The Evidential Basis of Linguistic Argumentation

Studies in Language Companion Series (SLCS)

This series has been established as a companion series to the periodical
Studies in Language.
For an overview of all books published in this series, please see
http://benjamins.com/catalog/slcs

Editors
Werner Abraham Elly van Gelderen
University of Vienna / Arizona State University
University of Munich

Editorial Board
Bernard Comrie Christian Lehmann
Max Planck Institute, Leipzig University of Erfurt
and University of California, Santa Barbara
Marianne Mithun
William Croft University of California, Santa Barbara
University of New Mexico
Heiko Narrog
Östen Dahl Tohuku University
University of Stockholm
Johanna L. Wood
Gerrit J. Dimmendaal University of Aarhus
University of Cologne
Debra Ziegeler
Ekkehard König University of Paris III
Free University of Berlin

Volume 153
The Evidential Basis of Linguistic Argumentation
Edited by András Kertész and Csilla Rákosi
The Evidential Basis
of Linguistic Argumentation

Edited by

András Kertész
University of Debrecen,
and MTA-DE Research Group for Theoretical Linguistics

Csilla Rákosi
MTA-DE Research Group for Theoretical Linguistics

John Benjamins Publishing Company

Amsterdam / Philadelphia
 TM
The paper used in this publication meets the minimum requirements of
8

the American National Standard for Information Sciences – Permanence

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

Library of Congress Cataloging-in-Publication Data

The evidential basis of linguistic argumentation / Edited by András Kertész, Csilla Rákosi.
p. cm. (Studies in Language Companion Series, issn 0165-7763 ; v. 153)
Includes bibliographical references and index.
1. Linguistic models--Data processing. 2. Linguistic analysis (Linguistics) 3. Linguistics--
Research--Methodology. 4. Corpora (Linguistics) 5. Computational linguistics.
I. Kertész, András, editor of compilation. II. Rákosi, Csilla, editor of compilation.
P128.M6E95   2014
410.1--dc23 2013048427
isbn 978 90 272 5918 9 (Hb ; alk. paper)
isbn 978 90 272 7055 9 (Eb)

© 2014 – John Benjamins B.V.

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

Chapter 1.
Introduction: The state of the art and the structure of the book 1
András Kertész & Csilla Rákosi

part i.  The methodological framework

Chapter 2.
The p-model of data and evidence in linguistics 15
András Kertész & Csilla Rákosi

part ii.  Object-theoretical applications

Chapter 3.
The plausibility of approaches to syntactic alternation of Hungarian verbs 51
Károly Bibok
Chapter 4.
Methods and argumentation in historical linguistics: A case study 71
Katalin Nagy C.
Chapter 5.
Hungarian verbs of natural phenomena with explicit and implicit subject
arguments: Their use and occurrence in the light of data 103
Enikő Németh T.
Chapter 6.
The development of a taxonomy of verbal disagreements
in the light of the p-model 133
Helga Vanda Koczogh
Chapter 7.
A case of disagreement: On plural reduplicating particles in Hungarian 179
György Rákosi
Chapter 8.
A plausibility-based model of shifted indexicals 199
Zoltán Vecsey
 The Evidential Basis of Linguistic Argumentation

part iii.  Metatheoretical applications

Chapter 9.
Thought experiments and real experiments as converging data sources
in pragmatics 221
András Kertész & Csilla Rákosi
Chapter 10.
Data and the resolution of inconsistency in Optimality Theory 271
Csilla Rákosi
Chapter 11.
Conclusions 309
András Kertész & Csilla Rákosi
Author index 315
Subject index 317
 chapter 1

Introduction
The state of the art and the structure of the book

András Kertész1, 2 & Csilla Rákosi2

1University of Debrecen, Institute of German Studies,
2MTA-DE Research Group for Theoretical Linguistics

1.  The aim of the volume

One of the major driving forces in the development of linguistics has been the
quest for its empirical foundations. For example, a prominent historian of linguis-
tics characterises the essence of this development as follows:
Modern theoretical linguistics cannot be understood if its struggle for scientific
status is not placed at the centre of its history. The main problem in this quest for
scientific status has always been that of empirical access, that is, some notion of
what constitutes reliable data, the formulation of one or more causal questions,
and some success at answering these questions by means of an explanatory
theory.(Seuren 1998: 140; bold emphasis added, italics as in the original)

Throughout the history of linguistics, the problem of empiricalness has manifested

itself in a complex network of methodological debates. Currently, the debate focuses
on the question of what kinds of data are allowed in different linguistic theories and
what subtypes of data can work as evidence for or against particular hypotheses.
Kertész & Rákosi (2012) contributed to this debate by putting forward a metatheo-
retical solution to this problem: the p-model of linguistic data and evidence which
conceives of linguistic theorising as a process of plausible argumentation.
The aim of the present volume is to show how the p-model can be applied to the
everyday problem-solving practice of the working linguist, and, what is of utmost
importance, how it can increase the effectiveness of linguistic theorising.

2.  On the state of the art

From the seventies on, there have been two interrelated sets of methodological
assumptions which have been accepted by the overwhelming majority of linguists
 András Kertész & Csilla Rákosi

from generativists to corpus linguists. The first set of assumptions is what we will
call the standard view of linguistic data and evidence and which we summarise as
follows:1
According to the standard view, the empiricalness of linguistic theories
depends crucially on the use of the appropriate data type. This means that the
origin of data is decisive: according to generativists, introspection is a reliable data
source, while corpus linguists are of the opinion that corpora are the only legiti-
mate data sources. There is general agreement on the view that the treatment of
data does not require special attention, because their handling can be sufficiently
controlled with the help of a few simple, general methodological rules.
The standard view of linguistic data and evidence was for the most part
shaped by a second set of assumptions, namely, the standard view of the analytical
philosophy of science.2 A central idea of the latter was that empirical theories must
consist of statements which can be tested on the basis of a particular subset of data,
namely, evidence. It is the requirement that evidence has to be directly control-
lable, thus objective and totally reliable that lies at the core of the interpretation of
the concept ‘evidence’. From these features it follows that evidence was supposed
to be capable of justifying the hypotheses of theories and providing a firm base for
the decision between rival theories.
From the second half of the nineties onwards, there was a turn in the meth-
odological debates on data and evidence in linguistics: a series of collections of
papers and conferences tackled aspects of linguistic theorising that had been
neglected and had previously remained unreflected.3 The quotation below illus-
trates the problem that has recently been in the focus of interest:
[…] there is a growing awareness among formal linguists that a sole reliance on
introspective data […] will no longer do. Not only should speakers’ intuitions be
collected in a systematic way, but also should the database of linguistic theories be
broadened as to include types of data that well go beyond introspective data as the
primary data of linguistic theorizing (such as e.g. psycholinguistic and historical

.  See Kertész & Rákosi (2008a, b, 2012) for detailed overviews.
.  A recent textbook characterises the standard view of the analytical philosophy of science
as follows: “The two philosophies, logical postivism and Popper’s falsificationism, are usually
taken together as forming what is known as the classical tradition. Sometimes the term
‘­positivism’ is used. At other times, it is called the standard or the orthodox view. These latter
expressions are apt.” (Hung 2014: 311)
.  See, among others, Schütze (1996), Borsley (ed.) (2005), Lehmann (2004), Kallmeyer &
Zifonun (eds) (2007), Kepser & Reis (eds) (2005), Kristiansen et al. (eds) (2006), Penke &
Rosenbach (eds) (2004, 2007), Stefanowitsch & Gries (eds) (2007), Sternefeld (ed.) (2007),
Featherston & Winkler (eds) (2009), Winkler & Featherston (eds) (2009).
 Introduction 

evidence). This recent concern of formal linguistic theory is reflected in other

conferences/workshops more or less explicitly touching on the issue of linguistic
evidence as well […].
  The topic as such is huge. Not only does it touch on the database of linguistic
theories but also, inevitably, on methodological issues (evidence, of whatever type,
is only as good as the methodology by which it has been ascertained).
(Penke & Rosenbach 2007: vii; emphasis added)

Against this background, raising questions regarding the useable types of data/
evidence and their treatment in linguistic theories are well-motivated method-
ological issues. Nevertheless, there are several reasons why it is extremely difficult
to answer such questions. One reason is that linguistics is not a homogeneous
­discipline. Thus, both the diversity and the common aspects of the treatment of
linguistic data and evidence have to be accounted for. Another reason is that the
object of linguistic research is not directly available for the linguist, but has to
be reconstructed from the directly accessible manifestations of the knowledge of
language (see e.g. Kepser & Reis 2005: 1). Therefore, theoretical tools for the repre-
sentation and treatment of the indirectness and the uncertainty of linguistic data
resulting from them are needed, too.
Against these fundamental difficulties, it is understandable that the first
attempts in the current literature to solve methodological problems about linguis-
tic data and evidence are double-faced. On the one hand, contemporary views still
contain elements originating from the standard view of the analytical philosophy
of science. For instance, linguistic evidence is still interpreted as a specific subset of
data which are true with certainty and whose main task is to test (verify, confirm or
falsify) the hypotheses of the theory. On the other hand, the very same views have
aspects which are no longer compatible with the standard view of the analytical
philosophy of science and the standard view of linguistic data and evidence, and
which indicate the need to break with these norms. Every important characteristic
of the concept of evidence adapted from the analytical philosophy of science has
been questioned, and the new proposals raise a series of innovative ideas.4 These
progressive insights are, however, not acknowledged by all authors; and even if
they are accepted by several of them, they are accepted in different forms, so they
cannot be generalised. The most important new insights are the following:

–– The database of linguistic research cannot be reduced to introspective data vs.

corpus data, but a wide variety of different data types is accepted and deemed
legitimate.

.  See Kertész & Rákosi (2008a, b, 2012) for detailed overviews.
 András Kertész & Csilla Rákosi

–– Linguistic theorising cannot rely only on one data type but different data types
have to be combined in order to increase the reliability of the results and to
capture the complexity of linguistic behaviour.
–– All data types have their own weaknesses. Thus, the control of the factors
which might influence the reliability of the given data type always requires
special care.
–– Linguistic data do not secure a neutral basis for linguistic theorising because
they are problem- and theory-dependent.
–– The function of data cannot be reduced to the testing of hypotheses. Rather,
there is a cyclic relationship between them and the theory, which means that
data are used in several cycles of the raising, checking, modifying, refining,
and extending of hypotheses and hypothesis systems.
–– Data can function not just as tools of problem solving but may also lead to the
emergence of new problems because the application of different data types
often generates contradictions.

No doubt, the progressive insights may pave the way for a new solution to the data/
evidence problem. The question is: how? To find the answer, let us emphasise that
the discrepancy witnessed by the above double facedness is rooted in the unsatis-
factory nature of metascientific reflection. This means the following.
First, raising questions related to the reliability, combinability, complexity and
treatment of data presupposes the clarification of more fundamental and general
methodological problems: What is a datum? What is evidence? What is the struc-
ture of data/evidence? What is the function of data/evidence in linguistic theories?
etc. The current literature does not address these questions overtly and the latently
presupposed answers seem to be fragmentary and unsatisfactory.
Second, most linguistic theories cannot be interpreted as deductive systems
of hypotheses but rather as complex networks of more or less sophisticated argu-
ments and counter-arguments. These arguments and counter-arguments usually
rely on not perfectly reliable data and regard them as pieces of evidence. Thus, it
is indispensable to reveal the relationship between the structure and the function
of data and evidence on the one hand and the argumentation structure of theories
on the other.
Third, the main shortcoming of the literature on the data/evidence prob-
lem is that linguists are not in possession of an elaborated, systematic and solid
metatheoretical (methodological) background. As we have shown in Kertész &
Rákosi (2008a, b, 2012), most contributions to the data/evidence debate are based
primarily on the naive reflection of the linguists on their everyday research practice
and contain several metatheoretical prejudices. These metatheoretical prejudices
are mostly remnants of the standard view of linguistic data and evidence and lead
to apparent inconsistencies with the (accepted) practice of linguistic research.
 Introduction 

Therefore, progressive insights might lead to a satisfactory solution to the

problem of linguistic data and evidence only through the development of a novel
metascientific model. Against the background of the considerations above, the
new model of linguistic theorising should focus on five key points: capturing the
uncertainty of the data that linguistic theories are built on; the tools which make it
possible to draw plausible inferences from uncertain premises; the cyclic and pris-
matic nature of linguistic theorising based on uncertain data; the way linguistic
theories treat the inconsistencies between the hypotheses and the data or between
rival hypotheses; and the pluralism of linguistic theorising.
In Kertész & Rákosi (2012) we called such a model the p-model. In order to
motivate the chapters which the present volume consists of, in the next section we
will elucidate the key features of the p-model just mentioned.

3.  On the p-model

In accordance with these features, the p-model rejects a series of methodological

prejudices which mainstream linguistics takes implicitly or explicitly for granted,
but which are at variance with the practice of research. The main innovations of
the p-model are as follows:

a. Uncertainty instead of certainty. In order to overcome the double-facedness

summarised in the previous section, the p-model claims that all kinds of lin-
guistic data – such as introspective, corpus, experimental and many more
types of data – are substantially uncertain. However, the p-model also claims
that by accounting for their uncertainty, it can make problem solving in lin-
guistics more effective. Whereas in the standard view the search for the firm
empirical basis of linguistic theories presupposed the certainty of data, the
p-model assumes that although the empirical basis which linguistics is built
on is uncertain, realising this uncertainty could yield better theories. The tenet
that by capturing the uncertainty of linguistic data and evidence the effec-
tiveness of linguistic theorising may increase, seems at first sight paradoxical.
However, it is not, for reasons to be explained below.
b. Plausibility instead of truth. The p-model reinterprets the notions of data and
evidence and places them in the process of plausible argumentation. Lin-
guistic data are neither certain ‘facts’ themselves nor true statements about
‘facts’. Rather, their basic property is their uncertainty. The p-model explicates
uncertainty as plausibility. Accordingly, linguistic data are assumed to consist
of a statement and a plausibility value assigned to it on the basis of a source
such as linguistic intuition, a corpus, an experiment, etc. Their main function
is to feed plausibility values into the theory at issue. Linguistic hypotheses
 András Kertész & Csilla Rákosi

often receive a plausibility value as the conclusion of a plausible inference one

premise of which is a datum in the above sense. Evidence is a special kind of
datum that, as a premise of a plausible inference, increases or decreases the
plausibility of a hypothesis obtained as the conclusion of such an inference.
Accordingly, the function of evidence is to support the decision between alter-
native hypotheses.
c. The cyclic and prismatic nature of linguistic theorising. However, a linguistic
theory is not simply a static structure consisting of hypotheses, data, evidence
and plausible inferences. Rather, the p-model characterises linguistic theories
as dynamic processes of plausible argumentation. The main property of such a
process is that previously accepted hypotheses, data, evidence, methodologi-
cal norms and other kinds of information are retrospectively re-evaluated at
later stages of the argumentation. Such re-evaluation takes place in argumen-
tation cycles that are prismatic in the sense that they evaluate the data, the
evidence, the hypotheses etc. from different perspectives. Neither data nor
evidence are immune to this changeability. During the dynamic, cyclic, pris-
matic, and retrospective re-evaluation of information, the plausibility values
of alternative hypotheses are evaluated and compared in order to find an at
least provisional solution to the problem raised.5 Accordingly, this process of
plausible argumentation may be effective even though it is built on the substan-
tial uncertainty of data, evidence and hypotheses in linguistics.
d. Inconsistency-tolerance instead of consistency at all costs. From the uncertainty
of data and evidence it also follows that the role that contradictions play in lin-
guistic theorising has to be captured. Namely, since both data and hypotheses
are statements with a plausibility value, it is usually the case that at some stage
of the argumentation process both a statement and its negation are plausible
to a certain extent. Then, there may be several techniques to handle such an
inconsistency. For example, if the plausibility values are different, then the
less plausible statement may give way to the more plausible one;6 or they may

.  According to the p-model, one never makes a decision between isolated hypotheses
but compares whole theory-variants built around the rival hypotheses – by examining, for
example, the plausibility values of the data and hypotheses included by them as well as their
completeness and consistency.
.  As we have already indicated in Footnote 5, this is not automatically so. That is, for
example, if the acceptance of the less plausible hypothesis leads to a consistent theory-version,
while the more plausible hypothesis generates a great number of contradictions, then the less
plausible hypothesis can be a better choice. It may also happen that the less plausible hypoth-
esis is consistent with a set of hypotheses with very high plausibility value, while the more
plausible is incompatible with them, etc. That is, the decision is well-founded only if it takes
every aspect of the possible alternatives into consideration.
 Introduction 

be separated by assigning them to different contexts in a well-motivated way;

or the contradiction may be temporarily tolerated until, in the light of new
information, one of the contradicting statements can be ruled out. Thus in
the p-model inconsistency is not necessarily disastrous, but a natural prop-
erty of the process of linguistic theorising. One main task of plausible argu-
mentation is to govern the dynamism of the emergence and the resolution of
inconsistencies.
e. Pluralism instead of absolutism. From the p-model’s view, according to which
linguistic theories are prismatic, cyclic and retrospective processes involving
the continuous re-evaluation of data, hypotheses and further information,
it follows that basically there will be no single correct solution to problems
which one is entitled to defend by all means. It is not the case that linguistic
inquiry starts with a set of certain premises from which true conclusions are
inferred by deductive inferences. Instead, linguists have to check and revise
their knowledge continuously through the retrospective, cyclic and prismatic
re-evaluation process of plausible argumentation. Thereby, they are expected
to consider all plausible hypotheses at their disposal and to find out which of
them is the most acceptable at a particular phase of the process. This picture,
which the p-model draws of linguistics, is at variance with the unreflected and
absolutistic defence of particular theories and the unreflected and mechani-
cal rejection of their rivals. The pluralism of linguistic theories is fruitful and
should be utilised in order to obtain more reliable and better-founded solu-
tions to the problems of linguistic research.

4.  The structure of the book

The volume consists of three parts. Part I outlines the methodological frame-
work. Chapter 2 in Part I entitled The p-model of data and evidence in linguistics
by András Kertész and Csilla Rákosi summarises the line of reasoning that has
motivated the elaboration of the p-model, introduces its central concepts, infers its
central claims, illustrates them by a series of examples and puts forward a possible
metatheoretical solution to the data/evidence problem in linguistics.
The case studies in Part II systematically apply the p-model, as outlined in
Chapter 2, to object-scientific problems and show that its application can further
the solution of the particular problem raised.
Chapter 3 is Károly Bibok’s contribution entitled The plausibility of approaches
to syntactic alternation of Hungarian verbs. With the help of Kertész and Rákosi’s
notion of plausible argumentation, the author attempts to re-evaluate the p ­ lausibility
of the tenets that several different approaches to syntactically alternating verb classes
 András Kertész & Csilla Rákosi

have proposed. After a discussion of the lexical and constructional approaches to

syntactic alternations, he offers a lexical-constructional account of these verb classes
and, especially, an analysis of Hungarian locative alternation verbs. He shows that
both his conception in general and the analysis carried out are more plausible than
the rivals and the analyses the latter proposed. With respect to the evaluation of the
lexical-constructional theory, investigations presented in Bibok (2010) are also rel-
evant. The hypothesis of the lexical-constructional theory of syntactic alternations
remains plausible in a much wider context, namely, in that of lexical pragmatics.
Chapter 4 Methods and argumentation in historical linguistics: A case study by
Katalin Nagy C. aims to examine various characteristics of historical linguistics
argumentation with the help of a case study, applying the metalinguistic frame-
work of Kertész and Rákosi’s p-model. It presents three approaches to the problem
of the morphological aspect of the grammaticalisation of the Catalan construction
“anar ‘go’ + infinitive” as a perfective past. These three accounts form a context in
which some statements and their negations with respect to the problem discussed
in the case study are simultaneously plausible. After taking into account metalin-
guistic considerations based on the p-model in order to decide between the three
competing hypotheses and judge their reliability, the author concludes that the
development of the Catalan “anar ‘go’ + infinitive” construction is best explained
by Juge’s (2006) proposal.
Enikő Németh T.’s case study in Chapter 5 is entitled Hungarian verbs of
­natural phenomena with explicit and implicit subject arguments: their use and
occurrence in the light of data. Hungarian verbs of natural phenomena have been
considered subjectless by widely accepted Hungarian grammars up until the pres-
ent, although there is a considerable amount of data from different direct sources
(such as ­written corpora, intuition, introspection and spoken discourses) that
testifies to occurrences of verbs of natural phenomena with explicit subjects. On
the basis of these data as well as plausible inferences built on them, the chapter
provides a novel explanation of the occurrence of Hungarian verbs of natural
phenomena without and with explicit subjects. According to the proposed expla-
nation, Hungarian verbs of natural phenomena have subject arguments in their
lexical-semantic representations that can (or even should) be left implicit in cer-
tain circumstances, but in other ones they can be lexically realised. Assuming such
a lexical-semantic representation and taking into consideration the contextual fac-
tors, the occurrences of Hungarian verbs of natural phenomena without and with
explicit subjects can be accounted for in a unified way. Relying on Kertész and
Rákosi’s p-model, the present chapter also intends to reflect on the argumentation
process in the course of which the proposed explanation has been reached, and
answers the question of why the new account can be considered more plausible
than the previous ones.
 Introduction 

In Chapter 6 entitled The development of a taxonomy of verbal disagreements

in the light of the p-model, Helga Vanda Koczogh raises the problem of which
category system allows for describing Hungarian disagreement based on a set of
criteria laid down in advance. She introduces four rival approaches to disagree-
ment of which none can be regarded as a solution to this problem, since – among
other things – they leave the relationship of functional categories and modification
devices unclarified by either applying only one of the two aspects or mixing them
without reflection. The author provides a step-by-step account of the multi-step
and multi-method strategy that is adopted in order to create a category system
of her own. In the course of the development of the framework, the steps of the
argumentation process are reconstructed highlighting its cyclic, retrospective, and
prismatic nature in compliance with the p-model. The author proposes a cate-
gory system which distinguishes between disagreement functions (strategies) and
modification devices (PFMs) and argues that this framework is more complete
and reliable than its previous version or other rival solutions.
Chapter 7 by György Rákosi bears the title A case of disagreement: On ­plural
reduplicating particles in Hungarian. The literature on the grammar of the so-
called reduplicating particle construction in Hungarian is divided over whether
the particle can be plural by plural oblique associates or not. This chapter offers a
critical overview of this literature and a possible reconstruction of the argumen-
tation presented by the respective authors. The author carries out a reconstruc-
tion process within the framework of the p-model of Kertész & Rákosi (2012).
After a discussion of data not noted so far in the literature, he concludes that,
contra general assumptions, the plural and the non-plural reduplicating varieties
are both grammatical but they are instances of two different construction types.
The paraconsistent solution he offers to the problem recognises and motivates the
dichotomy between non-plural and plural reduplicating constructions as the co-
existence of an unmarked and a marked construction within a particular pocket
of Hungarian grammar.
In Chapter 8, which is the final case study of Part II, Zoltán Vecsey proposes A
plausibility-based model of shifted indexicals. According to Kaplan’s (1989) original
theory of indexicals, contexts of use cannot in principle be shifted by sentential
operators. In the last few years, a number of empirical investigations have come to
the surprising conclusion that in some languages, such as Amharic or Zazaki, oper-
ators are able to act on certain contextual parameters. Thus, the current theoretical
situation concerning the crosslinguistic applicability of the Kaplanian theory has
become inconsistent. One reasonable way out of this inconsistent situation is to
adopt a metatheoretical point of view concerning the relation between Kaplan’s
theoretical principle and the linguistic data gathered from various sources. Seen
from this point of view, the controversy around the possibility of context shift can
 András Kertész & Csilla Rákosi

be reduced to the question of the plausibility of the Kaplanian model of indexicals.

Thereby, Kertész and Rákosi’s p-model is made use of.
The two chapters in Part III are intended to solve meta-scientific problems
with the help of the p-model.
Chapter 9 comprises Kertész and Rákosi’s metatheoretical case study entitled
Thought experiments and real experiments as converging data sources in prag-
matics. One of the most relevant, but still unclarified, methodological aspects
of current pragmatics is the relationship between thought experiments and real
experiments. In order to contribute to the clarification of this issue, the authors
apply the p-model to Searle’s (1969) thought experiment on Kennst du das Land,
wo die Zitronen blühn? and to Holtgraves & Ashley’s (2001) real experiments on
speech act recognition. The case study yields the finding that in current pragmat-
ics thought experiments and real experiments as data sources converge and in the
course of their convergence their interaction is dynamic, multifaceted and con-
tinuous. The reconstruction of the way thought experiments and real experiments
as data sources converge rests on the process of plausible argumentation in the
sense of the p-model.
In Chapter 10 entitled Data and the resolution of inconsistency in Optimality
Theory, Csilla Rákosi analyses the most frequent methods of inconsistency resolu-
tion applied in Optimality Theory. The author shows that basically, there are two
groups of inconsistencies in OT. Conflicts between “linguistic data” (acceptability
judgements) and applications of the model (results of the evaluation procedure)
are regarded as fatal failures. They are usually eliminated by the modification of
the theory. Certain conflicts between elements of the theory (the constraints and
the best output candidates), however, are deemed to be tolerable. The p-model
allows us to reconstruct the first strategy as a case of the “contrastive” strategy,
while the second strategy can be reconstructed as the joint application of the “con-
trastive” and “combinative” strategies in this metatheoretical framework.
Finally, Chapter 11 by the editors summarises the main findings of the case
studies and draws a series of conclusions that highlight the relevance of the
p-model. In particular, this concluding chapter shows that the p-model can be
applied to quite different linguistic theories in a sophisticated and flexible manner.
In addition, and most importantly, it is also exemplified that the application of the
p-model is capable of increasing the effectiveness of linguistic argumentation.

Acknowledgements

Work on the present volume was supported by the MTA-DE Research Group for
Theoretical Linguistics, the project OTKA K 77823 as well as the T
­ ÁMOP-4.2.2
 Introduction 

/B-10/1-2010-0024 project, which is co-financed by the European Union and the

European Social Fund. Csilla Rákosi’s work was also supported by the János Bolyai
Research Scholarship of the Hungarian Academy of Sciences.

References

Bibok, Károly. 2010. From syntactic alternations to lexical pragmatics. In The Role of Data at the
Semantics–pragmatics Interface, Enikő Németh T. & Károly Bibok (eds), 261–304. Berlin:
Mouton de Gruyter.
Borsley, Robert D. (ed.). 2005. Data in Theoretical Linguistics. Special issue of Lingua.
Featherston, Sam & Winkler, Susanne (eds). 2009. The Fruits of Empirical Linguistics, Vol. 1:
Process. Berlin: De Gruyter.
Holtgraves, Thomas & Ashley, Aaron. 2001. Comprehending illocutionary force. Memory &
Cognition 29: 83–90.
Hung, Edwin. 2014. Philosophy of Science Complete. A Text on Traditional Problems and Schools
of Thought. Wadsworth: Cengage Learning.
Juge, Matthew L. 2006. Morphological factors in the grammaticalisation of the Catalan “go” past.
Diachronica 23(2): 313–339.
Kallmeyer, Werner & Zifonun, Gisela (eds). 2007. New York: De Gruyter (Institut für Deutsche
Sprache – Jahrbuch 2006).
Kaplan, David. 1989. Demonstratives. In Themes from Kaplan, Joseph Almog, John Perry &
Howard Wettstein (eds), 481–563. Oxford: OUP.
Kepser, Stephan & Reis, Marga (eds). 2005. Linguistic Evidence. Empirical, Theoretical and Com-
putational Perspectives. Berlin: De Gruyter.
Kepser, Stephan & Reis, Marga. 2005. Evidence in linguistics. In Kepser & Reis (eds), 1–6.
Kertész, András & Rákosi, Csilla. 2008a. Conservatism vs. innovation in the debate on data in
generative grammar. In Kertész & Rákosi (eds), 85–108.
Kertész, András & Rákosi, Csilla. 2008b. Conservatism vs. innovation in the (un)grammaticality
debate. In Kertész & Rákosi (eds), 61–84.
Kertész, András & Rákosi, Csilla (eds). 2008. New Approaches to Linguistic Evidence. Pilot Stud-
ies / Neue Ansätze zu linguistischer Evidenz. Pilotstudien. Frankfurt: Peter Lang.
Kertész, András & Rákosi, Csilla. 2012. Data and Evidence in Linguistics. A Plausible Argumenta-
tion Model. Cambridge: CUP.
Kristiansen, Gitte, Achard, Michel, Dirven, René & de Mendoza Ibáñez, Francisco J. Ruiz
(eds). 2006. Cognitive Linguistics: Current Applications and Future Perspectives. Berlin: De
Gruyter.
Lehmann, Christian. 2004. Data in linguistics. The Linguistic Review 21: 175–210.
Penke, Martina & Rosenbach, Anette (eds). 2004. What Counts as Evidence in Linguistics?
­Studies in Language 28(3): 481–747. Special Issue.
Penke, Martina & Rosenbach, Anette (eds). 2007. What Counts as Evidence in Linguistics?
[­Benjamins Current Topics 7]. Amsterdam: John Benjamins.
Searle, John R. 1969. Speech Acts. An Essay in the Philosophy of Language. Cambridge: CUP.
Schütze, Carson T. 1996. The Empirical Base of Linguistics. Grammaticality Judgments and
­Linguistic Methodology. Chicago IL: The University of Chicago Press.
Seuren, Pieter A.M. 1998. Western Linguistics. An Historical Introduction. Oxford: Blackwell.
 András Kertész & Csilla Rákosi

Stefanowitsch, Anatol & Gries, Stefan T. (eds). 2007. Grammar Without Grammaticality [Corpus
Linguistics and Linguistic Theory 3(1)].
Sternefeld, W. (ed). 2007. Data in Generative Grammar [Theoretical Linguistics 33(3)].
Winkler, Susanne & Featherston, Sam (eds). 2009. The Fruits of Empirical Linguistics, Vol. 2:
Product. Berlin: De Gruyter.
part i

The methodological framework

 chapter 2

The p-model of data and evidence in linguistics

András Kertész1, 2 & Csilla Rákosi2

1University of Debrecen, Institute of German Studies,
2MTA-DE Research Group for Theoretical Linguistics

The p-model is a metatheoretical approach to linguistic theorising that has been

designed to capture the nature of linguistic data and evidence in an unorthodox
and novel, but well-motivated and effective way. Its starting point is the insight
that the latest literature on linguistic data and evidence acknowledged the
uncertainty of the latter as their key feature. Since the main characteristic of
plausible inferences is that they facilitate drawing conclusions from uncertain
premises, and since chains of plausible inferences constitute the main body of
the process of plausible argumentation, the basic idea of the p-model is that
the structure of linguistic theories is based on various techniques of plausible
argumentation. This chapter summarises the main hypotheses of the p-model,
which the case studies in the volume are expected to apply to various linguistic
theories.

1.  The problem

In the Introduction to the present volume, we put forward a brief analysis of the
current debate on linguistic data and evidence. Against this background, the prob-
lem to be solved is as follows:
(P) (a) What kind of metascientific framework is capable of providing a novel
and comprehensive model of linguistic theorising?
(b) How can the concepts of ‘linguistic data’ and ‘linguistic evidence’ be
defined in such a way that the shortcomings of the current literature
can be avoided?

Our hypothesis is that an argumentation theoretic model of linguistic theorising

can be expected to yield the solution to (P)(a) and (b) by transforming the rather
vague assumptions of the progressive insights put forward in the current debate
into a workable account of linguistic data and evidence.
 András Kertész & Csilla Rákosi

Our line of reasoning will be as follows:

In Section 2, we will give an overview of a model of plausible argumentation
meeting this expectation. We will call it the p-model and consider it as a possible
solution to (P)(a). In Section 3, we will show how the p-model yields our solution
to (P)(b). Finally, in Section 4, we will summarise our findings.
In Sections 2 and 3 we will illustrate the concepts which we introduce and
the claims which we make predominantly (but not exclusively) by examples taken
from the case studies in the present volume. Accordingly, this chapter prepares the
application of the p-model in the subsequent case studies.

2.  A possible solution to (P)(a): The p-model

2.1  Introductory remarks

Before introducing the p-model, let us make the following preliminary remarks.
The current state of the art in the philosophy of science clearly shows that the
standard view of the analytical philosophy of science has lost its former accep-
tance and dominance. As a consequence, it has been replaced by a variety of new
approaches. This pluralism legitimises the elaboration of a metascientific model
which intends to solve (P)(a) in an unorthodox and novel, but well-motivated and
effective way.
Second, the basic idea of the p-model is that the structure of linguistic theo-
ries is based on various techniques of plausible argumentation. Such an approach
has been put forward in Kertész and Rákosi (2012) and is called the p-model of
plausible argumentation.1
Third, we emphasised that the latest literature on linguistic data and evidence
acknowledged the uncertainty of the latter as their key feature. Since the main
characteristic of plausible inferences is that they facilitate drawing conclusions
from uncertain premises, they seem to provide an appropriate tool for modelling
this feature of the relationship between linguistic data/evidence and the hypoth-
eses of linguistic theories. Therefore, one of the p-model’s central notions will be
that of ‘plausible inference’.
Finally, we want to emphasise that plausible inferences do not constitute a
calculus. Their theory is no ‘logic’, since plausible inferences take non-formalisable
characteristics of statements into consideration, too.

.  Our approach has been motivated by the ideas of George Polya and Nicholas Rescher (cf.
e.g. Polya 1948, 1954; Rescher 1976, 1987), and prepared e.g. in Kertész (1993, 2004a, 2004b)
and Kertész & Rákosi (2006, 2009).
 The p-model of data and evidence in linguistics 

Thus, in the present chapter we will confine the discussion of the basic notions
on which the p-model rests to the elucidation of the ideas motivating them with-
out striving for formalisation. For the more complete and more precise – but, for
the substantial reason mentioned, still informal – presentation of the p-model see
Kertész and Rákosi (2012).

2.2  The uncertainty of information: Plausible statements

In most cases our hypotheses are not statements the truth of which is guaranteed
by experience, supporting evidence, a proof or whatever. Rather, they are plausible:
we are ready to accept them to some extent on the basis of suitable sources that
support them. Such – more or less reliable – sources are, for example, corpora, the-
ories, conjectures, the intuition of native speakers, experiments, fieldwork, histori-
cal documents, dictionaries, videotapes, or even inferences (cf. Rescher 1976: 6f.;
Rescher 1973: 63ff., Kertész & Rákosi 2012: 63ff.).
In order to capture this inherent uncertainty and the limited reliability of
information, the p-model – following Rescher (1976) – interprets scientific
hypotheses not as propositions but assigns them a structure consisting of an infor-
mation content and a plausibility value. This plausibility value does not belong to
the logical structure of the statements. It shows to what extent the given statement
is supported by a source; that is, how reliable the source makes it, and to what
extent one is willing to accept the hypothesis on the basis of the source at issue.
Plausibility is source-dependent. A statement may be very plausible according
to one source, and less plausible with respect to others. It may also happen that
some sources support the negation of the given statement and make it implausible.
Further, if several sources support a statement, then its plausibility value is higher
on the basis of all sources together than its plausibility value on the basis of any of
the sources alone.

Example 1. The statement

The lexical-semantic representation of weather verbs contains an internal
argument which cannot be expressed syntactically

is a plausible statement on the basis of a linguistic paper (Komlósy 1994) as a

source, because it is expressly supported by it. Nevertheless, this paper cannot be
regarded as a perfectly reliable source guaranteeing the truth of this statement.
The uncertainties of such a source can be formulated from different points of view:

–– The hypothesis at issue is an inductive generalisation covering several cases

(i.e. it makes a claim about all Hungarian weather verbs). Therefore, it has to
be checked whether there are weather verbs in Hungarian which may occur
with semantic arguments that can be explicitly realised.
 András Kertész & Csilla Rákosi

–– The linguistic intuition of the linguist (in this case, that of Komlósy) may be
influenced and rendered unreliable by the theory he relies on. Thus, he may be
mistaken about the analysis of the sentences at issue and state incorrectly that
they are subjectless in this sense.
–– It might happen that other native speakers judge the sentences containing the
given construction differently.

Németh T. (this volume) shows that after involving further sources (arguments
relying on introspective data) this hypothesis has to be regarded as implausible.

Example 2. The hypothesis

The reinterpretation of the construction “anar + infinitive” as a ­construction
with a present tense auxiliary in Old Catalan is based on the high frequency
of the 1Pl form anam ‘we go/we are going/we went’, which is ambiguous
­between past and present tense and allows an association to the present
tense paradigm of the verb anar.

can be deemed plausible, among others, on the basis of the frequency analysis
provided in Juge (2006) as a source. Nevertheless, Nagy C. (this volume) refers to
two factors which may diminish the reliability of this source: “First, in other cases
of analogical change the basicness of third persons is emphasised. Second, it is
questionable whether the frequency will remain the same, when investigating it
in a larger corpus.”

Plausibility can be represented numerically (cf. Rescher 1976). These values, how-
ever, do not have any exact meaning which could be applied in every context but
solely indicate different relative strengths of plausibility (reliability, supportedness,
acceptance) within one argumentation process (theory). The minimum value of
plausibility is indicated by 0, which means that the given statement is of neutral
plausibility on the basis of the source at issue. The maximal value is indicated by
1 meaning that the statement in question is true with certainty on the basis of the
given source.
To sum up, the plausibility value of the statement p on the basis of the source
S is such that

a. |p|S = 1, if p is true with certainty on the basis of S;

b. 0 < |p|S < 1, if p is plausible on the basis of S;
c. 0 < |~p|S < 1, if p is implausible on the basis of S;
d. |p|S = 0, if p is of neutral plausibility on the basis of S, i.e. if it is neither plau-
sible nor implausible on the basis of this source.
 The p-model of data and evidence in linguistics 

Example 3. In a theory on discontinuous constituents, for example, the following

plausibility values may be assigned to the statements in question:

a. |The sentence Elements moving rightward cannot be moved out of the clause
in which they originate contains the cognitive metaphor Clauses are
containers.|S = 0

where S is the Theory of Government and Binding (GB Theory). This statement
has neutral plausibility according to the source S, because neither it nor its nega-
tion is supported by the source S. The reason for this is that GB Theory cannot say
anything about the acceptability of this hypothesis because it does not make use of
concepts such as ‘cognitive metaphor’.

b. |Degree word complements form deep and surface structure discontinuous

constituents with degree words.|S = 0.2

If this statement has not been tested yet but is only a conjecture raised at this
moment, then it has to be assigned a low plausibility value on the basis of S which
is an intellectual source.

c. |Degree word complements form deep and surface structure discontinuous

constituents with degree words.|S = 0.4

It may be the case that the hypothesis in (b) has undergone some testing with the
help of a thought experiment conducted by the linguist. That is, if he/she has tried
to think over its consequences and examined its tenability on the basis of a few
acceptability judgements as evidence, then this hypothesis can receive a somewhat
higher plausibility value on the basis of a thought experiment as source S than in the
former case. Nevertheless, this plausibility value cannot be high, since the testing
was unsystematic and relied solely on the linguistic intuition of the given linguist.

d. |In sentences containing both a degree word complement clause and a relative
clause extraposed from subject position, the relative clause has to precede the
degree word complement clause.|S = 0.6

Suppose that this hypothesis has been checked by a linguist with the help of several
pairs of sentences such as People were so angry who knew John that they refused to
participate vs. *People were so angry that they refused to participate who knew John.
In this case, the above hypothesis can be judged to have a rather high plausibility
value on the basis of the linguist’s linguistic intuition as well as his/her linguistic
 András Kertész & Csilla Rákosi

analyses as a source S. Nevertheless, we must not forget that the judgements which
are rooted in this source may be controversial.

e. |The sentence People were so angry who knew John that they refused to par-
ticipate is grammatically correct, while the sentence People were so angry that
they refused to participate who knew John is ungrammatical.|S = 0.8

This statement concerns only the grammaticality of two sentences. Nevertheless,

even if S is the linguistic intuition of the linguist and he/she is convinced that
his/her judgement is doubtlessly correct, the above statement cannot be regarded
to be true with certainty. It may obtain a high plausibility value which may be
increased by involving further persons in the judgement of the two sentences’
grammaticality.
Of course, the highest plausibility value is provided by a well-designed experi-
ment with an appropriate statistical evaluation of the results.

It has to be stressed that low plausibility values do not mean improbability but rather,
a relatively small, limited amount of plausibility (reliability, acceptance). In such
cases, the source votes expressly for the given hypothesis. If a source is against a
hypothesis then it makes its negation plausible and the given hypothesis implausible.
We distinguish between direct and indirect sources. In the case of direct
sources, the plausibility of the statement at issue is evaluated with respect to the
reliability of the source. With indirect sources, however, the plausibility value of
the given statement is determined with reference to the plausibility of other state-
ments – that is, with the help of an inference.

Example 4. Nagy C. (this volume) judges the following hypothesis plausible:

The reinterpretation of “anar + infinitive” as a periphrasis with a present
tense auxiliary is realised due to the high frequency of the 1Pl form anam.

If she simply referred to Juge (2006) without presenting his argumentation, then
the plausibility value would be assigned to this statement on the basis of this paper
as a direct source. Nagy C., however, presents Juge’s argumentation in detail.
Therefore, the above hypothesis is supported by a series of plausible inferences as
an indirect source.

2.3  O
 btaining new information from uncertain information:
Plausible inferences
Deductive inferences the premises of which are true with certainty are capable of
guaranteeing the truth of the conclusion. That is, if the premises of an inference are
true with certainty, and there is a logical consequence relation between them and
the conclusion, then we are totally justified in accepting the conclusion. However,
 The p-model of data and evidence in linguistics 

very often either there is no logical consequence relation between the premises
and the conclusion or at least one of the premises, instead of being certainly true, is
only plausible in the light of the given sources. Moreover, the combination of these
two cases may also occur. Although such inferences are not capable of securing the
certainty of the conclusion (assured hypothetically or actually by the premises),
they may – under appropriate circumstances – make the conclusion plausible.
Plausible inferences are mappings associating a set of plausible and/or true
statements (premises) with a plausible statement (conclusion). In such inferences
the connection between the premises and the conclusion cannot be reduced to the
relationship between their logical structures, as is the case with deductive infer-
ences. Plausible inferences take into consideration not only the logical structure of
the premises and the conclusion but their plausibility values and semantic struc-
ture as well. They rest on a semantic relation: for example, causality, analogy, simi-
larity, sign, necessary or sufficient condition, part-whole relation etc.
The first type of plausible inferences consists of cases in which there is a logical
consequence relation between the premises and the conclusion but at least one of
the premises is only plausible and not true with certainty.

Example 5. Let us illustrate this by an example taken from Bibok (this volume):

0 < |If there is a verb with the semantic representation ‘X causes Y to

move into/onto Z’, then it can be converted into a verb with the semantic
representation ‘X causes Z to change state by means of moving Y into/onto
it’. |B < 1
0 < |The verb ken (X Y-t Z-re) ‘smear (X, Y on Z)’ is a verb with the semantic
representation ‘X causes Y to move onto Z’.|B < 1
0 < |The verb ken (X Z-t Y-nal) ‘smear (X, Z with Y)’ can be converted into a
corresponding verb with the semantic representation ‘X causes Z to change
state by means of moving Y onto it’.|I < 1

The first premise is a conditional sentence. It captures a lexical rule proposed for
the locative alternation of the verb ken ‘smear’ (cf. Pinker 1989: 79). This hypoth-
esis results from an inductive generalisation about the syntactic behaviour and the
semantic structure of a certain verb class. Although we cannot state that the first
premise is true with certainty, it is legitimate to deem it plausible on the basis of the
two sources mentioned. The second premise is plausible on the basis of theoretical
considerations. That is, it can be hypothesised but one cannot claim with certainty
that it is true. Due to the logical structure of the premises and the conclusion, there
is a logical consequence relation between the two premises of this inference.
It is clear that these premises are not sufficient to support the conclusion per-
fectly. If the premises were true with certainty, then the conclusion could also be
considered as certainly true. If the plausibility value of one of the premises were 0,
 András Kertész & Csilla Rákosi

then the conclusion would be of neutral plausibility, too. It is also clear that if the
plausibility of the premises increases (decreases), then the plausibility of the con-
clusion will also move in the same direction. From this it follows that if both prem-
ises are plausible statements, then one is allowed to suppose that the plausibility
of the conclusion lies between the two cases mentioned – that is, it is a plausible
statement, too.

The result of Example 5 can be easily generalised: If a logical consequence relation

and some semantic relation hold between the premises and the conclusion, the
premises make up a consistent set of statements, all premises have a positive plau-
sibility value (that is, they are either plausible or true with certainty) on the basis
of some set of sources S, while at least one of them is not true with certainty, then
the conclusion is plausible on the basis of this inference as a source. This means
that inferences belonging to the first type of plausible inferences can be regarded
as indirect sources that partially support their conclusion.
It may happen that the statements playing the role of the premises are consis-
tent, they are true with certainty according to a set of sources, and a semantic rela-
tion can be reconstructed that connects them with the conclusion – but there is
no logical consequence relation between them and the conclusion. Such instances
form the second type of plausible inferences. What matters in all these cases is
that the connection between the premises and the conclusion cannot be reduced
to the relationship between their logical structure, as with deductive inferences.
If there is no logical consequence relation between the premises and the conclu-
sion, then latent background assumptions are also needed (Rescher 1976: 60 ff.,
Polya 1948: 223). These background assumptions have to fulfil certain logical
and semantic criteria (see Kertész & Rákosi 2012: 88ff.), and they have to be
true, plausible, or at least not known to be implausible or false according to some
source.

Example 6. With inductive inferences, the set of the premises has to be supple-
mented by the background assumption that the cases not examined also possess
the characteristics that could be found in the investigated ones; the conclusion
states the presence of these characteristics as a general rule. Let us examine the
following example from Rákosi (this volume):
|In Axininca Campa, it is the output-form [noŋkomati] that belongs to the
input /no-N-koma-i/.|K = 1
1
|In Axininca Campa, it is the output-form [noŋkomataati] that belongs to
the input /no-N-koma-aa-i/.|K = 1
1
|In Axininca Campa, it is the output-form [noŋkomatakoti] that belongs to
the input /no-N-koma-ako-i/.|K = 1
1
 The p-model of data and evidence in linguistics 

|In Axininca Campa, it is the output-form [noŋkomatakotaatiro] that

­belongs to the input /no-N-koma-ako-aa-i-ro/.|K = 1
1
|In Axininca Campa, it is the output-form [noɲčhiki] that belongs to the
input /no-N-čhik-i/.|K = 1
1
|In Axininca Campa, it is the output-form [noɲčhikaati] that belongs to the
input /no-N-čhik-aa-i/.|K = 1
1
|In Axininca Campa, it is the output-form [noɲčhikakoti] that belongs to
the input /no-N-čhik-ako-i/.|K = 1
1
|In Axininca Campa, it is the output-form [noɲčhikakotaatiro] that belongs
to the input /no-N-čhik-ako-aa-i-ro/.|K = 1
1
[|For every other input form which contains a syllable without an onset in
Axininca Campa, the corresponding output form has a consonant inserted
into the onset position.|K = 0]
1
0 < |Syllables must have an onset Axininca Campa.|I < 1
1

Since Kager never indicates that statements of this type could turn out to be
false and be in need of revision, he seems to assign a very high or even maximal
plausibility value to the premises; in this reconstruction, we will choose the lat-
ter interpretation. At this point of his argumentation, he does not mention fur-
ther arguments which could support or refute the latent background assumption
which covers all cases not mentioned in the premises. Therefore, it has neutral
plausibility on the basis of K1, that is, the relevant passages of Kager (1999) as a
direct source. At a later stage of his argumentation (abbreviated as source K2),
however, Kager presents data which indicate that the latent background assump-
tion is implausible:
|In Axininca Campa, it is the output-form [osampi] that belongs to the
input /osampi/ and not the form *[tosampi].|K = 1
2
|In Axininca Campa, it is the output-form [iŋkomati] that belongs to the
input /i-N-koma-i/ and not the form *[tiŋkomati].|K = 1
2
[|For every other input form which contains a syllable without an onset in
Axininca Campa in word-initial position, the corresponding output form
has no onset word-initially.|K = 0]
2
0 < |Word-initial syllables do not have an onset in Axininca Campa.|I < 1
2

Thus, if we take into consideration the information provided by this source too,
and evaluate the plausibility value of the premises on the basis of the sources K1
and K2 together, then the first inference is no longer capable of supporting the
plausibility of the conclusion. This is due to the circumstance that it relies on a
latent background assumption which is implausible instead of being at least of
neutral plausibility.
 András Kertész & Csilla Rákosi

Example 7. In other cases the transformation of an inference of Type 2 into a

deductively valid inference is not confined to the addition of further statements,
but some premise-candidates have to be deleted or modified. This is the case with
inferences which are called ‘reduction’. Let us see the following instance of this
inference type:2
|If Axininca Campa is characterised by the constraint hierarchy Onset >>
Dep-IO, then the output-candidate [noŋkomati] is more optimal than the
output-candidate [noŋkomai].|K = 1
|The output-candidate [noŋkomati] is more optimal than the output-­
candidate [noŋkomai].|K = 1
0 < |Axininca Campa is characterised by the constraint hierarchy Onset >>
Dep-IO.|I < 1

There is no logical consequence relation between the premises and the conclusion.
This inference can be transformed into the following inference which is deduc-
tively valid:
[|If to every input form belongs the output form on the surface which
is ­optimal according to an evaluation based on the constraint ranking
Onset >> Dep-IO, then Axininca Campa is characterised by the constraint
­hierarchy Onset >> Dep-IO.|S = 1]
[|The output form [noŋkomati] is optimal according to an evaluation based
on the constraint ranking Onset >> Dep-IO.|S = 0]
0 < |It is the form [noŋkomati] that belongs to the input /no-N-koma-i/ at
the surface representation.|S ≤ 1
[|In every other case too, to the input form belongs the output form on the
surface which is optimal according to an evaluation based on the constraint
ranking Onset >> Dep-IO.|S = 0]
0 < |Axininca Campa is characterised by the constraint hierarchy Onset >>
Dep-IO.|I < 1

The transformation of the premises maintains the semantic relations between

the statements which occur in this inference. Thus, while the first premise
of the original inference presents a testable consequence (necessary condition) of
the application of the proposed constraint hierarchy (namely, the evaluation of two
output-candidates on the basis of this ranking), the first latent background assump-
tion of the transformed inference reverses this relationship and formulates a suf-
ficient condition of the correctness of this constraint hierarchy. Since at this point

.  See also inference (29) in Rákosi (this volume).

 The p-model of data and evidence in linguistics 

of the argumentation we do not possess further information about the evaluation

procedure (that is, about the evaluation of all possible output-candidates), the sec-
ond and third latent background assumptions have to be judged to be of neutral
plausibility. From this it follows that the conclusion can be assigned a low plausibil-
ity value on the basis of this inference as an indirect source. As is shown in Rákosi
(this volume), the next step of Kager’s argumentation revises the plausibility value
of the second latent background assumption, since it turns out to be implausible.

Our considerations in connection with Example 6 and 7 can be generalised as

follows. If there is no consequence relation between the premises and the conclu-
sion, then the semantic relation between the premises and the conclusion has to
be extendable to further statements which make the resulting inference deductive.
That is, the set of premises has to be transformed into another set of statements
which contains, besides one or more premises, some latent background assump-
tions as well. If the premises of the inference resulting from the transformation are
true, plausible, or at least of neutral plausibility, then an inference can be obtained
from the statements and sources that assigns a plausibility value to the conclusion
as an indirect source.
The combination of these two cases may also occur: with inferences belong-
ing to the third type of plausible inferences, there is no logical consequence relation
between the premises and the conclusion, and there is at least one premise which
is not certainly true but only plausible.

Example 8. Most inferences capturing the (positive) result of hypothesis testing

belong to this group, since usually not all premises can be regarded to be true with
certainty but are only plausible according to some source. For example, in B ­ altin
(1987) the following hypothesis is tested, among others (see Kertész & Rákosi 2013):

(H) Sentential complements are generated in the deep structure between the
­degree word and the adjective in the case of too, so and after the degree
word in the case of enough as sisters of the degree word. Then they get
moved rightward and adjoined to S’ in the surface structure (i.e. they are
‘extraposed’).

From (H) it follows that in sentences containing a relative clause extraposed from
subject position and a degree word complement clause (sentential complement),
the relative clause precedes the degree word complement clause – supposing that
relative clauses are adjoined to S’. (H) was tested with the help of the following data:

(D) The sentence People were so angry who knew John that they refused to
participate is grammatically correct, while the sentence People were so angry
that they refused to participate who knew John is ungrammatical.
 András Kertész & Csilla Rákosi

(D) can be regarded to be plausible – but not certainly true – on the basis of Baltin
(1987) as a source, since this source is basically nothing else but Baltin’s linguistic
intuition. From this we obtain the following plausible inference:
0 < |If (H), and relative clauses are adjoined to S’, then the sentence
People were so angry who knew John that they refused to participate is
­grammatically correct, while the sentence People were so angry that they
refused to participate who knew John is ungrammatical.|B < 1
0 < |The sentence People were so angry who knew John that they refused to
participate is grammatically correct, while the sentence People were so angry
that they refused to participate who knew John is ungrammatical.|B < 1
0 < |(H), and relative clauses are adjoined to S’.|I < 1

There is no logical consequence relation between the premises and the conclusion
but there is a semantic relation between them, both premises are plausible on the
basis of the same source and the inference can be transformed into a plausible infer-
ence of Type 1. Relying on this inference as an indirect source one may assign only
a relatively low plausibility value to (H). This plausibility value can be increased, for
example, if one checks (H) with the help of further “linguistic data” like (D).

The three cases also witness that plausible inferences are fallible. First, at the out-
set their conclusion is not true with certainty but only plausible to some extent;
that is, the conclusion gets only partial support from the premises. Second, plau-
sible inferences with latent background assumptions are especially liable to mislead
because if one of the latent background assumptions turns out to be implausible or
false, then the inference will be insufficient to establish the plausibility of the conclu-
sion even if the premises are true or plausible. Moreover, taking into consideration
a wider scope of sources, one may come into possession of information that makes
the premises, the latent background assumptions or the conclusion implausible or
false. Therefore, it is always the case that the conclusion is plausible only relative to
the premises, the latent background assumptions and the sources supporting them.

