou

he go-PERF not-have
"Did he go?"
b. ta qu-guo meiyou
he go-EXP not-have
"Has he been (there)?"
(36) a. ta hui/yinggai/neng qu bu
he will/should/can go not
"Will/should/can he go?"
b. * ta hui/yinggai/neng qu meiyou
he will/should/can go not-have
"Will/should/can he go?"
In (34a-b), the negation marker bu which appears as a question particle

(henceforth NEG-particle) cannot appear with the perfective aspect -le or the ex
periential aspect -guo. In contrast, we can use the NEG-particle meiyou with
these two aspects, as shown in (35).7 (36) further shows that in NPQs, bu can
appear with modals while meiyou cannot. Hence, in Mandarin, the agreement
that we have seen in Section 2.1 between negation and verb/aspect is maintained
in NPQs as well.
In Cantonese, however, the agreement between negation and aspect/verb
does not seem to hold in NPQs. As noted earlier, mei "not-yet" is the only nega
tion form that can be used in NPQs. Thus, if agreement were to hold in NPQs in
Cantonese, we would expect that NPQs cannot contain modals, as mei cannot
appear with modals, as we have seen in (18b). It tums out however that though
mei carries the interpretation of "not-yet", it can still appear with modals (37), as
well as the typical perfective and experiential aspects in NPQs (38).8
(37) a. ngo hoyi/yinggoi ceot-heoi mei

I can/should go-out not-yet
"Can/should I go out?"
b. keoi hai-dou se seon mei
he PROG write letter not-yet
"Is he writing the letter?"
(38) a. keoi sik-zo fan mei

he eat-PERF rice not
"Has he eaten?"
b. keoi heoi-go meigok mei
he go-EXP America not
"Has he been to America?"
The sentences in (38) and (37) show that even though Cantonese exhibits
agreement between negation and aspect/verb in typical negation environments,
such agreement is not observed in NPQs.
Taiwanese is similar to Cantonese in that there is no strict match
ing/agreement requirement in NPQs. We have shown earlier that Taiwanese is
similar to Cantonese and Mandarin in that there is agreement between negation
7
There are different proposals which address the question of why -le cannot appear with
mei(you) in regular negation contexts. We assume here that whatever the constraint is, it is not
a semantic incompatibility and that the structural description that leads to the non-co-occur
rence in this case is no longer met when the negation is in the C° position.
8 There is a yes-no question particle me in Cantonese which differs from mei in that it is not a
negation marker and does not have any restriction on the verb form. Thus, it is quite similar to
the ma question particle in Mandarin.
and aspect/verb in typical negation sentences. However, there is no such agree

ment in NPQs in Taiwanese, as shown in (39)-(40).
(39) . і e lai m/bo/buei/be

he will come not/not-have/not-yet/not-FUT
"Will he come?"
b. i gaN chu-ki m/bo/buei/be
he dare out-go not/not-have/not-yet/not-FUT
"Does he dare to go out?"
(40) a. i u ki hak-hao m/bo

he have go school not/not-have
"Did he go to school?"
b. і lai buei
he come not-yet
"Has he come?"
The examples in (39) and (40) contrast with the ones in (23)-(25). In (23), we
see that m, bo, and buei cannot appear with the modal e "will". However, in the
NPQ in (39a), all the negation markers can appear with e "will". Similarly, in
(24b), we see that m cannot appear with the perfective aspect. In contrast, in the
NPQ in (40a) we see that m can be used even though the perfective marker  is
present.
In short, the formation of NPQs differs in these dialects. Mandarin NPQs re
tain the same kind of agreement observed in typical negation environments while
Cantonese and Taiwanese NPQs do not.
5. Analysis
Two apparent questions arise given the above data in Classical Chinese and
in the three synchronic dialects of Chinese:
i. What is the relationship between Classical NPQs and syn

chronic NPQs in different dialects?
ii. Why do dialects differ with respect to the presence of
agreement in NPQs?
We have noted in Section 3 that the development in Classical Chinese NPQs

can be considered as incorporation of the question particle into the negation
marker ('incorporation' here is used in a non-technical sense). Hence, NEG takes
over the function of the question particle and the presence of NEG in the sentence
final position indicates a yes-no question. Turning now to synchronic NPQs in
the three dialects under investigation, we do not see the co-occurrence of NEG
and a Q-particle. 9 Instead, we see only NEG in the sentence final position.
Hence, one simple way of looking at the synchronic data is that the NEG in the
sentence final position is no longer a simple NEG but a NEG with whatever fea
tures that a Q-particle has. However, this simplistic view does not explain the
second question, that is, it does not offer an explanation for the dialectal differ
ence we have observed in terms of agreement in NPQs.
Before we address this question, we must first consider the position of the
NEG-particle. We assume, following T.-C. Tang (1989), that the sentence final
question particles in Mandarin (e.g. ne, ma) as well as those in other dialects are
in the C° position. Since negative particles are sentence final and they also mark
yes-no questions, we consider them on a par with other question particles in that
they are also in the C° position, though they clearly differ from typical question
particles in their ability to function as negation markers in a sentence.
5.1 Agreement vs. non-agreement dialects
Let us now turn to the second question, the question of dialectal difference
with respect to agreement. Consider first the dialects which lack agreement in
NPQs (i.e., Cantonese and Taiwanese). We consider these dialects to resemble
Classical Chinese in the formation of NPQs. In particular, as we have noted ear-
9
It should be noted that in Shanghai, the sentence final negative particle has a 'literal' incor
porated question particle, as shown below:
(i) yi φ le a (ii) yi φ le va (iii) yi ve qi
Һ gO PERF Q Һ gO PERF not-Q he not gO
"Did he go?" "Did he go?" "He is not going."
As we can see in (iii), the typical negation form is ve and the one used in NPQ is the combina
tion of ve and the question particle a. We need to examine Shanghai more closely to see the
pattern of NPQs. Furthermore, it is pointed out to us (Sybesma, p.c.) that in a Northern dialect
of Mandarin, NPQs with bu can co-occur with ma, though the ordering of the Q-particle and the
negation marker appears to differ from what we find in Classical Chinese:
(iv) ni qu ma bu
you go Q not
"Are you going?"
Further tests are needed to see whether these are genuine NPQs (see Appendix for some basic
tests).
lier, in Classical Chinese the negation markers were grammaticalized as question

particles. In other words, we propose that the negation markers in the non-
agreement dialects are base-generated in the C° position on a par with typical
question particles. The lack of agreement is thus naturally explained since the
base-generated negation markers which are used as question particles are not in a
position which can be construed with verb types and aspect types. Certain issues
related to these two non-agreement dialects as well as to the nature of negation
remain and we will come back to these issues in section 5.3.
Consider now Mandarin, the dialect which displays agreement in NPQs. We
have seen that Mandarin NPQs observe the same agreement restriction that holds
between negation and verb/aspect (as in (34)-(36)). That is, the use of bu vs.
meiyou depends on the verb/aspect or modal in the sentence regardless of
whether or not the negation markers are used as regular negative markers or
question particles. Suppose for the moment that the agreement that we see be
tween negation and verb/aspect is due to a selecţionai relation between negation
and verb/aspect. The agreement phenomenon in NPQs in Mandarin can be cap
tured if the negation marker moves to the C° position in overt syntax. The
agreement which holds for typical negation forms thus also holds for NEG-parti-
cles since they are in fact the same elements. Given this hypothesis, the differ
ence between Mandarin and Cantonese/Taiwanese in the formation of NPQs is
that the former involves movement of a negation marker to C° while the latter in
volves a base-generated negation marker in the C° position.
Some immediate questions arise given this analysis:
(41) a. Why can't negation markers in Mandarin be base-generated

in NPQs?
b. Why can negation markers in Mandarin be moved to the C°
position?
 Why can't negation markers in Cantonese/Taiwanese un
dergo movement in NPQs?
d. Why is it the case that some negation markers cannot be
used in NPQs?
These questions are related to the nature of negation as well as to the nature
of the NEG-particle in these dialects. Before we turn to these questions, we will
first consider some supporting evidence for the movement vs. base-generation
distinction.
5.2 Supporting evidence
We have so far examined simplex NPQs, which illustrate a basic dialectal dif
ference in terms of agreement. Below we present data involving verbs which
take clausal complements. In particular, we consider sentences in which the
agreement requirement of the matrix verb differs from that of the embedded
verb. We show that such cases provide further evidence for a movement analy
sis in the formation of NPQs in Mandarin Chinese. Consider first a very simple
case, where the matrix and embedded verbs share the same agreement require
ment:
(42) ta yiwei ni qu bu
he think you go not
(a) "Does he think or not think that you are going?"
(b) "Does he think that you are going or not going?"
(43) a. ta bu yiwei ni (hui) qu

he not think you will go
"He doesn't think that you will go."
b. ta yiwei ni bu qu
he think you not go
"He thinks that you are not going."
In (42), the NEG-particle is in the matrix C° indicating that the sentence is a

matrix yes-no question. (We will come back to cases with embedded questions.)
The matrix verb yiwei "to think" and the embedded verb qu "to go" can both oc
cur with the negation marker bu. As the (a) and (b) readings indicate, the sen
tence in (42) is ambiguous. We call the (a) readings the matrix reading and the
(b) reading the embedded reading. For the moment, we simply assume that the
embedded reading arises when the NEG-particle moves to the matrix (regardless
of whether or not the NEG-particle originates from the embedded NEG° or C°).
Compare (42) with (44) below:
(44) * ta hui yiwei ni yinggai qu meiyou

he will think you should go not-have
(a) "Will he think or not think that you should go?"
(b) "Will he think that you should go or not go?"
In contrast with the grammatical and ambiguous (42), (44) is ungrammaticai

(i.e., neither the matrix nor the embedded reading is available). The ungrammat
ically of (44) is in fact not surprising. Recall that the NEG-particle in Mandarin
must 'agree' with the verb/aspect of the sentence. The NEG-particle in (44) can
not agree with the matrix or the embedded predicate: the matrix contains the
modal hui "will" while the embedded clause has the modal yinggai "should".
Since meiyou cannot appear with modals, (44) is ungrammaticai.
The ungrammaticality of (44) is not at all surprising given our basic general
ization that the NEG-particle needs to agree with the verb/aspect type. Assume for
the moment, in contrast with the movement hypothesis presented earlier, that the
agreement requirement is a result of some non-local constraint. That is, let us
tentatively assume a constraint that requires the NEG-particle in C° to agree with
the verb/aspect type. Such a constraint may indeed account for the sentences in
(42) and (44). In (42), the embedded reading can be accounted for if we assume
that bu is base-generated in the embedded C° position and subsequently moves
to the matrix C° position (due to the fact that verbs such as yiwei "to think" do
not take embedded questions). The matrix reading will simply involve a base-
generated NEG-particle in the matrix C°. On the other hand, in (44), the con
straint will rule out both matrix and embedded readings since the NEG-particle bu
cannot be generated in either C° position because of the incompatibility between
the matrix and embedded verb/aspect.
Such a non-local constraint however runs into problems when there are
mixed verb/aspect types in the sentence. The data presented above are sentences
in which the matrix and the embedded predicate belong to the same type with re
spect to agreement with the NEG-particle. Consider now 'mixed' cases in (45)
and (46) below.
(45) ta yiwei ni qu-guo bu

he think you go-EXP not
a. "Does he think or not think that you have been (there)?"
b. * 'Does he think that you have been (there) or you haven't
been (there)?"
(46) ta hui yiwei ni qu-guo meiyou

he will think you go-EXP not-have
a. * 'Will he think or not think that you have been (there)?"
b. "Will he think that you have been (there) or you haven't
been (there)?"
In (45), the NEG-particle is bu and only the matrix verb satisfies the agree
ment requirement since the embedded one has the experiential marker attached to
it. As expected, the question does not have an embedded reading. On the other
hand, the NEG-particle in (46) is meiyou and only the embedded predicate can
agree with it since the matrix has the modal hui "will". And again, as expected,
the question does not have a matrix reading.
If the agreement constraint is some sort of non-local constraint on the C° and
the verb/aspect, it is possible to account for (45) but not for (46). For (45), it is
still possible to maintain that the NEG-particle cannot be base-generated in the
embedded C° due to the constraint. Thus, the only possibility is for it to be base-
generated in the matrix. However, (46) presents a problem for such an analysis.
It should be noted that even though the allowed reading in (46) is an embedded
reading, it is still a matrix question. Thus, though meiyou is allowed to be base-
generated in the embedded clause since it can occur with the experiential aspect
marker -guo, it has to move to the matrix C°. That is, the NEG-particle will
eventually end up in the matrix C°, even though the modal hui "will" is present
in the matrix. The non-local agreement constraint which rules out sentences such
as (44) will also rule out (46).
The above data show that if NEG-particles in Mandarin were to be base-gen
erated in C° positions, sentences such as (46) cannot be accounted for. On the
other hand, given a movement analysis, the grammaticality and the readings of
the sentences in (42), and (44)-(46) naturally follow. Let us consider the sen
tences one by one. In the grammatical and ambiguous (42), the NEG-particle bu
can be either generated in the embedded NEG or matrix NEG. In the embedded
case, the NEG-particle moves to the matrix C° via the embedded C°. In (44), the
NEG-particle meiyou cannot be generated in either the embedded NEG or the ma
trix NEG due to the selectionai restriction between the verb/aspect and the NEG
and thus the sentence is ruled out. On the other hand, in (45), even though the
NEG-particle bu cannot be generated in the embedded NEG due to the experiential
aspect marker -guo, it can be generated in the matrix NEG and subsequently
moves to the matrix C°. Turning now to the problematic case for the non-move
ment analysis, in (46) we see that the NEG-particle meiyou can indeed be base-
generated in the embedded NEG. As in the ambiguous case in (42), meiyou first
moves to the embedded C° and it subsequently moves to the matrix C°. The
movement from the embedded C° to the matrix C° does not involve the matrix
predicate, nor does it involve the matrix NEG. Hence, even though the
verb/aspect type of the matrix in (46) does not appear to agree with the NEG-par
ticle which ends up in the matrix C°, the sentence is still grammatical, with the
embedded reading.
A question that arises given such mixed cases is whether or not Cantonese
and Taiwanese data differ from the Mandarin data presented above. Given the
fact that Cantonese and Taiwanese do not display agreement in NPQs, it is ex
pected that even in mixed cases, ambiguous readings are allowed since NEG-par-
ticles are base-generated in C°s and can undergo movement. This prediction is
borne out, as (47)-(54) show.10
Taiwanese
(47) i giosi li (e) ki m
he think you will go not
(a) "Does he think or not think that you are leaving?"
(b) "Does he think that you are leaving or not leaving?"
(48) і e giosi li yinggai ki bo
he will think you should go not-have
(a) "Will he think or not think that you can leave?"
(b) "Will he think that you can leave or cannot leave?"
(49) і giosi li ki-gue m

he think you go-EXP not
(a) "Does he think or not think that you have left?"
(b) "Does he think that you have left or you have not left?"
(50) і e giosi li ki-gue bo

he will think you go-EXP not-have
(a) "Will he think or not think that you have been (there)?"
(b) "Will he think that you have been (there) or you haven't
been (there)?"
Cantonese
(51) keoi yiwai ni zau mei
he think you leave not-yet
(a) "Does he think or not think that you are leaving?"
(b) "Does he think that you are leaving or not leaving?"
(52) keoi wui yiwai ni hoyi zau mei
he will think you can leave not-yet
(a) "Will he think or not think that you can leave?"
(b) "Will he think that you can leave or cannot leave?"
10
It should be noted that even though the sentences are ambiguous, in some cases, there is a
preferred reading.
(53) keoi yiwai ni zau-zo mei

he think you leave-PERF not-yet
(a) "Does he think or not think that you have left?"
(b) "Does he think that you have left or you have not left?"
(54) keoi wui yiwai ni heoi-gwo mei

he will think you go-EXP not-yet
(a) "Will he think or not think that you have been (there)?"
(b) "Will he think that you have been (there) or you haven't
been (there)?"
As we can see, all the counterparts of (42)-(46) allow ambiguous readings,

further showing that agreement does not play a role in the formation of NPQs in
both Cantonese and Taiwanese. Before we conclude this section, we would like
to point out that in Mandarin, as well as Taiwanese, the neutral negation marker
(bu and m respectively) cannot indicate embedded questions, in contrast with the
other negation markers.
Mandarin
(55) ta xiang-zhidao ni lai-le meiyou
he wonder you come-PERF not-have
"He wonders whether you came."
(56) * ta xiang-zhidao ni qu bu
he wonder you go not
"He wonders whether you are going."
The contrast between meiyou and bu in their ability to indicate embedded yes-no
questions is illustrated in (55) and (56). The verb xiang-zhidao "to wonder"
requires an embedded interrogative and the ungrammaticality of (56) is due to
the fact that bu cannot indicate an embedded question. This property of bu is on
a par with the typical yes-no question particle ma, as we see in (57) and (58):
(57) huangrong zhidao hufei yijing zou-le ma

Huangrong knows Hufei already leave-PERF Q
(a) "Does Huangrong know that Hufei already left?"
(b) * "Huangrong knows whether or not Hufei left."
(58) * huangrong xiang-zhidao hufei zou-le ma

Huangrong wonder Hufei leave-PERF Q
"Huangrong wonders whether Hufei left."
It thus appears that bu is similar to ma in that both have a 'matrix property'. The
matrix property of ma has been attributed to the speaker-oriented interpretation
of ma. We suggest that bu is on a par with ma in its speaker-oriented property.
Taiwanese m also cannot indicate embedded questions, in contrast with the
other negation markers, as shown in (59).
(59) і shung-be-zaiyaN li e ki *m/bo/be/buei

he want-to-know you will go not/not-have/not-FUT/not-yet
"He wants to know whether you are going."
Hence, the speaker-oriented property is unrelated to the movement of the nega

tion marker to form NPQs.
5.3 Content of negation
We have shown that data involving embedded sentences with mixed

verb/aspect types present further support for our hypothesis of the dialectal dif
ferences in the formation of NPQs. We now tum to the questions raised earlier in
(41), repeated below as (60):
(60) a. Why can't negation markers in Mandarin be base-generated

in NPQs?
b. Why can negation markers in Mandarin be moved to the C°
position?
c. Why can't negation markers in Cantonese/Taiwanese un
dergo movement in NPQs?
d. Why is it the case that some negation markers cannot be
used in NPQs?
These questions all relate to the properties of negation and C°.

Following Cheng (1991), we assume that the clause type of interrogative
sentences can be marked by the insertion of a question particle or by movement
of an appropriate element to the C° position or to SPEC of  (see also Chomsky
1995). Consider first the non-movement dialects. In Cantonese and Taiwanese,
we maintain that some negation markers are base-generated in C° on a par with
typical question particles. In other words, these negation markers must carry the
formal feature that marks a sentence as a yes-no question. Let us assume it to be
[Q] (Chomsky 1995, among others). Thus, they are elements with a dual status,
as negation markers or as question particles. In other words, these elements
have a [Q, NEG] feature. Note that we maintain that only certain negation mark-
ers are both negative markers and question particles because not every negation
marker can be used in NPQs (for example, in Cantonese, only mei can be used in
NPQs). These 'dual status' negation markers are thus like some negation markers
in Classical Chinese in that they can function as a negation marker or as a ques
tion particle.
Let us turn now to the movement dialect, Mandarin. Following Chomsky
(1995), we assume that overt movement is triggered by unchecked features.
Hence, in Mandarin, movement of the negation markers is to check some formal
feature in C°. Note that the negation markers in Mandarin cannot be inserted as
question particles in C° to mark yes-no questions. In other words, negation
markers in Mandarin do not have a dual status, even though they do appear in
C°. We propose that Mandarin has a phonologically null C° with the formal fea
tures [Q, NEG]. This C° can be considered to be another residue of the historical
development of negation as question particles. That is, instead of having a full-
fledged negation marker functioning as a question particle, Mandarin has a C°
with [Q, NEG] features, with just a 'trace' of negation in it. This [NEG] feature
however has to be checked off in order for the sentence to be properly inter
preted as a yes-no question. Thus, negation markers bu and meiyou can and
must move to C° to check the [NEG] feature.
The picture that has emerged here provides answers to the questions in (60).
In particular, the answers all relate to the properties of  or NEG. With respect to
Mandarin, its negation markers cannot be base-generated in NPQs because they
are 'pure' negation markers rather than those with a dual status. And what
'allows' negation markers to move in Mandarin is the particular feature [NEG] in
the C°. The answer to the question raised in (60c) may have to do with
Economy. Given the fact that Cantonese and Taiwanese also have negation
markers, why is it the case that they cannot undergo movement, just as negation
markers do in Mandarin? There are indeed two different possibilities:
(a) Cantonese / Taiwanese also has a C° with a [Q, NEG] feature;
(b) Cantonese / Taiwanese does not have such a C°.
Consider possibility (b) first. If these two dialects do not have such a C°, there is
then no motivation for the negation markers to undergo movement. On the other
hand, if we have possibility (b), the trigger for movement is present, and the
question then is why movement is lacking. Note however that these dialects
have negation markers as question particles which can be directly merged into
the computation. Assuming that Merge is less costly than Move (see Chomsky
1995), the Merge option and thus the insertion of these negation markers always
rules out the movement possibility.
5.4 Extension
We have pointed out in the initial discussion of negation markers in these

three dialects that the distribution of negation markers in NPQs and in A-not-A
questions may differ. In the following summary of this distribution, * indicates
that the negation marker cannot appear while√indicates that it can.
Distribution of NPQs and A-not-A Questions

NPQs A-not-A Questions
Cantonese
m * v
mou * v
mei √ *
Taiwanese √
m √ (*)
bo √ √
be √ √
bue √ √
Mandarin
bu √ √
mei √ √
meiyou √ *
Note that in Mandarin, mei can be used alone in both NPQs and A-not-A
questions. The difference between mei and meiyou in A-not-A questions is
shown in (61) (see also foonote 4). We assume here that mei is a reduced form
of meiyou, which has a verbal element vow "to have" in it.
(61) a. ta lai-mei-lai
he come-not-have
"Did he come?"
b. * ta lai-meiyou-lai
he come-not-have-come
"Did he come?"
The pattern of distribution shown above also has a dialectal split: in

Cantonese and Taiwanese, the negation markers used in NPQs and those used in
A-not-A questions are in complementary distribution; in contrast, Mandarin
negation markers do not show complementary distribution (aside from meiyou,
which we will come back to immediately below). The complementarity in
Cantonese and Taiwanese as well as the dialectal split may appear to be mysteri
ous. However, we suggest that the analysis proposed in this paper together with
Huang's (1991) analysis of A-not-A questions provide an answer to the above
distribution.
Huang (1991) proposes that the formation of A-not-A questions involves a
PF-insertion of a negator. In other words, the negator in A-not-A questions does
not enter into the computation. At PF, the insertion of a negator is to occupy the
'not' slot in A-not-A questions. It is thus reasonable to assume that only 'pure'
negators can be inserted. The complementary distribution in Cantonese and
Taiwanese is thus explained. In both dialects, the negators that can be used in
NPQs have dual status. They thus do not qualify as 'pure' negators. On the other
hand, in Mandarin, negators differ from the ones in Cantonese/Taiwanese in that
they are simple NEG elements and thus they can also be inserted in A-not-A ques
tions. Note that meiyou cannot be inserted in A-not-A questions because meiyou
is a composite form consisting of both the negator mei and the verb you "to
have". Thus PF insertion will only see mei listed as a NEG.
The analyis of NPQs presented above thus provides a rather simple account
of the complementarity noted as well as the dialectal split in terms of A-not-A
questions.
6. Conclusion
We have argued that the main dialectal difference between Mandarin and
Cantonese/Taiwanese is due to the NEG°-to-C0 movement in the former and the
lack of it in the latter. We have seen a basic difference in terms of agreement
patterns in these two types of dialects. Sentences involving embedding further
support our claim that in Mandarin, there is NEG°-to-C° movement.
The dialectal variation we have seen may be traced back to historical devel
opment of negation markers as question particles. It appears that
Cantonese/Taiwanese maintains the grammaticalization of negation markers as
question particles and thus these negation markers can be simply inserted in the
C° position to form a yes-no question. On the other hand, Mandarin negation
markers are no longer question particles. The only 'trace' of the grammaticaliza
tion of negation markers as question particles that remains in this dialect can be
seen in the C° feature [Q, NEG], which triggers the movement of NEG in NPQs.
Lastly, we would like to point out that NPQs are not just found in Chinese
dialects. It appears that NPQs exist in languages such as Cambodian, Thai and
Vietnamese:
Cambodian (Griffith p.c.).

(62) a. Sowan min mәә1 sәphin nul tee
Sowan not read book that NEG
"John didn't read the book."
b. Sowan әә1 sәphin nul tee
Sowan read book that Q
"Didn't John read the book?"
(63) a. * Sowan min әә1 sәphin nul tee

Sowan not read book that Q
b. * Sowan min mәә1 sәphin nul tee tee
Sowan not read book that NEG Q
Thai (Noss 1964)

(64) thaan kaa-fεε iig: mãj
want coffee more yes/no
"Will you have some more coffee?"
(65) phom mâj-khəəj paj: 1әәj

I not-have go there
"I have not gone there."
Vietnamese (Huynh p.c.)

(66) a. John co hon Mary không
John has kiss Mary not
"Has John kissed Mary?"
b. John không có hon Mary
John not has kiss Mary
"John has not kissed Mary."
If our analysis is correct, it may be extended to explain these sentences in

languages other than Chinese. We leave this for future research.
Appendix
Comparison of NPQs with other kinds of yes-no questions
We examine here in detail the properties of NPQs and show that they are in
terpreted as other types of yes-no questions (such as Mandarin ma questions,
tag-questions, haishi-questions, A-not-A questions, VP-not-V questions, VP-VP
questions), but that they differ from other types of yes-no questions. We will
discuss Mandarin and Taiwanese, the former a representative of the movement
type in NPQs, the latter a representative of the base-generation type in NPQs.
Mandarin yes-no questions
Before we compare NPQs with the other types of yes-no questions, we first
provide a brief overview of some yes-no question types in Mandarin.
(67) Μα-question
ta lai-le ma
he come-PERF Q
"Did he come?"
(68) Tag-question
ta hui lai, bu shi ma
he will come not be Q
"He is coming, isn't he?"
Μα-questions are characterized by the sentence final particle ma. As we can see,
ma is also needed in tag-questions.
(69) Haishi-qaestion
ta lai haishi bu lai
he come or not come
"Is he coming or is he not coming?"
(70) A-not-A question
ta xihuan-bu֊xihuan ni
he like-not-like you
"Does he like you?"
(71) VP-not-V question
ta xihuan ni-bu-xihuan
he like you-not-like
"Does he like you?"
(72) VP-VP question

ni chi fan chi mian
you eat rice eat noodle
"Do you want to eatriceor noodle?"
VP-VP questions do not use any question particle or conjunction. The question is
indicated by the juxtaposition of VPs.
Comparisons
Besides the two differences noted in section 1.1. in the main text, there are
other differences between NPQs and the other types of yes-no questions.
Negated vs. non-negated verb
Some yes-no questions have a constraint on the verb form. In particular, the
constraint is that the verb has to be affirmative.11 For example, in NPQs, the
verb has to be affirmative as shown in (73).12
(73) a. ta qu bu
he go not
"Is he going?"
b. * ta bu qu bu
he not go not
"Isn't he going?"
c. ta ku-le meiyou
he cry-PERF not-have
"Did he cry?"
d. * ta meiyou ku meiyou
he not-have cry not-have
"Didn't he cry?"
The same restriction can be found in A-not-A questions and VP-not-V questions,
as in (74) and (75).
11
Some yes-no questions are irrelevant in this sub-section, such as ma-questions, tag-ques
tions andhaidhi'-questions.Themaz-questionsare irrelvant because the questions cannot be con
sidered to consist of a yes and a no part. As for tag-questions and haishi-questions, there is no
requirement on the co-occurrence of yes and no parts. VP-VP questions require that both VPs
consist of non-negated verb forms.
12
The NPQs with the negation marker m in Taiwanese are an exception. See example (86) for
details.
(74) a. ta qu-bu-qu
he go-not-go
"Is he going?"
b. * ta bu qu-qu
he not go-go
"Isn't he going?"
c. ta you-meiyou qu
he have-not-have go
"Did he go?"
d. * ta meiyou qu qu
he not-have go go
"Didn't he go?"
(75) a. ta xihuan ni-bu-xihuan

he like you-not-like
"Did he like you?"
b. * ta bu xihuan ni-xihuan
he not like you-like
"Did he like you?"
c. ta kan-wan nei-ben shu meiyou kan-wan
he read-finish that-cL book not-have read-finish
"Did he finish reading that book?"
d. * ta meiyou kan-wan nei-ben shu kan-wan
he not-have read-finish that-cL book read-finish
"Did he finish reading that book?"
This property alone may indicate that NPQs seem to be related to A-not-A
questions or VP-not-V questions. However, we saw earlier that NPQs cannot be
derived from A-not-A or VP-not-V questions: preverbal adjuncts and ne, which
may appear with these question types, may not appear in NPQs.
Co-occurrence with aspect markers
Each Chinese dialect has a number of aspectual markers. As already pointed
out in section 2.1., aspect markers can appear in NPQs, as shown in (76).
(76) a. ta qu-guo meiyou

he go-EXP not-have
"Has he been there?"
b. ta qu-le meiyou
he go-PERF not-have
"Has he gone?"
However, neither A-not-A questions nor VP-not-V questions take aspectual

markers.
(77) a. * ta you-mei-you-guo qian

he have-not-have-PERF money
"Did he ever have money?"
b. ta you-mei-you qu-guo meiguo
he have-not-have go-EXP America
"Has he been to America?"
 * ta you-mei-you-le qian
he have-not-have-PERF money
"Did he have money?"
Note that in (77b), at a first glance, there seems to be co-occurrence between A-

not-A questions and aspectual markers. However, comparison of (77b) and
(77a, c) shows that the verb which has the A-not-A form cannot take aspectual
marking. In (77b), it is the aspectual marker you "have" which carries the A-not-
A form while the main verb qu "go" takes the aspectual marker guo.
So far we have seen that NPQs differ from the other types of yes-no ques
tions. If NPQs are derived from reduced forms of the other yes-no questions, the
asymmetry noted above cannot be accounted for.
Conflict of presupposition
We have so far shown that NPQs differ from A-not-A, VP-not-V and VP-VP
questions. Now let us turn to the difference between NPQs and ma-questions. As
noted in Li & Thompson (1981), ma-questions serve to question the validity of a
statement (example (78) is from Li & Thompson 1981:549).
(78) Speaker A:
ni haoxiang shou-le yidian
you seem thin-PERF a little
"You seem to have lost some weight."
Speaker B:
shi ma? ni kan wo shou-le ma?
be Q yOU See I thin-PERF Q
wo ziji dao bu juede
I self on:the:contrary not feel
"Is that so? Do you think I've lost weight?
I haven't noticed it myself."
Speaker B:
* shi-bu-shi? ??ni kan wo shou-le meiyoul
be-not-be you see I thin-PERF not-have
wo ziji dao bu juede
I self on:the:contrary not feel
As we have seen in (78), to question the validity of speaker A's statement,

speaker  can use a ma question but not an A-not-A question nor an NPQ. The
latter question types are used in neutral contexts.
Co-occurrence with nandao and daodi
Another difference between ma-questions and NPQs is that adverbials like
nandao "really" only appear in ma-questions.
(79) a. nandao ta hui qu ma

really he will go Q
"Is he really going?"
b. * nandao ta hui qu bu
really he will go not
 * nandao ta lai-le meiyou
really he come-PERF not-have
"Did he really come?"
In contrast, adverbials like daodi "on earth" can only occur in NPQs.13
(80) a. * daodi ta hui qu ma

on-earth he will go Q
13
In addition to NPQs, daodi can also appear in haishi-questions, A-not-A questions and VP-
not-V questions. It however cannot occur in VP-VP questions.
b. daodi ta hui qu bu
really he will go not
c. daodi ta lai-le meiyou
really he come-PERF not-have
Taiwanese yes-no questions
Compared to Mandarin, Taiwanese has ma-type questions, tag-questions
and haishi-type questions, but not A-not-A questions (except in cases with the
copula si "be"), VP-not-V questions, and VP-VP questions. However, Taiwanese
has other kinds of yes-no questions like gam-questions and VP-a-VP ques
tions. 14 ՛ 15
(81 ) Μa-type question

і lai- hio
he come-PERF Q
"Did he come?"
(82) Tag-question
і m lai, si bo
he not come be not-have
"He is not coming, is he?"
(83) Haishi-type question
i e lai asi be lai
he will come or not-fut come
"Is he coming or is he not coming?"
(84) Gam-question
i gam e lai
he Q will come
"Is he coming?"
(85) VP-a-VP question
і jiak beng a jiak miN
he eat rice or eat noodle
"Does he want to eat rice or noodle?"
14
We will not discuss all kinds of Taiwanese yes-no questions here.
15
For a discussion of the pragmatic function of various Taiwanese sentence final question par
ticles, see Chen (1993).
Comparisons
Let us now turn to the comparison between Taiwanese NPQs and these other
types of yes-no questions.
Positive and negative
Except for m, in NPQs the verb has to be in a non-negation form, as shown
in (86).
(86) a. i ki/ m ki m
he go/not go not
"Is he going?/Is he not going?"
b. і u/*bo ki bo
he have/not-have go not-have
"Did he go?"
 і e/*be ki be
he will/not-FUT go not-FUT
"Will he go?"
Since Taiwanese has neither A-not-A nor VP-not-V questions, no comparison can
be made with such sentences. This constraint is irrelevant for /no-questions,
sibo-questions, asi-questions and gam-questions. The hio- and gam-questions
are irrelevant because they cannot be treated as consisting of a yes and no part.
As for sibo-questions and asi-questions, there is no requirement on the co-oc
currence of yes and no parts. VP-a-VP questions require both VPs to consist of
negation or non-negation verb forms.
Non-temporal preverbal adjuncts
As with Mandarin NPQs, in Taiwanese it is possible for NPQs to have pre-
verbal adjuncts, as in (87).
(87) . і tiaNüaN ki m
he often go not
"Does he go often?"
b. i yiting e ki be
he certainly will go not-FUT
"Will he certainly go?"
Note that, except for VP-a-VP questions, other types of Taiwanese yes-no ques
tions can also take non-temporal/locative preverbal adjuncts.
(88) . і yiging lai- hio

he already MR-PERF Q
"Did he already come?"
b. і jinjiaN m lai, shi bo
he really not come be not-have
"He is really not coming, is he?"
c. і yiting e lai asi be lai
he certainly will come or not-FUTcome
"Is he definitely coming or is he definitely not coming?"
d. і gam üaNtiaN e lai
he Q often will come
"Is he coming often?"
(89) ?* і tiaNtìaN jiak beng a jiak miN

he often eat rice or eat noodle
"Does he often eat rice or noodles?"
Aspectual markers
In Taiwanese NPQs, aspect markers can appear, as in (90).
(90) . і ki-gue bo
he go-EXP not-have
"Has he ever been there?"
b. i jiak-a buei
he eat-PERF not-yet
"Has he eaten?"
However, aspect markers cannot appear in VP-a-VP questions, though they can
occur in the other types of Taiwanese yes-no questions.
(91) * і ki-gue migok a ki-gue yinggok

he go-EXP America or go-EXP England
"Has he been to America or England?"
Co-occurrence with other particles

According to Chen (1993), there are at least nine kinds of sentence final
question particles in Taiwanese. They are -haN, -hio, -hoN, -le, -lio, ֊lo, -ne,
-ni, and -o. With respect to -hio, Chen claims that it cannot appear in WH-ques-
tions, disjunctive questions, hypothetical questions, truncated questions, and
confirmation questions. It seems that NPQs and non-gam-questions are all barred
from taking this question particle.
(92) a. * i ki m hio
he go not Q
"Does he go?"
b. * i u ki bo hio
he have go not-have Q
'Oid he go?"
c. * i e ki be hio
he will go not-FUT Q
"Will he go?"
(93) a. * і jinjiaN m lai, si bo hio

he really not come be not-have Q
"He really isn't coming, is he?"
b. * і yiting e lai asi be lai hio
he certainly will come or not-fut come Q
"Will he certainly come or not come?"
 ? і gam üaNtiaN e lai hio
he Q often will come Q
"Is he coming often?"
d. * і jiak beng a jiak miN hio
he eat rice or eat noodle Q
"Does he eat rice or noodles?"
Presupposition
According to Chen (1993), -hio is used to indicate strong assumption, but
NPQs are used in neutral contexts.
Co-occurrence with gamgong and daote
Lastly, the adverbials like gamgong "really" can appear only in A/ø-type
questions, whereas adverbials such as daote "on earth" can occur only in NPQs.
(94) . і gamgong m lai hio

he really not come Q
"Does he really not come?"
b. * i daote m lai hio
he on-earth not come Q
"Is he really not coming?"
(95) a. * і gamgong u lai bo

he really have come not-have
b. і daote u lai bo
he on-earth have come not-have
The comparisons discussed above have clearly shown that while NPQs and
certain other questions are semantically yes-no questions, they are syntactically
very distinct from one another.
REFERENCES
Chao, Yuen-Ren. 1968. A Grammar of Spoken Chinese. Berkeley and Los
Angeles: University of California Press.
Chen, Janet C.-W. 1993. "Taiwanese Sentence-final Question Particles".
Proceedings of the First Symposium on Languages in Taiwan, C4, 01-30. Taipei:
National Taiwan University.
Cheng, Lisa L.-S. 1991. On the Typology ofW/h-questions.PhD dissertation, MIT.
Cheung, Samuel H.-N. 1972. Cantonese as Spoken in Hong Kong. Hong Kong:
The Chinese University of Hong Kong.
Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press.
Huang, C.-T. James. 1991. "Modularity and Chinese A-not-A questions".
Interdisciplinary Approaches to Language: Essays in Honor of S.-Y. Kuroda,
ed. by Carol Georgopoulos & Roberta Ishihara, 305-332. Dordrecht: Kluwer.
Li, Charles & Sandra A. Thompson. 1981. Mandarin Chinese: A Functional
Reference Grammar. Berkeley: University of California Press.
Li, Paul. 1971. "Two Negative Markers in Taiwanese". Bulletin of the Institute of
the History of Philology 43.201-220. Institute of History and Philology,
Academia Sinica, Taiwan.
Noss, Richard . 1964. Thai Reference Grammar. Washington D.C.: Foreign
Service Institute, Department of State.
Tang, T.-C. 1989. Studies on Chinese Morphology and Syntax 2. Taipei: Student
Book Co.
. 1993. "Mingnanhua Foudingci de Yuji Neihan yu Jufa Biaoxian (A
Semantic and Syntactic Study of Negation in Taiwanese)". Publications of the
National Science Council, Part C: Humanities and Social Sciences. 3,2.224-243
Taipei, Taiwan.
Teng, Shou Xin. 1992. "Diversification and Unification of Negation in
Taiwanese". Chinese Languages and Linguistics 1. 609-629.
Wang, William. 1965. "Two Aspect Markers in Mandarin". Language 41,3.457-
470
Yue-Hashimoto, Anne. 1988. "A Preliminary Investigation into the Subclassifica
tion Problem of the Yue Dialects". Computational Analysis of Asian and
African Languages 30.7-38. Tokyo.
.1992. "Comparative Dialectal Grammar: Problems and Prospects". Paper
presented at the Fourth North American Conference on Chinese Linguistics,
University of Michigan, May 1992.
.1993. Comparative Chinese Dialectal Grammar: Handbook for
Investigators. Paris: Ecole des Hautes Études en Sciences Sociales, Centre de
recherches linguistiques sur ľAise Orientale.
Zhang, Min. 1990. A Typological Study of Yes-no Questions in Chinese Dialects:
in Diachronic Perspective. PhD dissertation, Peking University.
IMPERATIVE INVERSION
IN BELFAST ENGLISH
ALISON HENRY
University of Ulster at Jordanstown
Introduction
The variety of syntactic differences found between closely related dialects

presents considerable challenges for theories of language—such as the
Principles and Parameters model—which view the range of possible variation
between languages as highly constrained. Classical parameter theory (see for ex
ample Chomsky 1986, Roeper & Williams 1987) claims that differences be
tween language varieties reduce to the setting of a small number of parameters,
which have effects throughout the grammar. If the number of parameters is
comparatively small, the number of possible languages will also be constrained,
so that the theory may have problems in accounting for the observed variety of
dialects and indeed idiolects. Moreover, classical parameter theory, under which
the effects of parameter setting are seen in a range of areas of the grammar, pre
dicts that differences between dialects will be comparatively large-scale at the
surface level; a difference in the setting of one parameter will have a range of ef
fects throughout the grammar.
Recent work in linguistic theory within the Minimalist framework (Chomsky
1993, 1994, 1995) claims that the range of possible parameters is constrained;
the only way in which language varieties can differ from one another is in the
strength or weakness of functional categories, which in turn determines whether
movement occurs before or after Spell-out, that is, overtly in the syntax, or
covertly at LF. Moreover, optionality is to a large extent excluded in this frame-
work; under a principle of 'least effort', movement only occurs if forced.1 The
possibility of two dialects differing from one another in that one applies a pro-
Chomsky (1995) admits some optionality, but the concept of 'economy' still raises the
question of why movement takes place at all if it is optional.
80 ALISON HENRY
cess (for example verb raising) obligatorily while in the other it is optional, sits
uneasily with current approaches.
From the perspective of recent work in linguistic theory, then, we would ex
pect differences between language varieties to be of the following types:
a. differences should all be able to be accounted for in terms of

whether movement to a given functional category occurs before or
after Spell-out;
b. there should be no differences which apply only in a single con
struction or area of the grammar: all differences found should have
effects in a range of areas;
c. no optional/obligatory distinctions should exist: movement occurs
only if obligatory.
The question arises as to whether such an approach can in fact account for
thefine-graineddifferences we find between dialects, where there appear to be a
very large number of possible small-scale differences between language vari
eties. Two types of study outside the mainstream of work on linguistic theory
suggest that there may be problems. Work on sociolinguístics (see Labov 1994
for an overview) has argued, on the basis of studies of the output of a variety of
speakers, that language is characterised by a high degree of variability, subject to
both linguistic and extra-linguistic constraints. This view of language, where
speakers are seen to select between sociolinguístic variants made available by the
grammar, is essentially at variance with a model in which optionality is excluded
in principle.
There have also been findings in the study of historical syntax which have
called into question a straightforward parameter-setting model. Kroch (1994) ar
gues that historical change takes place via competing grammars which differ in
relation to one parameter setting, with the use of one grammar replacing the
other through a gradual shift. Again, this approach also seems to call into ques
tion a model in which there is no optionahty, and parameters are categorically set
in one direction or another.
In this paper, we will explore dialect differences in relation to overt-subject
imperatives between standard English and Belfast English, and within Belfast
English, in order to find out whether the kind of differences we find can in fact
be accommodated within current parameter theory.
It will be shown that the dialect variation found can to some extent be ac
commodated within the highly restrictive model of syntax envisaged by the
Minimalist program of syntactic theory (Chomsky 1994 ) : in particular, it can be
accounted for in terms of whether movement to a given functional projection is
IMPERATIVE INVERSION IN BELFAST ENGLISH 81
overt or covert. However, there are two respects in which the Belfast English
data presents problems for aspects of the theory. First, to account for the vari
ability found within the grammars of individual speakers, we need to admit a
degree of optionality into the syntax, allowing functional categories to be op
tionally strong. Secondly, to account for the fact that verb-raising can be specific
to certain constructions, we need to be able to specify the strength or weakness
of individual functional elements (e.g., the imperative morpheme) and not just a
functional category (e.g., C) in general. The theory thus on the one hand pre
dicts the type of variation which is found (overt versus covert movement) but
does not appear to allow sufficient flexibility in relation to the optionality of
movement and the level at which the strength or weakness of functional elements
may be specified.
1. Overt-subject imperatives in Belfast English: an overview
Like standard English imperatives, Belfast English imperatives can have null
subjects:
(1) Go away.
However, imperatives with overt subjects differ from standard English in that
they can have postverbal subjects. Thus, in addition to the standard English
version with a preverbal subject, as in:
(2) You take your time.
there is an alternative with verb-subject order:
(3) Take you your time.
In such sentences, the verb appears to have moved out of the verb phrase to a
position in front of the subject.
The order main verb-subject is restricted to imperatives; in other sentence
types the word order is as in standard English. Like standard English, Belfast
English does not allow the raising of main verbs, other than be and have, out of
the Verb Phrase.
(4) a. * Yesterday met Mary the students.
b. Yesterday Mary met the students.
82 ALISON HENRY
(5) a. * Go you to class every day?

b. Do you go to class every day?
(6) a. * John likes not that book.
b. John does not like that book.
Thus, it is not the case that Belfast English permits verb-subject order in general
more widely than does standard English.
There are two other major differences in word order between Belfast English
and standard English, in addition to that found in imperatives; these are inver
sion in embedded questions and subject position in for-to infinitives. The para
metric approach to syntax would lead one to expect that these differences might
derive from a single difference in parameter setting, but a study of the grammars
of individual speakers indicates that this cannot be the case; neither of these other
features of Belfast syntax necessarily co-occurs with imperative inversion in the
grammars of individual speakers.
Hiberno-English in general permits inversion in embedded questions
(McCloskey 1992, Henry 1995), so that sentences such as the following are
grammatical.
(7) They asked me was I going to the party.

(8) Every student wonders willsheipass the exam.
This structure is found throughout Ireland in the grammars of most speakers,

and is very robust; it shows no signs of disappearing from the language. On the
other hand, imperative inversion is restricted roughly to the North and East of
keland and is undergoing change: many younger speakers no longer use it.
Whatever permits embedded verb-second in questions, then, does not appear to
co-occur necessarily in the grammar with verb-subject order in imperatives.
Similarly, we find a small number of speakers in Belfast and other areas
who are able to use infinitives withafor-to sequence following the subject.
(9) I want her for to leave.

(10) The teacher likes them for to be in early.
Henry (1992) argues that in this structure for lowers to to, but one might be
tempted to seek an analysis of this which linked it to the imperative inversion
construction—though it is less obvious how the link might be construed than is
the case with inversion in embedded questions. However, such an effort would
again be misleading: the for-to construction does not necessarily co-occur in in-
dividual grammars with inversion in imperatives; speakers who permit impera

tive inversion do not necessarily have for-to infinitives and vice versa.
This illustrates the importance of ensuring that structures claimed to result
from a single parametric difference between dialects indeed necessarily co-occur
in the grammars of individual speakers, which is what parameter theory would
predict. Whether two structures occur within E-language in the same geographi
cal area is irrelevant; the key issue is whether the presence of one predicts the
presence of the other, or the absence of one the absence of the other, in individ
ual internalised grammars.
Returning to imperatives, Belfast English in fact does not have a single
grammar in relation to inverted imperatives; at least three sub-dialects can be
recognised—one which allows inversion with a wide range of verbs, which I
have termed the 'unrestricted inversion dialect'; one in which verb-subject order
in imperatives is restricted to unaccusative and passive verbs, which I have
termed the 'restricted inversion dialect'; and one which allows unrestricted in
version but raises the main verb as well as auxiliaries out of the verb phrase,
which I term the 'whole-verb inversion dialect'. A more detailed analysis of the
first two dialects can be found in Henry (1995).
2. The unrestricted inversion dialect
In this dialect verb-subject order can occur with all verbs. It seems that the
verb has moved out of the Verb Phrase to a position above the subject.
(11) Tell you the truth.

(12) Polish you that table.
(13) Take everybody their books.
The verb can move to a position above adverbs.
(14) Tell you always the truth.

(15) Polish you often that table.
(16) Take everybody always their books.
A particularly interesting feature of this dialect, again showing evidence that the
verb moves out of the verb phrase, is that it exhibits object shift. Where the verb
raises, weak object pronouns also raise to a position in front of the subject; as
we shall see below, the subject may remain in situ, with the verb and object
raising out of the VP to a position in front of the subject.
84 ALISON HENRY
(17) Throw it you to me.
Weak object pronouns must occur before adverbs.
(18) a. Throw it quickly you to me.

b. * Throw quickly it you to me.
c. * Throw you quickly it to me.
However, full DPs may not raise.
(19) a. * Throw the ball you to me.
b. * Throw the ball quickly you to me.
Neither may stressed or co-ordinated pronouns:
(20) * Throw THEM you to me.

(21) * Introduce him and her you to me.
This process applies to the same items—weak object pronouns—which un
dergo object shift in the mainland Scandinavian languages (Holmberg 1986). All
speakers of the unrestricted inversion dialect appear to allow object shift, pre
sumably to be characterised as movement of the object to SPECAGRO for check
ing.
Note that in these examples, the subject may appear either before or after the
object, apparently indicating that subject raising is optional; the subject may ei
ther raise out of VP to a position above that to which the object raises (probably
SPEC-AGRS), or it may remain in situ in VP. That the latter possibility exists is
indicated by the positioning of the subject in passives and in sentences with un-
accusative verbs. In imperatives with verbs of this type, and only in these, the
subject may appear after the main verb where there is an auxiliary present, rather
than, as is otherwise the case in imperatives, having to appear between the auxil
iary and the main verb.
(22) Be picked you for the team.
(23) Be going you out of the door by 9.
Object shift is obligatory for weak pronouns where the verb raises; all speakers
who have verb raising also have this object shift.
(24) a. Throw us quickly you your end there,

b. * Throw quickly you us your end there.
Here we see properties which do not appear to dissociate. Although no other

properties of the grammar seem to necessarily co-occur with verb raising in im
peratives, it is the case that where the verb raises, so also do weak pronouns. It
appears that object shift of weak pronouns, which occurs in mainland
Scandinavian languages, is in some sense latent in English. It does not occur be
cause main verbs do not raise, and it is dependent upon verbs raising through
AGRO; but where a dialect does allow verb raising, object shift automatically
follows.
A further indicator of verb position is generally taken to be its position in re
lation to negatives (see for example Pollock 1989). However the situation in re
lation to negation in overt-subject imperatives is rather complex. These impera
tives have, not the usual negative marker not, but a sentence-initial don 't, argued
by Beukema & Coopmans (1989) to be an invariant negative imperative marker
rather than a combination of do-support plus the usual negative marker not/n't.
Arguments for this include the fact that, unlike in questions, do and not cannot
occur separately in negative overt-subject imperatives, either in Belfast English
or in standard English.
(25) a. Don't you touch that.

b. * Do not you touch that.
 * Do you not touch that.
Compare the questions:
(26) a. Don't you like those books?

b. Do you not like those books?
And do-support is not available in imperatives even where required to fulfill the
NEG-criterion, nor is emphatic do available.
(27) * Under no circumstances do everybody leave.

(28) a. * Do you arrive on time.
b. * Do arrive you on time.
Moreover, don't appears with verbs which in general do not allow do support,
that is, be and auxiliaries.
(29) a. Don't you be late.

b. Don't be you late.
c. * Don't they be late?
86 ALISON HENRY
Thus, it seems thatdo-supportis unavailable in imperatives, possibly be

cause do is inserted under Tense and Tense is not instantiated in imperatives.
Thus, it seems that don t is an invariant negative marker, occurring sentence ini
tially, rather than a negative occurring in NEGP. On the basis of data from other
languages, Rivero & Terzi (1994) propose that imperatives are introduced by an
illocutionary force marker. Don't would appear to be the negative version of this
in English.
There is some evidence for the existence of a sentence-initial imperative
marker from some dialects of Hiberno-English other than Belfast English.
McCloskey (p.c.) points out that in some dialects of Hiberno-English there is an
overt sentence-initial imperative marker gon.
(30) Gon you make your mummy a cup of tea.
This marker cannot co-occur with don't, the negative imperative marker.
(31) a. * Don't gon you make your mummy a cup of tea.
b. * Gon don't you make your mummy a cup of tea.
It cannot co-occur with don't because it is occupying the same position.