2.4  The p-context and the p-context-dependency of plausible inferences

From the above characterisation of plausible inferences it is clear that the rela-
tion between the premises and the conclusion cannot be reduced to their formal
properties. Beyond their logical structure, we have to take into consideration all
information that may be relevant for judging the plausibility value of the premises
and latent background assumptions as well as the semantic relation between them
and the conclusion. This motivates the introduction of the notion of p-context,
which will serve as the background against which plausible inferences can be put
 The p-model of data and evidence in linguistics 

forward, used and evaluated. The p-context includes, first, a set of sources in terms
of which the plausibility value of statements can be judged. Second, it covers a
set of statements together with their relevant characteristics (in particular, their
plausibility values with respect to the sources in the p-context, and their logical
and semantic structure). Third, the accepted methodological norms related to the
components of the p-context (for instance, the permissible type(s) of inference,
the criteria of the acceptability of the statements, the methods of the treatment
of inconsistencies, the criteria for judging the reliability of the sources etc.) also
belong to the p-context.
Plausible inferences are in several respects p-context-dependent. First, the
p-context facilitates plausible inferences only in cases in which the premises are
plausible or true according to some sources in the context. In addition, plausible
inferences have to be legitimate inference types in the given p-context. Second,
if the p-context changes so that new sources become available which may influ-
ence the plausibility value of the premises (by making them more plausible or less
plausible or even implausible), then the plausibility of the conclusion will change
as well. Third, if there is no logical consequence relation between the premises and
the conclusion, then the plausibility of the conclusion does not only depend on
the plausibility of the premises, but is also influenced by every piece of informa-
tion related to the latent background assumptions. From this it follows that any
new information which strengthens or weakens the plausibility of the premises
and that of the latent background assumptions influences the plausibility of the
conclusion as well. This also means that plausible inferences are dynamic. They
are indirect sources which are sensitive to changes in the p-context. Fourth, there
are often many (direct or indirect) sources that allow a plausibility value to be
assigned to a statement. In such cases, the plausibility of this statement is higher
than its plausibility on the basis of the individual sources.

Example 9. To illustrate the p-context-dependency of plausible inferences, let us

examine an example. Vecsey (this volume) presents and analyses arguments for
and against Kaplan’s Fixity Thesis:

(FT) The denotation or propositional contribution of an indexical expression a is

fixed solely by the relevant parameters of its actual context of occurrence c,
and cannot be influenced by any sentential operators.

According to Kaplan’s (1989) argumentation, in English there is no context-shift.

As Vecsey (this volume) emphasises, the same holds, for example, for Hungarian
or German. Therefore, the following plausible inference (inductive generalisation)
presents itself:
 András Kertész & Csilla Rákosi

Inference 1:
0 < |In English, there is no context-shift.|S < 1
1
0 < |In Hungarian, there is no context-shift.|S < 1
1
0 < |In German, there is no context-shift.|S < 1
1
[|In all other languages, too, there is no context-shift.|S = 0]
1
0 < |There are no context-shifting languages.|I < 1
1

The latent background assumption has neutral plausibility on the basis of the
sources within this p-context (S1), since they do not contain more relevant
information.
Thus, if we test (FT) on the languages mentioned, we obtain the following
plausible inference:
Inference 2:
|If (FT) is crosslinguistically applicable, then there are no context-shifting
languages.|S1 = 1
0 < |There are no context-shifting languages.|I < 1
1
0 < |(FT) is crosslinguistically applicable.|I < 1
2

On the basis of acceptability judgements gained from native speakers of Amharic

and Zazaki presented in Schlenker (2003) and Anand & Nevins (2004), however,
plausible inferences 3 and 4 can be put forward:
Inference 3:
0 < |In Amharic, there is context-shift.|S < 1
2
0 < |In Zazaki, there is context-shift.|S < 1
2
0 < |There are context-shifting languages.|I < 1
3
Inference 4:
|If (FT) is crosslinguistically applicable, then there are no context-shifting
languages.|S = 1
1
0 < |There are context-shifting languages.|I < 1
3
0 < |(FT) is not crosslinguistically applicable.|I < 1
4

If one extends the p-context by further data which indicate the presence of
­context-shift in further languages, then the plausibility value of the second premise
increases, and, as a consequence, that of the conclusion increases, too. In contrast,
in this extended p-context the first inference is no longer capable of supporting
its conclusion. The reason for this is the circumstance that its latent background
assumption is not plausible on the basis of all available sources (S1 and S2). As a
consequence, the second inference loses its capability to support its conclusion,
 The p-model of data and evidence in linguistics 

because its second premise is not plausible on the basis of all available sources in
this extended p-context.
According to Vecsey (present volume), however, the situation changes again,
if we add the results of Stechow’s (2003) considerations to the p-context. He claims
that the argumentation presented by Schlenker (2003) and Anand & Nevins (2004)
is faulty and they cannot be regarded as reliable sources. In this p-context, the
source S2 loses its reliability and is no longer capable of supporting the premises of
Inference 3. Therefore, Inference 4 cannot be regarded as a reliable indirect source
either, which could make its conclusion plausible.

2.5  Problems, their solution and their resolution

In a p-context the sources may yield too much information in the sense that there
is a statement which is made plausible by some source while its negation is made
plausible by another. In such cases, the p-context is informationally overdetermined
(Rescher 1976: 2; Rescher & Brandom 1980: 3ff.) and the set of the plausible state-
ments in the p-context is p-inconsistent. Nevertheless, the p-context may be infor-
mationally underdetermined as well (Rescher & Brandom 1980: 3ff.). A typical case
of the informational underdetermination of the p-context is its p-incompleteness,
insofar as there are statements which are neither plausible (in the extreme case:
true with certainty) nor implausible (in the extreme case: false with certainty) with
respect to any source given. A p-context may be simultaneously informationally
under- and overdetermined.

Example 10. As Németh T. (this volume) shows, the p-context of Komlósy’s (1994)
lexical-functional approach becomes p-inconsistent by adding further data to it.
Namely, in the p-context of Komlósy (1994) the following statement receives a high
plausibility value, since it is supported by several arguments (plausible inferences):
(H) Hungarian verbs of natural phenomena cannot occur with an explicit
subject and do not license a formal subject, either.

Németh T., however, presents corpus data which support the opposite standpoint:
(~H) There are verbs of natural phenomena in Hungarian that can occur with
explicit subjects.

Both (H) and (~H) are assigned a high plausibility value by their sources. Thus, the
extended p-context is p-inconsistent.

Example 11. As Nagy C. (this volume) points out, Detges (2004) cannot be
regarded as a useable source on the basis of which it would be possible to assign a
plausibility value to the statement
 András Kertész & Csilla Rákosi

The Old French form (il) va dire is a present form [meaning ‘all of a sudden,
he says’].

or to its negation
The Old French form (il) va dire is not a present form [but it is in the
preterit, meaning ‘suddenly, he said’].

The reason for this lies in the circumstance that Detges refers in both cases to the
same source, namely to Gougenheim (1971 [1929]: 96f.). From this it follows that in
the p-context of Detges (2004), the above statement is of neutral plausibility; thus,
this p-context is informationally underdetermined, because it is p-incomplete.

We call instances of p-inconsistency and p-incompleteness p-problems. If a

p-­context is characterised by over- and/or underdetermination in this sense, then
it is called p-problematic.3 In order to solve a p-problem, we have to re-evaluate
the p-context by revising its elements (see Section 2.6 on this). A solution of a
p-problem is achieved if a p-context has been arrived at in which either

a. the statement in question is unanimously supported or opposed by the

sources, that is, it is either plausible or implausible on the basis of all sources
in the given p-context,4 or
b. the statements generating p-inconsistency can be represented in such a way that
they become separated systematically and this separation is well-motivated.5

It is possible, however, that a p-problem has several solutions. This necessitates

the introduction of the notion of the resolution of a p-problem. We resolve a
p-problem if we find a solution of the given p-problem which is, when compared
with other solutions, the best according to a particular set of accepted criteria
and according to the information available. It may be the case, however, that
in an informational state one can only show that for the time being there is no
resolution achievable.

Example 12. Nagy C. (this volume) presents the p-context consisting of investi-
gations into the morphological aspects of the history of the Catalan “anar ‘go’ +
infinitive” construction. This p-context is p-inconsistent, because the statement

.  These definitions can be generalised. Thus, cases of informational over-/underdetermina-

tion not necessarily related to plausibility are called problems.
.  Such cases will be called consistent solutions.
.  In such cases, we speak of a paraconsistent solution resulting from the superposition of
two solutions which contain only one member of the rival hypotheses. Cf. Subsection 2.6.2
and Rescher & Brandom (1980).
 The p-model of data and evidence in linguistics 

(H) The Catalan GO-past comes from the construction “GO-verb conjugated in
present tense + infinitive”

as well as its negation are plausible according to some source. Nagy C. analyses
three solutions proposed in the literature:
(S1) Colon (1978a, b) presents arguments which unanimously support (H). He
states that the current meaning of “anar + infinitive” can be derived from
the present auxiliary variant. The reason for the loss of preterit forms in the
auxiliary was a historical present use which pushed the preterit forms of the
auxiliary into the background. In this p-context version, it is the periphrasis
with present auxiliary which undergoes the semantic change. The spread of
the present auxiliary forms is interpreted as being the result of an extension
of historical present usage.
(S2) 
Detges (2004) deems (H) plausible and (~H) implausible and agrees with
Colon on the hypothesis that the current meaning of “anar + infinitive”
can be derived from the present auxiliary variant. He claims that the tense
of the auxiliary is optional, although the present-tense variant was used
more frequently at all times. He also maintains that the present auxiliary
periphrasis conventionalised as a foreground-marker due to its high
frequency.
(S3) 
Juge (2006) rejects (H) and supports (~H). According to this p-context
version, it is not the periphrasis with present auxiliary which undergoes
the semantic change, but the periphrasis with preterit auxiliary. Further,
the spread of the present auxiliary forms is the result of a reinterpretation,
­triggered by syncretism in one of the most frequent forms of the paradigm.

To sum up, while in the solutions provided by Colon and Detges a high plausi-
bility value is assigned to (H), and (~H) is rejected, Juge proposes the opposite:
giving up (H) and keeping (~H). Thus, all three proposals are consistent solutions
of the starting p-problem along the lines of point (a) mentioned above (which
says that the statement at issue is unanimously supported by the sources); that is,
according to these authors, one of the rival hypotheses has to be given up and only
the other one can be deemed to be plausible. With the help of the extension of
the p-­context by further sources and data, Nagy C. argues that the first two solu-
tions are untenable, because the plausibility of their hypotheses is lost or at least
decreases considerably and the p-inconsistency of these p-contexts increases. In
addition, the solution proposed by Juge (2006) can be regarded as the resolution
of this p-problem because it does not contain p-inconsistencies and its hypotheses
have a high plausibility value on the basis of the sources presented by Nagy C.

Example 13. Nagy C. (this volume) mentions a solution of a p-problem that is based
on the method described in point (b) above, that is, on the systematic separation of
the conflicting statements. As she points out, in Colon’s account both the hypothesis
 András Kertész & Csilla Rákosi

that the GO-periphrasis has a past tense value in Old Catalan and its rival, accord-
ing to which the periphrasis manifests a historical present use, are plausible. This,
however, leads to p-inconsistency. Namely, from the first hypothesis it follows that
the GO-verb does not refer to real motion but is grammaticalised to some degree.
If the second hypothesis holds, then the GO-verb should refer to real motion. Nev-
ertheless, we can maintain both hypotheses if we separate their field of application
and suppose that they refer to different periods in the development of Old Catalan.

2.6  The problem solving process

2.6.1  Plausible argumentation
To achieve the solutions or the resolution of a given p-problem, a process is con-
ducted which elaborates and compares as many solutions as possible. It is this heu-
ristic process that we will call plausible argumentation. In simple terms, plausible
argumentation is the gradual transformation of a p-problematic context into one
which is no longer (or at least, less) p-problematic. This involves the successive
re-evaluation of a p-problematic p-context by the elaboration of possible solutions
to its problems and the evaluation and comparison of the alternative solutions. Its
aim is to detect all available solutions and decide which of them is to be accepted
as the resolution of the given p-problem.
The above characterisation of plausible argumentation indicates that the argu-
mentation process is basically not linear, because the re-evaluation of a p
­ -problematic
p-context usually does not lead immediately to an unproblematic one but may raise
new problems. This may require the revision of previous decisions, the assess-
ment of other alternatives etc. Therefore, throughout the argumentation process
one returns to the problems at issue again and again, and re-evaluates the earlier
decisions about the acceptance or rejection of statements, the reliability of the
sources, the plausibility values of the statements, the workability of methodological
norms, the conclusions previously reached by inferences etc. In sum: one retrospec-
tively re-evaluates the information at issue (cf. Rescher 1976, 1987). Retrospective
re-evaluation is cyclic in nature. However, it is not only cyclic, it is also prismatic.
This means that the cycles continuously change the perspective from which the
pieces of information constituting the p-context are evaluated (cf. Rescher 1987).
The core of an argumentation cycle involves the following phases:

a. the extension of the p-context by new sources, new statements, new methods
etc.;
b. the coordination of the extended p-context (for example, comparing the plau-
sibility values of the statements stemming from the old and the new sources,
collecting and summarising information about the reliability of the sources,
checking the consistency of the set of the plausible statements etc.);
 The p-model of data and evidence in linguistics 

c. the modification of the extended and coordinated p-context. This means that
one has to elaborate the p-context that can be regarded as a re-evaluated ver-
sion of the starting p-context. For instance, one has to determine the plau-
sibility value of the statements in the p-context on the basis of the totality
of the sources in the p-context, one has to decide which sources have to be
rejected as unreliable or accepted as new, which methodological norms have
to be abandoned, modified or accepted, etc.; and
d. the comparison of the rival solutions.

An argumentation cycle is a phase of the plausible argumentation process which

develops a new solution to the p-problems included in the starting p-context and
examines whether with this solution their resolution has been achieved. It may
have subcycles which revise certain decisions made within the given argumenta-
tion cycle, and continue the argumentation process by turning back to an earlier
stage of the argumentation cycle. It consists of three main stages.
First, one sets up a new p-context version. One presents hypotheses,
sources, and methodological norms which one accepts and which are regarded
as the starting points of the next stage (cycle) of the argumentation process.
The new p ­ -context version may try, for example, to examine whether a solu-
tion can be achieved if one keeps only one member of the p-inconsistent
hypotheses and gives up the other one, or it may attempt to retain both rival
hypotheses and find out how they can be separated from each other in a well-
motivated way.
Second, the new p-context version is elaborated. This involves the extension
of the p-context by new pieces of information. To this end, one makes use of the
new direct and indirect sources (plausible inferences), methods, criteria, etc. in the
p-context. Then, the extended p-context is coordinated: the old and new informa-
tion related to the plausibility values of statements, the reliability of the sources,
the applicability of methods, or the fulfilment of the criteria are summarised. As a
next step, the coordinated p-context is modified. This means that one makes deci-
sions on the basis of the coordination of the old and new pieces of information in
order to elaborate a revised, re-evaluated version of the starting p-context which is
a solution of the starting p-problem.
Third, one compares the elaborated p-context version to its rivals obtained
in the earlier cycles. The question is whether it is a solution of the starting
p-problem(s). If not, then the argumentation process turns back and a new subcy-
cle is started in order to elaborate on the given p-context. If a solution has already
been arrived at, then it has to be examined to see whether it is a better solution
according to the accepted criteria than the ones elaborated earlier, and whether
further rival solutions are conceivable. The argumentation process can only ter-
minate if one is in possession of the best solution available within the practical
 András Kertész & Csilla Rákosi

limits given – that is, the resolution of the starting p-problem; otherwise, a new
argumentation cycle has to be initiated.
Figure 1 illustrates the main stages of the argumentation process.

Starting (p-problematic) p-context

Setting up a new p-context (version)

Extension of the p-context

Argumentation cycle
Elaboration of the
p-context

Subcycle
Coordination of the p-context

Modification of the p-context

Comparison of the elaborated

p-context to the rival solutions

Plausible argumentation process

(cyclic and prismatic re-evaluation of the starting p-context)

Final p-context (resolution of the starting p-problem)

Figure 1.  The process of plausible argumentation

Example 14. Kertész & Rákosi (this volume) and Cs. Rákosi (this volume) recon-
struct the argumentation cycles of the given author. Bibok (this volume), Németh T.
(this volume), Nagy C. (this volume), Koczogh (this volume), Gy. Rákosi (this
volume) and Vecsey (this volume) present and analyse rival solutions to a given
p-problem. Since in the first cycle it can only be shown that they are all problem-
atic and do not provide the resolution of the p-problem investigated, the authors
start a new argumentation cycle in order to find a resolution with the help of the
extension of the p-context with new sources.
 The p-model of data and evidence in linguistics 

2.6.2  Problem-solving strategies

As already mentioned in the preceding section, the decision as to whether the argu-
mentation process may terminate and the resolution of the starting p ­ -problem is
achieved is not absolute and not incontrovertible. The reason for this is, first, that
because of practical limits, the cyclic re-evaluation cannot be complete, cannot take
into consideration every piece of information and cannot examine every possibil-
ity but has to remain partial. The second reason is that in most cases, attempts at
the solution of the initial problems lead to the emergence of new problems which
should be solved – and so on ad infinitum. Third, the rival solutions obtained as
results of the argumentation cycles conducted are partial, too, since they have been
elaborated with the help of diverse heuristics. Moreover, the comparison of the
rival solutions cannot be reduced to the mechanical comparison of the plausibility
of their hypotheses but has to rely on a series of criteria. One must examine which
solution is, as a whole, the least p-problematic p-context, which solution contains
the highest number of hypotheses with a high plausibility value, which solution is
the most comprehensive etc. However, this is usually a difficult and complex task
because there may be conflicts among the evaluations obtained: one solution can
be optimal in respect to one criterion but not as satisfactory with others.
Therefore, it is of vital importance to find problem-solving strategies which
may lead to effective and reliable decisions. Such heuristics make it possible to
elaborate and compare a fair amount of p-context versions and arrive at a well-
founded resolution of the starting p-problem even though the fulfilment of these
tasks can be only partial.
An important subgroup of heuristics consists of strategies of the treatment of
p-inconsistency. Basically, one can follow three strategies:

a. The Contrastive Strategy. The essence of this strategy is that it treats the
p-­context versions containing the contradictory statements as rival alterna-
tives, compares them and strives for a decision between them.
b. The Exclusive Strategy. This strategy is the continuation of the Contrastive
Strategy in cases in which a decision has been reached between the rival
p-context versions elaborated. It fulfils a kind of control function insofar as it
examines whether the p-context version chosen can provide an explanation
for all phenomena that could be explained by the rejected p-context version.
This is important because the explanatory power of the resolution should be
as high as possible; therefore, information loss resulting from the rejection of
p-context versions should be avoided or at least minimised.
c. The Combinative Strategy. It may be the case that one wants to keep both rival
p-context versions because they illuminate some phenomenon from different
points of view which are equally important and cannot be given up. There-
fore, the two p-context versions are no longer deemed to be rivals but are
 András Kertész & Csilla Rákosi

rather co-existing alternatives which have to be maintained simultaneously.

The task is to elaborate both p-context versions, make them as comprehensive
and as free from p-problems as possible. Nevertheless, the separation of the
two p-context versions has to be systematic and well-motivated.

Example 15. Nagy C. (this volume) first applies the Contrastive Strategy and strives
for a comparison of the plausibility of the hypotheses and latent background
assumptions and via this, of the tenability of the rival solutions investigated. As
a result, she comes to the decision that one of the three solutions is considerably
better than the other two. At this point, she changes to the Exclusive Strategy and
examines whether the solution of Juge (2006) can solve all problems which have
arisen within the two rejected solutions.

Example 16. Bibok (this volume) adopts first the Contrastive Strategy, too. But after
the comparison of the rival solutions to problems related to the Hungarian loca-
tive alternation verbs presented by the lexical and the constructional approaches,
he changes to the Combinative Strategy. The reason for this decision lies in the
need to keep both perspectives, because lexical and constructional factors seem
to play a role in syntactic alternations. Therefore, neither of the two proposals can
be given up because relevant phenomena would remain without an explanation.
Thus, by the exact specification of the role of the lexical and the constructional
factors in the model, Bibok proposes a lexical-constructional explanation that is
capable of unifying the explanatory power of the two approaches without leading
to proliferation of p-inconsistencies in his approach.

The p-model allows us to describe the different strategies for the treatment of con-
tradictions, considering the latter to be an integral part of linguistic theorising, that
is, of a cyclic and prismatic argumentation process. These strategies make it possi-
ble to deem inconsistency not a fatal failure. They do not compel us to treat all con-
tradictions in a unique way but license several possible methods for their handling:

–– With the help of cyclic argumentation based on plausible inferences, one may
try to resolve contradictions by giving up one of their members in that the
Contrastive and the Exclusive Strategy are made use of.
–– One may tolerate a contradiction with the help of their paraconsistent repre-
sentation based on the application of the Combinative Strategy. In this case,
they may be regarded not as temporary phenomena but as constitutive ele-
ments of the given theory.
–– In the absence of an acceptable solution, when both kinds of strategies seem
to fail, the inconsistency may be temporarily tolerated until one is in posses-
sion of new information on the basis of which new argumentation cycles can
be initiated.
 The p-model of data and evidence in linguistics 

Consequently, the process of the emergence (and the resolution) of inconsistencies

is a constitutive characteristic of linguistic theorising. One of the most substantial
results of the p-model is that it systematically accounts for this characteristic.

2.7  The solution to (P)(a)

On the basis of the above considerations we may summarise our hypothetical
solution to (P)(a) as follows:
(SP) (a) The p-model provides a novel and comprehensive model of linguistic
theorising.

Naturally, the next step is to show how the p-model defines the notions of linguis-
tic data and evidence.

3.  A
 possible solution to (P)(b): The p-model’s concepts of ‘data’
and ‘evidence’

3.1  Data
The p-model suggests that data are not identical to “examples” such as the follow-
ing Hungarian sentences:
Az anya vizet önt a virágra.
*Az anya vízzel önti a virágot.
Rather, their structure consists of two components: a statement with an informa-
tion content and a plausibility value supported by a direct source (see also Rescher
1979: 69):
(D1) A datum is a statement with a positive plausibility value originating from
some direct source.

The above “examples” can be transformed into data as follows:

0 < |The sentence Az anya vizet önt a virágra is grammatical in Hungarian
while the sentence *Az anya vízzel önti a virágot is not.|S < 1

where S is the linguistic intuition of the linguist as a direct source. A second pos-
sibility is the transformation to a datum with a more complex structure whose
plausibility value originates from a compound of the linguist’s linguistic intuition
and linguistic knowledge:
0 < |The sentence Az anya vizet önt a virágra is grammatical in Hungarian
while the sentence *Az anya vízzel önti a virágot is not, and the meanings
and grammatical properties of the words and parts of words in these two
sentences can be characterised as follows:
 András Kertész & Csilla Rákosi

Az anya-∅ viz-et önt-∅ a virág-ra.

the mother-nom water-acc pour-prs.indf.3sg the flower-sub
‘The mother is pouring water onto the flower.’
*Az anya-∅ víz-zel önt-i a virág-ot.
   the mother-nom water-ins pour-prs.def.3sg the flower-acc
‘The mother is pouring the flower with water.’|S < 1

Accordingly, data are in most cases not claimed to be true with certainty, but they
are usually more or less plausible “truth candidates”. Their plausibility is usually
supported by the sources to some extent but the sources are not able to make them
certainly true (see also Rescher 1976: 8, 1977: 213, 1987: 307). Nevertheless, a datum
must possess a certain degree of initial plausibility, that is, it has to receive a plausi-
bility value from some reliable source. Statements which are of neutral plausibility
or implausible according to any sources in the p-context, are not data in this sense.

Example 17. Grammaticality/acceptability judgements are data since they are

plausible statements. Their source is the linguistic intuition of some native
speaker(s) and their information content consists of the evaluation of the correct-
ness (grammaticality or acceptability) of sentences. Accordingly, the statement
The sentence Hajnalodván, elindultunk hazafelé is correct in Hungarian,
while the sentence Hajnalodva, elindultunk hazafelé is not.6

is a datum in Németh T. (this volume: Section 2.4) because it receives its initial
plausibility value on the basis of the linguistic intuition of the author of Tóth
(2001). This plausibility value is not maximal, because linguistic intuition cannot
be regarded as a totally reliable source. As Németh T. (this volume) remarks, the
evaluation of the correctness of these sentences is not unanimous, because another
native speaker, for example, Sárik (1998) judges the acceptability of -ván-­participle
forms differently. That is, on the basis of this source, the following statement counts
as a datum:
Neither the sentence Hajnalodván, elindultunk hazafelé, nor the sentence
Hajnalodva, elindultunk hazafelé is grammatically correct in Hungarian.

.  For a better understanding, we repeat the author’s glosses here. It is important to remark,
however, that they do not belong to the datum at issue (although it is possible to construct a
datum that involves this information, too):

Hajnal-od-ván/*va, pro el-indul-t-unk haza-felé.

dawn-sfx-VÁN/-VA pvb-start-sfx-pst-indf.1pl home-towards
‘Dawn coming, we started for home.’
 The p-model of data and evidence in linguistics 

From this it follows that the p-context is p-inconsistent because it contains two
incompatible data. Despite this, the initial plausibility value of the first datum can
be increased by collecting judgements from other native speakers, too.
The statement
The va-participle *hajnalodva formed from the weather verb hajnalodik
‘to dawn’ is ill-formed, but the ván-participle form of it hajnalodván is
well-formed in Hungarian.
looks similar to the above datum but it is not a datum according to the terminol-
ogy of the p-model in the argumentation conducted by Németh T. (this volume).
The reason for this is that its plausibility value does not originate from a direct
source (a linguistic paper or the linguistic intuition of native speakers) but from
an indirect one (that is, from a plausible inference):
0 < |The sentence Hajnalodván, elindultunk hazafelé is correct, while the
sentence Hajnalodva, elindultunk hazafelé is not.|S < 1
0 < |If the sentence Hajnalodván, elindultunk hazafelé is correct, then the
ván-participle form of the verb hajnalodik ‘dawn’ is well-formed.|S < 1
0 < |If the sentence Hajnalodva, elindultunk hazafelé is not correct, then the
va-participle *hajnalodva formed from the weather verb hajnalodik ‘dawn’ is
ill-formed.|S < 1
0 < |The va-participle *hajnalodva formed from the weather verb hajnalodik
‘to dawn’ is ill-formed, but the ván-participle form of it, hajnalodván is well-
formed.|I < 1
If, however, this statement obtained its plausibility value from a linguistic paper
simply referred to by Németh T., then it would count as a datum.

Example 18. Corpora are often used and reliable – but not completely reliable –
sources, too, which make it possible to assign a plausibility value to statements
such as
The phrase esteledik a nap is grammatical in Hungarian.
The phrase sötétedik a látóhatár is grammatical in Hungarian.
The phrase hajnalodik az idő is grammatical in Hungarian.
The phrase zuhog az eső is grammatical in Hungarian.7
It may be the case that there are native speakers who judge these phrases to be
ungrammatical. Therefore, the presence of the phrases at issue in the Hungarian
National Corpus cannot guarantee that all speakers of Hungarian, or at least the
majority of them, accept them.

.  For the analysis of these data, see Németh T. (this volume).
 András Kertész & Csilla Rákosi

Example 19. Data used in linguistic theories often have a complex structure in the
sense that the source on the basis of which their plausibility can be judged has to
be interpreted as a compound of several sub-sources. For example, the plausibil-
ity value of the following statements is not just based on the linguist’s linguistic
intuition, native speakers’ judgements or corpus presence, but also on linguistic
analyses with the help of a semantic/syntactic/pragmatic theory:

a. In the sentence Mas con foren prop d’Agda, noves los van venir con havia pres,
lo dia passat, a aquells from Muntaner’s Crònica, van venir ‘arrived’ is a gram-
maticalised form, conveying a past tense meaning.
b. In Axininca Campa, to the input /no-N-čhik-i/ belongs the output-form
[noɲčhiki].
c. Villámlik az ég ‘the sky is lightning’ and sötétedik az ég ‘the sky is getting dark’
are well-formed expressions in Hungarian in which villámlik as well as sötéte-
dik are weather verbs and ég is an explicit subject.
d. In the Amharic sentence ȷ̌on ȷ̌əgna nə-ññ y yl-all, the attitude operator modi-
fies the content of the character of the pronoun.

The structure of the datum in (a), for example, is “The unit S can be identified
in corpus C and it had the characteristics X.” From this it is clear that the source
of its plausibility value is not the corpus (Muntaner’s Crònica) alone. Rather, it
is a compound of the historical document itself, the linguist’s linguistic intuition
and his/her skills in the application of linguistic analysis. The reliability of each
of these (sub-)sources influences the plausibility of the datum. Therefore, it is not
only factors which are decisive for the originality of the historical document that
are relevant; difficulties or uncertainties in the identification and interpretation of
the given linguistic phenomenon must be taken into account as well.

From what we have said about plausible statements in Section 2.2, the following
characterisation of data can be highlighted:

–– Data are “given” in a specific sense: their initial plausibility is not determined
with the help of plausible inferences conducted within the given argumenta-
tion process, but directly on the basis of the reliability of their source. Thus,
they function as starting points: plausibility values enter the argumentation
process through them – and, of course, they can spread towards other hypoth-
eses of the theory by plausible inferences.
–– Since data are defined as statements possessing a positive plausibility value, and
the plausibility value of statements depends on the p-context (see S­ ection 2.4),
data are p-context-dependent, which means that they are theory-­dependent
as well.
 The p-model of data and evidence in linguistics 

3.2  Evidence
The p-model defines three types of evidence in order to grasp the relationship
between data and other hypotheses of the theory.
Weak evidence for a hypothesis h simply means that we can build inference(s)
on the given datum that make(s) h plausible (in the extreme case true with cer-
tainty). Weak evidence against a hypothesis h means a datum on which we can
build inference(s) that make(s) h implausible (in the extreme case false with
certainty):
(D2) (a) A datum e is weak evidence for hypothesis h, if the p-context contains
statements that extend e into an indirect source on the basis of which a
positive plausibility value can be assigned to h.
(b) A datum e is weak evidence against hypothesis h, if the p-context
contains statements that extend e into an indirect source on the basis of
which a positive plausibility value can be assigned to ~h.

From this definition it follows that a datum can be weak evidence for a statement
and its rival simultaneously, although it may support them to different extents.

Example 21. As Bibok (this volume) shows, the lexical rule approach proposes the
following hypothesis for the description of the locative alternation of verbs:
(H) If there is a verb with the semantic representation ‘X causes Y to move
into/onto Z’, then it can be converted into a verb with the semantic
representation ‘X causes Z to change state by means of moving Y into/
onto it’.
(D1) is weak evidence for (H):
(D1) The sentence Az anya zsírral keni a kenyeret is grammatical in Hungarian.

First, (D1) is a plausible statement and its plausibility value originates from a direct
source (Bibok’s linguistic intuition). Second, we can construct a chain of plausible
inferences that make (H) plausible and one of them uses (D1) as a premise:
0 < |If (D1), then there is a verb ken with the semantic representation
‘X causes Z to change state by means of moving Y into/onto it’.|B < 1
0 < |(D1)|B < 1
0 < |There is a verb ken with the semantic representation ‘X causes Z to
change state by means of moving Y into/onto it’.|I < 1
1

|If (H), and there is a verb ken with the semantic representation ‘X causes
Y to move into/onto Z’, then there is also a verb ken with the semantic
­representation ‘X causes Z to change state by means of moving Y into/
onto it’.|B = 1
 András Kertész & Csilla Rákosi

0 < |There is a verb ken with the semantic representation ‘X causes Z to

change state by means of moving Y into/onto it’.|I < 1
1
0 < |There is a verb ken with the semantic representation ‘X causes Y to
move into/onto Z’.|B < 1
0 < |(H)|I < 1
2

while (D2) is weak evidence against (H):

(D2) The sentence Az anya vízzel önti a virágot is ungrammatical in Hungarian.

because the following chain of plausible inferences presents itself:

0 < |If (D2), then there is no verb önt with the semantic representation
‘X causes Z to change state by means of moving Y into/onto it’.|B < 1
0 < |(D2)|B < 1
0 < |There is no verb önt with the semantic representation ‘X causes Z to
change state by means of moving Y into/onto it’.|I < 1
3

|If (H), and there is a verb önt with the semantic representation ‘X causes
Y to move into/onto Z’, then there is also a verb önt with the semantic
­representation ‘X causes Z to change state by means of moving Y into/
onto it’.|B = 1
0 < |There is a verb önt with the semantic representation ‘X causes Y to
move into/onto Z’.|B < 1
0 < |There is no verb önt with the semantic representation ‘X causes Z to
change state by means of moving Y into/onto it’.|I < 1
3
0 < |(~H)|I < 1
4

Relative evidence for a hypothesis h also requires that the datum provides stronger
support to h than to its rivals:
(D3) (a) A datum e is relative evidence for hypothesis h, if
(i) e is weak evidence for hypothesis h;
(ii) the inference(s) connecting the premises and h provide(s) h with
a higher plausibility value than the plausibility values of h’s rivals
assigned to them by the inferences also using e as a premise.
(b) A datum e is relative evidence against hypothesis h, if
(i) e is weak evidence against hypothesis h;
(ii) the plausible inference(s) connecting the premises and ~h
provide(s) ~h with a higher plausibility value than the plausibility
value of h assigned to it by the inferences also using e as
a premise.
 The p-model of data and evidence in linguistics 

Example 22. The datum

(D) The sentence Az apa kávéval löttyenti az asztalterítőt is ungrammatical in
Hungarian.

is relative evidence for the hypothesis of the lexical-constructional approach

according to which
(H) The locative alternation is relevant for only those verbs whose meaning
representations are underspecified (in a relevant way).

The reason for this is that (D) can be treated with the help of (H)’s rivals – as Bibok
(this volume) shows – only as an exception. From this it follows that in this case,
the plausibility of the rival hypotheses can only be relatively low, because they can
receive only very weak support from inferences making use of (D). In contrast, the
plausibility of (H) can be high on the basis of plausible inferences involving (D)
as a premise.

The third type is strong evidence which means that the datum makes only hypoth-
esis h plausible and does not provide any support to its rivals:
(D4) (a) A datum e is strong evidence for hypothesis h, if
(i) e is weak evidence for hypothesis h;
(ii) e is not weak evidence for any of h’s rivals.
(b) A datum e is strong evidence against hypothesis h, if
(i) e is weak evidence against hypothesis h;
(ii) e is not weak evidence against any of h’s rivals.

Example 23. The datum referred to in Németh T. (this volume)

(D) Sötétedik az ég ‘the sky is getting dark’ is a well-formed expression.

is strong evidence against the hypothesis that

(H1) Verbs of natural phenomena in Hungarian cannot occur with explicit
subjects, and they do not allow even expletive, formal subjects.

The following plausible inference making use of (D) as a premise indicates that
(H1) has to be regarded as implausible:
|If (H1), and the verb sötétedik is a verb of natural phenomena and ég is an
explicit subject, then (~D).|N = 1
|The verb sötétedik is a verb of natural phenomena and ég is an explicit
subject.|N = 1
0 < |(D)|N < 1
0 < |(~H1)|I < 1
1
 András Kertész & Csilla Rákosi

(D) provides also weak evidence for all of (H1)’s rivals mentioned in Németh T.
(this volume), that is, for (H2) and (H3):
(H2) The verb sötétedik belongs to the group of verbs of natural phenomena that
have a subject argument position which can be filled optionally.
|If (H2), then (D).|N = 1
0 < |(D)|N < 1
0 < |(H2)|I < 1
2

(H3) Verbs of natural phenomena take a quasi-argumental subject.

|If (H3) and the verb sötétedik is a verb of natural phenomena,
then (D).|N = 1
|The verb sötétedik is a verb of natural phenomena.|N = 1
0 < |(D)|N < 1
0 < |(H3)|I < 1
3

In contrast, the datum

(D’) Villámlik az ég ‘the sky is getting dark’ is a well-formed expression.

does not provide strong evidence against (H1), because it is also weak evidence
against one of its rivals, namely (H2’), which is regarded as plausible by Keszler
(2000):
(H2’) The verb villámlik belongs to the group of verbs of natural phenomena which
do not have a subject argument position, while there are verbs of ­natural
phenomena which have a subject position that can be filled ­optionally.

It is easy to see that the concepts of weak, relative and strong evidence do not meet
the requirements laid down by the standard view of linguistic data and evidence.
First, evidence is i­nterpreted by the p-model not as a special subset of data but as
a datum with a special function relative to some hypothesis of the theory. From this
it follows that evidence is not objective, immediately given, theory-independent
and completely reliable but source- and theory-dependent and reliable only to a
certain extent. Second, data which meet the criteria laid down in (D2)–(D4) in
most cases do not perfectly support or refute the given hypothesis. The connec-
tion between the datum and the hypothesis is established by plausible inferences
relying on plausible premises. Third, the function of evidence is not restricted by
the p-model to the testing of hypotheses, that is, to the justification of theories, but
data and evidence play a role in every stage of the process of linguistic theorising.
 The p-model of data and evidence in linguistics 

4.  Conclusions

The solution to (P)(b) which we have obtained with the help of the p-model pro-
vides a radically new interpretation of linguistic theorising in general and of linguis-
tic data/evidence in particular. Its essence can be summarised as follows:
(SP)(b) Linguistic theorising is a retrospective, cyclic and prismatic plausible
argumentation process. The structure and the functions of linguistic data
and evidence emerge from, and are bound to, this argumentation process:
(i) The structure of linguistic data/evidence consists of a statement
(information content) and a plausibility value (relative strength of
acceptability) stemming from a direct source (linguistic intuition,
corpus, conjecture, experiment etc.). A datum is evidence for a
hypothesis if it is a premise of a plausible inference which makes this
hypothesis plausible.
(ii) The primary function of data is to supply the theory with plausibility
values. That is, data receive their initial plausibility value directly from
reliable sources. All other hypotheses obtain their plausibility value
indirectly (with the help of plausible inferences) from the data. The
function of evidence is that it contributes to the evaluation and the
comparison of the plausibility of rival hypotheses; thus, it is conducive
to the resolution of p-problems.

As we have seen, the peculiarities of the argumentation process can account for,
among others, the uncertainty, the variety and the combinability of linguistic data/
evidence as well as for the emergence and (provisional) existence of inconsisten-
cies triggered by data – that is, for those characteristics of data that could not be
explained with the help of other metatheoretical frameworks.
On the way to (SP)(b), our most important insights were as follows:

a. Data, evidence and the hypotheses supported by the latter are substantially
rooted in uncertainty.
b. Instead of linear progress, linguistic theory formation is built on the cyclic,
prismatic and retrospective re-evaluation of previously assumed data and
hypotheses.
c. There are several kinds of inconsistency, and not every kind is disastrous.
Among certain circumstances inconsistency may be tolerated, while the pro-
cess of plausible argumentation also governs the resolution of inconsistencies
as well as their temporary or long-term tolerance.
d. The pluralism of linguistic theories is fruitful because it allows linguists to
examine linguistic phenomena from different perspectives and must not give
way to the absolutistic defence of particular theories.
 András Kertész & Csilla Rákosi

e. There are no final solutions to linguistic problems; any solution can be pris-
matically and cyclically re-evaluated during later argumentation cycles.8
f. The argumentation structure of linguistic theories largely determines the effi-
ciency of linguistic research.
g. There is a complex relationship between the argumentation structure of lin-
guistic theories and linguistic data/evidence.
h. Data are not objective facts but plausible statements whose plausibility may
increase or decrease (or even disappear) dynamically during the argumenta-
tion process.
i. The source of linguistic data is not the objective observation of manifesta-
tions of linguistic behaviour. Data usually obtain their plausibility value from
a complex source which is a combination of the linguist’s linguistic intuition,
his/her analytic skills, and some other source such as corpora, psycholinguis-
tic experiments, judgements of native speakers, etc.

(SP)(b) does not provide final, all-embracing and infallible answers to questions
concerning the nature of linguistic data and evidence. Rather, it should be regarded
as a starting point which might open a new stage in the debate on linguistic data
and evidence by putting forward a novel and systematic metatheoretical model of
linguistic theorising.

Acknowledgements

Work on this chapter was supported by the MTA-DE Research Group for
Theoretical Linguistics, the project OTKA K 77823 as well as the TÁMOP-
4.2.2/B-10/1-2010-0024 project, which is co-financed by the European Union and
the European Social Fund. Csilla Rákosi’s work was also supported by the János
Bolyai Research Scholarship of the Hungarian Academy of Sciences.

References

Anand, Pranav & Nevins, Andrew. 2004. Shifty operators in changing contexts. In Proceedings
of the 14th Conference on Semantics and Linguistic Theory, Kazuha Watanabe & Robert B.
Young (eds), 20–37. Ithaca NY: CLC Publications.

.  A problem may, of course, have a clear-cut final solution under well-defined conditions,
within a given theoretical framework. Despite this, we must not forget that theoretical frame-
works may be revised, too, and give way to new, better (elaborated) ones.
 The p-model of data and evidence in linguistics 

Baltin, Mark R. 1987. Degree Complements. In Discontinuous Constituency [Syntax and seman-
tics 20], Geoffrey J. Huck & Almerindo E. Ojeda, (eds), 11–26. Orlando FL: Academic
Press.
Bibok, Károly. This volume. The plausibility of approaches to syntactic alternation of Hungarian
verbs.
Colon, Germà. 1978a. [1959]. El perfet perifràstic català “va + infinitiu”. (The Catalan periphras-
tic perfect “va + infinitive”). In La llengua catalana en els seus textos (The Catalan language
of texts),Vol.2, Germà Colon (ed.), 119–130. Barcelona: Curial.
Colon, Germà. 1978b[1976]. Sobre el perfet perifràstic ‘vado + infinitiu’, en català, en proven-
çal i en francès (On the periphrastic perfect ‘vado + infinitive’ in Catalan, Provençal and,
French). In La llengua catalana en els seus textos (The Catalan language of texts),Vol. 2,
Germà Colon (ed.), 131–174. Barcelona: Curial.
Detges, Ulrich. 2004. How cognitive is grammaticalization? The history of the Catalan perfet
perifràstic. In Up and Down the Cline – The Nature of Grammaticalization [Typological
Studies in Language 59], Olga Fischer, Muriel Norde & Harry Perridon (eds), 211–227.
Amsterdam: John Benjamins.
Gougenheim, Georges. 1971. [1929]. Étude sur les périphrases verbales de la langue française.
Paris: Nizet.
Juge, Matthew L. 2006. Morphological factors in the grammaticalization of the Catalan “go”
past. Diachronica 23(2): 313–339.
Kager, René. 1999. Optimality Theory. Cambrige: CUP.
Kaplan, David. 1989. Demonstratives. In Themes from Kaplan, Joseph Almog, John Perry &
Howard Wettstein (eds), 481–563. Oxford: OUP.
Kertész, András. 1993. Heuristik der deutschen Phonologie. Eine elementare Einführung in Strat-
egien der Problemlösung. Budapest: Akadémiai Kiadó.
Kertész, András. 2004a. Philosophie der Linguistik. Studien zur naturalisierten Wissenschafts-
theorie. Tübingen: Narr.
Kertész, András. 2004b. Cognitive Semantics and Scientific Knowledge. Case studies in the Cogni-
tive Science of Science [Converging Evidence in Languag and Communication Research 4].
Amsterdam: John Benjamins.
Kertész, András & Rákosi, Csilla. 2006. Inconsistency and plausible reasoning in an analysis
of German affricates. A case study in the philosophy of linguistics. Language Sciences 28:
386–423.
Kertész, András & Rákosi, Csilla. 2009. Cyclic vs. circular argumentation in the conceptual
­metaphor theory. Cognitive Linguistics 20(4): 703–732.
Kertész, András & Rákosi, Csilla. 2012. Data and Evidence in Linguistics. A Plausible Argumenta-
tion Model. Cambridge: CUP.
Kertész, András & Rákosi, Csilla. 2013. Paraconsistency and plausible argumentation in genera-
tive grammar: A case study. Journal of Logic, Language and Information 22(2): 195–230.
Kertész, András & Rákosi, Csilla. This volume. Thought experiments and real experiments as
converging data sources in pragmatics.
Keszler, Borbála (ed.). 2000. Magyar grammatika (Hungarian grammar). Budapest: Nemzeti
Tankönyvkiadó.
Komlósy, András. 1994. Complements and adjuncts. In Syntax and Semantics 27, Ferenc
Kiefer & Katalin É. Kiss (eds), 91–178. New York NY: Academic Press.
Nagy C., Katalin. This volume. Methods and argumentation in historical linguistics: A case study.
Németh T., Enikő. This volume. Hungarian verbs of natural phenomena with explicit and
implicit subject arguments: their use and occurrence in the light of data.
 András Kertész & Csilla Rákosi

Pinker, Steven. 1989. Learnability and Cognition: The Acquisition of Argument Structure.
­Cambridge MA: The MIT Press.
Polya, George. 1948. How to Solve It. Princeton NJ: Princeton University Press.
Polya, George. 1954. Patterns of Plausible Inference. London: OUP.
Rákosi, Csilla. This volume. Data and the resolution of inconsistency in Optimality Theory.
Rescher, Nicholas. 1973. The Coherence Theory of Truth. Oxford: Clarendon Press.
Rescher, Nicholas. 1976. Plausible Reasoning. Assen: Van Gorcum.
Rescher, Nicholas. 1977. Methodological Pragmatism. Oxford: Blackwell.
Rescher, Nicholas. 1979. Cognitive Systematization. Oxford: Blackwell.
Rescher, Nicholas. 1987. How serious a fallacy is inconsistency? Argumentation 1: 303–316.
Rescher, Nicholas & Brandom, Robert. 1980. The Logic of Inconsistency. Oxford: Blackwell.
Sárik, Pál. 1998. A határozói igenevek néhány problémája (Some problems of adverbial parti-
ciples). Magyar Nyelv 94: 423–435.
Schlenker, Philippe. 2003. A plea for monsters. Linguistics and Philosophy 26: 29–120.
von Stechow, Arnim. 2003. Feature deletion under semantic binding: Tense, person, and mood
under verbal quantifiers. In NELS 33: Proceedings of the Thirty-Third Annual Meeting of
the North East Linguistic Society, GLSA, Makoto Kadowaki & Shigeto Kawahara (eds),
379–404. Amherst MA: University of Massachusetts.
Tóth, Ildikó. 2001. Impersonal constructions and null expletives. In Argument structure in
­Hungarian, Istvan Kenesei (ed.), 51–78. Budapest: Akadémiai Kiadó.
Vecsey, Zoltán. This volume. A plausibility-based model of shifted indexicals.
part ii

Object-theoretical applications
 chapter 3

The plausibility of approaches to syntactic

alternation of Hungarian verbs

Károly Bibok
University of Szeged, Dept. of Russian Philology, MTA-DE Research Group
for Theoretical Linguistics

Several – lexical and constructional – approaches have been proposed in the

literature to theoretically treat syntactically alternating verb classes. In this
chapter I aim to assess their plausibility on the basis of Kertész and Rákosi’s
(this volume, 2012) notion of plausible argumentation and present a lexical-
constructional account. I argue that both my proposal in general and my analysis
of Hungarian verbs of locative alternation are more plausible than either lexical
or constructional accounts are separately. As for the assessment of the plausibility
of lexical-constructional theory itself, another current publication of mine (Bibok
2010) should be taken into consideration. The lexical-constructional framework
of syntactic alternations is supported by a much wider p-context, namely, by
lexical pragmatics, which duly highlights the unavoidable interaction between
lexicon and contexts.

1.  Introduction

1.1  Background
By syntactic alternation it is generally meant that a verb occurring in a type of syn-
tactic argument structure can be used in another one as well. As to the meanings
of these syntactic structures, they are different, but in a predictable way. Consider
the following Hungarian examples and verbs, which show syntactic alternations of
some kind, also known in several other languages.
(A) Locative alternation, i.e. change of place alternates with change of state. E.g.:
 Károly Bibok

(1) a. Az anya-∅ zsír-t ken-∅ a kenyér-re.

the mother-nom1 fat-acc smear-prs.indf.3sg the bread-sub
‘The mother is smearing fat on the bread.’
b. Az anya-∅ zsír-ral ken-i a kenyer-et.
the mother-nom fat-ins smear-prs.def.3sg the bread-acc
‘The mother is smearing the bread with fat.’

In a broader sense of syntactic alternation, the words occurring with different gov-
ernment patterns are not necessarily of the same form but it is sufficient if they are
connected to each other word-formationally. In (2) one can see various prefixed
verbs derived from ken ‘smear’. They themselves alternate similarly to the base verb.
(2) a. Az anya-∅ rá-ken-i a
the mother-nom onto-smear-prs.def.3sg the
zsír-t a kenyér-re.
fat-acc the bread.sub
‘The mother is smearing fat on the bread.’
b. Az anya-∅ meg-ken-i a kenyer-et zsír-ral.
the mother-nom pfv-smear-prs.def.3sg the bread-acc fat-ins
‘The mother will have smeared the bread with fat.’

The verbs which behave syntactically in the same way as ken ‘smear’ include the
following:
fest ‘paint’, fröcsköl ‘splash’, hint ‘dust; sprinkle’, locsol ‘water; sprinkle’, mázol
(3) 
‘paint’, öntöz ‘water; sprinkle’, permetez ‘spray; sprinkle’, spriccel ‘spray’, tölt
‘fill’, töm ‘cram’, etc.

(B) The verbs of manner of motion can take a directional phrase denoting directed
(proceeding) motion in a particular manner. E.g.:
(4) a. A gyerek-∅ úsz-ik.
the child-nom swim-prs.indf.3sg
‘The child is swimming.’
b. A gyerek-∅ a barlang-ba úsz-ik.
the child-nom the cave-ill swim-prs.indf.3sg
‘The child is swimming into the cave.’

.  The glosses are not intended to capture all morphological properties of Hungarian words
but indicate the necessary ones for the present purposes. The abbreviations used in the glosses
throughout this chapter are the following: 3sg = third person singular, acc = accusative, def
= definite (conjugation), ill = illative, indf = indefinite (conjugation), ins = instrumental,
nom = nominative, pfv = perfective (prefix), prs = present (tense) and sub = sublative.
 The plausibility of approaches to syntactic alternation of Hungarian verbs 

Other verbs which have a syntactic alternation resulting in a directed motion

sense are included in (5).
ballag ‘walk slowly’, biceg ‘hobble’, billeg ‘tip; walk swinging slightly from
(5) 
side to side’, csoszog ‘shuffle one’s feet’, csúszik ‘slide’, dülöngél ‘reel’, folyik
‘flow’, forog ‘spin’, fut ‘run’, gurul ‘roll’, himbálózik ‘swing’, hömpölyög ‘surge’,
imbolyog ‘totter’, kúszik ‘creep’, lebeg ‘float’, mászik ‘climb’, ömlik ‘pour’, pattan
‘bounce’, pattog ‘bounce (several times)’, pörög ‘spin’, repül ‘fly’, ring ‘swing’,
rohan ‘rush’, sétál ‘walk’, sántikál ‘hobble’, siklik ‘glide’, szalad ‘run’, száll ‘fly’,
szökdécsel ‘skip’, szökdel ‘skip’, szökken ‘skip (once)’, támolyog ‘stagger’, táncol
‘dance’, tántorog ‘wobble’, tipeg ‘waddle; toddle’, totyog ‘waddle; toddle’, ugrál
‘jump (several times)’, ugrik ‘jump’, etc.2

We can make plausible statements about the well-formedness of examples in (1),

(2) and (4) as well as about the existence of alternating verb classes in (3) and
(5). They are plausible in the sense that their reliability is “not guaranteed but
only partially supported by the source from which they originate; thus, they can
be regarded as acceptable only to a certain extent” (Kertész & Rákosi 2012: 63;
emphasis as in the original).3 In other words, if a source speaks for a statement,
then it makes it plausible, or, in an extreme case, true with certainty.4 Sources are
called by Kertész and Rákosi (this volume: Section 2.2, cf. 2012: 67) direct if “the
plausibility of the statement at issue is evaluated with respect to the reliability of
the source” and indirect if “the plausibility value of the given statement is deter-
mined with reference to the plausibility of other statements – that is, with the help
of an inference”.
In our case direct sources which support plausible statements about the accept-
ability of the examples and verb classes introduced above could be – besides my
intuition – dictionaries or corpora. It is important to call the reader’s attention to
the point that in Kertész and Rákosi’s (this volume: Section 3.1, 2012: ­Section 13.2)
view linguistic data are not linguistic expressions and structures alone, or mere

.  Several verbs of (5), such as dülöngél ‘reel’, forog ‘spin’, fut ‘run’, himbálózik ‘swing’,
­imbolyog ‘totter’, lebeg ‘float’, pattog ‘bounce (several times)’, pörög ‘spin’, ring ‘swing’, szökdécsel
‘skip’, szökdel ‘skip’, táncol ‘dance’, ugrál ‘jump (several times)’, ugrik ‘jump’, denote a manner
of motion which is not only displacement but also change of position. The expression (egy)
helyben ‘in (a) place’, added to these verbs, may unequivocally indicate change of position.
.  Mutatis mutandis, it also concerns the character of examples marked below with an
­asterisk as not well-formed. Consequently, the acceptability of linguistic expressions is not
conceived of as an issue treatable by a dual “yes–no” system in any case.
.  Throughout the present chapter, (im)plausibility should be understood with the following
proviso: in an extreme case true/false with certainty.
 Károly Bibok

acoustic or visual signs in various types of corpora, but plausible statements which
concern their occurrence and/or characteristics and whose plausibility stems
from direct sources. After all, it is not only a plausible statement about the well-
formedness of linguistic examples but also every hypothesis about their syntac-
tic or semantic structures and about their classification as one linguistic category
which is considered data if the plausibility value is assigned to it immediately on
the basis of the reliability of its source as, e.g., in the case of the existence of alter-
nating verb classes.
It is also worth noticing that if some plausible statements concerning (1)–(5)
became implausible, and even if a statement concerning an entire class of verbs
became implausible on the basis of one type of sources,5 it would not be an implau-
sible conclusion at all that the two above-mentioned syntactically alternating verb
classes exist and, consequently, the linguistic phenomenon of syntactic alternation
exists. Although, in the case of recognition of implausibility the datum concerning
the given class would become less plausible, or one should regard that source as
less reliable (or unreliable in some respects).

1.2  The aim and structure of the chapter

Despite the fact that there is a common agreement about the occurrence of syn-
tactically alternating verb classes, several approaches to the theoretical treatment
of their syntactic alternation have been proposed. In the present chapter I attempt
to re-evaluate the plausibility of the main tenets of these approaches with the help
of Kertész and Rákosi’s (this volume: Subsection 2.6.1, 2012: Section 10.3) notion
of plausible argumentation.
In the remainder of the chapter, following their proposal I will thoroughly
examine the plausibility of the central hypotheses of the various approaches to
syntactic alternation. In connection with them I will ask which is implausible or
exceeds the other as regards plausibility.

2.  Approaches to syntactic alternation

In lexicography there is a tradition that treats occurrences of a word with dif-

ferent syntactic patterns as instances of polysemy if they figure separately in
a dictionary. Although the locative alternation of the verb ken ‘smear’ can be

.  For instance, certain meanings are absent from dictionaries (see below).
 The plausibility of approaches to syntactic alternation of Hungarian verbs 

found as s­ eparate meanings in both the multivolume and the concise explana-
tory dictionaries of Hungarian (Bárczi & Országh 1959–1962; Pusztai 2003), the
directed motion sense of the verb úszik ‘swim’ is not listed in them but one can
infer it from an example: partra úszik ‘swim to the seaside/riverside’. Neverthe-
less, recent trends in theoretical linguistics seem to believe that an enumerative
conception of the lexicon which simply fixes various meanings of a word in a
lexical entry is insufficient. They all attempt to elaborate their own version which
exceeds the traditional lexicographic treatment of polysemy induced by syntac-
tic alternation.

2.1  Lexical rule approach

A lexical rule operates on the semantic representation of a lexical item and in
doing so it creates a new lexical item. The following lexical rule can be proposed
for the locative alternation of the verb ken ‘smear’ (cf. Pinker 1989: 79).

(6) If there is a verb with the semantic representation ‘X causes Y to move into/
onto Z’, then it can be converted into a verb with the semantic r­ epresentation
‘X causes Z to change state by means of moving Y into/onto it’.

One can get (6) as an inductive generalisation on the basis of the syntactic behav-
iour and the semantic structure of the verbs in (3). Since it is an induction, it can-
not be considered a truth. At the same time, (6) seems to be a plausible statement.
It is also a plausible assumption that
ken (X Y-t Z-re) ‘smear (X, Y on Z)’ contains the semantic representation ‘X
(7) 
causes Y to move onto Z’.

(6) and (7) together as premises lead to the following conclusion:

(8) There is an expression ken (X Z-t Y-nal) ‘smear (X, Z with Y)’ with the cor-
responding semantic representation ‘X causes Z to change state by means of
moving Y onto it’.

This conclusion is made by an inference scheme similar to the modus ponens well-
known in logic:
(9) If A, then B.
A
B

However, one has to remember that our premises are only plausible statements
and, therefore, the conclusion (8) can only be a plausible statement as well. Thus,
 Károly Bibok

the entire inference scheme is necessarily regarded as a plausible modus ponens

given in (10).6
(10) It is plausible that if a verb contains in its semantic representation ‘X causes
Y to move into/onto Z’ (= A), then the rule converts it into a verb with the
semantic representation ‘X causes Z to change state by means of moving Y
into/onto it’ (= B).
It is plausible that the verb ken (X Y-t Z-re) ‘smear (X, Y on Z)’ is a verb with
the semantic representation ‘X causes Y to move onto Z’ (= A).
It is plausible that there is a verb ken (X Z-t Y-nal) ‘smear (X, Z with Y)’
with the corresponding semantic representation ‘X causes Z to change state
by means of moving Y onto it’ (= B).

The inference in (10) explains the syntactic alternation of another verb, i.e. ken
‘smear’, not mentioned among the verbs in (3), on the basis of which the lexical
rule in (6) was hypothesised.
In connection with (10) two remarks are in order. First, the rule should not be
fixed in the direction indicated in the first premise of (10) but it should proceed in
the other direction as well: from the with-variant into the locative variant (Pinker
1989: 80). Second, the with-variant in Hungarian can mean that Z is not totally
affected by the activity as in the conclusion of (10) (cf. also (1)(b)) or is totally
affected as in (2)(b).
Plausible inferences such as (10) are put forward and evaluated against a
background which Kertész and Rákosi term the p-context, which includes, among
­others, “a set of sources in terms of which the plausibility value of statements can
be judged” (Kertész & Rákosi, this volume: Section 2.4, see also Kertész & Rákosi
2012: 122). Let us suppose that our initial p-context is extended by a new, reliable
source containing the following example:
(11) Az anya-∅ viz-et önt-∅ a virág-ra.
the mother-nom water-acc pour-prs.indf.3sg the flower-sub
‘The mother is pouring water onto the flower.’

Thus, extending the p-context, we gain a new datum, i.e. a plausible statement
about the occurrence of önt ‘pour’ in a source. Furthermore, it is also a plausible

.  Instead of the verbal indicator for plausibility of statements it is plausible that…, Kertész
and Rákosi (this volume: Section 2.2, Kertész & Rákosi 2012: 69f.) use the following notation
presented in the case of the first premise of (10):

(i) 0 < |If a verb contains in its semantic representation ‘X causes Y to move into/onto
Z’ (= A), then the rule converts it into a verb with the semantic representation
‘X causes Z to change state by means of moving Y into/onto it’ (= B).|S < 1
 The plausibility of approaches to syntactic alternation of Hungarian verbs 

hypothesis on the basis of this source that önt (X Y-t Z-re) ‘pour (X, Y onto Z)’ has
the semantic representation ‘X causes Y to move onto Z’. Let us apply our plausible
modus ponens scheme once again:
(12) It is plausible that if a verb contains in its semantic representation ‘X causes
Y to move into/onto Z’ (= A), then the rule converts it into a verb with the
semantic representation ‘X causes Z to change state by means of moving Y
into/onto it’ (= B).
It is plausible that the verb önt (X Y-t Z-re) ‘pour (X, Y onto Z)’ is a verb
with the semantic representation ‘X causes Y to move onto Z’ (= A).
It is plausible that there is a verb önt (X Z-t Y-nal) ‘pour (X, Z with Y)’ with
the corresponding semantic representation ‘X causes Z to change state by
means of moving Y onto it’ (= B).

At the same time, according to a more extended p-context containing (13) it is also
plausible that there is no such verb.7 Cf.:
(13) *Az anya-∅ víz-zel önt-i a virág-ot.
   the mother-nom water-ins pour-prs.def.3sg the flower-acc
‘The mother is pouring the flower with water.’

Furthermore, if the rule also derives a verb in the other direction, as mentioned
above, then the verb fed ‘cover’ should alternate. However, the contrary seems to
be the case. Consider (14).
(14) a. *Az anya-∅ csokoládémáz-at
   the mother-nom chocolate.coating-acc
fed-∅ a sütemény-re.
cover-prs.indf.3sg the cookie-sub
‘The mother is covering chocolate coating onto the cookie.’
b. Az anya-∅ csokoládémáz-zal
the mother-nom chocolate.coating-ins
fed-i a sütemény-t.
cover-prs.def.3sg the cookie-acc
‘The mother is covering the cookie with chocolate coating.’

Thus, we have an (extended) p-context which qualifies as p-inconsistent because

there is a statement that is made plausible by some source and its negation by

.  How negative evidence can be detected is an issue which deserves further research. It has
already been proposed that corpus-linguistic methods can provide access to negative evidence
and corpus linguists do not have to have recourse to acceptability judgments (Stefanowitsch
2008).
 Károly Bibok

another (Kertész & Rákosi this volume: Section 2.5, 2012: 130).8 This inconsis-
tency calls for the modification of the p-context which “means that one has to
elaborate the p-context that can be regarded as the re-evaluated version of the
starting p-context” (Kertész & Rákosi this volume: Subsection 2.6.1, cf. Kertész &
Rákosi 2012: 138). Since not only a set of sources belongs to a p-context but also
statements together with their relevant characteristics, we can gain a modified
p-context if the plausibility value of the lexical rule in (6) is revised. The new lexi-
cal rule should be formulated with respect to a narrow semantic class to whose
members – and only to those members – the given rule can be applied. In other
words, (6) is the necessary condition of the locative alternation; its sufficient con-
dition, however, seems to be that the verb at stake belongs to a narrow seman-
tic class. Pinker (1989: 126f.) lists six classes of English locative alternation verbs.
They are given in (15), which specially indicates where the locative alternation
cannot be attested in Hungarian.
Smear-class: ‘Simultaneous forceful contact and motion of a mass
(15) a. 
against a surface.’
Pile-class: ‘Vertical arrangement on a horizontal surface.’ In Hungarian
b. 
there is no alternation, unlike English: téglákat halmozott a székre –
*­téglákkal halmozta a széket vs. He heaped bricks on the stool – He
heaped the stool with bricks.
Splash-class: ‘Force is imparted to a mass, causing ballistic motion in a
c. 
specified spatial distribution along a trajectory.’
Scatter-class: ‘Mass is caused to move in a widespread or non-directed
d. 
distribution.’ In Hungarian there is no alternation, unlike English:
­magot szórt a földbe – *maggal szórta a földet vs. He scattered seeds onto
the field – He scattered the field with seeds.
Cram-class: ‘A mass is forced into a container against the limits of its
e. 
capacity.’
Load-class: ‘A mass of a size, shape, or a type defined by the intended
f. 
use of a container is put into the container, enabling it to accomplish its
function.’

Three remarks concerning (15) are in order. First, whereas in Hungarian there is
no alternation like (15)(d) a morphologically derived verb beszór ‘scatter’, contain-
ing the prefix be ‘in’, alternates syntactically in a broad sense, cf.: beszórja maggal a
földet ‘scatter the field with seeds’. Second, Pinker himself (1989: 126) admits that

.  In addition to önt ‘pour’ and fed ‘cover’, there are other verbs that may be characterised
with the semantic representation at issue but do not alternate. See (15)(b) and (15)(d) below.
 The plausibility of approaches to syntactic alternation of Hungarian verbs 

the with-variant of scatter in (15)(d) could be marginal for some English speak-
ers. Goldberg (1995: 239, Footnote 15) belongs to the latter group because, relying
on a certain source (presumably on her intuition) she doubts the existence of the
alternating scatter-class. However, according to The Collins Cobuild English Dic-
tionary for Advanced Learners (32001), the verbs sow and strew occur with both
syntactic patterns under consideration. Third, the full list of English locative alter-
nation verbs also includes two one-member sets: string (the Hungarian equivalent
does not alternate) and wrap (Pinker 1989: 127).
To explain that the verb ken ‘smear’ alternates syntactically as in (1) one needs
now to supplement the first premise of the plausible inference in (10) with the
indication of the semantic structure characteristic of the smear-class ken ‘smear’
belongs to. Furthermore, an account of the non-alternating behaviour of önt ‘pour’
and fed ‘cover’ simply means that their meanings do not correspond to any of the
semantic classes in (15). Rather these Hungarian verbs occur in non-alternating
verb classes similar to those established by Pinker (1989: 126f.) for English verbs,
namely, in the dribble-class (‘a mass is enabled to move via the force of gravity’)
and in the inundate-class (‘a layer completely covers a surface’), respectively. At
the same time, the p-context modified by statements about alternating and non-­
alternating narrow semantic classes seems to be inconsistent with its further
extension.9 Consider (16).
(16) a. Az apa-∅ (véletlenül/szándékosan) kávé-t
the father-nom (accidentally/intentionally) coffee-acc
löttyent-∅ az asztalterítő-re.
spill-prs.indf.3sg the tablecloth-sub
‘The father (accidentally/intentionally) spills coffee on the tablecloth.’
b. *Az apa-∅ (véletlenül/szándékosan) kávé-val
   the father-nom (accidentally/intentionally) coffee-ins
löttyent-i az asztalterítő-t.
spill-prs.def.3sg the tablecloth-acc
‘The father (accidentally/intentionally) spills the tablecloth with coffee.’

As (16) seems to indicate, löttyent ‘spill’ does not alternate. A plausible account
can consist in a statement that the meaning of this verb – like that of önt ‘pour’ – is
fairly similar to the semantic structure of the dribble-class (‘a mass is enabled to
move via the force of gravity’). However, one can realise that löttyent ‘spill’ means

.  Pinker (1989: 395f., Footnote 12) assumes that the verb sprinkle behaves as an alternating
verb on the basis of its phonological similarity to others. As to ditransitivisable verbs, Pinker
(1989: 118ff.) emphasises the role of morphological constraints.
 Károly Bibok

more than motion caused by gravity since a different force causes ballistic motion
of a mass. Thus, löttyent ‘spill’ could enter the alternating splash-class in (15)(c), cf.
‘force is imparted to a mass, causing ballistic motion in a specified spatial distribu-
tion along a trajectory’. It could be an objection that motion does not come about
in a sufficiently specified manner. This objection is eliminated by a (plausibly)
well-formed example containing the verb löttyent ‘spill’ with the prefix le ‘down’,
which does not effect how the mass moves.
(17) Az apa-∅ (véletlenül/szándékosan)
the father-nom (accidentally/intentionally)
le-löttyent-i kávé-val az asztalterítő-t.
down-spill-prs.def.3sg coffee-ins the tablecloth-acc
lit. ‘The father (accidentally/intentionally) spills down the tablecloth
with coffee.’

After all, putting a verb in a narrow semantic class does not correlate with the
alternating or non-alternating syntactic behaviour of this verb alone and with a
prefix not affecting the character of the movement.
To recapitulate where we have arrived with the help of lexical rules applied
to narrow semantic classes of verbs, we should say that even lexical idiosyncrasy
plays some role, in addition to lexical rules made more precise. Such a double char-
acter is acknowledged, at least implicitly, by the theory itself (see narrow seman-
tic classes vs. one-word classes and the factor of phonological similarity) and is
highlighted by our present investigation. It is also clear that in spite of extending
the p-context by new sources and statements, the methodological stance of the
treatment of syntactic alternations has not been changed. Overall, the meaning
brought about by a meaning shift is grasped as a lexical phenomenon but not as a
meaning occurring in a particular syntactic structure. Using a different methodol-
ogy in the starting, or initial, p-context, a constructional conception approaches
meaning shifts and syntactic alternations from the latter point of view.

2.2  Constructional approach

According to Construction Grammar (Goldberg 1995, 2006), constructions are
form−meaning pairs. Distinct constructions are defined as constructions some of
whose properties are unpredictable from knowledge of other constructions. They
exist independently of particular verbs and determine syntactic structures. In each
construction, the meaning of the verb is fused with that of the construction. One
part of its semantic representation plays a crucial role, namely, participant roles.
They are verb-specific roles which have to be distinguished from more general
argument roles, which figure in the semantic part of constructions, such as agent,
patient or goal.
 The plausibility of approaches to syntactic alternation of Hungarian verbs 

Applying the machinery of Construction Grammar to syntactic alternations,

I first cite Goldberg’s (1995: 176f.) own example of the locative alternation. The
lexical-semantic representation of the verb slather contains the following partici-
pant roles:
(18) slather 〈slatherer, thick-mass, target〉

The three participant roles of slather are compatible with the argument roles of
both the caused-motion construction and the causative-plus-with-adjunct con-
struction. The first construction has three argument roles: a cause, a theme and a
goal (directional). The two kinds of role sets can be fused with each other because
the slatherer is semantically construable as a cause, thick-mass as a theme – since
it undergoes a change of location –, and the target as a directional. Cf.:
(19) Sam slathered shaving cream onto his face.

In the case of the second construction, fusing the slatherer with the first argument
is the same as above. The target can be construed not only as a directional, but also
as a patient, in that the entity on which the substance is slathered is affected. Since
the third participant role of slather requires that it be expressed, a with-phrase
emerges even though in the framework of Construction Grammar it is an adjunct
of the corresponding construction. Cf.:
(20) Sam slathered his face with shaving cream.

Thus, one can posit the following hypothesis:

(21) If the participant roles of a verb are compatible with the argument structure
of two constructions, this verb occurs in syntactically alternating structures
(cf. Goldberg 1995: 179).
In order to test the plausibility of (21) let us apply Goldberg’s (1995, 2006) Con-
struction Grammar to the syntactic alternation exemplified by (4), which is
repeated here for convenience as (22).
(22) a. A gyerek-∅ úsz-ik.
the child-nom swim-prs.indf.3sg
‘The child is swimming.’
b. A gyerek-∅ a barlang-ba úsz-ik.
the child-nom the cave-ill swim-prs.indf.3sg
‘The child is swimming into the cave.’
According to Construction Grammar, both uses of the verb úszik ‘swim’, i.e. with
and without a directional phrase, should be analysable in constructional terms. In
the case without a directional phrase, the verb is integrated into a kind of simple
intransitive construction, and in the case with a directional phrase into the intran-
sitive motion construction. The verb úszik ‘swim’ can be associated with the latter
 Károly Bibok

construction and given a directed motion interpretation even if no participant

role is fixed with the verb which corresponds to the directional role. The direc-
tional role belongs to an argument which the intransitive motion construction
itself owns, independently of whether the verb has a participant role correspond-
ing semantically to a directional role.10
What is crucially important in connection with the constructional analy-
sis of the syntactic alternation characteristic of úszik ‘swim’ is that, according to
Construction Grammar, the first argument role of the intransitive motion con-
struction is nothing but a theme, similarly to the second argument role of the
caused-motion construction. Thus, the intransitive motion construction and, con-
sequently, the simple intransitive construction simply account for another use of
úszik ‘swim’, different from that in (22), where an agent subject figures. It is an
occurrence of that verb with a theme subject as in (23), which one can gain from a
new source by extension of the p-context.

(23) a. Az üveg-∅ úsz-ik.

the bottle-nom float-prs.indf.3sg
‘The bottle is floating.’
b. Az üveg-∅ a barlang-ba úsz-ik.
the bottle-nom the cave-ill float-prs.indf.3sg
‘The bottle is floating into the cave.’

Therefore, constructions somewhat different from the simple intransitive and

intransitive motion constructions have to be assumed to handle sentences with
agentive subjects as in (22), provided that the child really does act and he/she is
not interpreted as the inanimate subject of (23). Since the agent is an entity who

.  It should be worth noting that in a wider p-context one cannot avoid the question con-
cerning the motivation of the occurrence of a directional in the intransitive motion construc-
tion. Consider (i).

(i) a. Péter-∅ áll-∅.

Péter-nom stand-prs.indf.3sg
    ‘Péter is standing.’
b. Péter-∅ a szőnyeg-re áll-∅.
Péter-nom the carpet-sub stand-prs.indf.3sg
    ‘Péter steps onto the carpet.’

While in (22)(a) with úszik ‘swim’ there is a motion component which can motivate the
directional role in (22)(b), in (i)(a) one cannot refer to any motion which would make the
directional role plausible in (i)(b).
 The plausibility of approaches to syntactic alternation of Hungarian verbs 

carries out an activity, the intransitive activity and activity-motion constructions

can be assumed for (22).
Nevertheless, this duplication of constructions does not necessarily imply
the lexical proliferation of the Hungarian verb úszik ‘swim; float’. In construc-
tional terms, one can offer that the single participant role of the verb at stake
is construable either as a theme, or as an agent in both the directed motion
sense, i.e. in the intransitive motion and activity-motion constructions with
the directional argument role, and in the manner of motion sense, i.e. in the
simple intransitive and intransitive activity constructions without the direc-
tional argument role. However, in such a case the question of what the dou-
ble construal depends on should inevitably be accounted for in Construction
Grammar.
As to the fusion of a lexical entry with a construction or even with several
constructions, it has to be remarked that there are a number of constraints. Besides
semantic generalisations sensitive to contextual information and world knowledge
as well as narrow semantic classes known from the lexical rule theory, one should
pay special attention to idiosyncratic properties of lexical entries.
After all, after extending the p-context with new sources both the lexical
approach and Construction Grammar face lexical exceptions to general lexical
rules, and mechanisms of integration of lexical entries into constructions, respec-
tively. Although úszik ‘swim; float’ and other similar verbs in (5) show system-
atic polysemy and syntactic alternation, only some of those verbs which indicate
the manner of motion of inanimate entities capable of moving in the presence of
external effects are suitable for designating a directed motion (Komlósy 2000: 257).
Compare, for instance, the verbs pattog ‘bounce’ and inog ‘wobble’ in (24) and
(25), respectively.

(24) a. A labda-∅ a fal-∅ mellett pattog-∅.

the ball-nom the wall-nom by bounce-prs.indf.3sg
‘The ball bounces by the wall.’
b. A labda-∅ a fal-∅ mellé pattog-∅.
the ball-nom the wall-nom to bounce-prs.indf.3sg
‘The ball bounces to the wall.’

(25) a. A szék-∅ a fal-∅ mellett inog-∅.

the chair-nom the wall-nom by wobble-prs.indf.3sg
‘The chair wobbles by the wall.’
b. *A szék-∅ a fal-∅ mellé inog-∅.
   the chair-nom the wall-nom to wobble-prs.indf.3sg
‘The chair wobbles to the wall.’
 Károly Bibok

2.3  Lexical-constructional approach

In the preceding sections we have seen that both lexical and constructional factors
play a role in syntactic alternations.11 Therefore, it seems to be a plausible resolu-
tion of the rivalling lexical and constructional theories to build both of them into
our p-context as methodologies reconciled and supplemented with each other.
This hypothesis underlies an assumption that the advantages of the lexical and
constructional theories together provide a better way of investigating syntactic
alternations than each theory does separately. So, a lexical-constructional explana-
tion could be offered, which will be elaborated in two steps. First, conditions of the
fusion of verbs with constructions are not restricted to the indication of a single
constituent in lexical-semantic representations, namely, to participant roles, but
enriched meaning representations should be introduced for verbs. We should realise
that in doing so possibilities that were originally provided by frame semantics for
Goldberg’s (1995, 2006) Construction Grammar could be exploited. For example,
the enriched meaning structure requires that – besides the meaning representa-
tion without the activity element (cf. (23)(a)) – another meaning representation
be established, which includes the activity element (cf. (22)(a)). However, propos-
ing another meaning would be equivalent to an enumerative type of the lexicon
in which alternating verbs occur twice according to their meanings and we could
arrive at a contradiction with the basic attempts of both lexical and constructional
theories to go beyond the simple multiplication of meanings in the lexicon. Thus,
secondly, and also in accordance with a methodological requirement of lexical
economy (cf. Bierwisch 1997), a general meaning representation of a verb which
is semantically and pragmatically rich enough has to serve as a basis for double
construals of participant roles and, what is more, for both constructional meanings
coming about in syntactic alternations. To put it the other way round, lexically
verbs have underspecified representations with optional elements relevant to one
or another constructional meaning and not representations that are as specific as
constructional meanings.
At the same time, the lexical-constructional approach means not only the
unification of two different methodologies in a wider p-context but also pro-
vides a resolution of inconsistencies and open questions concerning general, or
grammatical, tools and lexical characteristics. From the point of view of syntactic
alternations it is a plausible hypothesis that verbs participate in them if they have
a general meaning which is compatible with all meanings occurring in alterna-
tions. If a verb does not have a lexical representation that can result in different

.  For a substantially more detailed analysis of the lexical and constructional approaches to
syntactic alternations, see Bibok (2010).
 The plausibility of approaches to syntactic alternation of Hungarian verbs 

i­nterpretations, that is, if a verb is lexically more specific, it is implausible that it

alternates syntactically. This will be shown by the re-analysis of locative alternation
in a lexical-constructional framework, which was first offered in Bibok (2008) and
was inspired by Iwata’s (2002, 2005) analysis of locative alternation.12
Recall that in accordance with (7) the verb ken ‘smear’ in a context X Y-t Z-re
‘X, Y on Z’ has a semantic representation ‘X causes Y to move onto Z’, to which
we can add the following specifications: Y = mass, Z = surface, and the causa-
tion includes smoothing movements of the hand. Furthermore, if we concretise
the change of state in meaning representation of the expression ken (X Z-t Y-nal)
‘smear (X, Z with Y)’ (see (8) above) as being covered partially or totally, we can
get the following semantic representation: ‘X causes Z to be covered partially or
totally with Y (= mass)’.13
Now, taking our methodological stance that the constructional meanings of
a verb condense into one lexical underspecified meaning and the two semantic
representations of ken ‘smear’ detailed above, one can come to a plausible enough
conclusion that the verb at stake has the following underspecified semantic
representation:
(26) ‘with smoothing movements of the hand, X causes a mass Y to move onto a
surface Z, and X causes a surface Z to be covered partially or totally with a
mass Y’.

Starting from the lexicon, containing the verb ken ‘smear’ with the underspecified
representation in (26), the constructional meanings of the given verb correspond
to the two possible interpretations of (26) alternating with each other. When a
mass is focused, or profiled, the constructional meaning is equal to the part of
(26) which is before the conjunction and, i.e. ‘with smoothing movements of the
hand, X causes a mass Y to move onto a surface Z’. This constructional meaning is
expressed in (1)(a), repeated here as (27).
(27) Az anya-∅ zsír-t ken-∅ a kenyér-re.
the mother-nom fat-acc smear-prs.indf.3sg the bread-sub
‘The mother is smearing fat on the bread.’

.  My lexical-constructional approach has some common features with Iwata’s (2002, 2005)
but also differs from it in the details of the characteristics of both lexical and constructional
meanings. A comparison between them would deserve a separate paper.
.  An attentive reader who may have noticed that the component of moving is missing here
from (8) is referred to Bibok (2008), where he/she finds a motivation to assume that such a
type of meaning be represented without a motion element.
 Károly Bibok

In the opposite case, when a surface comes into prominence, the constructional
meaning is ‘X causes a surface Z to be covered partially or totally with a mass Y’, i.e.
the part of (26) figuring after the conjunction and (if one leaves aside the manner
specification ‘with smoothing movements of the hand’). It is (1)(b), repeated here
as (28), that has this second interpretation.
(28) Az anya-∅ zsír-ral ken-i a kenyer-et.
the mother-nom fat-ins smear-prs.def.3sg the bread-acc
‘The mother is smearing the bread with fat.’

Consequently, the verb ken ‘smear’ can alternate syntactically because its under-
specified meaning in (26) provides access to two constructional meanings
expressed by the corresponding syntactic structures.14
Contrary, the verbs önt ‘pour’ in (11) and (13), i.e. Az anya vizet önt a virágra
‘The mother is pouring water onto the flower’ and *Az anya vízzel önti a virágot
‘The mother is pouring the flower with water’, as well as fed ‘cover’ in (14), i.e. *Az
anya csokoládémázat fed a süteményre ‘The mother is covering chocolate coating
onto the cookie’ − Az anya csokoládémázzal fedi a süteményt ‘The mother is cover-
ing the cookie with chocolate coating’, do not occur in syntactic structures of two
alternating types because their meaning is not underspecified in a way that would
allow two different interpretations. The verb önt ‘pour’ has a more specific mean-
ing representation which only contains (29) and does not contain (30):
(29) ‘in a given manner, X causes a mass Y to move onto a surface Z/into a
­container Z’;
(30) ‘in a given manner, X causes a surface Z to be covered/a container Z to be
filled (in) partially or totally with a mass Y’.

Also, the meaning representation of fed ‘cover’ is more specific but in the other
way. This verb simply means that
(31) ‘in a given manner, X causes a surface Z to be covered partially or totally
with a mass Y’

and does not mean that

(32) ‘in a given manner, X causes a mass Y to move onto a surface Z’.

Now, we can make the following plausible hypotheses concerning the locative alter-
nation in general. First, in accordance with principles of the lexical-constructional
approach, the locative alternation is relevant only for those verbs whose meaning

.  It is precisely because of the lexically unfixed character of (26) that one may speak of
­underspecification. For other forms of underspecification in meaning representations, see
Bibok (2004).
 The plausibility of approaches to syntactic alternation of Hungarian verbs 

representations are underspecified in a similar way as (26) is. In other words, both
constructional meanings are generally available if all locatively alternating verbs
are characterised with a meaning scheme (template) underspecified with regard
to their parts before or after the conjunction and. Thereby each of them can be
interpreted in two ways and expressed syntactically in two ways. Second, the verbs
which do not own such underspecified representations cannot occur in syntactic
structures alternating the locative variant with the with-variant. This entails that
the verbs not occurring in locative alternation do not have to be considered excep-
tions. The meaning of a verb itself determines – like narrow semantic classes but
more precisely – whether it may participate in the syntactic alternation at stake.
At this point of the present chapter we can explicitly show which data are
regarded as evidence for our lexical-constructional account and against its rivals,
i.e. lexical and constructional theories. Kertész and Rákosi (this volume: Sec-
tion 3.2, 2012: Section 13.3) generally view evidence as a datum whose function is
to contribute to the judgment and comparison of the plausibility of rival hypoth-
eses. They distinguish between three types of evidence: weak, relative and strong
evidence. Recall the examples in (16), given again here as (33).
(33) a. Az apa-∅ (véletlenül/szándékosan) kávé-t
the father-nom (accidentally/intentionally) coffee-acc
löttyent-∅ az asztalterítő-re.
spill-prs.indf.3sg the tablecloth-sub
‘The father (accidentally/intentionally) spills coffee on the tablecloth.’
b. *Az apa-∅ (véletlenül/szándékosan) kávé-val
the father-nom (accidentally/intentionally) coffee-ins
löttyent-i az asztalterítő-t.
spill-prs.def.3sg the tablecloth-acc
‘The father (accidentally/intentionally) spills the tablecloth with coffee.’
The plausible statement about the non-existence of the verb löttyent ‘spill’ with a
with-variant and the corresponding semantic structure, which was made in the
above discussion (cf. also (17)), should be considered weak evidence against the
hypothesis about the existence of a narrow semantic class, namely, the splash-
class in (15)(c) and, naturally, weak evidence for one of the rivals of that hypoth-
esis, i.e. for non-existence of that semantic class in the sense of Kertész & Rákosi
(this volume: Section 3.2, 2012: 178), provided – as said above – that the semantic
class includes verbs − and only those verbs − which undergo the locative alter-
nation. In the terms of Kertész & Rákosi (this volume: Section 3.2, 2012: 181)
such a plausible statement is strong evidence against a hypothesis, in the pres-
ent case: against the hypothesis regarding the existence of (15)(c). Moreover,
it is easy to see that the plausible statement at stake is relative evidence against
the hypothesis about the classification of relevant verbs into n ­ arrow semantic
 Károly Bibok

classes (cf. Kertész & Rákosi this volume: Section 3.2, 2012: 180). This is the case
because the plausible inferences connecting the premises and the negation of
such a hypothesis provide a higher plausibility value than the plausibility value
of the original hypothesis. Namely, the lexical-constructional account without
narrow semantic classes makes a more precise prediction concerning the set
of Hungarian verbs taking part in locative alternation while it does not have to
allow lexical idiosyncrasies which result in exceptions either (see our general
hypotheses concerning the locative alternation after (32)). At the same time,
the plausible statement about the non-existence of the verb löttyent ‘spill’ with
a with-variant and the corresponding semantic structure does not exclude that
the rivals of the lexical-constructional theory give an explanation with the help
of exceptions, i.e. the classification into narrow semantic classes can proceed
with some residue (maybe, in various languages with different lexical residue).
However, such a hypothesis loses against the lexical-constructional approach
from the point of view of plausibility.
Since the lexical-constructional approach rests on an idea which is not widely
accepted in regard to the form of the lexical-semantic representation − namely
that of underspecification −, it is important to gain some evidence to support it.
With this aim in mind, take into consideration the above plausible statements
about (11), (13) and (14), given here as (34)(a)–(b) and (35).

(34) a. Az anya-∅ viz-et önt-∅ a virág-ra.

the mother-nom water-acc pour-prs.indf.3sg the flower-sub
‘The mother is pouring water onto the flower.’
b. *Az anya-∅ víz-zel önt-i a virág-ot.
   the mother-nom water-ins pour-prs.def.3sg the flower-acc
‘The mother is pouring the flower with water.’

(35) a. *Az anya-∅ csokoládémáz-at fed-∅

   the mother-nom chocolate.coating-acc cover-prs.indf.3sg
a sütemény-re.
the cookie-sub
‘The mother is covering chocolate coating onto the cookie.’
b. Az anya-∅ csokoládémáz-zal fed-i
the mother-nom chocolate.coating-ins cover-prs.def.3sg
a sütemény-t.
the cookie-acc
‘The mother is covering the cookie with chocolate coating.’

What is crucial from the point of view of evidence for underspecificity is the double
direction of the non-occurrence of locative alternation (cf. (34)(b) vs. (35)(a)). The
possible lack of either one of both variants in alternation indicates that both scenes
 The plausibility of approaches to syntactic alternation of Hungarian verbs 

underlying the two constructional meanings seem to be necessary for a verb to alter-
nate syntactically. These scenes together can be captured at an underspecified level of
meaning representation (see (26) above). Nevertheless, once again we can only speak
about relative evidence because there are other ways to account for locative alterna-
tion even if they have to allow exceptions but thereby they have less plausibility.

3.  Summary and outlook

After a discussion of the lexical and constructional approaches to syntactic alterna-

tions I have offered their lexical-constructional account and, especially, an analysis of
Hungarian locative alternation verbs in such a framework. I hope that with the help
of Kertész and Rákosi’s (this volume: Subsection 2.6.1, 2012: Section 10.3) notion
of plausible argumentation I could show that both my conception in general and
the analysis carried out are more plausible than the rivals and the analyses they pro-
pose. In respect to the evaluation of the plausibility of lexical-constructional theory,
investigations presented in another recent publication of mine (Bibok 2010) are
also relevant. Here the following three points have to be highlighted. First, applying
the machinery of lexical-constructional theory to the other group of alternating
verbs, namely, to that of manner-of-motion verbs in (4) and (5) above, also gives
more plausible inferences in connection with the double agent–theme construal of
the first participant role of úszik ‘swim; float’ and similar verbs, than a proposal of
two separate constructions: one with an activity component and another without
it. Second, the lexical-constructional treatment of further types of syntactic alter-
nations and non-alternations makes it possible to assume a single meaning scheme
with the help of which one can plausibly generalise the appearance of the directed
motion sense not only in the case of manner-of-motion verbs but also with verbs
of spatial configuration introduced in Footnote 10 and with sound emission verbs.
Finally, the lexical-constructional conception of syntactic alternations remains
plausible in a much wider p-context, namely, in that of lexical pragmatics which
takes the contextualist stance in the debate on the semantics–pragmatics bound-
ary (cf. Jaszczolt 2012) and pays special attention to the necessary interaction of
lexicon and contexts. The lexical-constructional approach to syntactic alternations
naturally extends into lexical pragmatics and no inconsistency emerges.

Acknowledgements

Work on the present chapter was supported by the MTA-DE Research Group for
Theoretical Linguistics and the TÁMOP-4.2.2/B-10/1-2010-0024 project, which is
co-financed by the European Union and the European Social Fund.
 Károly Bibok

References

Bárczi, Géza & Országh, László (eds). 1959–1962. A magyar nyelv értelmező szótára (A Defining
Dictionary of the Hungarian Language) I–VII. Budapest: Akadémiai Kiadó.
Bibok, Károly. 2004. Word meaning and lexical pragmatics. Acta Linguistica Hungarica 51:
265–308.
Bibok, Károly. 2008. Az igék szemantikája és a szintaktikai alternáció (The semantics of verbs
and syntactic alternation). In Strukturális magyar nyelvtan 4: A szótár szerkezete (A Struc-
tural Grammar of Hungarian, vol. 4: The Structure of the Lexicon), Ferenc Kiefer (ed.),
23–70. Budapest: Akadémiai Kiadó.
Bibok, Károly. 2010. From syntactic alternations to lexical pragmatics. In The Role of Data at the
Semantics–Pragmatics Interface, Enikő Németh T. & Károly Bibok (eds), 261–304. Berlin:
Mouton de Gruyter.
Bierwisch, Manfred. 1997. Lexical information from a minimalist point of view. In The Role
of Economy Principles in Linguistic Theory [Studia Grammatica 40], Chris Wilder, Hans-
Martin Gärtner & Manfred Bierwisch (eds), 227–266. Berlin: Akademie Verlag.
The Collins Cobuild English Dictionary for Advanced Learners, 3rd edn. 2001. Glasgow: Harper
Collins.
Goldberg, Adele E. 1995. Constructions: A Construction Grammar Approach to Argument Struc-
ture. Chicago IL: University Press of Chicago.
Goldberg, Adele E. 2006. Constructions at Work: The Nature of Generalization in Language.
Oxford: OUP.
Iwata, Seizi. 2002. Does MANNER count or not? Manner-of-motion verbs revisited. Linguistics
40: 61–110.
Iwata, Seizi. 2005. Locative alternation and two levels of meaning. Cognitive Linguistics 16:
355–407.
Jaszczolt, Katarzyna M. 2012. Semantics and pragmatics: The boundary issue. In Semantics: An
International Handbook of Natural Language Meaning, Vol. 3, Claudia Maienborn, Klaus
von Heusinger & Paul Portner (eds), 2333–2360. Berlin: Mouton de Gruyter.
Kertész, András & Rákosi, Csilla. 2012. Data and Evidence in Linguistics. A Plausible Argumenta-
tion Model. Cambridge: CUP.
Kertész, András & Rákosi, Csilla. This volume. The p-model of data and evidence in linguistics.
Komlósy, András. 2000. A műveltetés (Causativisation). In Strukturális magyar nyelvtan 3:
­Morfológia (A Structural Grammar of Hungarian, vol. 3: Morphology), Ferenc Kiefer (ed.),
215–292. Budapest: Akadémiai Kiadó.
Pinker, Steven. 1989. Learnability and Cognition: The Acquisition of Argument Structure.
­Cambridge MA: The MIT Press.
Pusztai, Ferenc (ed.). 2003. Magyar értelmező kéziszótár (A Concise Dictionary of Definitions of
Hungarian), 2nd rev. edn. Budapest: Akadémiai Kiadó.
Stefanowitsch, Anatol. 2008. Negative entrenchment: A usage-based approach to negative evi-
dence. Cognitive Linguistics 19: 513–531.
 chapter 4

Methods and argumentation in historical

linguistics
A case study

Katalin Nagy C.
MTA-DE Research Group for Theoretical Linguistics

The present chapter aims to discuss some properties of historical linguistics

argumentation, through the analysis of examples taken from actual research
practice. Relying on Kertész and Rákosi’s (2012) p-model of scientific theorizing,
it compares three competing accounts of the historical development of the
Catalan construction “anar ‘go’ + infinitive”, namely, those provided in Colon
(1979a, b), in Detges (2004) and in Juge (2006). The simultaneous plausibility of
some statements and their negations in the starting p-context formed by these
three approaches leads to a p-inconsistency, which is eliminated by the extension
of the starting p-context and, then, the coordination of the extended p-context.
The analysis of historical research practice from a methodological point of view
suggests that the development of the Catalan “anar ‘go’ + infinitive” construction
finds the most satisfactory explanation under Juge’s (2006) proposal, which
uses sources, research methods and argumentation techniques traditionally
accepted in historical linguistics.

1.  Introduction

Linguistic inquiry always takes place in a context of previous research. At a cer-

tain stage of the investigation, practicing linguists have to summarise previous
results on the issue they are dealing with. In every linguistic field, some of these are
taken for granted. At the same time, linguistic inquiry takes place in a context of
argumentation traditions and commonly accepted methodological norms, which,
however, are not always considered consciously and sometimes are incompatible
with the research practice (cf. Kertész & Rákosi this volume a, 2012:  41ff.). ­Kertész
and Rákosi characterise scientific inquiry as a process of plausible ­argumentation
 Katalin Nagy C.

and include the above mentioned pieces of information in their so-called p­ -context
(Kertész & Rákosi, this volume b: Section 2.4, 2012: 121ff., for phases of scien-
tific inquiry see Kertész & Rákosi, this volume b: Section 2.6.1, 2012: 136ff.).
Researchers, when interpreting former literature on a certain problem, unavoid-
ably face rival claims, which they have to compare and to evaluate their reliability.
In other words, the p-context can contain statements and their negations, which
belong to competing approaches and are plausible simultaneously. In such cases
the researchers have to decide how to resolve this p-inconsistency (cf. Kertész &
Rákosi this volume b: Sections 2.5–6, 2012: Sections 10.2, 10.6). According to
­Kertész & Rákosi (this volume b: Section 3.1, 2012: 169ff.), data enter the argu-
mentation process with an initial plausibility value, which is determined by the
reliability of their source. But what does the reliability of different data sources
depend on? How can researchers decide between competing hypotheses?
What counts as a reliable source depends on the linguistic field we are work-
ing in. Sources which can support our statements include, for example, the “use
of special tools, methods, procedures […] certain methodological principles”
(­Kertész & Rákosi, this volume b: Section 2.2, 2012: Sections 9.1, 9.3). Linguists
usually apply the methods and argumentation techniques accepted in the lin-
guistics area in which they work automatically, without conscious reflection. It
also occurs that researchers accept previous claims made in the relevant literature
without asking about their reliability. The reputation of famous researchers also
plays a role in attributing plausibility values to their claims. In this way, the claims
of highly recognised authors are given high plausibility values when used as data
in other research. Statements often enter the argumentation process with a high
plausibility value, while it might be next to impossible to see where they come
from and whether they actually deserve such a high plausibility value.
The aim of the present chapter is to show the usefulness of taking into account
metalinguistic considerations in order to decide between competing hypotheses
in linguistic research and judge the reliability of certain claims. Using the metalin-
guistic framework presented in Kertész and Rákosi’s introductory chapter to the
present volume, some characteristics of historical linguistics argumentation are
examined through a concrete case study concerning a subquestion of the gram-
maticalisation of the Catalan construction “anar ‘go’ + infinitive” as a perfective
past. The present chapter does not aim to provide a full overview of argumentation
in historical linguistics but to illustrate the usefulness of metalinguistic consider-
ations in linguistic research.
The organisation of the chapter is as follows. After the present introduction, in
Section 2 two basic methods and argumentation techniques in historical linguis-
tics are discussed, namely, argumentation based on analogy and frequency count-
ing. In Section 3 their application is examined through a historical linguistic case
 Methods and argumentation in historical linguistics 

study in which three different accounts concerning the morphological development

of the Catalan “anar ‘go’ + infinitive” construction are presented and compared.
These three approaches form part of the starting p-context of the investigation.
The starting p-context is p-inconsistent, because it contains statements which are
plausible on the basis of some source (approach) but implausible on the basis of
others. The considerations presented in this section show how metalinguistic aware-
ness helps us decide between competing hypotheses and, in this way, eliminate the
­p-inconsistency. ­Section 4 deals with the role of universal tendencies in historical
linguistics argumentation. Finally, Section 5 discusses the conclusions of the chapter.

2.  Argumentation in historical linguistics

2.1  Quantitative and qualitative data in historical research

Frequency and analogy are key notions in historical linguistics argumentation in
the sense that argumentation methods which are widely applied in this linguistic
field rely on them. These argumentation methods are mainly based on data gained
from historical corpora. According to Fischer (2004: 730), the basis of historical
linguistic research is historical documents: “the historical linguist has only one
firm source of knowledge and that is the historical documents”. This quotation
uses a concept of data different from that accepted in our framework, which main-
tains that a datum is a statement with a plausibility value, and in most cases it is not
certainly true, only plausible to a certain degree. However, it shows how elemental
the corpus method is considered in historical linguistics.
Corpus data are regarded in Kertész and Rákosi’s framework as plausible
statements, which assert the presence of some linguistic structures in some corpus
or capture their characteristics (cf. Kertész & Rákosi, this volume b: Section 3.1,
Example 18, 2012: 173f.). These statements are only plausible, not certainly true,
because the reliability of corpus data is influenced by several factors, such as the
reliability of the historical documents, the researcher’s linguistic intuition and dif-
ficulties in the interpretation of the linguistic phenomenon under investigation
among others (cf. also Example 19 provided in Kertész & Rákosi, this volume b:
Section 3.1, 2012: 173f.). In any case, the corpus method should be considered
one of the main traditional tools of historical linguistics which presupposes the
use of statistics and frequency analysis. However, the above mentioned difficul-
ties should be taken into account when considering the plausibility values of data
gained with the help of these methods.
Naturally, beyond data originating in historical documents, historical linguis-
tics uses data from other types of sources as well (e.g. the contemporary language
 Katalin Nagy C.

system, a reconstructed language system, metalinguistic sources, and typological

sources among others). Quantitative data are important in each area of historical
linguistics, although they should be complemented with qualitative data, at least in
some fields (e.g. in historical semantics and historical pragmatics). The importance
of quantitative data in the study of some central phenomena of historical pragmat-
ics is emphasised, for example, by Fitzmaurice and Taavitsainen (2007: 27), who
highlight the importance of the quantitative analysis of large databases.

2.2  Frequency
Frequency is a central notion in historical linguistics in two respects: as a phe-
nomenon and as a method. On the one hand, several historical phenomena are
associated with the high frequency of the linguistic units concerned, and on the
other, frequency counting is a central method for investigating various historical
phenomena. Let us consider both concepts of frequency.
The definition of grammaticalisation as “the process of automatisation of
frequently occurring sequences of linguistic elements” (Bybee 2007: 969) shows
the importance of frequency also in grammaticalisation, especially in relation
to the generalisation of meaning. Some linguists consider the generalisation of
meaning as a precondition for more frequent use. Others think the opposite: it
is frequent use which facilitates the generalisation of meaning (Dér 2005: 48, cf.
Ladányi 1998: 420ff.), that is, according to this concept of frequency, it has a cogni-
tive impact. Actually, both claims seem to be maintainable. As Bybee (2005 [2003]:
602) points out, “frequency is not just a result of grammaticisation, it is also a
primary contributor to the process”. She argues that one of the main character-
istics of grammatical units is that, compared to lexical morphemes, they occur
extremely frequently. This frequency is not simply a result of grammaticalisation
but its main motive power. Examining the effects of repetition on constructions,
Bybee (2005: 10f.) finds that while low levels of repetition lead to conventionalisa-
tion only, extremely high frequency leads to the grammaticalisation of the new
construction, and that “the changes that take place in grammaticization are condi-
tioned at least in part by high frequency of use” (Bybee 2005: 11). These changes are
the following: phonological reduction, the gaining of autonomy of the new con-
struction, loss of the original meaning, and addition of a new, inferential meaning
from the context. Habituation processes cause semantic bleaching. Frequency also
has an impact on the strength of the exemplars and on category formation.
Let us consider the other concept of frequency. Frequency counting is a pow-
erful device of historical linguistics methodology, which provides quantitative
data and is applied in order to investigate the above mentioned cognitive pro-
cesses. Quantitative data should also be regarded as plausible statements of the
 Methods and argumentation in historical linguistics 

structure “There is a quantity Q of the linguistic structure or phenomenon X in

corpus C”. The use of this type of data raises the question of how representative
corpus C can be considered. Although the representativity of a corpus is desirable,
this requirement is very difficult to satisfy when we investigate earlier language
states, from which we do not have enough written documents. The frequency of a
given linguistic structure or phenomenon in such a corpus does not reflect its fre-
quency in the language state under study. In spite of this, there is an agreement in
the historical linguistics literature that quantitative data gained from – sometimes
relatively limited – available corpora can be successfully used in the investigation
of several historical phenomena. The researcher’s subjective judgements also influ-
ence the reliability of this type of data, above all when hardly accessible linguis-
tic phenomena are concerned. These uncertainties should be taken into account
when considering the plausibility of statements on the basis of such sources.
According to Bybee (2005 [2003]: 612), counting frequency provides us with
data about the process of grammaticalisation. She considers two relevant meth-
ods of counting frequency in linguistic studies, namely, token frequency and type
frequency, and defines token frequency as “the frequency of occurrence of a unit,
usually a word or morpheme, in running text”, while “type frequency refers to
the dictionary frequency of a particular pattern” (Bybee 2005 [2003]: 604), and in
another place she also notes that “the notion of type frequency can also be applied to
grammaticizing constructions by counting the different lexical items with which a
construction can be used” (Bybee 2005 [2003]: 605). According to Bybee, while high
token frequency corresponds to a very strong or highly entrenched local schema and
affects the nature of the cognitive representation, high type frequency corresponds
to the generality of the schema, that is, to a higher degree of grammaticalisation.
Type frequency and token frequency are closely related, since the increase in the
number of lexical units which can co-occur with a certain grammaticalising lin-
guistic item triggers an increase in token frequency. Bybee (2005 [2003]: 605–614)
applies these frequency counting methods in her study on the history of the English
auxiliary can. Table 1 shows the results from a Middle English corpus concerning the
different semantic groups of infinitives which can appear together with can.

Table 1.  Infinitives with can in Middle English (Bybee 2005 [2003]: 611f.)
1.  verbs of intellectual states or activities: 52 tokens, 18 types
  High token frequency: see 12, deem 6, understand 6, espy (discover) 5
  Type frequency: 18 distinct verbs
2.  verbs of communication: 102 tokens, 31 types
  High token frequency: tell 30, say 29, devyce ‘describe’ 8
  Type frequency: 31 distinct verbs
3.  verbs of skills (‘know how to’): 26 tokens, 18 types
 Katalin Nagy C.

Bybee compares these data with data from Old English. The comparison
shows that although these infinitives could also be used with can in Old English,
in Middle English more infinitives are documented, and also the token frequency
of some combinations in the first two groups of infinitive types increases. At the
same time, new infinitives appear with can.
To sum up, frequency analysis as a corpus method is a basic research tech-
nique in historical linguistics, since change in frequency is used to explain sev-
eral historical phenomena. However, it is worth noting that the research question
we are interested in and the theoretical framework we work in determine what
we count the frequency of. In this sense, the quantitative data obtained from fre-
quency counting is theory-dependent.

2.3  Analogy
Numerous historical changes discussed in the literature are attributed to analogi-
cal processes (Hock 1991: 167). In historical linguistics, the term ‘analogy’ tradi-
tionally covers processes whereby on the basis of a preexisting structural pattern
in a language, corresponding new patterns are introduced, or old patterns are
replaced (Jeffers & Lehiste 1979: 60, cf. also Bynon 1983). Analogical processes are
at work at every linguistic level. For instance, as is widely maintained in historical
linguistics, some changes in phonetic structure are influenced by factors which
are not phonetic in nature. One of them is analogy, whose main function is to
make phonetically and morphologically similar the form of lexical units which are
similar from a morphological, syntactic or semantic point of view. In morphology,
when a grammatical distinction is lost in some morphological classes as a result of
phonetic changes, the pattern of identity is usually extended by analogy (Jeffers &
Lehiste 1979: 71).
Hock (2005 [2003]) presents a wide range of phenomena which have been
considered analogical, namely, four-part analogy, leveling, morphophonemic
extension, blending, contamination, recomposition and folk etymology. He con-
cludes that
the entire range of analogical changes, as well as at least some aspects of semantic
change (mainly metaphorical changes) constitute points in a continuum of
changes which may be considered analogical in the larger sense, in that they
extend linguistic patterns. The differences in behaviour between these changes,
in terms of (potential) regularity or systematicity, are a consequence of the
differences in domain in which the changes can apply. (Hock 2005 [2003]: 455)

Let us consider two examples of how analogy works in language history. The cre-
ation of the medieval Spanish negation particle njn can be considered a simple
illustrative case of analogy. On the basis of direct sources it can be claimed that
 Methods and argumentation in historical linguistics 

the negation particles no (< non) and non (< non) ‘no’ alternate in medieval
­Spanish and that the forms ni (< nec) and njn ‘neither’ show a similar alternation.
­However, the final n of the form njn cannot be derived etymologically, that is, there
is no direct source which makes plausible that the form njn would be etymologi-
cal. Relying on the semantic similarity of the pairs no/non and ni/njn, it can be
supposed that the form njn was created by analogy of non, as shown in (1):
(1) ‘no’ no: non
‘neither’ ni: X = njn [nin]

The existence of the form njn in the medieval language state of Spanish has an
analogical explanation in the sense that it refers to the process of analogy at work
in its creation. Also, in the historical evolution of verbal systems a wide range of
­analogical developments can be traced. Let us consider an example from Alvar
and Pottier (1987 [1983]: 270). In medieval Spanish, several variants of the per-
fective past of the verb haber ‘have’ coexisted. In addition to etymological forms,
presumably analogically created forms were also used. Table 2 shows the perfec-
tive past forms of “habere > haber”, a verb of possession which also serves as
an auxiliary. Alvar and Pottier also enumerate some forms signalled as “strange
forms” which cannot be considered etymological, but cannot have an analogi-
cal explanation either. The forms in italics have survived and are used in current
Spanish. With one exception, they can all be regarded as analogical forms, that is,
their existence can be explained with reference to analogical processes at work at
an earlier language state.

Table 2.  Perfective past forms of haber (< HABERE) in Spanish

Etymological forms Analogical forms “Strange forms”

1sg (h)ove uve, hube, ovi of, off

2sg oviste, ovist(i) ovieste, hubiste
3sg ovo hubo ubu
1pl ovemos, ovimos oviemos, hubimos auimos
2pl oviste(i)s oviestes, hubisteis
3pl hubieron ovieron, ovioron

The phonetic changes which yielded the current perfective past forms of haber
can be explained by analogy. In this case it is difficult to imagine exactly how this
analogical process worked, since only one of the etymological forms contains the
vocal u (the 3Pl hubieron). In spite of this, it is the vowel of this verb stem which
extends to the other forms. The explanation may be that the speakers interpret
the vocal u as a marker of the perfective past tense, since it also marks perfective
 Katalin Nagy C.

past forms of other verbs in which it appears as an etymological vowel. Consider

examples in (2).
(2) au > o
SAPUI > sope
HABUI > ove
ou > u ~ o→u
POTUI > pude sope → supe
POSUI > puse ove → hube

In the perfective past forms of the verbs poder ‘can’ and poner ‘put’, pude ‘I could’
and puse ‘I put’, respectively, a verb stem appears which contains the vowel u
as a marker of the perfective past tense. Relying on an associative relationship
based on meaning, the vowel u can be supposed to extend analogically to other
verbs, including saber ‘know’ and haber ‘have’, yielding the perfective past forms
supe ‘I  knew’ and hube ‘I had’, respectively, and other analogical forms of the
paradigm.

2.4  Summary
The notions of frequency and analogy play a central role in historical linguistics
argumentation. Frequency-based argumentation makes use of plausible state-
ments which capture characteristics of some linguistic structures in quantita-
tive terms, while analogy-based argumentation contains references to analogical
processes which are supposed to work at a given linguistic level. Since they are
widely accepted methods and explanations in this field, statements from an argu-
mentation which relies on them can be assigned a high plausibility value. In the
following section let us examine how these notions appear in the research on a
concrete historical phenomenon and how the above considerations can help to
decide between competing hypotheses.

3.  A case study

3.1  Th
 e starting p-context: Three accounts of the morphological
development of the Catalan periphrastic perfective past
In this section let us examine some characteristics of historical linguistics argu-
mentation through a case study concerning the grammaticalisation of the periph-
rasis “anar ‘go’ + infinitive” in Catalan. Since this is a highly complicated process,
I discuss only a subquestion of this historical development.
The following hypotheses concerning the Catalan construction “anar +
infinitive” are unanimously accepted in the relevant literature. The c­ onstruction
 Methods and argumentation in historical linguistics 

“anar + infinitive” is currently used to express a perfective past meaning in

­Catalan. This verbal tense is formed with auxiliary forms that originate from the
full lexical verb anar ‘go’. The paradigm is shown in (3).