The position to which imperative verbs raise in Belfast English appears to be
the same position occupied by don't (and by gon in other dialects); don't and
verb raising cannot co-occur.
(32) a. * Don't lose you that book.

b Don't you lose that book.
If all movement takes place for checking, then the imperative verb must
move to check an imperative marker, which may be equated with the illocution
ary force marker of Rivero & Terzi (1994). A more detailed analysis of the con
struction is to be found in Henry (1995). For present purposes, the important
point to note is that imperatives in the unrestricted inversion dialect involve rais
ing of the verb out of the Verb Phrase, and this characteristic of the grammar
does not necessarily co-occur with any other feature relating to verb position
which differentiates the dialect from other varieties of English.
The verb in imperatives must, under Minimalist assumptions, raise for
checking, presumably of an imperative morpheme. Let us assume that this mor
pheme is in C, as argued in Henry (1995). Since raising takes place before
Spell-out in this dialect, the V-feature of the node to which the imperative raises
must be strong. On the contrary, verbs do not raise overtly in the syntax other
than in imperatives, so that this node must otherwise be specified as having a

weak V-feature. Thus, the grammar must be able to include a specification of the
strength of a particular morpheme, rather than only of a functional category as a
whole. The V-feature of  is usually weak in English, including Belfast English,
as shown by the fact that verb-subject order does not occur unless induced by
the WH-criterion or the NEG criterion (Rizzi 1992).
(33) * Yesterday went he to school.

(34) Read the students the books.
The specification of the imperative morpheme as strong is also optional. Verb

raising is not obligatory in this dialect.
(35) a. Read you that book.

b. You read that book.
(36) a. Bring everybody their raincoats.
b. Everybody bring their raincoats.
Verb raising is, then, confined to imperatives. It is not a reflection of a wider

availability of verb raising in general in Belfast English than in standard English.
Whatever differentiates the grammars of speakers of Belfast English from those
of standard English speakers must thus contain a statement about construction-
specific differences, however this is expressed. Thus, the prediction of a strong
version of parameter theory, which would see all differences between language
varieties as deriving from 'large-scale' parameter-setting differences which have
consequences throughout the grammar, does not seem to be able to deal with
these facts. Moreover, we can find speakers whose grammars differ only in the
single fact that one has inverted imperatives and the other does not. For many
younger speakers, especially those under 20, do not use inverted imperatives,
but have otherwise all the Belfast English constructions used by somewhat older
speakers. To account for this difference in grammars, we need to allow in some
way for construction-specific differences. This may be stated in terms of the
strength of a particular morpheme which occurs in imperatives, but this does not
alter the fact that it is construction-specific or morpheme-specific, rather than a
statement about the strength or weakness of a particular functional category in
the language in general.
It might be thought that speakers are bi-dialectal, speaking sometimes stan
dard English and sometimes Belfast English, and thus have two separate gram
mars, rather than a single grammar containing a rule which is optional.
However, there is no evidence that these speakers have two distinct grammars;
88 ALISON HENRY
although there is evidence of a shift towards standard English in more formal

circumstances, there is no evidence that speakers have two separate dialects
which they use on different occasions. Rather, local features, such as inverted
imperatives, co-occur with standard features such as uninverted imperatives,
even in the same dialogue, and much more frequently and unpredictably than in
the type of code-switching associated with bilingualism.
Thus, speakers' grammars must contain a specification for the imperative
morpheme, which is not only exclusive to that morpheme, but admits optional-
ity. These are two characteristics of parameters which have not generally been
permitted in current approaches to syntax, which envisage statements as re
stricted to functional categories as a whole, and specifically rule out optionality
of movement; indeed Chomsky (1994 ) argues that movement only occurs when
forced to do so, clearly excluding the possibility of optional movement.
There are thus two ways in which this dialect difference shows that models
of parameter setting based on different languages can be misleading. On the
other hand, note that the type of difference between the dialects is otherwise of
the same type as that found between different languages. It can be defined in
terms of whether the verb raises before or after Spell-out, a characteristic which
differentiates French from English in relation to whether or not the verb raises as
far as AGR (Pollock 1989) and both these languages from Verb Second lan
guages such as German where the verb raises to  in main clauses.
We also noted above that subject-raising in imperatives is optional, although
it is obligatory in other constructions. The grammar will also have to account for
this. There is in fact another dialect in which verb raising does not occur, but as
in the unrestricted inversion dialect subject-raising is optional, so that the picture
in relation to the nature of the subject-m-situ option is clearer. We now turn to
consider that dialect.
3. The restricted inversion dialect
In one of the sub-dialects, which appears to be used predominantly by

younger and middle-aged speakers, postverbal subjects occur in imperatives
only where the verb is unaccusative or passive. These speakers have lost the
possibility of verb movement, used by unrestricted inversion dialect speakers
who are in general middle-aged or older, and thus for them sentences like (35a)
and (36a) above are ungrammatical; they do however permit verb-subject order
in some imperatives:
(37) Go you away.

(38) Run everybody home.

(39) Be picked you for the team.
In this dialect, it seems that the subject can only occur postverbally where it
originates in a post-verbal position; the subject does not appear to be forced to
raise out of VP. Where there is an auxiliary, the subject occurs, not after the
auxiliary, but after the lexical verb.
(40) a. Be elected you president.

b. * Be you elected president.
(41) a. Be running you out of the door when they arrive.
b. * Be you running out of the door when they arrive.
Henry (1995) attributes the availability of this construction to the fact that subject
raising to SPEC-AGRS is not obligatory in Belfast English, a fact that is reflected
in the optionality of subject-verb concord.
(42) The eggs is cracked.
However, further study of speakers outside the inner Belfast area has shown
a dissociation between the availability of 'singular concord' and that of non-
raising of subjects in imperatives; there are speakers who allow non-raising of
subjects in imperatives, but do not have singular concord. Thus, the optionality
of subject raising seems again to be exclusive to imperatives, a property of the
strength of the NP-feature of some functional head in the imperative construc
tion. It is only in this construction that postverbal subjects are possible; they do
not occur generally in Belfast English.
(43) * Went he home.

(44) * Read John the book.
This variety of Belfast English, then, differs from standard English in that,
in imperatives, the subject raises at LF rather than before Spell-out, exactly the
type of difference we would expect to see under the Minimalist program.
However note that, like verb raising in the unrestricted inversion dialect, this is a
property of a single construction—perhaps of the imperative morpheme—and it
is also an optional property. Alongside imperatives which do not have subject
raising, the same speakers use imperatives in which the subject does raise.
(45) You leave on time.

(46) Everybodyi go to theiri places.
90 ALISON HENRY
We can tell that the preverbal DPs in these sentences are true subjects rather than
vocatives (which of course also occur with imperatives, as in (47), with an into
nation break), not only because an intonation break is not necessary, but also
because pronouns can refer back to third-person subjects. Beukema &
Coopmans (1989) point out that this is only possible with true subjects; where
the DP is a vocative, non-third-person pronouns cannot be co-referential with it.
(47) Everybodyi go to thehj *i places.
Note that the specification of one particular morpheme as having a particular

property (e.g., a weak NP-feature) as distinct from specifying a whole functional
category in this way provides a more complex, hence more marked, grammar
than one in which the property which determines when subject-raising occurs is
either always strong or always weak. We would expect such a grammar to be
vulnerable to change in historical terms, and this is what wefind;many children
of parents who are users of the restricted inversion dialect are not acquiring it,
but rather adopting a grammar without postverbal subjects in imperatives. This
is not a simple shift towards standard English; these children acquire many other
properties of the local variety, including for example inversion in embedded
questions. We know that parents address many imperatives to their children, but
presumably they have not used sufficient inverted imperatives for their chil
dren's grammars to be forced to adopt a marked option including a morpheme-
specific specification. Similarly, children of parents who are unrestricted inver
sion dialect users do not necessarily acquire a similar grammar: they adopt a re
stricted inversion dialect grammar, or one where inversion is not permitted.
Thus, we appear to have possible but comparatively marked options, adopted by
children in the acquisition process only if there is overwhelming evidence in
their favour.
4. The whole-verb inversion dialect
Finally, let us note that there are some speakers, mainly outside the inner-
city areas of Belfast, who show a pattern of inverted imperatives where the sub
ject occurs after the main verb in sentences with auxiliaries, even where the verb
is not unaccusative or passive. In Henry (1995) I argued that speakers in Belfast
did not in fact use such structures, but it turns out that they are used in some
other areas of Ireland which have inverted imperatives.
(48) Be doing you your homework when I get back.

(49) Have tried you that food before you leave.
For speakers of this dialect, this order is obligatory when the verb raises. In
all other respects, this variety is like the unrestricted inversion dialect; it permits
NEG criterion sentences and object shift, and disallows negation with inversion.
(50) On no account be jumping you on the furniture when your

mother gets back.
(51) Be telling it you to them when the others arrive
(52) * Don't be playing you a video game when I get back.
This dialect appears to have, like Spanish, movement of the main verb along
with the auxiliary to C.
(53) a. Esta hablando el profesor con el estudiante?

b. * Esta el profesor hablando con el estudiante?
Here again we see a dialect exploiting a possibility offered by UG. However, the
selection of this possibility does not appear to depend on other selections in the
grammar; the dialect of speakers who select this option does not differ from that
of the unrestricted inversion dialect in any other way.
In relation to imperatives, there appear in fact to be four different dialects in
the Belfast area: one which like standard English does not allow inversion; one
which allows verb movement to  in imperatives, and in which subject raising is
optional; a variant of this in which, where an auxiliary raises, the main verb
raises with it; and one which has subject raising as optional but does not allow
verb raising. Thus, even within a small area, there are several competing di
alects.
5. Conclusion
We have seen that in relation to imperatives, the kind of differences we find

between dialects are in one sense precisely what would be predicted by
Universal Grammar; they reduce to whether movement of verbs and subjects oc
curs before or after Spell-out.
Note that there has in effect been a shift in linguistic theory recently, though
one that has been largely unacknowledged, from the concept of parameters as
factors which affect a wide range of aspects in a language to factors which have
92 ALISON HENRY
features of various heads affects only whether an N or V will raise to a specific

head, but not other aspects of the language, and N- and V-raising are seen as
largely unrelated to each other. Word order is seen as determined, not by a gen
eral head direction parameter, but by the raising of elements for checking, with
various types of raising unrelated to each other. Thus, we have moved from a
conception of parameters as very wide-ranging to one where, while highly re
stricted in type, they cover a much smaller range of phenomena. What we have
found out about dialect variation suggests that this is the right approach. It is not
possible to account for the differences between Belfast English and Standard
English by means of a difference in the setting of one or two parameters which
would have a range of effects throughout the grammar. Since the differences
discussed in this paper between Belfast English and standard English, and
between the various sub-dialects of Belfast English, in relation to imperatives do
not necessarily co-occur with other differences between these dialects, we must
admit the possibility of variation which is essentially construction-specific, thus
allowing the specification of a morpheme (e.g., the imperative morpheme) as
strong even though the functional category it occupies, probably C, is in general
weak in these varieties. Secondly, we must allow optionality: either movement
itself must be optional in some cases (as with verb raising in imperatives in the
unrestricted inversion dialect), or elements must be able to be categorised as
optionally strong ('the imperative morpheme may be weak or strong'). However
we formalise these, we have to admit that something like construction-specific
differences between dialects, and optionality of movement, must be able to be
part of the variation between grammars permitted by UG, something which we
might have missed if study was confined to the larger-scale differences typically
found between those varieties generally characterised as different languages.
REFERENCES
Beukema, Fritz & Peter Coopmans. 1989. "A GB Perspective on the Imperative in
English". Journal of Linguistics 25.417-436.
Chomsky, Noam. 1986. Knowledge of Language. New York: Praeger.
. 1993. "A Minimalist Program for Linguistic Theory". The View from
Kenneth Hale & Samuel Jay Keyser, 1-52. Cambridge, Mass.: MIT Press.
. 1994 "Bare Phrase Structure". MIT Occasional Papers in Linguistics 5.
. 1995 The Minimalist Program. Cambridge, Mass.: MIT Press.
Henry, Alison. 1992. "Infinitives in a for-to dialect". Natural Language and

Linguistic Theory 10.279-301.
. 1995. Belfast English and Standard English: Dialect Variation and
Parameter Setting. Oxford: Oxford University Press.
Holmberg, Anders. 1986. Word Order and Syntactic Features in the Scandinavian
Languages and English. PhD dissertation, Stockholm University.
Kroch, Anthony. 1994. "Morphosyntactic Variation". Papers from the 30th
Regional Meeting of the Chicago Linguistic Society: Parasession on Variation
and Linguistic Theory, ed. by Katherine Beals, Jeanette Denton, Robert Knippen,
Lynette Melnar, Hisami Suzuki & Erica Zeifeld, 180-201. Chicago Linguistic
Society, University of Chicago.
Labov, William. 1994. Principles of Linguistic Change. Oxford: Blackwell.
McCloskey, James. 1992. "Adjunction, Selection and Embedded Verb Second".
ms, University of California at Santa Cruz.
Pollock, Jean-Yves. 1989. "Verb Movement, UG and the Structure of IP"
Rivero, Maria-Luisa & Antonio Terzi. 1994. "Imperatives and Illocutionary
Force". Paper presented at the GLOW Colloquium, Vienna.
Rizzi, L. 1992. "On WH-movement". University of Geneva Formal and Technical
Reports on Linguistics.
Roeper, Thomas & Edwin Williams. 1987. Parameter Setting. Dordrecht: Reidel.
SCANDINAVIAN POSSESSIVE CONSTRUCTIONS
FROM A NORTHERN SWEDISH VIEWPOINT
ANDERS HOLMBERG & GÖREL SANDSTRÖM
University of Tromsø & Umeå University
Introduction
We discuss two parameters which account for much of the variation in the
form of the noun phrase among the Scandinavian languages and dialects, in par
ticular possessive constructions, and more particularly, constructions with a
postnominal possessor. We will claim that one of the two parameters is a so-
called major parameter, that is a parameter which (a) concerns the feature values
of a functional category with a general distribution, (b) which is therefore resis
tant to change, and (c) therefore resistant to dialectal variation. The other pa
rameter is a so-called minor parameter, that is to say a parameter which (a) con
cerns the feature values of a functional category with a restricted distribution, (b)
which is therefore vulnerable to change, and (c) therefore subject to dialectal
variation.
1. Major and minor parameters
In Principles-and-Parameters (P&P) theory all grammars are special cases of

Universal Grammar (UG). There is grammatical variation among languages be
cause UG contains some parameters the values of which are left open by UG;
these are fixed in the process of language acquisition, on the basis of data in the
linguistic environment. It is understood, although perhaps not often explicitly
stated, that some parameters have a greater impact on the surface form of lan
guages than others. For some parameters, let us call them major parameters, a
change of value will have wide repercussions in the grammar, affecting many
96 ANDERS HOLMBERG & GÖREL SANDSTRÖM
constructions. Other parameters are minor parameters, in the sense that they af
fect few constructions, in the most extreme case only one.
The value of a minor parameter is much more likely to change from one gen
eration to the next than the value of a major parameter. All it takes to change the
value of a minor parameter is that the construction where the parameter has its
effects, and which hence provides the crucial trigger experience for fixing its
value, is pushed out of use by another construction, spontaneously created or
taken over from, say, a neighbouring dialect or language. In other words, a mi
nor change in the linguistic environment may be enough to change the value of a
minor parameter from one generation to the next. To change the value of a major
parameter, on the other hand, it is not enough that just one construction is
changed or replaced: since the value for a major parameter is selected on the ba
sis of information from several constructions, there have to be several more or
less simultaneous changes in the linguistic environment, affecting several of the
relevant constructions, to create the conditions for a change in the setting of the
parameter.
Since the value of a minor parameter is easily changed from one generation
to the next, it is likely to be subject to dialectal variation. A major parameter, on
the other hand, is less likely to exhibit dialectal variation. That is to say, what we
call dialectal variation typically concerns minor parameters.
One of the leading ideas within P&P theory is that all syntactic variation is
due to variation in the feature values of certain functional categories. In this per
spective a major parameter is one which concerns features of a functional cate
gory which is involved in many constructions, while a minor parameter is one
which concerns features of a functional category which may occur in only one
construction, in the most extreme case. We may summarized the properties of
major and minor parameters as follows:
(1) a MAJOR PARAMETER:

• concerns feature values of a functional category
with a general distibution,
• is (therefore) resistant to change, and
• (therefore) resistant to dialectal variation,
a MINOR PARAMETER:
• concerns feature values of a functional category
with a restricted distribution,
• is (therefore) vulnerable to change, and
• (therefore) likely to show dialectal variation.
SCANDINAVIAN POSSESSIVE CONSTRUCTIONS 97
These differences suggest a certain heuristic for associating syntactic forms

with major vs. minor parameters. Imagine a population  within which the social
conditions for more or less normal dialectal variation are met1, as shown by the
fact that various forms of dialectal variation in fact occurs. Imagine a syntactic
form F used within P, where F is subject to cross-linguistic variation (i.e., it is
not universal) but is not subject to any dialectal variation within P. Then F is
most probably dependent on a particular value of a major parameter. This
means, given (1), that there must be other forms in the language spoken by 
which are dependent on the same parameter value. If not, it is hard to explain the
lack of dialectal variation for F. That is to say, we should find other forms which
could in principle vary but which do not vary within P, and which can plausibly
be related to the same parameter value as F.
Conversely, if another syntactic form G is found to be subject to dialectal
variation within P, then G is probably dependent on the value of a minor parame
ter, and we do not necessarily expect to find other forms subject to variation cor
relating with the variation exhibited by G.
With this in mind, we move to examine the range of variation in possessive
constructions in the Scandinavian languages, especially constructions with a
postnominal possessor.
2. Postnominal possessors in Scandinavian
Within the Scandinavian languages, there are basically two forms of post-
nominal possessor constructions: one with a bare head noun, the other with a
head noun bearing the definite suffix ('N.D' = noun with definite suffix).
(2) a. bill Jons (Icelandic) b. bilen hans (Norwegian)

car Jon's car.D his
The construction exemplified in (2a), which we will call the N-POSS con
struction, is found in Icelandic and Old Scandinavian, but in no variety of
Mainland Scandinavian. The N.D-POSS construction of (2b) is common to
Icelandic, Norwegian, Northern Swedish, and, with a PP possessor, Farnese.
We will address mainly two issues. First, we will explain why the N-POSS con
struction (2a) is restricted to Icelandic and Old Scandinavian. We claim that this
is an effect of a major parameter in Scandinavian syntax, namely strong versus
1
Dialect variation may be restricted by various social factors, such as a strong normative tradi
tion, enforced in primary education and by various forms of 'language control'.
weak Case: Icelandic and Old Scandinavian have strong Case, while the other
Scandinavian languages all have weak Case. Second, we will discuss the N.D-
POSS construction (2b). This construction has been discussed in the literature on
the basis of facts from Icelandic, Norwegian, and some varieties of Northern
Swedish. In all of these languages/dialects the possessor can only be a pronoun,
or possibly a proper name in construction with a pronoun (examples will be
provided below). When we look at a wider range of Northern Swedish dialects,
however, it emerges that the variation in the categorial and morphological fea
tures of the possessor in this construction is much richer than is generally as
sumed. We will propose an analysis which is consistent with this variation. The
claim is that the definite form of the noun is moved to D, the highest functional
head in the nominal projection, and the possessor is moved to the specifier posi
tion of an AGR head situated between D and NP.2 The variation is due to a minor
parameter in the feature-content of the DP-internal AGR, where, however, the
features are subject to a universal hierarchy of argument features which is well
known from the typological literature.
3. Sentence structure and NP structure
Variation in sentential structure among the Scandinavian languages has been

studied in great detail; see Platzack (1987), Holmberg & Platzack (1991),
Holmberg & Platzack (1995), Holmberg (1994), Rohrbacher (1994), Vikner
(1994). The picture which emerges from these studies is clear enough: the main
division separates Icelandic and Old Scandinavian on the one hand from the
Mainland Scandinavian languages (Swedish, Danish, and Norwegian) on the
other.3 Between the two main groups there is plenty of variation; within the
groups there is by comparison little variation, and there is not much variation
cutting across the main division.4
When we look at the internal structure of noun phrases we can observe re
flections of the same division, with Icelandic and Old Scandinavian on one side
and Mainland Scandinavian on the other, but the picture is less clear here in that
2
The category we call D in this paper corresponds to the category called  in Sandström and
Holmberg (1994).
3
The position of Faroese in this dichotomy is more controversial, but according to Holmberg
(1994) it is truly in-between, in a well-defined sense, as described below fn. 6.
4
To put it more precisely, there is a good deal of variation which cuts across the 'main divi
sion', but it partitions the languages in all sorts of ways. In other words, the inter-group varia
tion is due to a number of different parameters, mostly with a quite limited syntactic effect—
what we call minor parameters.
(a) there is also some striking variation which cuts across this division, and (b)
there is a remarkable amount of variation within one of the main groups, namely
Mainland Scandinavian. In fact, it appears that there is at least as much word or
der and morphological variation in possessive constructions within Northern
Swedish alone as there is among all the Scandinavian standard languages. In (3)
are exemplified three possessive constructions found in the Scandinavian lan
guages. The table in (4) illustrates how they are distributed among the
Scandinavian languages.5
(3) a. bíll Jóns (N-POSS)

car Jon's
b. bilen hans (N.D-POSS)
car.D his
c. Jons bil (POSS-N)
Jon's car
(4) ICE NOR NSW STSW DAN

N-POSS +
N.D-POSS + + + - -
POSS-N - + + + +
Icelandic has two constructions with a postnominal possessor, N-POSS, i.e.,

a bare noun followed by a possessor, and N.D-POSS, i.e., a noun with a suffixed
definite article followed by a possessor. Norwegian and certain Northern
Swedish dialects have only N.D-POSS. In addition they have prenominai posses
sors. Standard Swedish and Danish have only prenominai possessors, as do
most varieties of Central and Southern Swedish. So, with regard to N-POSS and
POSS-N we do see the familiar pattern, with Icelandic on one side (in fact, again
together with Old Scandinavian) and the Mainland Scandinavian languages on
the other side. But the construction N.D-POSS straddles the boundary between
Icelandic and Mainland Scandinavian, being common to Icelandic, Norwegian,
and Northern Swedish, but excluded from most other dialects of Swedish, as
well as from Danish.
5
Mainly for ease of presentation we have left out Faroese from the table. Faroese has + in all
three rows. However, there are special restrictions on all the forms: N-POSS occurs only with
family relations, N.DEF-POSS only with a PP possessor, and POSS-N only with proper names.
We will not deal with Faroese in this paper, but see Holmberg (1994). The table represents an
idealization also in that not all Norwegian or Northern Swedish dialects have POSS-N as well as
N.DEF-POSS, but some do, including both standard varieties of Norwegian.
In a series of works Christer Platzack and Anders Holmberg have argued

that, as regards sentential syntax, there are two major parameters which distin
guish between the two main groups of Scandinavian languages: Icelandic and
Old Scandinavian have 'strong AGRS' and 'strong Case', while Mainland
Scandinavian (henceforth MSc) has 'weak AGRS' and 'weak Case'.6 This differ
ence is reflected in the richness of the morphological paradigms (although, as
discussed in Holmberg (1994), the relation between feature strength and overt
morphology is not a simple one): while Icelandic and Old Scandinavian have
fairly rich case and subject-verb agreement morphology, MSc has very poor
case morphology and no subject-verb agreement morphology. These parameters
are shown to be reflected in at least 15 syntactic differences, concerning imper
sonal constructions, verb placement, various scrambling phenomena, and so on.
Of the two morphological parameters claimed to be responsible for the varia
tion in sentential syntax, we do not, perhaps, expect to see any DP-internal ef
fects of strength of AGRS, but it seems not unlikely that the strength of Case
might have effects not just on the distribution of DPs in sentences, but also on
the internal syntax of DPs. This, we claim, is the case: the reason why no dialect
of MSc has N-POSS is that the construction requires strong Case, and among the
modern Scandinavian languages only Icelandic has strong Case. This notion will
be discussed in the next section.
4. Why MSc does not have N-POSS
Why is (2a), repeated here as (5a), not well formed in any MSc dialect, in
cluding those dialects which allow postnominal possessors in some construc
tions?
(5) a. bíll Jons (Icelandic) b.* bil Jons (Norwegian, NSw.)

car Jon's
In (6) there is a selection of postnominal possessive constructions found

among Norwegian and NSw dialects:
6
The claim made in these works, apart from Holmberg (1994), is that AGRS is actually absent
in MSc, reflected in the absence of any subject-verb agreement morphology. As mentioned
above in fn. 3, Faroese can be described as being truly in-between the two main groups, having
strong AGRS (like Icelandic) but weak Case (like MSc); see Holmberg & Platzack (1995),
Holmberg (1994).
(6) a. bilen hans

car.D his "his car"
b. bilen hans n Jon
car.D his ART Jon "Jon's car"
. bilen n Jons
car.D ART Jon's "Jon's car"
d. bilen läraren
car.D teacher.D "the teacher's car"
. bilen åt läraren
car.D to the.teacher "the teacher's car"
There is variation among these dialects in the categorial and morphological

features of the possessor: some dialects only allow pronominal possessors,
other dialects allow proper names as well, others still lexical (definite) DPs as
well, and still others allow pronouns and PPs but not lexical DPs; the possessor
may be marked with a possessive pronoun, an s-suffix, or not at all. However,
common to them all is that the head noun must be definite.
The Icelandic noun phrase differs from the MSc noun phrase in certain other
respects, too:
(7) · Icelandic has case morphology on nouns,

• Icelandic has no indefinite article,
• Icelandic has no free definite article.
These properties are exemplified in (8) and (9) (W = the so-called weak adjec
tival inflection).
(8) a. Ég sá bíl. (Icelandic) b. Jag såg * (en) bil. (Swedish)

I saw car+Acc I saw a car
"I saw a car"
(9) a. Ég sá nyja bílinn b. Jag såg *(den) nya bilen

I saw new.w car+ACC.D I saw the new.w car.D
"I saw the new car"
As far as we know, all MSc dialects have an obligatory indefinite article.

Likewise, although not all MSc dialects have a free definite article, it is neverthe-
less true (as far as we know) that no MSc dialect has a construction such as (9a)
as the unmarked alternative, the way Icelandic does.7
We assume the following principle, given here in two versions. Following
Delsing (1993) we call it 'the Argument Rule'.
(10) THE ARGUMENT RULE (two versions):

a. All arguments must have a filled determiner position at S-struc-
ture. (Delsing 1993: 65).
b. An argument must have a strong feature in D. (Holmberg 1994)
The role of this principle in Scandinavian syntax, conceived as in (10a), is

discussed in great detail in Delsing (1993). This is the principle which explains,
among other things, why argument noun phrases in for instance English are
normally headed by an article or some other visible determiner. The principle is
reformulated in Holmberg (1994) in the framework of a minimalist theory, as
suming no level of S-structure. In minimalist terms the Argument Rule must be
satisfied at LF, yet it is reflected in PF as a phonetically filled D. Following
Chomsky (1993) we assume that the strong feature has the effect of attracting
phonetic material to D. This is ensured by the following principles:
(11) a. A strong feature must be checked off (eliminated through

checking) before spell-out to PF.
b. A strong feature of a functional category F is checked off if
a phonetically licit category with a matching feature is ad
joined to F, or placed in SPEC-F.
A functional feature, such as Case or definiteness, can be morphologically

represented on a lexical head, for instance N, as well as being represented on a
7
Some MSc varieties have it as a marked, semantically restricted alternative; see Delsing
(1993:118). More interestingly, many Northern Swedish dialects lack a free definite article, but
nevertheless do not allow the construction (9a). Instead the adjective is incorporated in the defi
nite noun.
(i) Jag såg nybilen.
Í saw new.car.D
"I saw the new car."
According to Sandström and Holmberg (1994) the definite noun is moved to D (called  in that
work) across the attributive adjective, followed by adjective incorporation into the definite
noun, It is not inconceivable that the lack of a free definite article is one of the factors behind
the preference for the N.DEF-POSS construction over POSS-N in Northern Swedish. It is not,
however, the case that all dialects which have N.DEF-POSS also have Adjective Incorporation.
functional head determining the scope, and more generally, the syntactic role of
the feature, namely D.8 Both Case and definiteness are syntactically properties of
DP, although they may be morphologically realized on N, or some other head
other than D within DP. Checking is here regarded as a way to ensure that the
morphological feature and the syntactic feature have matching values. If the
syntactic feature is strong, in Chomsky's (1993) sense, it must be checked be
fore spell-out, either by insertion of an appropriately specified category in one of
the checking positions (cf. 1 lb), or by movement of an appropriately specified
lexical category. If the feature is weak, checking can wait until LF. We assume
that the need for checking is mutual between the morphological and the syntactic
feature. Let us say that the morphological feature cannot be interpreted at LF un
less it is checked. Following Chomsky (1993) we assume the syntactic feature
cannot be phonologically interpreted, and therefore must be checked off before
spell-out if it is strong (= 'visible'), assuming checking to result in deletion of
the syntactic feature. A weak (= 'invisible') feature is innocuous in PF, and
therefore need not be checked off before LF. Also following Chomsky (1993),
we assume the Greed principle:
(12) GREED
Movement of a category α is triggered only if the move
ment helps to satisfy the needs of α itself.
The strong feature required by the Argument Rule (10b) may be Case,
clearly a property of DP, hence its head D. Alternatively it is definiteness, also
clearly a property of D/DP. It follows from the assumptions made above that
strong Case in D requires that the language have either some kind of free Case
morpheme(s) which can be inserted in D, checking off the strong Case feature,
or bound Case morphology on N or some other nominal category within DP,
triggering movement of this category to D or SPEC-D. Correspondingly, strong
definiteness in D requires that the language have either free definiteness mor
phemes (articles) which can be inserted in D, checking off the strong definite
ness feature, or bound definiteness morphology on N or some other nominal cat-
8
This presupposes a lexicalist theory of inflectional morphology, according to which lexical
heads are inserted with (at least some) inflections already attached, as proposed by Chomsky
(1993).
egory within DP, triggering movement of this category to D or SPECD. A lexical

head without any functional morphology does not move.9 ,10
In (13), we give a rough analysis of an argument noun phrase in Icelandic, a
language with strong Case (represented as capitals, while the morphological
feature is represented as lower case letters).
(13)
Here the Case morphology on the noun triggers movement of N to D, checking

off the strong Case feature.
Structure (14) is the rough structure of an argument noun phrase in
Swedish, a language with weak Case but, by assumption, strong definiteness.
(14)
Here the definiteness morphology triggers N-movement to D, checking off the

strong feature.11
9
See Pollock (1994) for a similar view of checking.
10
The converse does not hold: a language may have, for example, Case-morphology on N, and
yet have weak Case in D, checked only in LF. According to Holmberg (1994) Faroese is such a
language.
11
N-to-D movement may be blocked by an adjective intervening between D and N: see Delsing
(1993: 116-134), Holmberg (1993), Kester (1993), Sandström and Holmberg (1994),
Santelmann (1993). In that case many varieties of Swedish, including Standard Swedish, have a
free article inserted in D, effecting the check-off of the strong definiteness feature. (Not all of
these varieties allow a postnominal possessor.)
We conclude that the reason why Icelandic unlike MSc can do without an in
definite article in (8), or a free definite article in (9) is that Icelandic has strong
Case. The strong Case feature satisfies the Argument Rule at LF, and the case
morphology on the noun triggers movement of the noun to D, checking off the
strong Case feature before spell-out. MSc, on the other hand, has to rely on
strong definiteness coupled with free or bound definiteness morphology. For the
same reason (15a) below is ungrammatical in MSc, while the corresponding
form is grammatical in Icelandic.12 (15b), on the other hand, is grammatical in
MSc as well as in Icelandic, subject to dialectal variation.
(15) a. * bil hans/Jons/ åt läraren/ etc.

b. bilen hans / Jons / åt läraren / etc.
Since MSc does not have strong Case, it has to rely on definiteness mor
phology triggering noun movement to D, eliminating the strong feature.
Therefore the noun in the postnominal possessor construction has to be definite,
as in (15b).
Insertion of a free definite article in (15a) would satisfy the Argument Rule,
but the result is nevertheless ungrammatical.
(16) * denbil(hans)
the car his
Note that (16) is ill-formed with or without the possessor: the definite form
of an unmodified DP can only be [DP bilen ]. We conjecture that this is an effect
(i) a. *nya bilen (Jons) b. den nya bilen (Jons)

new caro Jon ' s the new caro Jon's
The fact that adjectives have this blocking effect on noun movement strongly suggests that at
tributive adjectives are heads taking NP as complement; see the references mentioned above.
Many Northern Swedish dialects instead circumvent this blocking effect by incorporating the
adjective into the noun; see footnote 7. An attributive adjective apparently also blocks N-
movement in Icelandic. If it did not, we would expect the word order (iia) to be well formed,
(iib) shows that there is no need to insert a free head encoding Case in D or to incorporate A in
N in Icelandic.
(ii) a. *bíl nyja Jόns b. nyja bíl Jόns
car new Jón ' s new car Jón' s
Holmberg (1994) suggests that in this case the adjective moves to D, checking off the strong
Case feature, which is made possible by the fact that Icelandic has Case morphology on the ad
jective. See Sigurðsson (1993) for a different solution.
12
There are certain restrictions regarding the choice of head noun and the choice of possessor
which we will not discuss here: see Sigurösson (1993).
of economy of representation (Chomsky 1991 and subsequent work): where

there is a choice, a less specified category is preferred over a more specified one,
blocking the latter. In the case at hand, and possibly in general, this means that a
bound morpheme (the bound definite article) is preferred over a free morpheme
(the free definite article).13 The free definite article can be used only when the
bound form is not capable of satisfying the Argument Rule, movement to D be
ing blocked by  intervening adjective. Note, however, that (17) is also ill-
formed.14
(17) * en bil hans/Jons/ åt läraren/ etc.

a car his / Jon's / of the teacher / etc.
Here the corresponding possessorless form [DP en bil ] is fine. The Argument
Rule and strong feature checking is (potentially) satisfied in (17), so the problem
seems to be the possessor. We return to (16) and (17) below, after having es
tablished the more precise structure of DP in Swedish. 15
5. Postnominal possessors in Northern Swedish
Characteristic of the construction (2b), repeated here, is that there are con
straints on the categorial features of the possessor, subject to dialectal variation.
(2b) bilen min

car.D my
13
It is not obvious in which sense the free definite article is more specified than the bound
definite article, apart from containing one more phonetic segment. Tarald Taraldsen (p.c.) sug
gests that the free definite article may actually be bimorphemic, consisting of two parts:
[d [en]]. If so, it would indeed be structurally more complex and contain more feature specifica
tions than the bound article.
14
The construction with a postnominal PP (en bil åt läraren) is well formed as long as the PP
has a goal or benefactive reading: 'a car for the teacher'. In at least some dialects, including
standard varieties of Norwegian, it is ill formed if the PP is a possessor; see Taraldsen (1990).
See also footnotes 22 and 23, below.
15
An anonymous reviewer points out that formulation of the Argument Rule in terms of a
strong feature implies that there are languages which have a weak feature in D, given the usual
employment of the strong/weak dichotomy as a formalization of parametric variation. Insofar
as there are languages which make do without either articles or case-morphology, or any other
form of overt marking of argumenthood (Chinese is a possible candidate) this could be for
mally expressed as a weak feature in D. An interesting possibility is that predicative noun
phrases have a weak feature in D, perhaps universally.
In Standard Icelandic and Norwegian the possessor can only be a pronoun,

as in (18a), or a proper name in construction with a pronoun, as in (18b). 16
Crucially, the possessor cannot be a bare proper name or any other nominal cat
egory.
(18) a. bilen hans (Norwegian) b. bilen hans Jon

car.D his car.D his Jon
. * bilen Jons d. * bilen lærerens
car.D Jon's car.D teacher.D's
. * bilen hans læreren
car.D his teacher.D
In addition, Norwegian but not Icelandic permits a PP as possessor in this

construction. But in NSw dialects we find a wider range of categories as pos
sessors.17 Some dialects are like Norwegian and Standard Icelandic, permitting
only pronouns or proper names in construction with a pronoun in possessor
position, but other dialects permit bare proper names as well. Still other dialects
permit family relation terms, like mama and granny. Still other dialects permit
any definite DP in this position. However, there seems to be a pattern to the
combinations of possessor categories which occur in postnominal position. On
the basis of the data we now have, it seems that the postnominal possessors con
form to the following implicational hierarchy:
(19) pronoun < pronoun + name < bare name or family term < definite DP
If a dialect allows any of these categories as postnominal possessor, it al

lows all the categories to the left in the implicational hierarchy (19). The data
come from interviews with 17 informants from Västerbotten, Norrbotten, and
Österbotten.18 All informants except one accepted a postnominal pronoun. Ten
16
The pronoun + name construction, as in (18b), is usually regarded as instantiating not a
pronoun with a proper name in some kind of appositional relation, but rather as a form of arti
cle (a 'preproprial article'), homonymous with a pronoun, with a proper name as complement:
see Taraldsen (1990), Sigurösson (1993), and especially Delsing (1993). This is clearly not
correct for all NSw dialects, as will be shown below, hence possibly not correct for any dialect.
17
Certain Icelandic dialects, too, allow a wider range of categories as possessors, according to
Sigurösson (1993). The more precise properties of the possessors in these dialects have not
been investigated.
18
Västerbotten and Norrbotten make up roughly the northernmost third of Sweden.
Österbotten is a region in Finland where Swedish is widely spoken, primarily along the coast,
across the Baltic Sea from Västerbotten.
accepted a postnominal proper name in construction with a pronoun or a bare

proper name. Of these ten, five did not allow a postnominal bare proper name,
while three did not allow pronoun + name. Six accepted a postnominal (bare)
family term. Two accepted other bare DPs postnominally.19 In (20) we list some
examples, which also exemplify some of the variation in possessive morphol
ogy. 2 0 ' 2 1
(20) a. biln hans n Janne (Älvsbyn) b. biln Jannes (Piteå)

car.D his ART Janne car.D Janne's
"Janne's car"
c. huse n mormor (Ljusvattnet) d. biln prästn (Lövånger)
house.D ART .DAT grandmother car.D priest.D
"grandmother's house" "the priest's car"
e. biln hans kyrkoherdn (Fällfors)
car.D his Vicar.D
"the Vicar's car"
None of the dialects seem to allow an indefinite DP as postnominal posses

sor. We conjecture that indefinite DPs are lowest in the hierarchy.22 »
19
We also have a few examples of postnominal family terms and other DPs constructed with a
pronoun. One informant consistently constructed all postnominal arguments, names, family
terms and other definite DPs, with a pronoun.
20
The geographical name within parentheses denotes the village or region where the informant
(or one of the informants) who produced the example in question grew up. We do not know, at
this stage in our research, to what extent the informant's judgments are representative of the di
alect spoken in this region. For all we know they may represent only the informant's idiolect.
We will still refer to these varieties as 'the dialect of Älvsbyn', etc., for convenience.
21
The postnominal pronoun+name construction was discussed briefly in footnote 15 above.
(20a) shows that the pronoun is not necessarily a preproprial article in this construction, since
in (20a) the pronoun co-occurs with a preproprial article (usually (e)n for masculine, a for femi
nine). Five different dialects in our data combine a pronoun and a preproprial article in this
way. The morpheme n in (20c) is the dative singular of the feminine preproprial article. For
some reason, the corresponding masculine () seems never to be used as a possessive marker.
22
An indefinite DP may occur embedded in a postnominal possessor PP in various dialects.
We are indeterminate as regards the status of postnominal possessor PPs. On the one hand,
Taraldsen (1990) has presented arguments which appear to show that possessor PPs behave just
like postnominal possessor DPs, which in our terms would mean that they are subject to the
same movement-and-checking requirements as possessor DPs (see the text below). On the other
hand it is tempting to view postnominal possessor PPs as an analogue of other well known
cases where insertion of a 'dummy preposition' is an alternative to Case-triggered DP move
ment.
It is not uncommon for languages to have a special position for possessive

pronouns, family terms, and/or proper names. 23 There are a number of lan
guages which have pre- or postnominal possessives, and where the prenominai
possessors are restricted to just pronouns, or pronouns and proper names, and
possibly family terms. French and other Romance languages are a case in point,
allowing only pronouns in prenominai position.
(21) mon livre, le livre de Pierre (French)

my book, the book of Pierre
German allows pronouns, family terms, and proper names, but no other DPs
in prenominai position:
(22) mein Buch, Muttis Buch, Peters Buch

das Buch des Jungens, *des Jungens Buch
my book, mummy's book, Peter's book
the book of the boy, the boy's book
Russian, too, allows pronouns, family terms, and proper names, but no
other possessors in prenominai position:
(23) babushkina kniga, Petina kniga

kniga sekretarja, * sekretareva kniga
granny's book, Peter's book
book secretary's, secretary's book
So, in French, German and Russian there is a special prenominai position

where only pronouns (French), or pronouns, family terms, and proper names
(German, Russian) move, to have their morphological and/or categorial features
checked (the prenominai forms in both German and Russian have a special affix,
distinct from the ones found in postnominal position). In Scandinavian, too, we
find the same set of possessor categories occurring in a special position, only in
Scandinavian this position is superficially postnominal. Compare for instance
German, Russian, and the following two NSw dialects:
23
Postnominal indefinite possessors are found in the dialect of Överkalix (in Norrbotten),
documented in Källskog (1992). These are always marked Dative.
(i) ji hå fånne i bis s ino gåmal gu:b (Överkalix; Källskog 1992)
I have found a gun ,DAT old man
Possibly they should be grouped with postnominal possessor PPs, as found in many dialects
(see the previous footnote).
(24) Piteå:
boken hans, boken mormors, boken Jannes
* boken lärarens, lärarens bok
book.D his, book.D grandmother's, book.D Janne's
book.D teacher.D's, teacher.D's book
(25) Älvsbyn:
boken min, boken pappas, boken hans Janne
* boken lärarens, boken åt läraren
book.D my, book.D dad's, book.D his Janne
book.D the.teacher's, book.D to teacher.D
In both these dialects the possessor in the N.D-POSS construction can be a

pronoun, a family term, or a proper name (the name constructed with a pronoun,
though, in the second dialect), but it cannot be any other DP. In the dialect in
(24), other DPs are preposed, the head noun having the bare (indefinite) form. In
the dialect in (25), the construction is saved by insertion of a preposition.
We assume, however, that we are dealing with the same special pre-NP po
sition in Scandinavian as in the other languages, and that the reason why the
position is superficially postnominal in Scandinavian is that the definite noun has
moved even further to the left than the possessor. We henceforth refer to the
process placing the possessor in pre-NP position as Possessor Shift.
That the superficially postnominal possessor in the N.D-POSS construction is
in a specifier position to the left of (underlying) N, has been shown by Taraldsen
(1990), on the basis of ordering properties and binding properties between post-
nominal possessors and other noun complements. What is the landing site of
shifted possessors more precisely, though? The Scandinavian N.D-POSS con
struction is interesting in that it shows unequivocally that the position is not
SPEC-DP (the highest SPEC position in the nominal projection), but a lower posi
tion, since the definite noun occupies D.24 Taraldsen (1990) proposes that the
possessor is in SPECNP. We do not agree. Within the present theoretical frame-
work, the categonal-morphological restrictions on the possessor which obtain in
the construction in question indicate that the possessor is the specifier of a func
tional head, specified for certain features which must match those of its specifier;
the head and its specifier are in a checking relation. If it were the case that the
properties of the head N determined which categories can occur as possessor in
the German and Russian prenominai possessor construction and in the
24
In Italian, too, there is direct evidence of a prenominai possessor position between D and
NP, since the possessive pronoun co-occurs with an article: il mio libro, lit. "the my book" .
Scandinavian N.D-POSS construction, we would be justified in postulating that

Possessor Shift is to SPEC-NP (from, say, COMP-N position). But the choice of
prenominai possessor in French, German, and Russian, and postnominal pos
sessor in the MSc N.D-POSS construction is not dependent on the properties of
N, as far as we know. 25 Instead, what determines whether a nominal category
can undergo Possessor Shift or not is some combination of morphological and
categorial features of the possessor itself.
This all leads up to an analysis where the shifted possessor lands in the SPEC
of an abstract functional head between D and NP which checks the morphological
and categorial features of the possessor, i.e., the abstract functional head is a
form of AGR. We will call it AGRG ('genitive agreement', i.e., possessor agree
ment). In French and some NSw dialects only pronouns are checked in AGRG in
overt syntax, i.e., only pronouns move visibly to SPEC·AGRG. We may assume,
following Chomsky (1993), that all possessors are checked in AGRG at LF, but
this is not crucial. In German, Russian, and other NSw dialects not only pro
nouns but also proper names and family terms are checked in SPEC-AGRG in
overt syntax. In still other NSw dialects other definite DPs, too, but not indefi
nite DPs, are overtly checked in SPEC·AGRG.
Assuming that the Argument principle holds universally, the prenominai
possessor in French, German, and Russian ultimately moves to SPEC-DP,
checking off the strong feature in D in the process. But first, we now assume,
the possessor moves to SPEC-AGRG, checking its morphological and categorial
features.26 The analysis of the Scandinavian N.D-POSS construction is as in (26):
(26)
25
See, however, Sigurðsson (1993) for some observations regarding the interplay of noun
semantics and possessor constructions in Icelandic.
26
The fact that the prenominai possessor cannot co-occur with an article in German, indicates
that the possessor is in SPEC-DP (das Buch, Peters Buch, *das Peters Buch).
The definite noun moves to D, through AGRG. We postulate that movement

of the noun to AGRG has the effect of 'activating' AGRG, or, to use another
metaphor, it makes AGRG visible at LF by including AGRG in a visible head chain
(visible by virtue of the features of bilen, in this case). This makes it possible for
a possessor to move to SPECAGRG, having its features checked. This explains
why we have Possessor Shift (to abstract AGRG) only in conjunction with defi
nite noun movement to D (prenominai possessors are discussed below).27
As mentioned, in Standard Icelandic, Norwegian, and some NSw dialects
AGRG has only features checking pronouns (that is if we regard checking of the
pronoun + name construction as a special case of pronoun checking). But in,
say, the NSw dialect of Piteå, AGRG also has a feature checking family relation
terms and proper names, while in the dialect of Lövånger AGRG can check any
definite DP (see footnote 20 on our use of the notion 'dialect'). If so, the varia
tion is due to a low level morphological parameter, that is, in terms of the typol
ogy of parameters outlined in the introduction, a minor parameter. This means
that we do not expect to find any systematic covariation between the form of the
possessor construction and other variable properties in the dialects in question.
We do not, for example, expect the dialects which allow postnominal kin term
DPs to share some other property which is systematically absent from the other
dialects.
This minor parameter is, however, subject to a hierarchy, quite possibly a
universal hierarchy, where pronouns are ordered above family terms and proper
names, which, in turn, are ordered above other DPs. It is not clear what this hi
erarchy is a hierarchy of, or in other words, along which parameter the argument
categories are hierarchically ordered. Comrie (1989) suggests animacy; the hier
archy in (19) would form the upper end of an animacy hierarchy, the lower end
of which would consist of various inanimate noun categories. Comrie discusses
some other instantiations of this putative hierarchy. An alternative parameter,
also mentioned by Comrie, is definiteness: pronouns would be somehow inher-
27
Delsing (1993: 173-175) argues that the N.D-POSS construction does not involve N-move-
ment to D, but instead would be derived by movement of a DP headed by the definite noun to
SPEC-DP, and movement of the possessor (he only considers constructions with a pronominal
possessor) to D. The structure would be
(i)
This solves some descriptive problems which beset the N-to-D hypothesis, but creates other
problems. For one thing, it is far from obvious how this theory will account for the cases
where the possessor is not a simple pronoun, hence plausibly a head, but a complex DP, as for
instance in (6b, c, d).
ently more definite than names, which are more definite than definite DPs, which
are (obviously) more definite than indefinite DPs.
6. Prenominai possessors
What is the analysis of the Standard Swedish possessive construction (27),

which also figured in the dialect of Piteå in (24)?
(27) lärarens bok

teacher.D's book
Let us assume an analysis along the lines of Fiva (1985) for Norwegian and
Abney (1987) for English: The genitive -s is base-generated as a functional head
which takes NP as complement. More specifically we propose that -s is base-
generated in AGRG. That is to say, -s is a phonetically visible instantiation of
AGRG. The genitive -s blocks movement of the definite noun to D, hence the
noun has the bare form (but see below for a counterexample). Hence the only
way the strong feature in D, required by the Argument Rule, can be checked off,
is by movement of -s to D, the possessor ending up in SPEC-DP, by movement
from inside NP. Assuming Greed, this presupposes that -s as well as the posses
sor have some feature or features which require checking in D and SPEC-DP, re
spectively. Assuming strong definiteness in D, -s must have a definiteness fea
ture. The affixation of -s would be a late phonetic process suffixing -s to the fi
nal word of the possessor DP; see Fiva (1985).28
In Standard Swedish possessive -s is not restricted in any way, but accepts
any kind of DP as specifier (unlike for instance those varieties of English where
genitive -s takes only [+animate] DPs as its specifier: the boy's book, *the beer's
colour). The reason why Icelandic does not have prenominai possessors is,
28
This analysis of (27) is consistent with the fact that a noun phrase containing a prenominai
possessor can function as a predicate, as in
(i) Per är Jannes son. (Swedish)
Per is Janne's son
As discussed by Holmberg (1993) and Mandelbaum (1994), this sentence can have a predicative
reading equivalent with, say, John is a student. According to Holmberg (1993) and
Mandelbaum (1994) the predicative reading presupposes that the possessor does not occupy the
highest SPEC-position in the nominal projection. That is to say, in this construction -s and the
possessor have not moved to D/SPEC-DP, but remain in AGRG/SPEC·AGRG. Not being an ar
gument, the construction is not subject to the Argument Rule.
quite simply, that Icelandic does not have any overt AGRG, i.e., no morpheme
corresponding to MSc and English genitive -s.
Consider again the dialect exemplified in (24): as long as the possessor was
a pronoun, proper name, or family term, the 'abstract AGRG strategy' could be
employed, with definite noun movement through AGRG to D, followed by
Possessor Shift to SPEC-AGRG. However, for any other possessor category this
option is closed, by hypothesis because abstract AGRG does not have the features
required to check DPs other than pronouns, names, and family terms. For such
DPs, the dialect employs the 'overt AGRG strategy', inserting genitive -s, an un
restricted DP-checking head, in AGRG. The possessor DP moves to the SPEC of
-s.
In Standard Swedish as well as in Danish and Norwegian the head noun
must have the bare form in the presence of a prenominai possessor, as in (27).
However, in many varieties of NSw the head noun can be definite. Thus (28) is
sharply ungrammatical in Standard Swedish, Danish and Norwegian, but per
fectly well formed in many Northern Swedish dialects.29
(28) Jannes bilen

Janne's car.D
It seems to be the case in all, or nearly all, dialects where (28) occurs, that
the prenominai possessor cannot be a pronoun. This suggests that the construc
tion requires overt AGRG, i.e., genitive -s in AGRG. 30 Why this should be so, we
do not know. The construction is problematic in the framework articulated here.
In particular, how is checking of the definite noun ensured, without violation of
the Head Movement Constraint? We leave this construction for future research,
however.
This all implies that postnominal instances of genitive -s, as in (20b) or (24)
do not instantiate overt AGRG, but something closer to a genuine Case inflection,
presumably base-generated with the noun, checked in overt syntax. Note that in
several dialects, in fact most dialects in our data, postnominal lexical DPs have a
special morphology. Particularly common is the construction with a pronoun
plus lexical DP, but, as noted in passing, other possessive forms occur as well,
for instance (20c, d).
29
It is common also in Swedish dialects in South Finland.
30
Strikingly, one informant accepted a third person pronoun in prenominai position, but with
a 'double' genitive form, an -s added to the standard possessive form.
(i) hanses bilen
his's car
Consider again the ungrammatical construction (17), repeated here:
(17) * en bil hans/Jons/läraren(s) a car his/Jon's/the.teacher('s)
We can now conclude that the reason, or one of the reasons, why (17) is un
grammatical is that the possessor has not moved to SPEC AGRG to check its fea
tures. Then what about (29a, b)?
(29) a. * en hans / Jons / lärarens bil b. * den hans / Jons / lärarens

bü
a his / Jon's / the.teacher's car the his / Jon's / teacher's
car
These constructions, too, may be ruled out by economy of representation: we

know that the genitive -s can move to D, and the possessor DP to SPEC-DP,
checking off the strong feature required by the Argument Rule, as in (27), for
instance. If so, there are alternatives to (29a, b) which contain fewer features,
namely the corresponding constructions where -s and the possessor have moved
all the way to D/SPEC-DP. Consequently (29a, b) are blocked.
7. Possessor Shift and Object Shift
The Possessor Shift construction (26) is formally similar to another