(3) The Catalan perfective past tense: vaig (vàreig)

vas (vares)
va + infinitive
vam (vàrem/vem)
vau (vàreu/veu)
van (varen)

The verb anar also continues to be used as a full lexical verb with the meaning ‘go’
in modern Catalan. Although the paradigms of the full lexical verb anar and of the
auxiliary anar almost coincide, they are divergent from each other in some forms
(see Table 3 in Subsection 3.1.3). The var- forms were created by analogy, modeled
on the regular preterit forms of a first conjugation verb.
It is generally acknowledged that the source of the current Catalan preterit,
the medieval “anar ‘go’ + infinitive” construction was originally a purposive con-
struction which had the sense of ‘be in motion with the aim of doing something’.
The development of the Catalan preterit “anar + infinitive” raises a wide range
of problems, of which in the present chapter I discuss only one, which is the fol-
lowing. Although the auxiliary forms of the perfective past “anar + infinitive” in
current Catalan formally seem to be in the present tense, medieval data attest an
alternation between present and preterit auxiliary forms. In other words, the ques-
tion discussed in the present chapter concerns the morphological development of
the “anar ‘go’ + infinitive” construction: the mix of present and preterit auxiliary
use in early documents and the later extension of the use of present forms and the
loss of the preterit auxiliary versions. The following Examples (4) and (5) show the
medieval alternation.

(4) encontraren-se ab la host dels sarraÿns e anaren ferir

  go-pret-3pl1 attack-inf
en éls (Desc II 103) 2

‘they met the troops of the Saracens and attacked (literally: went [in order]
to attack) them’

.  The following abbreviations are used in the glosses throughout this chapter: 3 – third
person, inf – infinitive, pl – plural, prs – present tense, pret – preterit tense, and sg – singular.
.  The sources of the examples are abbreviated at the end of each example, with page and/or
line numbers (full details of sources can be found at the end of the chapter).
 Katalin Nagy C.

(5) E lo senyor rei brocà ab la senyera e va ferir

  go-prs-3sg attack-inf
entre els moros (Munt I 92)
‘And the Lord King charged with his banner and attacked (literally: goes
[in order] to attack) in the midst of the Moors’

Examples like (4) and (5) show the present–past alternation in the auxiliary forms
of the construction “anar + infinitive” in medieval Catalan. Two versions of the
periphrasis under consideration are documented: first, “anar ‘go/auxiliary’ conju-
gated in perfective past + infinitive” (Example 4) and, second, “anar ‘go/­auxiliary’
conjugated in present + infinitive” (Example 5). The question examined here is
which of these two medieval periphrases should be considered the antecedent of
the perfective past of modern Catalan. At the beginning of the investigation of
this question, we will summarise previous results, which belong to our starting
p-context. The three approaches I examine below provide statements which are
made plausible by some source and their negations by another, that is, they yield
a p-inconsistent p-context. In the following sections I discuss how we extend the
starting p-context with new information, re-evaluate the plausibility values of
contradicting statements comparing the data use and argumentation in the three
different accounts, and finally, modify the p-context in order to eliminate the
p-inconsistency. Let us start with a summary of three accounts of the morphologi-
cal development of the “anar ‘go’ + infinitive” construction, namely, those of Colon
(1978a, b), Detges (2004) and Juge (2006).

3.1.1  Colon (1978a, b)

Concerning the question of whether the origin of the Catalan GO-past is the
periphrasis formed with the present or past tense auxiliary, Colon (1978a, b)
provides a ‘narrative present account’, which derives the current Catalan “anar +
infinitive” construction from a historical present usage (cf. also Badia i ­Margarit
1981). From a semantic point of view, such an analysis is primarily based on the
medieval occurrences of “anar + infinitive” with a present tense auxiliary. Colon
does not deny the existence of preterit auxiliary occurrences in the historical texts,
but quite simply renders them invisible for the semantic and pragmatic analy-
ses. According to his analysis, the current meaning of “anar + infinitive” can be
derived from the present auxiliary variant. Colon (1978b: 161) argues that the
reason for the loss of preterit forms in the auxiliary was a historical present use
which pushed the preterit forms of the auxiliary into the background. He (Colon
1978a: 128f., Colon 1978b: 161) explains this in the following way: the use of the
GO-periphrasis and the historical present fulfill similar functions, inasmuch they
both make the description more vivid and attract the hearers’ interest, and, since
the construction “anar + infinitive” serves to animate the description, it is p
­ erfectly
 Methods and argumentation in historical linguistics 

understandable that its auxiliary was conjugated in the historical present instead
of a preterit, because its role is the same and both “unite their virtues”.
In order to illustrate the claim that the construction “anar + infinitive” serves
to animate the description, Colon (1978a: 127f.) mentions an interesting example
which shows the auxiliary in present tense. This example of “anar + infinitive”
appears in a general description in the present tense in a Catalan text, which has
a parallel text in Provençal. In the corresponding Provençal text a simple present
appears when Catalan uses the GO-periphrasis with a present tense auxiliary. Colon
attributes this use to the function of animating the narration. This occurrence does
not correspond to the formally similar occurrences which show the periphrasis with
a present tense auxiliary in a preterit context, because it appears in a general descrip-
tion in the present tense, that is, the context is different. This datum from a direct
source certainly makes plausible Colon’s hypothesis about the early function of the
GO-construction (animating the narration). At the same time, however, it makes
plausible the following claim too: in the earliest examples the auxiliary should
appear in a tense required by the context (cf. Colon 1978a: 128; Colon 1979b: 159).
Consequently, it can be supposed that in the earliest examples the auxiliary of the
periphrasis should be conjugated in the preterit in a preterit context.
Concerning the later morphological development of the construction, Colon
(1978b: 169) only refers to the appearance of the var- forms. The var- auxiliaries
(cf. (3)) were created by analogy, modeled on the regular preterit forms of first
conjugation verbs. Colon suggests that the formal similarity of var- auxiliaries to
regular perfective past forms attests to the temporal identity of the construction.
That is, he supposes that the construction is already grammaticalised as a perfec-
tive past tense at that time (about the 14th century).
Let us give a summary of the development of “anar + infinitive” in Colon’s
(1978a, b) account step by step:

1. “Anar + infinitive” is a periphrasis with a motion verb conjugated as the context

requires, used in order to animate the narration.
2. In a preterit context the use of auxiliaries in the historical present extends, and
this historical present usage displaces, the preterit auxiliary forms. The reason is
that the historical present and the GO-periphrasis have the same function and
“join their virtues”.
3. The preterit forms of the auxiliary disappear, relegated to the background by the
narrative present usage. The loss of the simple preterit tense3 also plays a role in
this process.

.  Reserved only for literary purposes in most Catalan dialects.

 Katalin Nagy C.

The statements presented above are explicit claims in Colon (1978a, b). However,
because the plausibility of a conclusion does not only depend on the plausibility
values of the premises, but is also influenced by the plausibility of the latent back-
ground assumptions, we have to reconstruct them before considering the plausibil-
ity of Colon’s hypotheses. Latent background assumptions are statements which are
also needed to secure the plausibility of the conclusion in certain cases (cf. Kertész &
Rákosi, this volume b: Section 2.3). The following latent background assumptions
can be reconstructed at the crucial 2nd and 3rd points of Colon’s argumentation.
First, he supposes that two things which have the same function can fulfill this
function better together. However, he does not support this claim by independent
evidence. Second, he supposes that the present auxiliary forms in medieval texts
are representatives of the historical present. Third, he supposes that the historical
present has the function of animating narration (cf. Colon 1978a: 129, 1978b: 161).
These latter two latent background assumptions show that Colon’s argumentation
concerning the morphological development of the “anar + infinitive” periphrasis
is based on an assumed historical present use. This view is, however, not properly
established: Colon does not express explicitly what he means by “historical pres-
ent” and does not support these assumptions by other data. Because none of them
is supported properly, low plausibility values can be assigned to these latent back-
ground assumptions, which also decreases the plausibility of the hypothesis.

4. The periphrasis – with the GO-verb in present – acquires a past tense value.

As to the shift between the 3rd and 4th points of the development, according to
Colon (1978a, 1978b: 156), we cannot examine the beginnings of the history of
the Catalan “anar + infinitive” construction, because they already appear with a
past tense value in the earliest texts. In contrast with Colon’s claim, on the basis
of corpus data it can be claimed that in the earliest text the range of contexts in
which the periphrasis can appear is very limited. The grammaticalisation of the
construction has not yet concluded and if there is a past tense value, it is not yet a
grammatical meaning.

5. As a consequence, the analogical var- forms appear, demonstrating the temporal

identity of the construction.

This claim can be assigned a high plausibility value on the basis of data from sev-
eral different language systems (see below, in 3.1.3).
Colon’s analysis is based on a wide range of data, not only from Old Catalan
but also from other Romance languages. However, his account also contains an
inconsistency between two of his statements, namely, (i) that the periphrasis has a
past tense value, and (ii) that it is about a historical present use. This inconsistency
results from the circumstance that both claims are based on the same sources of
 Methods and argumentation in historical linguistics 

data, that is, the same historical texts, which represent the same language state. It is
obvious that the two statements, inasmuch they concern the same language state,
cannot be true simultaneously. If the first statement is maintainable, the GO-verb
should already function as a grammatical marker or at least be grammaticalised
to some degree. However, this would contradict the statement that we are dealing
with a historical present usage, because in this case the GO-verb should refer to
real motion. Consequently, the simultaneous presence of both statements in the
argumentation would yield a p-inconsistent p-context. However, there is another
possibility: we can suppose that Colon intends the two statements to refer to two
different language states. In this sense, the second one concerns an earlier period
of the history of the Catalan, from which, as Colon also points out, we do not have
documents. For lack of direct sources from this language period, it is not possible
to support this claim properly. Consequently, it receives a low plausibility value,
which decreases the plausibility of Colon’s account.

3.1.2  Detges (2004)

Detges also claims that the present auxiliary variants of “anar + infinitive” are
representatives of a historical present usage. Detges’s (2004) argumentation is also
similar to that of Colon in that he attributes to the present tense auxiliary vari-
ant two effects: dynamism (due to the verb ‘go’) and presentification (due to the
historical present). He claims that the tense of the auxiliary is optional, although
the present-tense variant was used more frequently at all times. He also maintains
that the present auxiliary periphrasis conventionalised as a foreground-marker
due to its high frequency. Let us reconstruct Detges’s argumentation concerning
the present-past alternation in the auxiliary of “anar + infinitive” in more detail.

1. The narrative genres of medieval Western Romance show an unsystematic pres-

ent tense – past tense alternation, especially in earlier texts (Detges 2004: 218).

The problem with this claim is that, although it may be maintainable, it is hard
to see whether it is relevant enough to include in the argumentation. Although it
is possible that the formation of the past value of this Catalan construction took
place in the context of such an incoherent verbal tense use, Detges does not pro-
vide examples taken from these types of historical documents. Moreover, this
claim does not hold for the medieval Catalan chronicles, which contain the earliest
documented occurrences of the periphrasis under study. Detges himself presents
examples4 which show a coherent use of preterit tenses in narrative, only inter-
rupted by occurrences of the GO-periphrasis with an auxiliary in the present, but

.  These examples are provided at another point of his argumentation and with a different
aim from that of the present discussion.
 Katalin Nagy C.

with no other present tense forms. Example (11) provided in Detges (2004: 217,
taken from Colon 1978b: 135f.) is a passage from Desclot’s Chronicle. It contains
14 verbal forms, of which 13 are in the past tense, and the only one in the present is
the auxiliary of an occurrence of the GO-periphrasis. In his other example (again
taken from Colon 1978b: 169) he presents a passage from Muntaner’s Chronicle.
This contains 23 verbal forms. Two of them appear in a direct quotation and
another two in the expression addressing the public Què us diré? ‘What shall I tell
you?’ The rest of them belong to the real narrative part of the text, and 5 of them
are representatives of the “anar + infinitive” periphrasis. Of these five occurrences
only one has the auxiliary in the preterit, the other four are occurrences with a
present tense auxiliary. However, these present auxiliary forms are the only pres-
ent forms with a past tense reference in the text. That is, there are no other present
forms which could be representatives of the historical present: the coherent verbal
tense use in this narrative is only interrupted by the present auxiliary variants of
“anar + infinitive”. The illustrative examples do not show the unsystematic alterna-
tion of verbal tenses assumed by Detges. Consequently, this is not a felicitous argu-
ment, or at least it should be mentioned that the periphrasis also appears in texts
whose verbal tense use is systematic and coherent. The present tense forms in this
type of texts should be explained, too. That is, the p-context of Detges’s investiga-
tion is p-inconsistent in the sense that the statement concerning the “unsystem-
atic present tense – past tense alternation” is claimed to be plausible, while direct
sources make its negation plausible.

2. The historical present is dispensable for the “effect” of the periphrasis; the tense of
the auxiliary is optional.

Detges does not explain why, if this is the case, historical present would appear at
all, and why only in the case of this periphrasis. At this point a latent background
assumption can be reconstructed, that is, that the present auxiliary occurrences of
“anar + infinitive” are representatives of a historical present usage. However, since
he operates with an intuitive notion of the historical present, this statement is not
properly supported. Consequently, it has a low plausibility value, which weakens
the plausibility of his hypothesis as well.

3. The version which has the auxiliary in the present is the “stronger alternative”.
The present-tense realisation combines two effects, namely, the “dynamification
effect and, due to the historical present, “the actuality-effect” (Detges 2004: 218).

Detges bases this claim, again, on the latent background assumption that the pres-
ent auxiliary versions are representatives of a historical present usage. He argues
that they are stronger versions, since they combine two effects. But how can we
 Methods and argumentation in historical linguistics 

know what “effect” the periphrasis had in the Middle Ages? We should support
such a claim with evidence accepted in historical linguistics. If in the medieval
language there actually existed a stronger and a weaker version of the periphrasis,
such a difference should be reflected in the documented uses. However, contexts
in historical documents do not reflect a “weaker version ‒ stronger version” oppo-
sition: in early texts the present and the past tense auxiliary versions appear in
the same types of contexts and with the same infinitives. Consequently, we do
not have supporting evidence for the claim that the present auxiliary variant was
stronger. This statement is only based on the author’s subjective judgment and our
sources (i.e. historical documents) make it implausible by supporting its nega-
tion. Once again, the extension of the p-context of Detges’s analysis with new data
results in a p-inconsistency.

4. The present-tense variant was used more frequently at all times (Detges 2004: 218).

Detges does not support the higher frequency of the present auxiliary version with
(quantitative) data. Thus, the reader can misinterpret this statement, believing that
statistics have been used in the investigation, which is not the case. Moreover,
statistical data suggest the opposite hypothesis, according to which the preterit
auxiliary variant of “anar + infinitive” was more frequent at the beginning and was
displaced by the present auxiliary forms only later. In other words, the extension
of the p-context of his investigation with new information originating from these
statistics yields a p-inconsistency. We have to decide, then, which one of the two
simultaneously plausible statements to accept: the statement in 4. or its negation.

5. As a consequence of its higher frequency, the present-tense variant “finally came

to be conventionalized as the only possible form of the foreground-marker”
(­Detges 2004: 218).

Detges attributes the loss of the preterit forms to the higher frequency of present
auxiliary variants. Although frequency can have this effect, Detges does not sup-
port this claim with a frequency analysis of historical occurrences, neither does he
take into account results presented in previous literature, which show the oppo-
site: at the beginning the preterit auxiliary occurrences were more frequent. Con-
sequently, on the basis of information from relevant literature we have to judge
this statement to be implausible.
To sum up, Detges’s argumentation does not rely sufficiently on data whose
plausibility value originates from direct sources and in his argumentation there
appear statements which are not presented as results of usually accepted histori-
cal linguistics argumentation methods. At the same time, when he uses accepted
argumentation methods (e.g. based on frequency), his statements are either not
 Katalin Nagy C.

supported by data whose plausibility value originates from some direct source, or
further sources render them implausible.
In addition, a textual contradiction appears, confusing the reader. Detges
(2004: 213) translates the Old French form (il) va dire in the following way: “not
simply ‘he starts to say’, but more specifically ‘all of a sudden, he says’ (Gougenheim
1971 [1929]: 96)”. And elsewhere he says (Detges 2004: 217): “il va dire ‘suddenly,
he said’ (Gougenheim 1971 [1929]: 96–97)” with a translation in the preterit. We
do not know whether this contradiction between two different translations with
present and past tenses comes from Gougenheim or it is Detges’s own mistake. In
any case, it is very confusing, as the question of tense is a crucial one in the devel-
opment of this construction.
In the two accounts presented above, the morphological analysis of the devel-
opment of the Catalan GO-past amounts only to the “historical present explana-
tion”. However, it is in itself not properly established. Consequently, the p-context
of the research should be extended with new information about historical pres-
ent usage. In addition, these two analyses need to be complemented with more
detailed morphological and pragmatic analyses, because the plausibility of a state-
ment which is in accordance with analyses of other linguistic levels is higher. As
Kertész and Rákosi (this volume: Section 2.2, see also 2012: 76f.) point out, “if
several sources support a statement, then its plausibility value is higher on the
basis of all sources together than its plausibility value on the basis of any of the
sources alone”.

3.1.3  Juge (2006)

The third account of the grammaticalisation of the Catalan GO-past presented
here is Juge’s (2006) analysis, which examines the morphological aspect of the pro-
cess more thoroughly than the previous accounts. Juge (2006: 314) characterises
the history of “anar ‘go’ + infinitive” in the following way: “this seemingly puzzling
development results from the interaction of a number of commonly found dia-
chronic processes with language-specific constructional patterns”. In opposition
to the other two accounts, Juge takes the past auxiliary forms as the starting point
in his analysis. As a first step, he analyses the morphology of the auxiliary and
compares it with the paradigm of the full lexical verb anar ‘go’. The main difference
consists in the 1Pl and 2Pl forms, which showed a present–preterit syncretism in
Old Catalan. Juge’s initial data is shown in Table 3.
Juge (2006) claims that the present forms in the current paradigm result from
analogy and that this process only took place later, when the past meaning of the
whole construction had already consolidated. The syncretism of a very current
form of anar ‘go’ (anam ‘we go/we are going/we went’, and the second person plu-
ral anats) made it possible to reanalyze the periphrasis as a construction with a
 Methods and argumentation in historical linguistics 

Table 3.  Paradigms of anar considered in Juge (2006)

The Catalan preterit auxiliary and anar ‘go’ The present and preterit indicative of anar ‘go’
(Juge 2006: 314) in Old Catalan (Juge 2006: 320)
Preterit auxiliary ‘to go’ Present Preterit

vaig (vàreig) vaig vaig ané

vas (vares) vas vas anast
va va va anà
vam (vàrem/vem) anem anam anam
vau (vàreu/veu) aneu anats anats
van (varen) van van anaren

present tense auxiliary. As a consequence, the use of the present extended to the
rest of the paradigm of the auxiliary. Let us observe his methods and argumenta-
tion in more detail.

1. “Anar + infinitive” is a periphrasis with a preterit auxiliary.

This claim can be (and is) supported by frequency analysis, consequently, a high
plausibility value can be assigned to it.

2. A semantic change can already be detected in some preterit auxiliary forms.

That is, “anar + infinitive” with a preterit auxiliary becomes a past tense.

Juge (2006: 319) presents corpus data in support of this claim: in his corpus, 32 of
the 157 (20.4%) instances of anà ‘he went’ “are plausibly cases of the periphrastic
preterit” and, similarly, with anaren ‘they went’ there are 164 instances in all, with
40 (24.4%) of these “being good candidates for the periphrasis”.

3. A reinterpretation of the following type takes place: “anar (preterit) + infinitive”

→ “anar (present) + infinitive”. To put it differently, language users reinterpret
the periphrasis as a construction with a present tense auxiliary. Frequency and
analogy trigger this process. The result is that “anar + infinitive” is now a periph-
rasis with a present tense auxiliary.

The third point of the development described in Juge (2006) is crucial. At this
point, language users presumably reinterpret the periphrasis as a construction
with a present tense auxiliary. In this reinterpretation, Juge (2006: 320) attri-
butes a crucial role to the ambiguity between present and preterit found in the
first and second plural indicative (cf. Table 3). Because present and preterit are
formally identical in the first and second plural indicative, language users rein-
terpret the construction with a preterit auxiliary as one with a present auxiliary.
 Katalin Nagy C.

Juge (2006: 320) claims that “the influence is not directly on the other forms in the
paradigm; rather it is on the interpretation of the structure of the construction as
a whole”, that is, at this point Juge assumes an analogy with other constructions
which have the auxiliary in the present tense.
The analogy-based and frequency-based arguments are connected here. On
the one hand, according to Juge (2006: 321), “one possible reason for the reanalysis
of forms such as anam as present in this construction is the tendency for finite
verbs participating in periphrases to be morphologically present tense”. Juge’s
argumentation, relying on analogy with formally similar constructions, takes into
account the whole verbal system. His other argument (Juge 2006: 322) is based on
universal tendencies:
there is a strong crosslinguistic tendency for auxiliary marking to be in the
present, although this is certainly not an absolute universal […]. In various
Romance languages, for example, the perfect TAMCATS [Tense-Aspect-Mood-
CATegories] allow a wide range of auxiliary forms, and the progressive and GO-
future do so to a lesser extent. Even in these cases, the non-present marking on
auxiliaries strongly tends to be imperfective, not perfective.

That is, the reinterpretation of the construction “anar + infinitive” as a construc-

tion with a present tense auxiliary is triggered by the existence of other periphrases
with a past reference, which have a present, or at least a non-perfective auxiliary.
On the other hand, Juge claims that it is frequency which triggers this process.
As he (Juge 2006: 320) points out, “such forms, especially the 1Pl, are abnormally
frequent because they are first person narratives”. Table 4 shows the data provided
by Juge concerning the frequency of different forms of anar.

Table 4.  “Frequency of first person plural and other forms in Old Catalan”
Juge (2006: 321)
Preterit Present indicative

Form # Form #
1S vag 1
2S vas 1
3S anà 41 va 13
1Pl anam 209 anam (209)
3Pl anaren 71 van 10

209 of 346 occurrences represent the 1Pl form anam ‘we go/we are going/we
went’, which is a relatively high frequency. The frequency-based argumentation at
this point consists in claiming that the reinterpretation of “anar + infinitive” as a
periphrasis with a present tense auxiliary is realised due to the high frequency of
 Methods and argumentation in historical linguistics 

the 1Pl form anam. This form is ambiguous between past and present tense and
allows an association to the present tense paradigm of the verb anar.
Two problems should be mentioned, which can decrease the plausibility of
this statement. First, in other cases of analogical change the basicness of third per-
sons is emphasised. Second, it is questionable whether the frequency will remain
the same, when investigating it in a larger corpus. As to the first problem, the
extremely high frequency of this no-third person form can explain its central role
at this particular occasion, because it can be considered a basic form. Juge’s argu-
mentation at this point is in accordance with Kuryłowicz’s second “law” of anal-
ogy: “analogical developments follow the direction ‘basic form’ → ‘derived form’,
where the relationship between basic and derived forms is a consequence of their
spheres of usage” (Hock 1991: 212). Also: “of any given set of forms or morpho-
logical classes, the one which has a greater sphere of usage is more ‘basic’ than the
others” (Hock 1991: 214). That is, the 1Pl form anam, due to its high frequency,
can be considered a basic form. As to the second problem, at the moment, it is a
plausibility weakening factor, and needs more investigation.

4. Later, new forms are created by analogy of regular preterit forms. Juge (2006: 316)
argues that other forms are analogically created from the root va-, since the form
va ‘(he) goes’ is both monosyllabic and third person singular.

This claim can be supported by crosslinguistic data, since the basicness of the third
person can be found in many languages and often has diachronic consequences (cf.
Hock 1991: 220). Third persons tend to be more basic in analogical change than
other forms of the verb (Hock 2005 [2003]: 446), cf. also Kuryłowicz’s “sphere of
usage” provision (Hock 1991: 220ff.). Some forms created on the root va- resemble
regular preterit forms like cantares ‘you sang’. These are also given an explanation
in Juge (2006: 316). The marking of the periphrasis with a preterit meaning by a
preterit marker also has a parallel in another language. As Juge (2006: 316) explains
Russian presents a parallel for such a ‘double’ marking where the tense of the
auxiliary is the same as that of the construction as a whole. […] There may then
be a crosslinguistic tendency to make the auxiliary match the overall value of the
construction, thus leading to the forms of va- that look like preterits.

Finally, another statement crucial in the argumentation in Juge (2006) is the

following:

5. The present auxiliary occurrences do not show the characteristics of a historical

present use.

Juge (2006: 319) supports this claim by another one: all the present tense auxiliary
occurrences found in his corpus are representatives of the periphrasis (that is, they
 Katalin Nagy C.

do not retain the lexical meaning ‘go’ of anar). As he emphasises: “there is NOT A
SINGLE EXAMPLE of va or van used non-periphrastically with a past value” and
the present tense variants appear “with no other present tense forms nearby” (Juge
2006: 319; emphasis in the original). Similarly to the other two authors, Juge does
not explain exactly what the historical present is. Although his claims concerning
this issue are intuitively acceptable and seem to provide a correct description of
the phenomenon, with an extension of the p-context with new, relevant informa-
tion, the plausibility of his statements can also be modified.
To sum up, Juge’s (2006) argumentation takes into consideration the whole
verbal system and crosslinguistic tendencies, too. He often makes typological
comparisons and relies on examples and parallels from other languages. Juge bases
his claims on data originating from a historical corpus and provides an exhaustive
morphological analysis of the occurrences. He explains the morphological devel-
opment of the construction from the beginnings until the modern paradigm with
argumentation methods accepted in historical linguistics. References to analogical
processes and frequency are widely used in his argumentation, yielding a convinc-
ing analysis.

3.2  Extension of the starting p-context: The historical present

One of the main differences between the three analyses discussed above consists in
attributing the shift in the tense of the auxiliary to different reasons. The first two
accounts (i.e. Colon 1978a, 1978b and Detges 2004) attribute it to a historical pres-
ent usage, and the third one (i.e. Juge 2006) to morphological reasons. However,
they all have something in common: they operate with an implicit notion of the
historical present, that is, they do not say explicitly what they mean by this term.
If we want to argue for or against a historical present usage in medieval contexts of
“anar + infinitive”, we should have a notion about what the historical present is and
what function it has. That is, we have to extend the starting p-context with new
information and reconstruct the latent background assumptions concerning what
the historical present is, because the plausibility of the three hypotheses is also
influenced by the plausibility of the latent background assumptions of the infer-
ences which support them as indirect sources (cf. Kertész & Rákosi, this ­volume:
Sections 2.3–4).
We can suppose that Colon’s and Detges’s argumentations are based on an
intuitive notion of the historical present, which perhaps conceives of the historical
present as the present tense used to refer to past events and whose function is ani-
mating the narration. However, the use of the historical present is a more compli-
cated issue. Juge (2006: 319) gives a little bit more information about his notion of
the historical present when he argues that the present auxiliary versions cannot be
 Methods and argumentation in historical linguistics 

representatives of the historical present, because they appear “with no other pres-
ent tense forms nearby”. This suggests that historical present forms usually do not
appear in isolation. In any case, it is undeniable that if the historical present tense
forms had been grammaticalised in the Catalan GO-past, there should have been
also forms of anar in the historical present combined with an infinitive, in which
the verb anar retained its lexical meaning ‘go’. However, as Colon also emphasises,
the formally present forms all seem to have a past tense value and do not retain the
lexical meaning ‘go’ of anar.
Now I will attempt to summarise what is worth knowing about the historical
present for the purposes of the present chapter. The following hypotheses con-
cerning the historical present originate from the relevant literature, where they
are considered to deserve a high plausibility value. The historical present is tra-
ditionally defined as the present tense used in the narration of events set in the
past, and it has been suggested that it makes stories more vivid by bringing past
actions into the present (see relevant literature given in Wolfson 1979: 169 and in
Schiffrin 1981: 46). It occurs when relating the most crucial events in a narration.
Fludernik (1992) points out that the historical present use reflected in early texts
is related to an oral pattern of story telling. Later, however, as the written standard
is developed, historical present usage suffers some changes:
In the process of narrative development from the written codification of oral
storytelling via the written composition of texts on an oral model towards a
purely written conception of narrative structure, the shape and function of the
historical present tense necessarily undergo equivalent changes.
(Fludernik 1992: 1, cf. also Fludernik 1992: 21)

We can suppose that the use of the historical present may also be close to the oral
pattern in early Catalan texts.
Beyond the simple characterisation presented above, some interesting new
findings on the historical present have been put forward. The first is Wolfson’s
(1979) claim that the conversational historical present has no significance in itself.
To put it differently, it is not simply the historical present use that matters, but
the switching between present and past tenses. This present–past alternation is a
discourse phenomenon which organises and segments the narrative (cf. Wolfson
1979: 178): it separates narrative events from each other, and, at the same time,
highlights the turning points of a story. Wolfson (1979: 172) also emphasises that
“many stories are so organized that what seems to be the most important event is
given in the past tense”. That is, she does not accept the claim that the historical
present would have the function of dramatising certain acts. Rather, as ­Schiffrin
(1981: 59) explains, the shift to the historical present seems to be an “internal eval-
uation device: it allows the narrator to present events as if they were occurring at
 Katalin Nagy C.

that moment, so that the audience can hear for itself what happened, and can inter-
pret for itself the significance of those events for the experience”. Silva-­Corvalán’s
findings (1983: 774f.) offer Spanish support for Schiffrin’s claim for English that
historical present usage is an internal evaluation device.
The second finding is that historical present forms tend to occur more fre-
quently in specified points in the narrative but are missing in others. Schiffrin
(1981) analyzes how the use of the historical present depends on the organisation
of narratives. Fludernik (1992: 23) too, emphasises that the dynamics of the histor-
ical present should be discussed in relation to narrative episodes in their totality.
The present auxiliary versions of “anar + infinitive” actually can be found at
certain points of the narrative, namely, in the description of events which can be
considered “turning points” of the story. However, we should not conclude from
this that it is about a historical present usage. As Colon points out, both the periph-
rasis and the historical present share the same function, and this can be the reason
why they appear at the same points of the narrative: in the description of events
which the speaker wants to attract the hearers’ attention to. We cannot conclude
that it is about a historical present use because there are at least two other options
to explain the presence of the construction at these points of the narration. On the
one hand, it can be due to the value of the periphrasis in itself, if we regard it as a
pragmatically marked construction at that language state. On the other hand, the
present tense of the auxiliary can be also a simple morphological characteristic as
described in Juge (2006), that is, the tense of the auxiliary of an already grammati-
calised construction.
It is also worth noting Schiffrin’s (1981: 47) finding, according to which in
narrative clauses about 30% of the verbs are in the historical present. Moreover,
she notes that “there is a tendency for verbs in the same tense to cluster together.
[…] Thus sequences with rapid alternation between the HP [historical present]
and the P [past] are not typical” (Schiffrin 1981: 51) (for the clustering of his-
torical present forms cf. also Pinkster 1990: 73, 75). Investigating the historical
present/preterit alternation in oral Spanish narrative, Silva-Corvalán (1983: 767)
finds a very similar percentage for Spanish: “of 476 narrative clauses, 156 (32.7%)
verbs are in the historical present”. She claims that “this cross-linguistic similar-
ity may respond to a universal pragmatic function” (Silva-Corvalán 1983: 767,
­Footnote  13). These two characteristics, again, do not fit the present auxiliary
use in the medieval Catalan chronicles. In these texts the proportion of the pres-
ent variants of the construction does not reach 30%, and the occurrences do not
cluster together. Although in some passages of Muntaner’s Chronicle some occur-
rences appear close to each other, it is not a general characteristic of the use of the
­periphrasis in all chronicles, and it can also be attributed to the higher degree of
grammaticalisation of the construction. Moreover, what is typical for the use of
 Methods and argumentation in historical linguistics 

“anar + infinitive” in these medieval texts is precisely the rapid alternation of the
present auxiliary versions of the construction with verbal forms in the preterit,
which is not typical for a historical present usage. The main evidence, however, is
that “there is NOT A SINGLE EXAMPLE of va or van used non-periphrastically
with a past value” and the present tense variants appear “with no other present
tense forms nearby” (Juge 2006: 319). This finding argues contra the historical
present hypothesis, because a historical present use which affects only the occur-
rences of the “anar + infinitive” construction, but no other verbal forms, would be
very strange, and moreover, counterintuitive.
To sum up, the extension of the p-context with new information concerning
the historical present helps us to eliminate one of the p-inconsistencies present in
the starting p-context, namely, that which exists between the statement that there
is a historical present use in the historical documents under study (considered as
plausible in Colon’s and Detges’s accounts) and its negation (made plausible by
Juge’s approach). On the basis of the new information presented in the present
section we should modify the plausibility values of these statements: the statement
that it is about a historical present use should be evaluated implausible, while the
plausibility of its negation should be strengthened.

3.3  Coordination of the extended p-context

Consider a brief summary of the morphological development of “anar + infini-
tive” as characterised in the three accounts presented above:

1. The “historical present account” in Colon (1978a, b) and in Detges (2004):

“anar (conjugated in the preterit) + infinitive” → the loss of preterit forms of anar in
the construction, displaced by occurrences of anar in the historical present → “anar
(in the historical present) + infinitive” → semantic change → “anar (auxiliary, con-
jugated in the present) + infinitive”

2. The “morphological account” in Juge (2006):

“anar (conjugated in the preterit) + infinitive” → semantic change → “anar (aux-

iliary, conjugated in the preterit) + infinitive” → formal change of the auxiliary by
analogy → “anar (conjugated in the present) + infinitive”.

We can see that our starting p-context is problematic, because it is informationally

overdetermined: the simultaneous presence of certain statements and their nega-
tions in it results in a p-inconsistency. The main difference is that the two analy-
ses posit the semantic change at different points of the development. In the first
analysis it is the periphrasis with present auxiliary which undergoes the semantic
 Katalin Nagy C.

change, and in the second account it is the periphrasis with preterit auxiliary that
does. Similarly, the spread of the present auxiliary forms is accounted for differ-
ently in the two analyses. In the first, it is the result of an extension of the historical
present usage (which, very strangely, only affects the forms of this periphrasis). In
the second analysis, however, they are the result of a reinterpretation, triggered by
syncretism in one of the most frequent forms of the paradigm. According to Juge’s
theory, it was the change in the meaning which triggered formal change. This hap-
pens frequently in analogical changes: the users of a language give similar forms to
semantically similar linguistic units (cf. Hock 1991: 167).
The higher frequency of preterit forms of anar in the construction in early
texts in itself does not contradict the first analysis. The problem lies in the circum-
stance that the occurrences of the construction with the GO-verb in the present
show a higher degree of grammaticalisation in the sense that in medieval Cata-
lan chronicles we do not find any occurrence with a present form of anar which
would retain the lexical meaning ‘go’. In contrast, the preterit auxiliary versions
show a greater variety of meanings: some of them retain the old, lexical meaning
of motion, while others offer a new, more plausible interpretation. Table 5 shows
the distribution of supposed meanings of the occurrences of “anar + infinitive” in
two medieval Catalan chronicles.

Table 5.  Distribution of supposed meanings in the occurrences of “anar + infinitive”

in Desclot’s and Muntaner’s chronicles
Desclot (1283–88) Muntaner (ca. 1325)

Past Present Past Present

auxiliary auxiliary auxiliary auxiliary

‘motion’ is likely 61 (81.3%) – 35 (54.7%) –

‘motion’ is possible, but 10 (13.3%) – 4 (6.25%) –
the proposed action seems
to be more relevant
‘motion’ is unlikely 4 (5.3%) 26 25 (39.0%) 158

The proportions in Table 5 show a more advanced stage of grammaticalisation

of “anar + infinitive” in Muntaner’s chronicle, where it is unlikely to assign the
‘motion’ meaning more often than in Desclot’s chronicle. In addition, Muntaner’s
chronicle, written almost half a century later than Desclot’s, not only contains
more than twice as many occurrences of “anar + infinitive” as Desclot’s chroni-
cle, but Muntaner uses them several times in the same context. These data sug-
gest a change from the full meaning ‘go’ to an abstract, aspectual or grammatical
meaning.
 Methods and argumentation in historical linguistics 

As regards the occurrences with a present tense auxiliary, in Desclot’s chron-

icle there are 26 of this type, while in Muntaner’s work I found 158. All of them
can be regarded as examples of the periphrasis, that is, their auxiliary presumably
does not retain the lexical meaning of motion. In other words, occurrences with a
present form of anar can be supposed to be grammaticalised forms. Consider the
following Example (6).
(6) Mas con foren prop d’Agda, noves los van venir
  go-prs-3pl come-inf
con havia pres, lo dia passat, a aquells (Munt II 10)
‘But when they came near Agda news arrived (literally: go [in order]
to come) that those people had been taken on the previous day.’

In Example (6) the subject (‘news’) of the construction suggests that van venir
‘arrived’ should be a grammaticalised form, conveying a past tense meaning. This
claim is also supported by the tense of the verbal form havia pres ‘had been taken’
in the second clause. This past tense expresses an action, state or event that took
place before the start of another past action, that is, the form van venir should refer
to a past event. The meaning of the infinitive itself ‘come’ also suggests that the
form van does not retain its lexical meaning ‘go’, since it would yield the strange
and nonsense reading ‘the news come/are coming (in order) to go’. Several similar
occurrences documented in the medieval Catalan chronicles show that the require-
ment of an agentive, human subject and that of intentionality are disappearing.
The considerations presented above support the hypothesis according to
which present auxiliary occurrences of “anar + infinitive” in medieval texts are not
representatives of a historical present usage but rather forms which, in comparison
with the preterit auxiliary variants, show a more advanced stage of grammaticali-
sation. On the basis of the above considerations, we can attribute a high plausibil-
ity value to the “morphological account”, while the “historical present account”
seems to be unacceptable because it fails to explain adequately the data.

4.  Modification of the p-context and comparison of the rival solutions

At this point of the investigation, we have to compare the rival solutions, decide
which one to accept and, with the modification of the p-context, eliminate the
p-inconsistency.
Information concerning universal tendencies in linguistic change often helps us
to decide between competing hypotheses. Let us discuss how considerations about
universal tendencies in analogy and universal tendencies in semantic change in
grammaticalisation appear in historical linguistic argumentation. Various u
­ niversal
 Katalin Nagy C.

tendencies have been detected concerning the way in which meanings change in the
languages of the world (cf. Bybee et al. 1994, Heine & Kuteva 2002, among others),
but also in the case of analogy we find some tendencies concerning the way in which
analogical change usually occurs. Argumentations in linguistics which make use of
universal tendencies also rely on plausible inferences, because linguistic universals
cannot be regarded as certainly true statements, they can be only plausible. Raising
hypotheses about universal characteristics of languages starts from the investigation
of a limited number of languages when the presence of the characteristics at issue is
checked, and in a second step, an inductive inference is applied, where
the set of the premise-candidates has to be supplemented by the background
assumption that the cases not examined also possess the characteristics that
could be found in the investigated ones; the conclusion states the presence of
these characteristics as a general rule.  (Kertész & Rákosi 2012: 90)

Kertész and Rákosi examine various examples of plausible argumentation which

concern linguistic universals (see especially Example 12 in Kertész & Rákosi 2012;
for the testing of potential linguistic universals, see also Examples 4(1), 9(3), 19,
20(2) in Kertész & Rákosi 2012).
Statements pertaining to linguistic universals can also function as data in his-
torical linguistics and appear in the set of premises of inferences which aim to
judge the plausibility of some hypothesis of the theory (cf. Heine 2005 [2003]: 580,
585 f.). In relation to natural directionalities in analogical change, Hock (1991: 210)
also points out that “they might also be helpful in choosing between alternative
possible analyses. Everything else being equal, we would select that analysis which
better agrees with what is known to be more natural”. That is, statements concern-
ing universal tendencies of change can enter the argumentation process, and they
can make certain statements implausible.
Since I have presented above how tendencies in analogical change appear in
historical linguistic argumentation (see Subsection 3.1.3), in the present section I
only examine how considerations about universal tendencies in semantic change
can help us to choose between the competing accounts of the historical phenome-
non presented above. The question of how the development presented in the pres-
ent chapter fits universal tendencies of semantic development is twofold. First, we
can examine which lexemes past tense markers usually develop from, and, second,
what semantic change lexemes meaning ‘go’ usually undergo. Regarding the first
question, the “‘go’ conjugated in the present tense + infinitive” construction, past
tense development does not fit any universal tendency (cf. Heine & Kuteva 2002).
As to the second question, the lexeme ‘go’ can suffer different semantic develop-
ments in different constructions. Let us consider the following statements based
on Heine & Kuteva (2002).
 Methods and argumentation in historical linguistics 

a. The construction “GO-verb conjugated in present tense + infinitive” tends to

develop into a future tense in the world’s languages.
b. There is no language in which the construction “GO-verb conjugated in present
tense + infinitive” would have developed into a past tense.

We can use these general claims about the semantic development of lexemes
meaning ‘go’ in assigning plausibility values to the following two statements:

1. The Catalan GO-past comes from the construction “GO-verb conjugated in the
present tense + infinitive”.
2. The Catalan GO-past does not come from the construction “GO-verb conjugated
in the present tense + infinitive”.

In the case study presented above our starting p-context contained three hypoth-
eses concerning the morphological development of the Catalan “anar + infini-
tive” construction. Two of them made statement (1) plausible, while the third
made statement (2) plausible. Consequently, the p-context was p-inconsistent. We
extended the p-context with new information and re-evaluated the plausibility of
the three hypotheses. Now, we have to modify the p-context and decide between
the rival hypotheses. I want to show how taking into account universal tendencies
helps us to do that. A high plausibility value can be assigned to the statements
in (a) and (b), because they are supported by several case studies in historical
linguistics. Relying on (a) and (b), we can assign a high plausibility value to (2),
because it is in accordance with the general tendencies presented in (a) and (b),
while statement (1) can be considered implausible on the basis of (a) and (b). In
the present form, however, (2) is too general. As presented above in Section 3.1.3,
Juge (2006) claims that (3) the Catalan GO-past comes from the construction
“GO-verb conjugated in the preterit + infinitive”. This statement can be regarded
as a special case of (2). Moreover, this latter hypothesis can be supported by cross-
linguistic analogy with the Tucano language which shows a similar development.
When Bybee et al. (1994) examine possible paths for the formation of perfectives
and past tenses, they mention the completive aspect as one of the possible sources.
They define the meaning of completives as “to do something thoroughly and com-
pletely”, but they point out further semantic nuances and other uses completives
may have, among them that “the action is reported with some emphasis or sur-
prise value”. They report the ‘go’ completive in the Tucano language as one which
resembles “the English ‘went and did it’ construction: He went and told her the
whole story, […] which is usually used in the past to emphasise the deliberateness
and finality of an action” (Bybee et al. 1994: 57). The case of the Catalan “anar +
infinitive” is very similar: it also seems to refer to a completed action, and, at the
 Katalin Nagy C.

same time, it expresses emphasis. Thus, the development of this construction can
be supposed to follow the path described in Bybee et al. (1994: 104) as “that of
resultative or completive leading to anterior and then to perfective or simple past”.
That is, completives are a possible source for perfective aspect and then past mark-
ers, and the history of the Catalan “anar + infinitive” seems to instantiate this
process. Such an account is also in accordance with a historical pragmatic analysis
(see Nagy C. 2010).
Our initial question in the present case study was as follows: from the two
documented variants of the “anar + infinitive” construction, namely, (i) “anar ‘go/
auxiliar’ conjugated in the present + infinitive” and, (ii) “anar ‘go/auxiliar’ con-
jugated in the perfective past + infinitive”, which one should be considered the
ancestor of the perfective past of modern Catalan? Now, we can answer that it is
plausible that it is the second one. This decision favors Juge’s (2006) account and
makes Colon’s (1978a, b) and Detges’s (2004) claims concerning the morphologi-
cal aspect of the development of “anar ‘go’ + infinitive’ implausible. Juge’s (2006)
hypotheses deserve a high plausibility value primarily due to its complexity: his
argumentation relies widely on frequency, analogy and crosslinguistic similarities,
and the development he describes fits general tendencies of semantic and analogi-
cal change.

5.  Conclusions

In the present chapter I have discussed how metalinguistic considerations can

help the decision between competing hypotheses in linguistic research. I have pre-
sented three approaches concerning the morphological aspect of the history of the
Catalan “anar ‘go’ + infinitive” construction: those provided in Colon (1979a, b),
in Detges (2004) and in Juge (2006). I would like to highlight again that I have not
presented these approaches in their entirety. I have only taken into consideration
claims which concern the medieval present–past alternation in the auxiliary of the
construction. Even if some imperfections in their argumentations can be detected,
these authors all provide very valuable findings on different aspects of the his-
tory of this Catalan GO-periphrasis. The three accounts considered have formed a
p-context in which some statements and their negations concerning the subques-
tion discussed in the case study were plausible simultaneously, that is, our starting
p-context has been p-inconsistent. This p-inconsistency has been eliminated by
the extension of the starting p-context and, then, the coordination of the extended
p-context. We have re-evaluated and compared the plausibility values of different
statements present in the p-context, also taking into account some widely accepted
argumention methods in historical linguistics.
 Methods and argumentation in historical linguistics 

As regards the subquestion discussed in the present case study, the main prob-
lem with the historical present accounts has been exactly what Fischer (2007: 18)
describes in the following way: “There is a natural tendency to interpret an older
construction very much from the point of view of the modern system. This hap-
pens especially when the form of the construction has remained more or less the
same.” It seems that the formal similarity between the forms of the current para-
digm of “anar + infinitive” (which have the auxiliary in the present tense) and
the medieval present auxiliary forms of “anar + infinitive” resulted in a superfi-
cial analysis in this respect and distracted attention from preterit auxiliary occur-
rences of the medieval “anar + infinitive”. This has yielded the omission of a more
thorough morphological analysis, which appears only later, in Juge (2006). The
development of the Catalan “anar ‘go’ + infinitive” construction has found a more
satisfactory explanation under this latter proposal.
Relying on the considerations presented in this chapter, we can conclude the
following. In historical linguistics we should assign a higher plausibility value to
a statement if:
1. sources traditionally accepted in historical linguistics (e.g. historical corpora,
a properly reconstructed language system, etc.) are used in the research,
2. research methods and argumentation techniques traditionally accepted in
historical linguistics (e.g. analogy-based argumentation, frequency counting,
etc.) are used in the research,
3. the analysis is supported by, or at least does not contradict, accounts of other
linguistic levels,
4. it is supported by or at least does not contradict universal historical linguistics
tendencies detected in previous, reliable research.
It is important to note that the emphasis frequency counting has been given in the
present chapter does not mean that the corpus method and quantitative data would
be the only ones used in historical linguistics. The methods discussed in the pres-
ent chapter do not cover all the methodological tools used in historical linguistics.
In their introductory study to a volume of selected articles on historical pragmatics
methodology, Fitzmaurice and Taavitsainen (2007: 17) point out that “it is not clear
that corpus methodology is applicable to purely pragmatic research questions” and
that “corpus-based methods can be barriers to the investigation of pragmatic phe-
nomena like conversational implicatures, which are neither routinely nor conven-
tionally realised in lexical grammatical expressions” (­Fitzmaurice & Taavitsainen
2007: 18). To put it differently, in historical pragmatics studies qualitative methods
also have a great significance (cf. Navarro 2008). However, discussion of the limits
which the corpus method sets for the research of some, primarily historical, prag-
matics phenomena is beyond the scope of the present chapter.
 Katalin Nagy C.

Acknowledgements

Work on the present chapter was supported by the MTA-DE Research Group for
Theoretical Linguistics and the TÁMOP-4.2.2/B-10/1-2010-0024 project, which
is co-financed by the European Union and the European Social Fund. I am grate-
ful to Enikő Németh T. for valuable discussions about several topics which are
touched upon in this chapter, and to András Kertész and Csilla Rákosi for their
useful comments on earlier versions of the paper. I wish to thank Anna Fenyvesi
for improving my English. I alone am responsible for any remaining errors.

Historical sources

Desc: Desclot, Bernat: Crònica (Chronicle). 1283–88. Barcelona: Edicions 62. 1990.
Munt: Muntaner, Ramón: Crònica (Chronicle). Ca. 1325. Barcelona: Edicions 62. 1991. For the
English translation of the examples taken from the Chronicle of Muntaner I have mainly
relied on the following work: Goodenough, Lady, The Chronicle of Muntaner. 〈http://www.
yorku.ca/inpar/muntaner_goodenough.pdf〉

References

Alvar, Manuel & Pottier, Bernard. 1987[1983]. Morfología histórica del español (Historical mor-
phology of Spanish). Madrid: Gredos.
Badia i Margarit, Antoni M. 1981. Gramàtica històrica catalana (Catalan historical grammar).
València: Tres i Quatre.
Bybee, Joan. 2005[2003]. Mechanisms of change in grammaticalization. The role of frequency. In
The Handbook of Historical Linguistics, Brian D. Joseph & Richard D. Janda (eds), 602–623.
Oxford: Blackwell.
Bybee, Joan. 2005. From Usage to Grammar: The Mind’s Response to Repetition. LSA Presidential
Address, January 8, 2005. 〈http://www.unm.edu/~¡bybee/Bybee%20plenary.pdf〉
Bybee, Joan. 2007. Diachronic Linguistics. In The Oxford Handbook of Cognitive Linguistics,
Dirk Geeraerts & Hubert Cuyckens (eds), 945–987. Oxford: OUP.
Bybee, Joan, Perkins, Revere & Pagliuca, William. 1994. The Evolution of Grammar: Tense, Aspect,
and Modality in the Languages of the World. Chicago IL: The University of Chicago Press.
Bynon, Theodora. 1983. Historical Linguistics. Cambridge: CUP.
Colon, Germà. 1978a[1959]. El perfet perifràstic català “va + infinitiu”. (The Catalan periphras-
tic perfect “va + infinitive”). In La llengua catalana en els seus textos (The Catalan language
of texts), Vol. 2, Germà Colon (ed.), 119–130. Barcelona: Curial.
Colon, Germà. 1978b[1976]. Sobre el perfet perifràstic ‘vado + infinitiu’, en català, en proven-
çal i en francès (On the periphrastic perfect ‘vado + infinitive’ in Catalan, Provençal and,
French). In La llengua catalana en els seus textos (The Catalan language of texts), Vol. 2,
Germà Colon (ed.), 131–174. Barcelona: Curial.
 Methods and argumentation in historical linguistics 

Detges, Ulrich. 2004. How cognitive is grammaticalization? The history of the Catalan perfet
perifràstic. In Up and Down the Cline – The Nature of Grammaticalization [Typological
Studies in Language 59], Olga Fischer, Muriel Norde & Harry Perridon (eds), 211–227.
Amsterdam: John Benjamins.
Dér, Csilla Ilona. 2005. Grammatikalizációs folyamatok a magyar nyelvben – Elméleti kérdések
és esettanulmányok (Processes of grammaticalisation in Hungarian: Theoretical issues and
case studies). Ph.D. dissertation, Eötvös Loránd University, Budapest.
Fischer, Olga. 2004. What counts as evidence in historical linguistics? Studies in Language 28(3):
710–740.
Fischer, Olga. 2007. Morphosyntactic Change: Functional and Formal Perspectives. Oxford: OUP.
Fitzmaurice, Susan & Taavitsainen, Irma. 2007. Historical pragmatics: What it is and how to
do it. In Methods in Historical Pragmatics [Topics in English Linguistics 52], Susan M.
Fitzmaurice & Irma Taavitsainen (eds), 11–36. Berlin: Mouton de Gruyter.
Fludernik, Monika. 1992. The historical present tense in English literature. An oral pattern and
its literary adaptation. 〈http://freidok.uni-freiburg.de/volltexte/5343/pdf/Fludernik_The_
historical_present_tense.pdf〉
Gougenheim, Georges. 1971[1929]. Étude sur les périphrases verbales de la langue française
(A study on the verbal periphrases of French). Paris: Nizet.
Heine, Bernd. 2005[2003]. Grammaticalization. In The Handbook of Historical Linguistics, Brian
D. Joseph & Richard D. Janda (eds), 575–601. Oxford: Blackwell.
Heine, Bernd & Kuteva, Tania. 2002. World Lexicon of Grammaticalization. Cambridge: CUP.
Hock, Hans Henrich. 1991. Principles of Historical Linguistics. Berlin: Mouton de Gruyter.
Hock, Hans Henrich. 2005[2003]. Analogical change. In The Handbook of Historical Linguistics,
Brian D. Joseph & Richard D. Janda (eds), 441–460. Oxford: Blackwell.
Jeffers, Robert J. & Lehiste, Ilse. 1979. Principles and Methods for Historical Linguistics. ­Cambridge
MA: The MIT Press.
Juge, Matthew L. 2006. Morphological factors in the grammaticalization of the Catalan “go”
past. Diachronica 23(2): 313–339.
Kertész, András & Rákosi, Csilla. 2012. Data and Evidence in Linguistics. A Plausible Argumenta-
tion Model. Cambridge: CUP.
Kertész, András & Rákosi, Csilla. This volume a. Introduction: The state of the art and the struc-
ture of the book.
Kertész, András & Rákosi, Csilla. This volume b. The p-model of data and evidence in linguistics.
Ladányi, Mária. 1998. Jelentésváltozás és grammatikalizáció – kognitív és szerves nyelvelméleti
keretben (Meaning change and grammaticalization in a cognitive and organic theoretical
linguistic context). Magyar Nyelv 94: 407–423.
Nagy C., Katalin. 2010. The pragmatics of grammaticalization: The role of implicatures in
semantic change. Journal of Historical Pragmatics 11(1): 67–95.
Navarro, Federico. 2008. Análisis histórico del discurso. Hacia un enfoque histórico-discursivo
en el estudio diacrónico de la lengua (Historical discourse analysis. Toward a historical
discursive approach in the diachronic study of the language). In El valor de la diversidad
[meta]lingüística. Actas del VIII Congreso de Lingüística General (The value of the [meta]
linguistic diversity. Proceedings of the 7th Congress on General Linguistics), Antonio
Moreno Sandoval (ed.). Madrid: Universidad Autónoma de Madrid. 〈http://www.lllf.uam.
es/clg8/actas/pdf/paperCLG85.pdf〉
 Katalin Nagy C.