Scandinavian construction, namely Object Shift, shown in (30):
One property which Object Shift and Possessor Shift share is that they apply
only to pronouns in some varieties of Scandinavian, although not the same va
rieties. Thus, in MSc (including NSw) Object Shift is restricted to (weak) pro
nouns, while in Icelandic any definite object DP can shift:
(31) a. * Jeg så Jon ikke. (Norwegian)

b. Jeg så ham ikke.
I saw him not
The fact that only pronouns undergo Object Shift in MSc while all DPs do in
Icelandic has been explained as a consequence of the fact that only pronouns
have morphological case in MSc, while all DPs do in Icelandic (Holmberg &
Platzack 1995). The underlying assumption is that Object Shift only affects DPs
with overt Case. This hypothesis is confirmed by the Possessor Shift construc
tion, since the shifted possessor is either a pronoun, a DP headed by a pronoun,
or a DP with some form of overt possessive morphology, the only exception
being the apparently inflectionless possessor in the dialect represented in (20d).
At first sight, Icelandic seems to be a problem for this theory, though. The the
ory predicts that Icelandic should have Possessor Shift of a bare lexical DP,
since Icelandic has morphological Case, and strong Case, in the sense discussed
above. But as mentioned, Standard Icelandic allows only a pronoun or a pro
noun-plus-name as possessor in the N.D-POSS construction. This is not neces
sarily a problem, though. First, we have been informed by Halidór A.
Sigurðsson (p.c.)» that forms such as bókin Jóns, although not acceptable in
Standard Icelandic, are not infrequently encountered in spoken Icelandic.
Second, we conjecture that the fact that Icelandic has the N-POSS option for lexi
cal DP possessors means that the N.D-POSS construction with a lexical DP pos
sessor is blocked, as an effect of economy.31 This presupposes that Case is
cheaper than definiteness, and/or that (bound) case morphology is cheaper than
(bound) definiteness morphology (as discussed in section 3, what makes the N-
POSS construction work is strong Case coupled with bound case morphology).
At present we do not have any independent support for this hypothesis, except
the observation that bound case morphology is more common (perhaps even
31
There is a form of complementarity between the N-POSS and the N.DEF-POSS construction
in Standard Icelandic: N-POSS is used with lexical possessors but normally not pronouns, while
N.DEF-POSS is used with pronominal possessors only (including pronoun+name possessors);
see Sigurösson (1993). This suggests that, for some reason which we do not understand, the N-
POSS construction does not work when the possessor is a pronoun. In this case only, the
NDEF-POSS construction is allowed, in Icelandic.
much more common) among the languages of the world than bound definiteness
morphology.
Another property shared by Object Shift and Possessor Shift is that they
both presuppose movement of the lexical head, i.e., they both conform to
'Holmberg's Generalization'. Consider (32) and (33): in (32), Object Shift has
applied without verb movement (there is no verb movement in embedded clauses
in MSc). The result is ungrammatical. In (33) Possessor Shift has applied with
out definite noun movement. The result is ungrammatical.
(32) * ...at jeg ham (ikke) så (Norwegian)

...that I him not saw
(33) * hans Janne (nya) boken

his Janne new book.D
Above we proposed that the reason why the Possessor Shift requires noun
movement is that abstract AGRG must be included in a visible head chain to be
visible itself for grammatical processes, and thus able to check the features of a
possessor. This hypothesis can easily be extended to Object Shift, given the
analysis in (31): abstract AGRO needs to be included in a visible head chain to be
come visible, and thus able to check the features of an object.32 ,33
32
See Chomsky (1993) for a different account of why Object Shift requires verb movement.
Following Branigan (1992) Chomsky assumes that the reason why for instance (31) is ruled
out is that movement of the object across the subject position in SPEC.VP violates the
Shortest Movement condition.
(i) [.. Jegi [ A G R 0 P ham ... ikke [Vp e¡ [v- så e ]]]]
According to Chomsky (1993), the reason why verb movement helps to circumvent Shortest
Movement is that verb movement extends the domain of the verb, in a sense, so that
SPECAGROP and SPECVP count as equidistant from the object position, and therefore object
movement directly to SPECAGROP is technically as short as movement to SPECVP. It is not
obvious how to extend this explanation to Possessor Shift. We could postulate that SPEC-NP
of a definite noun contains an abstract category corresponding to the subject in SPECVP. This
category might be the abstract demonstrative pronoun postulated in Holmberg (1993). This
category would block possessor movement, unless the noun moves as well, making SPEC-NP
and SPEC-AGRGP equidistant from COMPN. However, as mentioned in the text, we do have
possessor movement to SPEC.AGRG, even across a definite noun (in some dialects), namely
when AGRG is overt.
33
Insofar as definite noun movement to D is obligatory, triggered by the needs of the definite
noun, (33) will be ungrammatical already because it contains an unmoved definite noun. In that
case it will not tell us anything about Holmberg's Generalization. Note, however, that in con
struction with an attributive adjective, the definite noun does not move to D (at least not
8. Conclusions
We have discussed two parameters affecting possessive constructions in the

Scandinavian languages: strong vs. weak Case, which accounted for the varia
tion in the N-POSS construction, and the feature content of AGRG, which ac
counted for the variation in the N.D-POSS construction. Both parameters are
'morphological', in the sense that they concern features of a functional head as
sociated with nominal morphology. In terms of the theory of parameters
sketched at the outset, strong or weak Case is a major parameter, exhibiting all
the diagnostic properties of major parameters listed in (1), namely
• It concerns the feature values of a functional category with a gen
eral distribution, namely the category D, present in all argument
noun phrases.
• Therefore it affects (potentially) all constructions involving the cat
egory D, that is all argument noun phrases; we mentioned the fact
that Icelandic, but not MSc, can do without an indefinite article and
a free definite article (but see footnote 7 on the free definite article),
and that Icelandic but not MSc has case morphology.
• It is resistant to change, and hence to dialectal variation: We are not
aware of any dialect in Mainland Scandinavia, however 'archaic'
or 'exotic', which would allow the N-POSS construction. Nor have
we heard of any dialect which would not have an indefinite article.
As regards Case morphology, there are several dialects in Northern
Sweden and Norway which have a morphological dative, but as
far as we know there is no dialect which would retain a
morphological distinction between subjective and objective Case,
which plausibly is a minimum requirement for strong Case to be
possible.
By contrast, the parameter responsible for the variation in the N.D-POSS
construction is a minor parameter, exhibiting all the diagnostic properties of mi
nor parameters listed in (1).
• It concerns the feature values of a category which occurs only in
noun phrases containing a possessor; therefore
• it does not affect any other constructions; and therefore
• it is vulnerable to change, hence to dialectal variation.
overtly). This is the case when a free definite article must be inserted: den nya boken 'the new
book.DEF'. Movement of a possessor to D is ruled out, as shown by (32), and as explained by
the theory presented here in terms of the visibility of AGRG.
We showed that the construction is, indeed, subject to rich dialectal varia
tion, especially in the Northern Swedish region, where, at first glance, it looks
as if every village has its own NP grammar. We showed that there is a pattern in
the variation though, the feature composition of AGRG being subject to a
(possibly) universal argument feature hierarchy. We also showed that the N.D-
POSS construction provides strong evidence of a functional head checking pos
sessors between D and NP. Finally we pointed out some striking similarities
between the N.D-POSS construction and Object Shift, another controversial
Scandinavian construction.34
REFERENCES
Abney, Steven. 1987. The English Noun Phrase in its Sentential Aspect. PhD dis
sertation, MIT.
Branigan, Philip. 1992. Subjects and Complementizers. PhD dissertation, MIT.
Chomsky, Noam. 1991. "Some Notes on Economy of Derivation and
Representation". Principles and Parameters in Comparative Grammar, ed. by
Robert Freidin, 416-454. Cambridge, Mass.: MIT Press.
____. 1993. "A Minimalist Program for Linguistic Theory". The View from
Kenneth Hale & Samuel Jay Keyser, 1-52. Cambridge Mass.: MIT Press.
Comrie, Bernard. 1989. Language Universals and Linguistic Typology. Oxford:
Blackwell.
Delsing, Lars-Olof. 1993. The Internal Structure of Noun Phrases in the
Scandinavian Languages. Department of Scandinavian Languages, University
of Lund.
Fiva, Toril. 1985. "NP-internal Chains in Norwegian". Nordic Journal of
Linguistics 8.25-47.
Holmberg, Anders. 1993. 'On the Structure of Predicate NP". Studia Linguistica
47.126-138.
____. 1994. "Morphological Parameters in Syntax: the Case of Faroese".
Reports from the Department of General Linguistics, University of Umeå, Report
35.21-62.
34
We wish to thank our informants Mona Marklund, Ann-Louise Marklund, Carita
Lundmark, Ingegerd Jonsson, Majken Widmark, Pelle Wållberg, Peder Alex, Johan Svenlin,
Britta Lundgren, Björn Nilsson, Belinda Backström, Malena Mattbäck, Daniel Gustavsson,
Carin Agerhäll, Gertrud Wallin, Anders Persson, Göran Boman. Thanks also to Tarald
Taraldsen, Phil Branigan, the audiences at the Atlantic Provinces Linguistic Association
Conference on Microparametric Syntax in Saint John, N.B., October 1944, the Scandinavian
Syntax Workshop in Oslo, January 1995, and the seminar of the Phonetics and Linguistics
Department in Bergen, February 1995, and last but not least, two anonymous referees, for their
comments on previous versions of this paper.
___. & Christer Platzack. 1991. "On the Role of Inflection in Scandinavian
Syntax". Issues in Germanic Syntax, ed. by Werner Abraham, 93-118. Berlin:
Mouton de Gruyter.
____&____. 1995. The Role of Inflection in Scandinavian Syntax. New
York and Oxford: Oxford University Press.
Kester, Ellen-Petra. 1993. "The Inflectional Properties of Scandinavian
Adjectives". Studia Linguistica 47.139-153.
Källskog, Margareta. 1992. Attityd, interferens, genitivsyntax. Studier i Nutida
Överkalixmål. [Attitude, Interference, Genitive Syntax. Studies in the Contempo
rary Överkalix Dialect] Publications of the Institute of Dialect and Folklore
Research, Uppsala, Ser. A:18.
Mandelbaum, Deborah. 1994. Syntactic Conditions on Saturation. PhD disserta
tion, City University of New York.
Platzack, Christer. 1987. "The Scandinavian Languages and the Null-Subject
Parameter". Natural Language and Linguistic Theory 5.377-402.
Pollock, Jean-Yves. 1994. "Checking Theory and Bare Verbs". Paths Towards
Universal Grammar: Studies in Honor of Richard S. Kayne, ed. by G. Cinque, J.
Koster, J.-Y. Pollock, L. Rizzi & R. Zanuttini, 293-310. Washington, D.C.:
Georgetown University Press.
Rohrbacher, Bernard. 1994. The Germanic VO Languages and the Full Paradigm.
PhD dissertation, University of Massachusetts at Amherst.
Sandström, Görel & Anders Holmberg. 1994. "Adjective Incorporation and the
Syntax of the Scandinavian Noun Phrase". Reports from the Department of
General Linguistics, University of Umeå, Report 35.81-97.
Santelmann, Lynn. 1993. "The Distribution of Double Determiners in Swedish:
Den support in D°". Studia Linguistica 47.154-176.
Sigurösson, Halldór Á. 1993. "The Structure of the Icelandic NP". Studia
Taraldsen, . Tarald. 1990. "D-projections and N-projections in Norwegian".
Grammar in Progress, ed. by Jóan Mascaro & Marina Nespor, 419-431.
Dordrecht: Foris.
Vikner, Sten. 1994. Verb Movement and Expletive Subjects in the Germanic
Languages. Oxford and New York: Oxford University Press.
THE OCCASIONAL ABSENCE OF ANAPHORIC
AGREEMENT IN LABRADOR INUTTUT*
ALANA JOHNS
Introduction
This article forms part of ongoing research into a parametric analysis of

some dialects of the Inuit language (Johns 1993, 1995). The Inuit language (in
most of Canada referred to as Inuktitut) is spokenfromAlaska to Greenland (see
Dorais 1990). As a result of the fact that migrations were from west to east, the
linguistically more conservative dialects are found in the west, while the more
innovative, i.e., those having undergone more phonological and morphological
change, are found in the east. In this work I will examine one property that dif
ferentiates Labrador Inuttut, an eastern dialect spoken on the northern coast of
Labrador, from other dialects. The property concerns the distribution of a par
ticular form of agreement suffixation (anaphoric agreement) which specifies that
one referent is an anaphor, taking another referent as antecedent. More specifi
cally, the issue revolves around whether or not anaphoric agreement may be
found in one syntactic context: possessed NPs in absolutive (nominative) case.
The research is conducted within the Principles and Parameters approach of
Generative Grammar (Chomsky 1991, 1993), where languages are viewed as
obeying a universal set of principles, and where language differences result from
Thanks to the organizers and participants of the Annual Meeting of the Atlantic Provinces
Linguistic Association held at the University of New Brunswick in Saint John in October
1994, where this paper was first presented at the special session on Microparametric Syntax
and Dialect Variation. Thanks also to the University of Toronto, where I worked on this paper
during my sabbatical leave, and to an anonymous reviewer for comments. This research was
made possible by funding provided by the Social Sciences and Humanities Research Council of
Canada (research grant number 410-94-570). Finally, thanks to all the speakers of
Qairnirmiutut and Labrador Inuttut whom I have consulted on this topic, especially Sally
Ikuutaq, Sybella Tuglavina and Harriet Lyall. All errors are mine.
122 ALANA JOHNS
differing lexical properties. Naturally languages vary as to the vocabulary items

that make up their inventory, but, vocabulary subsumes grammatical items, and
variance of these may produce different grammatical systems.1 Following the
work of Richard Kayne (see Kayne 1994b), I assume that language parameters
are more readily examined in closely related language groups, where the differ
ences are not so numerous as to make comparison nearly impossible.
The research on which this paper is based focuses on inflectional paradigms,
and the relation between paradigm differences and syntactic differences in related
dialects. The working hypothesis is that inflectional differences reflect syntactic
differences. If this hypothesis is correct, then a group of inflectional differences
may reflect a parameter.2 Much recent research on parametric differences be
tween language groups concerns movement differences (see Lightfoot &
Hornstein 1994 for examples). In languages such as Inuktitut surface movement
is not central, word order being almost completely free. In such languages,
much of the information signaled by movement is encoded in morphology, so
that a change in syntax is reflected by morpheme substitution, rather than mor
pheme displacement. As a result, we expect to see parameters reflected as varia
tion in inflection alone.
In this paper, we will examine one such variation in inflection which is im
posed by a parameter. The advantage of this approach to the study of inflectional
morphology is that it unites a number of widespread paradigmatic differences
under one explanation. A purely morphological account of these differences
would lead to numerous ad hoc morphological rules.
1. Anaphoric agreement in the Inuit language
It is beyond the scope of this paper to provide an explanation for the entire
phenomenon of anaphoric agreement in the Inuit language (see Bittner 1994;
Sadock 1994). Instead, the purpose is to show (a) that there is dialectal variation
1
The nature of this variance is not agreed upon by all generative grammarians. For instance
Chomsky (1995) seems to assume that all grammatical properties are found in all languages,
differences resulting from whether or not a particular feature of a lexical item is strong, giving
one set of properties or weak, giving another set of properties. Chomsky (1995:176) allows
that an agreement element may be "inert or perhaps missing". In my view, grammatical prop
erties form a universal set, but not all properties are observed in all languages. Variance results
from the set selected, and the interaction of this set with other sets.
2
That inflectional morphology reflects parameters does not mean that all parameters are mani
fested morphologically. For an example of parameterization within a language with little or no
morphology, see Cheng, Huang & Tang (1996) [this volume].
LABRADOR INUTTUT: ABSENCE OF ANAPHORIC AGREEMENT 123
in the facts of anaphoric agreement in the Inuit language, and (b) that this varia
tion is predicted within the parametric approach developed in Johns (1993) and
Johns (1995). Before describing the facts pertaining to dialectal variation, it is
necessary to review some of the general facts concerning anaphoric agreement in
the language.
Anaphoric agreement in Inuktitut is an explicit indication by means of
agreement morphology on either a possessed NP or a verb that an argument is
referentially co-indexed with a higher one. This contrast is illustrated in
examples (1) and (2), from the Qairnirmiut dialect of the Inuit language
(Qairnirmiutut or Q), spoken around Baker Lake, Northwest Territories.3
(1) anguü-p irni-a taku-jaa Q

man-REL Son-3S(ABS) See-TR-PART-3S/3S
"The mani sees hisj son"
(2) anguti-p irni-ni taku-jaa Q

man-REL Son-3RS(ABS) See-TR-PART-3S/3S
"The mani sees hisj son"
In example (1), we see a possessed NP irni-a "his/her son" (there is no gen

der in the Inuit language), which has an agreement marker -a, in bold. The
marker indicates third person possessor of a singular possessum. The entire NP
irni-a is in absolutive (null) case, and is the theme/patient of the verb. The tran
sitive verb is marked for agreement with the absolutive NP. In addition, the
transitive verb agrees with the other NP in relative case anguti-p "the man",
which we will for convenience call the agent. Note that word order is not fixed
in Inuktitut, although there are unmarked word orders, such as that shown in the
examples in (1) and (2), where the agent precedes the patient, and both precede
and agree with the verb.
We see that in the syntactic context in (1) (possessed NP in absolutive case),
third person possessor agreement is interpreted as disjoint in reference to the NP
in relative case. This is shown by the different indices in the English gloss ("the
man" # "his"). The example in (2) is exactly the same syntactic context—the
3
Abbreviations used in this article are the following: s for singular; Ρ for plural; D for dual; R
for reflexive; INTR for intransitive; TR for transitive; REL for relative (ergative) case; (ABS) for
absolutive case (null); SECOND for secondary case; PART for participial mood; INDIC for indica
tive mood; OPT for optative mood; NEG for negative, and 1, 2,  for first, second and third per
son. The notation 3S/1s indicates third person singular agent and first person singular
theme/patient.
124 ALANA JOHNS
possessed NP is in absolutive case. The only difference is that, in this example,

the possessor of "son" is interpreted as carrying the same referential index as
that of the NP in relative case (i.e., "the man" = "his"). This difference is indi
cated morphologically by the presence of a different possessive agreement
marker -ni, the so-called fourth person, or reflexive agreement morpheme,
shown in bold and glossed as 3R. English does not have an equivalent mor
pheme, and, as a result, such expressions are always ambiguous, as shown by
the necessity for subscripts in the English glosses.4
A slightly different syntactic context is the so-called antipassive construction
where the theme/patient is in oblique case, and the agent is in absolutive case.
Again, the possessor of the theme/patient may or may not be co-referential with
the agent, an absolutive NP, as shown in (3) and (4).
(3) angut irni-a-nik taku-juq Q

man(ABS) Son-3S-SECOND See-INTRPART-3S
"The mani sees hisj/*i son"
(4) angut irni-mi-nik taku-juq Q

man(ABS) Son-3RS-SECOND See-INTRPART-3S
"The mani sees hisi son"
In (3) and (4) we see examples from the Qairnirmiut dialect where the pos
sessed NP "his son" is in secondary case -nik, and the agent of the action "the
man" is in absolutive case. Only the latter agrees with the verb. As before, we
see that disjoint agreement between the possessor and the potential co-referent is
indicated by the possessive affix -a, while anaphoric agreement is indicated this
time by the suffix -mi.
The presence of the morpheme -nil-mi indicates obligatory co-reference be
tween an argument and its antecedent (see Finer 1985 for the seminal analysis of
switch reference phenomena under binding theory). In this property, anaphoric
agreement in the Inuit language resembles reflexive morphemes found other lan
guages, e.g., Scandinavian (Taraldsen 1996). Now consider examples of
anaphoric agreement on verbs.
(5) Miuri-up Jaani kunik-paa quviahu-ngmat Q

Mary-REL John(ABs) kiss-TRPART-3s/3s happy-because-3S
"Maryi kissed Johnj because she/hej.k *¡ was happy"
4
English could have The man sees his own son, but the addition of own adds a marked inter
pretation that is absent from the example in (2).
(6) Miuri-up Jaani kunik-paa quviahuk-kami Q

Mary-REL John(ABs) kiss-TRPART-3s/3s happy-because-3Rs
"Maryi kissed Johnj because she¡*j,*kwas happy"
In (5) and (6), we see that the verb in an adverbial clause may also display
anaphoric agreement, taking an argument of the matrix clause as antecedent. The
presence of third person agreement on the adverbial verb in (5), indicates that the
person who is happy is disjoint in reference (*i) to the agent of the kissing
(Mary), and must therefore have an independent index. Under this interpreta
tion, the one who is happy may be the patient of the kissing (John), or equally,
it may refer to someone other than the two NPs in the matrix clause.5 These
possibilities are indicated by the indices j or k
In example (6), however, the presence of the anaphoric agreement -mi on the
verb indicates obligatory co-indexing between the theme of this verb "to be
happy" and the agent of the kissing (Mary). No independent indexing is possi
ble, as indicated by *j,*k.
Examples (5) and (6) demonstrate two additional properties of anaphoric
agreement. The first is that co-indexing is not sufficient to license anaphoric
agreement. As stated above, one of the possible interpretations in example (5) is
that the one being happy is identical in reference to the one being kissed (John).
Even though these two arguments are co-indexed under this interpretation, the
morphological instantiation of co-indexing, i.e., anaphoric agreement, may not
appear. If anaphoric agreement were present in this construction, as in (6), it can
only indicate co-reference with the agent. In other words, anaphoric agreement
in an adverbial clause cannot take an absolutive NP as antecedent when a relative
case-marked NP agreeing with the verb exists. Nevertheless co-indexing of the
absolutive may take place without anaphoric agreement, and I will consider this
to be a separate process (see Grodzinsky & Reinhart 1993 for discussion and
references).
The second property that these examples demonstrate is that the third person
agreement is the unmarked (default) case. The fact that third person agreement
is found on the verb when the interpretation co-indexes the absolutive NP in (5)
shows that third person is not intrinsically marked for disjoint reference. In other
words, third person agreement is found either (a) when the licensing conditions
for possible anaphoric agreement are met (in which case, its presence signals
The latter interpretation is more likely when an explicit NP is mentioned, e.g.:

i) Miuri-up Jaani kunik-paa Piita quviahu-ngmat
Mary-REL John(ABS) kiss-TR-PART-3S/3S Peter(ABS) happy-because-3S
"Mary¡ kissed Johnj because Peterk was happy*:j, k*i "
126 ALANAJOHNS
disjoint agreement), or (b) when the licensing conditions for possible anaphoric
agreement are not met (in which case, its presence is neutral as to co-reference).
In this respect, third person is pronominal, while anaphoric agreement is an
anaphor (Finer 1985).
1.1 A paradigm gap in Labrador Inuttut
When we compare the examples in (1) and (2) with parallel constructions in
the Labrador dialect of the Inuit language (referred to as Labrador Inuttut), we
note the following difference.6
(7) anguti-up taku-janga inni-nga L

man-REL see-TR-PART-3s/3s son-3s(ABs)
"The mani sees hiSj ori son"
In (7) we see that where the possessor of the noun innik "son" has a differ
ent index from the agent of the action angutik "the man," the possessive agree
ment is the third person -nga, cognate of the -a that we saw in the Qairnirmiut
example (1); however, when the possessor of the noun has the same index as
the agent of the action, the agreement is still third person, and not the expected
anaphoric agreement affix -ni that we saw in (2). In fact, the presence of such an
affix makes the Labrador example unintelligible, as shown by the ungrammati-
cality of (8).
(8) anguti-up taku-janga * inni-ni L

man-REL See-TRPART-3S/3S Son-3RS(ABS)
"The ani sees hisi son"
As a result, the Labrador form in (7) is ambiguous in interpretation, as

shown by the indices in the English translation. The example in (7) might lead
us to hypothesize that Labrador Inuttut has lost anaphoric agreement, either
through historical change, or through contact with English. This is clearly not
the case, however, since Labrador Inuttut has anaphoric agreement morphemes
in a variety of other syntactic contexts, as shown in (9) through (12).
6
In the Labrador Inuit Standardized Spelling System  stands for q; e for ii;  for uu; and â for
aa.
(9) angutik inni-nga-nik taku-juk L

man(ABS) Son-3S-SECOND See-INTR-PART-3S
"The mani sees hisj/*i son"
(10) angutik inni-mi-nik taku-juk L

man(ABS) Son-3Rs-SECOND See-INTRPART-3S
In (9) and (10), which are parallel to the Qairnirmiut examples in (3) and (4),
we see that Labrador Inuttut clearly can license anaphoric agreement in oblique
case-marked NPs. Similar to the previous examples in (5) and (6), it can also li
cense anaphoric agreement on a verb in an adverbial clause, as shown in (11)
and (12).
(11) annak Kunga-juk angutik Kuviasu-mmat L

woman(ABS) smile-INTRPART-3s man(ABs) happy-because-3s
"The womani smiles because the manj is happyj"
(12) annak Kunga-juk Kuviasu-gami L

woman(ABs) smile-INTRPART-3s happy-because-3Rs
"The womani smiles because she is happyi"
What the above examples indicate is that agreement in Labrador Inuttut can
not be anaphoric in a particular syntactic context—on an absolutive NP.
Obviously syntax is involved to some degree. The morpheme is not lacking
from the language, and the restriction must be described in terms of a specific
case position. In the following sections I will show that this fact in Labrador
Inuttut not only is predicted by the parameter for Labrador Inuttut proposed in
Johns (1993 & 1995), but is one of many (seemingly unrelated) facts that fall
together under this parameter.
2. Clause structure and the Labrador Parameter
Based on the analysis in Johns (1992), we assume the analysis of a transi

tive clause in Inuktitut such as that shown in (13a) to be (13b).7
7
This tree could be changed to have all heads on the left of the complements, thereby ensuring
that all movement is leftward in keeping with Kayne (1994a). As far as I know, nothing in the
analysis discussed here hinges on this.
128 ALANAJOHNS
(13) a. anguti-up arnaq taku-ja-a Q

man-REL woman(ABs) see-TRPART-3s/3s
"The man sees the woman"
What (13b) indicates is that the transitive clause in (13a) contains two inflec
tional projections. The lower projection AGRPN contains the verb in the form of a
participial complement, an inflectional head AGRN and a specifier which gets rela
tive case. Note that AGRN marks the person of the agent plus NUM. The latter is
the number category of the participial. Third person theme/patient (unlike first
and second person) is a property of NUM.8 Above this projection is another
phrase AGRPV. Within this phrase, the entire AGRPN is the complement. The head
is AGRV, which marks the person of the theme/patient. The specifier of AGRPV
gets absolutive case. The participial becomes a full predicate (verb) through
movement upwards to the inflectional heads. First it moves up to join AGRN, and
then the two of them move up to join AGRV. Although this layering of inflection
cannot be seen in (13a), it is clear in (14) where the theme/patient is first person.
(14) anguti-up taku-ja-a-nga Q

man-REL see-TRPART-3S/1s
"The man sees me"
Here we see that the inflection attached to the verb is in the order predicted
by the clause structure in (13b). The third person agent inflection is internal to
the first person theme/patient inflection, due to the structure layering in (13b).
8
The idea that NUM can carry third person properties is based on ideas cited and developed in
Ritter (1995). Further work, beyond the scope of this paper, needs to be done on the structure
of AGRN, investigating, for example, whether NUM needs its own projection.
This analysis is independently motivated by a number of syntactic facts about

Inuit transitive clauses, and the reader is referred to Johns (1992) for details.
Along with the structure in (13b), our analysis will also assume that the specifier
of AGRPN must obey the following condition with respect to its head (AGRN).
(15) NOMINAL SPECIFIER CONDITION9

Nominal heads which agree with their specifiers may not move
up, stranding their specifiers; if such a head moves, the speci
fier must also move up.
This principle refers to nominal heads agreeing in feature (having a syntactic

link) with their specifier. This entails that in the absence of any such link, nomi
nal specifiers will be left behind if the head moves to a higher position.
As a result of the fact that specifiers in Inuktitut generally do agree with par-
ticipials, the movement of the participial+AGRN up to AGRV forces the specifier of
AGRPN to 'follow' its head by adjoining to the maximal AGRPV. This is shown in
(16)
In (16) we see that the participial moves up to AGRN, and subsequently the
participial+AGRPN moves to AGRV. In accordance with (15), the specifier (anguti-
up "the man") of AGRP N 'follows' its head by adjoining to AGRP V , a position in
which the Nominal Specifier Condition is met.
9
As pointed out by an anonymous reviewer, not all specifiers are subject to this restriction.
130 ALANA JOHNS
2.1 Participial vs. indicative mood in Qairnirmiutut
Not only participial mood morphemes, but indicative mood morphemes can
appear in this structure in Qairnirmiutut. The participial mood morphemes are
characterized by the presence of initial j/t, and the indicative mood morphemes
are characterized by the presence of initial v/p10, as in (17).
(17) a. anguti-up arnaq taku-va-a Q

man-REL woman( ABS) see-TRPART-3s/  s
"The man sees the woman"
Generally speaking, the presence of the indicative mood morpheme in a

Qairnirmiut declarative clause indicates some sort of surprise, or marked seman
tics. The most important characteristic of the indicative mood, for the present
purposes, is that unlike the participial mood which can function either as a
nominal verb or a matrix verb (18a), the indicative mood is always found in
main clauses, as shown by the ungrammaticality of the bracketed reading of
(18b).
(18) a. taku-jara Q
see-TR-PART-3s/s
"the one who I see" or "I see him/her/it"
b. taku-vara Q
see-TRINDIC-3s/s
"I see him/her/it" (*"the one who I see")
This restriction can be captured by defining a lexical property of the indicative

mood (see Johns 1987; 1993), such that it must always appear in AGR V , as in
(19).
(19) The indicative mood can only be in AGRV·11
10
The continuant forms /j/ and /v/ are found after stems ending in vowels, while the stop
forms /t/ and /p/ are found after stems ending in consonants.
11In fact the indicative mood is historically related to C(omp), involving properties including
narrative introduction, questions, etc. How it has become a mood of assertion is a question of
some interest.
2.2 The Labrador Parameter
As we saw in section 2.1, both the participial mood and the indicative mood
morphemes can appear in AGRV in Qairnirmiutut, this being obligatory in the case
of the indicative mood. Johns (1993 & 1995), shows that in Labrador Inuttut,12
there is a lexical restriction which does not allow the participial mood to move to
or project13 AGRV, stated in the form given in (20).
(20) LABRADOR PARAMETER:

The participial mood in the Labrador dialect may not move to
or project AGRV.
The lexical restriction in (20) effectively ensures that the only mood that will
appear in AGRV is the indicative mood in Labrador Inuttut. Although the restric
tion in (20) is a very simple and local restriction, it is equivalent to a parameter,
in that numerous dialect-particular facts of Labrador Inuttut are predicted by this
single property. Where it differs from some versions of parameters, e.g., Baker
(1996) is that it is not a property of the grammar, but of an individual lexical
item—the participial mood morpheme.
Briefly the effects of the Labrador Parameter are schematized in (21).
12 It should be noted that the Labrador region is not homogenous linguistically, and that the
parameter refers to only one speech group.
13 I leave aside an exact formulation of how a category projects, but refer the reader to recent
discussion along these lines in Grimshaw (to appear), Ghomeshi & Massam (1994), and
Chomsky (1995).
132 ALANA JOHNS
The participial mood morpheme cannot move to or project as far as AGRV-

Under the latter interpretation, the upper half of the structure in (21) would sim
ply be missing. As a result, any construction that needs an AGRV for person
features must appear in the indicative mood, which is not subject to the parame
ter in (20). Recall that not all person features are found in AGRV. Person features
of the agent of a transitive clause are found in AGRN. Only person features of first
and second theme/patient occupy the AGRV node. Third person theme/patient
features come from the category NUM, which is contained within AGRPN.
That third person theme/patient is a feature of NUM and not AGRV predicts an
asymmetry in the distribution of the participial mood and the indicative mood in
Labrador Inuttut. Any construction involving first or second person
theme/patient will necessarily appear only in the indicative mood (to project
AGRV, the indicative mood must be present), as shown below in (22a). In con
trast, third person theme/patient will optionally use either the participial or in
dicative mood, according to the semantic distinction that differentiates the two
moods (as we saw above in the Qairnirmiut examples). This is confirmed by
(22b).
(22) a. taku-vânga * taku-jânga L

See-TR-INDIC-3S/lS See-TR-PART-3S/lS
"She/he/it sees me" "She/he/it sees me"
b. taku-vaga taku-jaga L
See-TR-INDIC.lS/3 S See-TR-PART- IS/3 S
"I see him" (surprise) "I see him"
As predicted by the Labrador Parameter, first and second person agent are
not subject to mood restrictions. Because their features are in AGRN, the predicate
can, but need not, move to AGRV via the indicative mood.
Qairnirmiutut, not having the Labrador Parameter, displays no mood restric
tions with first or second person theme/patient, as shown in (23).14
(23) a. taku-vaanga b. taku-jaanga Q

See-TR-INDIC3S/lS See-TR-PART-3S/lS
"She/he/it sees me" (surprise) "She/he/it sees me"
14
The example in (23a) was colourfully translated with the English expression 'Yikes!' by
one speaker of Q.
As discussed in Johns (1995), another prediction of the Labrador Parameter

concerns the mood forms found with the negative morpheme. In Inuktitut in
general there is a restriction against the indicative mood appearing with the
negative morpheme, as shown by the Qairnirmiut examples in (24),15
(24) a. * taku-nngit-paanga Q
See-NEG-TR-INDIC3S/lS
"She/he/it doesn't see me"
b. taku-nngi-taanga Q
See-NEG-TR-PART-3S/lS
In (24a) we see that the negative morpheme -nngit- cannot be followed by
the indicative mood. Instead, after the negative, we find the participial in
Qairnirmiut declarative clauses, as in (24b). This generalization applies irrespec
tive of person.
Turning to Labrador Inuttut, we encounter a problem, in that the generaliza
tion that the indicative mood cannot follow the negative morpheme will intersect
and overlap with the Labrador Parameter, which states that the indicative mood
is required to project a level of structure wherein a subset of person features can
be expressed. Where the person features of AGRV are not required, e.g., third
person theme/patient, as in (22b), we observe the non-projecting participial
mood following the negative, as in (25).
(25) taku-nngi-taga L
See-NEG-TR-PART-lS/3S
"I don't see him/her/it"
In contrast, where the extra level of structure is required, e.g., first and sec
ond person theme or patient, neither the participial mood nor the indicative mood
is permitted following a negative, as shown in (26).
(26) a. * taku-nngi-tânga (Labrador Parameter) L

See-NEG-TR-PART-3S/lS
15
Further research needs to be done on examples such as (24a). Johns (1995) describes them
as ungrammatical, but subsequent research has shown that the combination of nega-
tive+indicative is possible in highly marked situations (see also Lowe 1988). Suffice it to say
that they are not grammatical for the reading given in (24a).
134 ALANA JOHNS
b. * taku-nngi-vânga (Negative Generalization) L

See-NEG-TR-INDIC3S/lS
The example in (26a) is ruled out by the Labrador Parameter, since the par
ticipial mood cannot project AGRV in order to accommodate first/second person
features. On the other hand, the indicative mood cannot 'rescue' the structure, as
shown in (26b), since the pan-arctic generalization restricts the appearance of an
indicative mood morpheme following a negative. As a result of this overlap,
Labrador Inuttut is forced to use another mood after the negative when first or
second person theme/patient is involved, as shown in (27).
(27) taku-nngi-lânga (Optative Mood)16 L

see-NEG-0PT-3S/is
3. The Labrador Parameter and anaphoric agreement.
In the preceding section we have seen that the Labrador Parameter described
in (20) accounts for a number of dialect-particular facts of Labrador Inuttut. In
this section, we will see that the Labrador Parameter is also responsible for the
lack of anaphoric agreement in absolutive case that was discussed in section 1.1.
Recall that in section 2, the adjunction of the NP in relative case to AGRPV
was triggered by the movement (or projection) of the predicate from head of
AGRN to head of AGRV position. In Labrador Inuttut, as we have just seen, the
predicate in the participial mood is never permitted to project or move to AGRV
(no matter what the person combination). This fact also predicts that specifier
adjunction cannot be triggered.
Consider the example in (28). Because the condition specified in (15) above
is not met (i.e., the head AGRN does not move), the specifier of AGRPN remains in
situ and does not move up to adjoin to AGRPV (where it would c-command the
specifier of AGRP V ). 17 As a result, the Labrador Parameter predicts that the abso-
16
Negative forms followed by -la- are not traditionally termed optative, but are considered to
be special negative forms (see Smith 1977). Here I use the term optative to suggest that the
-la- is related to the same form found in first and second person optative. This morpheme is
probably more accurately termed irrealis or subjunctive.
17
Alternatively, under a projection account, the AGRV projection is not there, and the absolu
tive is the subject of a nominal clause with no AGRV (see Johns 1987).
lutìve NP associated with a participial mood predicate will never be in a position

to have anaphoric agreement, since no potential antecedent c-commands it.
We now have an explanation for the absence of anaphoric agreement in

absolutive case possessive NPs, as was shown by (7) and (8), repeated here in
(29).
(29) a. anguti-up taku-janga inni-nga L

man-REL see-TRPART-3s/3s son-3s(ABs)
"The man sees hiSj or i son"
b. anguti-up taku-janga *inni-ni L

man-REL See-TRPART-3S/3S Son-3RS(ABS)
"The man¡ sees his, son"
We see then that anaphoric agreement is not possible on absolutive case NPs,
exactly the configuration where the relative does not c-command the absolutive
in Labrador Inuttut. As mentioned above in section 1, we can assume that third
person agreement is pronominal, and can therefore be co-indexed through a pro
cess of co-reference (see Grodzinsky & Reinhart 1993), so that (29a) is ambigu
ous between a co-indexed and a non-co-indexed interpretation (as in English).18
18
As mentioned above, word order is not a fixed property in Inuktitut; however, for some, but
not all, speakers of Labrador Inuttut the word order for the examples such as that in (29a) is of
ten given as [NPABS NP R E L Predicate] or "OSV" in Labrador Inuttut, rather than the unmarked
order [NPREL NP ABS Predicate] or "SOV" observed in other dialects. Word order preferences are
subtle facts in a language with mostly free word order, and further research is needed on this
topic.
136 ALANA JOHNS
At this point we might expect that the relative NP, which remains Iη situ' in
the structure shown in (28) should be able to bear anaphoric agreement, with the
absolutive NP as antecedent. After all, we have just seen that the relative NP can
not move up to c-command the absolutive one. Why then are examples such as
(30) not found?
(30) * angutik inni-mi taku-janga L

man(ABs) son-3RS.REL see-TR.3s/s
"HiS1 son sees the mani"
The fact that such sentences are also not possible leads us to conclude that
anaphoric agreement requires more than a particular structural configuration,
i.e., c-command. Following Finer (1985), Johns (1987) proposes that
anaphoric agreement also requires that there be a syntactic link between the po
tential anaphor and some agreement node of the antecedent. We can make this
notion of syntactic link more precise by stating that a syntactic domain must be
established between an antecedent and an anaphoric feature (i.e., an anaphoric
inflection) and mediated through some head. Syntactic links may be formed
between the antecedent and the potential anaphoric feature either by agreement or
by case assignment. In other words, anaphoric agreement involves more than
structure; a morphological chain must be constructed between the two elements.
Thus the conditions for anaphoric agreement can be characterized as in (31).
(31 ) ANAPHORIC FEATURE CONDITION

Anaphoric agreement can only appear when:
i) the antecedent c-commands the AGR, and
ii) the antecedent forms a syntactic link with AGR, where link
is defined as agreement or case.
We can now see that the condition stated in (31) will explain the set of
agreement facts in Labrador Inuttut. Recall that the core constructions under
consideration are examples such as (29b = 8) and (30), where anaphoric agree
ment cannot hold between a relative case NP and an absolutive case NP. As stated
above, the Labrador Parameter will prevent the potential antecedent (angutik "the
man") in (29b) from adjoining to AGRPV where it could c-command the potential
anaphor (the possessive agreement on innik "son"). At the same time, the
Anaphoric Feature Condition in (31) will prevent the SPEC of AGRN from form
ing a link with an antecedent in SPEC-AGRV, thus ruling out the possibility that
the former could bear anaphoric agreement, as shown by the ungrammaticality
of (30). This link cannot be formed because the Labrador Parameter in (20) pre
cludes movement or projection of the participial mood with AGRV, thus resulting
either in two non-linked projections, under the movement analysis, or only one
projection (under the projection analysis).
Now let us examine the structures in Labrador Inuttut where anaphoric
agreement is possible. Being grammatical, the examples from (10) and (12)
above, repeated below in (32), must have a syntactic link which is lacking in the
examples (29b) and (30).
(32) a. angutik inni-mi-nik taku-juk (=10) L

"The mani sees hisjison"
b. annak Kunga-juk Kuviasu-gami (=12) L

woman(ABs) smile-INTRPART-3s happy-because-3Rs
"The womani smiles because she is happy i '
The example in (32a) is an intransitive (or so-called antipassive) with an ob

ject in oblique case (here secondary). Crucially, in terms of this analysis,
oblique cases are marked with a bilabial consonant (thought to be cognate with
the relative case marker -(u)p on the possessor), which indicates dependency
between the oblique case and the predicate.19 Thus, although the possessed
noun inni-mi-nik does not agree with the predicate taku-juk in (32a), the NP is
syntactically linked to the predicate through oblique case marking. By transitiv
ity, the case link allows the agreement marker on the oblique NP to form a link
with the agreement marker on the predicate which in turns agrees with the sub
ject angutik "man". As a result, the possessive agreement marker on the oblique
NP may take angutik as antecedent for anaphoric agreement.
The example in (32b) is a case of an adverbial clause. Once again, it turns
out that there is a bilabial consonant (again considered to be cognate with the
relative case), this time on the adjunct predicate Kuviasu-gami. Note that in both
examples in (32), the bilabial consonant forms part of the reflexive agreement
markers (i.e., the two form a portmanteau morpheme), thus explaining the
switch from absolutive reflexive -ni to oblique reflexive -mi. The adjunct predi
cate therefore forms a direct link with the matrix predicate Kunga-juk "smiles",
and allows for a chain to be formed between its agreement and the NP annak
"woman" which agrees with the matrix predicate.
19
The oblique cases are -mik, -mut, -mit, -mi, -kkut, and -tut. Respectively, they indicate pa
tient, goal, source, location, path, and similarity.
138 ALANA JOHNS
Finally it is obvious that in languages without the Labrador Parameter, a

structure with a possessed absolutive NP bearing anaphoric agreement (such as
the Q example in (2) above) will be legitimate under the Anaphoric Feature
Condition. In such dialects, where the participial mood morpheme can move or
project to AGRV, the participial mood morpheme will be positioned in AGRV (see
the tree above in (16)), where it is linked by agreement both to the c-command-
ing NP in relative case (through AGRN which has moved/projected along with the
mood morpheme) and to the possessed NP in absolutive case (through AGR V ),
thus satisfying the linking condition in (31).
In summary, we have seen that the Labrador Parameter accurately predicts
the constructions in Labrador Inuttut in which anaphoric agreement may occur,
and those in which it may not. Moreover, these facts form part of a larger set of
facts that derive from the Labrador Parameter, including a dialect-specific re
striction on certain person/mood combinations. We have also seen that this ac
count supports an analysis of anaphoric agreement as requiring overt morpho
logical linking. This is not unexpected, given that anaphoric agreement is itself
an anaphor within a functional rather than lexical category.
3.1 NP adjunction and the indicative mood
In the preceding section, we saw that the Labrador Parameter, independently

needed for a wide variety of agreement facts dealing with the distribution of the
participial mood, predicts the absence of anaphoric agreement on NPs in absolu
tive case. As the Labrador Parameter prevents movement/projection to AGRV, this
means that not only will there be no adjunction of the relative case-marked NP to
AGR V , but as well, there is no possibility for the participial mood to serve as a
syntactic link between a relative case NP and a possessed NP in absolutive case.
In this section, we will examine the indicative mood, which as stated in (19) can
and must appear in AGRV in all dialects of Inuktitut, and therefore is not subject
to the Labrador Parameter. As a result, we expect that anaphoric agreement
should be possible on a possessed NP in absolutive case in a construction con
taining a predicate morphologically marked for indicative mood. This turns out
not to be the case, as shown in (33).
(33) anguti-up inni-nga taku-vauk L

man-REL son-3s(ABs) see-TRINDIC-3s/s (*-ni 3RS(ABS))
"The mani sees his¡ or j son"
As can be seen in (33), even though the predicate is in the indicative mood,
the anaphoric agreement marker -ni is not allowed. Once again, only an ambigu-
ous third person -nga is found. Assuming that the Anaphoric Feature Condition
in (31) is correct, there must be some explanation for this fact. Either the an
tecedent does not c-command the possessed NP, or there is a link missing from
the syntactic chain through AGRV which links the NP in relative case to the pos
sessed NP in absolutive case. Recall that the Nominal Specifier Condition in (15)
stipulates that a specifier must follow a nominal head with which it has formed a
syntactic link. In addition, the Anaphoric Feature Condition requires that there
be some syntactic means by which the potential anaphoric agreement can link up
with the NP in relative case. The question now is whether or not there is
morphological evidence that the indicative mood in Labrador Inuttut does not
achieve this linkage, thereby accounting for the lack of anaphoric agreement. Is
there some principled factor that allows us to distinguish between Qairnirmiutut
and Labrador Inuttut in this respect? It turns out that Labrador Inuttut is distinct
from all other dialects of Inuktitut in that the indicative mood agreement does not
reflect a syntactic relation with the agent NP. Consider the Labrador data in (34).
(34) a. taku-vauk L
see-TR-INDIC-3/3s
"He/she, they (2), they (many) see him/her/it"
b. taku-vâgik L
See-TR-INDIC-3/3D
"He/she, they (2), they (many) see those two"
 taku-vait L
see-TR-INDIC-3/3
"He/she, they (2), they (many) see them (many)"
As the examples in (34) show, the agreement markers in the Labrador in
dicative mood convey only the number of the theme/patient. In this regard, ex
amples (34a, b, and c) are singular, dual and plural respectively. These markers
do not convey number agreement with the agent NP in relative case. 20 What this
indicates is that the conditions of the Anaphoric Feature Condition in (31) are not
met; the possessive AGR on the NP in absolutive case cannot form a syntactic link
through AGRV to the NP in relative case. Anaphoric agreement cannot occur.
Compare similar examples in the indicative mood from a northern Quebec
dialect (NQ), shown in (35).
20
In fact the participial mood in Labrador Inuttut also lacks agent agreement. I leave it for fu
ture research to determine what rôle, if any, this has in the more general phenomenon of the
Labrador Parameter.
140 ALANA JOHNS
(35) Transitive indicative third person forms, from Dorais (1988)

a. taku-vaa 3s/3s NQ
b. taku-vaak 3D/3S
 taku-vaat 3P/3s
d. taku-vaangik /
e. taku-vait /
What (35) shows is that, even though the number of the agent is not ex
pressed by the inflection when the theme/patient number is more than one (e.g.,
35 d, e), the inflection reflects the number (singular, dual or plural) of the agent
when the theme/patient is singular. From this we can conclude that AGRV is
linked up to both the NP in relative case and the NP in absolutive case in this di
alect. Not surprisingly then, we find examples such as that in (36) taken from
Dorais (1988).
(36) ulu-ni tigu-vaa NQ

knife-3Rs(ABs) hold-TRINDIC-3s/3s [my gloss]
"Shei holds her¡ knife"
In this account then, the Anaphoric Feature Condition correlates the unique
ness of Labrador Inuttut having -vauk for 3s/3s and no third person agent
agreement in the transitive indicative paradigm with the lack of anaphoric agree
ment on possessed absolutive NPs. In all other dialects that can use the -v- as a
declarative mood, the 3s/3s form is -vaa . In addition, all other dialects inflect
the predicate for the number of the agent in relative case.21
In summary, we have seen that there are two properties whose existence
disallows anaphoric agreement in possessed absolutive NPs in Labrador Inuttut.
One property is the Labrador Parameter which prevents a participial mood from
moving to or projecting an AGRV node. As a consequence of this parameter, there
is no possibility of a link between the possessed NP in absolutive case and the
NP in relative case via AGRV. The second property is that the indicative mood
paradigms do not agree with third person agents. Without agreement, there is
again no possibility of the absolutive and the relative NPs forming a link that
would satisfy the Anaphoric Feature Condition.22
21
Certain issues remain to be explained, such as the role of number in this dialect difference,
i.e., is it a by-product or source of the difference? As well, the fact that the -vauk form can be
found in some other dialects in the transitive interrogative mood needs to be explained.
22
There may some common basis underlying this conspiracy (see footnote 20).
4. Conclusion and implications for other analyses
As we have seen, the absence of anaphoric agreement in some but not all
constructions in Labrador Inuttut is explained under an analysis whereby
anaphoric agreement must be both c-commanded by an antecedent, and syntacti
cally linked to the agreement of the antecedent (as proposed in Johns 1987).
Under this view, the basis of anaphoric agreement is overt morpheme to mor
pheme linking. In the case of the participial mood morpheme in Labrador
Inuttut, the requisite conditions are not met, since the Labrador Parameter pre
vents the participial mood morpheme from forming a link (via movement or
projection to AGRV) between the NPs in absolutive and relative case. In the case
of the indicative mood, the mood morpheme appears in AGRV, but lack of
agreement with the relative NP precludes a link being established with the abso
lutive NP.
As stated at the beginning of this paper, it is not the goal here to provide full
and detailed analysis of anaphoric agreement in general in Inuktitut.
Nevertheless, it is useful to review some of the facts and conclusions of this pa
per with previous analyses. For example, an analysis of anaphoric agreement
that is based purely on c-command of a 'syntacticized' argument structure, such
as that in Manning (1994) (see also Grimshaw 1990), will have difficulty ex
plaining why the 'transitive' form in Labrador Inuttut cannot have anaphoric
agreement, but the 'antipassive' can, as shown in examples (8) and (10), re
peated here in (37).
(37) a. anguti-up taku-janga * inni-ni L

man-REL See-TR.PART-3S/3S Son-3RS(ABS)
b. angutik inni-mi-nik taku-juk L
In terms of argument structure, both examples involve an agent and a pos

sessed theme/patient. Under the traditional assumption that it is the object and
not the subject of an antipassive that is affected by argument structure, any anal
ysis of anaphoric agreement based solely on c-command and argument structure
does not predict a grammaticality difference between the two examples, since the
agent angutik "man" will c-command in both. Moreover, recall that examples
such as (37a) are grammatical in all other dialects of Inuktitut. We would have
no justification in positing either (a) that the argument structure of these two ex
amples is different, or (b) that this difference in argument structure is not found
142 ALANA JOHNS
in similar constructions in related dialects. Only an account that involves syntax

(here overt morphology) to some degree can provide an explanation for why the
same argument may vary in properties from one construction to another.23
More syntactic accounts of anaphoric agreement in the Inuit language (see
Finer 1985; Bok-Bennema 1991; Bittner 1994; Sadock 1994; and Bittner & Hale
1996) must address the issue of why the structural relation between the relative
NP and possessed absolutive NP is different in Labrador Inuttut from that of
neighbouring dialects. Bittner & Hale (1996) Hcense anaphoric agreement based
on whether or not agreement is able to project a subject path. It is possible but
not clear that they could explain the absence of anaphoric agreement in the
Labrador examples by appealing to the lack of agreement between the relative NP
and the subject path (similar to the argument made in this paper that neither the
indicative nor the participial mood can form a link between the pertinent NPs). In
general then, further research on person and number indexing across dialects
should prove revealing.
REFERENCES
Baker, Mark. 1996. The Polysynthesis Parameter. Oxford University Press.
Bittner, Maria. 1994. Case, Scope, and Binding. Dordrecht: Kluwer Academic.
___ & Ken Hale. 1996. "Ergativity: Towards a Theory of a Heterogeneous
Class". Linguistic Inquiry21,4.
Bok-Bennema, Reineke. 1991. Case and Agreement in Inuit. Studies in Generative
Grammar 38. Berlin & New York: Foris Publications.
Cheng, Lisa L.-S., C.-T. James Huang & C.-C. Jane Tang. 1996. "Negative
Particle Questions: A Dialectal Comparison". Microparametric Syntax and
Dialect Variation, ed. by James R. Black & Virginia Motapanyane, 41-78.
Amsterdam and Philadelphia: John Benjamins Ltd. [this volume]
Chomsky, Noam. 1991. "Some Notes on Economy of Derivation and
Representation." Principles and Parameters in Comparative Grammar,ed. by
Robert Freidin, 416-454. Cambridge: MIT Press.
___. 1993. "A Minimalist Program for Linguistic Theory." The View from
Kenneth Hale & Samuel Jay Keyser, 1-52. Cambridge: MIT Press.
___. 1995. The Minimalist Program. Cambridge: MIT Press.
Dorais, Louis-Jacques. 1988. Tukilik: An Inuktitut Grammar for All. [Distributed
by Association Inuksiutiit Katimajiit Inc., Université Laval, Québec]
23
Argument structure may indeed play a role in anaphoric agreement. My point is only that it
cannot be the sole factor.
___. 1990. Inuit Uqausiqatigiit:Inuit Languages and Dialects. Arctic College,

Nunatta Campus. [Distributed by Association Inuksiutiit Katimajiit Inc.,
Université Laval, Québec]
Finer, Daniel. 1985. "The Syntax of Switch Reference". Linguistic Inquiry 16.35-
56.
Ghomeshi, Jila & Diane Massam. 1994. "Lexical/Syntactic Relations without
Projection". Linguistic Analysis 24.175-217.
Grimshaw, Jane. 1990. Argument Structure. Cambridge: MIT Press.
__-. to appear. "Minimal Projection, Heads and Optimality". Linguistic
Inquiry.
Grodzinsky, Yosef & Tanya Reinhart. 1993. 'The Innateness of Binding and Co-
reference". Linguistic Inquiry 24.69-102.
Johns, Alana. 1987. Transitivity and Grammatical Relations in Inuktitut. PhD
dissertation, University of Ottawa.
____. 1992. "Deriving Ergativity". Linguistic Inquiry 23.57-87.
____. 1993. "Symmetry in Labrador Inuttut". MIT Working Papers in
Linguistics 18.43-58. Cambridge: Dept. of Linguistics, MIT.
____. 1995. "On Some Mood Alternations in Labrador Inuttut". Grammatical
Relations: Theoretical Approaches to Empirical Questions, ed. by Clifford
Burgess, Katarzyna Dziwirek & Donna. Gerdts, 131-151. Stanford, California:
Center for the Study of Language and Information.
Kayne, Richard. 1994a. The Antisymmetry of Syntax. Cambridge: MIT Press.
____.1994b. "Micro-Parametric Syntax". Paper presented at the 18th Annual
Meeting of the Atlantic Provinces Linguistic Association, Saint John, New
Brunswick.
Lightfoot, David & Norbert Hornstein, eds. 1994. Verb Movement. New York:
Cambridge University Press.
Lowe, Ronald. 1988 "La forme déclarative et l'alternance des suffixes -vu- et -ju-
dans les dialectes inuit de l'arctique canadien de l'est". Revue québécoise de
linguistique 17.137-165.
Manning, Christopher. 1994. Understanding Ergativity. PhD dissertation, Stanford
University.
Ritter, Elizabeth. 1995. "On the Syntactic Category of Pronouns and
Agreement". Natural Language and Linguistic Theory 13.405-443.
Sadock, Jerrold. 1994. "Reflexive Reference in West Greenlandic".
Contemporary Linguistics 1.137-160. [formerly University of Chicago Working
Papers in Linguistics].
Smith, Lawrence. 1977. Some Grammatical Aspects of Labrador Inuttut (Mercury
Series No. 37). Ottawa: National Museum of Man [Canadian Museum of
Civilization].
Taraldsen, Knut Tarald. 1996. "Reflexives, Pronouns and Subject/Verb
Agreement in Icelandic and Faroese". Microparametric Syntax and Dialect
Variation, ed. by James R. Black & Virginia Motapanyane, 189-211.
Amsterdam and Philadelphia: John Benjamins Ltd. [this volume]
HYPOTHETICAL INFINITIVES AND
CROSSLINGUISTIC VARIATIONS IN
CONTINENTAL AND QUEBEC FRENCH1
FRANCE MARTINEAU & VIRGINIA MOTAPANYANE
Université d'Ottawa & University of New Brunswick-Saint John
1. Introduction
Hypothetical Infinitives (hereafter HIs) are non-finite clauses with conditional

interpretation that occur in Continental French (CF) and Quebec French (QF).
These constructions display several properties usually incompatible with non-
finite clauses: placement of HIs is restricted to dislocated positions, as in (1);
HIsdisallow extractions, as in (2); they license lexical subjects in Nominative
(see Vinet 1985 for arguments toward Nominative assignment in these clauses),
as in (3). These properties characterize HI clauses in both varieties of French2:
(1) a. Être nés riches, travailleriez-vous autant?3 QF

be-INF born rich work-coND you so much
"If you'd been born rich, would you work so much?"
b. * Travailleriez-vous autant être nés riches?
c. Pierre y aller seul, ce serait vraiment courageux. CF
Peter there go-INF alone that be-coND really brave
"If Peter went there alone, that would really be brave."4
d. * Ce serait vraiment courageux Pierre y aller seul.
lr
This paper is the result of a complete collaboration between the authors; the order of names is
alphabetical.
2
Although HI constructions occur in both CF and QF, they stand out as a productive option
only in QF. Villiard (1984:18) notices the absence of this construction in CF to the extent that
some native speakers do not recognize it; however, most speakers accept it, and show a passive
knowledge.
3
from Villiard 1984:18
4
from Vinet 1985:411
146 FRANCE MARTINEAU & VIRGINIA MOTAPANYANE
(2) a. * Quoi gagner,j'achèterais une maison. QF

what win-INF I buy-coND a house
b. * Quoi gagner, ce serait merveilleux? CF
what win-iNF that be-coND fantastic
(3) a. Mamère m'avoir fait ça, je lui parlerais plus. QF

my mother me have-INF done that I her talk-coND no more
"If my mother had done that to me, I would not have talked
to her any more".
b. La France battre le Brésil, ce serait inconcevable. CF
France beat-INF Brazil that be-coND inconceivable
"France beating Brazil would be inconceivable."5
The parallel behavior in (1) to (3) led Villiard & Vinet 1983, Villiard 1984,
Vinet 1985 to an analysis in which HIs have the same underlying structure in
both CF and QF. Along these lines (but in the framework of the Minimalist
Program), this paper will argue in § 2-5 that the properties in (1) to (3) follow
from the interaction between the features of the  Complementizer (C) and the
placement of the respective clause (Complementizer Phrase—henceforth CP) in
Topic. These two factors have an impact on the internal structure of HIs, which
qualify as 'absolute clauses'; that is, HI constructions behave like independent
(vs. subordinate) clauses and yield effects of finiteness, such as lexical subjects
in Nominative6.
In § 6, the paper will focus on facts of variation, which concern the referen
tial content of the null subject. More precisely, HIs present arbitrary null subjects
in CF, as in (Ic) above (see also Vinet 1985). On the other hand, the null subject
in QF may be co-referent with the subject of the matrix clause, as in (4a), disjoint
in reference, as in (4b), arbitrary, as in (4c), or may even occur with atmo
spheric verbs, as in (4d).
(4) a. Retourner en arrière, on trouverait ça vieux jeu. QF

return-INF back wefind-coNDit old game
"If we went back in time, we'd find it old-fashioned."7
5
from Vinet 1985:408
6
HI clauses seem to be a peculiarity of modern varieties of French. Old and Middle French also
had infinitive constructions with lexical subjects, but their distribution and interpretation were
different (see Martineau 1990, Junker & Martineau 1992). For example:
(i) il ne cuidoit point [sa fille estre tieulle.]
"He did not think that his daughter was like that."(from Martineau 1990:447)
7
Martel 1971: 987 [Québec 42f 121:6/ source: corpus de l'Estrie]
HYPOTHETIC AL INFINITTVES IN CONTINENTAL & QUÉBEC FRENCH 147
b. Avoir pas su parler l'anglais, la maison partait en feu.

have-INF not know speak English, the house go-IMPF on fire
"Had I been unable to speak Eng., the house would've burnt."8
c. De prendre un verre, ça passerait le temps,
of have-iNF a glass that pass-COND the time
"It would help pass the time, if we have a drink."9
d. Hier, avoir fait beau, nous finissions.
yesterday have-iNF made nice we finish-IMPF
"If the weather had been nice yesterday, we would have
finished it."10
The contrast between CF (lc) and QF (4a-d) revolves therefore around the
interpretation of null subjects: QF presents some flexibility in recuperating the
referential content of the null subject; it also includes internal identification, as in
(4b). Several questions arise from these observations: if we consider that HIs
have the same underlying structure in CF and QF, how does the structure allow
for different conditions on null subjects? Do HIs in QF meet the criteria for pro-
drop finite clauses? What property of the language would justify a pro-drop op
tion in HIs?
The analysis proposed here argues that HI configurations meet the criteria for
finite clauses. Hence, licensing and identification of null subjects (pro) at the in
ternal level is possible if the language has opted for the Null Subject Parameter;
on the other hand, the conditions on pro identification at the internal level are in
sufficient in non-pro-drop languages, such as Standard CF. In this sense, the
paper claims that QF behaves like a pro- drop language, in contrast to Standard
CF, and to a higher degree than Colloquial CF (as discussed in Roberge & Vinet
1989, Auger 1994).
2. Base generation vs. movement to Topic position
The sentences in (1), discussed in Villiard (1984) and Vinet (1985), show
that HI clauses surface only in Topic position. Further investigation of these
clauses will lead to the claim that HI clauses are base generated in Topic position,
not moved to this position. Tests confirming this analysis contrast HIs with non-
8
p.c. Françoise Mougeon, corpus Mougeon-Béniak
9
from Villiard 1984:32
10
from Dulong 1952:151
finite clauses that originate in a position subcategorized by the matrix verb. It

appears 11 that HIs cannot undergo clefting, as in (5a); they do not license para
sitic gaps, as in (6a); and they co-occur with WH-extraction to matrix CP, as in
(7a). On the other hand, fronted non-finite clauses, whose traces are governed
by a selecting head, allow for all the operations banned in HIs, as shown in (5b),
(6b), and rule out concurrent WH-movement, as in (7b).
(5) a. * C'est [gagner le gros lot] queje saurais quoi faire avec.
it is win-INF the big ticket that I know-coND what do-INF with
b. C'est [vivre dans cette ville]i que je ne veux plus ti.
it is live-iNF in this town that I not want any more
"It is living in this town that I don't want any more."
(6) a. * [Avoir un montant comme ça]i, on saurait garder ti sans

depenser ei.
have-INF an amount like that we know-coND keep-INF
without spend-INF
b. ? [Avoir vu les faitsJi je voulais déclarer ti à la police sans
commenter ei.
have-iNF seen the facts I wanted-IMPF declare-INF to the
police without comment-INF
"Having seen the facts I wanted to state [them] to the police
without commenting."
(7) a. [Rencontrer un pape], que lui dirais-tu?

meet-INF a pope what to-him say-coND you
"If you met a pope, what would you tell him?"
b. * [Revoir Pierre], où veux-tu?
see again-INF Peter where want-INDIc you
The contrastive behavior of the HIs and the fronted non-finite clauses which
originate in a position subcategorized by the verb indicates that only the latter
undergo a form of WH-movement. Fronting through WH-movement licenses
clefting (5b), parasitic gaps (6b), and disallows parallel WH-chains, as in (7b).
Since  clauses do not move, we consider them base generated in Topic.
11
Positional tests are conducted in QF, since HIs have a higher productivity in this language
and native speakers have clear judgments.
HYPOTHETICAL INFINITI VES IN CONTINENTAL & QUEBEC FRENCH 149
Furthermore, the types of chains that obtain in (5) to (7) are also different.
We adopt the typology of A'-chains proposed in Cinque (1990), who distin
guishes between: (i) chains headed by a constituent base generated in an A'-po
sition; and (ii) chains headed by a constituent which moved to an A'-position.
The latter, but not the former, has quantificational properties. Along these lines,
HI clauses form non-quantified A'-chains, which include resumptive pronouns,
such as ce, ça "that" in (1c), (3b) for CF, and (8a) for QF. Note that resumptive
pronouns are optional in QF, as shown in (8b).
(8) a. [Marie avoir trouvé un chum]i, çai m'aurait calmée.

Mary have-INF found a friend that me had-coND calmed
"Had Mary found a friend, that would have calmed me down."
b. [Marie avoir trouvé un chum], j'aurais été contente.
Mary have-iNF found a friend I have-cond been happy
"Had Mary found a friend, I would have been happy."
To sum up this section, the tests in (5) to (8) show that HI clauses are base
generated in Topic position, and head non-quantified A'-chains, which may in
clude resumptive pronouns.
3. Hypothetical Infinitives project to the Complementizer Phrase
3.1 Arguments from the internal structure
There is no empirical indication whether HIs project to IP or to CP.

Nevertheless, some operations allowed in HI clauses would rather favor a CP
analysis.
First, the presence of Nominative subjects in non-finite clauses, as illustrated
in (9), is generally associated with CP-clauses.
(9) [Camille avoir été malade], je serais restée à la maison.

Camille have-iNF been sick I be-coND stayed at the home
"If Camille had been sick, I would have stayed home."
Second, dislocation to Topic position within the HI, as in (10), shows that
the hierarchy of projections by-passes the IP level
(10) [ma belle-mèrei, li'amener ei à l'hôpital], dans une semaine

on l'enterre.
my mother-in-law her bring to the hospital in one week we
her bury
"If we brought my mother-in-law to the hospital, we would
bury her in a week."12
Finally, the complementizer de in HI clauses, as in (11), provides the crucial

piece of evidence for the definition of these clauses as CPs. In section 4.4. be-
low we will analyze de as an element in SPEC-CP.
(11) Tu sais, ma Lauri, ben des fois, j'ai l'air d'un sans coeur,
mais [d'íe voir comme ça] tomber en gondole, j'braillerais!
you know my Lauri many of times I have the air of one
without heart but of you see-INF like that fall off a gondola I
bawl-coND
"You know, my Lauri, I often behave like a rough man, but
if I were to see you like that, falling apart, I would cry." 13
The sentences in (9) to (11) would receive a natural account if we consider that
HIs project to CP.
3.2 Free alternation with si-conditional clauses
The definition of HIs as CP clauses would also explain their free alternation
with si-conditionals, which project to CP. The alternation is illustrated in (12):
(12) [Gagner de l'-], si on aurait de l'argent là, viens jamais qu'à

n'avoir un jour là, ça hein, j'aimerais ben ça aller en Europe.
make-INF some money if we had-coND some money there
come ever that to not have one day then that then I like-
coND well this go-INF in Europe
"If I made some mon-, if we had some money, then, if ever
the day comes when we have that, then, I would really like
to go to Europe."14
Martel 1971: 658 [Montréal 26f 24:35/source: projet de l'Estrie]

Martel 1971: [A. Dessureault-Descôt 162:heading/source: projet de l'Estrie]
Martel 1971: 207 [Estrie 36f 197:9/source: projet de l'Estrie]
HYPOTHETIC AL INFINITI VES IN CONTINENTAL & QUEBEC FRENCH 151
The hypothesis advanced in this section is that infinitives in Topic position

receive a hypothetical interpretation because they are dominated by a C-head with
'hypothetical' features. A C-head in HIs represents the non-lexical version of the
conditional complementizer si "if, which selects finite verbs.
Facts from Romance languages, such as presented in Kayne (1991), support
this analysis. Kayne (1991) demonstrates that the Romance morpheme si/se,
which occurs in indirect interrogatives and conditional clauses, has a comple
mentizer (C) status. This complementizer is incompatible with non-finite clauses
because of the conditions on the licensing of pro: si in  governs SPEC-IP, which
contains pro, and induces a violation of Binding Principles. However, V-to-I
movement may waive this effect in certain languages. That is, languages with
the word order infinitive-clitic allow for silse with infinitive clauses, whereas
languages with the clitic-infinitive word order do not:
(13) a. Gianni non sà se andare al cinema.

John not knows if to go to the cinema
"John does not know if he should go to the cinema."
b. * Marie ne sait pas si aller au cinema.
Mary not knows if to go to the cinema
The contrast between Italian (13a) and French (13b) follows from the fact
that the infinitive verb intervenes between se andpro in overt syntax, canceUng
the illicit government relation in Italian, whereas the same type of V-to-I move
ment does not apply in clitic-infinitive languages like French. Si is incompatible
with non-finite clauses in French, since it interferes with control on the non-lexi
cal subject15. Most HIs display control, in free alternation with lexical subjects;
therefore, it is expected that si be excluded from these constructions.
The analysis sketched above confirms that a C-head with [hypothetical] fea
tures dominates both finite and non-finite conditional clauses; the former exhibit
si, the latter present an empty head, which move from V-to-l-to-C in LF. Since
the clause with the latter configuration occurs in Topic, the hypothetical features
of  must interact with the non-finiteness of the embedded inflectional phrase
(IP), and lead to a configuration which changes the  into a finite clause.
15
In this paper we adopt the hypothesis initially proposed in Borer (1989) that all the prono
minal empty categories correspond to pro; the licensing conditions on pro differ in finite and
non-finite clauses.
4. Hypothetical Infinitives as finite clauses
4.1 Hypothetical Complementizers carry [+qu] features
First, we notice that a hypothetical Complementizer Phrase displays effects

of SPEC-head agreement as it does in interrogative CPs. More precisely,
SPECCP is blocked for WH-movement, as if it already contained a [+qu] ele
ment. Consider the examples in (14): WH-movement rules out both si-conditional
and HI clauses. One might suppose that (14a) violates the double filled CP con
straint. However, QF does not observe this constraint in interrogatives, as fur
ther shown in (14c); note that both si and que occupy   position (see Kayne
1991 for arguments). Moreover, WH-movement is also ruled out in HI clauses
(14b), which have their  position empty16.
(14) a. * Si Jean arrive quand, nous serons contents?

if John arrives when we be-FUT happy
b. * Jean arriver quand, on serait contents?
John arrive-INF when we be-coND happy
 Je me demande qui que tu vois en cachette.
I me ask who that you see in hiding
"I wonder who you're seeing secretly."
Again, hypothetical clauses (15c, e) are similar to interrogatives (15b) in so

far as they cannot be embedded under prepositional heads, such as pour. This
test contrasts hypothetical and interrogative clauses on the one hand, and CPs
with non-quantified features on C, on the other. The latter do occur with pour
(15a, d):
(15) a. Pour réussir à cet examen, on fera tout.

for succeed in this exam we do-FUT all
"We will do everything to pass this exam."
b. * Pour où réussir, on fera tout?
c. * Pour prendre un verre, ça passerait le temps.
for have-INF a glass that pass-coND the time
16
Tests such as (14b) have led Vinet (1985) to the conclusion that HIs project to IP, and lack
the necessary CP host.
HYPOTHETICAL INFINITIVES IN CONTINENTAL & QUÉBEC FRENCH 153
d. Pour qu'il réussisse à cet examen, on fera tout,

for that he succeed-SUBJ in this exam we do-fut all
"We will do everything in order for him to pass this exam."
e. * Pour si on prenait un verre, ça passerait le temps.
for if we had-IMPF a glass that pass-cond the time
The observations in (14) and (15) show that the hypothetical features on 
must imply [+qu] features. Hence the prediction that (i) C enters a local relation
with SPECCP, and marks it with [+qu] features; and (ii) lack of si triggers verb
movement to C, necessary to fulfill the checking on the hypothetical feature. The
mapping of these two operations depends on the properties of the embedded IP.
4.2 V-to-I-to-C and non-finitene ss
Let us consider the type of verb movement to  The order subject-negation-

clitic-verb in (16) indicates that V-to-I does not take place in overt syntax. It is
generally assumed (Chomsky 1993) that this type of movement applies in the LF
component. This assumption is based on previous analyses of French infinitive
inflection which proved that V-to-I is not triggered overtly in infinitive clauses
(with the exception of auxiliary verbs, as argued in Pollock 1989). Investigation
of other operations specific to infinitive clauses (i.e., licensing and identification
of pro) led to the hypothesis that I-to-C must apply covertly (see Borer 1989 for
an analysis of anaphoric AGR and obligatory I-to-C in infinitives). Therefore we
follow current analyses and assume that French infinitives do not present V-to-I
movement in overt syntax, but allow for V-to-I-to-C in LF.
(16) Marie pas y aller, je serais triste.

Mary not there go-INF I be-COND sad
"I would be sad if Mary did not go there."
Another general property of French infinitives is their negative value for

subject agreement and tense features. In consequence, both the subject and the
tense receive their referential content from the matrix clause; this operation is
mediated through a form of anaphoric binding (see Borer 1989 for anaphoric
binding of the null subject, and Terzi 1992 for anaphoric binding of the embed
ded tense). Apparently, HI clauses are not exceptional, since their temporal inter
pretation depends on the tense value of the matrix IP, as noted in Vinet
(1985:412):
(17) a. * La France battre le Brésil, cela a été inconcevable.

France beat-INF Brazil that has been unbelievable
b. * Le frigidaire tomber en panne, on a eu de l'air fin.
the fridge break-INF down we have had air fine = [looked
incompetent]
Vinet attributes the ungrammaticality of (17) to the clash between the unreal
ized tense in the infinitive and the past tense of the matrix clause; only a condi
tional verb would render (17) comprehensible.
A closer look will reveal, however, that the clash does not concern the tense
value in the matrix, as long as those tenses are included in conditional
paradigms:
(18) a. Avoir pas su parler l' anglais, la maison partait/serait partie

en feu.
have-INF not known speak-INF English the house
went/would have gone on fire
b. Ma belle-mère l'amener à l'hôpital, dans une semaine, on
l 'enterre/l' enterrerait.
her bury/we her would bury
The continuous past (18a) or the present tense (18b) can appear in the ma
trix, on a par with conditional forms, and the hypothetical interpretation of the
infinitive is maintained.
The tense and mood variations in (18) cast doubts on the realization of
anaphoric binding between the tense in the matrix and in the HI clause: the latter
does not copy, feature-by-feature, the mood/temporal specifications in the ma
trix. Moreover, anaphoric binding should be excluded for technical reasons: HIs
occupy dislocated positions, and they head chains on an optional basis—for ex
ample, the sentences in (18) do not exhibit chain formation. Lack of chains situ
ates the respective clauses outside the binding domain of the matrix clause: in
this configuration, one cannot suppose that binding involves some internal
empty category and becomes a property of the chain. An account of the realiza
tion of mood and tense values is therefore crucial for the understanding of the
internal structure of HI clauses.
5. Finiteness from abstract modals
5.1 Kayne's (1992) abstract modal
In order to account for the temporal value of Τ in hypothetical infinitives, we

start with the observation that this construction has the status of a quasi-inde
pendent clause, due to its placement in Topic position. From this point of view,
the hypothetical infinitive resembles the infinitive used as a suppletive imperative
in Romance languages, such as argued by Kayne (1992) and illustrated below:
(19) a. Non parlare a nessuno!

not to speak to nobody
Kayne makes the observation that infinitive inflections cannot appear in root
clauses, unless they are licensed in a way which ensures them the required
finiteness. The exact mechanism proposed in this analysis is triggered by the
presence of the negation: the negative marker non licenses an empty modal ele
ment, specified for positive tense value, as part of the inflectional chain; the in
finitive raises to the level of the empty modal head and acquires the tense fea
tures that enable it to appear in a root context. Arguments toward this analysis
come from Italian dialects where there is a lexical counterpart to the modal ele
ment, i.e. stà in Paduan (19b)—the element stà is restricted to negative impera
tives, as shown by the contrast with (19c):
(19) b. Non sta parlare!

not AUX to talk
. * Stà parlare!
AUX to talk
5.2 A temporal operator
Returning now to the hypothetical infinitive in QF, we follow Kayne's

(1992) line of argumentation and assume that positive tense features are licensed
in this type of inflection, and that is why the infinitive can appear in an indepen
dent context. However, we diverge from Kayne (1992) with respect to the
mechanism through which the tense features are licensed in the structure. First,
we cannot attribute the capacity to license a tense operator to the negation, be
cause the negation is optional in hypothetical constructions; moreover, this type
of infinitive projects to CP and displays marked features on C, which have hier
archical precedence over negation. Hence, we focus on the [+hypothetical] fea-
ture of C and consider it to be the licensing factor for positive temporal features
on L
We mustfirstobserve that hypothetical clauses (finite or non-finite) display a
variety of inflectional forms. Thus, indicatives, conditionals and infinitives al
ternate as grammatical tools for conveying the hypothetical reading: the resulting
reading depends, as already mentioned, on the presence of si or on the availabil
ity of I-to-C movement, in a specific context (i.e.,Topic-position in QF).
In hypothetical infinitives, the implication of temporality in the [hypothetical]
feature appears in the near obligatory conditional forms in the matrix clauses:
(20) a. [Gagner de l'argent), j'aimerais ben ça aller en Europe.

to make of the money I like-cond well this to go to Europe
"If I made some money, I would like to go to Europe."
The sentence in (20a) shows the hypothetical infinitive behaving as a corre

lated structure with respect to modal values, on a par with the non-standard
(20b), where the subordinated verb copies the values of the matrix verb.
(20) b. Je gagnerais de l'argent, j'aimerais ben ça aller en Europe.

I make-coND of the money I -nd well this to go to Europe
"If I made some money, I would like to go to Europe."
Temporal values of the hypothetical infinitives are also closely dependent on

the tense value in the matrix ('concordance des temps'). In (21a), the matrix
verb is present conditional and the present infinitive is used. In (21b), the matrix
verb is past conditional and the infinitive is also past:
(21) a. Gagner del'argent, j ' aimerais ben ça aller en Europe.

b. Avoir gagné de l'argent, j'aurais aimé ben ça aller en Europe.
The transfer of modal and temporal values of the matrix verb is not restricted
to conditional matrix verbs. As the indicative may also express the hypothetical
feature, we expect to find hypothetical infinitives when the matrix verb is indica
tive, as shown in (5b) and (22):
(22) a. Avoir pas su parler l'anglais, la maison partait en feu.

have not known speak English the housego-INDICon fire
"Had I been unable to speak English, the house would've burnt."
HYPOTHETICAL INFINITIVES IN CONTINENTAL & QUEBEC FRENCH 157
b. Ma belle-mère, l'amener à l'hôpital, dans une semaine on

l'enterre.
her bury-INDIC
"If we brought my mother-in-law to the hospital, we would
bury her in a week."
However, a matrix indicative verb, without the hypothetical value, turns the
sentence ungrammatical, as shown in (23):
(23) a. * Avoir pas su parler l'anglais, la maison partira en feu

have not known speak English the house go-IND/FUT in fire
b. * Ma belle-mère,l'amener à l'hôpital, dans une semaine on
l'enterrera.
my mother-in-law her bring to hospital in one week we her
bury-IND/FUT
Similarly, a conditional matrix verb which does not have the hypothetical inter
pretation rules out the sentence:
(24) a. * Avoir gagné de l'argent, je me demandais si j'irais en

Europe.
have made of the money I myself asked-IND/IMPF if I go-
coND to Europe
b. * Avoir gagné de l'argent, je me demandais si je serais allé en
Europe.
have made of the money I myself ask-iND/IMPF if I be-coND
gone to Europe
Note that in (24) the ungrammaticality is not due to the fact that the condi
tional verb is in a subordinate clause since sentence (25), with the tensed hypo
thetical conditional in a subordinate clause, is grammatical:
(25) Avoir de l'argent, il a dit qu'il resterait ici.

have of the money he has said that he stay-coND here
"He said that he would stay here if he had money."
The facts in (20) to (25) show that a transfer of temporal and modal values
takes place in hypothetical infinitive constructions, restricting the choice of in
flectional forms and the past/present interpretation. We relate this restriction to
158 FRANCE MARTINEAU & VIRGINIA MOT AP ANY ANE
the properties of the [hypothetical] feature in C, which projects a Specifier posi

tion, against which it can check the temporal value of the clause. The configura
tion is represented in (26):
(26)
The local SPEC-head relation in CP
We consider that the feature [+hypothetical] subsumes both conditional and

temporal interpretations, which would result from the co-indexation of Op and 
with embedded and matrix T. Co-indexation with matrix Τ is required because
the [hypothetical] feature does not display a 'hypothetical morphology'. Thus,
co-indexation is the means through which the hypothetical mood is identified
and then checked in syntax through the local SPEC-head relation represented in
(26).
The abstract Operator in (26), co-indexed with matrix T, is chain related to
embedded T, to which it transfers the positive temporal value. The structure in
(26) extends to hypothetical clauses in general, irrespective of the lexical/non-
lexical realization of  However, when the embedded form is an infinitive, the
temporal operator in SPEC-CP can also license an abstract modal, which confers
finiteness to the clause, along the lines of the analysis in Kayne (1992).
5.3 Hypothetical Infinitives with de
The examples in (27) occur in QF, and seem to contradict our analysis,
wherein  has both [+hypothetical] and [+qu] features: SPEC.CP in these exam
ples displays the element de usually compatible with [-qu] complementizers.
(27) a. De prendre un verre, ça passerait le temps.

of have a glass that pass-coND the time
"It would help pass the time if we had a drink."
HYPOTHETICAL INFINITIVES IN CONTINENTAL & QUEBEC FRENCH 159
b. De sortir un peu, ça lui changerait les idées.

of go out -INF a bit that him change-coND the ideas
"It would take his mind off things if he went out for a while." 17
We will argue that the examples in (27) represent a second type of HI con
struction, in which hypothetical  has [-qu] features. We follow the analysis in
Kayne (1991), where the complementizer de receives the definition of a
Specifier. Placement of de in SPEC.CP blocks WH-movement to this position on a
par with the presence of an empty Operator, as suggested in (26). On the other
hand, as a non-operator, de does not have the capacity to license abstract
modals, which are crucial for the status of subjects in HI clauses.
The following observations will indicate that a de-HI does not meet the crite
ria of an 'absolute clause'; thus, de-HIs must be typologically and functionally
different from HIs with empty CPs. First, clitic chain formation is obligatory in
de-Hls, as in (28a); the chain is represented in (28b).
(28) a. [De prendre un verre]i, çai lui passerait le temps.

of have-INF a drink this him pass-coND the time
b. CP i ... ç a i . . . ei
The empty category falls in the binding domain of the matrix IP. Binding and
control become properties of the chain, and are implemented through
'connectivity'. Note that HIs with empty CPs do not enter the chain in (28b),
unless optionally; therefore, they are not subject to connectivity.
Further examples in (28c,d) confirm that lack of resumptive pronouns rules
outde-HIs;the sentences become grammatical if we delete de.
(28) . * De prendre un coup, il menacerait sa famille.

of receive-INF a blow he threaten-coND his family
d. * Depenser à lâcher la job, tu devrais avertir le patron.
of think-INF to quit the job you should warn the boss 18
Second, lexical subjects cannot occur in de-Hls, as in (29), whereas this is

an important property of HIs with empty CPs; again, de deletion would render
the sentence grammatical.
17
from Villiard 1984: 32
18
from Villiard 1984: 32
160 FRANCE MARTTNEAU & VIRGINIA MOTAPANYANE
(29) a. * De Marie penser à lâcher la job, ça devrait plaire au patron.

of Mary think of quit the job this should please the boss
Third, the null subject in a de-HI is obligatorily co-referent with an argument

in the matrix, as in (30b); when co-reference is not available, the interpretation
of pro is arbitrary, as in (30a); on the contrary, the interpretation of null subjects
in HIs with empty CPs does not depend on control, as previously shown in (4).
(30) a. De rar/*je,il prendre un verre, ça passerait le temps.

b. De pr oil/*je sortir un peu, ça lui i changerait les idées.
The facts in (27) to (30) indicate that a de-HI behaves like a dislocated clause
which is chain related to a position in the matrix clause. The chain mediates the
processes of binding and control, and yields the 'connectivity' effects proper to
adjunct clauses. Therefore,de-HIs do not meet the criteria of 'absolute clauses',
which are finite, quasi-independent clauses.
To sum up sections 4 and 5, HI clauses project to CP, and are equivalent to
si-conditional clauses. The C-head carries the hypothetical feature, which allows
for the interpretation of HIs irrespective of the mood specifications in the matrix
(the latter may be conditional or indicative). The hypothetical feature pairs with
[+/-qu] features: [-qu] features allow for insertion of lexical elements in CPs
(i.e., de) and confer on HIs the properties of adjunct clauses; on the other hand,
[+qu] features license an Operator in SPECCP, which, in turn, licenses an ab
stract modal; the abstract modal confers finite features on HIs , and allows them
to function as 'absolute clauses'.
6. Cross-linguistic variation: the licensing of null subjects
Sections 1 to 5 have argued that HI clauses have the same underlying struc
ture in CF and QF. Nevertheless, variation arises within this common structure
with respect to the licensing of null subjects. Vinet (1985) shows that null sub
jects receive an obligatory arbitrary interpretation in CF, as in (31a); arbitrary
null subjects alternate with lexical subjects, as in (31b):
(31) . rr Pouvoir partir en vacances, ce serait merveilleux.

able-iNF go on holidays this be-coND wonderful
"If one could go on holidays, it would be wonderful."
HYPOTHETIC AL INFINITIVES IN CONTINENTAL & QUEBEC FRENCH 161
b. Pierre y aller seul, ce serait vraiment courageux.

Peter there go-INF alone this be-coND really brave
"If Peter went there alone, that would be really brave."
The restriction on subject interpretation in (31) follows naturally from the

definition of His as absolute clauses. Neither control nor positive agreement
features are available in absolute clauses: therefore the subject is either lexical
(licensed under case checking by positive T) or arbitrary pro (a configuration for
pro identification is not obtainable).
On the other hand, QF presents a variety of  clauses in which the null sub
ject may be controlled, arbitrary, or have disjoint reference. These instances are
illustrated in (4). Furthermore, the interpretation of the null subject in QF HI
clauses indicates that the null pronominal can be licensed and identified inter
nally. This observation is based on evidence that HIs display effects that gener
ally characterize pro-drop languages (see Rizzi 1982, 1986, 1990 for tests on
Italian dialects). For example: (i) pro occurs with atmospheric verbs as in (32)
and in impersonal expressions as in (33); (ii) pro has disjoint reference, as in
(34); and (iii) non-lexical pro alternates with emphatic pronouns as in (35a), as
opposed to clitic subjects as in (35b):
(32) a. Avoir fait mauvais, j'aurais resté à la maison.

have-INF done bad I have-coND stayed in the house
"If the weather was bad, I would have stayed home."
b. Avoir venté un peu moins, serais-tu venu en chaloupe avec
nous-autres?
have-iNF winded a little less be-coND you come in boat with
us others
"If it had been less windy, would you have come with us in
the boat?"19
(33) Avoir venu du monde pour pousser le char, on se serait

sorti de d'là ben plus vite.
have-iNF come the people for push-INF the car we ourselves
be-coND go out from there a lot faster
"If some people had come to push the car, we would have
been out of there a lot faster."
19
from Villiard 1984:38. Villiard mentions that judgments vary for the sentences (32): some
speakers accept them very readily, others less so.
(34) Si ils arriveraient puis [rils arrêter de monter ci]

if they arrive-coND then stop-iNF of raise-iNF this
puis [proils monter ça], [rils monter ça], on en aurait as
sez de notre salaire
then raise-iNF that raise-iNF that we of it have-coND enough
of our salary
"If they came and if they stopped raising (prices) for this
and for that, we would have enough salary (to live on)." 20
(35) a. Seigneur, [moi avoir dit ça à mon père], il m'aurait renfer

mée jusqu'à vingt et un an.
God I have-INF said that to my father he me have-coND
locked until twenty one year
"My Lord, if I had said that to my father, he would have
locked me up until I reached the age of twenty one."21
b. * Seigneur, [je avoir dit ça à mon père] ...
The properties in (i) to (iii) indicate that HI clauses behave like finite clauses
in pro-drop languages. This behaviour is compatible with the hypothesis put
forth in Roberge & Vinet (1989) and Auger (1994, 1995a,b) that QF presents the
characteristics of a pro-drop language in finite clauses.
6.1 Evidence for pro-drop in QF
Roberge & Vinet (1989) and more recently Auger (1994, 1995a, b) have
proposed an analysis of QF as a pro-drop language22. Roberge & Vinet suggest
that languages where subject clitics must be expressed, as it is generally the case
in QF, are pro-drop. They differ from languages such as Italian, which allow the
absence of subject clitics, in the way pro is licensed in subject position. In
Italian, pro receives its semantic content from the verbal ending; in French, pro
is licensed by the subject clitic, as in (36):
(36) pro SCL V
20
Martel 1971: 602 [Montréal 25h 18:23/source: projet de l'Estrie]
21
Martel 1971: 658 [Montréal 26f 4:13/source: projet de l'Estrie]
22
As shown further in this paper, the pro-drop character of QF is best argued for on the basis
of subject clitic doubling, a phenomenon which also occurs in Colloquial CF. Since Standard
CF does not exhibit subject clitic doubling, we prefer not to extend the pro-drop analysis to
this variety of French, without further evidence.
For Auger, who adopts a similar analysis, the subject clitic, as pronominal
affix, is not an argument which saturates the subcategorizing frame of the verb.
Thus, an overt lexical subject may co-occur with the subject clitic, as in (37):
(37) Les maringouins ils me suivent.

the mosquitoes they me follow
"Mosquitoes follow me."23
In Auger (1995a), it is shown that the NP in (37) enters a subject doubling

construction, and is not dislocated. For instance, bare quantifiers, which tend to
be excluded from subject dislocated positions, are found in QF, as in (38):
(38) En campagne, quand quelqu'un il dansait....

in country when someone he danced
"In the country, when someone danced..."(from Auger 1995a)
The arguments summed up in this section indicate that QF qualifies as a pro-

drop language with subject clitic doubling.
6.2 Expletive pro in QF
Roberge & Vinet (1989) also argue for the presence of pro in expletive ex
pressions, where the lexical expletive subject pronoun has been deleted, as in QF
sentences in (39)24.
(39) a. M'est avis que vous méritez ce qui vous arrive.

to-me is opinion that you deserve what to-you happens
"It is my opinion that you deserve what happens to you."
b. Me semble que Jean va venir.
to-me seems that John will come
"It seems to me that John will come."25
23
from Auger 1995a, 108-71:036
24
Roberge & Vinet (1989) leave aside cases where the lexical expletive is obligatory in em
bedded clauses (ib), although its deletion is possible in independent clauses (ia):
(i) a. Pleut un peu.
rains a little/ "It rains a little."
b. * Je sais bien que pleut un peu.
I know well that rains a little
For them, (ia) is the result of a possible phonological deletion.
25
from Roberge & Vinet 1989:110-11
164FRANCEMARTINEAU&VIRGINIAMOTAPANYANE
They propose that a syntactic rule may delete an expletive subject when an
object marker (me in (39)) or another element which could licence the expletive
subject is present. Explicitly, they propose the following principle:
(40) proexp is identified by INFLX

where χ represents a sub-set of grammatical features
Features under AGR and Tense would be found in this sub-set.

On the basis of deletion of the lexical expletive in (39), the authors further
propose that expletive pro may also be allowed in HIs (such as (32), (33)), when
certain criteria are met; that is, Tense has positive features, and the  is adjunct
to an INFL with [+tense/conditional], [+AGR] features.
Although the pro-drop analysis in Roberge & Vinet extends to both QF and
Colloquial (as opposed to Standard) French, the  phenomena, as illustrated in
(4) and (32-35), surface only in QF. This asymmetry is related to frequency:
sentences like (39) are more frequent in QF than in Colloquial French; it is then
expected that the option for the pro-drop setting be also more frequent in QF 
clauses. However, the data indicate a total absence of expletive and referential
pro in CF HI clauses. Therefore, an explanation is needed for the fact that the
pro-drop option is productive in QF vs. CF non-finite clauses.
6.3 The null subject parameter in QF HI clauses
The hypothesis advanced in this section is that reanalysis of HIs as finite

clauses extends the finiteness to AGR in QF. Increased frequency of the pro-drop
setting in independent finite clauses (as shown in § 6.1., 6.2.) favours the
speaker's pro-drop setting in  clauses, which meet the criteria for finiteness.
The background for our hypothesis comes from the analysis of pronominal
elements in Corver & Delfitto (1993). In this study, null pronominals are NPs
which need substantive features (identification) and visibility (case feature
checking). For null pronominals in subject position, there are two configurations
in which the conditions on identification and case can be implemented, as shown
in (41).
In (41),the D-head enters a local head-head relation with the complex
[AGR+T]. With AGR carrying substantive features, and T carrying case-checking
features, D mediates the transfer of both sets of features to SPEC-DP (which
hosts pro) because of the structural local relation of Specifier-head. The transfer
takes place when pro raises to SPEC-DP and enters a local relation with the me
diating D-head. In (41a) the D-head is lexical, because the [N] features on T are
weak and insufficient for case checking; note that il has an inherent Nominative
case. On the other hand, the configuration in (41b) is possible in a language

where the [N] features on Τ are strong and sufficient for D to case mark pro in its
SPEC position. Therefore, the representations in (41a) and (41b) account for a
contrastive pattern of licensing null pronominals in pro-drop languages (41b)
and non-pro-drop languages, such as Standard CF, represented in (41a).
The local SPEC-head relation in CP
If we apply the analysis in (41b) to  clauses, we obtain an account for the

extension of the positive value from Τ to AGR: these two functional heads merge
their features before they enter a local relation with the subject position.
Moreover, the subject position in HI clauses is locally related to the abstract
modal, to which the complex [AGR+T+V] raises, as in (42):
(42)
V-movement to abstract Modal in HI
According to (42), the abstract modal shares a positive value with the com
plex [AGR+T], as opposed to Τ only. Further effects on the subject position de
pend on the intrinsic features on Τ—that is, whether Τ has strong or weak [N]
features in the language.
As mentioned for (41a), Τ has weak [N] features in Standard CF, and re
quires the presence of a lexical D, such as iL The function of il, which is to
complement and transfer [AGR+T] features to pro, must be implemented in overt
syntax: il lacks lexical content and becomes invisible at LF. Therefore, if a DP
with the D-head il is inserted in SPEC-MP in (42), the structure would be ruled
out. Verb movement to the modal and further to  takes place at LF in  clauses,
but il is not visible to LF rules, and the assignment of substantive features to the
166 FRANCE MARTINEAU &VIRGINIAMOTAPANYANE
null pronominal cannot be implemented. Thus, the analysis in Corver & Delfitto
(1993), applied to  clauses in French, accounts for non-occurrence of clitic
pronouns in these constructions (35b), although lexical subjects and strong pro
nouns are allowed (35a).
The same analysis indicates, on the other hand, that licensing and identifica
tion of null pronorninals in (4), (32-35) must take place according to the pattern
in (41b). In this sense, the QF versions of HI clauses exhibit the properties of
pro-drop languages such as Spanish, and exclude subject clitics. Indeed, the
grammaticality of those examples can be accounted for only if we assume that
(41b) is the configuration for the null pronominal, and it is inserted in the subject
position in (42). The D-head is non-lexical, so it pairs with a T-head with strong
[N] features. The complex [AGR+T], with strong [N] features, raises to the ab
stract modal, which confers finiteness to it; thus, the complex [AGR+T] transfers
the required features to D, which in turn transfer these features to the pronominal
in SPEC-DP. This process is licit at LF.
To conclude this section, HI clauses in CF present the properties of Standard
non-pro-drop Continental French, and thus contrast with Hypothetical
Infinitives in Québec French with respect to the licensing of null subjects: (41a)
applies in Standard CF, (41b) applies in QF.
6.4 Final remarks
The properties of HI clauses, approached through the analysis in Corver &

Delfitto (1993), deepen the contrast between QF and CF with respect to the set
ting for the null subject parameter. Although a variety of CF (i.e., Colloquial
French) was shown to behave as a pro-drop language with subject clitic dou
bling, on a par with QF, productivity of the pro-drop option is higher in QF.
HI clauses takes this difference a step further, since the pro-drop option in
their context implies licensing configurations which exclude subject clitic dou
bling: at this point, CF speakers, who occasionally accept pro-drop in finite
clauses if they allow subject clitic doubling, reverse their pro-drop option in HIs,
where clitic doubling is not obtainable. On the other hand, QF speakers, who
have a higher exposure to pro-drop finite clauses, maintain this option in HIs.
HYPOTHETIC AL INFINITI VES IN CONTINENTAL & QUÉBEC FRENCH 167
REFERENCES
Auger, Julie. 1994. Pronominal Clitics in Quebec Colloquial French: A
Morphological Analysis. Ph.D dissertation, University of Pennsylvania.
__. 1995a. "A Morphological Analysis of Quebec Colloquial French
Pronominal Clitics". CLS 31-II: Papers from the Parasession on Clitics, ed. by
A. Dainora et al., 32-49. Chicago: Chicago Linguistic Society.
___. 1995b. "Les clitiques pronominaux en français parlé informel: une ap
proche morphologique". Revue québécoise de linguistique 24,1.21-60.
Borer, Hagit. 1989. "Anaphoric AGR". The Null Subject Parameter, ed. by O.
Jaeggli & K. Safir, 69-109. Dordrecht: Kluwer Academic Publishers.
Chomsky, Noam. 1993. "A Minimalist Program for Linguistic Theory". The
View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger,
ed. by Kenneth Hale & Samuel Jay Keyser, 1-52. Cambridge: MIT Press.
Cinque, Guglielmo. 1990. Types of A'-Dependencies. Cambridge: MIT Press.
Corver, Norbert & Denis Delfitto. 1993. Feature Asymmetry and the Nature of
Pronoun Movement. University of Utrecht: OTS Working Papers [OTS-WP-TL-
93-013].
Dulong, Gaston. 1952. "La langue franco-canadienne". Pédagogie Orientation
6.148-155.
Junker, Marie-Odile & France Martineau. 1992. "The Structure of Infinitives".
Probus 4,3.1-27.
Kayne, Richard. 1991. "Romance Clitics, Verb Movement and PRO". Linguistic
Inquiry 22.647-686.
___. 1992. "Italian Negative Infinitival Imperatives and Clitic Climbing".
Hommage à Nicolas Ruwet, ed. by L. Tasmowski & A. Zribi-Hertz, 300-312.
Ghent: Communication and Cognition.
Martel, Pierre. 1971. Corpus de VEstrie (includes Corpus de l'Estrie, de Québec,
de Montréal, de Saguenay Lac Saint-Jean). Sherbrooke, Québec: Université de
Sherbrooke.
Martineau, France. 1990. La montée du clitique en moyen français: une étude de
la syntaxe des constructions infinitives. PhD dissertation, Université d'Ottawa.
Pollock, Jean-Yves. 1989. "Verb Movement, Universal Grammar and the Structure
of IP". Linguistic Inquiry 20.365-424.
Rizzi, Luigi. 1982. Issues in Italian Syntax. Foris: Dordrecht.
__. 1986. "On the Status of Subject Clitics in Romance". Studies in
Romance Linguistics, ed. by O. Jaeggli &  Silva-Corvalan, 391-420.
Dordrecht: Foris.
___. 1990. Relativized Minimality. Cambridge: MIT Press.
Roberge, Yves & Marie-Thérèse Vinet. 1989. La Variation dialectale en gram
maire universelle. Montréal: Presses de l'Université de Montréal.
Terzi, Arhonto. 1992. PRO in Finite Clauses. A Study of the Inflectional Heads of
the Balkan Languages. PhD dissertation, City University of New York.
Thibault, Pierrette. 1984. Corpus Montréal 84: Projet de recherche sur le français
parlé à Montréal, Université de Montréal. See also Thibault, Pierrette & D.
Vincent. 1990. Un corpus de français parlé. Collection recherches sociolinguis-

tiques. Québec: Université Laval.
Villiard, Pierre. 1984. Infinitives hypothétiques disloquées, INFL et contrôle. MA
thesis, Université de Sherbrooke.
___ & M-T. Vinet. 1983. "Remarques sur l'expression de l'hypothèse en
québécois". Travaux de linguistique québécoise 4.209-221.
Vinet, Marie-Thérèse. 1985. "Lexical Subjects in French Infinitives". Selected
Papers from the 13th Linguistic Symposium on Romance Languages, ed. by L.
King &  Maley, 407-423. Amsterdam: John Benjamins.
THE SECOND PERSON SINGULAR
INTERROGATIVE IN THE TRADITIONAL
VERNACULAR OF THE BOLTON
METROPOLITAN AREA
GRAHAM SHORROCKS
Abstract
The dialect of the Bolton Metropolitan Area (Greater Manchester County,

England) contains a distinction between second person singular and plural in the
personal pronoun subsystem. In the case of the second person singular, inter
rogatives may be realized with a full form of the pronoun theaw "thou", or with
an enclitic form of the pronoun. The negative particle may also be cliticized in
negative interrogatives. The processes of reduction, assimilation and simplifica
tion have resulted in some extremely contracted interrogative forms, which have
not, on the whole, been described in the earlier literature. Several forms may
well be unique to an as-yet-geographically-undefined area around Bolton in the
North-West of England. My main purpose here is to describe the quite consider
able array of second person singular forms found in a large corpus of speech
(primarily free conversation, in which such forms occur fairly frequently, and
also naturally); to make a number of comments on the phonology and morpho
logical derivation of these forms; and to analyze their syntactic distribution. One
important fact to emerge is that negative tags may be used after negative as well
as positive statements in the Bolton Metropolitan Area dialect: this matter affects
the distribution of contracted as opposed to non-contracted forms of the second
person singular interrogative. All in all, the situation described here is very dif
ferent indeed from that which obtains in standard English, and is in certain re
spects different too from anything that has so far been described for other non
standard dialects. The study illustrates clearly the need for a corpus of tape-
recorded conversation in the investigation of non-standard syntax; and reveals
170 GRAHAM SHORROCKS
that there is more variation amongst English dialects at the morphological and
syntactic levels than has previously been supposed
1. Introduction
The following study is based in part upon data from my Grammar of the
traditional urban vernacular of the Farnworth area (1981), the fieldwork for
which was carried out in the 1970s (especially 1972-1974), as well as on further
data from fieldwork that took place in the 1980s within the Bolton Metropolitan
Area.1 Bolton itself is situated circa 12 miles north-west of Manchester. It is part
of a very densely populated area in what was formerly south-east Lancashire
(the South-East Lancashire Conurbation),2 and there is currently a wide variety
of industries there. In the past, however, the traditional industries were textiles
and mining. A consideration of historical, geographical, social, economic, ad
ministrative, and cultural factors shows the Bolton Metropolitan Area to be a pe
culiarly homogeneous unit under the urban field of influence of Bolton (see
Shorrocks 1981:1-27, § 0, especially pp. 15-21, § 0.5; Freeman, Rogers &
Kinvig 1966:218-219; Saxelby 1971:114-115; and Smith 1969:73). This unity
is reflected in the speech of the area, which is relatively homogeneous in charac
ter, and distinctive vis-à-vis that of other areas (see further Shorrocks 1981:21-
27,75-77, and 681). It is important to realize that the major cotton towns around
Manchester are not in any sense mere suburban satellites of Manchester, but that
each constitutes a relatively independent entity. Bolton is both the largest, and
least dependent on Manchester (see Freeman, Rogers & Kinvig 1966:218-219,
and 222). The data for the study are chiefly derived from tape recordings of
'free', i.e., unscripted, conversation made with informants representative of the
traditional vernacular of the area.
1
1 am grateful to the Social Sciences and Humanities Research Council of Canada for Standard
Research Grant #410-92-1137, which has enabled me to continue investigating these second
person singular forms as part of a wider study of the dialects of the North-West of England.
2
In 1972 the Local Government Act was passed, based upon the recommendations of the
Maud Commission, and the county boundaries were redrawn in 1974 in accordance therewith.
Bolton and the surrounding area then became a part of the newly-created Greater Manchester
County. For details of the history, geography, and socio-economic make-up of Greater
Manchester County, see Frangopulo (1977).
BOLTON VERNACULAR: SECOND PERSON SING. INTERROGATIVE 171
2. Second person pronouns
The dialect of the Bolton Metropolitan Area preserves a distinction between

second person singular and plural in the personal pronoun subsystem.3 The sin
gular forms are pronounced /ðε:, (ðæ:), ðε/ in the subjective case, the last of
these being somewhat reduced; and in the objective case /ði:, ðI/. Plural forms
are /jo:, (joə), jө/. In the case of the second person singular, interrogatives may
be realized either with a full form of the pronoun theaw, i.e., when the pronoun
is stressed, with a somewhat reduced form, /ðε/, or with enclitic forms of the
pronoun. The negative particle not is also sometimes cliticized in negative inter
rogative sentences, though see § 4, below, for restrictions thereon. The pro
cesses of reduction, assimilation and simplification have resulted in some ex
tremely contracted forms, which on the whole have not been described, let alone
explained, in the scholarly literature. Several of these forms may well be unique
to an as-yet-geographically-undefined area around Bolton. We shall not be con
cerned here with those interrogatives that contain a full or somewhat reduced
form of the pronoun, as /ðuz ðε(:)/ "does thou?", /kɔn ðε(i)/ "can thou?", ex
cept inasfar as their distribution affects that of the enclitic forms, but will focus
on the latter, where pronoun and verb, and the cliticized negative particle where
applicable, have coalesced, as:
/'dustө/, or even more contracted /dust/ "does thou?"