Pinkster, Harm. 1990. Is the Latin present tense the unmarked, neutral tense in the system? In
Latin in Use [Amsterdam Studies in the Pragmatics of Latin], Rodie Risselada (ed.), 63–83.
Amsterdam: Gieben.
Schiffrin, Deborah. 1981. Tense variation in narrative. Language 57(1): 45–62.
Silva-Corvalán, Carmen. 1983. Tense and aspect in oral Spanish narrative: context and mean-
ing. Language 59(4): 760–780.
Wolfson, Nessa. 1979. The conversational historical present alternation. Language 55(1):
168–182.
 chapter 5

Hungarian verbs of natural phenomena with

explicit and implicit subject arguments
Their use and occurrence in the light of data

Enikő Németh T.
University of Szeged, Dept. of General Linguistics, MTA-DE Research
Group for Theoretical Linguistics

The present chapter has two main aims. First, it attempts to provide a novel
account for the occurrence of Hungarian verbs of natural phenomena with
subject arguments. The problem which the present chapter wants to resolve
is that there is an inconsistency between the previous explanations of the
behaviour of verbs of natural phenomena with and without an explicit subject
and the data. The previous approaches have considered the Hungarian verbs of
natural phenomena subjectless, although data from various direct sources (such
as written corpora, intuition, introspection, and spoken discourses) testify to
occurrences of verbs of natural phenomena with explicit subjects. To eliminate
this inconsistency, the chapter argues for a new, unified approach, according to
which Hungarian verbs of natural phenomena have subject arguments in their
lexical-semantic representations which can (or even should) be lexically unrealised
in some contexts, while in others they can be explicitly expressed and contextual,
grammatical as well as lexical-semantic factors together license the implicit or
explicit occurrence of subject arguments. Second, applying Kertész & Rákosi’s
(2012, this volume) p-model, the chapter aims to reflect on a metalinguistic level
on how the proposed explanation has been reached in the course of the research
as well as how the new account can resolve the above mentioned problem.

1.  Introduction: Aims and the organisation of the chapter

In the present chapter I have two main aims. First, taking into account a wide
spectrum of data as well as relying on plausible inferences (cf. Kertész & Rákosi
2012: 56–128, this volume: Section 2.3) built on these data, I attempt to provide a
 Enikő Németh T.

novel explanation for the occurrence of Hungarian verbs of natural phenomena

with explicit and implicit subject arguments. The problem which I want to resolve
is that Hungarian verbs of natural phenomena have been considered subjectless
by widely accepted Hungarian traditional descriptive and theoretical grammars
up until the present (the only exception to this tradition is Tóth 2001), although
there is a considerable amount of data from different direct sources (Rescher 1976;
­Kertész & Rákosi 2012: 67, this volume: Section 2.2; Nagy C. 2008; Németh T. 2010)
such as written corpora, intuition, introspection and spoken discourses which tes-
tifies to occurrences of verbs of natural phenomena with explicit subjects. To elim-
inate the inconsistency between the previous explanations and the data regarding
the occurrences of Hungarian verbs of natural phenomena with explicit subjects I
want to argue for a new approach, according to which Hungarian verbs of natural
phenomena indeed have subject arguments in their lexical-semantic representa-
tions which can (or even should) be left implicit in certain circumstances, while
in other circumstances they can be lexically realised. Assuming such a lexical-
semantic representation as well as taking into consideration the contextual factors,
the occurrences of ­Hungarian verbs of natural phenomena with explicit subjects
and without them can be accounted for in a unified way.
Second, relying on Kertész and Rákosi’s (2012, this volume) model of
­plausible argumentation (henceforth p-model) – which they advance as a heu-
ristic tool for problem solving – I intend to reflect on a metalinguistic level on
how I have reached the proposed explanation in the course of the research as
well as how the new account can resolve the above mentioned problem. Since
my ­second aim is necessarily related to achieving my first aim, I attempt to
build the metatheoretical issues into the presentation of the object theoretical
discussion regarding the subjectlessness of the verbs of natural phenomena in
­Hungarian – i.e. I will present the object theoretical and metatheoretical issues
in their interrelationship.
The organisation of the chapter is as follows. In Section 2, I will briefly
review the previous approaches to the occurrence of Hungarian verbs of nat-
ural phenomena without or with a subject argument. Among the rivalling
descriptions and theories, Magyar grammatika (Hungarian Grammar) (­Keszler
2000) and the l­exical-functional grammar (Komlósy 1994, 2001) agree that
­Hungarian verbs of natural phenomena are subjectless, while Tóth’s (2001) gen-
erative syntactic analysis questions this claim, assuming that verbs of natural
phenomena have a quasi-argumental subject. These approaches serve as the
starting p­ -context for my research (Kertész & Rákosi 2012: 121ff., this volume:
Section 2.4). The ­starting p-context for my research is p-problematic, because
it is i­nformationally overdetermined, i.e. it contains too much information
 Hungarian verbs of natural phenomena with explicit and implicit subject arguments 

in the sense that it involves statements that are plausible according to some
source but cannot be true ­simultaneously (Kertész & Rákosi 2012: 130ff., this
volume: Section 2.5). The overdetermination of the ­starting p-context results
from incompatibility between, on the one hand, the proposals suggesting
the subjectlessness of verbs of natural phenomena, i.e. Magyar grammatika
[Hungarian Grammar] (Keszler 2000) as well as a lexical-functional grammar
(Komlósy 1994, 2001), and, on the other hand, a generative syntactic analy-
sis (Tóth 2001), which considers the verbs at stake to have a quasi-argumental
subject. Consequently, the starting p-context is p-inconsistent. In Section 3,
I will present a range of data capturing the characteristics of verbs of natu-
ral phenomena in Hungarian which provide evidence for the hypothesis con-
cerning the occurrence of verbs of natural phenomena with explicit subjects,
i.e. I will extend the starting p-context with data which were neglected earlier
and confront these data with the approaches in the starting p-context. By this
­coordination of the extended p-context (cf. Kertész & Rákosi 2012: 136–143,
this volume: ­Subsection 2.6.1), I will show that the uses of verbs of natural
­phenomena with explicit subjects cannot be accounted for by the approaches
in the extended p-context either. Consequently, the extended and coordinated
p-context is more p-inconsistent: there is p-­inconsistency inherited from the
starting p-context between the rivalling approaches and there is a new incon-
sistency in the extended p-context between the previous approaches and the
data newly ­introduced into the p-context. In Section 4, I will continue to extend
the p-context with the results of my previous research into implicit arguments
in Hungarian. In ­Section 5, I will modify the previous p-context (Kertész &
Rákosi 2012: Section 10.4, this volume: Subsection 2.6.1). Starting out from my
results regarding the verbs’ occurrence with implicit arguments in Hungarian
as initial hypotheses to explain the subjectless use of verbs of natural phenom-
ena, I will propose a new kind of lexical-semantic representation for the verbs
of natural phenomena, on the basis of which the occurrence of the verbs of
natural phenomena both without a subject and with a subject can be accounted
for in a unified way. By applying this kind of lexical-semantic representation as
well as taking into consideration contextual factors, the p-inconsistency can be
eliminated in the modified context at the end of the argumentation cycle. And,
finally, in Section 6, I will summarise my results and formulate my conclusions,
indicating why the suggested explanation is more plausible than the previous
approaches in the starting p-context, or, in other words, why my account for
the occurrence of Hungarian verbs of natural phenomena can be regarded as
the resolution of the starting p-problem (cf. Kertész & Rákosi 2012: 133f., this
volume: Section 2.5).
 Enikő Németh T.

2.  Th
 e rivalling approaches in the starting p-context: On the
subjectlessness of verbs of natural phenomena in Hungarian

2.1  M
 agyar értelmező kéziszótár (Concise Explanatory Dictionary
of Hungarian) (Pusztai 2003)
The majority of verbs of natural phenomena, e.g. esteledik ‘[for evening to] close
in’, alkonyodik ‘[for dusk to] set in’, sötétedik ‘[for dark to] grow’, hajnalodik ‘[for
day to] break, dawn’, virrad ‘dawn’, tavaszodik ‘[for spring to] come’, fagy ‘freeze’,
olvad ‘thaw’, villámlik ‘[for lightning to] strike’, havazik ‘snow’ are evaluated as
subjectless in the Magyar értelmező kéziszótár (Concise Explanatory Dictionary
of Hungarian) (Pusztai 2003). The feature of subjectlessness is assigned not only
to the above mentioned verbs but also to those which can be used with explicit
subjects quite easily, e.g. dereng ‘dawn’ – dereng az ég/a hajnal ‘the sky/the dawn
gets lighter’, zuhog ‘pour’ – zuhog az eső ‘the rain pours’, szakad ‘pour’ – szakad
az eső ‘the rain pours’. Furthermore, while items like the verbs borul ‘cloud over’,
dörög ‘thunder’, sötétedik ‘[for dark to] grow’ and be-sötétedik ‘pvb1-[for dark to]
grow’ contain the evaluation “subjectless”, verbs such as be-felhősödik ‘pvb-cloud
over’ and tisztul ‘clear up’ are not considered subjectless in the concise explanatory
dictionary although both groups of verbs can occur with the same explicit subject,
az ég ‘the sky’. The verbs ki-tisztul ‘pvb-clear up’ and pirkad ‘dawn’ are evaluated as
“also subjectless” in the dictionary. In the case of verbs with meanings e­ xpressing
the way the rain falls, the lexical items of szemerkél ‘drizzle’, csepereg ‘sprinkle’,
­csorog ‘trickle’, ömlik ‘pour’ do not include the evaluation “subjectless’, while the
verbs zuhog ‘pour’ and szakad ‘pour’ are considered subjectless, although all of
these verbs can occur with an explicit subject, az eső ‘the rain’. And, finally, among
verbs with meanings concerning the wind such as fúj ‘blow’, fújdogál ‘blow gently’,
fuvall ‘blow gently’, lengedezik ‘blow gently’, süvít ‘bluster’, fütyül ‘whistle’, only the
verb fúj ‘blow’ is evaluated as “also subjectless” although all of the named verbs can
be used without an explicit subject.
In the lexical entries of the above mentioned verbs of natural phenomena
there is no explicit clarification of the notion of subjectlessness and because of the
confusing evaluations regarding the subjectlessness of verbs of natural phenom-
ena there is no possibility to reconstruct this notion on the basis of the diction-
ary’s evaluations. Consequently, on the basis of the Magyar értelmező kéziszótár
­(Concise Explanatory Dictionary of Hungarian) one cannot make a plausible
statement about subjectlessness of Hungarian verbs of natural phenomena, i.e. it

.  The abbreviation pvb is used for preverb.

 Hungarian verbs of natural phenomena with explicit and implicit subject arguments 

cannot be considered a suitable source (Kertész & Rákosi 2012: 64, this volume:
Section 2.2).

2.2  Magyar grammatika (Hungarian Grammar) (Keszler 2000)

Magyar grammatika (Hungarian Grammar) (Keszler 2000: 85, 90f., 409) divides
Hungarian verbs of natural phenomena into two groups. The first group ­contains
verbs which are considered subjectless by the grammar, e.g. hajnalodik ‘[for day
to] break’, pirkad ‘dawn’, alkonyodik ‘[for evening to] close in’, tavaszodik ‘[for
spring to] come’, havazik ‘snow’, villámlik ‘[for lightning to] strike’, esteledik ‘[for
evening to] close in’. The structure of data concerning verbs in this group can be
reconstructed, for example, as follows:
The verb hajnalodik ‘[for day to] break’ is a verb of natural phenomena and is
subjectless.

The second group involves verbs with an optional explicit subject, e.g. esik az eső
‘the rain falls’, szemerkél az eső ‘the rain drizzles’, dörög az ég ‘the sky thunders’,
be-borul az ég ‘the sky [pvb-]clouds over’, ki-derül az ég ‘the sky [pvb-]clears up’,
sötétedik az ég ‘the sky darkens’, ki-világosodik az ég ‘the sky [pvb-]clears up’. The
structure of data concerning verbs in this group can be reconstructed, for example,
as follows:
The verb esik ‘[for precipitation to] fall’ is a verb of natural phenomena and is not
subjectless.

According to Keszler (2000), the filling of the subject argument position is not
permitted in the first group of verbs. Keszler (2000) argues that meanings of verbs
in the first group concern elements of the general natural environment; therefore,
there is no need or interest to name them explicitly. Furthermore, in some cases
the subject is missing because it is incorporated into the stem of the verb of natural
phenomena. However, it is not clear how to interpret the constraint of not filling
the subject argument position: either the verbs in the first group do not have a
subject argument position at all or they do have such a position but its filling is
prohibited. In the first case, the structure of data concerning verbs of natural phe-
nomena in the first group can also be reconstructed, for example, as follows:
The verb hajnalodik ‘[for day to] break’ is a verb of natural phenomena and it does
not have a subject argument position.

And, in the second case:

The verb hajnalodik ‘[for day to] break’ is a verb of natural phenomena, it has
a subject argument position but the filling of subject argument position is
prohibited.
 Enikő Németh T.

It must be noted that Keszler (2000) emphasises that verbs of natural phenomena
considered subjectless can and even must occur with explicit subjects in their
metaphorical uses, cf. e.g. havazik az élet ‘life is snowing’, az egész világ havazik
‘the whole world is snowing’, csend havazik ‘silence is snowing’, a szeme villám-
lik ‘his/her eyes flash’, villámlik a fekete nappal ‘the black daytime flashes [like
lightening]’.2 If verbs of natural phenomena are subjectless, i.e. one selects the
data structure first proposed in the previous paragraph, it must be explained how
a subject argument position arises in the metaphorical use. If one assumes that
these verbs always have a subject argument position, i.e. their data structure can
be reconstructed along the lines of the second proposal in the previous para-
graph, which must not be filled in non-metaphorical uses, then what constraints
predict the subjectless uses must be accounted for. It is obvious that these two
answers are inconsistent, and this inconsistency cannot be eliminated in Keszler
(2000).3 As for the verbs in the second group, Keszler (2000) allows them to be
used with explicit subjects, i.e. they certainly have a subject argument position.
But this position can be filled optionally, i.e. the verbs in second group can have
an explicit subject argument, and they can also be used without it. According to
this, the structure of data concerning the verbs in this group can also be recon-
structed, for example, as follows:
The verb esik ‘[for precipitation to] fall’ is a verb of natural phenomena and it has
a subject argument position which can be filled optionally.

The p-model argues that data must not be reduced to their information content,
instead, their structure consists of two components: their information content
and their plausibility value. The plausibility value of data stems from their direct
sources (Kertész & Rákosi 2012: 67, this volume: Section 2.2). Now, one can ask

.  These examples of metaphorical use can be found in the Magyar Nemzeti Szövegtár
(­ Hungarian National Corpus) 〈http://corpus.nytud.hu/mnsz〉.
.  Keszler (2000) argues that it is also characteristic of verbs in the first group that they
cannot have other complements either. The only exception mentioned by this grammar is the
verb esteledik ‘[for evening to] close in’, which can have an optional adverbial complement
ránk ‘upon us’. However, on the one hand, this optional adverbial complement can appear with
other verbs which have a meaning expressing the passing of days and seasons as well, e.g. ránk
alkonyodik ‘evening is closing in on us’, ránk hajnalodik ‘day is breaking on us’, ránk sötétedik
‘it is getting dark on us’, ránk virrad ‘dawn is breaking on us’, and ránk tavaszodik ‘spring is
breaking on us’. And, on the other hand, the adverbial complement in question can occur with
this group of verbs of natural phenomena in other forms as well, cf. Jól rátok e­ steledett ‘Evening
has really closed in on you’, Addig dolgoztak, míg rájuk virradt ‘They worked until dawn broke
on them’. I want to thank Anna Fenyvesi for drawing my attention to this p ­ henomenon and
also these two examples.
 Hungarian verbs of natural phenomena with explicit and implicit subject arguments 

what plausibility value can be assigned to the above reconstructed data? If one can
accept Magyar grammatika (Hungarian Grammar) as a relatively reliable source
on an authoritative basis, then the plausibility value of the above reconstructed
data is relatively high. But if one decomposes and analyses the information con-
tent of the reconstructed data, as has been done in the present article, one can find
inconsistencies. Because of these inconsistencies Magyar grammatika (Hungarian
Grammar) cannot be considered a reliable or even a usable source.

2.3  Lexical-functional grammar (Komlósy 1994)

In a lexical-functional approach, Komlósy (1994) also considers Hungarian
weather verbs4 to be subjectless. Here the lexical-semantic representation of
weather verbs contains an internal argument which cannot be expressed syntacti-
cally. Thus, weather verbs are not regents, they cannot occur with explicit subjects,
and they do not allow even expletive, formal subjects. Komlósy (2001) also empha-
sises that weather verbs are not predicates either. They are lexical items which do
not require any semantic arguments. Relying on the requirements of g­ rammatical
interpretability, semantic arguments cannot include internal ­arguments which
form a ­necessary, inherent part in the meaning of lexical items of weather verbs;
therefore, they cannot be explicitly realised (cf. lexical-semantic representations
in (1)(a)–(c)). Since weather verbs have only internal arguments and do not have
semantic arguments at all, they are neither regents nor predicates. Komlósy (2001)
considers Hungarian weather verbs to be similar to R ­ ussian weather verbs, which
do not have a controllable subject relation either. Both Hungarian and ­Russian
weather verbs differ from English weather verbs, which do require a formal
­subject, i.e. they are regents (cf. rain, snow). However, it must be noted that the
meaningless expletive subject it in English is not a semantic argument either, i.e.
rain and snow are not predicates, and their syntactic subject satisfies only formal
grammatical requirements.
Consider Komlósy’s (1994: 160f.) examples in (1)(a)–(c).
(1) a. Hajnal-od-ik.
dawn-sfx-prs.indf.3sg5
‘dawn is coming’ → ‘arise (dawn)’

.  The term ‘weather verb’ also occurs in the literature, but I prefer using the term ‘verb of
natural phenomena’ because of its wider scope. Komlósy (1994, 2001) and Tóth (2001) (cf.
Section 2.4) use the term ‘weather verb’.
.  The abbreviations used in the glosses throughout this chapter are the following: 1 – first
person, 2 – second person, 3 – third person, sg – singular, pl – plural, nom – nominative,
acc – accusative, gen − genitive, indf – indefinite (conjugation), ine – inessive, def – definite
 Enikő Németh T.

b. Villám-l-ott.
lightning-sfx-pst.indf.3sg
‘there was lightning’ → ‘appear (lightning)’
c. Sötét-ed-ik.
dark-sfx-prs.indf.3sg
‘it is getting dark’ → ‘arise (darkness)’

In (1)(a)–(c) weather verbs hajnalodik and villámlik have a morphologically com-

plex root as do most weather verbs in Hungarian.6 Their root is derived from
the nouns hajnal ‘dawn’ and villám ‘lightning’, respectively. The stem of the verb
­sötétedik has been derived from the adjective sötét ‘dark’.7 The meaning of the
­nominal or adjectival stem can be factored out as an internal argument in the
­lexical-semantic representation of the verb derived from the stem at stake, which
is why the proposition denoted by the weather verbs may be given a motivated
function-argument structure semantically. According to Komlósy’s (1994: 161)
proposal, the relation between the nominal or adjectival stems and affixes in the
case of morphologically complex weather verbs can be thought of as a relation
between a subject and a predicate. The subject slot of the affixes is filled in with
the nominal or adjectival stems; consequently, the resulting verbs ought to be
subjectless.
Although Komlósy (1994) attempts to provide an explicit explanation of the
subjectlessness of weather verbs in Hungarian, there are questions which it cannot
answer. First, let us reconstruct the data structure in Komlósy (1994) concerning
the verbs villámlik ‘[for lightning to] strike’ and sötétedik ‘darken’. On the basis of
Komlósy (1994) as a direct source, the following statement is plausible:
The verbs villámlik and sötétedik are weather verbs which cannot occur with an
explicit subject and do not license a formal subject either.

From this one can infer the following plausible statement:

Expressions villámlik az ég ‘the sky is lightning’ and sötétedik az ég ‘the sky is
getting dark’ are ill-formed.

(conjugation), sfx – suffix, sup – superessive, prs – present tense, pst – past tense, and pvb –
preverb.
.  However, the morphological complexity is not considered a definitive property of the
Hungarian weather verbs by Komlósy (1994).
.  Komlósy (1994: 160) argues that adjectival stems might be conceived as nouns in the
roots of weather verbs based on the morphologically unmarked adjective→noun derivation
in Hungarian.
 Hungarian verbs of natural phenomena with explicit and implicit subject arguments 

However, according to my linguistic intuition this statement is implausible. My

intuition as a direct source supports the following data:

Villámlik az ég ‘the sky is lightning’ and sötétedik az ég ‘the sky is getting dark’ are
well-formed expressions in which villámlik as well as sötétedik are weather verbs
and ég is an explicit subject.

If these weather verbs are subjectless, i.e. they do not have a subject slot in their
lexical-semantic representation, it should not be possible to use them with overt,
syntactically realised subjects in metaphorical (cf. a szeme villámlik ‘his/her eyes
are flashing’, villámlik a fekete nappal ‘the black daytime flashes [like lightning]’)
and non-metaphorical senses (cf. villámlik az ég ‘the sky is lightning’ sötétedik az
ég ‘the sky is getting dark’ in Section 2.1 and 2.2). If we take into consideration
these latter data as well and extend and coordinate the p-context of the lexical-
functional approach to Hungarian weather verbs, i.e. confront these data with
­Komlósy’s (1994, 2001) proposals, Komlósy’s lexical-functional approach to
­Hungarian weather verbs becomes p-inconsistent.

2.4  A generative syntactic analysis (Tóth 2001)

The generative grammatical analysis proposed by Tóth (2001) does not accept
the views of lexical-functional grammar that weather verbs are subjectless in
­Hungarian and 3sg agreement in the verbs is taken to be the morphological reflex
of default agreement. On the basis of theoretical considerations as well as relying
on the various occurrences of weather verbs with subjects in other languages, Tóth
(2001) argues that in null subject languages weather verbs are not subjectless but
instead have null-quasi-argumental subjects, their 3sg agreement is real agree-
ment, and they can take part in control relations. Strong evidence (cf. Kertész &
Rákosi 2012: 181, this volume: Section 3.2) for the non-subjectlessness of weather
verbs in Hungarian is provided by their behaviour in participial adjunct clauses.
Participial adjunct clauses either contain va-participles or ván-participles. Tóth
(2001: 63) emphasises that although these two forms have a number of common
syntactic properties, there is an important difference between them. Va-participles
cannot have a lexical subject due to lack of case, while a ván-participle can have a
nominative marked DP or a pronominal subject, cf. (2)–(3).

(2) János-Ø meg-érkez-vén/∗ve, pro

János-nom.sg pvb-arrive-VÁN/-VA
el-ind-ul-t-unk haza-felé.
pvb-start-sfx-pst-indf.1pl home-towards
‘János having arrived, we started for home.’
 Enikő Németh T.

(3) Hajnal-od-ván/∗va, pro el-indul-t-unk haza-felé.

dawn-sfx-VÁN/-VA pvb-start-sfx-pst-indf.1pl home-towards
‘Dawn coming, we started for home.’

The utterance in (3) makes it possible to form the following plausible statement:
The va-participle *hajnalodva formed from the weather verb hajnalodik ‘dawn’ is
ill-formed, but the ván-participle form of it, hajnalodván is well-formed.

According to Tóth’s intuition as a direct source, hajnalodik can grammatically be

combined with -ván, and only with -ván. This contrast is unpredictable and inex-
plicable for the approaches in which weather verbs cannot have a subject but is
accommodated under Tóth’s (2001: 63) quasi-argumental subject analysis, accord-
ing to which ván-participles assign nominative case to the quasi-argumental pro
and, thus, make it licit, while va-participles cannot license pro because no case is
available. Consequently, since weather verbs select -ván constructions, i.e. they
can be combined with -ván, resulting in well-formed utterances, the plausible
statement about (3) serves as strong evidence against the hypothesis that weather
verbs are syntactically subjectless in the context of Tóth’s (2001) generative syntac-
tic analysis.8
It must be noted that some native speakers of Hungarian cannot accept or
hardly accept weather verbs in -ván constructions. According to their intuition the
following plausibility statement can be formed:
Ván-participle forms from the weather verbs are ill-formed.

Sárik (1998: 426) also shares this opinion. To consider ván-participle forms

formed from weather verbs ill-formed contradicts Tóth’s (2001) argumentation,
i.e. data concerning the ill-formedness of the utterances containing ván-­participle
­constructions formed from the weather verbs causes further p-inconsistency
in the p-context. However, my own intuition as well as the intuition of other
­Hungarian native speakers supports Tóth’s intuition. Since the statement about
the well-formedness of ván-constructions formed from weather verbs is plausible
while the statement about ill-formedness of ván-constructions formed from the
weather verbs is implausible for me and other native speakers, I will rely on the
former statement in the subsequent argumentation.
If we continue thinking in the way suggested by Tóth (2001) and also reverse
the direction of the argumentation in lexical-functional grammar concerning
the relationship between internal arguments, semantic arguments and syntac-
tic complements, the following statements can be made. If weather verbs have

.  For the verbs’ behaviour in participial adjunct clauses cf. also Sárik (1998).
 Hungarian verbs of natural phenomena with explicit and implicit subject arguments 

a ­syntactic subject, it cannot be an expression of an internal argument, because

internal arguments cannot be syntactically realised. Syntactic subjects of weather
verbs cannot be considered realisations of semantic arguments either. There are
cases when lexical items require grammatically syntactic complements which are
not expressions of semantic arguments (cf. rain and its expletive formal subject it:
It is raining). However, the comparison of (2) and (3) convincingly testifies that
Hungarian weather verbs occur with syntactic subjects which are not empty exple-
tives. Therefore, the conclusion of Tóth’s (2001) argumentation that weather verbs
take a quasi-argumental subject is more plausible than the conclusion of previous
approaches according to which weather verbs are subjectless in Hungarian.

2.5  The evaluation of the starting p-context

The thorough investigation of the rivalling approaches in the previous
­subsections reveals that the starting p-context is p-problematic, it contains
p-incompleteness and p-inconsistencies (cf. Kertész & Rákosi this volume:
Section 2.5, 2012: 130ff.). First, the Magyar értelmező kéziszótár (Concise
Explanatory Dictionary of Hungarian) (Pusztai 2003) and Magyar grammatika
(Hungarian Grammar) (Keszler 2000) have inconsistent evaluations regarding
the occurrence of Hungarian verbs of natural phenomena both without and with
an explicit subject (cf. Sections 2.1 and 2.2). Komlósy’s (1994, 2001) proposal in
lexical-functional grammar is also p-problematic because of its incompleteness
(cf. Section 2.3). Second, there is an inconsistency between, on the one hand,
the Magyar értelmező kéziszótár (Concise Explanatory Dictionary of ­Hungarian)
(Pusztai 2003), Magyar grammatika (Hungarian Grammar) (Keszler 2000) as
well as Komlósy’s (1994, 2001) lexical-functional grammatical proposal and, on
the other hand, Tóth’s (2001) generative syntactic analysis. The first three con-
sider Hungarian weather verbs subjectless, while the fourth non-subjectless.

3.  Extending the starting p-context with new data

The Magyar értelmező kéziszótár (Concise Explanatory Dictionary of ­Hungarian)

(Pusztai 2003) and Magyar grammatika (Hungarian Grammar) (Keszler 2000)
have collected their examples mainly from various corpora: previous dictionar-
ies, grammars and works of fiction,9 while the lexical-functional grammar and
the generative grammatical analysis through intuition. These approaches all

.  For the problems of classification of direct data sources cf. Németh T. (2010).
 Enikő Németh T.

regard their own sources of examples as unquestionable. On the basis of their

examples, Keszler (2000) and Komlósy (1994, 2001) consider the hypothesis that
the verbs of natural phenomena are subjectless acceptable with a high plausibility
value, while Tóth’s (2001) analysis argues for the opposite hypothesis, accord-
ing to which verbs of natural phenomena are not subjectless. In order to resolve
this contradiction, I have taken into account a new spectrum of data based on
various new direct sources, namely the Magyar Nemzeti Szövegtár (Hungarian
National Corpus) 〈http://corpus.nytud.hu/mnsz〉, my own intuition, and spoken
discourses. The aim of the extension of the starting p-context with new data from
these direct sources is to examine whether we can find evidence for one or the
other hypothesis.
The data regarding the subjectlessness of verbs of natural phenomena referred
to in Sections 2.1 and 2.2 seem to be counterintuitive on the basis of my own
intuition, therefore I consider them implausible. If the Magyar Nemzeti Szövegtár
(Hungarian National Corpus) is also taken into account as a direct data source in
addition to my intuition, it turns out that a great number of occurrences of the
verbs of natural phenomena with explicit subjects can be found, c.f. e.g. (4)(a)–(h).

(4) a. este-l-ed-ik/alkony-od-ik
evening-sfx-sfx-prs.indef.3sg/dusk-sfx-prs.indef.3sg
a nap-Ø
the day-nom.sg
‘the day is coming to the evening / the day is coming to dusk’
b. sötét-ed-ik a látóhatár-Ø/ az ég-Ø
dark-sfx-prs.indef.3sg the horizon-nom.sg/ the sky-nom.sg
‘The horizon/sky is getting dark.’
c. hajnal-od-ik az idő-Ø
dawn-sfx-prs.indf.3sg the time-nom.sg
‘The day is breaking. [lit. The time is coming to dawn.]’
d. virrad a nap-Ø/ az éjszaka-Ø/
dawn.prs.indef.3sg the day-nom.sg/ the night-nom.sg/
a fény-Ø
the light-nom.sg
‘The day/night/light is dawning.’
e. villám-l-ik az ég-Ø
lightning-sfx-prs.indef.3sg the sky-nom.sg
‘The sky is lightning.’
f. zuhog az eső-Ø
pour.prs.indef.3sg the rain-nom.sg
‘Rain is pouring down.’
 Hungarian verbs of natural phenomena with explicit and implicit subject arguments 

g. szakad az eső-Ø
pour.prs.indef.3sg the rain-nom.sg
‘Rain is pouring down.’
h. szakad a hó-Ø
pour.prs.indef.3sg the snow-nom.sg
‘It is snowing heavily.’

Based on the occurrences of verbs of natural phenomena with explicit subjects in

the Magyar Nemzeti Szövegtár [Hungarian National Corpus], a high plausibility
value can be assigned to the following statement:
The verbs esteledik/alkonyodik, sötétedik, hajnalodik, virrad, villámlik, zuhog,
szakad are verbs of natural phenomena and they can occur with explicit subjects.

Of course, one can say that the Hungarian National Corpus is not a reliable source
of data, or the occurrence in a corpus cannot be evaluated as sufficient positive
evidence (Bibok 2009). However, my own intuition, as well as the testing of data by
other native speakers, have made the following statement plausible:
Hungarian verbs of natural phenomena can occur with explicit subjects.

Furthermore, examining everyday spoken discourses, I have also collected occur-

rences of verbs of natural phenomena with explicit subjects (cf. (5)(b), (10)(a) and
(10)(c) in Section 5). Relying on these occurrences with explicit subjects and tak-
ing into consideration my own intuition, I have also assigned a high plausibility
value to the above statement. Consequently, on the basis of data from these various
direct sources, a high plausibility value can be assigned to the hypothesis that:
Hungarian verbs of natural phenomena are not subjectless.

If we also take into account metaphorical uses, then all verbs of natural phenom-
ena can occur with explicit subjects. Let us consider only two verbs – havazik
‘snow’, villámlik ‘[for lightening to] strike’– in their metaphorical uses: havazik
az élet ‘life is snowing’, az egész világ havazik ‘the whole world is snowing’, csend
havazik ‘silence is snowing’, a szeme villámlik ‘his/her eyes are flashing [like light-
ning]’, villámlik a fekete nappal ‘the black daytime is lightning’.
In the light of the data newly introduced into the starting p-context, it can be
seen that verbs of natural phenomena considered subjectless by Keszler (2000) can
occur with explicit subjects. Thus, on the basis of these data one can conclude the
following: (i) verbs of natural phenomena considered subjectless by the H ­ ungarian
grammatical tradition are not subjectless, instead, they can occur with explicit
subjects, and, (ii) Hungarian verbs of natural phenomena cannot be divided into
two groups regarding whether they can occur with explicit subjects or not.
 Enikő Németh T.

If we assume lexical-semantic representations for weather verbs similar to

the ones proposed by Komlósy’s (1994, 2001) lexical-functional approach (cf.
(1)(a)–(c)), i.e. lexical-semantic representations without a subject argument
position, then we cannot account for the uses of weather verbs with explicit
­non-metaphorical and metaphorical subjects either.
However, on the basis of data from the above mentioned direct sources
we can go further even than Tóth’s (2001) quasi-argumental analysis in her
generative syntactic approach. Since the hypothesis that weather verbs can
­
occur with explicit, lexicalised non-metaphorical and metaphorical subjects is
­plausible, their subject can be thought of as a real semantic argument which must
be involved in their lexical-semantic representation, i.e. instead of Tóth’s quasi-
argumental analysis a real-argumental analysis can be suggested which can be
considered more complete and plausible than Tóth’s analysis, since it can account
for the weather verbs’ occurrence with explicit non-metaphorical and metaphori-
cal ­subjects as well.
By the confrontation of new data with the rivalling approaches in the starting
p-context, i.e. by the coordination of the extended p-context (cf. Kertész & Rákosi
2012: Section 10.4, this volume: Subsection 2.6.1), it has been revealed that the use
of verbs of natural phenomena with explicit non-metaphorical and m ­ etaphorical
subjects cannot be accounted for by the approaches in the extended p-context
either. Consequently, the extended and coordinated p-context is also p­ -incomplete.
At the same time the extended and coordinated p-context is p-inconsistent as well.
There is inconsistency inherited from the starting p-context between the rivalling
approaches and there is further inconsistency in the extended p-context between
the previous approaches and the data newly introduced into the p-context. Finally,
at this stage of the argumentation it can be concluded that the hypothesis according
to which Hungarian verbs of natural phenomena are subjectless is less plausible,
and the opposite hypothesis, i.e. that verbs of natural phenomena in Hungarian
are not subjectless, is more plausible.

4.  E
 xtending the p-context with results of previous research into implicit
arguments in Hungarian

4.1  Definition of implicit arguments and their occurrence in Hungarian

As evidenced by my previous research, implicit arguments can be defined as
­arguments involved in the lexical-semantic representations of verbs but which
are lexically unrealised, and whose implicit presence in utterances is attested
by ­lexical-semantic, grammatical (phonological, morphological, syntactic and
 Hungarian verbs of natural phenomena with explicit and implicit subject arguments 

s­ emantic), discourse, and/or pragmatic evidence (cf. e.g. Németh T. 2000, 2001,
2008, 2010; Bibok & Németh T. 2001; Németh T. & Bibok 2010). Such evidence
testifies to the implicit presence of an argument and attests that the particular
utterance does not have a performance error. In Hungarian there are ­various
types of implicit arguments and their occurrence in utterances is licensed by
various factors. The separate mention of these various factors in the above men-
tioned definition of implicit arguments indicates the diverse character of the
different types of implicit arguments. However, although the possibilities of the
occurrence of implicit arguments in Hungarian can be explained by various fac-
tors, their intensive interaction should be emphasised. An argument can be left
implicit in Hungarian in the following three cases (Németh T. 2000, 2001, 2008,
2010; Bibok & Németh T. 2001; Németh T. & Bibok 2010):

A. If some element of the lexical-semantic representation of a verb licenses the

lexically unrealised occurrence of an argument involved in this r­ epresentation,
according to the principle of relevance (cf. Rita eszik ‘Rita is eating’).
B. If the rest of the utterance, i.e. the immediate context in which the argument
occurs provides a relevant, typical interpretation (cf. Ági adott a koldusnak
‘Ági gave [money] to the beggar’).
C. If extending the immediate context of the argument results in an ­interpretation
consistent with the principle of relevance (cf. Látod? ‘Can you see [that]?’).

4.2  Compatible rivalling proposals

The above mentioned definition of implicit arguments is compatible with the
approaches of relevance theory (Groefsema 1995) and centering theory (Cote
1996), according to which an argument included in the conceptual-semantic
representation of a verb should not be lexically realised under certain lexical-
semantic and contextual conditions. Functional grammars provide a similar
definition of implicit arguments licensing optional realisation to an argument in
the predicate frame (cf. Dik 1985; Groot 1985). For example, in functional gram-
mars the verb eat is always considered transitive and in its intransitive use, i.e.
without an object argument, the second argument position in the predicate frame
remains unspecified. Functional discourse grammar replaces the predicate frame
in functional grammar by a combination of predication templates, lexemes pro-
vided with abstract meaning definitions and a linking mechanism joining them
together (García Velasco & Portero Muñoz 2002). In functional discourse gram-
mar the omission and the realisation of an argument of a lexical item is accounted
for as the joining of the abstract meaning definition of the particular lexical item
with different predicate templates, i.e. the verbs’ occurrence with an implicit or an
 Enikő Németh T.

explicit argument is treated as a syntactic alternation. In this respect, functional

discourse grammar has an approach similar to that of construction grammar (cf.
Goldberg 1995, 2005) and the lexical-constructional framework (cf. Bibok 2008).
The proposals treated in the previous paragraph can be thought of as
­arguments for the definition suggested in Section 4.1 and they increase the plau-
sibility of the hypotheses based on the definition. At the same time, since they are
rivalling approaches, I have indicated in my previous research (cf. e.g. Németh T.
2008, 2010) why their suggestions are incomplete and how my proposal results in
a better solution to the problem of implicit arguments.

4.3  Non-compatible rivalling approaches

However, the definition of implicit arguments in Section 4.1 differs from the
­definition widely accepted by generative grammars (cf. Radford 1997; Kocsány
1997; Bartos 2000; É. Kiss 2002). The definition suggested in 4.1 allows us to
­consider a phonetically unexpressed empty pronoun (pro and even PRO) to be a
kind of implicit argument which has a position of its own in the syntactic struc-
ture, different from the other types of lexically unrealised arguments.
The handling of omitted arguments in lexical functional grammar also differs
from the above mentioned definition (cf. Komlósy 1994, 2001). For example, lexi-
cal functional grammar lists two lexical items for the verb eat, a transitive and an
intransitive one, and a lexical rule is introduced to relate these two items to each
other, changing their argument structure in order to account for the occurrence
of eat with or without an object argument. In the case of Hungarian weather verbs
such as esik ‘[for precipitation to] fall’ lexical functional grammar assumes a pro
subject argument in the lexicon instead of the occurrence with an implicit subject
argument (cf. Komlósy 2001).
In the context of these approaches my proposal generates inconsistency. In my
previous research I have resolved this inconsistency by showing that argumenta-
tion built on my definition can explain a wider spectrum of data in a coherent
way, i.e. it is more plausible in the light of data than the rivalling approaches (cf.
Németh T. 2008, 2010).10

.  It is worth mentioning that my definition of implicit arguments contains, but termino-
logically does not differentiate between, the notions of implicit arguments and some kinds
of unarticulated constituents recently discussed in the semantics and pragmatics literature
(cf. e.g. Taylor 2001; Martí 2006; Recanati 2007). The confrontation of the approaches in the
debate concerning unarticulated constituents with my proposal will be the task of a further
argumentation cycle (cf. Kertész & Rákosi 2012: Section 10.4, this volume: Subsection 2.6.1).
 Hungarian verbs of natural phenomena with explicit and implicit subject arguments 

5.  M
 odification of the p-context: The occurrence of verbs of natural
phenomena with implicit subject arguments in Hungarian

On the basis of the theoretical conclusions in the previous sections and also ­relying
on various kinds of data from different direct sources we can assume that verbs of
natural phenomena considered subjectless in Hungarian grammars are not sub-
jectless. They can all occur with (or without) explicit subjects but differ in what
manner and in what context. The default use of the verbs esteledik ‘[for evening to]
close in’, be-esteledik ‘pvb-[for evening to] close in’, alkonyodik ‘[for dusk to] set in’,
be-alkonyodik ‘pvb-[for dusk to] set in’, hajnalodik ‘[for day to] break’, tavaszodik
‘[for spring to] come’, ki-tavaszodik ‘pvb-[for spring to] come’, fagy ‘freeze’, olvad
‘thaw’, villámlik ‘[for lightning to] strike’, havazik ‘snow’, be-havazik ‘pvb snow in’,
virrad ‘dawn’, pirkad ‘dawn’ etc. is the occurrence without an explicit, lexicalised
subject. However, this does not necessarily mean that these verbs are zero argu-
ment verbs, as is widely assumed in the literature. On the contrary, these verbs
have an argument position in their lexical-semantic representation which can be
filled in with an explicit, lexicalised syntactic subject, but what lexemes can appear
in this subject position is strictly constrained. In the literature concerning implicit
object arguments in various languages a hypothesis has been formulated accord-
ing to which if a verb can have only one specific object or a very limited number
of specific objects, then the object(s) can be predicted on the basis of the particu-
lar verb meaning, and, therefore, be easily omitted (Rice 1988; García Velasco &
Portero Muñoz 2002). For example, in English the verbs of bodily emission such
as sneeze, blow, cry, spit, ejaculate, piss, and vomit, and verbs of contribution such
as contribute, donate, and give (‘contribute’) can occur with implicit direct object
arguments in this manner. In construction grammar Goldberg (2005) argues that
these verbs can be fused with the implicit theme construction since their second
argument position can be filled in by a unique object. The omitted syntactic object
can be identified with a particular kind of bodily emission or contribution which is
indicated in the lexical-semantic representation of these verbs as selection restric-
tions. Pragmatic factors, especially politeness considerations, also motivate l­ eaving
these arguments lexically unrealised. Lexical-functional grammar also applies a
semantically determined, unique selection restriction to account for characteris-
ing idioms such as to crane one’s neck (Komlósy 2001: 121).
There is no theoretical argument against the extension of this hypothesis to syn-
tactic subject arguments. By means of selection restrictions, the lexical-semantic
representation of verbs of natural phenomena enumerated above unambiguously
and uniquely determines those kinds of natural phenomena (e.g. the Hungarian
equivalents of sky, dawn, day and horizon) that can occupy the s­ yntactic subject
position in non-metaphorical uses. To put it the other way around, it directly
 Enikő Németh T.

f­ ollows from the selection restriction what kinds of natural phenomena can serve
as syntactic subjects with verbs of natural phenomena; therefore, there is no need
to express them explicitly. Thus, if the verbs in question occur without an overt,
lexically realised subject, then this implicit subject can be unambiguously and
exclusively identified with the information about the particular kind of natural
phenomenon involved in the selection restriction. Consider (5)(a)–(b).
(5) a. Nagy a vihar-Ø. Villám-l-ik.
big the storm-nom.sg lightning-sfx-prs.indef.3sg
‘The storm is big. It is lightning.’
b. Nagy a vihar-Ø. Az egész
big the storm-nom.sg the whole
ég-Ø villám-l-ik.
sky-nom.sg lightning-sfx-prs.indef.3sg
‘The storm is big. It is lightning [in the entire sky].’

In (5)(a) the verb villámlik ‘[for lightning to] strike’ occurs with an implicit subject
argument, while in (5)(b) with a lexically realised one. The occurrence of the verb
villámlik ‘[for lightning to] strike’ with an implicit subject argument is licensed by
the unique selection restriction on argument x in its lexical-semantic representa-
tion: argument x must be az ég ‘the sky’ among the natural phenomena (cf. (6)).
Similarly, in the semantic representation of the verbs esteledik ‘[for evening to]
close in’ and hajnalodik ‘[for day to] break, dawn’ the argument x which can be
realised as a syntactic subject must be day/time and time, respectively: esteledik a
nap/az idő,11 hajnalodik az idő (cf. (7)–(8)).
(6) villámlik: ‘x: ég erős cikázó fényjelenséget mutat’
‘x: sky is showing very bright flashes of light’

(7) esteledik: ‘x: nap/idő (24 órányi időegység) az esti időszakhoz közeledik’
‘x: day/time (24 hour time unit) is approaching evening time’

(8) hajnalodik: ‘x: idő napkeltéhez közeledik’

‘x: time is approaching daybreak’

The possibility to use the verbs villámlik ‘[for lightning to] strike’, esteledik ‘[for
evening to] close in’, hajnalodik ‘[for day to] break, dawn’ with implicit subject
arguments follows from the double interpretability of their lexical-semantic
­representations. If they occur with an explicit subject, then the subject argument
is also in the focus of attention, while if these verbs are used with an implicit

.  Esteledik a nap (24 hour time unit) and esteledik az idő are dialectal variants.
 Hungarian verbs of natural phenomena with explicit and implicit subject arguments 

s­ubject, only the particular natural (weather) event is in the focus of attention,
but the lexically unrealised subject argument can also be accessed by means of the
unique ­selection restriction as background information. The two kinds of occur-
rence of verbs v­ illámlik ‘[for lightning to] strike’, esteledik ‘[for evening to] close
in’, ­hajnalodik ‘[for day to] break, dawn’ with an overt or covert subject argument
can be considered a syntactic alternation similar to the occurrence of transitive
verbs’ with or without an explicit object argument (Bibok 2008, 2010). Because of
the uniqueness of the selection restriction on argument x and also the permanent
accessibility of argument x in the lexical-semantic representation as background
information, the verbs villámlik ‘[for lightning to] strike’, esteledik ‘[for evening
to] close in’, hajnalodik ‘[for day to] break, dawn’ as well as be-esteledik ‘pvb-[for
evening to] close in’, alkonyodik ‘[for dusk to] set in’, be-alkonyodik ‘pvb-[for dusk
to] set in’, tavaszodik ‘[for spring to] come’, ki-tavaszodik ‘pvb-[for spring to] come’,
fagy ‘freeze’, olvad ‘thaw’, havazik ‘snow’, be-havazik ‘pvb-snow in’, virrad ‘dawn’,
pirkad ‘dawn’ can be used with implicit subject arguments very economically,
which is why their o ­ ccurrence with a lexically unrealised subject argument can be
considered their default use.
In the semantic literature the meaning of the utterance Villámlik ‘It is light-
ning’ in (5)(a) is accounted for as a thetic statement (Maleczki 2001; Martí 2006;
Recanati 2007). Thetic statements are about a particular situation, they have no
topic constituent. But if a component of a particular situation e.g. a temporal or
locative constituent becomes prominent enough, then it can play the role of the
sentential topic (Maleczki 2001: 169, 179f.), cf. (9)(a)–(b).
(9) a. Tegnap hav-az-ott.
Yesterday snow-sfx-pst.indef3sg
‘Yesterday it snowed.’
b. Galyatető-n hav-az-ik.
Galyatető -sup snow-sfx-prs.indef3sg.
‘At Galyatető it is snowing.’

The utterances in (9)(a)–(b) are already categorical judgments which make a

predication on the prominent temporal and local components of the particular
situation which are now in the sentential topic position. Similarly, in (5)(b) in
the course of the occurrence of the verb villámlik ‘[for lightning] to strike’ with
the explicit subject argument az ég ‘the sky’ the participant required by the verb’s
unique selection restriction becomes prominent, therefore it can occupy the topic
position in the sentence, and, consequently, the statement can be considered a
categorical judgment.
Now, let us turn to the metaphorical use of the verb villámlik ‘[for lightning]
to strike’ in (10)(a)–(c).
 Enikő Németh T.

(10) a. (In the family there is a big row between the father and kids beause of
the kids’ bad behaviour when the mother enters the room.)
Lát-om, nagy a vihar-Ø,
see-prs.def.1sg big the storm-nom.sg
villám-l-ik apa-Ø tekintet-e-Ø.
lightning-sfx-prs.indef.3sg father-gen.sg glance-poss.3sg-nom.sg
‘I can see that there is a big storm, father’s glances are flashing
[like lightning].’
b. ∗Lát-om, nagy a vihar-Ø,
 see-prs.def.1sg big the storm-nom.sg
villám-l-ik [apa-Ø tekintet-e-Ø].
lightning-sfx-prs.indef.3sg   father-gen.sg glance-poss.3sg-nom.sg

c. (Later, kids among themselves.)
− Lát-t-ad, milyen volt
   see-pst-def.2sg what kind be.pst.3sg
apa-Ø tekintet-e-Ø?
father-gen.sg glance-poss.3sg-nom.sg
‘Did you see what father’s glances were like?’
− Igen. Villám-l-ott
   yes lightning-sfx-pst.indef.3sg
[apa-Ø tekintet-e-Ø].
  father-gen.sg glance-poss.3sg-nom.sg
‘Yes. They [=father’s glances] were flashing [like lightning].’

In (10)(a) az apa tekintete ‘father’s glances’ serves as the explicit subject of the
verb villámlik ‘[for lightning to] strike’ metaphorically used. The metaphorical use
of villámlik with an explicit subject is also supported by the metaphorical use of
vihar ‘storm’ in the utterance context. On the basis of (6) it is obvious that by
means of its selection restriction the verb villámlik requires a unique subject argu-
ment (ég ‘sky’). The lexically realised subject argument in (10)(a) is not identical
with the one required by the selection restriction; consequently, it cannot be left
implicit, which is why (10)(b) is not grammatical. (10)(b) cannot be used with an
implicit subject argument metaphorically either. Keszler (2000: 409) referred to
this ­phenomenon by formulating a constraint according to which in the metaphor-
ical uses weather verbs must be supplied with explicit, lexically realised subjects.
However, in (10)(c) one can see that the verb villámlik ‘[for lightning] to strike’ can
occur with an implicit subject argument in a metaphorical sense. But this occur-
rence with an implicit subject argument in a metaphorical sense is not licensed
by the verb’s unique selection restriction on argument x and, consequently, by the
 Hungarian verbs of natural phenomena with explicit and implicit subject arguments 

possibility of double interpretability, instead, it is licensed by discourse grammati-

cal factors, it is a kind of zero anaphor.
Let us also examine another group of verbs of natural phenomena, namely,
verbs with meanings expressing how rain, snow etc. falls. With the verbs szemerkél
‘drizzle’, csepeg/csöpög ‘[for rain to] fall in big drops’, csepereg/csöpörög ‘sprinkle’,
eseget ‘rain sporadically’, csorog/csurog ‘trickle’, zuhog ‘pour’, szakad ‘pour’, ömlik
‘pour’ the noun eső ‘rain’, with the verb szállingózik ‘fall softly’ the noun hó ‘snow’,
with the verb szitál ‘mizzle’ the nouns köd ‘haze’ and eső ‘rain’, and with the verb
hull ‘fall’ the nouns eső ‘rain’, hó ‘snow’ as well as zúzmara ‘rime’ can serve as
explicit, lexically realised subjects. However the verb esik ‘[for precipitation to]
fall’ can be used with a range of nouns in the subject position, e.g. eső ‘rain’, hó
‘snow’, hódara ‘hoar-frost’, dara ‘sleet’, jégeső ‘hail’, jég ‘hail’, köd ‘haze’, permet ‘fine
rain’, harmat ‘dew’, dér ‘frost’, and zúzmara ‘rime’. Verbs expressing meanings relat-
ing to how rain, snow etc. fall can occur with implicit subject arguments in various
ways. Consider (11).
(11) – Milyen az idő-Ø?
   What.like the weather-nom.sg
‘What is the weather like?’
a. – Szemerkél. ‘It is drizzling.’
– Csepeg/csöpög. ‘It is dripping rain.’
– Csepereg/csöpörög. ‘It is sprinkling rain.’
– Eseget. ‘It is raining sporadically.’
– Csorog/csurog. ‘It is trickling rain.’
– Zuhog/szakad/ömlik. ‘It is pouring with rain.’
b. – Szállingózik. ‘It is snowing softly.’
c. – Szitál. ‘It is mizzling.’
d. ∗– Hull. ‘It [precipitation] is falling.’
e. – Esik. ‘It [precipitation] is falling.’

Let us examine the utterances in (11)(a)–(e) answering the question Milyen az

idő? ‘What is the weather like?’ Answers in (11)(a) contain verbs with implicit
subject arguments. In all of them the implicit subject argument can be identi-
fied unambiguously and uniquely with the noun eső ‘rain’. Verbs in (11)(a) have
a shared meaning component in their lexical-semantic representation, namely, a
selection restriction which predicts that their argument x must be eső ‘rain’: x: eső
‘rain’. This unique selection restriction licenses the verbs’ occurrences with implicit
subject arguments in (11)(a). There are native speakers of Hungarian who can use
the verbs zuhog and szakad in (11)(a) with the explicit subject hó ‘snow’ as well,
cf. zuhog a hó ‘it is pouring with snow’, szakad a hó ‘it is pouring with snow’. In
this case in the lexical-semantic representation of the verbs zuhog and szakad the
 Enikő Németh T.

unique selection restriction constraining the subject argument is a two-element-

set from which one has to select according to the contextual requirement: x: {eső
‘rain’, hó ‘snow’}. (Cf. also the lexical-semantic representation of the verb szitál
‘mizzle’ below.)
The meanings of verbs in (11)(a) also differ from each other. The differ-
ences concern the manner in which the rain falls. While in the case of szemerkél
‘drizzle’ this manner is ‘lightly falling in fine drops’, in the case of zuhog ‘pour’
it is ‘very heavily raining’. There are various proposals in the literature regarding
the status of this kind of information relating to the lexical-semantic representa-
tions of verbs. Conceptual semantics (Jackendoff 1990) and the theory of genera-
tive lexicon (Pustejovsky 1995, 1998) do not distinguish between linguistic and
encyclopaedic knowledge in lexical (conceptual) semantic representations, thus
the pieces of information related to the manner of actions and events must be
built into them. Two-level conceptual semantics (Bierwisch 1983, 1996), as its
name suggests, however, makes a difference between these two types of knowl-
edge and, consequently, between the linguistic and conceptual meanings. In this
theory the characteristic manner of “falling events” can be captured by the lexical
stereotype belonging to encyclopaedic knowledge. Although in relevance theory
(Sperber & Wilson 1995) the linguistic and encyclopaedic kinds of information
are distinguished, they are treated in the same way, rather than being provided
with a separate, independent status. In relevance theory the conceptual node of a
verb involves three types of information: lexical, encyclopaedic and logical. If the
verbs’ conceptual node becomes active, all these three kinds of information can
be accessed simultaneously. The encyclopaedic information on the characteristic
manner of actions or events is added to the lexical information in the course of its
use or interpretation. In functional discourse grammar (García Velasco & Portero
Muñoz 2002) the encyclopaedic information is included in the cognitive compo-
nent of the grammar. When a verb appears in a sentence, the representational level
and the cognitive component interact with each other, so both the linguistic and
encyclopaedic information is taken into account.
In this brief overview we have seen that the rivalling solutions differ in terms
of whether they distinguish between linguistic knowledge and encyclopaedic
knowledge or not. But it is indisputable that the lexical-semantic representation
of verbs must be extended by this information, or this information must somehow
be added to it (cf. Bibok & Németh T. 2001: 296). So, it is plausible to conclude that
the information concerning the typical, characteristic manner of actions or events
forms a necessary part of verb meanings.12

.  The present chapter does not aim to confront these solutions. This can be the task of another
argumentation cycle (Kertész & Rákosi 2012: Section 10.4, this volume: Subsection 2.6.1).
 Hungarian verbs of natural phenomena with explicit and implicit subject arguments 

Now, let us return to the answers to the question Milyen az idő? ‘What is
the weather like?’ In (11)(b) the verb szállingózik ‘snow softly’ also puts a unique
­selection restriction on its subject argument, it must be hó ‘snow’, i.e. x: hó ‘snow’.
This unique selection restriction makes it possible to use the verb szállingózik
without an explicit subject. However, in (11)(c) it is not unambiguous what the
exact reference of the lexically unrealised subject argument is. In the lexical-
semantic representation of the verb szitál ‘mizzle’ the unique selection ­restriction
concerning argument x is a two-element-set from which one has to select: x: {köd
‘haze’; eső ‘rain’}. The particular occurrence of the verb szitál with an implicit
­subject is licensed by the interaction between the verb’s lexical-semantic repre-
sentation, including the selection restriction and information from the observ-
able physical context. If one does not know the physical environment, then one
cannot select from the two possibilities predicted by the selection restriction.
However, this does not mean that (11)(c) becomes a syntactically ill-formed
­utterance, but rather that two meanings can be assigned to it. There are Hungarian
native speakers according to whose intuition the acceptability of (11)(b) and (c)
is questionable. Their intuition always requires an explicit subject with the verbs
szállingózik ‘snow softly’ and szitál ‘mizzle’ without a specific context. In this case
the lexical-semantic representation of the verbs szállingózik and szitál do not con-
tain a unique selection restriction requiring the particular kind of precipitation
they can occur with, i.e. this piece of encyclopaedic information has not yet been
built into the lexical-semantic representation of these verbs. It must be noted that
this ­compulsory-subject-analysis supports the hypothesis that weather verbs can
occur with an explicit subject to a great extent.
The utterance in (11)(d) is not grammatical. Although in the use of the verb
hull ‘[for precipitation to] fall’ as a weather verb it is obvious what kind of pre-
cipitation can fall, i.e. e.g. eső ‘rain’, hó ‘snow’ and zúzmara ‘rime’, it cannot be
used without an explicit subject in a sentence environment. It can be assumed
that there is no unique selection restriction built into the lexical-semantic repre-
sentation of hull ‘[for precipitation to] fall’ which can predict what precipitation
can fall, therefore it cannot occur with an implicit subject argument in the first
(A) manner (cf. Section 4.1), only relying on its lexical-semantic representation.
However, if one answers the question Milyen az idő? ‘What is the weather like?’
with the utterance Nagy pelyhekben hull [a hó] ‘It is snowing in large flakes’, then
the use of hull ‘[for precipitation to] fall’ with implicit subject argument becomes
acceptable. The encyclopaedic information stored under the conceptual address of
hull ‘[for precipitation to] fall’ and the other lexemes in the utterance license the
omission of the subject argument a hó ‘the snow’; i.e. hull ‘[for precipitation to]
fall’ can occur with an implicit subject argument in the second (B) manner. Hull
‘[for p­ recipitation to] fall’ can be used with an implicit subject argument in the
third (C) manner as well, cf. (12).
 Enikő Németh T.

(12) (The weather forecast predicts snow.)

− Száll-ingóz-ik már [a hó-Ø]?
    fly-sfx-prs.indef.3sg already    the snow-nom.sg
‘Is it [the snow] snowing softly yet?’
− Már hull [a hó-Ø].
   already fall.prs.indef.3sg    the snow-nom.sg
‘It [the snow] is falling already.’

In (12), in the question the verb szállingózik ‘snow softly’ occurs with an implicit
subject argument in the first (A) manner by means of the selection restriction.13
This question and the implicit subject argument [a hó ‘the snow] within it provide
the discourse context for the answer to the question. In the answer the implicit
subject argument [a hó ‘the snow] occurs with the verb hull ‘[for precipitation
to] fall’ in the third (C) manner as a zero anaphor, co-referential with the implicit
subject argument in the question.
Let us continue the analysis examining the utterance in (11)(e). The lexical-
semantic representation of the verb esik ‘[for precipitation to] fall’ does not put a
unique selection restriction on its subject argument, instead the selection ­restriction
constrains the type of the subject argument: x: csapadék ‘precipitation’, cf. (13).
(13) esik: ‘x: csapadék hull’
‘x: precipitation falls’

In (11)(e) the lexically unrealised subject argument can be identified with the type
csapadék ‘precipitation’. If a particular token of the type csapadék, i.e. a s­ pecific
kind of precipitation is referred to, then it must be lexicalised, cf. (14).
(14) Es-ik az eső-Ø/ hó-Ø/ jégeső-Ø/
fall-prs.indef.3sg the rain-nom.sg snow-nom.sg hail-nom.sg
hó-dara-Ø/ köd-Ø/ permet-Ø/ harmat-Ø/
snow-sleet-nom.sg haze-nom.sg sprinkle-nom.sg dew-nom.sg
dér-Ø/ zúzmara-Ø.
frost-nom.sg rime-nom.sg
‘Rain/snow/hail/hoar-frost/haze/a sprinkle of rain/dew/frost/rime is falling.’

The particular kinds of the type of precipitation enumerated in (14) show that the
type csapadék ‘precipitation’ has a prototypical structure. This prototypical struc-
ture also influences the occurrence of the verb esik ‘[for precipitation to] fall’ with
implicit subject arguments. The more typical the type of precipitation is, the easier

.  Of course, the information from the weather forecast may also influence the use of
szállingózik ‘snow softly’ with implicit subject argument.
 Hungarian verbs of natural phenomena with explicit and implicit subject arguments 

it is to use the verb esik ‘[for precipitation to] fall’ with an implicit s­ ubject argument
without a specific physical or discourse context. Since in H ­ ungarian ­geographical
circumstances the prototypical csapadék ‘precipitation’ is rain, without a specific
physical or discourse context the reference of the implicit s­ubject argument in
(11)(e) can be identified with az eső ‘the rain’.14
Finally, it is worth analysing the verb havazik ‘snow’ in some more detail.
The verb havazik ‘snow’ can occur with an explicit subject argument as well,
cf. (15)(a)–(c).
(15) a. ?A hó-Ø hav-az-ik./
  the snow-nom.sg snow-sfx-prs.indef.3sg/
?Hav-az-ik a hó-Ø.
  snow-sfx-prs.indef.3sg the snow-nom.sg
‘The snow is snowing.’
b. A rég vár-t hó-Ø meg-áll-ás-Ø nélkül,
the long wait-ptcp snow-nom.sg pvb-stop-sfx-nom.sg without
gyönyörű-en hav-az-ik.
beautiful-sfx snow- sfx-prs.indef.3sg
‘The long awaited snow is snowing beautifully and continuously.’
c. A múl-t század-i hó-Ø még nagy
the last-ptcp century-sfx snow-nom.sg still large
pelyh-ek-ben hav-az-ott.
flake-pl-ine snow-sfx-pst.indef.3sg
‘The snows of the last century were snowing in large flakes.’

While the acceptability of (15)(a) is questionable, (15)(b)–(c) are acceptable.

There is an old observation in connection with the behaviour of cognate objects
according to which if a cognate object or the verb itself is complemented by cer-
tain constituents, then the emerging constructions become acceptable. While the
utterance in (16)(a) is rather strange, the utterance in (16)(b) is quite acceptable.
(16) a. ?József-Ø ebéd-et ebéd-el.
  József-nom.sg dinner-acc.sg dinner-sfx.prs.indef.3sg
‘?József has a dinner for dinner.’
b. József-Ø meleg ebéd-et ebédel.
József-nom.sg warm dinner-acc.sg dinner-sfx.prs.indef.3sg
‘József has a warm dinner for dinner.’

.  In winter the prototype of precipitation can be snow.

 Enikő Németh T.

This analysis can be extended to the verb havazik ‘snow’ and the noun hó ‘snow’.
The acceptability of (16)(b)–(c) can be accounted for as follows: the verb havazik
is complemented by an adverb, and the noun hó is complemented by an adjective.
The verb havazik ‘snow’ cannot occur with an implicit subject argument other
than a hó ‘the snow’. In its lexical-semantic representation the verb havazik puts a
unique selection restriction on its argument x: hó ‘snow’, cf. (17).
(17) havazik: ‘x: hó esik’
‘x: snow falls’

Since the verb havazik ‘snow’ is morphologically derived from the noun hó ‘snow’,
predicted by the selection restriction to be the unique subject argument, the mor-
phological structure of havazik ‘snow’ is transparent. This transparent morpho-
logical structure, as well as its phonetic form, easily recall the subject argument
hó predicted by the selection restriction. Therefore, the explicit occurrence of hó
‘snow’ is disturbingly redundant with the verb havazik ‘snow’. Of course, in meta-
phorical uses the verb havazik ‘snow’ also occurs with explicit subjects as we have
seen above in Section 3: havazik az élet ‘life is snowing’, az egész világ havazik ‘the
whole world is snowing’, csend havazik ‘silence is snowing’.
Consider also (18).
(18) Hav-az-ik [a hó-Ø]. Nagy-on szép-en
snow-sfx-prs.indef.3sg   the snow-nom.sg very-sfx nice-sfx
es-ik [a hó-Ø].
fall-prs.indef.3sg    the snow-nom.sg
‘[Snow] is snowing. [The snow] is falling very nicely.

The possibility that in the second utterance in (18) one can refer to a hó ‘the snow’
by a zero anaphor supports the plausibility of the analysis assuming an implicit
subject argument in (18) and weakens the plausibility of another widely accepted
analysis in traditional Hungarian grammars according to which the noun hó
‘snow’ is incorporated into the verb havazik ‘snow’.

6.  Th
 e resolution of the starting p-problem in the modified p-context:
The advantages of the analysis of verbs of natural phenomena
with implicit and explicit subject arguments

Starting out from the wide spectrum of data neglected by previous approaches, in
this study I have proposed a lexical-semantic representation with a subject argu-
ment position for the verbs of natural phenomena analysed as subjectless verbs
in Hungarian grammatical tradition as well as in lexical-functional grammar.
 Hungarian verbs of natural phenomena with explicit and implicit subject arguments 

The verbs of natural phenomena constrain their subject argument very strictly
by means of the elements in their lexical-semantic representation such as selec-
tion restrictions, information on the characteristic manner of natural events, and
the prototypical structure of the categories. The lexical-semantic representation of
these verbs can be interpreted in two ways. Either only the natural events denoted
by the verbs are in the focus of attention, or their subject is as well. When only the
natural events are in the focus of attention, the subject arguments can be left lexi-
cally unrealised. Verbs give access to their lexical-semantic representations with
their argument structure. If the subject argument is left implicit in the first (A)
manner, it can be identified with the information provided by the constraints the
selection restriction imposes on the subject argument as background informa-
tion. Since in the case of verbs of natural phenomena what can fill in the subject
position is strictly, or even uniquely, constrained by the selection restrictions, the
subject argument should not be lexically realised. Therefore, the economical and
default occurrence of verbs of natural phenomena is the use without an explicit
subject, i.e. with implicit subjects. At the same time, if the subject argument is also
in the focus of attention or the particular subject argument is not the prototype
of the category required by the selection restriction or the speaker wants to refer
to a specific token of the category, then the subject argument must be explicitly
expressed. Similarly, in metaphorical uses, the verbs of natural phenomena cannot
occur with a lexically unrealised subject argument in the first (A) manner either.
Sometimes the background information in the lexical-semantic representation
of a verb of natural phenomenon by itself cannot license an argument to be left
implicit. In these cases the lexical-semantic representation of the verb interacts
with the context.
The main advantages of the proposed solution are as follows. First, it can
explain the occurrences of verbs of natural phenomena without and with an
explicit subject argument in a unified way. Second, it can account for a wider
spectrum of data than the rivalling approaches in the starting p-context could, i.e.
it cancels the p-inconsistency in the extended p-context. Third, assuming a sub-
ject argument position in the lexical-semantic representation of verbs of natural
phenomena, the metaphorical uses can be handled in the same way as the non-
metaphorical occurrences, i.e. further p-inconsistency and p-incompleteness in
the starting and extended p-context can be eliminated. Fourth, the occurrences of
verbs of natural phenomena with implicit subject arguments can be treated simi-
larly to the uses of verbs with implicit direct and indirect objects. And, fifth, the
occurrences of verbs of natural phenomena with explicit and implicit subjects can
be considered syntactic alternations similarly to the non-objectless and objectless
uses of transitive verbs. This supports the proposition of this kind of a lexical-
semantic representation for verbs of natural phenomena from the syntactic point
 Enikő Németh T.

of view. All these advantages increase the explanatory power of the implicit subject
argument analysis of verbs of natural phenomena.
To summarise: confronting the proposed implicit subject analysis with
the rivalling approaches in the starting p-context, it can be established that the
new analysis is more plausible than the previous ones on the basis of the newly
­introduced data from various direct sources as well as theoretical statements listed
above as advantages. Therefore, the new solution can be considered as a resolution
of the p-problem in the starting p-context and the newly introduced data from dif-
ferent direct sources serve as strong evidence for it (cf. Kertész & Rákosi 2012: 133,
this volume: Section 2.5).

Acknowledgements

The research reported on in the present chapter was supported by the MTA-
DE Research Group for Theoretical Linguistics and the TÁMOP-4.2.2/B-10
/1-2010-0024 project, which is co-financed by the European Union and the
European Social Fund. I would like to express my gratitude to András ­Kertész,
Csilla Rákosi and Károly Bibok for their comments and suggestions which
helped me to clarify some points. I wish to thank Anna Fenyvesi for improving
my English.

References

Bartos, Huba. 2000. Az inflexiós jelenségek szintaktikai háttere (The syntactic background
of inflectional phenomena). In Stukturális magyar nyelvtan, 3: Morfológia (Structural
­Hungarian Grammar, 3: Morphology), Ferenc Kiefer (ed.), 653–762. Budapest: Akadémiai
Kiadó.
Bibok, Károly. 2008. Az igék szemantikája és a szintaktikai alternáció (The semantics of verbs and
syntactic alternation). In Strukturális magyar nyelvtan, 4: A szótár szerkezete ­(Structural
Hungarian Grammar, 4: Structure of the Lexicon), Ferenc Kiefer (ed.), 23–70. Budapest:
Akadémiai Kiadó.
Bibok, Károly. 2009. Az evidencia problémája az elméleti nyelvészetben (The problem of evi-
dence in theoretical linguistics). Talk presented at the conference Kutatócsoportjaink (Our
research groups), Szeged.
Bibok, Károly. 2010. From syntactic alternation to lexical pragmatics. In The Role of Data at the
Semantics-Pragmatics Interface, Enikő Németh T. & Károly Bibok (eds), 261–304. Berlin:
Mouton de Gruyter.
Bibok, Károly & Németh T., Enikő. 2001. How the lexicon and context interact in the mean-
ing construction of utterances. In Pragmatics and the Flexibility of Word Meaning, Enikő
Németh T. & Károly Bibok (eds), 289–320. Amsterdam: Elsevier.
 Hungarian verbs of natural phenomena with explicit and implicit subject arguments 

Bierwisch, Manfred. 1983. Semantische und konzeptuelle Repräsentation lexikalischer

­Einheiten. In Untersuchungen zur Semantik [Studia Grammatica 22], Rudolf Růžička &
Wolfgang Motsch (eds), 61–99. Berlin: Akademie Verlag.
Bierwisch, Manfred. 1996. How much space gets into language? In Language and Space, Paul
Bloom, Mary A Peterson, Lynn Nadel & Merrill F. Garrett (eds), 31–76. Cambridge MA:
The MIT Press.
Cote, Sharon A. 1996. Grammatical and Discourse Properties of Null Arguments in English.
Ph.D. dissertation, 〈http://falcon.jmu.edu/∼cotesa/research.html〉 (20 December 2009).
Dik, Simon C. 1985. Formal and semantic adjustment of derived constructions. In Predicates
and Terms in Functional Grammar, A. Machtelt Bolkestein, Casper de Groot & J. Lachlan
Mackenzie (eds), 1–28. Dordrecht: Foris.
É. Kiss, Katalin. 2002. The Syntax of Hungarian. Cambridge: CUP.
García Velasco, Daniel & Portero Muñoz, Carmen. 2002. Understood Objects in Functional Gram-
mar [Working Papers in Functional Grammar 76], Amsterdam: University of Amsterdam.
Goldberg, Adele E. 1995. Constructions: A Construction Grammar Approach to Argument Struc-
ture. Chicago IL: The University of Chicago Press.
Goldberg, Adele E. 2005. Constructions, lexical semantics and the correspondence principle:
Accounting for generalizations and subregularities in the realization of arguments. In The
Syntax of Aspect, Nomi Erteschik-Shir & Tova Rapoport (eds), 215–236. Oxford: OUP.
Groefsema, Marjolein. 1995. Understood arguments: A semantic/pragmatic approach. Lingua
96: 139–161.
de Groot, Casper. 1985. Predicates and features. In Predicates and Terms in Functional Grammar,
A. Machtelt Bolkestein, Casper de Groot & J. Lachlan Mackenzie (eds), 71–84. Dordrecht:
Foris.
Jackendoff, Ray. 1990. Semantic Structures. Cambridge MA: The MIT Press.
Kertész, András & Rákosi, Csilla. 2012. Data and Evidence in linguistics. A plausible argumenta-
tion model. Cambridge: CUP.
Kertész, András & Rákosi, Csilla. this volume. The p-model of data and evidence in linguistics.
Keszler, Borbála (ed.). 2000. Magyar grammatika (Hungarian Grammar). Budapest: Nemzeti
Tankönyvkiadó.
Kocsány, Piroska. 1997. Pro-drop und Textkohärenz: Interagierende Regeln im Gebrauch des
Subjektspronomens im Ungarischen. Acta Linguistica Hungarica 44: 385–411.
Komlósy, András. 1994. Complements and adjuncts. In Syntax and Semantics 27, Ferenc Kiefer
& Katalin É. Kiss (eds), 91–178. New York NY: Academic Press.
Komlósy, András. 2001. A lexikai-funkcionális grammatika mondattanának alapfogalmai [Basic
notions in the syntax of lexical-functional grammar]. Budapest: Tinta Könyvkiadó.
Maleczki Márta. 2001. Indefinite arguments in Hungarian. In Argument Structure in Hungarian,
István Kenesei (ed), 157–199. Budapest: Akadémiai Kiadó.
Martí, Luisa. 2006. Unarticulated constituents revisited. Linguistics and Philosophy 29: 135–166.
Nagy C., Katalin. 2008. Data in historical pragmatics: A case study on the Catalan periphrastic
perfective past. In New Approaches to Linguistic Evidence: Pilot Studies, András Kertész &
Csilla Rákosi (eds), 171–197. Frankfurt: Peter Lang.
Németh T., Enikő. 2000. Occurrence and identification of implicit arguments in Hungarian.
Journal of Pragmatics 32: 1657–1683.
Németh T., Enikő 2001. Implicit arguments in Hungarian. In Argument Structure in Hungarian,
István Kenesei (ed.), 113–156. Budapest: Akadémiai Kiadó.
 Enikő Németh T.

Németh T. Enikő. 2008. Az implicit alanyi és tárgyi argumentumok előfordulásának lexikai-

szemantikai jellemzői (The lexical-semantic properties of the occurrence of implicit sub-
ject and direct object arguments). In Strukturális magyar nyelvtan, 4: A szótár szerkezete
(Structural Hungarian Grammar, 4: Structure of the Lexicon), Ferenc Kiefer (ed.), 71–128.
Budapest: Akadémiai Kiadó.
Németh T., Enikő. 2010. How lexical-semantic factors influence the verbs’ occurrence with
implicit direct object arguments in Hungarian. In The Role of Data at the Semantics–­
Pragmatics Interface, Enikő Németh T. & Károly Bibok (eds), 305–348. Berlin: Mouton de
Gruyter.
Németh T., Enikő & Károly Bibok. 2010. Interaction between grammar and pragmatics: The
case of implicit arguments, implicit predicates and co-composition in Hungarian. Journal
of Pragmatics 42: 501–524.
Pustejovsky, James. 1995. The Generative Lexicon. Cambridge MA: The MIT Press.
Pustejovsky, James. 1998. Generativity and explanation in semantics: A reply to Fodor and
Lepore. Linguistic Inquiry 29: 289–311.
Pusztai, Ferenc (ed.). 2003. Magyar értelmező kéziszótár (Concise Explanatory Dictionary of
Hungarian), 2nd rev. edn. Budapest: Akadémiai Kiadó.
Radford, Andrew. 1997. Syntax: A Minimalist introduction. Cambridge: CUP.
Recanati, François. 2007. It is raining (somewhere). Linguistics and Philosophy 30: 123–146.
Rescher, Nicholas. 1976. Plausible Reasoning. Assen: Van Gorcum.
Rice, Sally. 1988. Unlikely lexical entries. Berkeley Linguistic Series 14: 202–212.
Sárik, Pál. 1998. A határozói igenevek néhány problémája (Some problems of adverbial parti-
ciples). Magyar Nyelv 94: 423–435.
Sperber, Dan & Wilson, Deirdre. 1995. Relevance: Communication and Cognition, 2nd edn.
Oxford: Blackwell.
Taylor, Kenneth. 2001. Sex, breakfast, and descriptus interruptus. Synthese 128: 45–61.
Tóth, Ildikó. 2001. Impersonal constructions and null expletives. In Argument Structure in
­Hungarian, István Kenesei (ed.), 51–78. Budapest: Akadémiai Kiadó.
 chapter 6

The development of a taxonomy of verbal

disagreements in the light of the p-model

Helga Vanda Koczogh

University of Debrecen, Institute of English and American Studies

The main objective of this chapter is to find a category system that is the most
capable of describing Hungarian disagreement based on certain criteria set
beforehand. As none of the rival approaches to disagreement to be introduced
in this chapter seem to satisfy each criterion, a new taxonomy is proposed.
Distinguishing between strategies and pragmatic force modifiers (PFMs), this
new category system is more reliable and more capable of capturing the complex
nature of (Hungarian) disagreements that the rival models. A detailed description
of the metatheoretical research is provided including a step-by-step description
of the analytical procedure(s) employed. By describing each step of the research
along with the decisions and re-evaluations of the information at hand as well as
the p-context, the chapter demonstrates the cyclic, retrospective and prismatic
nature of the plausible argumentation process.

1.  Introduction: The goal and structure of the chapter

To date, the majority of research carried out on disagreement has focused on the
English language and investigated the linguistic manifestation of disagreement
while ignoring its functional spectrum. The primary aim of the present chapter
is to find a category system that is the most capable of describing Hungarian dis-
agreement. The specific aim of my chapter is twofold: (1) to examine how ver-
bal disagreement is accomplished linguistically in the mixed-sex conversations of
Hungarian undergraduate students and (2) to identify the functional categories
of disagreements occurring in the research corpus. The present study proposes
a framework that is more capable of capturing the complex nature of (Hungar-
ian) disagreements and can account for a wider range of occurrences than other
rival models. In the course of the development of the framework, the steps of the
argumentation process are reconstructed highlighting its cyclic, retrospective, and
prismatic nature (cf. Kertész & Rákosi, this volume: Subsection 2.6.1).
 Helga Vanda Koczogh

At this point it is important to explain what I mean by the term verbal dis-
agreement. Verbal disagreement is defined in this chapter as a situated activity
whose function is to express an opinion (or belief) the propositional content or
illocutionary force of which is – or is intended to be – partly or fully inconsistent
with that of a prior (non-verbal) utterance.1
The structure of the chapter is as follows. Section 2, provides a brief over-
view of major taxonomies of disagreement. These rival models serve as the start-
ing p-context (ibid.: Subsection 2.6.1) for my research, i.e. they constitute the
first step of the argumentation process. The starting p-context is problematic,
since it contains too much information about the categorisation of disagree-
ment. Thus, the starting p-context is informationally overdetermined (ibid.:
Section 2.5). S­ ection 3 gives a step-by-step account of the multi-step and multi-
method strategy that is adopted in order to address the research issues. The
starting p-context is extended by a corpus of spoken discourses and the section
provides a detailed description of a proposed taxonomy of disagreement that is
set up based on data from several direct sources. This extended p-context could
not resolve the problems identified in the starting p-context. Thus, in Section
4, a new argumentation cycle is started and the p-context is extended further
with new research data. This extended p-context is not satisfactory either, as it
inherited some weaknesses of the previous models. I prismatically re-evaluate
the p-context, revisit my taxonomy and analyze disagreements from a different
perspective, which results in a modified version of the framework. I argue for
this new version of the taxonomy in Section 5, which distinguishes between
disagreement functions (strategies) and certain linguistic devices (PFMs). This
framework is more complete and reliable than its previous version or the other
p-context versions described earlier. Consequently, the new taxonomy can be
regarded as a resolution of the starting problem in the starting p-context (ibid.:
Section 2.5).

2.  The starting p-context: Rival frameworks of disagreement

As already mentioned in the previous section, the global aim of the chapter is to
find the model that is the most capable of describing Hungarian disagreement.
Thus, I raise the following problem:
(P) Which category system allows for describing Hungarian disagreement?

.  For more details see Koczogh (2012).

 The development of a taxonomy of verbal disagreements in the light of the p-model 

In my view, a model like that should meet the following criteria:

1. it should be comprehensive with a number of categories in order to capture

the complex nature of disagreement
2. it should contain easily distinguishable, clear-cut categories
3. the categories should be as reliable as possible; thus, they should not be based
solely on the linguist’s intuition2

In this chapter, I will look for the category system that can satisfy the above-listed
criteria to the greatest extent; that is, the model that offers the best solution to (P).
In the literature of conflict talk, various types of disagreement have been identified
by different researchers based on different criteria. In what follows, I will introduce
the four most detailed and presumably most influential models of disagreement
from contrastive and evaluative points of view. This means that I will introduce,
analyse and evaluate the four models separately as rival p-context versions within
the starting p-context.

2.1  P-context version 1: Muntigl & Turnbull’s model (1998)

Based on structural and pragmatic criteria, Muntigl and Turnbull (1998) identify
four types of disagreements:

1. irrelevancy claims: “meta-dispute-acts that comment on the conversational

interactions” (p. 229). They indicate that the preceding utterance is irrelevant
to the topic discussed using expressions like It doesn’t matter or It’s nothing to
do with it.
2. challenges: frequently preceded by markers displaying disagreement with prior
turn and normally have the syntactic form of questions. They imply that the
other interlocutor cannot provide evidence for his/her claim. (ibid., pp. 229f.)
3. contradictions: negate the propositional content of the previous utterance
often marked by a negative particle such as no or not. This is what Rees-Miller
(1995) refers to as Not P. (ibid., p. 231)
4. counterclaim: with counterclaims “speakers propose an alternative claim
that does not directly contradict nor challenge other’s claim” (ibid., p. 232).
They often co-occur with hesitation markers, mitigating devices and token
agreement.

.  I am aware that using our intuitions in linguistic analysis is indispensable and unavoid-
able; however, it might undermine the reliability of the analysis as it cannot be considered as
a completely reliable source of data. Hence, using intuition as the sole basis of analysis is not
viable.
 Helga Vanda Koczogh

Muntigl and Turnbull also note that these acts can co-occur, with the most fre-
quent combination being contradictions followed by counterclaims. In my view,
irrelevancy claims cannot be considered disagreements, as they do not contradict
(part of) the previous utterance, but rather call into question its relevance to the
topic under discussion. Another weakness of the model is that it consists of a very
limited number of categories, as a result of which it cannot possibly account for all
occurrences of disagreement. For instance, the frequently used partial agreement
(Yes, but…) does not fit any of the categories listed by Muntigl and Turnbull. Thus,
p-context version 1 does not fulfil criterion (a) described in the previous section.

2.2  P-context version 2: Rees-Miller’s (1995, 2000) taxonomy

Rees-Miller (1995, 2000), investigating the expression of disagreement in aca-
demic setting, organises the acts of disagreement into three broad categories
“based on the presence or absence of identifiable linguistic markers” (2000: 1993):
‘softened disagreement’, ‘not softened or strengthened disagreement’, and ‘aggra-
vated disagreement’ (see Figure 1 for the taxonomy and distribution of types of
linguistic markers).

Type of disagreement Type of linguistic marker

Positive comment
Positive politeness Humor
(21% of turns) Inclusive 1st person
Partial agreement
Softened disagreement
(62% of turns)
Questions
I think/I don’t know
Negative politeness
Downtoners (maybe,
(41% of turns)
sort of )
Verbs of uncertainty

Disagreement not
softened or Contradictory statement
strengthened
(30% of turns) Verbal shadowing

Rhetorical question
Aggravated
Intensifiers
disagreement
Personal, accusatory you
(8% of turns)
Judgemental vocabulary

Figure 1.  Taxonomy of disagreements (Rees-Miller 2000: 1095)

Rees-Miller (1995, 2000) takes Brown and Levinson’s (1987) model of polite-
ness as the underlying point of departure for her classification of disagreement
 The development of a taxonomy of verbal disagreements in the light of the p-model 

whereby the category of ‘softened disagreements’ corresponds to Brown and

Levinson’s ‘off record’ and ‘on record with redress’ strategies, and the groups of
‘aggravated’ and ‘neither softened nor strengthened’ disagreements correspond to
‘bald-on-record’ strategies.
The category of softened disagreements is further subdivided into positive
politeness and negative politeness strategies, again based on the Brown and Levin-
sonian distinction. Positive politeness strategies include linguistic markers (e.g.
humour, positive comment, etc.) that serve to soften the face-threatening force of
disagreement and to increase solidarity with the addressee. In negative politeness
strategies, softeners (e.g. questions, downtowners, etc.) are employed in order to
indicate uncertainty (or its facade) on the part of the speaker and thus protect the
negative face-needs of the addressee.
No overt linguistic markers can be identified in disagreements that are nei-
ther softened nor strengthened, still they are recognisable as disagreements as they
contradict the previous utterance. Besides contradiction, verbal shadowing, a rep-
etition of “a previous speaker’s utterance with altered words or intonation to indi-
cate disagreement” (Rees-Miller 2000: 1094), was used to express disagreement.
The category of aggravated disagreements goes beyond Brown & L ­ evinson’s
(1987) bald on record politeness in the sense that disagreements are not only
expressed explicitly with no redress, but their face-threatening force is also strength-
ened by the use of linguistic markers such as intensifiers or accusatory you.
Rees-Miller (1995, 2000) explores the complex ways of how the Brown and
Levinsonian factors of power and rank of imposition affect the linguistic choice
of strategies for disagreeing. This study is highly valuable, since it provides an
in-depth analysis of the speech act of disagreement and thoroughly investigates
the linguistic features by which it is expressed in American university settings.
Nevertheless, Rees-Miller’s model is not without flaws either. First, she claims
to have categorised disagreements based on the presence and absence of linguistic
markers, yet her linguistic markers include some types (e.g. partial agreement,
contradictory statement) that seem to be rather functional than linguistic catego-
ries. Second, although this taxonomy is undoubtedly more elaborate than that of
Muntigl and Turnbull (1998), it does not make mention of the most direct way of
disagreeing, that is simply saying I don’t agree or I disagree. Furthermore, some
of the categories are fuzzy, they seem to be based on linguistic intuition and the
author does not give any guidance on how to distinguish them, which means that
the model does not meet criteria (b) and (c). To illustrate this point, let’s con-
sider questions, for example. According to the model, questions mark softened
disagreement, while rhetorical questions are used in aggravated disagreement. It
is not clear when a question should be regarded as a softening or an aggravating
device. Also, the question raises how to categorise a disagreeing utterance that is
 Helga Vanda Koczogh

expressed by a question with the intention of challenging the other person. Rees-
Miller’s model cannot account for an utterance like that, thus, it does not meet cri-
terion (a) listed in Section 2. Moreover, the taxonomy disregards the circumstance
that a given linguistic item can have several functions. For instance, I think can
be a hedge or a booster depending on such factors as context and/or intonation.
Considering all the above, we can see that p-context version 2 does not offer a
solution to (P) either.

2.3  P-context version 3: Locher’s (2004) model

Another equally significant study of disagreement was carried out by Locher
(2004) focusing on the interface of power and politeness in the realisation of dis-
agreements in three different settings: (1) an informal dinner among family and
friends, (2) a business meeting, and (3) spoken discourses collected during the
2000 US presidential election featuring a radio interview of President Clinton, a
US Supreme Court hearing on the 2000 presidential election, and a televised presi-
dential debate.
In the paper analysing the linguistic realisation of disagreements in the din-
ner conversation, Locher (2004) identifies the following strategies through which
disagreement was expressed: (1) hedges (well, just, uhm, uh, I think, I don’t know),
(2) giving personal or emotional reasons for disagreeing, (3) modal auxiliaries, (4)
shifting responsibility, (5) objections in the form of a question, (6) the use of but,
(7) repetition of an utterance by a next or the same speaker, and (8) unmitigated
disagreement (p. 113). According to Locher, this list includes the main varieties
of expressing disagreement during the dinner and the categories are presented
in order of frequency. However, some of these categories (hedges, modal auxilia-
ries, discourse connectives) are ways in which disagreements are mitigated rather
than expressed. Hence, similarly to Rees-Miller’s (1995, 2000) taxonomy, Locher’s
model contains of mixture of functional and structural categories, which makes
it less reliable. In addition, we can easily think of examples that can be regarded
as belonging to several categories at the same time. For example, the utterance
But uh I think it might be fair only in certain cases. fits into categories (1), (2), (3)
and (6), too. This might weaken the reliability of the model. Also, Locher’s model
cannot account for partial agreement and some aggravated forms of disagreement
(e.g. evaluations/judgements). As a result, the model does not seem to be elabo-
rate enough to categorise all utterances of disagreement. Thus, p-context version 3
does not meet the requirements laid down in Section 2 either.

2.4  P-context version 4: Bándli’s (2009) framework

At this point, I wish to turn to Bándli’s (2009) categorisation of disagreement
strategies in Hungarian. She claims to have identified the disagreement ­strategies
 The development of a taxonomy of verbal disagreements in the light of the p-model 

employed by Hungarian people (aged between 20–40) in naturally occurring

spontaneous verbal interactions. However, the methodology she utilised resulted
in guided conversations using oral discourse completion task. In the 120 dialogues
she analyzed, the following disagreement strategies were identified listed from the
most direct one to the most indirect one (ibid., 107ff.):

1. judgement/qualification: a strategy in which the speaker makes an explicit

evaluative statement on the other speaker (e.g. Nem vagy normális! ~ You are
insane!), the propositional content of the other speaker’s utterance (e.g. Ezek
nem jó szemüvegek. ~ These glasses aren’t good.) or the other speaker’s opinion
(e.g. Ez hülyeség! ~ That is nonsense!).
2. statement of completed action (e.g. Odatettem. ~ I did put it there.)
3. explanation/listing of reasons (e.g. Tönkreteszi a szemedet. ~ It ruins your eyes.)
4. alternative suggestions (e.g. Először orvoshoz kellett volna menned. ~ You
should have gone to the doctor first.)
5. doubts/uncertainties: a strategy which criticises/questions the validity of the
other speaker’s stance (e.g. Ezt most komolyan gondolod? ~ Are you serious?,
Biztos vagy benne, hogy jó? ~ Are you sure it’s good?) or expresses the speaker’s
own unwillingness to do something (e.g. Én nem vennék be olyat. ~ I wouldn’t
take something like that.).
6. partial acceptance: a strategy in which the speaker emphasises his/her partial
agreement with the previous speaker’s utterance and implies that s/he is on a
different opinion (e.g. Azért maradj életben, majd mesélj róla! ~ Stay alive then
and tell me about it!)

Whilst the above categorisation must be credited for being the first attempt to
identify the linguistic strategies used for expressing disagreement in Hungarian,
it cannot be denied that it has several flaws. First, some of the categories (e.g.
strategies (2)–(4)) are used as if they were self-evident without any explanation of
their meaning. Second, the difference between some of the strategies is not clear
and the examples provided by the speaker give little help differentiating among
the categories and clarifying the boundaries. For instance, the utterance Először
orvoshoz kellett volna menned. (~ You should have gone to the doctor first.) can
also be interpreted as criticism which includes a negative evaluation on the past
action/behaviour of the other interactant. Thus, it could be easily categorised as a
judgement/qualification as well. Furthermore, the functional categories seem to
be intuitive. Intonation can have a meaning differentiating role in the examples
listed above; however, Bándli (2009) does not discuss the prosodic features of her
examples. Moreover, the examples she provides for illustrating the identified strat-
egies are out of context, which makes it almost impossible to judge the accuracy
of the categories, since the reader is not familiar with the preceding utterance with
 Helga Vanda Koczogh

which disagreement is expressed.3 In addition, some of the categories (e.g. state-

ment of completed action, alternative suggestions) are likely to be specific to the
contexts investigated by the researcher. Finally, the model is not elaborate enough
to enable the categorisation of disagreements that express challenge or simple con-
tradiction. Thus, Bándli’s model does not satisfy any of the criteria set in Section 2
and therefore it cannot be considered to be a solution to (P).

2.5  Evaluation of the starting p-context

The starting p-context can be considered problematic from several aspects. The
major limitation of most of the above-mentioned p-context versions is that instead
of applying a single consistent criterion, disagreements are categorised on the basis
of a combination of semantic, pragmatic and structural criteria. More importantly,
the majority of these studies focus on the linguistic realisation of disagreements
regardless of their functions. Thus, they do not provide a comprehensive picture
of the ways disagreement can be expressed (consequently, they do not satisfy cri-
terion a) mentioned in Section (2). We have also seen that some utterances can
be categorised as belonging to two or three categories at the same time within the
same model. This might be due to the fuzzy categories (criterion b)) and intui-
tive categorisation (criterion c)) these models are based on. Consequently, the
starting p-context versions do not fulfil the criteria laid down in Section 2 and
contain statements with a low plausibility value (Kertész & Rákosi, this volume:
Section 2.2). Moreover, the starting p-context is problematic (ibid.: Section 2.5)
in another respect as well. As already discussed, there are certain utterance types
that are considered as disagreement based on the definition provided in the Intro-
duction; however, none of the models can account for all these types of disagree-
ment. Thus, the starting p-context versions are informationally underdetermined
(ibid.). However, several rival approaches to disagreement constitute the starting
p-context, making it informationally overdetermined (i.e. containing too much
information) as well (ibid.).
It must be mentioned that we are also facing a problem in the starting
­p-context which is of a more basic type than the plausibility value-related informa-
tional over- and underdetermination (Kertész & Rákosi 2012: 131). That is, based

.  It has to be mentioned that mapping the pragmatic competence of Roma and Hungarian
primary school children, Mászlainé Nagy (2010) also touches upon the differences in the ex-
pression of disagreement utilised by the two groups. Based on Brown and Levinson’s (1987)
politeness theory, she identifies three broad categories of disagreements (‘bare disagreement’
strategy, ‘disagreement with recovery’, ‘indirect disagreement’ strategy), but provides no defini-
tion of them.
 The development of a taxonomy of verbal disagreements in the light of the p-model 

on a set of criteria, we try to find a category system which is capable of describing

a linguistic phenomenon. This type of problem “preceeds cases of informational
over/underdetermination in the problem-solving process” (ibid.). Since all four
­p-context versions are burdened with several problems and cannot serve with
a satisfactory solution to (P), it seems to be highly motivated to seek for a new
solution.

3.  Argumentation cycle 1

3.1  Setting up a new p-context

As the contrastive and evaluative perspectives failed to lead to a solution of the
problems in the starting p-context, I prismatically re-evaluated the p-context and
investigated the problems from a different perspective. I attempted to apply a com-
binative perspective by creating a category system of my own on the basis of the
four models, taking also into consideration new sources and data. A step-by-step
account of the multi-method strategy that was adopted in order to find a possible
solution to (P) is provided in the following section.

3.2  The elaboration of the p-context

This section describes the methodology and research design4 employed to exam-
ine the research objectives and provides the details of the multi-step and multi-
method research strategy adopted to find another possible solution to (P). In order
to resolve the problems mentioned in Section 2.5., I have taken into account data
from two direct sources (Kertész & Rákosi, this volume: Section 2.2), namely my
own linguistic intuition and spoken discourses.
The core corpus of my research is comprised of audio-recorded face-to-face
conversations of mix-sex dyads. The interactants were undergraduate students
at the University of Debrecen (aged 18–24) and native speakers of Hungarian.
The corpus is amounted to 68,194 words, which is approximately 444 minutes
(7hrs 24mins) of speech. This corpus was used as the basis for analysis. I identified
525 tokens of verbal disagreement in the corpus altogether. It must be noted that
although non-verbal ways of expressing disagreement (e.g. gestures, posture, etc.)
are not within the scope of this research, paralinguistic features often contribute to
the identification of an utterance as disagreement and help assign functions to it.

.  For a more detailed account of the research design including methodological consider-
ations see Koczogh (2012).
 Helga Vanda Koczogh

Therefore, suprasegmental characteristics will also be discussed where applicable.

Since intonation and other prosodic features can differentiate meaning and func-
tion, on examples quoted and used for illustration in this chapter, intonational
patterns5 were also marked.
The models described in the p-context versions had difficulty with the rela-
tionship between linguistic markers and functional categories as they either used
only one of the two aspects or mixed them without reflection. I attempt to clarify
the relationship between linguistic markers and functional categories and I pro-
pose a model that makes use of both groups at the same time. In other words, I
will use a combination of the models that are based on function and the one which
consists of linguistic markers. Naturally, in my model, I strived for using the cat-
egories that seemed to be the most relevant and appropriate ones in order to avoid
the flaws of the previous models described earlier. Despite their imperfections,
previous frameworks of disagreement types serve as a starting point for mapping
the functional spectrum of disagreements. Namely, the functional categories were
created and assigned to the tokens of disagreement using a mixture of Muntigl &
Turnbull’s (1998), Rees-Miller’s (1995, 2000), Locher’s (2004), and Bándli’s model
with some modifications and additions.6 Similarly to Rees-Miller (1995, 2000), the
disagreement strategy in my analysis is organised first into three broad categories,
labelled as ‘softened’, ‘neutral’ (equivalent to Rees-Miller’s category of ‘not softened
or aggravated strategies’) and ‘strengthened’ strategies. As in my analysis the focus
is on the functions of disagreement rather than on the linguistic markers, I created
subcategories based on the functions and assigned (a) function(s) to each token of
disagreement. After manually indexing all tokens of disagreement in the corpus, I
examined them thoroughly and I managed to identify the linguistic markers that
usually help with the expression of, and co-occur with, the given strategy. The
strategies and the linguistic markers are illustrated in Table 1 and will be defined
and further illustrated on the following pages.

3.2.1  Softened disagreement strategies

Similarly to Rees-Miller’s (1995, 2000) taxonomy, the category of softened dis-
agreements includes strategies that are softened with one or more mitigating
devices. The illocutionary force of disagreement can be softened by the following

.  The following intonational symbols are adapted from Varga (1985) and Németh T. (1996)
in a simplified form: /● – rising intonation, \● – falling intonation, /\● – rise-fall intonation,
\/● – fall-rise intonation, /\/● – rise-fall-rise intonation, ║ – intonation phase boundary.
.  The functions of disagreement reported in previous research were revised and comple-
mented on the basis of the patterns found in the research corpus.
 The development of a taxonomy of verbal disagreements in the light of the p-model 
Table 1.  Disagreement strategies and linguistic markers
Disagreement strategies
1. Partial agreement, token
agreement (avoid disagreement)
2. Explanation (give/ask for
reasons, give example)
Softened disagreement
3. Joke
4. Ask questions
5. Hedge
6. Impersonalise S to H, point-
of-view distancing
7. Interject
8. Contradictory statement
Neutral disagreement
9. State disagreement
10. Clarify S’s meaning
11. Verbal shadowing
12. Use mockery (irony, sarcasm)
13. Intensify disagreement
Strengthened disagreement
14. Ask rhetorical questions
15. Challenge
16. Judge
17. Show disbelief, play down
previous statement
Linguistic items
(Hát) igen/jó/oké, de… / (Well) yes/fine/OK, but…
Valamennyire igen, de… / To some extent yes, but…
… mivel/(azért) mert… / … since/because…
Miért…? / Why…?
none
none
Hát…, / Well,
-hat, -het / can, could
(Hát) szerintem / (Well), I think
valamennyire / to some extent
lehet, hogy / maybe/perhaps
m:, ö: / uhm:, u:h
Azt mondják… / It is said that
Egyesek szerint… / According to some people…
M:?, Ha!, Á:! / U:hm?, Ha!, A:h!
negation (Ez nem igaz!, Ne!, Nem. / This is not true!,
No!, Not…)
linking word expressing contrast (De/Viszont…, /
But/However,…)
Én ezzel nem értek egyet. / I don’t agree with this.
Ebben nem értünk egyet. / We don’t agree on this.
Nem erre gondoltam. / I didn’t mean that.
Nem úgy… /Not that way…
Nem arra értem. / I didn’t mean it that way.
Én nem erről beszélek. / I’m not talking about that.
none
emotive language (Ó:, igen! / Yea:h, right., Az igen!
/ Wow!)
egyáltalán nem / not at all
sokkal / much more
abszolút / absolutely
teljesen / totally
akkor is / still
mondom / I’m telling you
none
interjections
Judgemental vocabulary
(hülyeség/marhaság/baromság, butaság / nonsense,
rubbish, bullshit, silly,
seggfej / asshole)
Á: / A:h
Ugyan már! / Come on!
 Helga Vanda Koczogh

strategies: partial or token agreement, giving or asking for examples or reasons,

joking, asking questions instead of using statements, hedges,7 point-of-view dis-
tancing, and the use of interjections. In softened disagreements, one of these or
a combination of some of these strategies are used. The category of softened dis-
agreements corresponds to the Brown and Levinsonian categories of ‘on record
with redress’ and ‘off record’. Softened disagreements typically occur in contexts
where face considerations of the other take priority over the need to convey dis-
agreement. In what follows, I will provide the definition and illustration of the
softened disagreement strategies identified in the research corpus.
Partial agreement occurs when the speaker makes a concession before express-
ing disagreement in order to soften the force of the utterance. The most common
forms of partial agreement are expressed by Igen/Jó/Oké, de… (~ Yes/Fine/OK,
but…) optionally prefaced by the discourse marker hát (~ well). For example, in
the discussion of the characters of a story the female interlocutor (F03)8 disagrees
with her boyfriend the following way:

Example 19
M03 [Hát a Vőlegény és a Bölcs között vacilál]tam, de a Vőlegény
szerintem az, aki aki igazából az egész történetben semmi semmi
rosszat nem tett.
[Well I was hesitating between the Wiseman and the Fiancé], but
I think the Fiancé is the one who who actually didn’t didn’t do
anything wrong in the whole story.
→ F03 \●Jó de a \●Bölcs se║. \●Sőt ő ő \●jó tanácsot adott neki║.
Fine, but neither did the Wiseman. Actually, he he gave him good
advice.

.  Linguistic devices with the function of softening the force of face threat to the addressee.
.  Speakers in the research corpus were given a code consisting of a capital letter indicating
the sex of the speaker, and a two-digit number that is identical in the case of conversational
partners.
.  The transcription conventions (modified from Tannen 1984 and Du Bois et al. 1992) used
in the examples are as follows:. – falling intonation with a conclusion point;, – continuing
intonation; ? – rising or question intonation; ! – strong feelings, high volumes, sentence–final
exclamatory intonation; ??? – unclear or unintelligible words;:vowel/consonant – lengthening
of sound; … – short pause; … … – medium or long pause; @@@ – laughter; TEXT – emphatic
stress; “text” – direct quotations read out loud from the task sheet; [text] – speech overlap;
(text) – transcriber’s descriptive comments on speech quality (i.e. irony, sarcasm, etc.); {text} –
transcriber’s comments added for clarity; → – lines relevant to the analysis at hand.
 The development of a taxonomy of verbal disagreements in the light of the p-model 

In this example, F03 begins her utterance with an agreeing jó (~ fine) before con-
tradicting M03’s utterance and elaborating on this contradicting statement.
Token agreement is a polite disagreement disguised as an agreement. It occurs
when the speaker appears to agree with the previous speaker in order to save
the addressee’s positive face. It is reportedly one of the most common strategies
(Bándli 2009; Brown & Levinson 1987; Holtgraves 1997; Locher 2004; Rees-Miller
1995, 2000) and often expressed the same way as partial agreement.10
Another softened disagreement strategy identified in my corpus is explana-
tion. This umbrella term refers to an utterance that gives or asks for a reason or
example to indicate that the speaker cannot accept the previous speaker’s proposi-
tion. In Example 2, the female speaker contradicts the previous utterance indi-
rectly, by giving a counter-example:
Example 2
Talking about gender differences in cooking skills
M05 A nők sokkal jobban el tudják találni szerintem az ízeket is.
I think women are much more sensitive to flavours.
F05 \●Hát mondjuk … /●például a halászlét azt \●nálunk általába a
→
\●Fater főzi║.
Well, let’s say … for example fish soup is cooked by my father in
my family.

When asking for a reason, the speaker’s interrogation is frequently preceded by the
contrastive marker de (~ but) and/or accompanied by a critical or doubtful tone.
Although the contrastive marker de expresses opposition in itself, these utterances
would generally be identified as disagreement either because of the intonation or
the divergent propositional content. The following example illustrates this point:
Example 3
M12 Én egyetértek ezzel az állítással.
I agree with this statement.
→ F12 \●De MIÉR║?
But WHY?

In the last line the emphatic MIÉR (~ WHY) alone clearly conveys the message of
disagreement, which is strengthened by the use of de (~ but). The falling intona-
tion on the question also reinforces disagreement.

.  Since token agreements and partial agreements are closely-related concepts with only a
thin borderline between them, I grouped them under the same category.
 Helga Vanda Koczogh

Providing a reason or explanation can also mitigate disagreement, as it is ori-

ented to the hearer’s positive face, and is often marked by conjuctions such as mert
(~ since, because) or mivel (~ as), as illustrated by Example 4 below:
Example 4
F08 „Mindenkinek feltétlenül be kellene oltatnia magát az új járványok
ellen.”
“Everybody should definitely get themselves vaccinated against
new diseases.”
→ M08 Hát szerintem /●nem║ mivel \●én se \●oltattam be magam║.
Well, I don’t think so, since I haven’t had myself vaccinated either.

Giving an explanation or listing reasons is noted as a frequently used disagreement

strategy in Bándli’s (2009) study as well.
Joking is another positive politeness strategy used for softening disagreement.
It is included in Brown & Levinson’s (1987) and Rees-Miller’s (1995, 2000) frame-
work, but other researchers have also investigated the relationship of humour and
disagreement in verbal interactions. Joking can prevent conflict, as it puts the
addressee at ease and often generates laughter on the part of the speaker and/or
addressee. Consider the following example:
Example 5
F05 Mér, a férfiak azok meg versenyeznek! A nők nem versenyeznek
annyira.
Why? Men race. Women don’t race so much.
→ M05 \●Figyelj már║ \●szerinted egy egy \●hatvan éves tata /\●szerinted
versenyzik║? Örül hogyha /\●beül a kocsiba║. (nevetve)
Listen, do you think a a 60-year-old fogey races? He is happy if he
can get in the car. (laughing)

In this extract speaker M05 uses humour to buffer the effect of his disagreement
and utters his opinion in a non-serious, joking manner accompanied by laughter.
This utterance creates a light-hearted atmosphere and results in the laughter of the
listener, which indicates that an atmosphere of camaraderie has been established
through the use of joking. There is no identifiable linguistic item that generally co-
occurs with the strategy of joking, but the analyst can rely on other features such
as laughter and tone.
Expressing objection in the form of a question rather than a direct statement
is a softened disagreement strategy noted by both Rees-Miller (1995, 2000) and
Locher (2004). As explained by the latter (2004: 133), the indirect strategy of using
an interrogative form for implying disagreement can “express relational work and
hence face considerations (for S and H) of various degrees if it lacks the quality of
 The development of a taxonomy of verbal disagreements in the light of the p-model 

a challenge” (emphasis added). In the following example, the interrogative form

is considered to be more polite than simply stating that the Fiancé has nothing to
do with it.
Example 6
M08 Szóval szerintem ez á ugyanazért adtam rossz pontot neki, mint a
Szépleánynak, mert balfasz (nevetve). Szerintem legalább [is].
So I think this ah I gave him a bad score for the same reason as
for the Pretty Girl, because he is a screw-up (laughing). At least
according to [me].
→ F08 [Hát] de most ő \●mit tehet arról║?
[But] what could he do about it?

In Rees-Miller’s (1995) study, the use of interrogative form is the second most
common mitigating device. It occurs in two forms: yes/no questions and seem-
ingly information-seeking wh-questions used to express disagreement. The sec-
ond type, as well as negative questions, is reported to have been used for presenting
contradictory evidence or competing alternatives (pp. 100–105). I also consider
the use of questions (other than requests for reasons or examples and rhetorical
questions) a softened disagreement strategy, however, I categorise negative ques-
tions as challenges since they express a higher degree of adversarialness and they
question the other interlocutor’s stance. Similarly to joking, there is no identifiable
linguistic marker of the strategy of using a question, only the opposing proposi-
tional content and sometimes the tone of the question yield some clues.
Hedges can also be used to mitigate disagreement. The classification of miti-
gating devices proposed by Rees-Miller (1995) outlines the following four groups:
personal pronoun I followed by a cognitive verb such as think, question forms,
downtoners (e.g. sort of, a little, sometimes), and verbs of uncertainty (e.g. may,
seem). Locher (2004) reports hedging to be the most common softened disagree-
ment strategy in her research and provides the following exhaustive list of hedges
as an extension of Aijmer’s (1986) list:
actually, anyway, as it were, basically, a bit, certainly, honestly, I mean, I think, in
a way, in fact, just, kind of, let me, little, maybe, more or less, of course, perhaps,
probably, say, see, so-called, somehow, sort of, stuff, suppose, type of, uh, uhm, well,
whatever, what you call; or and and in final position. (p. 115)

It is important to note here, however, that although these linguistic items do tend
to lessen the impact of opinion and disagreement on the listener, there is no direct
correspondence between form and function. Some of these devices (e.g. discourse
markers) are multifunctional, so they can fulfil several functions based on the con-
text. Hence, it is crucial to interpret the meaning and function of the given item in
 Helga Vanda Koczogh

its context in each case. To illustrate this point, there are linguistic devices which
can act both as a hedge and as a booster, a point also noted by Holmes (1995: 78).
Hedges and boosters can combine with each other to soften or strengthen the
force of the utterance more effectively, a phenomenon reported by other research-
ers (e.g. Holmes 1995; Locher 2004; Rees-Miller 1995, 2000).
The last but one softened disagreement category includes the strategies of
impersonalise speaker and hearer and point-of view distancing. These strategies are
similar in the sense that they both aim at shifting responsibility, that is, making an
utterance appear as coming from a different source. This can be achieved by using
passive structures (see Example 7) or general subjects (see Example 8) instead of
the first person pronoun I. These negative politeness strategies mitigate disagree-
ment as well as protecting the face of the speaker. Locher (2004: 130) calls this
strategy ‘shifting responsibility’ and argues that its advantage in a conversation is
that “the content of what a speaker just said might be debatable, but the person as
such is not as exposed to criticism as when he or she had reported the content as
his or her own point of view”.