/,ðus(ө)nt/ "does thou not?"
/kɔnt/ remember? "can thou remember?"
Theaw can remember Joe Breawn, /katnt/?
'Thou can remember Joe Brown, can thou not?"
3
In dialect literature, of which there has been and still is a great deal in this area, the second
person singular pronoun is spelled theaw or theau in the subjective case, and thee in the objec
tive case; the plural pronoun is spelled yo. Theaw is always singular; but the singular-plural
distinction is complicated by traces of yo as a polite singular, and also by the fact that un
stressed yo and the unstressed form of the standard pronoun you—which has both singular and
plural reference—have the same pronunciation in the Bolton Metropolitan Area, viz., /jө/.
The reader should note that [Θ] does not have the value currently assigned to it in the latest re
vision of the alphabet of the International Phonetic Association (see Journal of the
International Phonetic Association 23/1 (1993):32-34 and unpaginated centre-page chart), but
that it rather represents a vowel that is typically rounder, fronter, and tenser than the /ə/ of
British Received Pronunciation.
3. Phonology and morphology of cliticized second person

singular interrogative forms incorporating the pronoun
The first verb, CON or CAN, is presented in greater detail. The presentation
of the remaining verbs follows the same basic pattern. Variants within round
brackets are much less frequent in occurrence.
CON/CAN "can". From a historical point of view, the dialect under discussion
is of a northern west-Midland type, so that /o/ often appears before a nasal
where other dialects of English have /a/. Even within residual traditional ver
nacular, both vowels are encountered in this lexeme. The declarative forms are
included for purposes of comparison, viz., non-past affirmative declarative: /ðε:
kon, kan/ "thou can". Non-past affirmative interrogative forms incorporating the
pronoun: /kontto, kont; 'kantte, (kant)/ "can thou?".4 The former of each inter
rogative pair here is the older—a fact borne out by observation, dialect literature,
and informants' judgments alike. The even-more-contracted form /kont/ occurs
morefrequentlythan its more residual counterpartkɔonttө/.The general import
of these remarks will apply with equal force to the other verbs, below. Non-past
negative declarative: /ðε: kɔænt, komt,(kæmt)/"thou can't". Non-past negative
interrogative: /'kɔtnt, 'katnt, (kɔənt,koint,kaeint)/"can thou not?/can't thou?".
The first two forms, which are both traditional and more frequent, are presum
ably to be derived from modal + cliticized pronoun + cliticized negative particle.
This derivation would entail a hypothetical interstage */kɔntnt, 'kantnt/, phoneti
cally "can thou not?". Final consonant clusters tend to be
glottalized, and medial clusters to be geminated or glottalized, in this dialect (see
further Shorrocks 1988). Subsequent simplification of the cluster has resulted in
the loss of the final consonant of the base.5 Preterite affirmative declarative: /ðε:
kuð/ "thou could". Preterite affirmative interrogative:
4
/tt/ represents a geminated consonant. For reasons of space, I will not debate here the appro
priateness or inappropriateness of this phonemicization, though I have suggested under could
thou? that a contrast is at stake in that particular case.
5
Trudgill (1990:87) mentions a number of interrogative forms from areas considerably to the
south of the one under discussion here:
" Traditional working-class Bristol dialect [...] has cassr't? = "can't you?" from canst
thee not?, and dissn't? = "didn't you?" from didst thee not? The Potteries area of
northern Staffordshire has ast? = "have you?" from hast thou?, and thee coost = "you
could" from couldest,"
The form cassn't? also shows loss of/n/. Trudgill (87) additionally cites asta? "have you?",
dusta? "do you?", and tha cossna "you can't" from western Derbyshire after a folksong in
Scollins & Titford (n.d.). However, he does not attempt any systematic presentation or analy
sis.
"could thou?" There is a problem of phonemicization here, and in some other

forms (below): the heavy pre-glottalization of the final /t/ in the second form,
phonetically , will, under clarity-norm conditions, be sufficient to distin
guish "could thou?" from /kut/ "cut". In allegro speech the distinction might be
blurred. I have therefore adopted the solution /(t)/. Preterite negative declarative:
/öe: kunt, 'kudnt, ('kudn)/ "thou couldn't". Preterite negative interrogative:
/kutnt/ "could thou not?" Here we can perceive the second person singular pro
noun in the devoicing of the final /d/ of the base morpheme. This is not evident
in all verbs, of course, as they do not all end with lenis stops.
MET "might". Affirmative declarative /öe: met/ "thou might". Affirmative inter
rogative: mεttө, (mε(t)t)/ "might thou?". Negative declarative: /öe: met not/
"thou might not" (there is no declarative second person singular contracted form
in my corpus). Negative interrogative: /mεtnt/ "might thou not?" Here the pro
noun is completely assimilated.
MUN "must". As far as the choice of modal auxiliary is concerned, there is no
distinction in this dialect between compulsion due to logical necessity and com
pulsion on account of moral obligation. MUN, a modal of Old Norse origin, is—
nowadays, at least—a stylistic variant (cf. MUST, below). Base form: /mun/. No
contracted affirmative second person singular interrogative forms were recorded
in my corpus. Negative declarative: /öe: meint/, and—much less usual—/öe:
munt/ "thou mustn't". Negative interrogative: /'motnt/ "must thou not?". As
with CON/CAN, we note the loss of thefinal/n/ of the base form.
MUST A stylistic variant (ef. MUN, above); though obviously more characteris
tic of speech modified towards the standard, it is nevertheless well entrenched
within the traditional vernacular too, as cliticized forms incorporating the pro
noun attest. Affirmative declarative: /öe: must/. Affirmative interrogative:
/"muste, (must)/ "must thou?". Negative declarative: /ðe: 'mus(ө)nt/. Negative
interrogative: /mus(ө)nt/ "must thou not?".
OWT "ought". Affirmative declarative: /öe: 3Yt/ "thou ought". Affirmative in
terrogative: Yttө/ "ought thou?" Negative declarative: /öe: '3Ytnt/ "thou
oughtn't". Negative interrogative: Ytnt/ "ought thou not?"
SHOULD There are no non-past interrogatives in the second person: the tradi
tional vernacular has for present purposes preterite forms only of the modal
shall Affirmative declarative: /öe:∫ud/"thou should". Affirmative interrogative:
"should thou?" Negative declarative: /öe: 'Judnt, Junt/ "thou
shouldn't". Negative interrogative "should thou
not? (shouldn't thou?)". There is a problem here with the variant /∫utntө/, which
is anomalous in terms of the general derivational patterns (cf. COULD and
WOULD, which are also historical preterites with afinal/-d/). is the regu-
lar development for "should thou not?", and presupposes an earlier */'∫udtnt/.
/'∫utntө/ might be due to informant error—possibly somewhat analogical in na
ture, or a different derivational route: /'∫udnt/ "shouldn't" + cliticized pronoun +
devoicing of/d/ under the influence of heavy glottalization of the medial conso
nant cluster. There is too little evidence to resolve this matter here.
WILL Non-past affirmative declarative: /ðε: wil/ "thou will". Non-past affirma
tive interrogative (cf. forms with a stressed pronoun: /wíl ðε/ "will thou?"; /wilt
dei/ "wilt thou?" also occurs): /"wilttө, wil(t)t/ "will/wilt thou?". Non-past nega
tive declarative: /δε: weint, weint/ "thou won't". Non-past negative interroga
tive: /'wөtnt, 'witnt/ "will/wilt thou not?". It is probable that the lateral (like the
nasal of CON/CAN and MUN, above) has been lost through simplification of the
glottalized consonant cluster. The vowel /ө/ is explicable in terms of rounding
and lowering of /I/ in the historical environment /w-(l)/. Preterite affirmative
declarative: /δει wud/ "thou would". Preterite affirmative interrogative: /'wutte,
'wudtө, wu(t)t, (wudt)/ "would thou?". Preterite negative declarative: /δει
'wudnt, wunt, ('wudn)/ "thou wouldn't". Preterite negative interrogative:
/'wutnt, 'wudnt/ "would thou not?".
BE Non-past affirmative declarative: ðε: æ(r)t, at/ "thou art". Non-past affir
mative interrogative: /attө, a(t)t/ "art thou?". In allegro speech we also meet
such highly-contracted sequences as /wiə(r)t/ "where art thou ...?"; /wot 'le
):(r)nin/ "what art thou learning?"; /eit 'fiilin/ "how art thou feeling?". Non-past
negative declarative: /ðε: æ(r)nt/ "thou aren't". Non-past negative interrogative:
/*atnt, æi(r)nt/ "art thou not?, aren't thou?". Preterite affirmative declarative: /δει
wθ(r), wθ(r)/ "thou were".6 Preterite affirmative interrogative: /"wθttθ, wθ(t)t/
"were thou?". Preterite negative declarative: /ðε: wө:(r)nt/ "thou weren't".
Preterite negative interrogative: /wө(R)nttө, 'wetnt, wθ(r)nt/. The second vari
ant has the morphological derivation "were thou not?", and the first and third
"weren't thou?". Whilst the former is the predominant pattern in the dialect, the
latter is probably more frequent in this particular tense of this particular verb—a
state of affairs that may well relate to the fact that /'wetnt/ is homophonous with
the commoner variant of "will thou not?" The homophony is unsurprising, as
historical /r/ is optionally retained after /ө/, but has been lost without exception
in the environment /θ/ + fortis consonant.
HAVE Non-past affirmative declarative: /ðε: az/ "thou has". Non-past affirma
tive interrogative: /aste, ast/ "has thou?", with the latter often /as/ before a con
sonant. Though the third form is highly contracted, the pronoun remains de
tectable in the fortis /s/ resulting from an earlier assimilation. In allegro speech
we meet such highly-contracted forms as /wots bakt/ "what has thou backed?"
Preterite forms in this dialect have been regularized to were throughout the paradigm.
[horse racing]; /wɔəss dun/ "what has thou done?". Non-past negative declara
tive: /ðει 'aznt/ "thou hasn't". Non-past negative interrogative: /as(ə)nt/ "has
thou not". Again, the pronoun is perceptible in the devoicing of a lenis conso
nant, the /z/ of /az/ "has". Preterite affirmative declarative: /δει ad/ "thou had".
Preterite affirmative interrogative: /'atto, 'adtө, a(t)t, (adt)/ "had thou?". Note
how the first and third variants are homophonous with the forms of the non-past
affirmative interrogative of BE! Preterite negative declarative: /ðει 'adnt/ "thou
hadn't". Preterite negative interrogative: /'adnt, 'atnt/. Note how the latter, "had
thou not?", though the regular development, is homophonous with the chief
variant of the non-past negative interrogative of BE. This fact may explain why it
seems to occur less frequently than /adnt/ "hadn't thou?/had thou not?".
DARE This verb is sometimes a modal auxiliary. It is a generally difficult verb,
at least from the standpoint of the analyst, with preterite and non-past forms not
clearly distinguished. I have recorded: /dust, dest/ "does thou dare?" The loss of
historical /r/ in these particular phonological environments has resulted in a form
/dust/ that is homophonous with a variant of the non-past affirmative interroga
tive of DO.
DO Non-past affirmative declarative: /δει duz/ "thou does". Non-past affirma
tive interrogative: /dustө, dust/ "does thou?". The final /t/ of the latter is quite
often elided before a consonant, though again the pronoun is still perceptible in
the devoicing of the fricative that has resulted from an earlier assimilation. In al
legro speech we meet such highly-contracted forms as /wots mi:n/ "what does
thou mean?". Non-past negative declarative: /ðε: 'duznt/ "thou doesn't". Non-
past negative interrogative: /'dus(ө)nt/ "does thou not?". Preterite affirmative
declarative: /δει did/ "thou did". Preterite affirmative interrogative: /dittө, dI(t)t,
(didt)/ "did thou?". Preterite negative declarative: /ðει 'didnt/ "thou didn't".
Preterite negative interrogative: ditnt, 'didnt/ "did thou not?, (didn't thou?)".
4. Distribution of cliticized interrogative forms
It is important to note that at least three factors condition the distribution of

these contracted forms, (i) In affirmative interrogative sentences, the choice be
tween enclitic and full forms of the pronoun relates to considerations of stress,
whereas in negative interrogative sentences (i.e., not tags), it does not: the di
alect, like many others in the North, inclines very strongly and across the
paradigm towards the construction verb + subject + noan/not, as in Con Ah
noan 'a' one? "Can I not have one?", as opposed to English English standard
Can t I have one?1 The same preference carries over into 'modified' speech that
is in many, or even most, other respects relatively standardized. Thus, con
tracted negative interrogative forms are mostly found in tags. (ii) Only the
anomalous finite BE has these second person interrogative enclitic forms when
functioning as a full (lexical) verb; otherwise the forms are restricted to auxil
iaries. HAVE is followed by getten "got" (past participle of get) when possession
is indicated affirmatively; and in negative questions the dialect uses HAVE
(usually + getten) + noanlnone rather than the usual English English standard
form HAVE + contracted negative + any. (iii) In tags after negative propositions,
affirmative contracted forms do not necessarily occur in this dialect. This is so
important a contrast with other varieties of English that it warrants further com
ment.8
Lester (1971:161) is representative of grammars of standard varieties of
English generally in furnishing the following examples of question tags:
(1) It is hot isn't it?

(2) It isn't hot is it?
(3) It rained didn't it?
(4) It didn't rain didit?
7
Trudgill (1984:33) notes:
Speakers of Standard English English in the south of England tend to use, in their
speech, contracted negatives of the type:
I haven't done it
I won't do it
In the north of England, the alternative contraction is, in some areas, more common:
I've not done it
I' ll not do it.
Trudgill & Hannah (1994:97) have drawn attention to the same tendency in Scottish English:
There is a tendency not to contract the negative element not in Scot Eng, especially
in yes-no questions. If an auxiliary is present in a negated sentence, the auxiliary
usually contracts.
ScotEng EngEng
Is he not coming? Isn't he coming?
Did you not see it? Didn' t you see it?
He' 11 not go He won't go
You've not seen it You haven't seen it.
Shorrocks (1981:595; and in press, and the references cited there).

and then commenting: "In short, given the main sentence we can always predict
exactly what the tag will be." (164).9 My impression is that linguists and dialec-
tologists alike assume that other varieties are in accord with the standard as far as
the syntactic subsystem of interrogative tags is concerned. In the dialect of the
Bolton Metropolitan Area, however, this is certainly not the case. There are two
types of negative tag, the one of which obligatorily follows an affirmative
proposition and the other of which optionally follows a negative proposition. In
the Bolton Metropolitan Area equivalents of sentences (1) and (3), above, a
negative tag is used, the tag taking the form:
(1) It's 'ot /ιnt, *ιznt, ιn/ ŋ /ιt/ "isn't it?"

(3') It rained /dιnt, 'didnt, dιn/ ŋ /ιt/ "didn't it?"10
i.e., a contracted negative form is used in the tag. In (2) and (4), we get:
(2') It's not'ot ιz ιt/ "isit?"

(4') It didn't rain /ιdd ιt/ "did it?"
i.e., an affirmative tag, as in standard English; or, at least as likely, and with just
the same propositional content:
(2') It's not'ot /ιz ιnɔt/ "is it not?"

(4') It didn't rain /did It nɔt/ "did it not?"11
This latter type of tag is pronounced with a 3 1 i intonation contour.12 More
importantly, we should note that the negative tag that optionally follows a nega-
9
There are, of course, exceptions to this pattern, as in It's hot, is it?, where the statement
preceding the tag is essentially a quotation.
10
Whether the deletion of the /t/ of the cliticized negative is due to assimilation, or is part of
a wider pattern of final-/t/ deletion in English, is uncertain. See Britton (1992:39) on this
point.
11
The dialect makes extensive use of multiple negation or negative concord. See Shorrocks
(1981:631-633, § 6.8.7.3; 1985). In some contexts, the multiple negation is perhaps cumula-
tively emphatic in function, as researchers in English dialectology and the history of the
English language have often asserted; on other occasions, though, it seems to me, at least,
more a question of agreement. I will not labour what is clearly a moot point here, but merely
note that what is going on in these tags is probably part of a wider pattern of multiple nega-
tion or negative concord.
12 Contrast the rising terminal in standard It's hot, is it not?—a pattern that I take to be im-
possible in the dialect, where a negative tag must involve a contraction after an affirmative
proposition. However, if such a tag were produced, say, in quoting direct speech or in irony,
tive proposition must take the form verb + subject pronoun + /not/ "not". Thus
it is kepi entirely distinct from the contracted negative tags that obligatorily fol
low affirmative propositions. For the Bolton Metropolitan Area dialect, then, we
cannot "predict exactly what the tag will be" after a negative proposition, but
rather that it will take one of two forms.
5. Historical background
In the Middle English period, afirmative interrogatives with cliticized forms

of the second person singular pronoun are well-attested. Wright & Wright
(1923:154, §372) note:
Nom. θu (θu) beside the unaccented form θu, which became tou (tu)
when attached enclitically to a verb, as hastou, -tǔ hast thou, wiltou, -tǔ wilt
thou. This form with t- has also been regularly preserved in interrogative and
subordinate sentences in many of the modem dialects.13
Similarly, Mossé (1952:40, § 47.1) observes under the heading
"Assimilation": "Total or partial sandhi assimilation of dentals in unaccented
words: [...] art өu > artu, wilt өu > wiltu"; and later (55, § 64):
III. Second person, singular.—Alongside Ou,  there was an enclitic
form tu, töw (and even te) often attached to the verb (cp. § 47.1), for
it would be distinguished by its intonation contour from the negative tag that optionally fol
lows a negative proposition.
13
I have changed the thorn symbol of the original to [Θ]. Use of reduced /tə/ in affirmative in
terrogatives is noted in various parts of the North, as in Wright's Windhill (West Riding of
Yorkshire) monograph (1892) and Brilioth's Lorton (Cumberland) study [1913]. Brilioth com
ments as follows,rightlydrawing attention to the possibility of partial assimilation:
§ 408. 2nd Person. The weak form to of the 2nd pers. sing, is only used interroga
tively in unaccented position; the t of this form has probably arisen through assimi
lation with the t-ending in verb forms like āt art, aut ought, dušt durst, and partial
asimilation with the s-ending of the 3rd pers. sing., which ending is often extended
to all three persons of the pres. sing. Professor Wright (A Grammar of the Dialect of
Windhill, p. 118) ascribes this transition of θ into t of the form ta, tö solely to the
unaccented position of the pronoun, basing his theory on the fact that there are only
three verbal forms ending in t (at art, out oughtest, d s t durst), but he has not
taken into consideration that partial assimilation with the s-ending of the 2nd pers.
sing, may also have been at work: the effort required in pronouncing the sibilant fol
lowed by the dental spirant in combinations like wants-өə, dis- is undoubtedly
much greater than in the case of s + the dental stop in wants-t, dis-tô. (104-
105, with the original thorn symbol changed to [Θ].)
example artow "art thou", has tow "hast thou" (21/311), wiltow, woltou
"wilt thou", wenest tǔ "thinkest thou" (3/25), wilte "wilt thou" (12/528).14
Jordan (1974:149-150, § 151) comments specifically on the weakening of
the vowel of the pronoun: "With enclisis  appears weakened in seiste, woste
"do you say, know" RG, wilte Hav. 528, 1135, thenkeste Hav. 578 (Klaeber
1926:321)."
We may note a not dissimilar development in the history of the German lan
guage during the Middle High German period, with enclitic formations evident
in second person singular interrogatives under conditions of weak stress, as
similation, and a blending of verb and second person singular pronoun that was
very likely a contributing factor in the re-analysis of the verb inflection from -s
to -st:
Das von der Wortfügung der natürlichen Rede her häufig dem Verbum
nachgestellte Personalpronomen pflegen die Schreiber angesichts seiner
satzrhythmischen Untertonigkeit gern enklitisch dem Verbum anzuhängen:
vgl. bistu oder in »Christus und die Samariterin« 27 mahttu; nimistu. In
solchen Verschmelzungsformen ist du wegen der assimilatorischen
Angleichung seines Anlauts an den voraufgehenden stimmlosen
Konsonanten nicht mehr ohne weiteres erkennbar; so begreift sich, daß es in
falscher Formauflösung als zum Verbum gehörig aufgefaßt wurde.(Tschirch
1989:30)
Early Modern English has affirmative interrogative forms that are clearly di
rect descendants of the Middle English forms cited above. For instance, in
Hamlet, Act 5, Scene 1, we find the lines: "Woot weep? Woo't fight? Woo't
fast? Woo't tear thyself? | Woo't drink up eisel, eat a crocodile?" (Shakespeare
[1600]:265-266). Some versions of the play follow up these lines with "Dost
come here to whine?" The editor's Note to line 265 in the edition cited here
reads: "Woo't wilt (a colloquial form of the second person singular of will (OED
v.l Α 3δ)). The form seems to have been associated with challenges and the like
in Shakespeare's mind. Compare 2 Henry IV 2.1.54-5; Anthony 4.2.7." These
remarks agree with the interpretation in Onions (1986:322).
Cliticizaüon of the negative particle is a later development. Britton (1992:38)
observes that the negative enclitic /nt/ is a contraction of the negative particle
under reduced stress which was contemporaneous with the process of cliticisa-
tion and which was first recorded in English in the second half of the seven
teenth century. Subsequent phonological changes affecting some of the cliti-
cised verb forms in Standard English have made the /nt/ redundant in certain
instances as a marker of negation because in a number of forms the contrast
I have changed the thorn symbol of the original to [Θ].

between negative and non-negative is adequately made by root-modification

in the cliticised negative.
Such historical observations help to explain why the majority of the negative
second person singular interrogatives in the dialect of the Bolton Metropolitan
Area follow the pattern verb + cliticized pronoun + cliticized negative.
6. Earlier scholarship
In a very early essay on the dialect of south Lancashire, Picton (1865:43)

drew attention to cliticized affirmative interrogatives in this region and their early
literary antecedents:
Hearsto, hearest thou.

Seesto, seest thou.
This form of expression is found in our old writers—
"Sone, slepestow?
Sestow this people?"— Piers Ploughman s Vision, 468.15
The standard glossary of the Lancashire dialects is Nodal & Milner (1875-
1882). Here I note only:16 arto? "art thou?"; artono?, artrito? "art thou not?";
asto? "hast thou?"—again with supporting quotations from early literature. The
first of the more localized glossaries is Cunliffe (1886), dealing with Rochdale
and Rossendale to the east and north-east of Bolton, and again only a few, af
firmative forms are recorded: arto? "art thou?"; hasto? "hast thou?"; wilto "wilt
thou?"; and the adverb wilto-sholto "whether or not" [the gloss itself is regional;
"whether orno"].
Ellis (1889) contains only one second person interrogative in the
Comparative Specimen, and none in the Dialect Test. The style of the relevant
sentence is formal, and no instructions are given about number or stress. There
are four renderings of the Comparative Specimen for the Variety that includes
what is now the Bolton Metropolitan Area, and the plural pronoun is used in
three of these. (See further p. 334 [1,766].)
In an early south Lancashire glossary (1901), dealing with the area between
and to the east of Bolton and Manchester, Taylor gave a "Conjugation of the
Verb, to have" (n.p.), which includes the following second person singular in-
Langland composed the A Text oí Piers Ploughman in [1362/1363].

One has to look for these forms under any dialectal spellings of which one can conceive.
BOLTON VERNACULAR: SECOND PERSON SING. INTERROGATTIVE 181
terrogatives incorporating the pronoun:17 non-past affirmative Hast?, Hasto?;

non-past negative Hastno'?; preterite affirmative Hadst?, Hadsto?; preterite
negative Hadstno' ? The inclusion of the future 'tense' incidentally yields these
forms of WILL: non-past affirmative Wilt ha'?, Wilto ha'?; non-past negative
Wiltno' ha' ? Looking in the glossary itself under conceivable dialectal spellings,
I note additionally: art'?, arto?, arta? "art thou?"; art'no?, artunt? "art thou not?";
asto?, asta? "hast thou?"; astno?, astna? "hast thou not?"; did't?, didto? "didst
thou?"; const?, con't?, con to? "can'st [sic] thou?"; conino' "can'st thou not?";
shal't?, shal'to? "shalt thou?"; wilt'?, wilta? "wilt thou?" Although Taylor's
glossary is not especially scholarly, and is not usually consulted by scholars, it
does in fact contain a number of entries that show the considerable contraction
that some second person singular interrogative forms were undergoing.
Wright, in his English Dialect Grammar (1905:271-272, § 404 (a)), ob
serves in re the nominative form of the second person singular pronoun in
English dialects: "But in the northern and north-midland counties the ő has
gen[erally] become t in interrogative and subordinate sentences."18 And further:
"The unstressed form is gen[erally] . But in the northern and north-midland
counties the unstressed form is tə in interrogative and subordinate sentences"
(272). These comments suggest that the cliticized forms of the type under dis
cussion here are regionally restricted—a fact that would be in keeping with the
conservative tendencies of the Northern and north-Midland dialects. Indeed, the
pronoun thou itself is regionally restricted nowadays.
However, Wright did not attempt any systematic presentation or further dis
cussion of these contracted forms in the English Dialect Grammar (1905);
though some individual forms are listed in his English Dialect Dictionary (1898-
1905), for dialect literature informally captures a number of these interrogatives,
and a few forms are recorded in older glossaries. The lack of further discussion
in the English Dialect Grammar is not surprising, as the grammar of the day was
chiefly morphological and word-bound. In the English Dialect Dictionary, I find
no trace at all anywhere in the United Kingdom of certain forms such as /'katnt,
'kotnt, 'wotnt, 'witnt/. A number of our Bolton forms would appear to be
unique, or at any rate previously unrecorded. Narrowing the search for second
17
In common with many dialect writers and glossarists, Taylor spells have and other words
with the h of the standard, even though /h/ is not a phoneme of the dialects of south
Lancashire. His use of the apostrophe is inconsistent, as will be evident from a comparison of
the forms cited.
18
Strictly it is historical /ө/ that became /t/, /ð developed in atonic syllables in the fourteenth
century after the sandhi development whereby /Θ/ became /t/. See further Jordan (1976:187, §
207).
person singular interrogative forms incorporating the pronoun to Lancashire, I

find nothing at all under the headwords DARE, MAY/MIGHT, MUN, NEED,
OUGHT, SHALL/SHOULD, WOULD. Under HAVE we find: Lan[cashire]19 hasta?,
hasto?, asto?' wots to? "what have you?", ., s.e. Lanfcashire] hast' ?, hasto?;
s.e. Lanfcashire] hastno?, hastn't?, hasn't?; s.e. Lanfcashire] hadst?, hadsto?;
hadstno'? Under DO: Lanfcashire] dost?, dusta? "does thou?"; dustno? "does
thou not?"; didto? "did thou?". Under BE we find: Lanfcashire] and e.
Lanfcashire] arto?, s. Lanfcashire] art to?, art?; s. Lanfcashire] (are not theaw?),
art no?, artn't?; Lanfcashire] wurt?, s. Lanfcashire] wert to?; s. Lanfcashire]
wert no?; wern't to? Under CAN we find just Lanfcashire] con ta? And under
WILL: Lanfcashire] wilt'na?, s. Lanfcashire] wiltno?; (and Lanfcashire] winnot
tho?). Wright does not offer any phonetic transcriptions for these particular in
terrogative forms that are recorded in the dictionary. It is worth noting, how
ever, that the majority of Wright's negative interrogatives follow the pattern verb
+ cliticized pronoun + cliticized negative particle.
An early grammar of the School of Wright is that by Hargreaves on
Adlington (1904), which is no great distance from Bolton. He notes (p. 80, §
91) that 'tə is only used after the auxiliary verb, es tə fon id? hast thou found
it?" (See further § 118, p. 100 and p. 101.) A few similar forms are also
recorded, as kɔn to? "can thou?" (p. 96), ar ðă and or to? "are thou?" (p. 103).
The enclitic forms are not systematically recorded for all the relevant verbs; the
ones that are recorded do not show the same degree of contraction as our Bolton
forms.
The other early south Lancashire grammar is that by Schilling on Oldham
dialect (1906). He gives " wiər ărt(o) guín? = Where art thou going (to)? [The
form ta, -to is only enclitic.]"; kon to? ["can thou?'!; kontno? ["can thou not?"];
Salto? ["shall thou?"]; osto? = hast thou?; astno? ["has thou not?"]; arto? ["are
thou?"]; artno? ["are thou not?"]; wilta, wiltə? ["will thou?"]; wiltna? ["will
thou not?"]; dust(o)? ["does thou?"]; өiŋksta? - dost thou think?20 The number
19
This is a terribly vague designation, since pre-1974 Lancashire was geographically exten
sive, had a big population, and the boundary between the Northern and north-Midland dialects
ran right through it.
20
Schilling's Grammar is generally thought to be of limited reliability. The forms cited here
are from the following pages: § 108, p. 113; § 124, p. 123; p. 124; p. 125; p. 126; p. 127; p.
129; p. 130. I have added some glosses, since Schilling's paradigmatic layout has been lost
here; also the diacritic to the shall-förm, which otherwise seems anomalous; and have omitted a
negative interrogative form of DO, where there seems to be a fairly obvious misprint. I have
replaced the thorn symbol of the original by [ө]. Schilling says that the enclitic interrogative is
preferred in Oldham (p. 123, under CAN), and draws attention in a note to the contracted second
and third person singular negative interrogative forms of CAN.
of contracted forms recorded here is greater than in Hargreaves' monograph,

and the possibility of dropping the schwa of the pronoun with at least BE and DO
is reminiscent of the Bolton forms. Otherwise these forms are not so contracted
as the present-day Bolton forms—and the negative particle does not end in /t/.
A rather later study of a pre-1974 south Lancashire dialect is Hunt's (1959)
Heywood Grammar. Hunt gives [ti, te] as enclitic forms of the second person
singular pronoun that have arisen through assimilation, and that occur only in
interrogatives (169). In the historical phonology (140, § V.18), the following
examples are to be found: ['a:ti] "art thou?"; ['dosti] "dost thou?"; ['asti] "hast
thou?"; ['wiltι] "wilt thou?". Unfortunately, no additional forms are given in the
grammatical section in the discussion of 'minor groups'. It is therefore uncertain
whether there are more contracted forms still, and in particular what the negative
interrogative forms are like.
The Survey of English Dialects (SED) Basic Material contributes disappoint
ingly little to our understanding of this area of syntax. First, there is a general
lack of appropriate questions, so that the syntax of interrogative forms is poorly
covered. Only SED questions IX.5.4 do you?, did you?, IX.7.2 are you?,
IX.7.3 aren t you?, IX.7.5 aren t you? and IX.7.6 weren't you? are designed to
sample this area of syntax. There is therefore no prospect of eliciting forms such
as /"kotnt/ "can thou not?". Second, there are cases of unanswered questions, ir
regular responses, pressured responses, etc., all of which remind us of the diffi
culties of eliciting syntax by questionnaire techniques generally, and more par
ticularly of the difficulty of the SED's 'converting questions'. Third, there are
critical failures to record relevant information: for instance, the pronouns were
not recorded at all at a few localities; yet it is essential that we know in every
case whether the pronoun has been accounted for, since we are dealing with
such highly contracted forms! For instance, /a:(r)t/ could in theory be simply a
verb art or verb + pronoun combination are/art thou? Fourth, there is a quite
alarming failure to control for stress. See, for example, Note 2 to question
IX.5.4 in the Northern Region responses: "It is clear that the f[ield] w[orker]
often recorded the stressed forms of the pronouns" (Orton & Halliday 1962-
1963). Cliticized forms of pronouns occur under conditions of weak stress, of
course. This means that we do not have comparable material from all localities.
And fifth, there is also a disastrous failure to control strictly for number! These
considerations leave us singularly short of clearcut cases of contracted second
person singular interrogatives incorporating unstressed pronouns, and make ge
ographical comparison difficult, if not ultimately impossible.
7. Conclusions
This study has shown that there are second person singular interrogative
forms in the dialect of the Bolton Metropolitan Area that are highly distinctive
both phonologically and morphologically in comparison to interrogative forms in
most other varieties of English spoken today. Extremely contracted enclitic
forms have developed through the processes of reduction, assimilation and
simplification under conditions of weak stress. The contractions have sometimes
been of such an order that certain second person singular interrogative forms of
different verbs are now homophonous, even within such a small and sharply
delimited subset of forms. This homophony may account for an apparent prefer
ence for /'adnt, 'wө:(r)nttө/ over more usual variants. Negative forms are more
often than not of the type verb + cliticized pronoun + cliticized negative particle.
This is because the cliticized forms of the singular pronoun developed in Middle
English times, whereas cliticization of the negative particle was a later develop
ment. There are also important syntactic differences between this dialect and the
standard in the preference for the construction verb + pronoun + negative parti
cle rather than verb + cliticized negative particle + pronoun; and in the existence
of two types of negative tag, the one of which obligatorily follows an affirmative
statement and the other of which optionally occurs after negative statements. Our
review of the previous scholarly literature reveals that certain forms such as
/'katnt, 'kɔtnt, 'mөtnt, 'wөtnt, 'witnt/ have not been recorded before, even im-
pressionistically, and that there has been no systematic treatment of second per
son singular interrogative forms.
The syntactic subsystem treated here has implications for our methodology
and fieldwork: at the linguistic level of syntax, many phenomena cannot be reli
ably elicited by questionnaire techniques. Whole sentences, themselves ques
tions rather than responses, are needed to elicit the interrogative patterns.
Furthermore, the syntactic subsystem described here manifests itself typically in
conversation rather than narrative, so that a free exchange of views between in
formant and fieldworker or between two (or perhaps more) informants should
be encouraged. These interrogative forms emerge fully only if the informant is
free to question the fieldworker (or another informant), and to do this in a rela
tively informal atmosphere—otherwise the plural form or the standard pronoun
might be used as a polite singular. Interviews in which thefieldworkerasks all
the questions, and the informant simply provides answers, are inappropriate
where the interrogative is concerned. Whilst dialectologists have generally rec
ognized that questionnaires do not elicit syntax effectively, they have been slow
to advocate the methodological corollary: namely, that monographs be written
based upon prolongedfieldworkin a locality and a large corpus of tape-recorded
conversation. Some, perhaps influenced by the monographs of the School of

Wright, have obviously felt that there was little to describe at the syntactic level;
at any rate, most have preferred to carry out work within the dominant
paradigms of diachronic phonology, synchronic phonology, dialect geography,
and sociolinguistics, and have neglected the syntax of non-standard speech. The
work that I have carried out in Greater Manchester County shows that the estab
lished view of grammatical variation among English dialects is incorrect. The
dialect of the Bolton Metropolitan Area differs appreciably from standard
English at the grammatical level—and there is no reason at all why that particular
dialect should be unique in this respect. Rather, it is the theoretical preconcep
tions and associated methods of dialectologists that have led to the misconcep
tion that English dialects do not vary much at the grammatical level, and to a
concomitant lack of work at that level.
REFERENCES
Brilioth, Börje. [1913.] A Grammar of the Dialect of Lorton (Cumberland).
Historical and Descriptive, with an Appendix on the Scandinavian Element and
Dialect Specimens and a Glossary. Publications of the Philological Soc. 1.
London: Oxford Univ. Press.
Britton, Derek. 1992. "Secondary Contraction and Deletion of Negative Enclitics
in English Dialects". Zeitschrift für Dialektologie und Linguistik 59,1.38-49.
Cunliffe, Henry, comp. 1886. A Glossary of Rochdale-with-Rossendale Words and
Phrases. Manchester & London: John Heywood.
Ellis, Alexander J[ohn]. 1869-1889.On Early English Pronunciation, with
Especial Reference to Shakspere and Chaucer, Containing an Investigation of
the Correspondence of Writing with Speech in England, from the Anglosaxon
Period to. the Existing Received and Dialectal Forms, with a Systematic Notation
of Spoken Sounds by Means of the Ordinary Printing Types. Including a Re-
Arrangement of Prof. F.J. Child's Memoirs on the Language of Chaucer and
Gower, Reprints of the Rare Tracts by Salesbury on English, 1547, and Welsh,
1567, and by Barcley on French, 1521, Abstracts of Schmeller's Treatise on
Bavarian Dialects, and Winkler's Low German and Friesian Dialecticon, and
Prince L.L. Bonaparte's Vowel and Consonant Lists. 5 vols. Publications of the
Chaucer Society. 2nd Ser. 1, 4, 5, 11, 25. Publications of the Early English Text
Society. Extra Ser. 2, 7, 14, 23, 56. And without numbers in the publications of
the Philological Society. London: 1869-1889. Pag. cont; vol. 5 also has sepa
rate pag. Title: Parts I-III, title stops at Barcley on French, 1521. Imprint: Parts I-
IV: Asher & Co. for the Philological Society; Trübner & Co. for the Chaucer
Society and the Early English Text Society. Part V: Trübner & Co. for the
Philological Society, the Early English Text Society, and the Chaucer Society,
1889. Part V. Existing Dialectal as Compared with West Saxon Pronunciation.
With Two Maps of the Dialect Districts. [The subtitle is used much more fre
quently: The Existing Phonology of English Dialects Compared with That of
West Saxon Speech. Forming Part V. of Early English Pronunciation.] Repr.
New York: Greenwood Press, 1968.
Frangopulo, N[icholas] J[oseph]. 1977. Tradition in Action. The Historical
Evolution of Greater Manchester County. East Ardsley (Wakefield): EP
Publishing Limited.
Freeman, T[homas] W[alter], H.B. Rogers & R.H. Kinvig. [1966.] Lancashire,
Cheshire and the Isle of Man. Regions of the British Isles. [London: Thomas]
Nelson [& Sons Ltd.].
Hargreaves, Alexander. 1904. A Grammar of the Dialect of Adlington
(Lancashire). Anglistische Forschungen 13. Heidelberg: Carl Winter. Repr.
Amsterdam: Swets & Zeitlinger NV, 1967.
Hunt, Jean. [Listed in some bibliographies under her married name, Jean Wade.]
1959. A Grammar of the Dialect of Heywood, Lancashire. MA thesis, Dept. of
English, Univ. of Leeds. Publ, in microfiche, East Ardsley (Wakefield): Micro
Methods Limited (later SR Publishers Ltd./EP Microform).
International Phonetic Association's Alphabet. 1993 revision. Journal of the
International Phonetic Association 23,1.32-34 and unpaginated centre-page
chart.
Jordan, Richard. 1974. Handbook of Middle English Grammar: Phonology.
Transl. & rev. Eugene Joseph Crook. Janua Linguarum. Series Practica 218. The
Hague & Paris: Mouton.
Lester, Mark. 1971. Introductory Transformational Grammar of English. New
York: Holt, Rinehart & Winston.
Mossé, Fernand. 1952. A Handbook of Middle English. Transl. James A. Walker.
Baltimore & London: The Johns Hopkins Univ. Press. (9th printing, 1987.)
Nodal, John H[oward] & George Milner. 1875-1882. A Glossary of the
Lancashire Dialect. Publications of the Manchester Literary Club. Publ, in 2
Parts. A Glossary of the Lancashire Dialect. Part I, A-Ε. Publications of the
Manchester Literary Club. Manchester: Alexander Ireland & Co.; London:
Trübner & Co. for the Manchester Literary Club, 1875. Part II, F-Z.
Manchester: Abel Heywood & Son for the Manchester Literary Club, 1882.
Repr. in one vol. in a limited ed. of 100 copies. Bath: Olivers, 1972. Also publ,
as Part /.— A to E. Part IL— F to Ζ. Publications of the English Dialect Society
10 & 35. Series  Original Glossaries. Vol. 14. London: Triibner & Co. for the
English Dialect Society, 1875-1882. Repr. Vaduz: Kraus Reprint Ltd., 1965.
Onions, C.T. A Shakespeare Glossary. Enl. & rev. Robert D. Eagleson. Oxford:
Clarendon Press, 1986.
Orton, Harold & Wilfrid J. Halliday, eds. 1962-1963. Survey of English Dialects.
(B). The Basic Material. Volume I. The Six Northern Counties and the Isle of
Man. 3 Parts. Leeds: E.J. Arnold & Son Limited for the Univ. of Leeds.
Picton, J.A. 1865. "The South Lancashire Dialect". Proceedings of the Literary
and Philosophical Society of Liverpool 54 (1864-1865):17-56. Extracted by
permission & repr. as Notes on the South Lancashire Dialect. Liverpool:
Privately printed.
Saxelby, C[harles] H[arold], ed. 1971. Bolton Survey. General ed. C.H. Saxelby.
Subject eds.: Geography, Alice Dyson; History, W,E. Brown; Natural History and
Pre-History, A. Hazlewood. Pub, under the auspices of the Bolton branches of
the Geographical and Historical Associations and Bolton Field Naturalists
Society. [Bolton: Printed by Tillotsons (Bolton) Limited, Bolton & London, for
the Bolton Survey Committee,] 1953. Repr. with a new introduction by W.E.
Brown. County History Reprints. Wakefield: SR Publishers Ltd.
Schilling, Karl G. 1906. A Grammar of the Dialect of Oldham (Lancashire).
Doctoral diss. Univ. of Gießen, 1906. Publ. Darmstadt: G. Otto's Hof»
Buchdruckerei.
Scollins, R. & J. Titford. N.d. Ey Up, Mi Duck. Notes to record produced by Ram
Productions, Ilkeston, Derbyshire.
Shakespeare, William. [1600.] Hamlet. Ed. G.R. Hibbard. The Oxford
Shakespeare. Oxford & New York: Oxford Univ. Press, 1987.
Shorrocks, Graham. 1981. A Grammar of the Dialect of Farnworth and District
(Greater Manchester County, Formerly Lancashire). 2 vols. Ph.D. diss. Univ. of
Sheffield, 1980-1981. Ann Arbor: University Microfilms international; London:
Information Publications. 81-70, 023.
___. 1985. "Aspects of Affirmation and Negation in the Dialect of Farnworth
and District (Greater Manchester County, Formerly Lancashire)". Journal of the
Lancashire Dialect Society 34.20-29.
___. 1988. "Glottalization and Gemination in an English Urban Dialect". The
Canadian Journal of Linguistics/La Revue canadienne de Linguistique 33,1.59-
64.
— _ . In press. 'Fieid Methods and Non-standard Grammar". Methods VIII.
Proceedings of the International Conference on Dialectology ¡Colloque interna
tional de dialectologie, 2-7 August, 1993, University of Victoria, B.C. Ed. James
Arthurs. [With a summary in French.]
Smith, David M. 1969. The North West. Industrial Britain. Newton Abbot: David &
Charles.
Survey of English Dialects. See under Orton & Halliday, eds. (1962-1963).
Taylor, Francis Edward, comp. 1901. The Folk-Speech of South Lancashire: A
Glossary of Words Which Are, or have Been During the Last Hundred Years, in
Common Use in That Portion of the County Palatine Situate Between Bolton
and Manchester, Including Dialect Words, Children's Words, Local
Mispronunciations, Colloquialisms, and Local Slang, with an Appendix of
Quaint Sayings. Manchester & London: John Heywood.
Tschirch, Fritz. 1989. Geschichte der deutschen Sprache. Zweiter Teil:
Entwicklung und Wandlungen der deutschen Sprachgestalt vom Hochmittelalter
bis zur Gegenwart. Grundlagen der Germanistik 9. 3te, ergänzte und überar
beitete Auflage bearbeitet von Werner Besch. Berlin: Erich Schmidt Verlag.
Trudgill, P[eter]. 1984. "Standard English in England". Language in the British
Isles,ed. by Peter Trudgill, 32-44. Cambridge: Cambridge Univ. Press.
—-. 1990. The Dialects of England. Oxford, & Cambridge, Ma.: Basil
Blackwell.
_—— & Jean Hannah. 1994. International English. A Guide to Varieties of

Standard English. 3rd ed. London, etc.: Edward Arnold.
Wright, Joseph. 1892. A Grammar of the Dialect of Windhill, in the West Riding of
Yorkshire. Illustrated by a Series of Dialect Specimens, Phonetically Rendered;
with a Glossarial Index of the Words Used in the Grammar and Specimens.
Publication of the English Dialect Society 67. London: Kegan, Paul, Trench,
Trübner & Co. for the English Dialect Society.
___. ed. 1898-1905. The English Dialect Dictionary. Being the Complete
Vocabulary of All Dialect Words Still in Use, or Known to have Been in Use
During the Last Two Hundred Years. Founded on the Publications of the
English Dialect Society and on a Large Amount of Material Never Before
Printed. 6 vols, London & Oxford: Henry Frowde, Publisher to the English
Dialect Society; New York: G.P. Putnam's Sons. Repr. Oxford Univ. Press,
1961. [Vol. 6 contains The English Dialect Grammar, which was also publ, sepa
rately. See below.]
__. 1905. The English Dialect Grammar. Comprising the Dialects of
England, of the Shetland and Orkney Islands, and of Those Parts of Scotland,
Ireland and Wales Where English is Habitually Spoken. Oxford, London,
Edinburgh, Glasgow, New York & Toronto: Henry Frowde. Repr. Norwich:
Fletcher, 1970.
——— & Elizabeth Mary Wright. 1923. An Elementary Middle English
Grammar. London, etc.: Humphrey Milford, Oxford Univ. Press.
REFLEXIVES PRONOUNS AND SUBJECT/VERB
AGREEMENT IN ICELANDIC AND FAROESE*
KNUTTARALD TARALDSEN
University of Troms0
1. Introduction
In this article, I will study two seemingly unrelated topics in Icelandic and
Faroese syntax—the properties of subject /verb agreement and the distribution of
reflexives and pronouns in the two languages. It will turn out, however, that the
analysis I propose to explain certain properties of subject/verb agreement easily
extends to provide an account of the distributional constraints on reflexives and
pronouns, whence I will ultimately draw the conclusion that the relevant proper
ties in the two domains actually reflect the same set of syntactic principles. These
principles include a general licensing condition on empty arguments and an
economy principle preferring unspecified feature values over specified ones.
I will first present an analysis of subject/ verb agreement and extend this
analysis to object/verb agreement. This analysis is subsequently shown to ex
plain an otherwise unexpected restriction on reflexives in Icelandic. The notion
of Economy of Representation embodied in the analysis is then shown to predict
the complementary distribution of reflexives and non-reflexive pronouns.
Finally, I show that this approach leads to a unified analysis of a set of contrasts
between Icelandic and Faroese.
I am grateful to Anders Holmberg and an anonymous reviewer for comments on an earlier

version of this paper, and to Margrét Jónsdóttir and Höskuldur Práinsson for judgments on
some of the Icelandic examples. Usual disclaimers apply.
190 KNUT TARALD TARALDSEN
2. Subject /verb agreement
I will begin by considering the question why certain DPs move to SPEC-po-
sitions. In particular, I will use certain properties of subject /verb agreement in
Icelandic and Faroese to argue against the currently fashionable view that
movement to SPEC-positions is in general driven by feature checking. Instead, I
will suggest that such movement more plausibly reflects the way predication is
realized in syntactic form. The appearance of agreement features on the verb will
be taken to be forced by movement rather than the other way around. This as
sumption will be the key to my analysis of subject/verb agreement, and even
tually also of object/verb agreement, in Icelandic and Faroese.
2.1. Feature checking and movement to SPEC
Chomsky (1993) proposes that a DP moves to some SPEC-XP just in case

either the DP bears a Case-feature which must be licensed by being checked off
against a corresponding feature of X, or X has person / number features that can
only be licensed by being matched against features of SPECXP. On this view, a
nominative subject DP, for instance, may move to SPEC-AGRSP for two reasons:
the DP must justify its nominative Case by checking the corresponding feature(s)
against AGRS (containing T), or AGRS needs the DP to license its person /number
features. But in addition to nominative (N) subjects, Icelandic and Faroese have
oblique subjects, i.e., subject DPs with accusative (A), genitive (G) or dative (D)
case. Some Icelandic examples (from Sigurðsson 1989:200-201) are given in
(l)-(3):
(1) Hana dreymdi illa

she-Α dreamt badly
"She had a bad dream."
(2) Gunnars getur oft i fornum sögum

Gunnar-G mentions often in old stories
"Gunnar is often mentioned in old stories."
(3) Mér líður vel

I-D feel well
"I am feeling well."
As Sigurösson (1989:204-209) shows, oblique subjects are exactly like all

other subjects except that they do not have nominative Case, and do not induce
REFLEXIVES, PRONOUNS & AGREEMENT: ICELANDIC & FAROESE 191
subject/verb agreement. In particular, they appear in exactly the same positions

as nominative subjects, including SPECAGRSP. But it appears fairly unlikely that
oblique subjects move to SPECAGRSP because of any need for feature checking.
On the one hand, the particular Case assigned to an oblique subject clearly de
pends on the main verb, and should not be taken to be licensed by the T+AGRS
combination standardly assumed to license nominative Case. 1 On the other
hand, the verb never agrees (overtly) with an oblique subject. It always either
has the 3SG inflection, presumably by default, or actually agrees with a direct
object, as in (7) below, a fact which clearly speaks against the hypothesis that
oblique subjects move to AGRS to license person/number features on AGRS2:
(4) Pá dreymdil*-u illa

they-A dreamt-3SG/*-3PL badly
"They had a bad dream."
(5) Sona Gunnars getur/*-a oft i fornum sögum

sons-G Gunnar-G mention-3SG/*-3PL often in old stories
"Gunnar's sons are often mentioned in old stories."
(6) Okkur líður/*-un vel

we-D feel-3SG/*-iPL well
"We are feeling well."
(7) Okkur l*\íkarl-al*-um hestarnir

we-D ?*like-?*3SG/-3PL/*-IPL horses-the-N
"We like the horses."
In addition to this empirical problem confronting an application of the feature

checking approach to DP-movement in languages like Icelandic, there are certain
conceptual problems, at least with respect to the way this approach is imple
mented in Chomsky (1993). In Chomsky (1993) and much recent work, feature
checking is viewed as a process erasing certain features from functional heads,
1
Technically, it is of course possible to say that oblique subjects have 'abstract' nominative
Case, on top of whatever Case is reflected in their morphology, but this clearly has no ex
planatory value in the absence of independent motivation.
2
A finite verb always has overt number- and person-agreement with a nominative subject; cf.
2.4. below. Hence, if one were to posit covert agreement for sentences with oblique subjects,
one would have to explain why the agreement features are invisible in just this case. Moreover,
such covert agreement would have to be restricted to person, since the verb may have visible
number-agreement with a nominative object.
192 KNUTTARALD TARALDSEN
and is enforced by Full Interpretation on the assumption that these features re
ceive no interpretation, either at the PF- or the LF-interface. Thus, the relevant
features stand out as vestiges of an autonomous syntactic component in a theory
where syntax is taken to be determined by interface conditions. From this point
of view, the current use of feature checking to fuel movement to SPEC-positions
seems curiously non-minimalist.
2.2. The syntax ofρ redication
On the basis of the preceding observations, I will assume that DP-movement

to SPEC-positions is not in general motivated by the need for feature checking.
Rather, I will adopt the position that such movement is required to form struc
tures that can be interpreted as predications.3 More precisely, I will assume that
predication must take the following form at LF:
(8)
In structures of this form, X' is predicated of the constituent in SPECXP, which

thereby is assigned the θ-role corresponding to the position of the empty cate
gory (ec). Assume that the ec must be the trace of the constituent in SPECXP.
Then, if certain constituents can only be assigned θ-roles under predication, in
the sense just defined, movement to SPEC-positions is ultimately enforced by the
principle of Full Interpretation applying at LF.
Suppose now that we require the ec in (8) to be uniquely identified within
the predicate, i.e., X'. The following is a precise definition of what 'uniquely
identified' will be taken to mean:
(9) ec is uniquely identified in X' iff (a) or (b) holds:

(a) ec is Case-licensed by some Y in X'
(b) the person /number features of ec are copied on X.
The intuition behind this proposal is that if some head Y licenses a Case K, and
there is no overt DP bearing K, then the existence of a K-marked ec is deduced.
3
See Holmberg (1993:128) for a similar proposal, which, however, does not include an identi
fication condition on the predicate-internal empty argument.
Otherwise, the presence of the ec must be flagged by copying its person /number
features onto the head of the relevant domain.4
We can now treat subject/verb agreement as a side-effect of movement to
SPECXP in (8). If the DP moved to SPECXP in (8) is not Case-licensed in X', its
trace, the ec in (8), is uniquely identified in the predicate X' only if its person/
number features are copied onto X. Since the person / number features of the
trace must have the same values as those of the moved DP, the effect of specifier
/head agreement will arise, although, on this view, there is no feature checking
relation between X and SPECXP. If the DP in SPECXP is Case-licensed within
X', however, the identification condition does not require the person / number
features of its trace to appear on X, which may then take on default values.
We shall now see how this proposal can provide an account of the subject/
verb agreement pattern in Icelandic and Faroese.
2.3. Oblique subjects and subject/verb agreement
As mentioned above, the verb never agrees with an oblique subject in