Example 7
M05 Szerintem a férfiak SOKKAL intelligensebbek a nőknél.
I think men are MUCH MORE intelligent than women.
→ F05 De ö: /●azt mondják║ hogy azér egy nő /●tud több dologra figyelni
egyszerre║ … \●jobban használja agyának a ké /●mindkét felét║…
But u:h it is said that women are able to pay attention to more things
at the same time … they use both of both sides of their brain…

Example 8
F01 Szerintem mindenkinek be kellene oltatnia magát H1N1 ellen.
I think everyone should get themselves vaccinated against H1N1.
M01 \●Egyesek szerint /●csak azt oltsák be aki aki aki /●veszélynek van
kitéve║ tehát a /●terheseket║ az /●időseket║ és aki \●veszélyeztetett
területen dolgozik║.
According to some people only those should be vaccinated who
who who are at risk, that is, pregnant women, the elderly, and those
who work in a risky area.

The last softened disagreement strategy is interject, which is conventionally

expressed by words or phrases that convey emotion on the part of the speaker (e.g.
Ha! ~ Ha!, Á: ~ A:h, M: ~ U:hm) typically placed at the beginning of an utterance.
The meaning of these interjections depends on their intonation contour as well
as context. Using interjections rather than stating disagreement directly is a safer
and more polite way of expressing disagreement. Consider the following example:
 The development of a taxonomy of verbal disagreements in the light of the p-model 

Example 9
F03 És nézd meg a világ legjobb … ö:… legjobb szakácsait, [szinte
mind férfi].
And look at the world’s best … u:h … best chefs, [almost all of
them are male].
M03 [Na jól van]
[Ok fine]
F03 És ott van XY milyen jól főz!
And there’s XY, what a great cook he is!
→ M03 /●M:? (szkeptikus)
U:hm? (sceptical)

Here the sceptical indirect answer m: (~ u:hm)11 uttered with a rising intonation is
considered to be more polite than simply stating Nem. (~ No, he isn’t).

3.2.2  Neutral disagreement strategies

The second broad category of disagreement strategies I label neutral strategies,
indicating that no facework is done in order to make the utterances sound more
polite or impolite. In Rees-Miller’s (1995, 2000) framework this group is referred to
as ‘neither softened nor strengthened’ and sometimes as ‘not softened or strength-
ened’ disagreement, while Locher uses the term ‘non-mitigating disagreement
strategies’. These strategies include straightforward disagreements, where there is
no attempt on the part of the speaker either to lessen or strengthen the force of dis-
agreement. Neutral disagreement strategies include the following categories, each
of which will be discussed and illustrated below: contradictory statement, stating
disagreement, clarifying speaker’s meaning, and verbal shadowing.
Contradictory statement occurs when a speaker disagrees by uttering the
negated proposition expressed by the previous claim. This can be expressed by
the single word utterance Nem. (~ No.), by repeating the previous utterance in a
negated form, or by uttering a direct statement such as Ez nem igaz. (~ This is not
true.). This type of disagreement corresponds to the category of ‘direct disagree-
ment’ in Rees-Miller’s (1995, 2000) taxonomy. Once again, it must be noted that
we cannot always equate one form with a single function. For instance, no does
not always mean disagreement and neither does yes mean agreement in every case.

.  Hums in Hungarian serve a broad range of discourse functions such as signalling agree-
ment, disagreement, surprise, hesitation, doubt, scepticism, urge, etc. depending on their pro-
sodic features and the context of talk. For instance, while affirmative hums are associated with
a descending-rising intonation, interrogative hums feature rising intonation (Markó 2007: 30).
For a functional analysis of humming in Hungarian see Markó (2005, 2007).
 Helga Vanda Koczogh

Another common means of expressing disagreement is uttering a contradic-

tory statement prefaced by a linking word expressing contrast such as de (~but)
or viszont (~ however), as in the following exchange about gender differences in
intelligence:
Example 10
F09 Hát igen, de ha egy fiú hiába ilyen elektronika felé érdeklődik, ha
kell, akkor ő ugyanúgy meg tudja tanulni az irodalmat vagy egy
nyelvet vagy bármit. De a lány, mint például én is, én nem tudnám
megtanulni se a fizikát, se a kémiát.
Well yes, but even when a guy is interested in electronics, if he
has to, he can study literature or a foreign language or anything.
However, a girl like me, I wouldn’t be able to study either physics
or chemistry.
→ M09 De ebbe /\●nem lehetsz biztos szerintem║. /●Tehát║
But you can’t be sure of that, I think. So

In the above example, speaker M09 expresses his disagreement with the contra-
dictory marker de (~ but) in utterance-initial position followed by a statement
whose propositional content expresses disagreement. Contrastive markers, by
themselves, express opposition (ebbe nem lehetsz biztos szerintem ~ you can’t be
sure of that, I think). In this case de (~ but) signals contrast, but without its use the
utterance would still be recognised as disagreement.
The third subtype of contradictory statement is when the utterance, by itself,
“does not bear any markers of disagreement. However, in the context in which
it occurs, it contradicts a previous utterance” (Rees-Miller 1995: 116). I would
add to this definition that, naturally, this type of disagreement does not neces-
sarily have to contradict the whole of the previous proposition; the contrasting
information can refer to only a part of it. In this type of disagreement, contradic-
tion is not stated but implied and so understanding the illocutionary force often
involves more mental processing on the part of the listener. Consider the following
example:
Example 11
M15 Főzési tudományom a tojásrántottában ki is merül.
My knowledge of cooking ends with scrambled eggs.
F15 De abba teszel jó sok mindent, igaz?
But you put a lot of things in it, right?
→ M15 \●Tojást║.
Eggs.

In the last line of this interchange the disagreement is expressed with a single word
(Tojást. ~ Eggs.) without any linguistic markers of opposition. It is an indirect way
 The development of a taxonomy of verbal disagreements in the light of the p-model 

of saying No, I put only eggs in it, and is understood by virtue of the Gricean maxim
of relevance.12
Another neutral disagreement strategy is what I term stating disagreement.
Stating disagreement, as it is used here, means an utterance that explicitly states that
the speaker disagrees with the previous proposition, as illustrated by Example 12.
Example 12
F02 “Az új járványok ellen mindenkinek feltétlenül be kellene oltatnia
magát.”
“Everyone should definitely get themselves vaccinated against new
diseases.”
→ M02 Hm /●nem értek vele egyet \●egyáltalán║.
Uhm I don’t agree with this at all.

This type of disagreement always contains the expression nem ért egyet, which
contains the performatives agree or disagree in its English translation, thus con-
veying the message with maximum efficiency. Interestingly enough, this category
does not appear in either Rees-Miller’s (1995, 2000) or Locher’s (2004) schema at
all. This might be due to the relative infrequency of the use of this strategy in the
context of verbal interactions in English.13 However, what is more remarkable is
that Bándli (2009) and Mászlainé Nagy (2010) do not make mention of this dis-
agreement strategy in Hungarian either.
Clarifying speaker’s meaning is another recently discovered disagreement
strategy emerging from the research corpus that does not appear in previous liter-
ature. It has the function of clarifying the misunderstood meaning of the speaker’s
previous utterance, which therefore contradicts or corrects the other interlocu-
tor’s previous proposition. This strategy has a constructive effect on conversation,
as it helps to avoid communication breakdowns and facilitates the smooth flow
of interaction. In the following example taken from a dispute on euthanasia, the
female speaker (F04) disagrees with her boyfriend (M04) twice because she wants
to explain what she meant.

.  Speaker M15’s brief but informative utterance also observes the maxims of quantity and
manner.
.  I investigated the occurrence of the expressions I don’t agree, I do not agree and I disagree
in the British National Corpus (BNC) Spoken and in the corpus of US TV Talk (with the help
of Corpus Concordance English (CCE) program) with a volume of 100 million and 2 million
words respectively. My queries resulted in zero hits in CCE for each of the searches, while I
found no tokens for the first, a single token for the second, and only 20 tokens for the third
expression in BNC Spoken. The contexts of hits were mainly council meetings, teacher’s con-
ferences, court hearings and public debate programs.
 Helga Vanda Koczogh

Example 13
M04 De én azt mondom, hogy ehhez egy nagy adag pesszimista
­világnézet is kell. Le kell tojni önmagunkat néha ahhoz, hogy ö:
… hogy hogy ilyen ilyen estekben hogy úgymond ‘meg akarok
halni, mert izé ronda vagyok, meg itt a púp a hátamon’. (irónikus)
But I say that for this you need a great deal of pessimistic
worldview. Sometimes we need to get over ourselves so that u:h …
that that in such such cases like ‘I want to die because I’m ugly and
there is a hunch on my back’. (ironical)
→ F04 De /\●nem úgy Maci║ én \●nem erről beszélek║.
But it’s not that Honey Bear, I’m not talking about that.
M04 Tudom, ilyen rokkant esetekbe is, hogy … például…mi miről is
akar …
I know in the cases of disabilities that … for example… wha what
did I want to
→ F04 \●Nem erre gondoltam \●Macika║! Ö: \●mondjuk egy /\●fél even
belül /\●rákban meghalsz║. […]
I didn’t mean that, Honey Bear. U:h let’s say you are going to die of
cancer in half a year. […]

Similarly to the previous strategy, this one is not mentioned in other taxonomies of
disagreement strategies either. In the cases of Rees-Miller (1995, 2000) and Locher
(2004), this might be explained by the circumstance that their studies are con-
cerned with how disagreement is (linguistically and non-linguistically) expressed
rather than with the functions of particular utterances.
The last neutral disagreement strategy identified in the research corpus is
labelled verbal shadowing by Rees-Miller (1995, 2000). By this term I mean the
act of repeating the previous speaker’s words, phrases or entire utterance (some-
times with slight alteration) to question the content of the previous utterance.
I take verbal shadowing in a narrower sense than Rees-Miller (1995), whose
interpretation allows for significant alterations in single words in the opposing
utterance. It is noteworthy that in some cases only the intonation contour is
changed to express disagreement and often criticism, as exemplified in the fol-
lowing interchange from the dispute about gender differences in people’s sense
of direction:

Example 14
M05 Menjünk tovább! „A férfiak intelligensebbek a nőknél.”
Let’s move on. “Men are more intelligent than women.”
 The development of a taxonomy of verbal disagreements in the light of the p-model 

F05 Hát SZERINTEM ez EGYÁLTALÁN nem igaz. Például most

mondják azt is, hogy a férfiaknak jobb a tájékozódó képességük
oszt mégis eltévedsz, én meg odatalálok.
Well I DON’T THINK this is true AT ALL. Now for example they
also say that men have a better sense of direction but you still get
lost while I find my way there.
M05 Ó: nem igaz! Jó, most néha eltévedek, [igen], de alapból a
férfiaknak
O:h that’s not true! Fine, I do lose my way sometimes, [yes], but
men fundamentally have
→ F05 [\●Né:ha║?]
[Someti:mes?]
M05 jobb a tájékozódó képessége.
a better sense of orientation.

In this heated debate speaker F05 expresses her disagreement several times, but
in the marked line she does this by simply picking up and repeating a single word
from the previous utterance (Né:ha? ~ Someti:mes?) and adding a critical tone to it
with the vowel lengthening and falling intonation.

3.2.3  Strengthened disagreement strategies

The third group of disagreement strategies include disagreements that not only
lack mitigation, but whose force is strengthened with the aims of making dis-
agreement direct, defending the speaker’s point of view more successfully, and
sometimes even assaulting the interlocutor. Strengthened disagreement strategies
include the following: (1) use mockery (irony, sarcasm), (2) intensify disagree-
ment, (3) ask rhetorical questions, (4) challenge, (5) judge, (6) show disbelief,
play down previous statement. Brown & Levinson (1987) did not consider the
possibility of aggravating an FTA, so they did not include strengthened strate-
gies in their framework. They only list bald-on-record strategies, which are direct
ways of expressing an FTA. However, there are significant qualitative differences
between bald-on-record strategies and the strengthened disagreement strategies
listed above. The latter go beyond Brown and Levinson’s bald-on-record strate-
gies, as they have a higher face-threatening force due to their aggravated nature
and because they often cause offense. Strengthened disagreements can occur in
contexts where defending one’s point of view is more important than protecting
the addressee’s face, or when the speaker wishes to be rude or hurtful (Culpeper
1996, 2011). The rest of this section aims at describing and illustrating strength-
ened disagreement strategies and their linguistic markers, if any.
 Helga Vanda Koczogh

Use mockery is a strengthened disagreement strategy in which the real intent

of the speaker is “concealed or contradicted by the literal meaning of words or
a situation” (Encyclopedia Britannica Online, 2006), but is indicated by tone of
voice. This disagreement strategy is manifested in the research corpus in the forms
of irony and sarcasm. Irony in early politeness theories (e.g. Brown & Levinson
1987; Grice 1975; Leech 1983) was considered as an indirect face-saving tool; how-
ever, since then it has been suggested that irony can also be used to attack face.
The use of irony is an impoliteness strategy called ‘mock politeness’ in Culpeper’s
(1996, 2005, 2011; Culpeper et al. 2003) framework, whose primary aim is to cause
social disharmony. In contrast, ‘mock impoliteness’ is employed with the goal to
establish and maintain intimacy. Accordingly, irony is not always used antisocially
and has different purposes regarding politeness or impoliteness. Its interpreta-
tion depends on various factors, such as the social distance and power relations
between the interactants. Generally, it can be said that irony tends to serve as a
device reinforcing solidarity among equals, typically friends, while it usually is
a means of attacking face among unequals such as strangers or in situations of
one-upmanship (Bousfield 2008; Culpeper 2011). Thus, irony is a highly complex
phenomenon which can hardly be defined exhaustively.14 Its central element is a
discrepancy between intended meaning and actual wording as well as the expres-
sion of (mostly negative) emotion or attitude (Nauke & Braun 2011: 1450). In my
analysis I follow Komlósi’s (2013: 412) definition of irony, according to which with
ironic utterances the speaker intends to raise the hearer’s metapragmatic aware-
ness and draws attention to the deflection and difference between what has been
said and at least one suggested meaning, while maintaining the ambiguity of the
pragmatic meaning.
The terms irony and sarcasm are closely-related; they are frequently used
interchangeably and they sometimes even overlap (e.g. sarcastic irony) (Toplak &
Katz 2000). They are often confused with each other since both are indirect forms
of speech. Irony and sarcasm are special types of mockery. Delienating the differ-
ence between sarcasm and irony would require a separate paper, but here it will
be sufficient to say that irony is hidden mockery, while sarcasm is an open one. It
is hard to differentiate irony and sarcasm even on the basis of prosodic features,
since they both share the common characteristics of a sudden rise (or peaking) –
fall intonation, faster rate of speaking, prolonged syllables, and intensified stress
on syllables (Fónagy 1971; Fónagy & Magdics 1967).

.  For a comprehensive overview of approaches to irony and other related phenomena (e.g.
sarcasm, cynycism, mocking) see, for example, Komlósi (2007, 2010, 2013) or Svindt (2007,
2008).
 The development of a taxonomy of verbal disagreements in the light of the p-model 

Irony and sarcasm are frequently expressed or accompanied by emotive lan-

guage such as Ó: (~ O:h) or Ó: igen (~ Yea:h, right.), as illustrated in the following
example:
Example 15
F05 @@@ Hát jó, de a Révésznek akkor sem kellett volna
@@@ Well fine, but still the Ferryman shouldn’t have
M05 De lehet, hogy a Révész és a Haramiák haverok voltak.
But maybe the Ferryman and the Thieves were friends.
→ F05 /\●Ó: igen║!
Yea:h, right.

In this extract about two characters of a story, speaker M05’s non-serious proposal
of the two characters being friends is rejected by speaker F05’s ironic utterance
(Ó: igen! ~ Yea:h, right.).
Intensify disagreement is another strengthened disagreement strategy,
one which makes use of certain words or expressions (e.g. egyáltalán ~ at all,
abszolút ~ absolutely, teljesen ~ totally, sokkal ~ much more, mondom ~ I’m tell-
ing you, etc.) to emphasise and strengthen the effect of disagreement, a strategy
also identified by Rees-Miller (1995, 2000). I refer to these linguistic items as
intensifiers as they increase the degree of intensity of disagreement. Expres-
sions such as szerintem (~ I think), csak (~ just), hát (~ well) also belong here
when they are used as boosters, since they intensify or emphasise the force
of the utterance and express certainty on the part of the speaker. Intensified
disagreements are frequently accompanied by paralinguistic features such as
volume amplification of the intensifier and quick tempo applied to the open-
ing phrase (e.g. hát ~ well, szerintem ~ I think). In some instances, intensifiers
are used in clusters. The second line (Hát SZERINTEM ez EGYÁLTALÁN nem
igaz. ~ Well I DON’T THINK this is true AT ALL.) of Example 14 is a case in
point. In this utterance three tokens of intensifiers (hát ~ well, szerintem nem
~ I don’t think, egyáltalán ~ at all) are adopted with emphatic stress on two of
them to reinforce the speaker’s oppositional intent and make her disagreement
prominent.
Another form of strengthened disagreement is the use of a rhetorical question,
a question asked not to evoke an actual reply but to imply that there is no answer to
the question or that the answer is too obvious. In some cases it is used for the sake
of persuasive effect and the speaker provides an answer. “What makes a rhetorical
question an aggravated form of disagreement is that its basic purpose is to score
points against an opponent, not to further understanding or build consensus”
(Rees-Miller 1995: 137f.). Let us consider Example 16, taken from a discussion
about abortion and adoption.
 Helga Vanda Koczogh

Example 16
F02 Akkor örökbe lehet adni vagy akármi.
Then it can be given up for adoption or whatever.
M02 De az is szívtelenség eléggé.
But that’s pretty heartless too.
→ F02 \●Örökbe adni║? \●Bolond vagy║? /●Legalább lesz valaki║ aki
\●szereti║!
To give it up for adoption? Are you crazy? At least there’ll be
somebody to love him/her!

In the interchange two rhetorical questions are used one after the other with a fall-
ing intonation, followed by an explanation (Örökbe adni? Bolond vagy? Legalább
lesz valaki, aki szereti. ~ To give it up for adoption? Are you crazy? At least there’ll
be somebody to love him/her.). The adoption of rhetorical questions is intended to
send the other interlocutor the unmistakable signal that his previous statement
about giving up a baby for adoption is heartless is unequivocally rejected. Hence,
the deployment of this strategy clearly displays overt disagreement.
Challenge is a disagreement strategy which displays strong disagreement with
the prior proposition and typically has the syntactic forms of either imperative or
interrogative (both rhetorical and response-seeking questions). Rather than mak-
ing a specific claim, this type of disagreement questions the addressee’s stance and
implies that (s)he cannot provide evidence for his/her claim. This strategy is often
accompanied by phonetic stress, rapid tempo and an ironic tone, all of which high-
light the speaker’s oppositional stance. The following example, taken from a dispute
on gender differences in driving skills, illustrates how hostile challenges can be:
Example 17
M05 „A nők biztonságosabban vezetnek, mint a férfiak.” Az igen!
(irónikus)
“Women drive more safely than men do.” Wow! (ironical)
→ F05 \●Miért║? Szerinted /\●nem igaz║? (irónikus)
Why? Don’t you think it’s true? (ironical)
M05 Nem igaz! … A nők nem tudnak odafigyelni a vezetésre!
It’s not true! … Women can’t focus on driving!
→ F05 \●HA║…! És /●elmondanád hogy \●MIÉRT║?
HA…! And could you tell me WHY?

As this scene opens, speaker M05’s interjection Az igen! (~wow) uttered in an

ironic way clearly indicates his reluctance to accept the statement and creates an
antagonistic atmosphere. It evokes his girlfriend’s first challenging response imply-
ing strong disapproval (Miért? Szerinted nem igaz? ~ Why? Don’t you think it’s
 The development of a taxonomy of verbal disagreements in the light of the p-model 

true?). Then, he makes a blunt contradictory statement (Nem igaz! ~ It’s not true!)
followed by a categorical proposition (A nők nem tudnak odafigyelni a vezetésre! ~
Women can’t concentrate on driving!), which generates an even more hostile atmo-
sphere and elicits an antagonistic challenge (És elmondanád, hogy MIÉRT? ~ And
could you tell me WHY?) preceded by an interjection (HA!) expressing resentment.
The challenge in the last line of the conversation questions the stance of the other
and implies that he cannot back up his claim, thus, directly attacks his compe-
tence. Hence it can be stated that challenges direct a high degree of aggravation
to the other’s rationality, which also explains why this disagreement strategy is
claimed to be the second most face aggravating disagreement type in Muntigl &
Turnbull’s (1998) categorisation.
Judging is a strengthened disagreement strategy in which the speaker expresses
a negative evaluation of the previous proposition, often to the point of condem-
nation.15 In this strategy, interlocutors often use offensive language (‘judgmental
vocabulary’ in Rees-Miller 1995, 2000) such as hülyeség (~ rubbish), marhaság
(~ bullshit), seggfej (~ asshole), etc. to show their strong disagreement. The aggra-
vating adjectives are frequently utilised with word stress and/or falling or rise-fall
intonation to exert an emphatic effect on disagreement. Consider the following
example on higher education tuition fee:
Example 18
F06 „A felsőoktatási intézményekben be kellene vezetni a kötelező
tandíjat, de költségét a hallgató tanulmányi átlagától kellene
függővé tenni.” Szerintem ez teljesen korrekt és helyénvaló.
“Compulsory tuition fees should be introduced in higher
education, but the amount for a semester should depend on the
student’s grades”. I think this is absolutely fair and right.
→ M06 De hát ez /\●BAROMSÁG║.
But that’s RUBBISH.

In the above interchange speaker F06 expresses her agreement with the statement
she has just read out. The other interlocutor’s response in the subsequent turn
clearly conveys the message that he strongly disagrees with both the proposal and
the previous speaker and directly confronts her. The use of prefatory de (~but), the
word choice of baromság (~ rubbish) and the emphatic stress and rise-fall intona-
tion on the latter all magnify the opposition and make the disagreement explicit.
The last strengthened disagreement strategy identified in the corpus is show
disbelief, play down previous statement, a strategy in which the speaker expresses

.  Bándli (2009) also identifies this disagreement strategy and calls it ‘Judgement/­Qualification’.
 Helga Vanda Koczogh

that (s)he does not believe or doubts the previous proposition and therefore can-
not accept it.16 Disbelief is often accompanied and/or expressed by interjections
expressing scepticism (e.g. Á: ~ A:h) and emotive phrases such as kizárt (~ no way)
or Az ki van zárva (~ That’s out of the question), as exemplified in the following
extract, which is the continuation of Example 6.
Example 19
F08 [De] hát most ő mit tehet arról? {hogy balfasz}
[But] what could he do about it? {that he is a screw-up}
M08 M: olvasni kellett volna a sorok között.
U: hm he should have read between the lines.
→ F08 Jaj \●UGYAN [\●MÁ:R║!]
Oh, COME [O:N!]

Here, the functions of the last utterance are clearly to make the previous state-
ment appear absurd and to urge speaker M08 to abandon his supposedly foolish
proposition. The imperative force of the idiom ugyan már (~ come on), the use of
interjection jaj (~ oh), and the emphasis on the last two words with the prolonged
vowel articulation and the falling intonation contour on the phrase highlight the
speaker’s rejection of the previous statement and overtly express her opposing
stance. Similarly to challenges, this strategy often questions the other interlocu-
tor’s rationality, thus aggravating his/her face.

3.2.4  Odd-one-outs
It must be mentioned that I encountered some tokens of disagreement in my cor-
pus that could not be categorised based on my category system either. Consider
the following example taken from a discussion about two characters of a story:
Example 20
F03 […] Nekem ő {a Szépfiú} volt a legellen[szenvesebb.]
[…] For me he {the Pretty Boy} was the most dis[tasteful].
M03 [Nekem pedig] a Szépleány. Azért mert mert szerintem ő volt
a leggonoszabb azért, mert mert [megcsalta a … megcsalta a
vőlegényét.]…
[For me it was] the Pretty Girl. Because because I think she was the
meanest because because [she cheated cheated on her fiancé.] …

.  Bándli’s (2009) categorisation of disagreement strategies also includes a category called
‘doubts/uncertainties’, however, it accounts for utterances that express uncertainty and attempt
to make the other speaker unsure of his/her stance.
 The development of a taxonomy of verbal disagreements in the light of the p-model 

→ F03 [\●Látszik látszik hogy \●ellenkezően gondolkodunk║.] És a

\●Szépfiú║?
\●Hogy mondhat olyat neki║? /\●Mindent megtesz érte║ és \●akkor
azt mondja másnap hogy \●‘bocs’.
[It’s clear it’s clear that we are thinking differently.] And the Pretty
Boy? How could he say something like that to her? She does
everything for him and then the next day he says ‘sorry’.
In the above dispute speaker F03 dislikes the Pretty Boy the most and refutes the
proposal that the Pretty Girl is the most distasteful character in the story. First,
she makes a metastatement that she and the other interlocutor are on a differ-
ent opinion (Látszik, látszik hogy ellenkezően gondolkodunk. ~ It’s clear it’s clear
that we are thinking differently) followed by two rhetorical questions (És a Szépfiú?
Hogy mondhat olyat neki? ~ And the Pretty Boy? How could he say something like
that to her?) aimed at proving that she is right. The falling intonation instead of a
rise-fall pattern at the end of the two questions and the statement following them
provide us with the rhetorical interpretation of the questions. Using a metastate-
ment which implies that the participants disagree is a strategy that does not fit
readily into any of the identified disagreement strategies. The categorisation of
such instances of disagreement warrants further research.

3.3  Comparison to other solutions

Section 3 described and widely illustrated the framework used for identifying pat-
terned disagreement strategies deployed by Hungarian undergraduate students.
It also examined what linguistic items typically co-occur with each strategy. The
p-contexts versions described in Sections 2.1–2.4 have been taken as points of
departure for the classification of disagreement strategies and a few other cate-
gories (e.g. stating disagreement, clarify speaker’s meaning), in addition to those
identified in p-context versions 1–4, were created/introduced17 during the manual
indexing process of tokens of disagreement. With the help of these new catego-
ries the informational underdetermination decreased, although did not disappear
entirely. Considering all the above, the model proposed in the previous section
can be considered more elaborate than the ones in the starting p-context and can
account for more types of disagreement. Hence, it seems to be a better solution to
(P) than the p-context versions discussed earlier. However, as seen in the previ-
ous section, there are some occurrences of disagreement in the research corpus

.  motivated by data found in my corpus.

 Helga Vanda Koczogh

that even the newly proposed framework cannot account for. Thus, the extended
p-context is still informationally underdetermined. This finding triggers further
re-evaluation of the extended p-context and a new argumentation cycle.

4.  Argumentation cycle 2

4.1  Setting up a new p-context

In the course of indexing the tokens of disagreement, while applying both the
categorising system described in Subsections 3.2.1–3.2.3 and using my linguis-
tic intuitions as a native speaker, I encountered some difficulties. For instance,
I noticed that the vast majority of disagreements are multifunctional, so I could
often easily assign two and sometimes even three functions to a single token.
This might be partly explained by the realisation that there is a degree of overlap
between the categories of my framework and that boundaries are sometimes fuzzy.
To illustrate this point, let us consider Example 12 one more time, repeated here
for convenience.
Example 12
F02 „Az új járványok ellen mindenkinek feltétlenül be kellene oltatnia
magát.”
“Everyone should definitely get themselves vaccinated against new
diseases.”
→ M02 Hm /●nem értek vele egyet \●egyáltalán║.
Uhm I don’t agree with this at all.

In Subsection 3.2.2, the function of speaker M02’s response to the previous pro-
posal was categorised as stating disagreement; however, egyáltalán (~ at all) magni-
fies the opposition, so it deploys the intensify disagreement strategy as well. Along
similar lines, apart from the softened disagreement strategy of interject, the dis-
agreement (M:? ~ U:hm?) expressed in Example 9, also makes use of the strength-
ened strategy of showing disbelief, playing down previous statement. In a similar
vein, besides the discussed function, the disagreements uttered in Examples 518
(Figyelj már, szerinted egy egy hatvan eves tata szerinted versenyzik? Örül, hogyha
beül a kocsiba. ~ Listen, do you think a 60-year-old fogey races? He is happy if he
can get in the car.) and 1419 (Né:ha? ~ So:metimes?) can equally be interpreted

.  Cf. Subsection 3.2.1.

.  Cf. Subsection 3.2.2.
 The development of a taxonomy of verbal disagreements in the light of the p-model 

as examples of mockery. Furthermore, Example 720 illustrates a clear-cut case of

impersonalisation (De ö: azt mondják, hogy azér egy nő tud több dologra figyelni
egyszerre…jobban használja agyának a ké mindkét felét… ~ But u:h it is said that
women are able to pay attention to more things at the same time…they use both sides
of their brain…), yet this opposing utterance can also be understood as a counter-
example of men’s mental superiority over women.
As a result of all of the above considerations, a cyclic, prismatic, and retrospec-
tive re-evaluation of the model was needed. In order to decide on the refinements,
I extended the p-context further with the intuitions of other native speakers of
Hungarian.

4.2  Elaboration of the p-context

4.2.1  Extension of the p-context: Research on the intuitions of native speakers
It was important to undertake the research with participants of a similar back-
ground to that of the subjects in the research corpus. Therefore, I analysed the
answers gained from 100 Hungarian university students21 aged between 18 and
24, equally divided by gender. Their task was to simultaneously read and listen
to extracts from the research corpus each of which contained occurrences of dis-
agreement and to assign these tokens of disagreement to the strategies listed on
their answer sheet. Informants were also asked to indicate the level of confidence
of their answer in each case on a 4-point Likert-type scale ranging from uncertain
(that is, just a guess – value 1) to absolutely sure (value 4).
Some minor changes were done in the coding scheme for ease of understand-
ing. These involved the renaming of some functional categories (e.g. verbal shad-
owing to repetition), reducing the number of categories with the help of merging
some loosely related functions (e.g. mockery, interjection and repetition into the
category echo,22 challenge, label/evaluate and doubt/disbelief under the umbrella
term meta-communication) while omitting others (e.g. impersonalise S to H, point-
of-view distancing and rhetorical questions on the basis of frequency of occurrence
in the corpus). As a result of the above-mentioned steps, I managed to reduce the
number of functional categories from 18 to 11 (including the category other). The

.  Cf. Subsection 3.2.1.

.  Informants were majoring in English (some of them Linguistics track students) or in
Communication at the time of research.
.  Echo is a relevance-theoretic term here and it includes the listed strategies since they
achieve their relevance “by expressing the speaker’s attitude to the views she tacitly attributes
to someone else” (Sperber & Wilson 2006: 621).
 Helga Vanda Koczogh

resulting categorisation system showing the abbreviations of the categories used

for indexing and a short description of the functions is presented in Table 2.

Table 2.  Categories for indexing in the research

Disagreement Abbreviation Category
type

P/TA Partial agreement, token agreement

EXPL Explanation (reason, example, question for
Softened clarification/reason)
HUM Humour
HED Hedge
CST Contradictory statement
Neutral STD Stating disagreement
CS’M Clarification of speaker’s meaning
ECHO Echo (mockery, interjection, repetition)
ID Emphasise/Intensify disagreement
Strengthened
META Meta-communication (challenge, doubt/disbelief,
label/evaluate,)
ELSE Other

The informants were presented with 23 excerpts, which included at least one
example of each function as well as two odd-one-outs (items 8 and 19); the latter
made use of strategies (impersonalise S to H and rhetorical question) that were not
listed among the categories. The aim was to discover whether the participants can
identify the functions of these tokens of disagreement, and, if not, what categories
they assign to them from the list.
Although I do not question the value of the answers with low confidence level
values, ignoring this information ensures the elimination of answers that might
have resulted from pure guesswork. Hence, only the answers with confidence level
values of 3 and 4 were included in the analysis presented below.
The research results are presented in two parts. First, the findings are dis-
cussed with reference to the answers that were given with a high level of confi-
dence indicating the individual items’ perceived level of difficulty. In the second
part, the research participants’ categorisation is compared with my intuition to
find out whether they match with each other. Some of the differences and the
problematic cases are discussed in the second half of this section.
As Table 3 illustrates, items 2, 22, 1, 18, 20, and 16 (in decreasing order)
received the highest number of answers (over 70) with a high confidence level
value. Naturally, this simply means that in the case of these tokens of disagree-
ment, participants were more confident about the function(s) they assigned to the
token, but it does not necessarily mean that their interpretations were correct. The
 The development of a taxonomy of verbal disagreements in the light of the p-model 

answers given to these items are relatively homogenous in terms of the assigned
functions. These disagreements are either expressed explicitly (e.g. extract 2) or
deploy the strategies of mockery (e.g. extract 18) and disbelief (e.g. extract 20).
The tokens of disagreement that were answered with a higher degree of per-
ceived difficulty are the following (presented in increasing order): items 5, 8, 13,
19. These tokens received fewer than 50 answers with a high level of confidence
and, with the exception of number 8, are more diverse with regard to the assigned
categories. Each of these items will be further investigated in the second half of
this section.

Table 3.  Research results (%) with confidence level values of 3 and 4
Extract Categories*
no.
P/TA EXPL HUM HED CST STD CS’M ECHO ID META ELSE

1 2.6 5.3 82.9 7.9 1.3

2 1.3 3.75 95
3 88.6 2.9 8.6
4 5.7 41.4 38.6 14.3
5 62 2.4 28.6 2.4 4.8
6 67.9 26.4 5.7
7 6.9 36.2 3.4 53.4
8 13.7 4.6 7 2.3 2.3
9 38.2 48.5 13.2
10 3.8 55.8 40.4
11 3.2 91.9 4.8
12 1.5 18.5 78.5 1.5
13 4.3 2.1 59.6 34
14 18.2 3.6 74.5 3.6
15 79.1 4.5 16.4
16 53.5 2.8 43.7
17 76.8 8.9 14.3
18 4.1 1.4 87.7 6.8
19 12.5 33.3 47.9 6.6
20 2.8 4.2 93.1
21 52.3 40 6.2 1.5
22 93.7 2.5 3.8
* The figures show the number of answers given to the item at hand (in percentages) with confidence level
values of 3 and 4. The numbers in bold show the function assigned to the given token by most participants.
The numbers in italics indicate the category/(ies) assigned to the given token by the researcher.
 Helga Vanda Koczogh

Table 3 provides the answers with confidence level values of 3 or 4 in percent-

ages. It shows which function(s) has/have been assigned to which disagreement
and by how many subjects (in percentages). The figures displayed in Table 3 match
with my indexing in the vast majority of cases (19 cases, 86%). This is remarkable
given that the research participants received only a brief introduction to the cat-
egories used in my framework. What follows is the discussion of examples where
a mismatch was found between the assigned functions.
As already mentioned, the fewest answers of high confidence level were given
to excerpt 5 illustrated by Example 21. According to 26 participants (62%), this
token of disagreement (De történhetnek is balesetek. ~ But accidents could still
happen.) deploys the strategy of explanation, that is, the force of disagreement is
softened by providing a reason. This interpretation seems acceptable, although
the other function of hedge is also manifested in the token expressed by the affix
-het, which corresponds to the English modal verb could. Only (28.6%) of the par-
ticipants assigned the latter function to this token, which indicates that laypeople
might be less aware of this function and/or it is less salient.
Example 21
On abortion
M05 És mi van ha félnének a gyerekvállalástól?
And what if they were scared of having a baby?
F05 Hát jó, de hát most aki nem akar, az védekezik, nem? Jó esetbe.
Well fine, but now those who don’t want it {to have a baby} they
use protection, don’t they? In an ideal case.
M05 Jó.
Fine.
F05 És ez a lényeg!
And that’s the point!
→ M05 De \/\●történhetnek is balesetek║!
But accidents could still happen!

Item 8 (Example 7) illustrates the strategy impersonalise S to H based on my intu-

ition and I was interested to learn what function participants would assign to it, as
this particular category was not listed on the answer sheet. Only one participant
(2.3%) marked its function under the category other, but unfortunately she did not
specify her interpretation. Most participants were unsure about the disagreement
strategy adopted in the example, but most of those who were more confident (n=36,
83.7%) assigned the category explanation to the utterance. Indeed, while disagree-
ing the speaker provides a reason why women are more intelligent, expressed as if
were coming from a different source. Therefore, I added the participants’ interpre-
tation to the analysis, while also keeping my own reading of the utterance.
 The development of a taxonomy of verbal disagreements in the light of the p-model 

Moving on to item 13, illustrated by Example 22, we can see that although the
function assigned by most participants (59.6%) corresponded to my categorisa-
tion (interjection), a considerable number of them (34%) interpreted the interjec-
tion M:?(~ U:hm?) uttered by speaker F04 as a way of expressing doubt or disbelief.
I must admit that in its broader sense disbelief can include such indirect utterances
as the one under discussion, especially when uttered with rising intonation.
Example 22
On death penalty
M04 Ez ebben egyet értek, hogy nem kéne alkalmazni. Ö: igaz, ö… van
egy fajta megfélemlítő ereje a halálbüntetésnek a bűnözők számára.
That this I agree with that it shouldn’t be used. U:h it’s true that
uh… death penalty has a kind of deterrent power for criminals.
→ F04 /●M:║?
U:hm?

Extract 19 (see Example 23) illustrates a heated debate on the character of the
Pretty Boy and contains a token of disagreement that was in the focus of my
interest in terms of its categorisation. According to my analysis,23 the function
of speaker M12’s utterance (MI? Az az, hogy kihasználta egy éjszakára? Átment,
[visszajött] ~ WHAT? That he had her for a one-night stand? He went over, [came
back]) in the last line of the excerpt is that of a rhetorical question; however, this
strategy was not listed among the categories on the task sheet.
Example 23
M12 Az nem lett volna gonoszság? Átmegyek aztán visszajövök
Wouldn’t it have been a mean thing? I go over then I come back
F12 De nem [ne:m]
But no [no:]
M12 [és otthagyja]. ME:RT?
[and he leaves her there.] WHY:?
F12 Hát mert akkor Ő tesz áldozatott a Szép[lányért]
Well because then HE makes sacrifices for the Pretty [Girl]
M12 \●MI║? Az az hogy /\●kihasználta egy éjszakára║? /●Átment║
→
[\●visszajött║]
WHAT? That he had her for a one-night stand? He went over,
[came back]

.  It should be stressed that my analysis is primarily based on the functions of disagree-
ment that were identified in previous literature and it is not based on intuition alone.
 Helga Vanda Koczogh

Only three subjects assigned the function other to this instance of disagreement,
but two of them interpreted it as a rhetorical question. Most of the participants
(47.9%) placed it into the meta category, identifying its function as a challenge.
Those subjects who assigned the function echo to it (33.3%) justified it with the
mocking tone. In my view, the tone adds a hostile attitude to the utterance, which
makes it a challenge rather than mockery. Also the interrogatives question the
stance of the other speaker in an antagonistic way, which is a characteristic feature
of challenges.

4.2.2  Coordination of the p-context

In general, it can be stated that there is a strong correlation between the functions
assigned to the instances of disagreement by myself (based on previous literature
and my linguistic intuition) and by the 100 research participants in most of the
cases. Nevertheless, the research did not constitute the final phase of the develop-
ment of my taxonomy. In some cases a significant number of the research par-
ticipants assigned an additional function to disagreements that I had disregarded.
Therefore, when I judged the new function to be justified, I added it to my analysis,
thus increasing the number of multifunctional instances of disagreement. How-
ever, this also makes my model less objective and more intuitive.
It is also worth mentioning that although most participants (53.4%) assigned
the same function (challenge from the meta category) to the token of disagree-
ment (HA! És elmondanád, hogy MIÉRT? ~ HA! And could you tell me WHY?)
in Extract 7 (see Example 17) as I did, a remarkable number of subjects (36.2%)
interpreted it either as an interjection (25.9%) or as mockery (10.3%). This drew
my attention to the possibility that participants focused on different parts of the
disagreement; while some felt the first part to be important, the others concen-
trated on the second half. This led me to conclude that my framework needed to
be reconsidered.
As a result of the additional functions yielded by the results of the research,
I reannotated the corpus one more time. While thinking about the findings
of the research and indexing the tokens of disagreement again, I came to the
following conclusion: as my taxonomy was based on previous frameworks on
disagreement, it inherited some of their weaknesses. The most important of
these shortcomings is that most researchers use fuzzy categorisation systems
due to the lack of differentiation between function and linguistic realisation.
In most of the cases when a token of disagreement was interpreted in two ways
in the research, only one of them is its genuine function while the other one is
the result of focusing on a linguistic item by which disagreement is expressed.
The latter does not change the token’s original function; it simply serves as a
modifying device.
 The development of a taxonomy of verbal disagreements in the light of the p-model 

Furthermore, the disagreements in my corpus are often expressed by a com-

bination of strategies, a finding noted by other researchers as well (e.g. Culpeper
2011; Culpeper et al. 2003; Muntigl & Turnbull 1998). In other words, in many
cases the first part of the disagreeing utterance makes use of a strategy and the
second part employs another one. Moreover, as mentioned earlier, the functional
categories do not seem to be transparent so they are subject to individual subjects’
interpretations. This also justifies the distinction between function and linguistic
items that only modify the force of the utterance. Within an evaluative perspec-
tive, we can conclude that the model proposed in the extended p-context does not
meet the criteria laid down in Section 2; thus, it cannot be considered as a solution
to (P). The above-mentioned issues motivated the continuation of the argumenta-
tion process in order to obtain a more reliable and refined framework. Thus, as the
following step, the taxonomy used for analysing disagreements was reconsidered
one more time. The new categorising system I set up is discussed in the following
section.

4.2.3  Modification of p-context

As a first logical step, a new analytical framework had to be set up that distin-
guished between content and form. For this purpose the coding system used
in CCSARP (Blum-Kulka et al. 1989) for analysing requests and apologies was
considered. Although the analytical framework of CCSARP coded both con-
tent (termed strategies) and linguistic markers such as conditionals, hedges, and
intensifiers, the coding scheme used for indexing requests and apologies was not
suitable for my corpus. This can partly be explained by the circumstance that the
strategies used for coding the above-mentioned speech acts were specific to them,
and thus could not be applied to disagreement. However, the taxonomy of lin-
guistic markers used in CCSARP provided a useful starting point for the further
development of my coding scheme.
The categories used in the previous version of my taxonomy were reconsid-
ered and divided into two groups: disagreement strategies and so-called pragmatic
force modifiers (hereafter PFMs). I use the term disagreement strategy in a meaning
synonymous to function and, similarly to Locher (2004), I do not wish to imply
that the use of these strategies was the result of conscious planning. Disagree-
ment strategies include the functional categories of disagreement types coding the
implied meanings of disagreements. The group of PFMs, on the other hand, con-
tains linguistic items (e.g. hedges, intensifiers) and other devices (e.g. rhetorical
question, humour) that modify the force of disagreement by either aggravating or
mitigating it. As a subsequent step, the list of PFMs was complemented with addi-
tional categories based on the taxonomy of linguistic markers used in CCSARP
and Brown & Levinson’s (1987) framework.
 Helga Vanda Koczogh

As the vast majority of the categories comprising my new system for coding
disagreements have already been introduced, what follows is only a short defini-
tion of each for convenience and some prototypical examples of reactions to the
utterance Men can cook better than women.24

Disagreement strategies:

–– partial agreement/token agreement: partial agreement is an utterance that

makes a concession before expressing disagreement in order to soften the
force of disagreement (e.g. Igen, de vannak kivételek. ~ Yes, but there are excep-
tions.; Jó, de vannak kivételek. ~ Fine, but there are exceptions.). Token agree-
ment is a polite disagreement disguised as an agreement (e.g. Lehet. ~ Maybe.;
Talán. ~ Perhaps.).
–– explanation (give/ask for reason/example): an utterance that gives (De azt
mondják, hogy a nagymamád és az anyukád főztje a legjobb a világon. ~ But
it is said that your grandmother’s and your mother’s cooking is the best in the
world.; Ha ez így lenne, akkor ők lennének a háziasszonyok. ~ If it was so,
they were the housewifes.; De ha megkérdeznek, akkor te is az anyukád főztjét
szereted jobban, nem az apukádét. ~ If someone asks you, you also like your
mother’s cooking more not your father’s.) or asks for (MIÉRT?~ WHY?) a rea-
son or example to indicate that the previous speaker’s proposition cannot be
accepted by the speaker. When asking for a reason, the speaker’s interrogation
is usually accompanied by a critical or doubtful tone.
–– contradictory statement: an utterance that expresses contradiction by either
negating the proposition expressed by the previous claim (e.g. Nem. ~ No.;
Nem főznek jobban, mint a nők. ~ They can’t cook better than women.; A nők
főznek jobban, mint a férfiak. ~ Women can cook better than men.) or directly
stating that it is not true (e.g. Nem igaz. ~ That’s not true.).
–– implied contradiction: an utterance that, by itself, “does not bear any markers
of disagreement. However, in the context in which it occurs, it contradicts a
previous utterance”25 (e.g. Teát! ~ Tea!; Én meg tudok a vízen járni! ~ And I can
walk on water!). In this type of disagreement, contradiction is only implied,
and thus understanding the implicature requires more mental processing on
the part of the listener.

.  Except for the last one (clarification of speaker’s meaning), where an additional turn
needed to be inserted the disagreeing utterance reacts to.
.  Rees-Miller (1995: 116).
 The development of a taxonomy of verbal disagreements in the light of the p-model 

–– stating disagreement: an utterance that explicitly states that the speaker dis-
agrees with the previous proposition (e.g. Ezzel nem értek egyet. ~ I don’t agree
with this.).
–– challenge: an utterance that displays strong disagreement with the prior propo-
sition by questioning the addressee’s position and implying that (s)he cannot
provide evidence for his/her claim26 (e.g. Miért főznének jobban? ~ Why would
they cook better?; És elmondanád, hogy miért? ~ And could you tell me why?;
Akkor te miért főzöl olyan pocsékul? ~ Why do you cook so badly, then?; Akkor
mondj már egy olyan férfit, akire ez igaz! ~ Then tell me a man of whom it’s true.).
–– disbelief: an utterance that indicates that the speaker does not believe or
doubts the previous proposition and thus cannot accept it (e.g. Á:! ~A:h!;
Ugyan már! ~ Come on!; Ki van zárva! ~ That’s out of the question!).
–– evaluation: an utterance that expresses a negative evaluation of the previous
speaker’s proposition, indicating strong disapproval (Hülyeség! ~ Nonsense!;
Ez marhaság! ~ That’s bullshit!).
–– clarification of speaker’s meaning: an utterance that clarifies the usually mis-
understood meaning of the speaker’s previous utterance (e.g. Szóval te jobban
főzöl, mint én? ~ So you can cook better than me?; Magyarul én béna vagyok a
főzéshez. ~ So I am lame at cooking.), which, therefore, contradicts or corrects
the other interlocutor’s previous proposition (e.g. Én nem erről beszélek. ~ I’m
not talking about that.; Nem arra értem. ~ I didn’t mean that.; Nem úgy. ~ Not
that way.).

As is clear from the above, no new categories have been added to the disagreement
strategies. The only modification made to this new version of the coding scheme is
the division of contradictory statement and implied contradiction into two different
categories. This is due to these categories being different from each other in terms
of their directness.
The disagreement strategies listed above exhibit varying degrees of in/direct-
ness. Direct disagreement strategies are utterances in which there is a match
between the propositional content and the illocutionary force of the acts, so the
disagreement is uttered in a way that clearly communicates to the hearer that a dis-
agreement is being produced. They express disagreement explicitly,27 and therefore
require no or little interpretative work on the addressee’s part. Obviously, the most
direct disagreement strategy is stating disagreement, and other direct ­strategies

.  It typically has the syntactic form of interrogative or imperative.

.  The terms ‘direct’ and ‘explicit’ as well as ‘indirect’ and ‘implicit’ are used as synonyms in
this chapter.
 Helga Vanda Koczogh

include contradictory statement, disbelief and evaluation.28 In contrast, in the case

of indirect disagreement strategies, the propositional content does not clearly con-
vey a disagreement and therefore their illocutionary force is not transparent. As a
result, indirect disagreements generally require more inferential effort on the part
of the hearer and it may be necessary to rely on the prosodic and/or non-verbal
context to recognise/interpret the utterance as a disagreement. Indirect disagree-
ments make use of the following strategies: partial agreement/token agreement,
explanation, implied contradiction, challenge, and clarification of speaker’s meaning.
It has to be noted that the disagreement strategies within the two broad categories
described above also display different levels of in/directness.
Turning back to my taxonomy, the category of PFMs has been further divided
into two groups, depending on their effect on the pragmatic force of disagree-
ments: mitigators and aggravators. The subgroup of mitigators contains linguistic
items and devices that soften the force of disagreements, each of which is listed
and briefly defined below.29

Pragmatic force modifiers:

Mitigators

–– humour: a mitigating device that softens the force of the disagreement usually
engendering laughter on the part of the speaker, listener, or both. The utter-
ance is made in a non-serious, joking manner (cf. Example 5).
–– hedge: mitigating devices used to (1) soften the face-threatening force of dis-
agreement by indicating uncertainty on the part of the speaker (e.g. szerintem
~ I think, lehet ~ maybe, esetleg ~ perhaps, hat, -het ~ can, could), (2) lower the
effect on the meaning of another element (e.g. valamennyire ~ to some extent,
egy kicsit ~ a little) or (3) fill in the time used for thinking (e.g. ö ~ uh, m: ~
u:hm, hát ~ well).
–– impersonalisation: the use of impersonal/passive structures to shift responsi-
bility. It is a mitigating device as it makes the utterance appear as coming from

.  This categorisation of disagreement strategies in terms of in/directness is based on the

opinion of 8 PhD students of linguistics, including myself. In my conceptualisation, disagree-
ment strategies form a continuum as regards their in/directness, but their exact place on the
continuum warrants further research. Therefore, they are only divided into the two broad
categories of ‘direct strategy’ and ‘indirect strategy’ in this chapter.
.  Due to space considerations, linguistic devices that would require a broader context are
not illustrated here. However, cross-references are provided.
 The development of a taxonomy of verbal disagreements in the light of the p-model 

a different source (e.g. Egyesek szerint ~ Some say, Azt mondják, hogy… ~ It is
said that…).
–– tag question: a short question added to a statement which requests assurance
or affirmation regarding what is expressed in the main clause. It serves to seek
confirmation either explicitly or rhetorically and to involve the hearer in the
discourse (e.g. ugye? ~ right?).
–– conditional: a mitigating linguistic device expressing a hypothetical situation,
thus softening the pragmatic force of the utterance.
–– term of endearment: a word or a phrase used to address a person in a way that
conveys solidarity and affection and thus softens the propositional content of
the utterance (e.g. Maci, Macika ~ Honey Bear).30

As is apparent from the list above, the last three categories are new additions to the
taxonomy. Tag questions have several functions (e.g. epistemic, facilitative, chal-
lenging) and there is no one-to-one correspondence between individual occur-
rences and functions. As a result, tag questions need to be interpreted with regard
to the given context. Here only those tag questions are considered as mitigators
that soften the force of the utterance by showing positive face considerations and
facilitate conversation.
The category of conditional refers to l’art pour l’art conditional, which softens
the force of disagreement by expressing a hypothetical situation. Other polite uses
of conditional are included among hedges.
The last category of mitigators includes the use of terms of endearment as in
the research corpus they were uttered in a soft tone without special emphasis and
thus they soften the force of disagreement by shifting the focus from the disagree-
ment per se to the camaraderie and ties existing between interlocutors. This is
illustrated in Example 13, where the last two utterances of speaker M04 are miti-
gated by the use of Maci and Macika (~ Honey Bear) emphasising the special bond
between the speaker and the hearer.
The other group of PFMs termed aggravators include those linguistic items
and devices that aggravate the pragmatic force of disagreement. As can be seen
from the list below, this group contains only one new category when compared to
the previous versions of the taxonomy. It is the category of tag question with the
function of challenging the other speaker.

.  Naturally, I do not intend to imply that terms of endearment always reflect immediacy.
These terms can have either positive or negative polarity which is determined by intonation
and context. In the research corpus, both occurrences were uttered with a soft tone and falling
intonation, which conveys solidarity rather than distance or annoyance.
 Helga Vanda Koczogh

Aggravators:

–– mockery: an aggravating device (indicated by tone of voice) in which the real

intent of the speaker is “concealed or contradicted by the literal meaning of
words or a situation” (Encyclopedia Britannica Online, 2006). The two types
of mockery employed in the research corpus are irony (hidden mockery) and
sarcasm (open mockery) (cf. Example 15).
–– intensifier: words or phrases (e.g. egyáltalán ~ at all, teljesen ~ totally,
mondom ~ I’m telling you) used to emphasise and strengthen the effect of
disagreement.
–– interjection: words (e.g. ha! ~ ha!, m: ~u:hm, á: ~ a:h) or phrases (e.g. ugyan
már! ~ come on!) used to convey emotion (e.g. disbelief, surprise, anger, irri-
tation, etc.) on the part of the speaker typically placed at the beginning of an
utterance. Their meaning depends on their intonation contour.
–– verbal shadowing, repetition: repetition of a previous speaker’s words, phrases
or entire utterance (and sometimes intonation) by another speaker to ques-
tion the content of the previous utterance. Intonation usually entails criticism
(cf. Example 14).
–– rhetorical question: “a question to which no answer is expected because there
can be no answer, the answer is obvious (implication: to the meanest intel-
ligence), or because the speaker provides an answer. What makes a rhetori-
cal question an aggravating device is that its basic purpose is to score points
against an opponent” (Rees-Miller, 1995: pp. 137f.). It generally occurs with a
falling intonation (cf. Examples 16 and 20).
–– tag question: a short question added to a statement which has the function of
strengthening the aggravating force of the utterance by challenging the posi-
tion of the other interlocutor (e.g. ugye? ~ right?, nem? ~ negative tag such as
isn’t it?).