Icelandic and Faroese. We have also seen that the Case-marking of an oblique
subject is determined by the main verb in these languages. In fact, Sigurðsson
(1989:189-241) argues convincingly that oblique subjects always originate as
internal arguments of the verbs that select their Case-features. Hence, sentences
like (4)-(7) instantiate the schema in (8) with the ec Case-licensed within X',
when X is taken as AGRS and the ec is in SPECVP.5 Accordingly, the identifica
tion requirement on predication, as formulated above, does not require copying
the person/number features of the oblique subject's trace onto AGRS. Hence, we
predict, correctly, that subject/verb agreement need not apply with oblique sub
jects.
To complete the account of (the lack of) subject/verb agreement with oblique
subjects, however, we must also explain why subject / verb agreement is not
only not required in this case, but in fact impossible. To this end, I would like to
appeal to a notion of Economy of Representation. In general, the principle of
Economy of Representation is thought of as favoring smaller trees over bigger
ones in the same candidate set. I would suggest that the relevant evaluation met
ric actually counts specified features rather than nodes. Thus, in a set of com-
4
In this formulation, my identification requirement for predicate-internal empty arguments
bears a certain resemblance to various disjunctive formulations of the ECP, taking (9b) to cor
respond to lexical proper government, while (9a) is a form of antecedent-government. The pos
sible significance of this relationship remains to be explored.
5 The exact nature of Case-licensing seems largely irrelevant here.
petitors that are equally well-formed with respect to other constraints, the prin
ciple of Economy of Representation will select the one with the least number of
specified feature-values and discard the others:
(10) When two otherwise well-formed structures A and  are non-

distinct, and A has fewer specified features than B, only A is
grammatical.6
Given (10), a representation with the features of an oblique subject copied on

AGRS will always be weeded out, since there is always a licit non-distinct repre
sentation where AGRS has unspecified (default) person/number features. Hence,
the verb never agrees with an oblique subject.
2.4. Nominative subjects and subject/verb agreement
A finite verb always agrees with a nominative subject in Icelandic and

Faroese, both in person and number, as illustrated by the following Icelandic
example:7
(11) Við fórum /*-ø /*-u til Noregs

we-N went-IPL / I S G /*-3PL to Norway-G
"We went to Norway."
This follows immediately from the proposal in 2.2 if nominative subjects are not
Case-licensed in a position lower than SPEC·AGRS. Then, the trace left behind in
AGRS', corresponding to the ec in (8), is uniquely identified, according to (9),
only if its person/number features are copied onto AGRS. Hence, we have the ef
fect of obligatory subject/verb agreement with nominative subjects.
6
The term 'non-distinct' is to be understood in such a way that (a) the phonetic matrix is not
taken into consideration and (b) an unspecified occurrence of a feature is non-distinct from any
specified occurrence of the same feature. This has the consequence that two representations A
and  where A has X in position Ρ and  has Y in P, may sometimes be considered non-distinct
from one another even when X and Y are generally regarded as different lexical items. This
property becomes crucial in 3.2. below. Thus, I disagree with Chomsky (1995), whose notion
of 'numeration' allows two representations to belong to the same competitor set only if they
are built from exactly the same lexical items.
7 Actually, the finite verb overtly inflects only for number in Modern Faroese. I have not ex
plored the possible relevance of this for the text analysis except for the suggestion in footnote
22.
In fact, most current accounts have nominative Case licensed in SPECAGRS,

with finite Τ combining with AGRS, i.e., outside AGRS' as is required for the
analysis just given to work properly. Yet, I would like to take a somewhat more
unusual position, namely that nominative Case is simply the default value
assigned at PF to an unspecified Case-feature. Obviously, this view also leads to
the expectation that nominative subjects require subject/verb agreement, since an
unspecified Case-feature is not licensed at all, and so, in particular, is not
licensed within AGRS'.
There are two considerations that would appear to favor the interpretation of
the nominative as a default Case, when we restrict our attention to Icelandic and
Faroese. The first has to do with the Case assigned to the direct object in sen
tences with oblique subjects. In Icelandic, though not in Faroese, this is the
nominative, as we already saw in (7) (repeated below as (12)), unless the main
verb happens to select an oblique object:8
(12) Okkur ?*likar/-a/*-um hestarnir

we-D ?*like-3SG/-3PL/*-iPL horses-the-N
The nominative object in sentences like (12) could not be licensed in

SPECAGRSP, filled by the oblique subject at all relevant levels, hence not by a
combination of finite Τ and AGRS . And in fact, the licensing of nominative ob
jects in sentences with oblique subjects does not depend on finiteness, since
nominative objects occur in all sorts of infinitival clauses, as exemplified in
(13):9
(13) Hann taldi okkur lika hestarnir

he-N believed-3SG we-D like horses-the-N
"He believed us to like the horses."
8 The distinctive property of an object with lexically selected Case is that it retains its Case
when promoted to subject, e.g., in passives.
9
This seems true for the majority of Icelandic speakers, although there are quite a few speakers
who do not accept infinitival clauses with oblique subjects and nominative objects in ECM-
contexts. According to Sigurðsson (1993: footnote 4), some of these speakers actually have ac
cusative Case on the object in this situation. Possibly, for the speakers who cannot have nom
inative Case on the embedded object in ECM-contexts, the infinitival AGRO cannot host a num
ber-feature in this configuration, as would be required for the trace of a nominative object to
satisfy (9) under the analysis in 2.5.
If we assume that no element licenses 'structural' accusative Case in sen

tences with oblique subjects (in Icelandic), the object can only have unspecified
Case, which does not require Case-licensing, and the nominative is simply the
default value chosen for an unspecified Case-feature at PF. In contrast to the
standard treatment of nominative Case, this approach has the virtue of allowing
for a unified analysis of nominative objects and nominative subjects. Suppose
that non-finite Τ (raised to AGRS) licenses null Case, forcing the subject to be
PRO, in control sentences, as in Chomsky & Lasnik (1991). Otherwise, no Case
is licensed above VP. Then, a non-oblique subject will be PRO in control sen
tences, but will have accusative Case, licensed by the matrix V, in ECM-con-
texts. In other infinitival clauses, the subject raises to the matrix SPECAGRSP
(independently of Case-licensing, since oblique subjects raise as well). If a sub
ject DP eventually winds up in the SPECAGRSP of a finite clause, further A-
movement is disallowed, again independently of Case-considerations, and so
must have unspecified Case, realized as the nominative at PF. The only signifi
cant properties of finite T, from this point of view, are that it does not license
null Case, and that it disallows Α-movement from the associated SPECAGRSP. 1 0
Since objects are never in the domain of a head licensing null Case (AGRS) and
never move into the matrix clause anyway, finiteness has the appearance of af
fecting the distribution of nominative subjects, but not that of nominative ob
jects.
The second relevant observation concerns a surprising restriction on direct
objects in sentences with oblique subjects. Even though (14) is perfectly fine,
the sentence in (15) is ungrammatical in Icelandic:
(14) Mér lkar hόkin (15) * Mér likar stulkan

I-D Hke-3sG book-the-N I-D like-3sG girl-the-N
"I like the book." "I like the girl."
To make (15) grammatical, one has to insert a preposition:
(16) Mér likar við stulkuna

I-D like-3SG with girl-the-A
"I like the girl."
In general, the preposition is obligatory, in Icelandic, whenever the object of lika

is taken to denote a human being.
10
Like most other current accounts, this analysis cannot use Case-theory to rule out sentences
like *There seems a man to be sleeping under the table; cf. Chomsky (1995) for a possible ex
planation not based on Case-licensing.
In Faroese, there is no such restriction. The Faroese counterpart to (15) is

grammatical even without the supporting preposition:
( 17) ær dámar gentuna

I-D like-3SG girl-the-A
"I like the girl."
This suggests that the restriction found in Icelandic is linked to the fact that ob
ject has nominative Case in sentences with oblique subjects in Icelandic, but has
accusative Case in Faroese, a minimal difference between the two languages,
which will be exploited again at various points below.
The conjecture that the ill-formedness of (15) is to be traced back to the di
rect object's having nominative Case is also supported by the observation that
whenever the direct object itself has oblique Case, it may perfectly well denote a
human being even in Icelandic:
(18) Hana vantar vini

she-Α lack-3sG friends-Α
"She lacks friends."
(19) Hann saknadi dóttur sinar

he-N missed daughter-G his-G
"He was missing his daughter."
(20) Barninu var lofað góðri dagmömmu

child-the-D was promised good-D day-mommy-D
"The child was promised a good nanny."
At the same time, we learn from (19)-(20), that the relevant distinction is not
simply accusative vs nominative, but rather nominative vs all other Cases. This
now becomes entirely natural if the nominative differs from all other Cases by
being the default value assigned to an unspecified Case-feature. We can then
characterize the condition relevant to (15) as follows:11
11
(21) is obviously not more than a descriptive generalization. A more explanatory account
might try to link (21) to the appearance of a preposition in front of direct objects denoting hu
man beings in Spanish and various other varieties of Romance. Such an account should also
explain why verbs carrying the -st suffix (at least historically an affixed reflexive) are systemat
ically exempt from (20) in Icelandic. Finally, we will eventually have to explain why (21) does
not apply to subjects.
(21) A [+ human] direct object must have a specified Case-feature.
2.5. Icelandic object/verb agreement
As mentioned above, the verb agrees with the direct object in Icelandic sen
tences with oblique subjects; cf. (7) (repeated below as (22)) and (23):12
(22) Okkur 1*lír/-/*- hestarnir

we-D ?*like-?*3SG/-3PL/*-IPL horses-the-N
(23) Mér ?*þykir> /-ja peir vera gáfaðir

I-D ?*think-?*3SG /-3PL they-N be gifted-N
"I think they are gifted."
In Faroese, however, the verb always takes the default 3SG form in such
sentences. This contrast between Icelandic and Faroese with respect to object
agreement would appear to follow from the fact that the object has accusative
Case in sentences with oblique subjects in Faroese. Not even in Icelandic does
the verb agree with an object which does not have nominative Case (because of
selectionai properties of the verb, the infinitive leiðast in (24), overriding default
Case assignment):
(24) Mér þykir I*-ja peim leiðast bókin

I-D think-3SG /*-3PL they-D be-bored-with book-the-N
"I think they are bored with the book."
If an object must move to SPECAGROP, by LF,13 the analysis of subject / verb

agreement proposed in 2.2 easily extends to object/verb agreement as well. If
the object is Case-licensed within AGRO', the identification condition does not
require a copy of its features on AGRO, and therefore Economy of Representation
12
In general, object agreement is strongly preferred over the default option (always 3SG) in
sentences with nominative objects. Nevertheless, many speakers find the default form relatively
acceptable. The text analysis is based on the assumption that such speakers allow pure number
agreement not to be spelled out at PF, taking verbal agreement with nominative objects to be
obligatory at least at LF.
13 On the basis of the proposals made in 2.2., we may conjecture that all direct arguments of a
verb are linked to their θ-roles via predication at LF. Then, AGRO, like AGRS, must be the head
of a predicate at LF. Word order seems to indicate that raising to S P E C A G R O P generally only
occurs after Spell-out, a property I have no explanation for.
disallows it. 14 But if the object is nominative, it really has no specified Case-
feature, outside of PF, and its trace in AGRO' is uniquely identified only if its
features are copied onto AGRO.
Notice that object agreement is restricted to agreement with respect to num
ber. Person agreement with the object is not possible:15
(25) Mér ?þi /?*-ja /*-j ið pið vera gáfaðir

I-D ?think-3SG /?*-3PL /*-2PL you-PL-N be gifted-N
"I think you are gifted."
This suggests that AGRO, unlike AGRS, can only host number features, a hypoth
esis which will become important in the next section.16
2.6. Summary
In this section, I have argued that movement to SPEC-positions is ultimately

driven by the need to form predication structures at LF, rather than by feature
checking of the conventional sort. The argument was based both on conceptual
considerations and certain empirical observations about Icelandic and Faroese. It
was then shown that an analysis which treats the nominative as the PF-reflex of
unspecified Case, and incorporates an identification condition on the empty ar
gument slot within predicates and a principle of Economy of Representation,
14
As a special case, objects assigned 'structural' accusative Case in sentences with non-
oblique subjects will never induce object / verb agreement, assuming that the 'structural' ac
cusative, like lexically selected Case, is licensed within AGRO', albeit by general rule.
15 Some Icelandic speakers do not accept 1st and 2nd person nominative objects at all. This is
not predicted by my analysis, unless, for these speakers, forms like ég "I" vs við "we" have
only person features, being distinguished as, say, 1 person vs 4 person rather than as lSG vs
IPL; cf. Kayne (1989). Another fact for which I have no explanation is that speakers who do
accept 1st and 2nd person nominative objects apparently prefer the default form of the verb over
object / verb agreement in sentences like (25), although the judgment goes in the opposite di
rection with 3rd person nominative objects.
16 Suppose number features can only be represented on AGRO, and that, moreover, a specified
person feature on AGRS needs the support of a specified number feature agreeing with the same
DP. Then, placing a nominative DP in SPEC-AGRSP, by requiring a specified person feature on
AGRS, will force AGRS to attract AGRO and the number feature of AGRO must agree with the
nominative subject rather than the object. Consequently we would predict correctly that the ob
ject cannot be nominative in a sentence with a nominative subject. Licensing of 'structural' ac
cusative Case for the object could then be viewed as a 'last resort' option (distinct from the de
fault assignment of nominative Case at PF, since, crucially, its effect must be LF-visible).
successfully accounts for the properties of subject/verb agreement and object/

verb agreement in Icelandic and Faroese.
In the next section, we shall see that the various ingredients of this analysis
receive independent support from the fact that they also lead to an explanatory
account of certain puzzling facts about the distribution of pronouns and simple
reflexives in Icelandic and Faroese.
3. Reflexives and pronouns
Icelandic and Faroese both have a so-called simple reflexive pronoun, sig in
Icelandic and seg in Faroese, which, unlike the non-reflexive pronouns, is refer-
entially dependent, and needs to be bound by a suitably local antecedent. Below,
I will first examine a surprising gap in the distribution of Icelandic sig, and then
discuss the nature of the complementary distribution of sig/seg and the non-re
flexive pronouns. In each case, I will argue that the Binding Theory does not
explain the pertinent observations, whereas the assumptions underlying the
analysis in Section 2 actually do.
3.1. sig cannot be a nominative object
In the Icelandic sentence (26), the relevant binding conditions, discussed in

more detail below, should be satisfied exactly as in the structurally similar (27):
(26) * Mariu fannst sig vera gáfuð

Mary-D thought-3SG sig-N be gifted-N
"Mary thought she was gifted."
(27) Maria taldi sig vera gáfaða

Mary-N believed-3SG sig-A be gifted-A
"Mary believed herself to be gifted."
Yet, (26) is ungrammatical, although (27) is fine. A priori, one might of course
claim that oblique subjects are not licit antecedents for sig. But the grammatical
y of (28) (where the embedded verb has a dative subject) immediately shows
this to be false:
(28) Mariu fannst sér fara aftur i norsku

Mary-D thought-3SG sig-D go backward in Norwegian
"Mary thought her Norwegian was getting worse."
Even more strikingly, the Faroese counterpart of (26) is grammatical:
(29) Maríu tókti seg vera klóka

Mary-D thought֊3SG seg- be wise-A
"Mary thought she was wise."
The only other difference between Faroese and Icelandic potentially relevant
to this contrast concerns the Case assigned to the object in sentences with
oblique subjects: in Faroese, it is accusative, but in Icelandic, nominative.
Hence, we are led to conclude that (26) is ungrammatical because sig has nomi
native Case. But why should precisely sig be allergic to nominative Case ?
In fact, the analysis developed in the preceding section provides an immediate
answer to this question. By previous assumption, sig ¡seg must eventually raise
to the matrix SPECAGROP in (26)-(29). By the identification condition on the
predicate internal ec, and the definition of unique identification in (9) (repeated
below), the trace of sig /seg must either be Case-licensed within AGRO', or else
have a copy of its features placed under AGRO:

(b) the person /number features of ec are copied on X.
Since AGRO, as seen in 2.5 may host only number features, the latter option is
open only to DPs that in fact have specified number features. Since the nomina
tive Case is simply the PF-value for unspecified Case, we therefore predict that a
nominative DP can successfully raise to SPECAGROP only if it has a specified
number feature.
Now, it so happens that sig /seg is unique within the pronominal system of
Icelandic and Faroese in that it does not have both singular and plural forms. In
fact, even when the antecedent is plural, sig /seg has the morphological shape of
a singular pronoun.17 We may take this to mean that sig /seg is actually unspec
ified for number and is assigned the singular form at PF by default. It follows
17
Compare the following paradigms : By contrast, the 1PL and 2PL forms are:
1SG 2SG sig 1PL 2PL
G mín fl ín sín G okkar ykkar
D mér flér sér D okkur ykkur
A mig flig sig A okkur ykkur
Faroese seg is morphologically related to the 1SG and 2SG forms in the same way.
then from the preceding observations that a nominative sig cannot raise to
18
SPECAGROP, and hence, sig cannot be a nominative object.
The success of this analysis provides independent support for the assump
tions it is based on, viz., the identification requirement associated with predica
tion and the claim that nominative DPs have unspecified Case (outside of PF).19
3.2. Sig and Economy of Representation
In general, the Icelandic sig, rather than a non-reflexive pronoun, must be

used when the pronoun would be co-referential with the closest subject. Thus,
h "her" cannot co-refer with the subject in (30):
(30) María taldi sigi /*hanai vera gáfa ða

Mary-N believed-3SG sig /her- be gifted-A
"Mary believed herself/her to be gifted."
This observation might suggest that sig is subject to the Binding Theory's
Principle A, while non-reflexive pronouns go by Principle B. However, sig too
seems subject to Principle  rather than Principle A. On the one hand, its an
tecedent need not be included in its binding domain, as shown by the well-
known 'long distance' cases. On the other hand, it does not normally even tol-
18
Sig cannot occur as a nominative subject either. Although it is somewhat more difficult to
eliminate competing explanations for this fact, I would like to suggest that here too, we see
the effect of the identification requirement on predicate-internal empty arguments. In particular,
I would suggest that sig not only lacks specified number, but in fact also has no person-fea
ture. Providing adequate motivation for this claim would take us outside the bounds of this ar
ticle, but notice that if sig has no specified number-feature, the claim made in footnote 16 that
a specified person-feature needs the support of a specified number feature entails that sig has no
specified person-feature either.
19 It is potentially embarrassing for the analysis presented here that it does not extend to (i),
since reciprocal hvor annar "each other" would not appear to be unspecified for number :
(i) * Börnunum leiddist hvorum annar
children-the-D got-bored-3SG each-D other-N
However, it appears that the reciprocal facts are somewhat less clear-cut than those involving
sig. Thus, Margret Jónsdóttir (p.c.) (unlike Höskuldur Práinsson, p.c.) considers (ii) as quite
acceptable. In (ii), the each-part of the reciprocal expression, like , also has nominative
Case, and does not agree in Case with its antecedent. Corresponding examples with nominative
(sjálf) sig are strongly rejected:
(ii) Börnunum leiddist hvor annar
children-the-D got-bored-3SG each-N other-N
For the moment, then, the significance of (i) is hard to assess.
erate an antecedent within its binding domain. In simple clauses like (31), for
instance, sig is impossible unless sjálfur "self is added:
(31) Maríai elskar *(sjálfa) sigi

Mary-N loves֊3SG self- sig-A
"Mary loves herself."
The reason (30) is grammatical without sjálfur is that sig here is part of an em
bedded clause, so that it may be assigned a binding domain which does not in
clude its antecedent.
In this perspective, sig is a pronoun, not an anaphor, in the standard sense
of this term. It is however referentially dependent, possibly because it has no
specified value of the number feature.
Assuming, therefore, that Binding Theory provides no explanation for the
complementarity of sig and non-reflexive pronouns, we are led back to the
question why the non-reflexive pronoun cannot co-refer with the matrix subject
in sentences like (30). xA.gain, the answer is supplied by one of the ingredients
entering into the analysis developed in the preceding section, the principle of
Economy of Representation.
As we saw in 3.1, the contention that sig has no specified number feature is
supported both by morphological considerations and the fact that this assump
tion allows a straightforward account of the absence of nominative sig. The non-
reflexive 3rd person pronouns, on the other hand, both have separate singular
and plural forms and can appear as nominative objects, as we shall see shortly.
Thus, they must have specified number features. By Economy of Representation
then, a non-reflexive pronoun is excluded from any context where it can licitly
be replaced by sig, which has fewer specified features.20 In (30), for instance,
20
When sig inside an infinitival or a subjunctive clause is bound to the subject of a higher
clause, it can be replaced by a non-reflexive pronoun. My analysis requires a structural differ
ence between the two sentences that result from this: the one with sig must have a structural
property allowing sig to link up with a distant antecedent, while the one with a non-reflexive
pronoun must not have this property. Then, the two will not be in the same competitor set.
This view is supported by the following observation (Höskuldur Práinsson, p.c.) : Consider a
sentence of the form indicated in (i), where IP2 and IP3 both are infinitival or subjunctive com
plements allowing for long distance binding of sig, and DP 2 and DP3 are both non-subjects :
(i) DP1 ... [IP2 ...DP2 ... [ I p 3 ... DP3 ...
Suppose both of DP 2 and DP3 are bound by DP 1 . Placing sig or the appropriate form of hann
in the positions of DP 2 and DP3, seems to yield the following acceptability pattern :
h cannot be co-indexed with the matrix subject, although this does not violate
Principle B, since this indexing licenses sig, also a non-anaphoric pronoun, in
contrast with the corresponding indexing in (31).21
If this analysis is correct, it provides independent evidence for our notion of
Economy of Representation as well as the assumption that sig has no specified
number feature, two important components of the analyses presented above. In
the next subsection, I provide a piece of striking evidence that the analysis is in
deed on the right track.
3.3. A surprising case of bound non-reflexive pronouns
We have seen that sig cannot occur in the nominative Case, according to the
analysis in 3.1 because it would then fail to satisfy the identification requirement
on predicate-internal empty arguments as defined in (9). Since objects have
nominative Case in Icelandic sentences with oblique subjects, provided the verb
does not select an oblique Case for its object, there will therefore be sentences
where the object cannot be sig, even though it would be bound by a suitable an
tecedent, as shown in 3.1. The analysis in 3.2 now predicts that exactly in these
(ii) (a) ok DP2= hann DP3 = Խոո

(b) * DP2= hann DP3 = sig
(c) ok DP2 = sig DP3 = sig
(d) ok DP2 = sig DP3 = Խոո
The unacceptability of (iib) follows from the text analysis if DP3 = sig bound by DPļ requires
(i) to have a structural property allowing the binding domain Δ relevant for licensing sig to in
clude DP1 Then, since IP3 is embedded in IP2, DP2 and DPļ must also be inside Δ, so that
DP2 = sig is possible, and DP2 = hann is impossible, by Economy of Representation.
21
The claim that co-indexing with the matrix subject does not violate Principle  in (30) (or
in (32) below), forces us to assume either that the embedded subject position is not governed
(only) by the matrix verb, or that the theory should not require an element to be governed
within its binding domain, as, for instance, in Hestvik (1990). A complication arises from the
fact that 'object shift' (see Holmberg & Platzack (1995) and references therein) may move the
embedded pronominal subject into the matrix clause, as seen more clearly in (i)-(ii), where the
pronoun precedes the matrix negation :
(i) María taldi sig /hana ekki vera gáfaða
Mary-N believed-3SG sig /her-A not be gifted-A
"Mary didn't believe herself/her to be gifted."
(ii) Manu fannst hún ekki vera gáfuð
Mary-D thought-3SG she-N not be gifted
"Mary didn't think she was gifted."
Yet, the pronoun may remain co-indexed with the matrix subject. At this point, I have no pre
cise proposal to make as to why object shift does not affect the assignment of binding domain
to the pronoun.
sentences, the object can be a non-reflexive pronoun even when it is co-referen

tial with the subject, as long as Principle  is not violated. This expectation is in
fact fulfilled. In (32), hún may be co-referential with the subject:
(32) Maríui fannst húni vera gáfuð

Mary-D thought֊3SG she-N be gifted
On my analysis, (32) is grammatical with hún co-referring with the subject,

since the corresponding sentence with sig, i.e., (26), is impossible because of
(9). Thus (32), with co-reference between the subject and the object, contrasts
with (30), where sig /seg can be used when co-reference obtains, since the sub
ject is not a nominative.
The pronoun in (32) may be bound by the subject because it has nominative
Case, and sig is thereby excluded. Confirmation of this analysis is provided by
(33): henni cannot be bound by the subject, in our analysis, because it is not a
nominative. Sig is therefore possible, as shown by (28), repeated below:
(33) * Maríui fannst hii fara aftur i norsku

Mary-D thought֊3SG she-D go backward in Norwegian
(28) Maríui fannst séri fara aftur i norsku

Mary-D thought-3SG sig-D go backward in Norwegian
Likewise, the object pronoun cannot be bound by the subject in the Faroese
counterpart to (32):
(34) * Maríui tókti hanai veraklóka

Mary-D thought-3SG she- be wise-A
This follows in our analysis because the object has accusative Case in Faroese,
making seg possible:
(29) Maríui tókti seg vera klóka

Mary-D thought-3SG seg- be wise-A
These observations show quite conclusively that the possibility for the object
pronoun to be bound by the subject in (32) has nothing to do with the fact that
the subject is oblique, but rather depends on the Case-marking of the object.
Notice, finally, that the effect of nominative Case relevant to (32) is not
simply that of lifting the normal Principle  requirements. If nominative pro
nouns were immune to Principle  effects, we would not understand the con
trast between (32) and (35), where the object pronoun cannot be bound by the
subject, although it has nominative Case:
(35) * Maríui leiddist húni

Mary-D got-bored-with-3SG she-N
"Mary got bored with her."
If on the other hand Principle  holds for nominative pronouns too, the contrast
is expected, since the pronoun has a binding domain (the infinitival clause) not
including its binder in (32) but not in (35).
I conclude that attributing the complementary distribution of sig and non-re
flexive pronouns to the principle of Economy of Representation makes correct
predictions that are not matched by alternative approaches.
3.4. Summary
In this section, we have seen that the distributional properties of sig /seg and
the non-reflexive pronouns in Icelandic and Faroese provide independent moti
vation for the crucial assumptions that the analysis in Section 2 was based on. In
particular, the absence of a nominative object sig was derived from the identifi
cation requirement on predicate-internal empty arguments and the assumption
that the nominative is the PF-spell-out of unspecified Case. The complementarity
of sig /seg and the non-reflexive 3rd person pronouns, on the other hand, was
taken to follow from the same notion of Economy of Representation that was
used in Section 2, a hypothesis that was perhaps most convincingly supported
by the facts discussed in 3.3.
Putting together the analyses in Sections 2 and 3 then, we have also reached
the surprising conclusion that the unavailability of nominative object sig and the
obligatoriness of verbal agreement with nominative subjects and objects are the
same fact at an appropriate level of abstraction, both reflecting the identification
condition on predicate-internal empty arguments and the status of the nominative
as a default Case. The complementarity of sig /seg and the non-reflexive pro
nouns, on the other hand, follows from the same principle of Economy of
Representation that excludes verbal agreement with non-nominative subjects and

objects.
4. Icelandic vs Faroese
At various points, we have seen contrasts between Icelandic and Faroese

playing an important role in the argumentation leading up to the analyses applied
in Sections 2 and 3. In this section, I will return to these contrasts and show that
they all fall out from the analysis, assuming a single structural difference be
tween the two languages.
4.1. The contrasts
I pointed out in 2.4 that Faroese differs from Icelandic with respect to the
Case assigned to the direct object in sentences with oblique subjects. Whenever
the Case is not lexically selected by the main verb, it is nominative in Icelandic,
but accusative in Faroese. The second difference we observed was that Icelandic
líka, but not its Faroese counterpart, rejects an object denoting a human being:
(14) * Mér likar stulkan

I-D like-3SG girl-the-N
"I like the girl."
(16) Mær damar gentuna

I-D like-3SG girl-the-A
"I like the girl."
We then saw that a finite verb shows number agreement with the object in sen
tences with oblique subjects in Icelandic, but not in Faroese:
(7) Okkur ?*likar/-a/*-um hestarnir

we-D like-?*3SG/-3PL/*-IPL horses-the-N
(36) Okkum dámar/*-a hestarnar

we-D like-3SG/*-3PL horses-the-A
The next contrast concerns the possibility of using the simple reflexive sig
/seg in the object position of a verb with an oblique subject, when the verb does
not select an oblique Case for its object. In Faroese this is perfectly possible,
and in Icelandic, it is not:
(26) * Mariu fannst sig vera gáfuð

Mary-D thought-3SG sig-N be gifted-N
(29) Maríu tókti seg vera klóka

Mary-D thought֊3SG seg- be wise-A
Finally, we observed that a non-reflexive 3rd person pronoun is free to co-

refer with the matrix subject in Icelandic sentences like (32), but not in their
Faroese counterparts:
(32) Maríui fannst huni vera gáfuð

Mary-D thought-3SG she-N be gifted
(34) *Maríui tókti hanai vera klóka

Mary-D thought-3SG she- be wise-A
Thus, we have isolated five different contrasts between Icelandic and Faroese.
4.2. A single parameter ?
One might take all the five contrasts listed above to follow from the first of
them, the contrast with respect to the Case-marking of the direct object in sen
tences with oblique subjects. From this, the remaining four contrasts follow,
within the framework of assumptions adopted in Sections 2-3. To recapitulate,
the contrast illustrated by *(15) vs (17) is a consequence of the difference with
respect to Case-marking together with the assumption that the nominative is a
default Case, and the condition in (21):
(21) A [+human] direct object must have a specified Case-feature.

The fact that object agreement is impossible in Faroese, although it is possi

ble in Icelandic, again follows from the difference in Case-marking, which now
interacts with the identification condition based on (9) and Economy of
Representation as interpreted in (10):

(b) the person /number features of ec are copiėd on X.
(10) When two otherwise well-formed structures A and  are non-

distinct, and A has fewer specified features than B, only A is
grammatical.
Likewise, the difference in Case-marking predicts the contrast *(26) vs (29),

given the analysis of the nominative as the PF-interpretation of unspecified Case,
and the identification condition based on (9). Finally, (32) vs (34) follows from
the same assumptions plus Economy of Representation.
This result does not mean that we have succeeded in isolating a single pa
rameter distinguishing Icelandic from Faroese in the relevant domain. It might
well be that the difference between the two languages with respect to the Case-
marking of the direct objects of verbs with oblique subjects follows from the in
terplay of various parameters with different settings in the two languages. In
fact, it is far from clear how to formulate a single parameter which would control
the distribution of accusative vs nominative Case in the relevant environments.
Alternatively, we take the contrast with respect to object agreement as basic.
Suppose that AGRO may carry a number feature in Icelandic, but not in Faroese.
This accounts directly for the contrast with respect to object agreement.
Assuming that the object must always raise to SPECAGROP, it also predicts that
Faroese must assign accusative Case to the object even in the contexts where the
object would be nominative in Icelandic. If the nominative is the default value
assigned to unspecified Case-features at PF, it follows from the identification
condition based on (9) that the trace of a raised nominative object is identified
within AGRO' only if its number feature is copied onto AGRO. Thus, if Faroese
AGRO cannot host a number feature, the object can never have nominative Case
in Faroese.
From this point of view, licensing of 'structural' accusative Case could be
considered a 'last resort' mechanism, in the sense of footnote 16, available in
both languages, but applicable only where a more highly favored option does
not exist, with copying of the object's number feature onto AGRO to be preferred
over 'structural' Case-licensing. To preserve my account of *(26) vs (29), we
must then make sure that the 'last resort' mechanism does not become available
for (26), even though its failure to apply actually causes (26) to be ill-formed.
This can be done by taking 'last resort' mechanisms to come into play only if
none of the members in a reference set is grammatical when more highly favored
options are taken. Assuming as before that (26) and (32) are in the same refer
ence set, the possibility for the object in (32) to be licensed without 'structural'
accusative Case makes 'last resort' Case-marking illicit in (26) as well. It is un
clear, however, how to extend this approach to (15), which is perhaps unlikely
to be in the same reference set as (16). Hence, the preceding remarks must be
left as suggestions for future research.22
4.3. Summary
In this section, I have listed five contrasts between Icelandic and Faroese,
and shown how they can be related to one another on the analysis developed
earlier. We have seen that there are at least two possibilities with respect to
which of the five contrasts should be considered basic. In either case, all five
contrasts are deduced from a single source in this analysis.
5. Conclusion
We started out by providing an analysis of subject /verb agreement and ob

ject/verb agreement in Icelandic and Faroese. That analysis derives the relevant
agreement patterns from principles and assumptions which subsequently turned
out to explain some striking distributional properties of reflexive and non-reflex
ive pronouns aş well. If the analysis is correct, a specific notion of predication
and a principle of economy of derivation play an important role in the syntax of
natural language.
Finally, we saw that five differences between Icelandic and Faroese, two
closely related languages, are reduced to one, on the approach taken here. The
question whether there really is a single basic parameter, however, has been left
open, although some speculations were offered.
22
Deriving the relevant contrasts between Icelandic and Faroese from the properties of AGRO
would seem particularly attractive, if the stipulated impossibility of associating a number fea
ture with AGRO in Faroese could be linked with the lack of person features on AGRS in Faroese
(see footnote 7).
REFERENCES
. 1995. "Bare Phrase Structure". Government and Binding Theory and
the Minimalist Program, ed. by Gert Webelhuth, 385-439. Oxford & Cambridge:
Blackwell.
& Howard Lasnik 1993. "Principles and Parameters Theory". Syntax:
Ein internationales Handbuch zeitgenös sicher Forschung = An International
Handbook of Contemporary Research, ed. by Joachim Jacobs, Arnim von
Stechow, Wolfgang Sternefeld & Theo Vennemann, 506-569. Berlin: Walter de
Gruyter.
Hestvik, Arild. 1990. LF-Movement of Pronouns and the Computation of Binding
Domains. PhD dissertation, Brandeis University.
Holmberg, Anders. 1993. "On the Structure of Predicate NP". Studia Linguistica
47.126-138.
& Christer Platzack. 1995. The Role of Inflection in Scandinavian Syntax.
New York & Oxford: Oxford University Press.
Kayne, Richard S. 1989. "Notes on English Agreement". Unpublished ms., City
University of New York.
Sigurðsson, Halldor A. 1989. Verbal Syntax and Case in Icelandic. PhD disserta
tion, Lund University.
. 1993. "Agreement as Head Visible Feature Government". Studia
ADVERBIAL QUANTIFIERS AND DIALECTAL
VARIATION IN A MINIMALIST FRAMEWORK*
MARIE-THÉRÈSE VINET
Université de Sherbrooke
Introduction
The question addressed in this paper is how best to characterize dialectal

variation with adverbial quantifiers in a minimalist program as developed by
Chomsky (1993, 1994). Within this framework, variation in general is consid
ered a property of lexical items and their features, and of sets constructed out of
them. Moreover, it is specified that variation must be visible for the young child
who develops his/her grammar. It is then all but natural to observe that variation
should essentially be found at the Phonetic Form (PF) interface and in the lexi
con. In other words, as assumed by Chomsky (1994), parameters between
grammars are limited to the lexicon. The same pattern applies to micro-paramet
ric syntax.
In an important work on variation, Cheng (1991) discusses lexical ambigui
ties of WH-words in Mandarin Chinese which can be interpreted in various ways
depending on a polarity trigger. She assumes that a lexical difference between
English and Mandarin Chinese WH-words is responsible for a series of other
derived properties in these grammars. A similar view on variation is developed
in Watanabe (1992). For a study of crosslinguistic variation of reflexives along
similar lines, cf. Jakubowicz (1994).
This study was partially funded by the Social Sciences and Humanities Research Council of
Canada [SSHRC grant number 410-93-0838) and FCAR (94ER0401). I would like to thank
the members of the FCAR research group (UQAM-Sherbrooke) for their comments and helpful
suggestions as well as the audiences in Tarragona (17/3/94), at the University of Toronto
(4/3/94), the University of Ottawa (28/9/94), Université du Québec à Montréal (20/10/94) and
the Saint John campus of the University of New Brunswick. Special thanks also to Guglielmo
Cinque, Esther Torrego and Anne Zribi-Hertz. The usual disclaimers apply.
214 MARIE-THERESE VINET
In a related fashion, this study focuses on lexical properties of adverbial

quantifiers in Quebec French (QF) such as the reduplicated form benben "many,
a lot" in (1), which is interpreted as a negative polarity item (NPI) in certain con
texts. It can be observed that the adverbial quantifier appears postnominally only
when it is in a 'triggering' environment, as in (1c):
(1) a. J'ai pas (vraiment) rencontré benben d'amis.

b. J'ai pas (vraiment) benben rencontré d'amis.
I didn't (really) a lot meet of friends
c. J'ai pas rencontré d'amis benben/*J' rencontré d'amis benben
"I didn't meet a lot of friends /I met a lot of friends."
A distributional contrast is observed with the Standard French (SF) quantifier

beaucoup "a lot", for instance, which is not a polarity item and where the post-
nominal position, as in (2c), is always ruled out:
(2) a. Je n'ai pas rencontré beaucoup d'amis.

b. Je n'ai pas beaucoup rencontré d'amis.
I didn't a lot meet of friends
c. * Je n'ai (pas) rencontré d'amis beaucoup (SF)
"I didn't meet a lot of friends."
Within Chomsky's minimalist framework, operations are driven by morpho

logical necessity only. Adverbial-quantifiers in general do not need to check their
features. Hence, by the Principles of Economy (Greed), they cannot move. A
movement of Q at S-structure into the adjoined VP position [pas VP[benbeni
y՝[rencontré]] ti ď amis] is therefore not tenable within the proposed framework.
I rather propose that such adverbial-quantifiers are generated in situ in various
positions in the domain of the predicate, depending on their lexical properties.
The aim of this study is to demonstrate that it is clearly the lexical properties
of adverbial Qs along with general principles of Universal Grammar (UG), which
are responsible for the choice of location illustrated in (1c / 2c).
This article is organized as follows. In section 1, the relevant facts from
Quebec French are introduced and benben is compared to other quantifiers. In
section 2, the lexical properties of various measure quantifiers in this dialect are
related to the possibility of a post-NP or sentence-final position of the adverbial
Q. I then suggest an analysis to account for the different behavior of adjectival-
quantifiers. In the next section, the important role of NEG in this structure is ana
lyzed. I illustrate how my analysis is compatible with Cinque (1994), who has
hypothesized, through his study of past participle movement in Italian, that ad-
ADVERBIAL QUANTIFIERS AND DIALECT VARIAΉΟΝ 215
verbs occur in the SPEC of various intermediate functional phrases. Moreover, it

is demonstrated how post-nominal benben and other quantifiers of the same type
are licensed in the grammar of QF. It is shown that such Qs obey locality condi
tions, c-command and the notion of syntactic domain.
1. Adverbial quantifiers in Quebec French
Adverbial quantifiers of the type studied here have received very little atten
tion in the literature. For some recent related studies, cf. Léard & Amyot (1993),
Vinet (1994), Vinet & al. (1994).
Benben can be loosely compared to beaucoup in SF in the sense that both are
quantifiers and when they appear with Negation they both show scope interac
tion effects. Benben in (la) presents scope ambiguities just as beaucoup would
in the same position (cf. Rizzi 1990:18). It can mean "many friends are such that
I did not meet them" or "not many friends are such that I met them". The sen
tences (lb, c), however, are unambiguous and allow only the second reading. It
can be observed, as a result, that benben in a pre-past participle position and
benben in a post-NP position both receive the same unambiguous internal scope
interpretation.
However, there exist many obvious differences in behaviour between the
two Qs (cf. Vinet & al. 1994). First, it can be observed that benben can appear
before a noun and it can be used as a quantifier or an intensifier for a verb, an
adjective or an adverb (cf. ). On the contrary beaucoup is more restricted. It
can quantify over a noun and it can quantify a verb but it is ruled out with most
adjectives and adverbs (cf.(3b)), except for comparative adverbs, as illustrated
in (3c) below:
(3) a. pas benben ďamis / benben mangé / drôle / souvent

"not a lot of friends / eaten a lot / very funny / very often"
b. beaucoup d'amis / mangé; *beaucoup drôle / *beaucoup souvent
c. beaucoup (*benben) trop / plus / moms doux
"much too / more / less soft"
A second difference is related to the fact that, in a positive context, benben

can appear with a partitive determiner while beaucoup is uniquely followed by
the bare partitive preposition de (cf. Vinet & al. 1994):
(4) a. J'ai benben dela misère,

b. * J'ai benben de misère.
216 MARIE-THÉRÈSE VINET
c. * J'ai beaucoup dela misère.

d. J'ai beaucoup de misère.
"I have a lot of difficulties."
The contrast in (4b) and (4d) clearly indicates that de is a negative polarity
item when it appears with benben (cf. Horn 1978) and a partitive preposition
when it appears with beaucoup. In other words, de is not in the scope of ben-
ben. Though beaucoup can license an existentially quantified N, benben cannot.
Note that benben is always optional in both positive or negative contexts, as il
lustrated in (5a, b). Nevertheless, in a positive context, benben is followed by
an indefinite des and it can never appear sentence-finally, as (5c) illustrates:
(5) a. J'ai pas (benben) d'amis.

b. J'ai (benben) des amis.
c. * J'ai des amis benben.
"I have a lot of friends."
Note that if benben does appear with the indefinite in certain negative envi
ronments, it then has the meaning of the adverb vraiment "really" as in C'était
pas des cours benben parce qu'on faisait rien "They were not really courses be
cause we weren't doing anything" (P. Thibault: corpus sur la langue parlée à
Montréal, Université de Montréal).
Benben, as well as beaucoup, can quantify at a distance (cf. Obenauer 1983,
1984). However, when benben appears in a positive context, acceptability
judgements are not always clear: quantification at a distance (QAD) seems more
acceptable with certain verbs only:1
(6) a. ? J'ai benben lu / vu de livres.

"I have read / seen a lot of books."
b. * J'ai benben aimé / apprécié de films.
"I have enjoyed / appreciated a lot of movies."
On the other hand, QAD is acceptable with all predicates in a negative context, as
illustrated in (7):
1
Obenauer (1983, 1984) claims that QAD is excluded with predicates where an iterative (X
times) interpretation of the adverb is found. According to his hypothesis, it is the iterative in
terpretation over the event which licenses QAD. Doetjes (1995) questions this hypothesis and
demonstrates that it is rather a count event which is responsible for the iterative reading. QAD
is possible with a mass event, for instance, as illustrated in Julien a beaucouplbenben dormi
"Julien has slept a lot".
ADVERBIAL QUANTIFIERS AND DIALECT VARIΑΉΟΝ 217
(7)  a pas benben lu / vu / aimé de livres

"He did not read / see / enjoy many books."
All the facts presented in this section tend to illustrate that benben is a polar
ity sensitive item whose distribution and interpretation is sensitive to negative
versus affirmative contexts (cf. Progovac 1993). This situation is reminiscent of
a relatively similar phenomenon with the quantifier not much which is normally
in suppletive variation with a lot. As noted by Klima (1964:283), only a lot can
appear in a positive sentence, as exemplified in (8) below:
(8) a. Writers don't accept suggestions much these days,

b. * Writers accept suggestions much these days.
 Writers accept suggestions a lot these days.
To conclude this section, it can therefore be hypothesized that negation and

more particularly negative polarity adverbial-quantifiers are necessary triggers
for the post-nominal position. However, if they are necessary triggers in the
process, can they be claimed to be sufficient triggers as well?
2. Lexical properties of adverbial quantifiers
Other adverbial Qs, like benben, can also appear postnominally in QF, as il
lustrated in (9b):
(9) a. us ont pas benben / tellement / trop trop / le diable ďargent.

"They don't have a lot/so much/too much/the devil (of) money."
b. Ľs ont pas d' argent benben / tellement / trop trop / le diable...
Adverbial Qs which appear postnominally in (9) denote scalar endpoints (cf.

Fauconnier 1980) and some of them, namely trop trop and le diable, are strict
NPIs. Positive polarity Qs in both SF and QF, such as pas mal "quite a lot", never
show up in a postnominal position when they bind a non-variable i.e., an
anaphor as in (10):2
2
An apparent counterexample which could be raised is the case of en masse "masses/a lot"
which can be used postnominally in a positive context as exemplified below:
(i) a. J'ai en masse d'amis. b. *J'ai d'amis en masse,
. J'ai des amis en masse.
"I have masses / a lot of friends."
(10) * Ils ont bu e de bière pas mal.

"They drank (of) beer quite a lot."
If it does, it is then claimed that the representation will not satisfy the condi
tion of Full Interpretation, it will not converge at Logical Form (LF) and the
derivation will crash.
Note, incidentally, that the presence of post-nominal adverbial-quantifiers is
not totally excluded from the grammar of French spoken in France.
Interestingly, there exists the negative form du tout "at all" in SF as in Elle n'a
pas (du tout) rencontré (du tout) ď amis (du tout) "She did not (at all) meet (at
all) friends (at all)". One can also find the following data with the negative po
larity quantifier bezef"a lot", from dialectal Arabic bezzaf"a. lot", in the gram
mar of colloquial French. However, bezef follows a different pattern, as can be
observed in (11):
(11) a. J'aipas bezef ď argent. /J'en aipas bezef.
"I don't have a lot of money."
b. * J'aipas d'argent bezef.
c. * J'aipas bezef trouvé d'argent
"I have not found a lot of money."
Cellar (1991) mentions that bezef is used more frequently in the negative.
Note that bezef in colloquial French behaves differently from bezzaf in dialectal
Arabic which can be used in a positive context and also sentence-finally as in:
Ahmed mrid bezzaf"Amhed (is) ill a lot". The behavior of bezef seems to follow
more closely the behavior of adjectival-quantifiers in QF and we now turn to the
study of such quantifiers.
2.1 Adjectival-quantifiers
Interestingly, a certain type of quantifier in QF, namely adjectival-quantifiers

which can always take a reduplicated form, such as gros I gros gros, épais,
grand, etc., never seem to appear postnominally. This is observed even if they
surface obligatorily in a negative context, as is the case with gros gros. Contrary
to adverbial-quantifiers also, they can only appear with quantified NPs, never
with adjectives or adverbs, as in *pas gros gros drôle I longtemps "not big big
However, when en masse appears postnominally, it does not quantify over the N any more.
The partitive de is then obligatorily transformed into an indefinite determiner des, as in (ic).
ADVERBIAL QUANTIFIERS AND DIALECT VARIATION 219
funny / a long time". These QF adjectival quantifiers, or quantifiers with adjecti

val forms, are exemplified in (12):
(12) a. On avait pas gros gros d'argent.

"We didn't have big big (of) money."
b. * J'ai pas d'argent gros gros.
"I don't have (of) money big big."
c. Elle a (pas) mis épais (épais) de confiture sur ma toast.
"She (didn't) put (heavy heavy of) jam on my toast."
d. * Elle a pas mis de confiture épais (épais) sur ma toast.
"She didn't put jam heavy heavy on my toast."
e. T'as pas grand (grand) de cuisine.
"You don't have big big (of) kitchen."
f. * T'as pas de cuisine grand (grand).
"You don't have of kitchen big big."
g.  m'a pas laissé large (large) de terrain.
"He didn't leave me large large (of) field."
h. *  m'a pas laissé de terrain large (large).
"He didn't leave me offieldlarge large."
This situation clearly demonstrates that the distinction between the quanti
fiers in (9) and in (12) is not uniquely triggered by negation, it is also morpho
logical in nature. Furthermore, the non-occurrence of examples like (12a, c, e,
g) in SF must be explained. Note that SF also has negative adjectival-quantifiers
but in examples without de as in Je n'ai pas grand temps "I don't have big / (a
lot) of time" or with idiosyncratic expressions where a full NP never surfaces: Il
n'en reste pas lourd "There isn't much left", Je n'en menais pas large "My heart
was in my boots". There also exists the case of the quantifier lerche "expensive /
a lot" in French slang, as in J' ai pas trouvé lerche ď hôtels "I didn't find a lot of
hotels". These adjectival Qs also cannot appear postnominally. Adjectival-quan
tifiers can therefore not be considered the sole property of QF. They are also
found in different dialects of French.
It is important to note that adverbial-quantifiers seem to be particular to (the
various grammars of) French. They are not found in many Romance languages,
such as Italian and Spanish. Longobardi (1994) indicates that molto/-a/-i/-e "a lot
of' is inflected for gender and number in agreement with the head noun:
(13) a. Ho trovato molti amici. (Longobardi 1994)

"I found a lot of friends."
b. Ho trovato molta acqua.

"I found a lot of water."
From this point of view, molto can be said to behave more like an adjective
or a determiner. However, quantification over individuals, when the head noun
is in the singular, is always ruled out:*Hotrovato molto amico "I found a lot of
friend". In the same fashion, quantification over individuals, with adjectival-
quantifiers in QF, never allows a head noun in the singular:*J'ai pas gros gros
de cheval à vendre "I do not have big big of (a) horse to sell".
The invariable nature of adverbs in general explains the impossibility of gen
der and number agreement between the adjectival-quantifier and the head noun:
(14) a. * J'ai pas grosses grosses de robes à me mettre.

"I do not have big big of dresses to put on (me)."
b. * J'ai pas grande grande de cuisine.
"I do not have large large of kitchen."
These last examples clearly indicate that adjectival-quantifiers in French cor

respond more closely to French (adverbial) quantifiers than to the Italian quanti
fier molto. In other words, they are not adjectival in character. As indicated in
Baker (1988), when two elements are incorporated they then form a new mor
phological category which cannot be subdivided.
2.2. The inner structure of adjectival-quantifiers
Why are there adjectival-quantifiers? What is the structural difference be

tween a plain adjective—un livre intéressant "an interesting book"—and an ad
jectival quantifier as in (12) above? Recall that Abney (1987) has already pro
posed that APs, QPs and ADVPs are identical in internal structure. His claim was
that they are subvarieties of the same syntactic category [+N, +ADJ]. In a similar
fashion, I want to hypothesize that morphologically complex Qs which present
[+N] or [+A] features cannot quantify at a distance. Only the forms which are
interpreted as morphologically non-complex or bare Q forms can therefore ap
pear preverbally or postnominally.
What about adjectival-quantifiers? Should they be distinguished structurally
from adverbial-quantifiers? I would like to assume that adjectival-quantifiers are
morphologically complex and present the structure in (15) where Cinque's
(1993b) analysis of adjectives as specifiers of a Functional projection is taken
into account:
The Specifier-Head agreement between A and Q (an agreement between a

lexical Specifier and a functional head) would then account for the lack of overt
agreement features on this type of adjective, as the examples in (14) above illus
trate.
Moreover, it can be observed that only a subclass of lexical adjectives would
be possible candidates, namely adjectives with a polarized contrast which ex
press a (negative) gradable antonymy having to do with spatial and temporal ex
tension (cf. Lyons 1977). These semantically unmarked adjectives can be listed
as follows and the range of lexical forms selected can vary from one dialect to
another: gros "big", grand "tall", épais "thick", large "large", lourd "heavy", etc.
However, one cannot find adjectives like petit "small", mince "thin" as in *J'ai
(pas) petit de temps "I don't have small of time". Other forms such as beau
"beautiful", bon "good", intéressant "interesting", for instance, which are rather
evaluative are not acceptable in this context
The important aspect of structure (15) for my analysis is that it establishes a
clear difference between morphologically complex quantifiers such as adjectival-
quantifiers and the non-complex Q forms listed in (9) above which correspond to
'true' adverbial forms. As previously observed, certain derived properties are
attached to these two different morphological shapes. Adjectival-quantifiers,
contrary to adverbial forms, for instance, must always be locally related to the
NP they quantify:
(16) * J'ai pas gros (gros) invité de gens.

"I didn't big (big) invite (of) people."
I therefore suggest that Q adjectives form a word class with distinct gram
matical properties. In the next section, I illustrate how the syntactic representa
tion of such adverbial-quantifiers, namely benben, depends on a trigger in the
sentence.
3. NEG as a licenser for benben
The syntactic properties of the adverbial-quantifier benben differ, just like

the quantifier any, according to whether it is in the scope of negation or not.
222 MARIE-THERESE VINET
Kayne (1981) has shown that French [QP[e] de N] structures were headed by an
empty initial QP and that beaucoup could fill this position. However, as illus
trated in (17) below, there is a sharp contrast between the two forms since ben-
ben in (17b,c) is a negative polarity item which must be licensed by a sentential
NEG:
(17) a. Beaucoup d'amis sont venus,

b. * Benben d'amis sontpas venus.
"Many friends did not come over."
 ?* Pas benben d'amis sont venus.
"Not many friends came over."
The negation forms which can license benben are very few. There is the
minimal negation quantifier pas, the aspectual quantifier plus (pronounced pu
[py] in QF) and jamais the temporal quantifier:
(18) a. J'ai pas /pu /jamais benbende vingt-cinqcennes.

b. J'ai pas / pu / jamais de vingt-cinq cennes benben.
"I have not / no more / never (a lot) of quarters (=25¢) (a lot)."
The NEG licenser can never be a head such as ne or non, as in (19):
(19) a. * Je n'ai benben d'amis, (cf. Elle n' aime personne)

"I NEG have a lot of friends." ("She does not like anyone.")
b. * J'ai rencontré non benben d'amis, mais benben d'ennemis.
"I have met not a lot of friends, but a lot of ennemies."
Cinque (1994) has proposed that adverbs in general are specifiers and not
adjuncts, as in the Emonds / Pollock approach. His main argument is connected
to the relatively rigid order which is found among adverbs. He therefore hypoth
esizes that adverbs occur in the SPEC of various intermediate functional phrases
(FPs) such as NEGP (pas), TP (toujours), ASPP (benben), etc., an hypothesis
which explains the rigid word order among adverbs:
(20) a. Je n'ai pas toujours beaucoup tout bien nettoyé.

I did not always a lot all well clean,
b. * Je n'ai pas beaucoup tout bien toujours nettoyé.
I did not a lot all well always clean.
This situation can be represented as in (21):

(21) ...FP1[pasF'FP2[toujoursҒ'FP3[ Ғ' ...VP[...nettoyé]]]]
Benben and similar adverbial-quantifiers can appear as the most deeply em
bedded element of the VP because, as negative polarity items, they are focus
sensitive. These all appear with at least two syllables and there is always a
falling intonation on the last syllable when it appears sentence-finally. Adverbs
with one syllable only, such as ben or trop, can usually not bear the greatest
prominence of the sentence (cf. Cinque 1993a), as illustrated in (22) below:
(22) a. J'ai pas ben aimé ça. /*J'ai pas aimé ça ben.
"I didn't (much) like it (much)."
b. Il a pas trop vu la différence.
c. * Il a pas vu la différence trop.
"He didn't (too much) see the difference (too much)."
Note that there must also be a rigid order between the -commander and the
polarity item, as claimed by Baker (1970), Lasnik (1993) and many others. This
explains the unacceptability of the following sentence in (23):
(23) * J'ai benben pas / pu /jamais de vingt-cinq cennes.

"I have a lot not / no more / never of quarters."
It can therefore be observed that the proper NEG -commanders for benben
usually correspond to negative adverbial forms and not to syntactic morphemes
such as ne I non, a class distinction also noted in Di Sciullo & Tremblay (1993).
I leave aside the preposition sans which also acts as a -commander in sans ben-
ben réfléchir I sans réfléchir benben "without much thinking / without thinking
much".
3.1 The syntactic domain of benben
Sportiche (1988) has proposed that 'floating' Qs are generally partitive and
that tous "all" is treated as a partitive structure, without de insertion. Even
though adverbial Qs like benben are also partitive and require an association with
a plural NP just like tous, benben behaves nonetheless quite differently. First,
the scope of the domain of benben is VP only, contrary to tous where a syntactic
dependency between Q and the NP in SPEC-IP position is possible. This contrast
is illustrated below in (24):
(24) a. Les enfantSi ont tous ei vu ce film.

"The children haveallmasc/plseen this movie."
b. * D'enfantsi ont pas benben e¡ vu ce film.
"Children have not a lot seen this movie."
Secondly, benben is a negative polarity item when it appears in conjunction
with the partitive negative polarity item de. It is then the negation quantifier pas
which can quantify at a distance.3 It can appear in all positions where adverbs
are allowed, namely pre-PP, post-PP or sentence-finally as observed with the
adverb souvent in (25):
(25) a. Ils ont pas (souvent) ramassé (souvent) de fraises (souvent).

"They have not (often) picked up (often) strawberries (often)."
b. Ils ont pas (benben) ramassé (benben) de fraises (benben).
I suggest that the NEG item is both a trigger and a binder for the polarity of Q
and de. It is also hypothesized that the trigger must be in the appropriate domain
since negative polarity is a clause-bound process. As often noted, volitional
verbs can create a larger domain for negative polarity contrary to others like
epistemic and factive verbs, as illustrated by the very clear contrast in the follow
ing examples:
(26) a. Elle veut pas que j'aye ďamis benben après Fècole.
"She does not want me to have-SUBJ friends much after school.
b. * Elle pense pas que j'ai d'amis benben après ľécole.
c. * Elle regrette pas que j 'aye d'amis benben après l'école.
"She does not think / regret that I have (of) friends much..."
In (26a), as predicted, postnominal benben is interpreted in the scope of the

matrix verb {vouloir), which is assumed to carry no truth value. Moreover, the
3
Furthermore, note that there can be no minimality effects with benben since (contrary to
beaucoup) it cannot bind by itself de N expressions. Minimality effects, as in (ib), are rather
linked to the presence of the Neg item pas which binds the existentially quantified N, as (ic,d)
illustrate:
(i) a. Combien de films que t'as pas benben aimés?
b. * Combien que t'as pas benben aimé de films?
c. * Combien que t'as pas aimé de films (benben)?
"How many movies did you not like a lot?"
d. Combien que t'as aimé de films?
"How many movies did you like?"
NEG licenser cannot bind more than one polarity item at a time in the same do
main, as (27) illustrates:
(27) a. ?* Les enfants ont pas benben ramassé de fraises trop trop.
"The children have not a lot picked up strawberries too
much."
b. * Les enfants ont pas benben ramassé de fraises benben.
4. Conclusion
In sum, the paper proposes that different factors can explain the syntactic
representations of adverbial Qs in QF and in other varieties of French as well.
First, it was indicated that certain adverbial features within a lexical form such as
benben were more dominant than in the case of adjectival-quantifiers. Moreover,
all these quantifiers which can easily appear sentence-finally always bear a
reading that falls within the standard negative polarity environments. These lexi
cal characteristics thus explain why adjectival-quantifiers cannot quantify at a
distance.
It was also demonstrated that the quantifier benben can receive several inter
pretations in the grammar of QF and that its interpretation as a negative polarity
item always depends on a trigger, as also observed by Cheng (1991) for the
various interpretations of WH-words in Mandarin Chinese. On the other hand,
adverbial quantifiers like beaucoup in SF are not interpreted as negative polarity
items depending on a trigger. They are rather independent items and therefore
they are not focus sensitive or polarity sensitive items.
Secondly, since benben and other quantifiers of the same type appear in a
negative polarity environment, it was indicated how they obey principles of UG,
such as locality conditions, c-command and the notion of syntactic domain.
Furthermore, it was shown that such Qs simply appear in various A positions in
the structure, as proposed by Chomsky (1993) and Cinque (1994). All of these
principles allow one to identify invariants in the model, i.e., elements which are
considered part of conceptual necessity for developing any human language.
REFERENCES
Abney, Steven. 1987. The English Noun Phrase in its Sentential Aspect. PhD dis
sertation, MIT.
Baker, Leroy . 1970. "Double negatives". Linguistic Inquiry 1,2.169-186.
Baker, Mark. 1988. Incorporation: a Theory of Grammatical Function Changing.
University of Chicago Press.
Cheng, Lisa. 1991. On the Typology of WH-Questions. PhD dissertation, MIT.
Cellar, Jacques. 1991. Dictionnaire du français non conventionnel. Paris:
Hachette.
. 1994. "Bare Phrase Structure". MIT Occasional Papers in Linguistics 5,
Cinque, Guglielmo. 1993a. "A Null Theory of Phrase and Compound Stress".
Linguistic Inquiry 24,2.239-298.
. 1993b. 'On the Evidence for Partial N Movement in the Romance DP".
ms., University of Venice.
—. 1994. "Romance Past Participle Movement and Clause Structure". Paper
read at "Table-ronde: linguistique comparée des langues et dialectes maro
cains", Rabat University, Marrakesh, Morocco, February 3-6, 1994.
Di Sciullo, Anna-Maria & Mireille Tremblay. 1993. "Négation et interfaces".
Toronto Working Papers in Linguistics 12.1, ed. by Carrie Dyck, 75-89.
Toronto: Linguistics Graduate Students Union.
Doetjes, Jenny. 1995. "Quantification at a Distance and Event Relatedness".
Linguistics in the Netherlands 1994, ed. by Reineke Bok-Bennema & Crit
Cremers, 13-24. Amsterdam & Philadelphia: John Benjamins Ltd.
Emonds, Joseph. 1985. A Unified Theory of Syntactic Categories. Dordrecht:
Foris.
Fauconnier, Gilles. 1980. Étude de certains aspects logiques et grammaticaux de
la quantification et de l' anaphore en français et en anglais. Presses de l'Uni
versité de Lille.
Horn, Laurence R. 1978. "Remarks on Neg-Raising". Syntax and Semantics 9:
Pragmatics, ed. by Peter Cole, 129-220. New York: Academic Press.
Jakubowicz, Célia. 1994. "On the Morphological Specification of Reflexives:
Implications for Acquisition". Proceedings of NELS 24, 205-219. Graduate
Students Linguistic Association: University of Massachusetts at Amherst.
Kayne, Richard. 1981. "ECP Extensions". Linguistic Inquiry 12,1.93-133.
Klima, Edward S. 1964. "Negation in English". The Structure of Language:
Readings in the Philosophy of Language, ed. by Jerry Fodor & Jerrold Katz,
246-323. Englewood Cliffs, New Jersey: Prentice-Hall.
Lasnik, Howard. 1993. "Lectures on Minimalist Syntax". University of
Connecticut Occasional Papers in Linguistics 1.
Léard, Jean-Marcel & Denis Amyot. 1993. "Observations sur la grammaire de

Bien(F) et Ben (Q), quantifieurs nominaux", Paper presented at the meeting of
the Canadian Linguistic Association, Carleton University, Ottawa.
Longobardi, Giuseppe. 1994. "Reference and Proper Names". Linguistic Inquiry
25,4.609-665.
Lyons, John. 1977. Semantics. Cambridge: Cambridge University Press.
Obenauer, Hans-Georg. 1983. "Une quantification non canonique: la
'quantification à distance"'. Langue française 58.66-88.
·. 1984. "On the Identification of Empty Categories". The Linguistic
Review 4.153-202.
Pollock, Jean-Yves. 1989. "Verb Movement, Universal Grammar and the Structure
of IP". Linguistic Inquiry 20.365-424.
Progovac, Ljiljana. 1993. "Negative Polarity: Entailment and Binding".
Linguistics and Philosophy 16.149-180.
Rizzi, Luigi. 1990. Relativized Minimality. Cambridge: MIT Press.
Sportiche, Dominique. 1988. "A Theory of Floating Quantifiers and its
Corollaries for Constituent Structure". Linguistic Inquiry 19,3.425-449.
Thibault, Pierrette. 1984. Corpus Montréal 84: Projet de recherche sur le français
parlé à Montréal. Université de Montréal. See also Thibault, Pierrette & D.
Vincent. 1990. Un corpus de français parlé. Collection recherches sociolinguís
tiques. Québec: Université Laval.
Vinet, Marie-Thérèse. 1994. "Adverbial Quantifiers in Quebec French". Paper
presented at the Fourth Colloquium on Generative Grammar, March 16-18,
1994, Tarragona, Spain.
, Marie-Odile Junker, Noëlla Roy & Alain Villeneuve. 1994. "Les pro
priétés lexicales des quantifieurs adverbiaux en français québécois". Paper read
at meeting of ACFAS [Association canadienne française pour l'avancement des
sciences], Université du Québec à Montréal.
Watanabe, Akira. 1992. "WH in Situ, Subjacency, and Chain Formation", ms.,
MIT, Cambridge, Mass.
VERB CLUSTERS IN CONTINENTAL
WEST GERMANIC DIALECTS
JAN-WOUTER ZWART
University of Groningen
1. Introduction
The Continental West Germanic Languages include the standard varieties of

Dutch, Frisian, and High German, as well as a large number of non-standard
varieties, the more familiar of which are the dialects spoken in Belgium and the
South of the Netherlands (Flemish, Brabantish, Limburgian), Northern
Germany (Low German), the Rhine Valley (Luxemburgish), South-Eastern
Germany and Austria (e.g., Bavarian), and Switzerland (Swiss German). In this
paper, both the standard and the non-standard varieties will be referred to as di
alects.
All these dialects differ from English in having the finite verb occupy the
position after the first constituent in main clauses (a property the Continental
West Germanic dialects share with the North Germanic dialects), and differ from
both English and North Germanic in having the verb follow its noun phrase
complement in embedded clauses and infinitival constructions. The latter prop
erty is illustrated in (1) for Dutch:
(1) a. (Ik wil) dat hij het boek leest

I want that he the book reads
"I want him to read the book."
b. * (Ik wil) dat hij leest het boek
c. (Hij wil) het boek lezen
He wants the book read-INF(INITIVE)
"He wants to read the book."
d. * (Hij wil) lezen het boek
230 JAN-WOUTER ZWART
On the basis of the word order facts in (1), the Continental West Germanic
languages have been classified as OV-languages in the tradition of generative
syntax (Bach 1962, Koster 1975). The word order in (la) and (lc) was consid
ered to reflect the deep structure ordering of the meaningful elements, the SVO
order of main clauses being derived by movement of the finite verb to the second
position in the sentence (see also den Besten 1983). This analysis presupposes a
phrase structural split dividing the Germanic languages into English and North
Germanic on the one side, and Continental West Germanic on the other. The
split was related to the position of the head in the phrase: North Germanic and
English would have the structure in (2), Continental West Germanic the one in
(3):
Recently, however, it has been argued that phrases in all Germanic lan
guages, including Continental West Germanic, should be construed as in (2)
(Kayne 1994, Zwart 1994). This presupposes that the word order in (la) and
(lc) does not immediately reflect the deep structure ordering of the meaningful
elements. Rather, (la) and (lc) may be derived by movement of the object noun
phrase to the left, as demonstrated by Vanden Wyngaerd (1989). This does not
affect the explanation for the SVO word order in main clauses, which is derived
from the embedded clause word order by the verb movement identified by
Koster (1975) and den Besten (1983).
Immediate evidence for the object noun phrase movement to the left is pre
sented by the phenomenon that the object noun phrase and the verb in OV-con-
structions need not be adjacent:
(4) a. (Ik wil) dat Jan het boek snel leest

I want that John the book quickly reads
"I want John to read the book quickly."
VERB CLUSTERS IN WEST GERMANIC DIALECTS 231
b. (Ik wil) het boek snel lezen

I want the book quickly read-INF
"I want to read the book quickly."
The phenomenon in (4), often referred to as scrambling or object shift, is

present in all Continental West Germanic dialects. Assuming that the object noun
phrase must be generated as a sister of the verb, the OV-order in (4) presents no
indication of the deep structure ordering of the verb phrase. Neither, then, does
the OV-order in (la) and (lb) (if we assume, as seems reasonable, that the
grammatical factor triggering object shift in (4) may be present in (1) as well).
If the possibility of noun phrase movement makes the word order facts in (1)
irrelevant for the question whether phrases in Continental West Germanic adhere
to the structure in (2) or to the one in (3), the only way to study this question is
by looking at other instances of complementation to the verb.
There are two cases to consider here. One case is presented by finite com
plement clauses. These invariably follow the verb, also in embedded clauses and
infinitival constructions:
(5) a. (Hij denkt) dat ik wil dat hij het boek leest
He thinks that I want that he the book reads
"He thinks that I want him to read the book."
b. * (Hij denkt) dat ik dat hij het boek leest wil
c. (Je moet) willen dat hij het boek leest
You must want that he the book reads
"You've got to want him to read the book."
d. * (Je moet) dat hij het boek leest willen
In Zwart (1994), it is argued that the word order in (5a) and (5c) does reflect
the basic ordering of the meaningful elements in the Dutch VP. This is because
the factor triggering noun phrase movement in (1) and (4)—Case assignment by
assumption (see Vanden Wyngaerd 1989)—has no effect on clauses (which are
not assigned Case). This deviates from the traditional analysis of (5), in which
the clause is assumed to move to the right, by a process called extraposition
(Reuland 1981). Extraposition, however, has a number of curious properties,
leading Kayne (1994) to conclude that this movement process actually does not
exist.
In this paper, we will discuss the remaining type of verbal complementation
to be studied in connection with the position of the V in the verb phrase in
Continental West Germanic: complementation by a verbal or infinitival con
stituent. This type of complementation typically gives rise to the construction of
verbal clusters in Continental West Germanic (cf. Evers 1975). The order of the
verbs in the verbal cluster shows a bewildering variation across Continental
West Germanic dialects. It will be argued that this variation can best be analyzed
by assuming that all phrases involved are structured as in (2).
Having established this, the paper explores the possibilities of participle
placement in Continental West Germanic, based on the analysis of auxiliary
constructions as possessive constructions proposed in Kayne (1993).
2. Verb clusters in Continental West Germanic
The Continental West Germanic dialects show a large variety of verb clus
ters, especially in embedded clauses, where the verb movement which puts the
finite verb in the second position does not apply. The most elementary cases are
those in which an auxiliary verb has a past participle in its complement domain,
or when a modal, causative, or perception verb has an infinitive in its comple
ment domain. In these situations, clusters are created that consist of two verbs
only. More complex clusters arise by iteration of the processes that give rise to
these simple clusters. (The infinitives in the complement domain of raising and
control verbs, generally marked by a prefix cognate of EngUsh to, do not appear
to give rise to cluster formation, and will be kept out of the discussion.)
Examples of simple clusters are given in (6)-(8), from Standard Dutch. The
organization of the clusters is represented numerically on the extreme right,
where a verb β originating in the complement domain of a verb α gets a higher
number than α (so if α = 1, β = 2 or higher):
(6) a. (Ik denk) dat Jan het boek heeft gelezen 1 -2

I think that John the book has read-PART(ICIPLE)
"I think that John has read the book."
b. (Ik denk) dat Jan het boek gelezen heeft 2-1
(7) a. (Ik denk) dat Jan het boek kan lezen 1-2
I think that John the book can read-INF
"I think that John is capable of reading the book."
b. (Ik denk) dat Jan het boek lezen kan 2-1
(8) a. (Ik denk) dat ik Jan het boek laat/zie lezen 1 -2

I think that I John the book let/see read-iNF
"I think that I'm letting/seeing John read the book."
b. (Ik denk) dat ik Jan het boek lezen laat/zie 2-1
The 1-2 order in (6a) is prominent in written Dutch, whereas the 2-1 order in
(6b) is prominent in spoken Dutch (Stroop 1970). In (7) and (8), the 1-2 order
is slightly favored in both written and spoken Dutch, but the 2-1 order is not im
possible.
Standard Dutch is presumably a composite of several systems present in the
various dialects of Dutch (Stroop 1970). The freedom of word order in the verb
clusters in (6)-(8) is generally absent from the dialects. Nevertheless, both the 1-
2 order and the 2-1 order are represented in the dialects, so that (6)-(8) is a
proper representation of the word order possibilities in Continental West
Germanic verb clusters in general.
Looking at the dialects of Dutch, it appears that in the Southern dialects, the
participle has a tendency to precede the auxiliary, as in (6b) (Verhasselt
1961:153, Vanacker 1969, but cf. Shepherd 1946:61 on the dialect of
Maastricht, who gives examples of the 1-2 order only). This tendency is less
strong in the dialects spoken in the Eastern part of the Netherlands (Stroop
1970:250), whereas the dialects spoken in the North appear to employ the 2-1
order exclusively. The latter is also true of Frisian, High German, and the di
alects of German, as far as I have been able to ascertain (Bruch 1973:93 men
tions some Luxemburgish vestiges of the 1-2 order, which was a possibility of
Middle High German).
The 2-1 order in (7)-(8), where the finite verb has an infinitive in its com
plement domain, is used exclusively in High German, Frisian, and the dialects
spoken in the North of the Netherlands. However, many German dialects show
the 1-2 order in this case (Bruch 1973:94 on Luxemburgish, Baur 1988:157 and
Lötscher 1978:4 on Swiss German), which is also used very prominently in the
Dutch dialects spoken in Belgium, Limburg, and the dialects spoken in the West
and the South of the Netherlands. In the East of the Netherlands, a mixed situa
tion seems to exist (Stroop 1970:254).
In more complex verb clusters, tendencies tend to become rule. Thus, in
Standard Dutch verb clusters, the general word order becomes 1-2-3 (see (9)),
while in High German and Frisian the general word order becomes 3-2-1 (see
Frisian (10), from Tiersma 1985:139):
(9) a. (Ik denk) dat Jan het boek moet kunnen lezen 1-2-3
I think that John the book must can-INF read-INF
"I think John must be capable of reading the book."
b. * (Ik denk) dat Jan het boek lezen kunnen moet 3-2-1
(10) a. ... wêrom 'tik de hiele dei sitten bliuwe moatten ha 4-3-2-1
why that I the whole day sit-INF stay-INF must-PART
have-FIN(ITE)
"... why I have had to remain sitting all day."
ե *... wêrom 't ik de hiele dei ha moatten bliuwe sitten 1-2-3-4
But many exceptions exist. In Standard Dutch, for instance, when the sec
ond verb is an auxiliary, the participle in its complement may show up in three
positions:
(11) a. (Ik denk) dat Jan het boek moet hebben gelezen 1-2-3
I think that John the book must have-INF read-PART
"I think John must have read the book."
b. (Ik denk) dat Jan het boek moet gelezen hebben 1-3-2
c. (Ik denk) dat Jan het boek gelezen moet hebben 3-1-2
The general tendency is confirmed, however, in that the order of the modal
and the auxiliary in the complement domain of the modal is fixed:
(11) d. ?? (Ik denk) dat Jan het boek gelezen hebben moet. 3-2-1
e. * (Ik denk) dat Jan het boek hebben gelezen moet. 2-3-1
f. * (Ik denk) dat Jan het boek hebben moet gelezen. 2-1-3
In Frisian, only the 3-2-1 order is allowed in this case, whereas High German
allows both the 3-2-1 order and the 1-3-2 order.
In clusters of three verbs, the first of which is an auxiliary, the verb in its
immediate complement domain often takes the shape of an infinitive, instead of
the expected participial shape. This phenomenon, referred to as the infìnitivus
pro participio or IPP-phenomenon is present in all those dialects of Continental
West Germanic in which the auxiliary precedes its complement in the verbal
cluster (Hoekstra 1994; the IPP effect is also attested in clusters in which the
participle-turned-infinitive and its complement verb both precede the auxiliary
(2-3-1), as in (35); see also Lange 1981, Hoeksema 1988, Haegeman 1995, and
references cited there). (12a) is from Standard Dutch, (12b) from High German:
(12) a. (Ik denk) dat Jan het boek heeft kunnen/*gekund lezen 1-2-3
I think that John the book has can-INF/pART read-INF
"I think John could have read the book."
b. (Ich glaube) daß J. das Buch hat lesen können/*gekonnt 1-3-2
I think that John the book has read-iNF can-INF/PART
(Notice that the modal is in the complement of the auxiliary, in spite of what
the English translation suggests.) German (12b) again deviates from the ex
pected 3-2-1 pattern, the 1-2 order setting up the context for the IPP-effect. In the
3-2-1 order, no IPP-effect occurs (van Dam 1972:146; another order in which
the IPP-effect shows up in verbal clusters in German dialects is the 3-1-2 order,
as in kommen hat können"come-INFhas -INF", again displaying the 1-2 or
der):
(13) (Ich glaube) daß ich Johann das Buch lesen

gesehen/*sehen habe 3-2-1
I think that I John the book read-INF seen-PART/INF have-FIN
"I think I saw John read the book."
If the third verb in a cluster is replaced by the combination of an auxiliary

and a participle, the 1-2-3 dialects (like Standard Dutch) show a cluster consist
ing of one participle, two infinitives, and the matrix auxiliary verb (see (14a),
where the placement of the participle is in fact as liberal as in (11)). Certain 3-2-
1 dialects, on the other hand, show a cluster containing two participles and one
infinitive next to the matrix auxiliary verb (see (14b), from Stellingwerfs,
(Bloemhoff 1979:37, cf. den Dikken & Hoekstra 1995)). This is because the
IPP-effect is absent in strict 3-2-1 dialects:
(14) a. (Ik dacht) dat Jan het boek gelezen had kunnen hebben 4-1-2-3
I thought that John the book read-PART had can-INF have-INF
"I thought John could have read the book."
b. ... omdat zi'j et wel es daon hebben kund had 4-3-2-1
because she it PARTICLE PARTICLE done-PART
have-iNF could-PART had
"... because she may very well have done it."
Other double participle constructions do not seem to involve two auxiliary

verbs. The following are examples from Dutch (15a), High German (15b) and
Swiss German (15c):
(15) a. (Ik wist niet) dat Jan ontslagen was (?geworden) 3-1-2
I knew not that Johnfired-PARTwas become-PART
"I did not know John had been fired."
b. Grad wo է abgfaare gsy bisch... 3-2-1
just when you taken-off-PART ЄЄN-PART are-FIN
"Just after you had left."
. (Ich möchte wissen) warum das Buch nicht

gelesen worden ist 3-2-1
I would-like know why the book not
read-PART become-PART is
"I would like to know why the book has not been read."
These constructions are curious, in that in all dialects, no matter what the
preferred word order in the verbal cluster, the most deeply embedded participle
has to precede the hierarchically higher participle. Thus, whereas the placement
of the participle is particularly Uberal in Dutch, (16) is completely ungrammatical
(cf. (15a)):
(16) (Ik wist niet) dat Jan was (*geworden) ontslagen 1-2-3
I knew not that John was ЄMЄ-PART fired-PART
"I didn't know John had been fired."
A final descriptive generalization about verb clusters in Continental West

Germanic dialects is the following (due to Zwart 1994). Verb clusters in a num
ber of dialects may be broken up by material properly belonging to the most
deeply embedded verb in the cluster. Such material can be a complement of the
verb, a particle belonging to the verb, a secondary predicate associated with the
verb, an adverb modifying the verb, a stranded preposition belonging to a com
plement or adjunct PP associated with the verb, etc. (see Vanacker 1970 for an
overview of the phenomena in Flemish dialects; see also Haegeman & van
Riemsdijk 1986, Haegeman 1994). The phenomenon, called Verb Projection
Raising in the generative literature, is illustrated in (17a), from East Flemish
(Vanacker 1970:145; der is extracted from the PP headed by voor), and (17b),
from Swiss German (Baur 1988:157; the example also shows the IPP-effect):
(17) a. We zullen der moeten voor zorgen 1..2-3

we will there must-INF for -INF
"We will have to take care of that."
b. (I bi stolz,) das і ha chöne über de see schwüme 1-2-3
I am proud that I have-FIN can-INF across the lake swim-INF
"I'm proud that I have been able to swim across the lake."
The generalization regarding Verb Projection Raising is that the material

breaking up the cluster must be situated to the left of the verb which the material
belongs to (in the sense just described). This implies that the phenomenon is ab
sent from those languages that keep to a strict 3-2-1 order in the verbal cluster.
Thus, it does not occur in Frisian and in the Northern Dutch dialects (though it
does in West Frisian, which behaves more like High German in this respect, cf.
Hoekstra 1994), and it shows up in High German only in the 1-3-2 cases
(again, the auxiliary is hierarchically superior to the modal, contrary to what the
English translation suggests):
(18) (Ich bin der Meinung) daß er das Buch hätte

genau durchsehen sollen 1-3-2
I am of-the opinion that he the book had
exactly through-look-iNF shall-INF
"I feel that he should have looked the book through carefully."
The Verb Projection Raising phenomenon in (17) suggests that, properly

speaking, the 'ascending' orders (such as 1-2, 1-2-3, 1-3-2, etc.) do not present
clusters, since there is a way for independent material to intervene between the
members of the string of verbs. If so, the term 'cluster' in this connection
should be understood in a pretheoretical sense, as referring to a string of verbs,
rather than to an adjunction structure of some sort (see also section 5, and Zwart
1995).
Though many more cases exist, this may suffice as a general survey of the
properties of verb clusters in Continental West Germanic.
3. Analysis of the verb clusters from an OV point of view
Let us now examine how the traditional analysis of Continental West

Germanic has approached the problem of how to derive the various word orders
in the verb clusters. Recall that this analysis starts from the hypothesis that the
verb phrase in Continental West Germanic is structured as in (3) (in contradis
tinction to the structure in (2), employed in English and North Germanic).
Assuming all the verbs to occupy their basic positions, one would expect in
variant 'descending' orders (3-2-1, 2-1, etc.). As we have seen, however, few
Continental West Germanic dialects employ 'descending' clusters uniquely.
Frisian and the dialects spoken in the North of the Netherlands appear to be the
mostrigidvarieties in this respect. All other dialects use either mixed orders (like
the High German 1-3-2 order, or some of the orders found in Dutch (11)), or
strictly ascending orders (like the Dutch construction in (9a), or actually use both
ascending and descending orders (like Dutch in (6) and Luxemburgish, which
has the 2-1 order with auxiliary-participle constructions and the 1-2 order with
modal-infinitive constructions). To account for this bewildering variation, then,

some movement processes have to be assumed.
In Evers' (1975) classic analysis, verb custers are created by moving em
bedded verbs to the right and adjoining them to the embedding verb. The ad
junction can take place to the left and to the right of the embedding verb.
Adjunction to the right yields the Dutch order in (6a), and adjunction to the left
yields the High German order and the order of Dutch (6b).
As argued in Zwart (1994), it is impossible in this approach to keep the di
rection of adjunction constant even within a single language. That is, the varia
tion in (6a) and (6b) must be expressed in terms of variation in the direction of
adjunction. Similarly, the 1-3-2 order of High German (12b) must be derived by
adjoining the most deeply embedded verb (3) to the left of the immediate higher
verb (2), followed by subsequent adjunction of the cluster 3-2 to the right of the
matrix verb (1).
In addition to the variation of the direction of adjunction, the analysis has to
allow for variation of the phrase structural status of the category which is ad
joined. This addition is needed to account for the Verb Projection Raising facts
(see (17)). Following den Besten & Edmondson (1983), it has been assumed
that these phenomena are the result of movement of (part of) the embedded verb
phrase to the right. This movement is followed by adjunction of the raised verb
projection, either to the right of the higher verb, or to the right of some projec
tion of the higher verb. The exact nature of this operation has been the subject of
much discussion, which I will not go into here (see Rutten 1991 for a survey).
(The fact, however, that Verb Projection Raising occurs both in 1-2-3 clusters,
as in (17), and in 1-3-2 clusters, as in (18), suggests that the direction of ad
junction in Verb Projection Raising is also subject to variation. In particular, the
l-ADV+3-2 order in (18) can only be derived by adjoining the ADV+3 phrase to
the left of the higher verb 2, followed by adjunction of the complex head
ADV+3+2 to the right of the matrix verb 1.)
This analysis is unsatisfactory for the following reasons:
a. There is no consistent direction of adjunction, either across

Continental West Germanic dialects, or even within particular
Continental West Germanic dialects;
b. There is nofixedphrase structure level of the category adjoined;
 It is not clear what triggers the various movements, in the sense
that there is no understanding how particular asymmetries are to
be explained (for instance the asymmetry between infinitives and
participles, the former adjoining to the left and the latter to the
right in a number of dialects (e.g., Luxemburgish.));
d. The parametrization makes no reference to the timing of the

movements (i.e., either in overt syntax or in covert syntax),
which is generally considered to be a major source of parametric
variation (see most recently Chomsky 1993);
e. It is unclear why in the Germanic SOV languages multi-verb con
structions show such a variety of word orders within the cluster,
whereas in the Germanic SVO languages multi-verb constructions
invariably show strictly 'ascending' orders;
f. It is unclear why certain phenomena (the IPP effect, Verb
Projection Raising) are sensitive to the surface order of the mem
bers of the verb cluster (see below, section 5);
g. The analysis relies on a phrase structural split among the
Germanic languages for which there is no independent empirical
basis (see section 1).
In the following, we will approach the questions of word order in the

Continental West Germanic verb cluster from ¿he perspective of Kayne (1994)
and Zwart (1994), according to which verb phrases in all Germanic languages
have the structure in (2).
4. Analysis of the verb clusters from a VO point of view
If the verb phrase in Continental West Germanic is structured as in (2), em

bedded verbs are generated in a position to therightof embedding verbs. As be
fore, it is assumed that the noun phrase complement of a verb is moved to the
left. In cases of multi-verb constructions, this implies that the complement of the
most deeply embedded verb will appear to the left of the verb cluster, as in (19b)
(see below for movement to a licensing position in between the auxiliary and the
participle, yielding the Verb Projection Raising effect):
(19) a. ... dat Jan heeft gelezen [het boek] (cf. 6)
that John has read-PART the book
b. ... dat Jan [het boek]i heeft gelezen ti
In Zwart (1994:399), it is argued that a similar process applies to embedded

predicates:
(20) a. ... dat Jan heeft gelezen [het boek] [uit]
that John has read-PART the book out
"... that John finished the book."
b. ... dat Jan [het boek]i [uit]j heeft gelezen ti tj
These elements, then, will not normally interfere with the verb movements
giving rise to verb clustering.
There are certain restrictions on the movement of the object and the embed
ded predicate in (19) and (20). For example, the movements must give rise to
crossing paths, rather than nesting paths:
(21 ) *... dat Jan [uit]j [het boek]i heeft gelezen ti tj (cf. 20)
Assuming, with Chomsky (1993), that the relevant movements are directed
to the specifier position of a functional projection, (21) can be excluded if the
functional projections have to be ordered in some way.
Such an ordering, however, in principle leaves open the possibility of gen
erating the relevant functional projections in various positions. This is not ex
cluded, as long as the result does not violate the restrictions on the ordering of
the functional projections involved.
More concretely, if the functional projection that hosts the object must pre
cede the functional projection that hosts the embedded predicate, it is not ex
cluded that the latter functional projection is generated somewhere between the
auxiliary and the participle. This would yield the word order in (22):
(22) ... dat Jan [het boek]i heeft [uit]j gelezen ti tj
Although Standard Dutch appears to prefer (20b) over (22), constructions

like (22) are found in several Continental West Germanic dialects. (22), for ex
ample, instantiates the Verb Projection Raising phenomenon, illustrated in (17).
This approach to Verb Projection Raising, involving licensing positions in be
tween the verbs of the 'verb cluster', is first found in Kaan (1992:111).
Thus, it appears that, in the VOapproach, Verb Projection Raising can be
described independently of the verb movement that gives rise to the formation of
verb clusters. This eliminates one of the problems mentioned above in connec
tion with the assumption that the VP in Continental West Germanic is organized
as in (3) (i.e., with the head following its complement). Under this assumption,
the Verb Projection Raising facts (cf. (17)) can only be derived by moving (part
of) the verb phrase to the right. Starting from the structure in (2), with the head
preceding its complement, there is no need to allow for variation of the phrase
structure level of the category moved. In other words, we can maintain that verb
clustering never involves movement of verb projections. This significantly re
duces the range of variation that the system of grammar allows. (In section 5,
we will argue that participles move as phrases to a specifier position. But, unlike
the Verb Projection Raising proposal, this does not increase the number of pos
sible movements, as the participle movement is of the well-known type of move
ment of a phrase to a specifier position.)
A further restriction on Verb Projection Raising is that the material breaking
up the cluster (such as uit in (22)) cannot be placed to the right of the verb which
it properly belongs to. Thus, in (22) uit "out (i.e., finished)" is the predicate of a
Small Clause het boek uit "the book out", occurring in the complement of the
verb gelezen "read", In this sense, uit belongs to gelezen (cf. Neeleman 1994,
who analyzes the combination of uit and gelezen as a complex predicate). We
know that gelezen may appear both to the right and to the left of the auxiliary
heeft (cf. (6)). But when gelezen appears to the left of heeft, so must uit:
(23) a. ... dat Jan [het boek]i [uit]j gelezenk heeft tk ti tj
that John the book out read-PART has
b. *... dat Jan [het boek]¡ gelezenk heeft [uit]j tk ti tj
c. *... dat Jan [het boek]i gelezenk [uitį heeft tk ti tj
In Standard Dutch, (23b) and (23c) contrast sharply with the marginally ac
ceptable (22). In Verb Projection Raising dialects like West Flemish and Swiss
German, examples like (23b) and (23c) are never found.
Apparently, the intervening material in Verb Projection Raising constructions
must be construed (in a pretheoretical sense) with the verb it belongs to.
'Construing with' is apparently sensitive to directionality, as is also suggested in
Kayne (1994). Below, we will provide a more technical description of
'construing with' in Verb Projection Raising constructions.
So far, we have seen that the VO-approach makes it unnecessary to refer to
the phrase structural status of the category moved (in the process of construing
of verb clusters). It does not seem to be the case that there is more going on than
object movement, predicate movement, and verb movement.
Let us next consider the question of the direction of adjunction. In the OV-
approach, the variation in (6), reflecting the range of possibilities in two-verb
clusters in Continental West Germanic, is analyzed as involving a choice be
tween adjunction to the right and adjunction to the left. As can be seen, there is
no way of keeping the direction of adjunction constant, even within a single lan
guage (e.g., Standard Dutch and High German).
As a first approximation, we could propose that the order in (6a) results
from there being no movement at all, while the order in (6b) is the result of ad
junction of the participle to the left of the auxiliary:
(24) a. ... [het boek]j... [VP heeft [ү gelezen tj]] = (6a)
b. ... [het boek] j ... [VP gelezeni-heeft [ү ti tj]] =(6b)
In other words, the variation in (6) is the result of the absence vs. presence
of a particular movement (or, in Chomsky's (1993) framework, of the timing of
the movement of the participle, taking place before Spell-Out in (24b), and after
Spell-Out in (24a)). Similarly, the variation in (7) and (8) could be taken to re
sult from the absence vs. presence of raising and adjunction of the infinitive to
the modal, perceptive, or causative verb.
This approach, which was taken in Zwart (1993), has the immediate advan
tage that no language or construction specific statements need to be made about
the direction of adjunction. It is also in accordance with Kayne's (1994) obser
vation that adjunction seems to always take place on the left.
A similar approach could not be as successful if we were to start from the
OV-structure in (3). In principle, the 2-1 orders in (6a), (7a), and (8a) could be
obtained by abstaining from verb movement (instead of raising of the embedded
verb and adjoining it to the left of the embedding verb):
(25) a. [het boek]j [ү [ү tj ti ] heeft-gelezeni ] = (6a)
b. [het boek]j [ү [VP tj gelezen ] heeft ] = (6b)
But this generates several problems. For one thing, the Verb Projection
Raising order heeft uit gelezen must be derived by raising a (part of a) verb
phrase, as discussed above. For another, it is unclear why the 2-1 cluster is im
penetrable in all Continental West Germanic dialects. Finally, this would lead to
problems if clusters with more than two verbs are taken into account.
Consider, for instance, the three-verb clusters in the sentences (11) from
Standard Dutch. Allowed are the orders 1-2-3, 1-3-2, and 3-1-2. Very marginal
is 3-2-1, and unacceptable are 2-3-1 and 2-1-3. From the VO-perspective, it
looks like the basic 1-2 order is fixed (i.e., the auxiliary always follows the
modal verb), while the 3 verb (the participle) can choose to not move at all,
move part way, or move all the way:
(26) . [VP moet [ү hebben [VP gelezen ]]] =(lla)
b. [ү moet [ү gelezenrhebben [VP ti ]]] =(11b)
c. [ү gelezenrmoet [ү hebben [VP ti ]]] = (11c)
If necessary, the 3-2-1 order of (11d) can be derived by adjoining the com
plex gelezen-hebben in (26b) to the left of the matrix verb moet:
(26) d. [VP [gelezen-hebben]j-moet [ү tj [ү ti ]]] =(11d)
Since 3-2-1 clusters appear in various Continental West Germanic dialects,

such a derivation may have to be allowed. The remaining two orders, which are
ungrammatical in Standard Dutch, are harder to derive. The only way to derive
the 2-3-1 order in (1 le) seems to be to move the verb phrase to the left:
(26) . [ү [ү hebben [ү gelezen ]]j moet tj ] = (1 le)
This movement clearly deviates from the ones considered up to now. (11e)
appears to be excluded in Continental West Germanic, although the 2-3-1 order
does occur in several dialects when the 1 verb is an auxiliary (e.g., in West
Flemish zien gebeuren heęft [see happen-INFINITIVE has] "has seen happen"). The
process by which (26e) is derived, then, cannot entirely be excluded, but it does
not seem to apply in constructions in which the 1 verb is a modal verb (see
section 5 and Zwart 1995 for further discussion of this asymmetry between
modal verbs and auxiliaries). Finally, the 2-1-3 order can only be derived by
moving the middle verb out to the left:
(26) f. [VP hebbenj-moet[V tj [ү gelezen]]] =(11f)
The extremely sharp ungrammaticality of (11f), across all Continental West

Germanic, as far as I have been able to check, is clearly reminiscent of the un-
grammaticality of (23b) (where the embedded predicate appears to be left
stranded). This suggests that for licensing purposes the participle is dependent
on the auxiliary in the same way as the embedded predicate (or another element
in the relevant complement domain) is on the verb selecting it. The generalization
then seems to be that the participle must be licensed somewhere to the left of the
auxiliary, which leaves the 1-2 order in (6a) and the 1-2-3 order in (1la) a very
curious exception indeed. Nevertheless, the parallellism between (23b) and (11f)
strongly suggests that this generalization is correct, and that some additional ex
planation is needed to account for the final position of the participle in (6a),
(11a), and many other constructions across Continental West Germanic.
On the OV-approach, the general picture of the participle 3 moving gradually
leftward (yielding the 1-2-3, 1-3-2, and 3-1-2 orders of (lla-)) is lost. Three
entirely different derivations are needed to reach the three fully grammatical
word orders in (lla-). The 1-2-3 order is derived by adjoining the participle 3
to the right of the auxiliary 2, followed by adjunction of the auxiliary-participle
cluster to the right of the matrix verb 1 (27a). The 1-3-2 order is derived by ad-
joining the participle 3 to the left of the auxiliary 2, followed again by adjunction
of the participle-auxiliary cluster to therightof the modal verb 1 (27b). Finally,
the 3-1-2 order is derived by leaving the participle 3 in its place, while moving
the auxiliary 2 out to theright(27c):
(27) . [VP [VP [VP ti ] tj ] moet-[hebben-gelezeni] = ()

b· [vp [vp [vp tj ] tj ] moet-[gelezenrhebben]j ] = (11b)
c. tvp [VP tvp gelezen ] tj ] moet-hebbenj ] = (1 le)
Again, it is unclear why the cluster in (27a) is penetrable in several

Continental West Germanic dialects, as is illustrated in (28) (cf. (22)):
(28) a. ... dat Jan het boek moet hebben uit gelezen
that John the book must-FIN have-INF out read-PART
"... that John must have finished the book."
b. ... dat Jan het boek moet uit hebben gelezen
As (28b) shows, the penetration of the cluster cannot be the result of some
complex uit-gelezen moving and adjoining to the auxiliary as a single participle.
Such an analysis is generally not feasible in Verb Projection Raising construc
tions, since the material breaking up the cluster may very well be a complete
phrase.
Returning to the question of how to derive the various orderings in (11)
from an OV-point of view, the 3-2-1 ordering of (11d) is unproblematic. It can
be derived by abstainingfrommovement entirely:
(27) d. [VP [VP [VP gelezen ] hebben ] moet ] = (1 ld)
But of the remaining two orders, the first cannot be blocked without utter
stipulation. Thus, the 2-3-1 order of (11e) could be derived by adjoining the
participle 3 to the auxiliary 2, as in (27a), and by abstaining from further move
ment (27e). It is entirely unclear what blocks this process (apart from the stipu
lation that once the movement process has started, it has to go all the way). The
completely ungrammatical 2-1-3 order of (11f) is derivable by moving the par
ticiple 3 all the way to therightand adjoining it to the modal 1 (27f):
(27) . [ү [ү [үtj] hebben-gelezeni ] moet ] = (1le)

f· [vp [VP[vptj]hebben ] moet-gelezeni ] = (11f)
This derivation can be blocked by some version of the Head Movement

Constraint, which does not allow a head to move across a governor (Travis
1984). There is a way, however, of deriving (11f) while adhering to the Head
Movement Constraint, namely by adjoining the auxiliary 2 to the left of the
modal verb 1, followed by adjunction of the participle 3 to the right of the auxil
iary-modal complex. The definition of the Head Movement Constraint ensures
that in this case, the trace of the auxiliary verb does not count as an intervening
governor (see Baker 1988):
(27) f '. [VP [VP [VP ti ] tj ] [hebbenj-moet]-gelezeni ] = (11f)
Given the possibility of adjunction to the left (as needed in the derivation of
(27b)), this derivation should in principle be allowed. Yet (11f) is hopelessly
ungrammatical in all dialects of Continental West Germanic.
In conclusion, the various word orders discussed here can be derived in a
simpler and more restrictive way if it is assumed that the verb phrase in
Continental West Germanic is structured as in (2).
5. Patterns of optional movement
The analysis of the Continental West Germanic verb clusters from a VO-
point of view, as presented in section 4, has one unsatisfactory aspect. The phe
nomenon that in certain dialects both the auxiliary-participle order and the par
ticiple-auxiliary order are possible is described in terms of optional movement of
the participle to the left. But the ungrammaticality of (11f) suggests that such
optionality in fact does not exist (see the discussion around (26f)). (As
Chomsky 1993 points out, optionality ought not to be part of a system of gram
mar. However, it cannot be excluded that the optionality in (6) in fact reflects a
limited form of bilingualism among speakers of the relevant dialect—perhaps not
accidentally a standard dialect. (11f), however, strongly suggests that there is
always participle movement in Continental West Germanic.)
If participle movement is optional, movement of the auxiliary to the left of
the modal should not interfere with the placement of the participle. Yet this is
what happens, as the following paradigm shows:
(29) a. ... dat Jan kan komen

that John can-FIΝ come-INF
b. ... dat Jan korneni kan ti
(30) a. ,.. dat Jan het boek kan hebben gelezen

that John the book can-FIN have-INF read-PART
"... that John may have read the book."
b. *... dat Jan het boek hebbeni kan ti gelezen
Hence, participle movement cannot be optional. The ungrammaticality of

(30b) shows that the participle must always be licensed in close proximity to the
auxiliary. The optionality underlying (6) must relate to the position the participle
is moved to: either preceding or following the auxiliary. When the auxiliary
moves, only the position preceding the auxiliary remains as a potential landing
site for the participle.
At this point, it becomes necessary to look into the structure of auxiliary-
participle constructions in more detail. The use of a verb referring to possession
to mark the past tense in the Indo-European languages is too striking to be over
looked (cf. Vendryes 1937). Kayne (1993) therefore proposes to analyze auxil
iary-participle constructions as possessive constructions.
Possessive verbs like have are often considered to be a composite of be and
a functional element (mostly a preposition) (see e.g., Benveniste 1966, Freeze
1992). Following Kayne (1993), I will assume that the auxiliary have should
also be treated as a composite of two heads, which I will call BE and OF. Each of
these heads projects a (verb) phrase, yielding two specifier positions. The com
plement of the lower head, OF, is a Small Clause-like structure, consisting of an
Agreement Phrase (as in Kayne 1993) and a lexical projection in the complement
of AGR (slightly deviating from Kayne 1993):
I assume that in possessive constructions, like (32a), the lexical projection

XP in (31) is an NP, whereas in auxiliary-participle constructions, like (32b), XP
equals VP. Finally, in constructions containing a secondary predicate, like (32c),

XP is a Small Clause:
(32) a. ... dat Jan [een boek]¡ heeft ti

that John a book has
"... that John has a book."
b. ... dat Jan [het boek]i heeft [VP gelezen ti ]
that John the book has read-PART
"... that John has read the book."
 ... dat Jan [het boek]i [uit]j heeft[SCtitj]
that John the book out has
"... that John has finished the book."
The movement of the object het boek in (32) (in fact, a Small Clause subject
in (32c)) and of the predicate uit in (32c) takes place under the conditions dis
cussed in section 4. The landing site of these elements is generally located to the
left of the BE-OF composite have. (We will return to Verb Projection Raising
below.)
The representation in (32b) has the participle gelezen inside the VP embedded
under OF. But, as we have seen, the participle must also move to the left. When
we zoom in on the construction in (32b), we find that the structure must be
something like (33):
(33) ... [het boek]i [VP-- BE [VP- OF [AGRP [VP gelezen ti ]]]]
In (33), OF incorporates into BE , yielding have. Have being a composite,

there are two specifier positions associated with have, indicated by the double
hyphens in (33). I would like to propose now that the participle gelezen can be
licensed in each of these two specifier positions. This yields either (34a) or
(34b):
(34) a. ... [het boek]i [VP-- BE [VP- OF [AGRP [VP gelezen ti ]]]]
b. ... [het boek] i [VP g e l e z e n BE [VP - OF [ AGR p[ VP tj tį ]]]]
Assuming now that BE +OF (=have) is spelled out in the position of BE, (34)
immediately yields the two possible word orders in (6):
(6) a. ... dat Jan het boek heeft gelezen = (34a)

that John the book has read-PART
b. ... dat Jan het boek gelezen heeft = (34b)

that John the book read-PART has
(The optionality of moving the participle to either the specifier position of OF

or the specifier position of BE can be derived from Chomsky's (1993:17)
Equidistance Principle. As a result of the incorporation of OF into BE , both
specifier positions are in the minimal domain of the composite BE +OF, hence,
according to Chomsky, equidistant from any position lower in the tree.)
The variation among the Continental West Germanic languages with respect
to the position of the participle can now be described in terms of which of the
two specifier positions in (33) may be occupied by the participle.
This analysis differs from the one entertained in section 4, in that participle
movement was described as adjunction to a head in section 4, and as movement
to a specifier position in this section. There is actually evidence that the auxiliary
licenses its complement (i.e., the participle) in a specifier position, rather than
via head-adjunction. This evidence is based on a curious construction from West
Flemish, discussed in Hoekstra (1994), den Dikken (1994), and Haegeman
(1995). This construction shows the order 2-3-1, where 1 is a form of the auxil
iary have ((35a) from personal observation, (35b) quoted from Liliane
Haegeman, p.c., in den Dikken 1994:83):
(35) a. ... da Jan 't zien gebeuren eet 2-3-1

that John it see-INF happen-INF has
"... that John saw it happen."
b. ... da Valère zou willen dienen boek kuopen een 1-3-4-2
that Valery should want-INF that book buy-INF have-INF
"... that Valery would have wanted to buy that book."
In (35a), the cluster of infinitives zien-gebeuren has apparently been moved

to the left of the auxiliary eet. This could in principle be an instance of head
movement. But (35b) shows that it is not. In (35b), the preposed cluster, willen-
kopen, is broken up by the object dienen boek, expressing the internal argument
of the most deeply embedded verb. This, then, is an instance of the Verb
Projection Raising phenomenon, suggesting that a licensing position for the di
rect object is created to the immediate left of the verb kuopen (see section 2, and
the discussion around (22) above). This licensing position must be the specifier
position of a functional projection, leading to the conclusion that the combination
dienen boek kuopen is a phrase rather than a head. (It will be assumed without
demonstration here that the construction in (35b) does not have the properties of
noun incorporation constructions.) Hence, willen-dienen boek kopen must be a
phrase as well, moving around the auxiliary to what must be analyzed as a spec
ifier position. Generalizing this result, it must be the case that auxiliaries license
the verbs in their complement in a specifier position.
(This might lead to the conclusion that what is moved in (34) is actually not
the participle itself, but the VP containing the participle as a whole. This, how
ever leads to a problem in cases where the participle has a clausal complement.
Contrary to what one would expect under the scenario under consideration here,
the clausal complement does not appear to the left of the auxiliary, as pointed out
to me by Daniel Biiring (p.c.). I will not discuss this issue here, leaving the
phrase structure status of the participle in (34) open.)
The observation that participles move to a specifier position is not easily ac
commodated under the assumption that the VP in Continental West Germanic is
structured as in (3). Since specifiers do not appear to the right, participle-final
orders can only be described as the result of head-adjunction under that ap
proach. As a further disadvantage of the OV-approach, it should be mentioned
that there is no way of relating the optionality of participle placement to the pos
sessive structure of auxiliary constructions as argued for in Kayne (1993). To be
more precise, movement to each of the two participle constructions would al
ways yield a participle-auxiliary order, as (36) shows:
(36) . [VP-- [VP gelezeni [VP ti 1 OF] BE ]

b. [VP gelezen i [ү - [VP ti] OF ] BE ]
This should count as a further argument against the analysis of the

Continental West Germanic dialects as headfinallanguages.
As the facts in (1 la-) show, the optionality in the placement of the participle
is not exhausted by the two variants yielded by (34). In particular, the participle
may appear at the far left of a multi-verb cluster, and at various positions in be
tween if the cluster contains more than three verbs. These possibilities are illus
trated in (37).
(37) a. .., dat Jan het boek gelezen moet hebben = (11c)
that John the book read-PART must-FIN have-INF
"... that John must have read the book."
b. ?... dat Jan het boek moet gelezen kunnen hebben 1-4-2-3
that John the book must-FiN read-PART can-INF have-INF
"... that John must have been able to read the book."
The 1-2-3 and 1-3-2 orders of (11a-b), not repeated here, can now be de
scribed as the immediate result of the participle movement illustrated in (34). The
participle occupies either the lower or the higher specifier position associated
with the composite have. But this leaves the 3-1-2 order in (37a)/(llc) unac
counted for. The same can be said about the 3-2-1 order in (11d), marginal in
Dutch, but grammatical in several Continental West Germanic dialects (High
German, among others).
In a pretheoretic sense, we could say that the modal verb moet in (37a)
'takes over' the licensing of the participle from the auxiliary hebben. Similarly
for kunnen in (37b). A first approximation therefore could be to assume that in
(37) the participle is licensed in the specifier position associated with the modal
verb.
This analysis cannot work if the modal verb itself needs the specifier posi
tion involved to license the verb in its immediate complement domain (i.e.,
hebben in (37)). However, it is not clear whether the infinitive in the comple
ment domain of the modal is licensed by movement to a specifier position or by
head-adjunction to the modal. Constructions like the ones in (35), showing that
participles are licensed in a specifier position, are typically absent from
Continental West Germanic dialects when the 1 verb is a modal verb (as far as I
have been able to ascertain). (See Zwart 1995 for more discussion of the differ
ence between verb clusters involving modal verbs and verb clusters involving
auxiliaries, in particular in relation to the Stellingwerf dialect.)
Moreover, on the basis of the pair in (38), one would expect to find the pah
in (39) also, if the complement of the modal were to move as a phrase:
(38) a. ... dat Jan het boek moet lezen

that John the book must-™ read-INF
"... that John must read the book."
b. ... dat Jan het boek lezen moet
(39) a. ... dat Jan het boek moet hebben gelezen = (1 la)
that John the book must-FIN have-INF read-PART
"... that John must have read the book."
b. *... dat Jan het boek [hebben gelezen] moet = (11e)
But the 2-3-1 order in (39b) is apparently only found where 1 is an auxiliary
instead of a modal. This leads me to conclude that infinitives, if they move, ad
join to a head, instead of moving to a specifier position.
Following a suggestion by Eric Hoekstra, I will assume that a modal may
'take over' from an infinitival auxiliary (i.e., license the participle in its specifier
position) on the basis of the licensing relation that exists between the modal and
the infinitive. Even if the infinitive does not move to the modal overtly, as in
(37), the licensing relation between the two verbs must be said to exist. In terms
of Chomsky (1993), we may assume that the infinitive adjoins to the modal in
the hidden component of syntax LF. In a more representational approach, as ad
vocated by Groat & O'Neil (1994), we may assume that the infinitive has in fact
adjoined to the modal in (37), leaving a copy in its original position behind. The
language may then choose which of the copies of the infinitive to spell out:
(40) [VP gelezeni (hebbenj)-moet [VP (hebbenj [VP ti ]]]
Spelling out the higher copy yields the High German order 3-2-1, spelling
out the lower copy yields the Standard Dutch order 3-1-2.
We may now assume that the auxiliary hebben transfers its capacity to li
cense the participle in a specifier position to its sister in the adjunction configu
ration in (40). In this way, the modal can take over from the auxiliary. The op
tionality of this process (attested by the possibility of the various word orders in
(lla-)) remains unaccounted for under this approach.
This analysis of the word order possibilities in (37) carries over to the Dutch
constructions with several infinitivals, where the participle is allowed to appear
to the extreme left of the cluster:
(41) ... dat Jan het boek gelezen zou moeten kunnen hebben 5-1-2-3-4
that John the book read-PART should-FIN must-INF
can-INF have-INF
"... that John should have been able to read the book."
In the covert part of the structure underlying (41), hebben is adjoined to

kunnen, the cluster kunnen-hebben is adjoined to moeten, and the cluster kun-
nen-hebben-moeten is adjoined to zou. The capacity to license the participle in a
specifier position is transferred with each adjunction.
This analysis of patterns like (37) and (41) is supported if we consider cases
where the higher copy of the adjoined infinitive (cf. (40)) is actually spelled out.
In those cases, only the specifier position of the modal (as we have analyzed it)
is available as a licensing position for the participle. In other words, the option
of selecting the specifier positions associated with hebben itself disappears when
hebben is spelled out in the adjoined position in (40). This can be seen from the
radical ungrammaticality, across all Continental West Germanic, of the 2-1-3 or
der in (11f) (analyzed in (42)):
(11) f. *... dat Jan het boek hebben moet gelezen 2-1-3
that John the book have-INF must-FIN read-PART
(42) . [VP hebbenj-moet [VP --tj [VP gelezeni (OF)[AGRPti]]]]
b. [VP hebbenj-moet [VP gelezeni tj [VP -- (OF) [AGRPti ]]]]

The ungrammaticality of (1 If) leads us to conclude that when the adjunction
of the infinitive auxiliary to the modal verb is overt, the auxiliary can no longer
license the participle itself, but has to transfer its capacity to license to its sister,
the modal verb. (In other words, in (1 la) and (1 lb), the covert adjunction of the
auxiliary to the modal can be ignored, but the overt adjunction in (11f) can not.)
This analysis of the possible word orders in (1 la-) finds an obvious parallel
in the analysis of Verb Projection Raising constructions. As was illustrated in
(23), the material breaking up the verb cluster cannot be situated to the left of the
verb which this material properly belongs to (see (43a)). It can however be situ
ated further to the left than expected:
(43) a. <.. dat Jan het boek (uit) gelezen (*uit) moet (*uit) hebben 3-1-2
that John the book (out) read-PART (out)
must-FIN (out) have-INF
b. ... dat Jan het boek (uit) moet (uit) hebben (uit) gelezen 1-2-3
that John the book (out) must-FIN (out)
have-INF (out) read-PART
Uit being a predicate in the complement domain of gelezen, the instructions
to create a functional projection for licensing uit must derive from gelezen.
Apparently, however, it is not necessary to create the relevant functional projec
tion in the immediate vicinity of gelezen (see (43b)). This we can describe along
the same lines as we described participle placement above (mutatis mutandis).
Two differences between the placement of 'Verb Projection Raising-mate-
riaľ (such as uit in (43)) and participles must be taken into account. First, a
predicate like uit may have to be licensed in the specifier position of a functional
head, rather than in the specifier position of a verb (see also Zwart 1994). This
is because the specifier position of a verb is used for the licensing of embedded
verbs, in particular participles. Secondly, it is unclear whether a participle, when
it is moved to the specifier position of a hierarchically higher verb (whether OF,
BE, or a modal verb) is still able to create the functional projection needed for li
censing the predicate embedded in its complement domain.
I would like to explore the possibility that the participle, when moved to a li
censing position, transfers the capacity to project a phrase for licensing the em-
bedded predicate to the head with which the participle is in construction. In

(43a), for example, this is moet, and in (43b), the OF-head (cf. (40)). The ca
pacity to project a licensing phrase now resides with a verb in a head position.
Now the parallel with the analysis of participle placement becomes apparent: in
multiple verb constructions, the verbs adjoin to each other, transferring the ca
pacity to project a phrase for the embedded predicate with each adjunction.
As before, the overt position of the verb in which the capacity to license
originates (in this case, the participle) determines the range of licensing posi
tions. In (43b), the licensing phrase for uit can be projected from each verbal
position, based on the covert syntactic representation in which all infinitives are
adjoined to each other. In (43a), the licensing phrase for uit can only be pro
jected from the highest verbal position, because that is where the participle is
overtly realized.
Thus, the distribution of intervening material in Verb Projection Raising
constructions and the distribution of participles seem to be regulated in essen
tially identical ways. This insight is entirely due to the VO-approach to the syntax
of the verb clusters in Continental West Germanic, which has made it possible to
analyze participle placement as movement to a specifier position.
6. Patterns in the word order variation.
So far, we have been studying the range of possible word orders across
Continental West Germanic. In this concluding section, I would like to concen
trate on a particular area where Continental West Germanic dialects are spoken,
to see whether there is a system in the actual patterns of word order that corre
sponds to a particular group of speakers or dialects.
The area under consideration is the territory of the Netherlands, and the data
and generalizations are derived from Stroop (1970). I will limit myself to con-
tructions involving a participle, an auxiliary, and zero or more modal verbs.
Stroop (1970:264) identifies three systems in the verb clusters used among
the speakers of Netherlandic dialects. The relevant examples are given below
(the auxiliary worden is comparable to hebben in relevant respects; in particular,
I assume that worden is a composite of a directional and a stative head, as in the
paraphrase 'come to be'):
(44) System I
a. ... dat hij gehaald werd 2-1
that he fetched-PART became
"... that he was fetched."
b. ... dat hij gehaald worden moest 3-2-1

that he fetched-PART become-INF must-FIN
"... that he had to be fetched."
c. ... dat hij gehaald worden moeten zou 4-3-2-1
that he fetched-PART become-INF must-INF should
"... that he ought to be fetched."
(45) System 
a. ... dat hij gehaald werd 2-1
that he fetched-PART became
b. ... dat hij gehaald moest worden 3-1 -2
that he fetched-PART must-FIΝ become-INF
 ... dat hij gehaald zou moeten worden 4-1 -2-3
that he fetched-PART should-FIΝ must-INF become-INF
(46) System 
a. ... dat hij werd gehaald 1-2
that he became fetched-PART
b. ... dat hij moest gehaald worden or worden gehaald 1-3-2 or 1-2-3
that he must-FIN fetched-PART become-INF or become fetched
(Stroop does not present a c. example for system .)

In system I, the infinitives are consistently adjoined to the modals, and the
participle is obligatorily moved to the highest available position. For (44b-c), the
adjunction of the infinitives leaves the participle no other choice than to move to
the specifier position of the highest verb. All lower potential licensing positions
are disqualified by the adjunction of the infinitives. For (44a), we have to say
that the option of licensing the participle in the specifier of OF (yielding a 1-2 or
der) is not chosen.
In system II, the infinitives consistently refuse to adjoin in overt syntax,
yielding 'ascending' orders, apart from the participle. The participle, as in sys
tem I, consistently moves to the highest available licensing position. That is, the
system consistently takes the highest copy of the auxiliary to determine the li
censing position for the participle.
In system III, as in system II, the modal and the auxiliary show the
'ascending' order. Unlike system , the participle is licensed in a specifier posi
tion directly associated with the auxiliary. As predicted by the analysis presented
here, there are two such positions (assuming worden to be a composite, like
hebben), each of which can apparently be taken optionally. This optionality is
predicted in the analysis presented here, but it is likewise predicted that (46a)
could optionally be realized as gehaald werd (2-1). The absence of this possibil
ity in Stroop's system ΠΙ remains a subject for further study.
7. Conclusion
This paper leaves many questions concerning verb clusters in Continental

West Germanic unanswered. For example, the mechanism of the IPP-effect and
the word order generalizations associated with it have been glossed over (see
Vanden Wyngaerd 1994 for an interesting approach). Also, the question of
'Verb Projection Raising material' intervening between an auxiliary and a par
ticiple has not been discussed. (In the line of the analysis pursued here, we are
led to conclude that BE and OF can be separated by the licensing phrase for the
intervening material.)
The Continental West Germanic dialects display a wealth of material to be
discussed in this context. It was not my intention here to do justice to all the rel
evant data. At points I had to abstract away from particular phenomena in the
individual dialects covered (as one can infer from recent publications on the
same subject, e.g., Haegeman 1995). I can only make up for the omissions by
referring the reader to the vast literature on the subject, some of which may be
accessible via the bibliographical references included here.
Be that as it may, I hope to have shown that progress can be made if the
syntax of the Continental West Germanic verb clusters is approached from a re
strictive starting point, namely that all Germanic languages should be considered
to be structured alike.1
1
This paper could not have been written without previous exposure to presentations and
publications on the subject of verb clusters in Continental West Germanic by many Dutch,
Belgian and German colleagues. I would like to mention in this context (without intending to
leave anyone out) Hans den Besten, Hans Broekhuis, Marcel den Dikken, Arnold Evers, Liliane
Haegeman, Eric Hoekstra, Jan Koster, Karen Lattewitz, Eric Reuland, Henk van Riemsdijk,
Jean Rutten, and Manuela Schönenberger. For this paper in particular, comments and data from
Liliane Haegeman, Eric Hoekstra, Karen Lattewitz, and an anonymous reviewer for the APLA
Conference Proceedings have been extremely helpful. I also would like to thank the audiences
at the APLA conference on Microparametric Syntax and Dialect Variation, at the 10th
Comparative Germanic Syntax Workshop in Brussels (January 1995), and at the linguistics
colloquia at the University of Cologne, the University of Frankfurt, and the University of
Groningen. The author is supported by a grant from the Netherlands Organization for Scientific
Research, NWO.
REFERENCES
Bach, Emmon. 1962. "The Order of Elements in a Transformational Grammar of
German". Language 38.263-269.
Baker, Mark  1988. Incorporation: A Theory of Grammatical Function
Changing. Chicago: The University of Chicago Press.
Baur, Arthur. 1988. Schwyzertüütsch. Winterthur: Gemsberg Verlag.
Benveniste, Emile. 1966. "Être et avoir dans leurs fonctions linguistiques".
Problèmes de linguistique générale 1, ed. by Emile Benveniste, 187-222. Paris:
Gallimard.
Besten, Hans den. 1983. "On the Interaction of Root Transformations and Lexical
Deletive Rules". On the Formal Syntax of the Westgermania, ed. by W.
Abraham, 47-131. Amsterdam: John Benjamins.
& Jerold Edmondson. 1983. "The Verbal Complex in Continental West
Germanic". On the Formal Syntax of the Westgermania, ed. by W. Abraham,
155-216. Amsterdam: John Benjamins.
Bloemhoff, Hein. 1979. "Heranalyse van een Stellingwerver oppervlaktestruc
tuur". Us Wurk 28.31-38.
Bruch, Robert. 1973. Précis populaire de grammaire luxembourgeoise.
Luxembourg: Éditions de la Section de Linguistique de l'Institut gr.-d.
View from Building 20. Essays in Linguistics in Honor of Sylvain Bromberger,
ed. by Ken Hale and Samuel J. Keyser, 1-52. Cambridge: MIT Press.
Dam, Jan van. 1972. Syntax der deutschen Sprache. Groningen: Wolters-
Noordhoff.
Dikken, Marcel den. 1994. "Minimalist Verb (Projection) Raising". Minimalism
and Kayne's Asymmetry Hypothesis. Groninger Arbeiten zur germanistischen
Linguistik 37, ed. by C. Jan-Wouter Zwart, 71-88. Department of German,
University of Groningen.
& Eric Hoekstra. 1995. "Parasitic Participles". Paper presented at the 10th
Comparative Germanic Syntax Workshop, Brussels, January 18.
Evers, Arnold. 1975. The Transformational Cycle in Dutch and German. PhD dis
sertation, University of Utrecht.
Freeze, Ray 1992. "Existentials and Other Locatives". Language 68.553-595.
Groat, Erich & John O'Neil. 1994. "Spell-out at the LF-Interface". ms., Harvard
University.
Haegeman, Liliane. 1994. "Verb Raising as Verb Projection Raising: Some
Empirical Problems". Linguistic Inquiry 25.509-522.
. 1995. "IPP Constructions and Verb Movement in West Flemish", ms.,
University of Geneva.
& Henk van Riemsdijk. 1986. "Verb Projection Raising, Scope, and the
Typology of Rules Affecting Verbs". Linguistic Inquiry 17.417-466.
Hoeksema, Jack. 1988. "A Constraint on Governors in the West Germanic Verb
Ouster". Morphology and Modularity, ed. by Martin Everaert, Arnold Evers &
Mieke Trommelen, 147-161. Dordrecht: Foris.
Hoekstra, Eric. 1994. "Analysing Linear Asymmetries in the Verb Clusters of

Dutch and Frisian and their Dialects", ms., P.J. Meertens Institute, Amsterdam.
Kaan, Edith. 1992. "A Minimalist Approach to Extraposition". MA thesis:
University of Groningen.
Kayne, Richard S. 1993. "Toward a Modular Theory of Auxiliary Selection".
Studia Linguistica 47.3-31.
. 1994. The Antisymmetry of Syntax. Cambridge: MIT Press.
Koster, Jan. 1975. "Dutch as an SOV Language". Linguistic Analysis 1.111-136.
Lange, Klaus-Peter. 1981. "Warum Ersatzinfinitiv"? Groninger Arbeiten zur ger
manistischen Linguistik 19, 62-81.
Lötscher, Andreas. 1978. "Zur Verbstellung im Zürichdeutschen und in anderen
Varianten des Deutschen". Zeitschrift für Dialektologie und Linguistik 45.1-29.
Neeleman, Ad. 1994. Complex Predicates. PhD dissertation, University of Utrecht.
Reuland, Eric J. 1981. "On Extraposition of Complement Clauses". Proceedings
of NELS 77, ed. by Victoria Burke & James Pustejovski, 296-318. Amherst,
Mass.: Graduate Linguistics Students Association.
Rutten, Jean. 1991. Infinitival Complements and Auxiliaries. PhD dissertation,
University of Amsterdam.
Shepherd, P.H.M. 1946. Van Taol naar Taal. Nederlands voor Maastricht en om
streken. Maastricht: Goffin.
Stroop, Jan. 1970. "Systeem in gesproken werkwoordsgroepen". Nederlands di
alectonderzoek, ed. by J. Stroop, 247-264. Amsterdam: Huis aan de drie
grachten.
Tiersma, Peter Meijes. 1985. Frisian Reference Grammar. Dordrecht: Foris
Publications.
Travis, Lisa. 1984. Parameters and Effects of Word Order Variation. PhD disserta
tion, MIT.
Vanacker, Valeer F. 1969. "Een paar punten uit een vergelijkend syntaxisonder-
zoek van enkele Zuidnederlandse dialecten". Gentse Bijdragen 1969 (Album E.
de Bock), ed. by W. Pée, 291-302. Uit het seminarie voor Nederlandse taalkunde
en Vlaamse dialectologie van de Rijksuniversiteit Gent XXVI.
. 1970. "Een 'Zuidnederlandse' constructie in een paar Zuidnederlandse
dialecten". De Nieuwe Taalgids 1970, special Van Haeringen issue, 140-157.
Vanden Wyngaerd, Guido. 1989. "Object Shift as an -Movement Rule". MIT
Working Papers in Linguistics 11.256-271.
. 1994. "IPP and the Structure of Participles". Minimalism and Kayne's
Asymmetry Hypothesis. Groninger Arbeiten zur germanistischen Linguistik 37,
ed. by . Jan-Wouter Zwart, 265-276. Department of German, University of
Groningen.
Vendryes, Joseph. 1937. "Sur l'emploi de ľauxiliaire avoir pour marquer le
passé". Mélanges de linguistique et de philologie offerts à Jacq. van Ginneken,
ed. by Jan Wils et al., 85-92. Klincksieck, Paris.
Verhasselt, J. 1961. "Verschillen tussen Noord en Zuid inzake de volgorde hulp
werkwoord-hoofdwerkwoord". Taal en Tongval 13.153-157.
Zwart, C. Jan-Wouter. 1993. Dutch Syntax. A Minimalist Approach. PhD disserta
tion, University of Groningen.
. 1994. "Dutch is Head Initial". The Linguistic Review 11.377-406.

. 1995. "A Note on Verb Clusters in the Stellingwerf Dialect". Linguistics
in The Netherlands 1995, ed. by Marcel den Dikken & Kees Hengeveld, 215-
226. Amsterdam: John Benjamins.
GENERAL INDEX
Note՝. This index does not claim to be exhaustive. For instance, very familiar subjets with no
special relevance to the discussion have been left out. When the terms selected are frequent
enough, the page number is followed by/., indicating that this item is referred to frequently
throughout the article. The abbreviation n. indicates that the item is referred to in a footnote.
A. number agreement in Inuit 139

A position 29, 33 object agreement in number 199
-barchain 27, 31,37, 149 object-verb agreement 189
-bar movement 28, 32 overt AGRG strategy 114
-bar position 29, 31, 33 past participle agreement in Romance
A-chain 32 languages x-xi
A-not-A questions 43, 47, 65, 71-73 possessor agreement 111
Abney, Steven 113,220 SPEC-head agreement 152, 193, 221
absolute clauses 159, 160 subject agreement 153
absorbeur de Cas 18 subject-verb agreement 189
absorption 16, 17 with direct object in Icelandic 191
abstract modals 155, 158 Aissen, Judith & David Perlmutter 3
, Peter & Maaike Schoorlemmer allegro speech 173, 174
21 anaphoric binding 153
accomplishment verbs 48 Anaphoric Feature Condition 136, 138,
adjunct clauses 160, 164 139
adjunction 134, 238 anglais 9, 14
directionality 241 animacy hierarchy 112
NP adjunction in Inuit 139 antipassive 124, 137, 141
structure 237 antonymy
adjuncts 28, 32, 222 negative gradable antonymy 221
agent 5, 8, 11, 123-142 Arabic
agreement 38, 42, 53, 56-66, 161 bezzaf in dialectal Arabic 218
abstract AGRG strategy 114 argument interne 5
AGRG (=genitive agreement) 111, 112, Argument Rule 102-105, 113-115
119 argument positions, see A-position, A-bar
anaphoric agreement 121, 135, 137 position
Anaphoric Feature Condition 136 arguments
checking 34 argument explicite 13, 14, 15
fourth person (reflexive) agreement argument externe 5-18
morpheme in Inuit 124 argument externe actif 11-13
nominative subjects & obligatory argument features 98
agreement in Icelandic 194 argument implicite 13, 14
in NPQs 57 empty arguments 189, 206
inert / missing 122/7. aspect markers 76
260 GENERAL INDEX
aspectual markers 47, 71 C-movement 35

Assignation 16 c-sélection 5, 16, 17, 19, 20
assimilation 171, 184 Cambodian 67
atmospheric verbs 146 CAN 172
Auger, Julie 147 Canada 3, 121
Authier, J.-Marc 17 Cantonese 41-47, 53-66
auxiliaire 9 Cas 11,17 see also Case
Cas accusatif 16-20
. Cas nominatif 11, 16
Bach, Emmon 230 Case 35, 100/, 121 see also Cas
Baker, Mark 131, 220, 245 accusative subjects 190
Baker, C.L. 223 case-features on DPs 190
Baker Lake, Northwest Territories 123 dative subjects 190
Baur Arthur 233 ergatif 10
BE 174 ergative 123
Belfast English 80-92 genitive subjects 190
Belletti, Adriana 2, 3, 6, 13, 15, 16 inflection 114
benben (Québec French) 214 Inuit
quantifier I intensifier 215 absolutive (nominative) 121
Benincà, Paola xv relative (ergative) case in Inuit case
enveniste, Emile 246 123
Besten, Hans den 27, 230 secondary case 124
— & Jerold Edmondson 238 neutre 9, 10
Beukema, Fritz & Peter Coopmans 85 neutre/ergatif 8, 9
bezef (Colloquial French) 218 nominative case 34, 190
bilingualism 245 as default value at PF 195
binding , 124, 154-160, 200, 203, 206 of direct object in Ss with oblique
anaphoric binding 153 subjects 207
Binding Principles 151 position 35
Binding Theory 200, 203 strong Case in Icelandic and Old
Bittner, Maria 122, 142 Scandinavian 98f.
— & Ken Hale 142 structural accusative case 196, 199n.,
Bloemhoff, Hein 235 209
-ennema, Reineke 142 weak Case in Scandinavian 98/
Bolton 169f. Catalan xiii
Borer, Hagit 25, 153 catégorie vide 17, 20
bounding theory 30 causatives (passives of) xi
Branigan, Philip 34 Cellar, Jacques 218
bridge verbs 29 chain uniformity 32
Brilioth, Börje 178. chaîne thématique 16
Britton, Derek 179 chains
Browning, Marguerite 33 A-chains 32
Bruch, Robert 233 A'-chains 27, 31,37, 149
Burzio, Luigi 2 dislocated clause 160
uniform chains 27
. uniformity constraint 26
 (complementiser) position 27 WH-chains 148
, primary 35 X°-chain 31
, secondary 35 Chao, Yuen-Ren 45
c-command 136, 139 checking see also features 86, 92, 103
GENERAL INDEX 261
checking domain 30, 31, 32, 35 construction moyenne 2-19

Chen, Janet C.-W. 76, 77 control 151, 159, 161, 196, 209
Cheng, Lisa L.-S. 63, 122., 213 contrôle 14
Cheung, Samuel H.-N. 47 contrôle argumentai 14
Chinese 41f. contrôle thématique 14
Cantonese 41f. Corver, Norbert & Denis Delfitto 164,
Classical Chinese 42f. 166
Mandarin 4 1 f . . , 2 1 3 covert movement 29, 34
Shangai 56n. CP (Complementiser Phrase)
Taiwanese 41f. CP barrier 30
Chomsky, Noam ix-xvii, 2 - 1 3 , 25-34, 64, CP recursion 30, 34
79, 88, 102, 121, 153, 190, 239 Critère-Ө 16
— & Howard Lasnik 33, 36, 196 Culicover, Peter W. 34η.
Cinque, Guglielmo 2, 11, 20, 149, 214, Cunliffe, Henry 180
222
Classical Chinese 42, 51-56 D.
clefting 148 Dam, Jan van 235
clitic climbing xi, xii Danish 27-35, 98, 114
clitic splitting xi Danish/German contrast 28
code-switching 88 DARE 175
Colloquial French 166 de
comparative syntax ix-xvii negative polarity item 216
complementation 231 partitive preposition 216
complementisers definiteness 103, 116
at complementiser 27 Delsing, Lars-Olof 102
if complementiser 26 DFC (Doubly-filled Complementiser) 36
that complementiser 26, 33 dialect vs. language xiv֊xv
Comrie, Bernard 112 Dikken, Marcel den 248
CON/CAN "can" 172 — & Eric Hoekstra 235
concordance des temps 156 Di Sciullo, Anna-Maria & Mireille
Condition de visibilité 16 Tremblay 223
connectivity 159 dislocation 149
constraints DO 175
chain uniformity constraint 26 do-support 85
contrainte aspectuelle 5, 18 Doetjes, Jenny 216n.
contrainte générique 2, 19, 20 double object constructions χ
doubly-filled CP constraint 36, 152 Dorais, Louis-Jacques 121, 140
extraction constraints 29 Dulong, Gaston 147n.
Head Movement Constraint 114, 245 Dutch 229f.
LF constraint on chains 30 dialects 233
non-local agreement constraint 60 spoken Dutch 233
on distribution of reflexives and Standard Dutch 232
pronouns in Scandinavian 189
on movement of NEG-particle in E.
Chinese 59 E-language 83
on representations (ECP) 30 ec (empty category) 33, 192
on traces (RES(ECP)) 31 ECM (Exceptional Case Marking) ix-x
Residue of the ECP 30 ECM complements 31
construal Economy 64, 116, 214
in Verb Projection Raising 241
262 GENERAL INDEX
economy of representation 105, 115, 189, strong Case 100, 103

193, 199, 203 substantive 164
ECP (Empty Category Principle), 26, 30, visible 103
33-37 weak xiv
Ellis, Alexander J. 180 checking at LF 103
empty category 33, 192 weak AGRS 100
English x֊xi, 27, 33, 34, 102, 113, 126, weak Case 100, 104
213, 229 [+human] 198
Belfast English 80-92 [+hypothetical] 151, 155
Hiberno-English 82, 86 [+qu]153
of Bolton 169f. [-qu] 159
Lancashire dialects 180 Fellbaum, Christiane 3, 21
Middle English 178 — & Anne Zribi-Hertz 2, 5, 13
standard English 80, 87 Finer, Daniel 124, 136
Equidistance Principle 248 Fiva, Toril 113
ergatif 10 Flemish 236
ergative 123 for + lexical subject construction x
espagnol 4, 9, 12 français see also French 2, 9, 10, 12, 17
de Madrid, madrilène 1, 13, 17, 20 français canadien 5, 13, 15, 16, 17
Evers, Arnold 232, 238 français continental 2
expletive 29 Frangopulo, Nicholas 170n.
extraction 28 Freeman, Thomas, H. B. Rogers & R. H.
extraction constraints 29 Kinvig 170
extraction from SPECC 30 Freeze, Ray 246
French ix֊xvii, 88, 109, 111, 151
F. adverbial quantifiers 213
Fagan, Sarah 2, 5, 13 Canadian French 1
Farnworth 170 Colloquial French 164
Faroese 97, 189, 197,207 Continental French 145, 166
distribution of reflexives and pronouns Québec French 145, 166, 214
189 Standard French 164, 214
subject/verb agreement 189 Frisian 233, 237
Fauconnier, Gilles 217 Full Interpretation 30, 192, 218
features functional categories
Case change in feature values 96
morphological Case 116 functional heads 25
strong Case 116
checking & erasure from functional G.
heads 191 gam-questions 73
functional German 27, 28, 30, 31, 88, 109, 110
Case 102 Middle High German 179
definiteness 102 Germanic 27
hypothetical feature 151 auxiliary-participle constructions 246
invisible 103 Bavarian 229
matching values 103 Brabantish 229
N feature 166 Continental West Germanic languages
person/number in Icelandic 190 229
strong xiv Dutch 229
checking before spell-out 103 finite complement clauses 231
strong AGRS 100 Flemish 229
GENERAL INDEX 263
Frisian 229 His (hypothetical infinitives) 145, 151,

High German 229 159, 164
infinitivus pro participio phenomenon historical syntax 80
234 Hoeksema, Jack 234
Limburgian 229 Hoekstra, Eric 234, 250
Low German 229 Hoekstra, Teun & Rene Mulder 34.
Luxemburgish 229, 238 Holmberg, Anders 84, 98, 99, 100, 102,
Middle High German 179, 233 192.
multi-verb constructions 239 — & Christer Platzack 98, 116, 204 .
North Germanic 230 Holmberg's Generalization 117
object shift 231 Horn, Laurence R. 216
possessive constructions 246 Huang, C.-T. James 66
scrambling 231 Hunt, Jean 183
Swiss German 229 hypothetical Complementizer Phrase 152
verbal clusters 231 hypothetical feature 151
West Flemish 241 hypothetical infinitive hypothetical
West Frisian 237 infinitives
West Germanic 230 absolute clauses 146
GETTEN (as past part, of GET) 176 characteristics 146
Ghomeshi, Jila & Diane Massam 13 ln. null subject 146
government 30 pro-drop 147
proper government of traces 26 variation 160
grammaire universelle see also Universal
Grammar 20 I.
grammaticalité 13, 20 Iatridou, Sabine & A. Kroch 29
grammaticalization Icelandic 97-118, 189,207
of negation markers as question particles distribution of reflexives and pronouns
66,67 189
Greed 103, 113,214 object/verb agreement 198
Grimshaw, Jane 141 spoken 116
Groat, Erich & John O'Neil 251 subject/verb agreement 189
Grodzinsky, Yosef & Tanya Reinhart 125, if complementiser 26
135 illocutionary force marker 86
Gross, Maurice 2, 6 imperatives 45, 79f.
imperative marker 86
H. imperative morpheme 88
Haegeman, Liliane 234, 236, 255 non-raising of subjects in imperatives
— & Henk van Riemsdijk 236 89
haishi'-questions 68 overt-subject imperatives 81
Hargreaves, Alexander 182, 183 restricted inversion dialect 83
Harley, Heidi 34 singular concord 89
HAVE 174 unrestricted inversion dialect 83
head direction parameter 92 vocatives in imperatives 90
Head Movement Constraint 114, 245 whole-verb inversion dialect 83, 91
head-adjunction 248 incorporation
head-movement 35 OF & BE 247
Henry, Alison 82, 86, 89, 91 indicative mood 130, 133
Hestvik, Arild 204. infinitives
Hiberno-English 82, 86 as suppletive imperative 155
High German 233, 250 infinitivus pro participio 234
264 GENERAL INDEX
Inuit 121f, 142 Kayne, Richard S. ix-xvii, 19, 122, 151,

adverbial clauses & anaphoric agreement 158, 222, 230, 246
125 Källskog, Margareta 108n.
anaphoric agreement 121 Kester, Ellen-Petra 104ft.
anaphoric agreement on verbs 124 Keyser, Samuel Jay & Thomas Roeper 2,
antipassive construction 124 6, 13
chain formation 138 Klima, Edward 217
gender 123 Koster, Jan 230
indicative mood 130, 138 Kroch, Anthony 80
inflectional paradigms 122
Inuktitut 121 Լ.
irrealis 134ft. L-marked 30
Labrador Inuit Standardized Spelling L-related 31
System 126ft. Labov, William 80
Labrador Inuttut 121f. Labrador 121f.
Labrador Parameter 131, 136, 138, 141 Labrador Inuttut 126, 134, 138, 141
Negative Generalization 134 Labrador Parameter 131, 136, 138, 141
Nominal Specifier Condition 129 Lancashire dialects 180
nominal verb 130 Lange, Klaus-Peter 234
Northern Québec dialect 140 langues romanes 2-20
oblique cases 137ft. Larson, Richard 32
optative mood 134, 134ft. Lasnik, Howard 223
participial mood 130f. Law, Paul 29
NUM (number category of participial in Léard, Jean-Marcel & Denis Amyot 215
Inuit) 128 Least Effort (see also Economy) 79
portmanteau morpheme 137 Lepschy, Giulio 3
Qairnirmiut dialect 123 Lester, Mark 176
Qairnirmiutut 123 Lewis, David 18, 19
secondary case 124 LF (Logical Form), 31, 36, 79, 102, 111,
word order 122, 123 151, 192,218,251
Inuktitut 121f., 127, 142 Li, Charles & Sandra A. Thompson 45
IPP (infinitivus pro participio) effect 234 Li, Paul 49
Italian x-xvii, 151, 162, 214, 219 licensing condition on traces 31
Gardenese xii Lightfoot, David & Norbert Hornstein 122
Milanese xi locative inversion construction 34
Northern Italian dialects xv Longobardi, Giuseppe 219
Sardinian xii Lötscher, Andreas 233
Trentino 5ft. Lowe, Ronald 133
italien 9, 10, 12 Luxemburgish 229, 238
Lyons, Christopher 2, 6
J. Lyons, John 221
Jaeggli, Osvaldo 8, 14, 17, 19, 20
Jakubowicz, Célia 213 M.
Johns, Alana 121, 127, 130, 136, 141 Mainland Scandinavian (MSc) 97, 98, 100
Jordan, Richard 179 major parameter 95, 118
Junker, Marie-Odile & France Martineau Mandarin see also Chinese 41-64
146ft. Mandarin ma questions 68
Mandelbaum, Deborah 113n..
K. Manning, Christopher 141
Kaan, Edith 240 Manzini, Maria-Rita 8
GENERAL INDEX 265
Martel, Pierre 146η., 150η., 162η. NEG-criterion 85, 87,91

Martineau, France 146η. NEG-particle 53-60
Martinon, Philippe  agreement requirement 60
Massam, Diane 21η. NEG°-to-C° movement 66
matrix property 63 negation in Chinese 45
matrix verb 29, 30 negation in overt-subject imperatives 85
McCloskey, James 82 negation markers 41-67
Merge 65 negation markers as question particles 52,
MET "might" 173 65,66
Middle English 178 neutre 9, 10
Middle High German 179 neutre/ergatif 8, 9
minimal pair 27 NOAN (NOT) 175
Minimalist theory 27, 80, 102, 146, 192, Nodal, John H. & George Milner 180
213 Nominal Specifier Condition 129, 139
minor parameter 95, 112, 118 nominative
modals 54 as default case in Icelandic & Faroese
montée 5 see also subject raising 195
Mossé, Fernand 178 assigned to direct object 195
Move 65 nominative Case 34
movement 2 6 / , 42, 57, 7 9 / , 103/, 122, non-local agreement constraint 60
148/, 190/, 215, 230 ηοη-θ position 33
A-movement 196 non-uniform chains 33
agreement features 190 North Germanic 229
and predication 192 Northern Swedish 98, 99
DP-movement in Icelandic 191 Norwegian 98, 99, 107, 112, 113, 114
driven by feature checking 190 Noss, Richard B. 67
extraposition 231 NPI (Negative Polarity Item) 214
head movement 248 NPQs (Negative Particle Questions) 42-77
I-to-C 153 NSw (Northern Swedish) 107-114
NEG°-to-C° movement 66 Null Subject Parameter 147
object shift 231 null subjects xi, 160, 166
of DPs to SPEC position 190 NUM (number category of participial in
optionality 245 Inuit) 128
overt vs. covert movement 81 numeration 194n.
predication 190
V-to-I 153 O.
V-to-I movement 151 Obenauer, Hans-Georg 2, 216
V-to-I-to-C 153 object
WH-movement 26, 27, 28, 31, 32, 35, nominative object 195
148, 152, 159 object shift 83, 84, 91, 115, 116, 117,
movement vs. base-generation 57 231
moyen see constructions moyennes objet thématique 16, 18
MSc (Mainland Scandinavian) 100, 111, oblique subjects 190
118 lack of subject/verb agreement 193
MUN "must" 173 Occitan xii
MUST 173 Old Norse 173
Old Scandinavian 97, 98
N. Onions, C.T. 179
Neeleman, Ad 241 opérateur non-sélectif 19
NEG 56 optionality 80
266 GENERAL INDEX
in syntax 81 Possessor Shift 110-117

inverted vs. uninverted imperatives 88 Postal, Paul 19
of categorisation of morpheme as strong predicat monadique 5
or weak 92 predication 8, 10, 192/
of movement 79 and Full Interpretation 192
of subject raising 89 identification requirement 193
of subject-verb concord 89 structures at LF 199
Orton, Harold & Wilfred J. Halliday 183 vs. feature checking 199
overt vs. covert movement 81 prédication 8, 10, 11
overt-subject imperatives 80 preposition stranding χ
OWT "ought" 173 preverbal adjuncts 43
primary  35, 37
P. Principles and Parameters 79, 95, 121
Paduan 155 pro
parameter setting 79 arbitrary 161
parameters 25, 79f, 95f, 121f, 147, 164, expletive expressions 163
208, 213 non-lexical 161
and lexicon 213 pro-drop 162, 166
change in value 96 reference of 161
finiteness of parameters xiv Progovac, Ljiljana 217
head direction parameter 92 pronouns 26, 84, 98 f, 146f, 169f, 189f.
individuation of parameters xiii-xvii clitic pronouns 166
Labrador Parameter 131 emphatic pronouns 161
major parameter 95, 118 reflexive pronouns (Icelandic, Faroese)
minor parameter 95, 112, 118 189f.
Null Subject Parameter 147 relative pronouns 26
Principles and Parameters model 79 resumptive pronouns 149, 159
Principles and Parameters theory 95 second person singular
parasitic gaps 148 enclitic forms 175, 179
participial mood 130, 133 objective case 171
particle constructions xvi subjective case 171
partitive clitic xiii strong pronouns 166
passif 8, 14, 20 subject pronouns 164
passive 32, 83, 91 weak object pronouns 84, 86, 90
Contrôle, italien & espagnol 14η. proper government 30, 31, 33
Picton, J. A. 180 proper government of traces 26
Platzack, Christer 98 proper governor 33
Poletto, Cecilia xv proposition infinitive 14, 15
Pollock, Jean-Yves 85, 88, 103η., 153
portmanteau morpheme 137 Q.
possessive constructions 95f. Q-particle 56
implicational hierarchy 107 QAD (quantification at a distance) 216
N.D-POSS construction 97 Qairnirmiut 123
Possessor Shift 110-117 Qairnirmiutut 130
postnominal possessor 95 quantifiers 163
postnominal possessors in Northern adjectival quantifiers 220
Swedish 106 adverbial (French) 214
PP possessor 97 aspectual 222
pronoun plus lexical DP 114 beaucoup (Standard French) 214
Scandinavian languages and dialects 95 benben (Québec French) 214
GENERAL INDEX 267
bezef (colloquial French) 218 Icelandic 97-118, 189f.,207

lerche (colloquial French) 219 MSc (Mainland Scandinavian) 97, 98,
measure quantifiers 214 100
negation 222 Northern Swedish 98, 99
negation & scope 215 Norwegian 98, 99, 107, 112, 113, 114
positive polarity items 217 NSw (Northern Swedish) 107-114
quantification at a distance 216 Old Norse 173
temporal 222 Old Scandinavian 97, 98, 100
Québec French 214 Swedish 98
question particles 42-67 Central Swedish 99
Northern Swedish 97
R. Southern Swedish 99
reduction 171, 184 standard Swedish 99
reflexive clitic xiii-xiv Schilling, Karl G. 182
registres rhétoriques  Schwartz, Bonnie & Sten Vikner 28.
relative clause 26, 36 Shlonsky, Ur 34n.
relative pronouns 26 Scollins, R. & J. Titford 172.
Renzi, Լ. xv scrambling 231
RES(ECP) (Residue of the ECP), 31-37 se moyen 17, 20
restricted inversion dialect 83 se/si (Romance reflexive marker) 2,11,
resumptive pronouns 149, 159 16,20
Reuland, Eric J. 231 se/si impersonnel 11
Rivero, María-Luisa & Antonio Terzi 86 se/si moyen 11, 16, 19, 20
Ritter, Elizabeth 128. secondary  36, 37
Rizzi, Luigi 2, 33, 87, 161, 215 seg (Faroese reflexive pronoun) 200
Roberge, Yves 17 sentence-final question particles 43
— & Marie-Thérèse Vinet 147, 162 Shepherd, P. H. M. 233
Roberts, Ian 1,5, 11, 15,20 Shorrocks, Graham 170
Roeper, Thomas 2, 6 SHOULD 173
— & Edwin Williams 79 si/se (Romance complementizer) 151
Rohrbacher, Bernard 98 sig (Icelandic reflexive pronoun) 200
rôle thématique externe 16, 17, 20 complementarity with non-reflexive
Romance languages x-xiv, 1f., 109, 151, pronouns 203
155, 219 pronoun vs. anaphor 203
auxiliary selection xii-χίν Sigur>sson, Halldór Á. 116, 190
Russian 109, 110 simplification 171, 184
Rutten, Jean 238 Small Clause 241
Ruwet, Nicolas 2, 3, 6 Smith, David M. 170
Smith, Lawrence 134.
S. Sobin, Nicholas 25, 33
Sadock, Jerrold 123, 142 sociolinguístic variants 80
Sandfeld, Kristian 19 Spanish see also espagnol 4, 9, 12, 91,
sandhi assimilation of dentals 178 166, 219
Sandström, Görel & Anders Holmberg SPEC-C 28, 30, 31,33, 34, 35
98., 102n., 104n. Spell-out 79, 80, 87, 88, 89, 92, 102, 242
Santelmann, Lynn 104. Sportiche, Dominique 223
Saxelby, Charles H. 170 standard English 80, 87
Scandinavian 95f., 189f. Standard French 164, 214
Danish 27-35, 98, 114 Stéfanini, Jean 3
Faroese 97, 189f, 197, 207 stress
268 GENERAL INDEX
weak stress 184 Trudgill, Peter 172., 176.

Stroik, Thomas 8 — & Jean Hannah 176.
Stroop, Jan 233, 253
subjacency 26 U.
subjects UG (Universal Grammar) ix, , xiii, xiv,
oblique subjects 190 27, 34, 38, 91, 92, 95, 214
subject clitic doubling 163, 166 unaccusative 84, 88, 91
subject extraction 26, 37 uniform chains 27
subject raising 84 uniform A-bar chain 30, 33, 37
subject relative clause 33 uniform -chain 31
subject raising in imperatives 88 uniformity constraint 26
verb agreement & oblique subjects 191 uniformity, strong 33
sujet thématique 6 uniformity, weak 33
Survey of English Dialects 183
Swedish 98 V.
Central Swedish 99 vacuous movement effects 36.
Northern Swedish 97 V- complex 29, 30, 31,35
Southern Swedish 99 V/2 (verb second) 27, 28, 30, 35
standard Swedish 99 Vanacker, Valeer 233, 236
Swiss German 233, 241 Vanden Wyngaerd, Guido 230, 255
switch reference 124 Vendryes, Joseph 246
Vanelli, Laura xv
T. verb movement to  28
tag-questions 68 Verb Projection Raising 236, 238, 255
tags 176 verb raising 28, 81
Taiwanese 42-74 verb raising (in imperatives) 87
Tang, T.-C. 49, 56 Verb Second languages 88
Taraldsen, Knut Tarald 110, 124 verbal clusters 232
Taylor, Francis Edward 180 Verhasselt, J. 233
Teng, Shou Xin 49 Vietnamese 67
Tense 86, 164 Vikner, Sten 27, 29, 98
Terzi, Arhonto 153 Villiard, Pierre 146
Thai 67 — & Marie-Thérèse Vinet 146
that complementiser 26, 33 Vinet, Marie-Thérèse 146, 153, 160, 215
that-trace 25, 27, 37 voix moyenne 2-20
that-trace effect 25, 33 voix passive 6, 8, 14, 17
théorie des Cas 18 volitional verbs 224
theory of barriers 30 VP-a-VP questions 75
Ө-bar 32 VP-not-V questions 43, 44, 68, 71, 73, 74
Ө-marking 32 VP-VP questions 68, 73
Ө-position 32, 33
Thibault, Pierrette 216 W.
Tiersma, Peter Meijes 233 Wang, William 45, 46
topic 28, 146, 147, 155 Watanabe, Akira 34., 213
topic projection 34. weak object pronouns 83
topicalisation 29, 33 weak subjacency effects 33
topicalisation of S 32 Wehrli, Eric 2, 5, 13
Travis, Lisa 28., 245 West Flemish 241
Trentino (Italian dialect) 5. West Frisian 237
Tschirch, Fritz 179
GENERAL INDEX 269
WH .
WH-chains 148 X°-chain 31
WH-criterion 87
WH-extraction 26, 28, 148 Y.
WH-movement 26, 27, 28, 31, 32, 35, yes-no questions 43, 56, 62, 64, 68, 69,
148, 152, 159 73, 76, 77
WH-phrase xiii, 28, 29 yes-no questions in Chinese 41
WH-questions 76 yes-no question types in Mandarin 68
whole-verb inversion dialect 83 Yue-Hashimoto, Anne 43, 47
WILL 174
Williams, Edwin 2, 5, 13 Z.
word order Zhang, Min 51, 52
Germanic 230 Zribi-Hertz, Anne .
Wright, Joseph 181 Zubizarreta, María-Luisa 2, 6
— & Elizabeth Mary Wright 178 Zwart, Jan-Wouter 28η., 230, 236, 242,
250

Kayne 1996

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Kayne 1996

Uploaded by

Copyright:

Available Formats

MICROPARAMETRIC SYNTAX AND DIALECT VARIATION

AMSTERDAM STUDIES IN THE THEORY AND

Series IV - CURRENT ISSUES IN LINGUISTIC THEORY

Advisory Editorial Board

Henning Andersen (Los Angeles); Raimo Anttila (Los Angeles)

James R. Black and Virginia Motapanyane (eds)

Microparametric Syntax and Dialect Variation

JOHN BENJAMINS PUBLISHING COMPANY

Library of Congress Cataloging-in-Publication Data

The papers in this volume originated as presentations at the 18th annual

Comparative syntax can be thought of as that facet of syntactic theory di­

prepositional complementizers,2 and ultimately be derived from an abstract dif­

languages: in this sense, the hypothesis testing task generated by French-Italian

3 Cf. Kayne (1985).

individuate at least 25 syntactically distinct languages/dialects solely by studying

the number of varieties of English/ distinct grammars of English must be at least

This paper introduces novel micro-parametric data drawn from Canadian

0. De la problématique de deux propriétés dites

(1) a. Ces sakés japonais se boivent frais en été.

(2) a. Ces sakés japonais se boivent frais en été.

(3) a. * Ces sakés japonais se boivent frais par les gens,

Personne ne semble cependant avoir remarqué l'existence de dialectes parmi

(4) a. En général, ces débats s'enregistrent par Anne, qui est

De plus, notre collègue, Madame María-Luisa Rivero, nous a informé que

(6) a. Eso se dice (por los militares).

1. De deux analyses divergentes de la voix moyenne

On trouve dans la littérature deux grands types d'hypothèses qui tentent de

(7) a. L'analyse de Williams (1981:94, 106-voir aussi Fagan

b. L'analyse de Wehrli (1986:274-voir aussi Cinque 1988:561,

D'un autre côté, Belletti (1982:5), Chomsky (1981:270-271), Keyser &

(8) a. L'analyse de Roberts (1987:273-277-qui s'inspire en partie de

b. L'analyse de Chomsky (1981:270-271-voir aussi Keyser &

2. De quatre tests qui distinguent les arguments externes qui

b. * Les porte-avions de l'ennemi ont coulé pour PRO mettre fin

(10) a. Le portaerei nemiche sono state affondate per mettere fine

(11) a. Los portaviones del enemigo fueron hundidos para poner

Comme l'indique le contraste de grammaticalité qui oppose l'exemple du

'neutres' ou 'ergatifs' en italien et en espagnol. Nous nous contenterons donc de

De la même manière, le contraste en (13a,b) démontre que seuls les argu­

(13) a. Cette machine à bouillir le linge est automatique.

Enfin, le contraste en (14a,b) illustre le fait que seules les constructions

Ces quatre tests sont bien entendu applicable à la construction moyenne et

3. Application des tests aux moyens dans les langues romanes

(15) a. Cette racine se mange pour maigrir. (Zubizarreta 1987:150)

(16) a. Le mele si mangiano per perdere peso.

(17) a. Una fabrica, eso se incendia para cobrar el seguro.

(18) a. Vu que ces manifestants sont aussi méprisés qu'ils sont

c. Estos museos militares tienen la desventaja de no poderse

En effet, la grammaticalité de phrases comme celles en (18a,c), (19a,c) et

4. De ľhypothese d'un argument externe explicite dans

(21) a. John wants PRO to be loved by everyone.

(22) De tels mensonges se racontent parfois pour PRO être ad­

(23) Tali mensogne si raccontano spesso per PRO essere ammi­

(24) Esas mentiras se cuentan á veces para PRO ser admirado.

En résumé, nous avons défendu l'hypothèse que toute construction moyenne

5. Une analyse microparamétrique des moyens dans les

(25) Le SE/SI moyen dans les langues romanes: Belletti (1982:5)

(26) Le SE/SI moyen dans les langues romanes: Roberts (1987:273)

Comme on le voit en (25), l'analyse de Belletti suppose que le morphème

(27) a. Le SE moyen du français et de l'espagnol madrilène

b. Le SE/SI moyen des autres langues romanes

Examinons maintenant quelques-unes des conséquences qui découlent de ces

Passons maintenant à l'entrée lexicale du SE/SI des moyens correspondant

(28) a. "Nous ne nous vendrons jamais, nous autres," disait Zola

Comparative syntax can be thought of as that facet of syntactic theory di

prepositional complementizers,2 and ultimately be derived from an abstract dif

De la même manière, le contraste en (13a,b) démontre que seuls les argu

(22) De tels mensonges se racontent parfois pour PRO être ad

(23) Tali mensogne si raccontano spesso per PRO essere ammi

In principle, describing the grammars of two dialects of a language is no dif

In Danish, this type of extraction is impossible, as example (4) demon

In this example, the S complement to that is topicalised to SPECC. If we as

According to Zhang (1990), the sentence-final question particle later disap