It goes without saying that neither the list of disagreement functions nor that of
PFMs is exhaustive by any means. They include the categories that were identified
in my research corpus but other disagreement strategies and even PFMs might be
discovered, for example, when examining disagreement in other contexts.
As for the construction of disagreements, I identified the following internal
structure in the course of indexing illustrated by Figure 2.
As the figure shows, disagreement can be expressed via a single disagreement
strategy or optionally (indicated by brackets) with a combination of strategies, and
can be preceded by a pre-sequence. Optionally, the pragmatic force of disagree-
ments can be altered by the use of PFMs.
 The development of a taxonomy of verbal disagreements in the light of the p-model 

Disagree- (Disagree- (Disagree-

(Pre- Disagree-
ment ment ment
sequence) ment
strategy 1 strategy 2) strategy 3)

(PFMs)**

** Naturally, it is possible to assign more than one PFM to a single disagreement strategy.

Figure 2.  Structure of disagreement

4.3  Comparison to other solutions

As was mentioned before, the new coding scheme for analysing disagreement
described in the previous section provided a solution for both the multifunction-
ality of a single token and the differences in categorisation discussed in Section 4.2.
Due to space limitations, I will only discuss some examples here.
If we reconsider Example 15, repeated below for convenience, in terms of the
new taxonomy we can see that the last utterance (Ó: igen! ~ Yea:h, right.) employs
the disagreement strategy of disbelief and its pragmatic force is strengthened by
the aggravator mockery. Originally, Ó: igen was assigned two functions (disbelief
and mockery) but the division between disagreement strategy and PFM provides
a more plausible analysis.
Example 15
F05 @@@ Hát jó, de a Révésznek akkor se kellett volna
@@@ Well fine, but still the Ferryman shouldn’t have
M05 De lehet, hogy a Révész és a Haramiák haverok voltak.
But maybe the Ferryman and the Thieves were friends.
→ F05 /\●Ó: igen║!
Yea:h, right.

In order to interpret the last utterance as a disagreement one needs more than
mere lexical knowledge. On the surface level Ó: igen! can be understood as an
agreement and it is the ironical tone (rise-fall intonation) that helps us arrive at
the right interpretation. As Kotthoff (1993) states, in the context of a dispute inter-
locutors use short assent as a means of being ironic and expressing disagreement.
As a result of the new taxonomy for analysing disagreements, the disagreement
expressed by speaker F05 in Example 7 below can be categorised clearly without
having to assign several functions to it, unlike in the previous version of the model.
 Helga Vanda Koczogh

Example 7
M05 Szerintem a férfiak SOKKAL intelligensebbek a nőknél.
I think men are MUCH MORE intelligent than women.
→ F05 De ö: /●azt mondják║ hogy azér egy nő /●tud több dologra figyelni
egyszerre║ … \●jobban használja agyának a ké /●mindkét felét║…
But u:h it is said that women are able to pay attention to more
things at the same time … they use both of both sides of their
brain…

In speaker F05’s utterance the initial de (~ but) indicates that she would like to
reject the previous proposition (a férfiak sokkal intelligensebbek a nőknél ~ men are
much more intelligent than women) and she chooses to express disagreement in a
mitigated way. The contrastive marker de (~ but) is followed by a hesitation marker
(ö: ~ u:h) and a counter-claim that provides an explanation and supports why the
speaker is right. Giving a reason or explanation attenuates disagreement and so
does the use of the generic statement azt mondják, hogy (~ it is said that) serving
as a means of distancing the speaker from the truth value of the proposition. Thus,
in the new framework the utterance under scrutiny utilises the strategy of explana-
tion and it is accompanied by two mitigators (hedge and impersonalisation).
The following example illustrates how disagreement strategies can co-occur in
order to express disagreement in a more effective way.
Example 24
F09 … Hát szerintem nem kellene kötelező tandíjat bevezetni, mert
most én azért fizessek, mert tanulni akarok? Nekem ez így nem
… kapcsolódik össze.
… Well I don’t think compulsory tuition fees should be introduced
because should I pay because I want to study? For me this doesn’t
… make any sense.
→ M09 Hát /●jó║ de azt azért \/●figyelembe kell venni║, hogy a \●háttérben
azért van egy apparátus║ amit \●biztosítanak számodra║ /\●azért
hogy tanulhass║.
Well fine, but it needs to be taken into consideration that there is
some apparatus provided (for you) in the background so that you
can study.

Example 24, taken from a discussion about compulsory tuition fees in higher edu-
cation, illustrates the most frequently used combination of disagreement strategies
in my research corpus. The disagreement starts with a hedging device (hát ~ well),
followed by a partial agreement (jó, de ~ fine, but) and an explanation (azt azért
figyelembe kell venni, hogy… ~ it needs to be taken into consideration that…) of how
 The development of a taxonomy of verbal disagreements in the light of the p-model 

tuition fees are used for the benefit of students. As Kotthoff (1993: 208) points out,
“partial agreement can help to develop the main point of disagreement” and “is
often a prestep to further disagreement”.

5.  The final p-context

As can be seen from the examples discussed in the previous section, the new tax-
onomy distinguishing between disagreement functions (strategies) and linguis-
tic devices that either mitigate or aggravate the force of the opposing utterance
(PFMs) seems to serve as a suitable analytical tool for providing us with a com-
prehensive analysis of disagreements. In addition, this framework allows for an
analysis of utterances that contain a combined use of disagreement strategies, thus
the new taxonomy is more complete and sophisticated than its previous version
or the models described in the starting p-context. Also, the functional catego-
ries are less subjective than those of the other p-context versions, since they are
assigned typically co-occurring PFMs and, when applicable, acoustic features as
well, which make the categories more distinguishable and the model more robust.
Therefore, this model leads to statements with a higher plausibility value than the
other p-context versions. Moreover, the model also resolves the informational
underdetermination of the previous p-context versions, since it can account for
all instances of disagreement found in the research corpus. One can conclude that,
based on the information available and the set of criteria laid down in Section 2,
this modified category system seems to be a much better solution to (P) than the
rival models (cf., 2.1–2.4); thus, it can be regarded as a resolution to (P) in the
information state at our disposal. Consequently, on the basis of this case study it
can be stated that the resolution to (P) is as follows:
(SP) The model distinguishing between strategies and PFMs allows for
­describing Hungarian disagreement the most.

However, it must be noted that (SP) is a fallible solution which, in light of new
data, can be reconsidered in the course of a new argumentation cycle.

6.  Conclusion

This chapter has introduced and contained a discussion of the methodological

approach and research design adopted to meet the research objectives presented
in Section 1. After a brief overview of rival models of disagreement, I provided a
detailed description of the metatheoretical research, along with an account of the
 Helga Vanda Koczogh

changes made in the instrument as the research proceeded. I explained the multi-
step method used for analysis including coding schemes and extensive illustration,
where applicable. A step-by-step description of the analytical procedure(s) was
aimed at guiding the reader through the stages of the research whose findings
constantly shaped the original framework until it reached its ultimate form. By
describing each step of the research along with the decisions and re-evaluations
of the information at hand as well as the p-context, I demonstrated the cyclic and
prismatic nature of the plausible argumentation process.
As a primary goal, the study attempted to find the model that is the most
capable of describing Hungarian disagreement. It succeeded in mapping the
functional spectrum of disagreements employed by Hungarian undergraduate
students by identifying specific disagreement strategies. Additionally, PFMs,
which either strengthen or soften the illocutionary force of the utterance, were
also identified. This study, thus, managed to investigate disagreement not only in
type of strategy but also in terms of modification devices. Furthermore, it pro-
poses a framework of verbal disagreement that is more reliable and sophisticated
than other rival models. It seems to meet all the criteria laid down previously and
thus can be regarded as a resolution to (P), which might be re-evaluated based
on new data.

Acknowledgements

Work on the present chapter was supported by the MTA-DE Research Group for
Theoretical Linguistics and the TÁMOP-4.2.2/B-10/1-2010-0024 project, which is
co-financed by the European Union and the European Social Fund. I would like to
I express my gratitude to András Kertész and Csilla Rákosi for their helpful com-
ments and insightful views on earlier drafts of this paper.

References

Aijmer, Karin. 1986. Discourse variation and hedging. In Corpus Linguistics, II, New Studies in
the Analysis and Exploitation of Computer Corpora, Jan Aarts & Willem Meijs (eds), 1–18.
Amsterdam: Rodopi.
Bándli, Judit. 2009. Az egyet nem értés pragmatikája. PhD dissertation, University of Pécs, Pécs,
Hungary.
Blum-Kulka, Shoshana, House, Juliane, & Kasper, Gabriele. 1989. The CCSARP coding man-
ual. In Cross-cultural Pragmatics: Requests and Apologies, Shoshana Blum-Kulka, Juliane
House, & Gabriele Kasper (eds), 273–294. Norwood NJ: Ablex.
Brown, Penelope, & Levinson, Stephen. 1987. Politeness: Some Universals in Language Usage.
Cambridge: CUP.
 The development of a taxonomy of verbal disagreements in the light of the p-model 

Bousfield, Derek. 2008. Impoliteness in Interaction [Pragmatics & Beyond New Series 167].
Amsterdam: John Benjamins.
Culpeper, Jonathan. 1996. Towards an anatomy of impoliteness. Journal of Pragmatics 25: 349–367.
Culpeper, Jonathan. 2005. Impoliteness and entertainment in the television quiz show: The
Weakest Link. Journal of Politeness Research: Language, Behaviour, Culture 1: 35–72.
Culpeper, Jonathan. 2011. Impoliteness: Using Language to Cause Offence. Cambridge: CUP.
Culpeper, Jonathan, Bousfield, Derek, & Wichmann, Anne. 2003. Impoliteness revisited: With
special reference to dynamic and prosodic aspects. Journal of Pragmatics 35: 1545–1579.
Du Bois, John W., Cumming, Susanna, Schuetze-Coburn, Stephan, & Padino, Danae (eds).
1992. Discourse transcription. In Santa Barbara Papers in Linguistics 4. Santa Barbara CA:
University of California.
Fónagy, Iván. 1971. Synthèse de l’ironie. Basel: Karger.
Fónagy, Iván, & Magdics, Klára. 1967. A magyar beszéd dallama. Budapest: Akadémiai Kiadó.
Grice, Herbert Paul. 1975. Logic and conversation. In Syntax and Semantics, 3: Speech Acts, Peter
Cole & Jerry Morgan (eds), 41–58. New York NY: Academic Press.
Holtgraves, Tom. 1997. Yes, but…: Positive politeness in conversation arguments. Journal of
Language and Social Psychology 16: 222–239.
Holmes, Janet. 1995. Women, Men and Politeness. London: Longman.
Kertész, András, & Rákosi, Csilla. This volume. Evidence and the argumentation structure of
linguistic theories.
Kertész, András, & Rákosi, Csilla. 2012. Data and Evidence in Linguistics. A Plausible Argumen-
tation Model. Cambridge: CUP.
Koczogh, Helga Vanda. 2012. The Effects of Gender and Social Distance on the Expression of
Verbal Disagreement Employed by Hungarian Undergraduate Students. Ph.D. disserta-
tion, University of Debrecen, Debrecen, Hungary.
Komlósi, Boglárka. 2007. Az iróniakutatás elméleti dilemmái a pragmatikában. Nyelvtudomány
3–4: 95–128.
Komlósi, Boglárka. 2010. Irony in the semantics-pragmatics interface: A reconstructional
model. 〈http://conference.clancorpus.net/?p=1422〉
Komlósi, Boglárka. 2013. A polarizációs tendencia szerepe az irónia nyilvánvalótételében. A
hiperbola és lilotész mint nyomok az irónikus interpretációhoz. Magyar Nyelv 4: 409–420.
Kotthoff, Helga. 1993. Disagreement and concession in disputes: On the context sensitivity of
preference structures. Language in Society 22: 193–216.
Leech, Geoffrey. 1983. Principles of Pragmatics. London: Longman.
Locher, Miriam. 2004. Power and Politeness in Action: Disagreements in Oral Communication.
Berlin: Mouton de Gruyter.
Markó, Alexandra. 2005. A spontán beszéd néhány szupraszegmentális jellegzetessége: Mono-
logikus és dialogikus szövegek összevetése, valamint a hümmögés vizsgálata. Ph.D. disser-
tation, ELTE, Budapest, Hungary.
Markó, Alexandra. 2007. A Special Conversational Device: Humming in Hungarian. The
­Phonetician 95: 28–35.
Mászlainé Nagy, Judit. 2010. A jászapáti Magyar és roma kisiskolások nyelvhasználati stratégiái-
nak kultúraközi elemzése. PhD dissertation, University of Debrecen, Debrecen, Hungary.
Muntigl, Peter & Turnbull, William. 1998. Conversational structure and facework in arguing.
Journal of Pragmatics 29(3): 225–256.
Németh T., Enikő. 1996. A szóbeli diskurzusok megnyilatkozáspéldányokra tagolása [Nyelvtudo-
mányi értekezések 142]. Budapest: Akadémiai Kiadó.
 Helga Vanda Koczogh

Nauke, Astrid, & Braun, Angelika. 2011. The production and perception of irony in short
­context-free utterances. ICPhS 7: 1450–1453.
Rees-Miller, Janie. 1995. Linguistic Features of Disagreement in Face-to-Face Encounters in
University Settings. Ph.D. dissertation, Stony Brook, NY: State University of New York.
Rees-Miller, Janie. 2000. Power, severity, and context in disagreement. Journal of Pragmatics 32:
1087–1111.
Sperber, Dan, & Wilson, Deirdre. 2006. Relevance Theory. In The Handbook of Pragmatic,
­Laurence R. Horn & Gregory Ward (eds), 607–632. Oxford: Blackwell.
Svindt, Veronika. 2007. Iróniaértelmezés elméletben és gyakorlatban. Első Század, 1: 171–208.
〈http://www.epa.hu/01600/01639/00003/pdf/elso szazad_2007_1_171-208.pdf〉
Svindt, Veronika. 2008. Az irónia szupraszegmentális ismertetőjegyei. Félúton 4: 1–15.
Tannen, Deborah. 1984. Conversational Style: Analyzing Talk Among Friends. Norwood NJ:
Ablex.
Toplak Maggie & Katz, Albert N. 2000. On the uses of sarcastic irony. Journal of Pragmatics
32(10): 1467–1488.
Varga, László. 1985. Intonation in the Hungarian sentence. In Approaches to Hungarian, István
Kenesei (ed.), 205–224. JATE: Szeged, Hungary.
 chapter 7

A case of disagreement
On plural reduplicating particles in Hungarian

György Rákosi
University of Debrecen, Institute of English and American Studies, MTA-DE
Research Group for Theoretical Linguistics

The chapter discusses the so-called Hungarian reduplicating particle verb

construction that has generated some interest in the pertinent generative literature
on Hungarian. This literature is divided over whether reduplicating particles can
bear plural morphology in the presence of a third person plural oblique associate
or not: some accept and some reject the resulting agreeing reduplicating particle
verb construction, thus creating a par excellence context of data inconsistency. The
chapter offers a detailed and critical overview of the literature, and presents some
novel arguments in an effort to find a paraconsistent solution to this problem
within the framework of the p-model of Kertész and Rákosi (2012). This solution
rests and on the claim that the plural and the non-plural reduplicating particle
verb constructions are radically different in their grammar, since the particle only
acts as a pronominal in the case of the former construction type. The pronominal
use of the particle is a marked option, rendering the agreeing reduplicating
particle verb construction a marked phenomenon of Hungarian grammar.

1.  Introduction

Hungarian verbal particles fall into several, grammatically distinct classes. One type
doubles the case suffix that appears on the oblique associate of the particle verb. The
resulting reduplicating particle verb construction (RPVC for short) is illustrated in (1):1

.  Diverging from Hungarian orthographical conventions, I write the verb and the particle
that immediately precedes it as two different words, and I do this consistently throughout this
article. It is partly for expository purposes, and partly for the reason that these conventions
are not fully consistent and they would dictate the two-word spelling in some of the cases to
be discussed here anyway. Native speakers are also often uncertain about whether a particular
 György Rákosi

(1) Peti rá nézett a lány-ra.

Peti onto.3sg looked the girl-onto
‘Pete looked at the girl.’

In this example, the oblique associate of the particle verb bears sublative case, and
the case suffix is reduplicated as the verbal particle rá.
The reduplicating particle verb construction has been studied quite exten-
sively in the literature on Hungarian, including, but not restricted to, Ackerman
(1987, 1990, 2003), Ackerman & Webelhuth (1993), Ackerman & LeSourd (1997),
Hegedűs (to appear), É. Kiss (1998, 2002), Kálmán & Trón (1999), Rákosi & L ­ aczkó
(2011, to appear), Surányi (2009a, b) and Ürögdi (2003). The construction has
many intriguing properties and my aim in this article is to scrutinise one of these:
the behaviour of the particle when the oblique associate is plural. In the standard,
the particle takes what looks like a default 3sg form even in the presence of a plu-
ral associate.2 There is general consensus in the literature on the existence, i.e. the
grammaticality, of the resulting non-agreeing RPVC. Consider (2a) for illustration.
(2) non-agreeing RPVC
a. Peti rá nézett a lány-ok-ra.
Peti onto.3sg looked the girl-pl-onto
‘Pete looked at the girls.’
agreeing RPVC
b. Peti rá-juk nézett a lány-ok-ra.
Peti onto-3pl looked the girl-pl-onto
‘Pete looked at the girls.’

Much less obvious is the answer to the question whether the hypothetical alterna-
tive with a plural form of the particle is grammatical or not (2)(b). Most of the
literature listed above mentions the agreeing RPVC, but they are in disagreement
as to whether it is to be considered grammatical or ungrammatical. Some explicitly
reject it, and mark constructional equivalents of (2)(b) with the asterisk (É. Kiss
1998, Hegedűs to appear, Kálmán & Trón 1999; Kenesei 2005). But for others, the
agreeing RPVC is grammatical and they generally take it to be so without e­ xplicitly
entertaining the possibility that it need not be (É. Kiss 2002; Surányi 2009a, b). This
basic inconsistency and its possible resolution is the focus of this chapter.

particle-verb combination should be spelt as one or two words – an uncertainty that is well-
reflected in the variation in written forms found in corpora.
.  Since the purposes of this chapter do not require a decision, I remain agnostic about
whether the particle in (2)(a) is underspecified for number or whether it is a morphologically
unmarked 3sg form. I gloss it as 3sg for expository reasons.
 A case of disagreement 

To be able to execute the resolution process in a manner that introduces

new insights to the argumentation, I apply the p-model of Kertész & Rákosi (this
­volume) as a framework of plausible argumentation. Within this model, the above
described problem is a p-problem in the technical sense of the term as it origi-
nates from a particular case of informational overdetermination they refer to as
p-inconsistency (Kertész & Rákosi, this volume: Section 2.5). The issue underlying
the problem itself can be formulated as follows:3
(P1) How can the following p-inconsistency be treated?
(i) 0 < |The agreeing RPVC is grammatical in Hungarian.|linguistic intuition of
some native speakers < 1
(ii) 0 < |The agreeing RPV‌C is ungrammatical in Hungarian.|linguistic intuition
of some native speakers < 1

In pursuing an answer to this question, the chapter stays at a relatively descrip-

tive level, even if I try to make explicit note of certain background assumptions
where they form important part of the plausible argumentation process. This is
partly justified by the ultimate metalinguistic nature of this inquiry, and partly by
the lack of attention to (P1) as a theoretically relevant problem in the pertinent
literature. In other words, the assumed grammaticality or ungrammaticality of
the agreeing RPVC has not played a significant role as a decisive argument for or
against particular analyses.4
The problem itself represents the much-attested case of inconsistent grammat-
icality judgements in modern linguistics (see Kertész & Rákosi 2012 and Kertész
2012 for a general discussion, as well as Németh T. (this volume) for a thorough
discussion of another such case) and as such, it merits attention on its own right.
I argue below that the usual solution, which treats such judgement inconsistencies
as instances of dialectal or idiolectal variation, is not enlightening and is mislead-
ing in certain respects. I try to substantiate the claim that the ­agreeing RPVC does
exist, albeit as a construction that is marked in certain ways, unlike the default
non-agreeing construction. It is not my intention here to spell out a detailed anal-
ysis that captures the marked nature of this construction. Rather, my aim is to

.  The notation |p|S introduces the plausibility value of a statement p relative to a (possibly
plural) individual S (see Kertész & Rákosi 2012, this volume). This notation is described in
more detail in 3.1.
.  It is beyond the scope of this chapter to discuss the Hungarian reduplicating particle verb
construction from a comprehensive cross-linguistic perspective. For recent discussions of
P-doubling constructions in German, see Noonan (2010) and Abraham (2010).
 György Rákosi

pursue a cyclic and prismatic argumentation process that hopefully arrives at a

resolution to (P1) that subsequent linguistic research may find useful.
The elaboration of a new p-context to a resolution to (P1) will include the
introduction of an important underlying issue, formulated as (P2) below:
(P2) Modulo the difference in number marking, are the agreeing and the non-
agreeing RPVCs instances of the same construction type?

Interestingly, all the pertinent literature tacitly assume that the answer to (P2) is
yes, irrespective of whether they take the agreeing RPVC to be grammatical or not.
This assumption is tacit in the sense that no-one has claimed explicitly that the
two respective RPVC varieties are distinct constructions. I attempt to show here
that the answer to (P2) turns out to be negative under closer inspection, and this
recognition will help us find a solution to (P1) that is hopefully more satisfactory
than previous accounts – that is, it can be regarded to be the resolution of (P1) on
the basis of the information at our disposal.
The structure of the chapter is as follows. In Section 2, I set up the minimal
descriptive background as a prerequisite for the construal of the initial p-context
of argumentation. This p-context is essentially the reconstruction of the existing
accounts that altogether give an inconsistent answer to (P1). I describe the key
features of this p-context in Section 3, discussing arguments that this literature
offers in support or in rejection of the grammaticality of the agreeing reduplicating
particle verb construction. In Section 4, I extend this p-context by the introduction
of new pieces of evidence favouring a positive answer to (P1) as well as coordinate
it via synthesising the old and the new information that is relevant for the argu-
mentation. In Section 5, I provide a summary by presenting the resulting final
p-context that crucially relies on a joint resolution to (P1) and (P2).

2.  The descriptive background

It is well-known that Hungarian is an extensively agglutinating language. One

reflex of this is the use of suffixal markers where other languages have adpo-
sitional elements. Consider the oblique expression Katira ‘onto Kate’ in (3) for
illustration:
(3) János fel nézett Kati-ra.
John up looked Kate-onto
‘John looked up to Kate.’

If these locative case markers take a pronominal complement, than the case suf-
fix assumes the relevant agreement morphology (4)(a), which is zero in 3sg (see
 A case of disagreement 

Footnote 2). The pronoun itself is normally not present in compliance with the
pro-drop nature of Hungarian (4)(b), and it is there only if it receives a discourse
function. In (4)(a), it is the focus of the clause.
(4) a. János ő-rá nézett fel.
John (s)he-onto.3sg looked up
‘It is her that John looked up to.’
b. János fel nézett rá.
John up look onto.3sg
‘John looked up to her.’

For ease of exposition, I simply refer to both inflected case suffixes in (4)(a) and
(4)(b) as pronominal forms, without intending to imply any theoretical commit-
ment with this term.
It is this pronominal form that apparently occurs in the reduplicating con-
struction, but only in its ‘pro-dropped’ version. I repeat (1) as (5)(a), and compare
it with (5)(b):
(5) a. Peti rá nézett a lány-ra.
Peti onto.3sg looked the girl-onto
‘Pete looked at the girl.’
b. *Peti ő-rá nézett a lány-ra.
 Peti (s)he-onto.3sg looked the girl-onto
‘Peter looked at the girl.’

Since there is general consensus about the ungrammaticality of (5)(b), and since
my own intuition is in agreement with this consensus, I treat the u
­ ngrammaticality
of (5)(b) as certain. The contrast between (5)(b) and (5)(a), and the potential
analogy of the particle in (5)(a) with the pronominal in (4)(b) indicate the essen-
tial challenge that these constructions pose for a linguist: in what sense can the
particle in (5)(a) be treated as a pronominal of some sort? As I argue in Section 4,
this important issue constitutes a crucial element in the elaboration of (P2).
Before embarking on the establishment of the initial p-context for our case,
I give an exhaustive list of the inflected suffixal pronominal forms that can serve
as reduplicating particles.
(6) a. bele ‘into (it)’
b. benne ‘in (it)’
c. érte ‘for (it)’
d. hozzá ‘to (it)’
e. neki ‘to/against (it)’ (dative case)
f. rá ‘onto (it)’
g. rajta ‘on (it)’
h. vele ‘with (it)’
 György Rákosi

Some of these are more productive in this use than others, and they share the
property that they are either directional or (stative) locative in their semantics.5
These 8 elements represent the class of particles that constitutes the empirical
­coverage of the forthcoming investigation.

3.  Setting up the p-context: The literature on agreeing RPVCs

3.1  The naysayers

I refer to those authors who consistently deny the grammaticality of the agreeing
reduplicating particle verb constructions as naysayers. I quote all of the relevant
data they include in their papers in (7)–(9). The judgements are of the respective
authors.6
– 
É. Kiss (1998: 126)
(7) a. János hozzá vágta a labdá-t a fiú-k-hoz.
John to.3sg threw the ball-acc the boy-pl-to
‘John threw the ball at the boys.’
b. *János hozzá-juk vágta a labdá-t a fiú-k-hoz.
   John to-3pl threw the ball-acc the boy-pl-to
‘John threw the ball at the boys.’
(8) a. János bele nyúlt a kosar-ak-ba.
John into.3sg reached the basket-pl-into
‘John reached into the baskets.’
b. *János belé-jük nyúlt a kosar-ak-ba.
   John into-3pl reached the basket-pl-into
‘John reached into the baskets.’
–  Kálmán & Trón (1999: 208)
(9) a. *Rá-juk nézt-em a fiú-k-ra.
   onto-3pl looked-1sg the boy-pl-onto
‘I looked at the boys.’
b. *Belé-jük szeretett a lány-ok-ba.
   into-3pl loved the girl-pl-into
‘He fell in love with the girls.’

.  See É. Kiss (1998, 2002) and Surányi (2009a, b) on the linguistic relevance of this.
.  Kenesei (2005: 422) also denies the grammaticality of the construction, but since he
does this in a review discussion of É. Kiss’s (2002) examples, I come back to his remarks in
Section 3.3 below.
 A case of disagreement 

‒  Hegedűs (in prep: 145)

(10) a. *Péter rá-juk nézett a lány-ok-ra.
   Peter onto-3pl looked the girl-pl-onto
‘Peter looked at the girls.’
b. *Az autó nek-ik ment a kerítés-ek-nek.
   the car dat-3pl went the fence-pl-dat
‘The car bumped into the fences.’

These authors all seem to treat these examples implicitly as representatives of what
I call here the agreeing reduplicating particle construction, and they describe it as
one that does not license (plural) number agreement between the particle and the
associate.
In short, the following plausible inferences present themselves as possible
reconstructions of their argumentation:
|The respective sentences (7)(b), (8)(b), (9) and (10) are
ungrammatical.|S = 1.
NAY
[0 < |The respective sentences are representative of the agreeing
RPVC.|S < 1]
NAY
[0 < |All other instances of this construction type are ungrammatical,
too.|S < 1]
NAY
0 < |The agreeing RPVC is ungrammatical as a construction.|I <1
NAY

Here and below I use the notation that Kertész & Rákosi (2012, this volume:
­Section 2.2–3) introduce in their p-model with respect to plausible inferences.
The first line is an instance of the notation that denotes the plausibility value
of the relevant statement, where 1 is certainty on the basis of a given source S,
and 0 is neutral plausibility on the basis of a given source S. The second and third
line above introduce background assumptions with a plausibility value between 0
and 1.7 I is the plausible inference itself. I use NAY as a subindex on S as a short-
hand for the group of naysayers, i.e. those authors who consider the agreeing
RPVC ungrammatical.
The above background assumptions are taken to have a relatively high
plausibility value, since the description that the authors provide of the data
strongly implies that they share them. But since it is not an explicitly presented
­commitment  – that is, none of the authors claim that their examples generally

.  As Kertész and Rákosi (this volume: Section 2.3) point out, “With inductive inferences,
the set of the premises has to be supplemented by the background assumption that the cases
not examined also possess the characteristics that could be found in the investigated ones …”.
This is a general assumption in linguistic practice.
 György Rákosi

r­ epresent the ­construction type itself –, the plausibility value is not 1. It follows
that the conclusion also has a plausibility value that is lower than 1.

3.2  The yea-sayers

Analogically extending the previous naming convention to the other camp, I refer
to those who find the agreeing RPVC grammatical as yea-sayers. Accordingly,
S stands for the intuition of yea-sayers as a data source. This group of linguists
YEA
include É. Kiss (2002), Surányi (2009a, b) and myself.
I quote Surányi’s data as they appear in his two papers on particles:

Surányi (2009a: 72)
– 
(11) Az orvos hozzá / hozzá-juk érintett egy műszer-t
the doctor to.3sg / to-3pl touched an instrument-acc
a gyerek-ek-hez.
the child-pl-to
‘The doctor had an instrument touch the children.’

Surányi (2009b: 231)
– 
(12) János nek-i / nek-ik ment a járókelő-k-nek.
John dat-3sg / dat-3pl went the passer.by-pl-dat
‘John bumped into the passers-by.’

There is no indication in the papers that his intuitions dictate any grammaticality
difference between the agreeing and the non-agreeing RPVCs, therefore I take it
to mean that Surányi finds both equally grammatical.
É. Kiss (2002: 197) mentions the following two minimal pairs:
–  É. Kiss (2002: 197)
(13) a. János nek-ik ment a járókelő-k-nek.
John dat-3pl went the passer.by-pl-dat
‘John bumped into the passers-by.’
b. János nek-i ment a járókelő-k-nek.
John dat-3sg went the passer.by-pl-dat
‘John bumped into the passers-by.’

(14) a. ??János nek-ik vágta a labdá-t az üvegtáblá-k-nak.

 John dat-3pl threw the ball-acc the window.pane-pl-dat
‘John threw the ball against the window-panes.’
b. János nek-i vágta a labdá-t az üvegtáblá-k-nak.
John dat-3sg threw the ball-acc the window.pane-pl-dat
‘John threw the ball against the window-panes.’
 A case of disagreement 

There are two noteworthy aspects of these examples. First, while É. Kiss (1998)
finds (7)(b) and (8)(b) ungrammatical, (13)(a) is presented in É. Kiss (2002) as
grammatical. Admittedly, the three respective sentences are non-identical, but
both texts suggest that the author describes properties of the construction and not
of particular examples only.8
The second interesting aspect of these examples is the contrast between
(13)(a) and (14)(a). The difference between the two examples is that the oblique
associate in (13)(a) has a human referent but the one in (14)(a) does not. This
latter is marked by É. Kiss as “??”, indicating that she does not find it either fully
grammatical or ungrammatical. She does not, however, make explicit note of this
contrast beyond presenting the data. The oblique associates in Surányi’s examples
each have human referents, and it is not clear whether the lack of examples with
non-human referents is an intentional feature of the presentation or not.
I interpret this as a sign that each author assumes latently that the agreeing
RPVC is only grammatical with human associates. Since this is my own intuition,
too, at this stage of the argumentation process I will set aside any examples where
the oblique associate has non-human reference. This also helps reconstruct the
initial p-context in such a way that “noisy” data are filtered out. In other words, I
background certain types of data that are not promising to be useful for the subse-
quent plausible argumentation process, and this will help concentrate on and the
core issue that constitutes the essence of (P1). In Section 4, I return to this refer-
ential restriction only to show that this strategy of backgrounding is justifiable at
this stage of the argumentation.
The yea-sayers’ approach is characterised by the plausible inference pattern
below.
|The respective sentences (11), (12) and (13)(a) are grammatical.|S = 1.
YEA
[0 < |These sentences are representative of the agreeing RPVC that includes
an oblique with +human referents.|S < 1]
YEA
[0 < |All other instances of this construction type are grammatical,
too.|S < 1]
YEA
0 < |If the oblique associate has +human referent(s), then the agreeing
RPVC is grammatical.|I <1
YEA

.  Two illustrative quotes:

É. Kiss (1998: 142): “A preverbal element of the hozzá type does not agree in number

with the postverbal argument bearing the same case ending.”
É. Kiss (2002: 197): “[…] when coindexed with a plural lexical noun phrase, the 3rd

person PP [i.e. the particle] does not necessarily bear a plural marker.”
 György Rákosi

Since the background assumptions (the second and the third lines) are, of course,
not explicit in É. Kiss (2002) and Surányi (2009a, b), the plausibility of the conclu-
sion itself is also lower than 1.

3.3  The initial p-context

The initial p-context has been restricted above to cases where the oblique associate
of the agreeing RPVC has human referents. Converting the two previous plausible
inferences into two complex statements describing this restricted p-context, the
result can be characterised as follows (remember that these statements are not
certain, but they have a high plausibility value):
(G) [→ I ] If the oblique associate has +human referents, then the
YEA
agreeing RPVC is grammatical.
(∼G) [→ I ] Even if the oblique associate has +human referents, the
NAY
agreeing RPVC is not grammatical.

This is the basic p-inconsistency that underlies (P1) and which is in need of
resolution.
Before continuing, I would like to point out that this problem has been explic-
itly noted by Kenesei (2005) and Hegedűs (in prep.) in their discussion of the
relevant data. Kenesei (2005) is in fact a review of É. Kiss (2002), and I quote an
illustrative passage discussing É. Kiss’s data that appear in (13) and (14) above.
Correction of, or comment on, the Hungarian examples themselves is of course
not the remit of a copy-editor. One would expect this from a reviewer, but although
Katalin É. Kiss thanks an anonymous reviewer, it can only be supposed that this
person was not a native Hungarian linguist, since otherwise one would have
expected comment on some of the more controversial examples. To illustrate,
compare the following pair on p. 197, rejected by all my students [reference to
examples (13)(a) and (14)(a) above] […].
Both of these sentences are totally unacceptable: the grammatical versions …
must have singular nek-i [dat-3sg, cf. (13)(b) and (14)(b)], […].
(Kenesei 2005: 423)

Kenesei (2005) thus provides a solution that, primarily on the basis of his and his
students’ intuitions, gives a negative answer to (P1), supporting (∼G). In what fol-
lows I try to argue that the total elimination of (G), which Kenesei seems to sug-
gest, is much less plausible as a solution than its maintenance.

3.4  Further remarks on the initial p-context

Beyond their own intuitions, the authors quoted above bring relatively little moti-
vation for the acceptance of either (G) or (∼G). What they do provide is some
remarks that point toward an explanation that either takes (G) or (∼G) as the basic
 A case of disagreement 

datum to be explained. These remarks are, however, somewhat occasional and do

not add up to a complete and coherent analysis. In this section, I try to collect them
to enrich the initial p-context for the purposes of the subsequent argumentation.
As for (∼G), the only specific analytical suggestion that I have been able to
locate is mentioned in É. Kiss (1998: 142). She analyses the reduplicating particle
as a synchronically inert verbal prefix, which is not expected to show agreement
by its very nature. This argument, however, is somewhat circular: we only know
that the particle has special status because of the lack of plural agreement (taking
É. Kiss 1998 now as a source of the respective judgement); and, in turn, the particle
is not supposed to agree because it is a special, grammaticalised form.
As for (G), the following two components of an explanation can be recon-
structed. First, É. Kiss (2002) takes both the agreeing and the non-agreeing RPVCs
to be appositive constructions. In her analysis, the particle is a full pronoun, and
the oblique associate is an adjunct that is coindexed with the particle. That is,
the Hungarian structure in (15)(a), which is (1) repeated, is not what its English
­translation suggests, but it is analogical with the English structure in (15)(b):
(15) a. Peti rá nézett a lány-ra.
Peti onto.3sg looked the girl-onto
‘Pete looked at the girl.’
b. Pete looked at her, the girl.

Under this analysis, we naturally expect the particle to be grammatical in the

­plural with plural associates (= the agreeing RPVC), as is grammatical the English
sentence I looked at them, the girls. Notice, however, that what we do not expect
in this analysis is the grammaticality of the non-agreeing construction, since the
corresponding *I looked at her, the girls is ungrammatical. Therefore the apposi-
tive hypothesis, in the absence of further provisions, generates a further piece of
p-inconsistency.
É. Kiss herself notes this problem (2002: 197) and adds a remark that the non-
agreeing construction is probably grammatical, even in the light of the appositive
analysis, because plural marking of both the particle and the associate would be
redundant. In other words, it is enough if the associate is marked plural since that
will take care of the intended plural reference. Therefore the particle can, but need
not necessarily bear a plural feature. Coming from an entirely different perspec-
tive, Surányi (2009a: 729, 2009b: 231) also makes the same redundancy argument,
assuming that the plural marking on the particle is a possibility that grammar
optionally provides but does not require.
The initial p-context thus contains opposing intuitions supporting (G) and
(∼G), respectively. The linguistic analyses that the literature offers in ­explanation
of either (G) and (∼G) are partial at best and may actually generate further
inconsistencies.
 György Rákosi

4.  E
 laborating the p-context: The differential analysis of agreeing
and non-agreeing RPVCs

4.1  The background to a new direction

I repeat example (2) as (16) and (P2) to remind us of the direction where I would
like to proceed at this point.
(16) non-agreeing RPVC
a. Peti rá nézett a lány-ok-ra.
Peti onto.3sg looked the girl-pl-onto
‘Pete looked at the girls.’
agreeing RPVC
b. Peti rá-juk nézett a lány-ok-ra.
Peti onto-3pl looked the girl-pl-onto
‘Pete looked at the girls.’
(P2) Modulo the difference in number marking, are the agreeing and the non-
agreeing RPVCs instances of the same construction type?

As stated in the Introduction to this chapter, (P2) is not raised explicitly in the
literature. The common latent background assumption is that we are dealing with
instances of one and the same construction.
The existing analyses fall into two different types along a parameter that is
relevant for us. É. Kiss (2002), as we have seen in the previous section, argues that
the particle in the RPVC is a full pronoun. As such, it is the head of an appositive
construction to which the oblique associate is an adjunct. No other account makes
this assumption in the literature. Others take this particle to be a verbal ­prefix
(É. Kiss 1998), a reduced copy of the associate (Ürögdi 2003; Surányi 2009a, b),
or an incorporated agreement marker of a special sort (Rákosi & Laczkó 2011, to
appear; and see also the series of papers by Ackerman and co-authors for a similar
approach). Cutting this opposition to its essentials, the difference is between treat-
ing the particle either as a full pronoun or as something else.
The claim that I intend to substantiate here is that both sides are right, albeit
for different domains of data. The two hypotheses that I want to defend below are
as follows:

(H1) The particle is not a pronoun in the non-agreeing RPVC.

(H2) The particle is a pronoun in the agreeing RPVC.

(H1) and (H2) together entail a negative answer to (P2), and they elaborate the
initial p-context in a way that facilitates the establishment of a resolution to
 A case of disagreement 

(P1).9 This is a paraconsistent solution in the sense of Kertész & Rákosi (2012)
and K­ ertész (2012) inasmuch as two contradictory statements are shown to
be ­co-existing at a higher level of analysis by restricting their application to
distinct empirical domains. Below I discuss three sets of data in support of this
claim. The source of these data is my own intuitions and the intuitions of the
speakers that I have asked for judgements.

4.2  Evidence 1: Binding data

The particle in the non-agreeing construction does not interfere with binding
dependencies, consider (17)(a):
(17) a. A gyerek-ek rá néztek egymás-ra.
the child-pl onto.3sg looked each.other-onto
‘The children looked at each other.’
b. *A gyerek-ek rá-juk néztek egymás-ra.
   the child-pl onto-3pl looked each.other-onto
‘The children looked at each other.’

In (17)(a), the antecedent of the oblique reciprocal anaphor is the subject noun
phrase, and the presence of the non-agreeing particle does not influence the gram-
maticality of the construction. (17)(a) is grammatical with or without the particle. In
(17)(b), the presence of the agreeing particle renders the sentence ungrammatical.10
It is a plausible assumption that (17)(b) is ungrammatical for the same reason
as (18):
(18) *A gyerek-eki bemutatták őketi egymás-naki.
   the child-pl introduced them each.other-dat
‘*The childreni introduced themi to each otheri.’

That is, we can treat (18) and (17)(b) as instances of the same phenomena: a
­pronoun intervenes between the antecedent and the anaphor, and its presence cre-
ates ungrammaticality. Crucially, this explanation works if we treat the agreeing
particle as a pronoun of some sort (H2), but we do not consider the non-agreeing
particle as a pronominal (H1).

.  (H2) can be equally plausible both in the context of (G) and (~G). It is because the
agreeing construction can in principle be either grammatical or ungrammatical because of
(H2). I argue here that the former is the case.
.  In terms of the traditional Binding Theory of Generative Grammar, the pronominal par-
ticle creates a Principle B violation.
 György Rákosi

4.3  Evidence 2: The verbless construction

Verbal particles can form a special construction with their oblique associates in
verbless clauses. Imagine a context in which a classroom is suddenly being flooded
and the teacher tells the students to jump on their parents, who are present. The
two sentences in (19) function as short commands in such a situation, but the two
in (20) are ungrammatical.
(19) a. Fel a szülő-k-re!
up the parent-pl-onto
‘Up onto the parents!’
b. Rá a szülő-k-re!
onto.3sg the parent-pl-onto
‘Onto the parents!’
(20) a. *Rá-juk a szülő-k-re!
   onto-3pl the parent-pl-onto
‘Onto the parents!’
b. *Ő-rá az apá-d-ra!
   onto-3sg the father-poss.2sg-onto
‘Onto your father!’

(19)(a) includes an adverbial particle, which is non-pronominal without doubt.

The non-agreeing construction in (19)(b) can be treated in parallel: it is grammati-
cal because it contains a non-pronominal particle.
On the other hand, (20)(b), which contains an overt pronoun as part of the
particle, is ungrammatical. I simply interpret this as evidence that the verbless
construction is not licensed with pronominal particles. If that is the case, the
ungrammaticality of (20)(a) is explained under the assumption that the agreeing
particle is a pronoun.
Notice that no explanation is required for these data in the context of the
­current argumentation. All we need is the parallel that we can draw between appar-
ently non-pronominal particles (19)(a) and the non-agreeing particle (19)(b) on
the one hand, and between apparent pronominal particles (20)(b) and the agree-
ing particle (20)(a) on the other.

4.4  Evidence 3: On the +human restriction

In 3.3, I have restricted our elaboration of the p-context to agreeing RPVCs in
which the oblique associate has +human reference. This was occasioned by (13)
and (14), two pairs of sentences from É. Kiss (2002). I repeat these two as (21)
and (22).
 A case of disagreement 

(21) a. János nek-ik ment a járókelő-k-nek.

John dat-3pl went the passer.by-pl-dat
‘John bumped into the passers-by.’
b. János nek-i ment a járókelő-k-nek.
John dat-3sg went the passer.by-pl-dat
‘John bumped into the passers-by.’
(22) a. ??János nek-ik vágta a labdá-t az üvegtáblá-k-nak.
 John dat-3pl threw the ball-acc the window.pane-pl-dat
‘John threw the ball against the window-panes.’
b. János nek-i vágta a labdá-t az üvegtáblá-k-nak.
John dat-3sg threw the ball-acc the window.pane-pl-dat
‘John threw the ball against the window-panes.’

The semantic properties of the associate do not seem to matter in the non-agreeing
construction ((21)(b) & (22)(b)). Once we switch to the agreeing construction
((21)(a) & (22)(a)), judgements get markedly worse if the associate (and therefore
the particle itself) has -human reference. What may lie behind this?
I have been suggesting that non-agreeing particles are not pronouns. As such,
they are not expected to show any special selectional restrictions on their obliques,
which is the case by (21)(b) and (22)(b). Interestingly, true pronouns do not have
such referential restrictions either, consider (23):
(23) János nek-ik vágta a labdá-t.
John dat-3pl went the ball-acc
‘John threw the ball against them.’

The pronoun nekik ‘to them’ can refer to humans and non-humans, alike.11 That
seems to create a problem if we, just like in the previous two cases, want to treat
the a-examples above and (23) analogously, and analyse the agreeing particle in
(21)(a) and (22)(a) as a pronoun.
One possible solution to this particular problem can be found along the
strong pronoun – weak pronoun distinction of Cardinaletti & Starke (1999).
Their proposal, which is aimed at cross-linguistic coverage, draws a systematic

.  The editors of this volume do not find the -human reading of this pronoun grammatical.
Arguably, this is indeed not the default reading, but many native speakers, including myself,
do find it a possible reading in a facilitating context. For example, I find (23) a perfect answer
to the question What happened to the windowpanes? Nevertheless, it could be the case that
there is a split between native speakers in accepting or rejecting the -human reading of such
pronouns, which is yet another instance of data inconsistency in this p-context. I put this issue
aside because it does not influence the main line of argumentation.
 György Rákosi

distinction between strong pronouns, weak pronouns and clitics. It is the first
two categories that are relevant for us. Weak pronouns are deficient in several
ways, and they are underspecified for the +/–human feature by default. Strong
pronouns, on the other hand, are not deficient structurally and they may not refer
to non-human entities.
I take the pronoun in (23) to be a weak pronoun, which is a natural assump-
tion given that the ‘real’ pronominal complement of the case marker is absent (or
is pro-dropped). This explains why it shows no referential restrictions. If the plural
agreeing particle in (21)(a) and (22)(a) is a pronoun and it forms an appositive
construction with the associate as É. Kiss (2002) suggests, than it has to be a strong
pronoun, since appositives require a strong pronoun head. The +human restriction
follows from the strong pronoun status if we accept the account of ­Cardinaletti &
Starke (1999). Therefore the data in (21) and (22) make the pronominal analysis of
the agreeing particle even more plausible rather than implausible.12

4.5  Interim conclusions

I have briefly overviewed three constructions above. What is common to each
is that agreeing plural particles seem to pattern up with real pronouns in these
constructions, while non-agreeing particles do not. I interpret this parallelism
as strong evidence in favour of (H2), and the lack of such parallelism as strong
­evidence in favour of (H1).
(H1) The particle is not a pronoun in the non-agreeing RPVC.
(H2) The particle is a pronoun in the agreeing RPVC.

This reasoning relies on the background assumption that grammatical elements

that show parallel behaviour are of the same category, and those that do not are
categorically distinct. The evidence we have summoned is not enough in itself to
establish either (H1) or (H2) as certain, though it does render both quite plausible.
The reason is that other constructions not investigated here may break the parallel
that underlies behind (H2) or they may establish it against (H1). Given that I am
not aware of such evidence at the moment, I assume that both (H1) and (H2) have
high plausibility.
This also means that my answer to the question that constitute (P2).
(P2) Modulo the difference in number marking, are the agreeing and the non-
agreeing RPVCs instances of the same construction type?

.  One could argue that the non-agreeing particles are also weak pronouns, and that is why
they are not subject to a referential constraint. However, even weak pronouns should show
Principle B affects, which non-agreeing particles do not, as we have seen above in Section 4.2.
 A case of disagreement 

is in the negative. In the final section, I conclude this chapter by pointing out
how this result helps to create the final p-context for the purposes of the current
argumentation.

5.  By way of summary: The final p-context

I have established the initial p-context of this argumentation in Section 3, where

we have seen that the initial setup that the literature on the agreeing reduplicating
particle construction creates is p-inconsistent as far as the grammaticality judge-
ments are concerned. Some simply reject such examples, whereas they pass with
flying colours for others. I formulated this p-problem as (P1):
(P1) How can the following p-inconsistency be treated?
(i) 0 < |The agreeing RPVC is grammatical in Hungarian.|linguistic intuition of
some native speakers < 1
(ii) 0 < |The agreeing RPVC is ungrammatical in Hungarian.|linguistic intuition
of some native speakers < 1

An easy solution of such inconsistencies is the assumption that there is dialectal

or idiolectal variation across the speakers. The former is not likely to be the case,
since the known dialectal background of the linguists involved in the debate does
not seem to correlate systematically with their judgements. It could still, however,
be the case that what accounts for the problem expressed in (P1) is accidentally
distributed idiolectal variation.
I have tried to argue that this assumption is not very fruitful in and of itself.
I have argued in Section 4 that a successful elaboration of the initial p-context
by questioning an underlying assumption brings important insights. The new
­p-problem identified was (P2), and I have suggested a negative answer.
(P2) Modulo the difference in number marking, are the agreeing and the non-
agreeing RPVCs instances of the same construction type?
(H1) The particle is not a pronoun in the non-agreeing RPVC.
(H2) The particle is a pronoun in the agreeing RPVC.

The negative answer to (P2) has followed from a plausible argumentation process
that has rendered (H1) and (H2) statements of high plausibility.
Thus the final solution to (P1), I argue, rests on the acceptance of (H2) as a
very plausible statement. (H2) requires an account of the agreeing reduplicating
particle verb construction in which the relation between the particle and the asso-
ciate is appositive in nature. This is the account that É. Kiss (2002) proposes for
both RPVCs, but I have argued that it has to be restricted to the agreeing type. The
plural particle, as a referential pronoun, can only be in an appositive relation with
 György Rákosi

the likewise ­referential associate. This construction is marked in the usual sense of
the term: because of the appositive structure, it is more complex and ‘heavier’ than
its unmarked counterpart, the non-agreeing RPVC; and it is also less frequent in
corpora or in speech.
Thus the issue, I believe, is not simply or only whether the agreeing RPVC is
grammatical or ungrammatical. The issue is, as I have tried to show, is that it is a
marked construction and native speakers may often show uncertain judgements
towards it and they may often try to avoid it altogether. This is not to deny that
there may be idiolectal variation. But if there is one, then it should also follow
from the structural difference between the agreeing and the non-agreeing RPVC
that I have established here. That is, some speakers may strongly disprefer or reject
the pronominal use of the reduplicating particle, and I suggest it is exactly these
speakers that disprefer or reject the agreeing RPVC.
This line of reasoning represents an instance of what Kertész & Rákosi (2012,
this volume: Subsection 2.6.2) call the Combinative Strategy, applied to the current
case in an attempt to tolerate the inconsistency that constitutes (P1) and to arrive at a
paraconsistent solution. We have found that it is much more motivated to treat state-
ments (i) and (ii) of (P1) as co-existing alternatives that can be ­maintained simulta-
neously. This requires the systematic separation of the two inconsistent p-contexts as
follows. First, it needs to be recognised that (i) and (ii) of (P1) characterise distinct
groups of native speakers. Second, the assumption that the ­agreeing and the non-
agreeing RPVC are instances of essentially the same construction type must be given
up: the particle is pronoun in the agreeing RPVC, but it is not in the non-agreeing
RPVC. Third, the pronominal use of the particle is a marked phenomenon and that
explains why even favourable native speakers may find it somewhat unusual, and
why many of them vacillate between ­dispreferring or outright rejecting it.
Needless to say, this account raises many further questions, the issue of
what it means to be data and evidence in linguistics being a prominent one. But
my intentions have been more modest in this chapter. I have tried to show that
­conscious reliance on the p-model of Kertész and Rákosi (2012, this volume) as
a metalinguistic framework of analysis helps us find a potential resolution to
an inconsistency that characterises the literature on the agreeing reduplicating
­particle construction of Hungarian.

Acknowledgments

The research reported here is supported, in part, by OTKA (Hungarian Scientific

Research Fund), grant number NK 100804; the TÁMOP-4.2.2/B-10/1-2010-0024
project, which is c­ o-financed by the European Union and the European Social
Fund; and by the MTA-DE Research Group for Theoretical Linguistics.
 A case of disagreement 

References

Abraham, Werner. 2010. Misleading homonymies, economical PPs in microvariation, and P as

a probe. In Mapping Spatial PPs. The Cartography of Syntactic Structures, Vol. 6, Guglielmo
Cinque & Luigi Rizzi (eds), 261–294. Oxford: OUP.
Ackerman, Farrell. 1987. Pronominal incorporation: The case of prefixal preverbs. In Approaches
to Hungarian 2, István Kenesei (ed), 213–260. Szeged: JATE.
Ackerman, Farrell. 1990. The morphological blocking principle and oblique pronominal
incorporation in Hungarian. In Grammatical Relations: A Cross-Theoretical Perspective,
­Katarzyna Dziwirek, Patrick Farrell & Errapel Meijas-Bikandi (eds), 1–19. Stanford CA:
CSLI.
Ackerman, Farrell. 2003. Lexeme derivation and multiword predicates in Hungarian. Acta
­Linguistica Hungarica 50(1–2): 7–32.
Ackerman, Farrell & LeSourd, Philip. 1997. Toward a lexical representation of phrasal predi-
cates. In Complex Predicates, Alex Alsina, Joan Bresnan & Peter Sells (eds), 203–246.
­Stanford CA: CSLI.
Ackerman, Farrell & Webelhuth, Gert. 1993. The composition of (dis)continuous predicates:
Lexical or syntactic? Acta Linguistica Hungarica 44: 317–340.
Cardinaletti, Anna & Starke, Michal. 1999. The typology of structural deficiency: A case study
of three classes of pronouns. In Clitics in the Languages of Europe. Empirical Approaches to
Language Typology, Henk C. Riemsdijk (ed.), 145–233. Berlin: Mouton de Gruyter.
É. Kiss, Katalin. 1998. Verbal prefixes or postpositions? Postpositional aspectualizers in
­Hungarian. In Approaches to Hungarian 6, Casper de Groot & István Kenesei (eds), Szeged:
JATE.
É. Kiss, Katalin. 2002. The Syntax of Hungarian. Cambridge: CUP.
Hegedűs, Veronika. In preparation. Predicative PPs and particles. Ch. 4 of Ph.D. dissertation in
preparation.
Kálmán, László & Trón, Viktor. 1999. A magyar igekötő egyeztetése (Verbal particle agreement
in Hungarian). In A mai magyar nyelv leírásának újabb módszerei IV, László Büky & Márta
Maleczki (eds), 203–211. Szeged: SZTE, Általános Nyelvészeti Tanszék.
Kenesei, István. 2005. Hungarian in focus. Review Article. Journal of Linguistics 41(2): 409–435.
Kertész, András. 2012. The ‘Galilean style in science’ and the inconsistency of linguistic
­theorising. Foundations of Science 17: 91–108.
Kertész, András & Rákosi, Csilla. 2012. Data and Evidence in Linguistics. A Plausible Argumenta-
tion Model. Cambridge: CUP.
Kertész, András & Rákosi, Csilla. This volume. The p-model of data and evidence in linguistics.
Németh T., Enikő. This volume. Hungarian verbs of natural phenomena with explicit and
implicit subject arguments: their use and occurrence in the light of data.
Noonan, Máire. 2010. À to zu. In Mapping spatial PPs. The Cartography of Syntactic Structures,
Vol. 6, Guglielmo Cinque & Luigi Rizzi (eds), 161–195. Oxford: OUP.
Rákosi, György & Laczkó, Tibor. 2011. Inflecting spatial particles and shadows of the past in
Hungarian. In Proceedings of the LFG ‘11 Conference. Miriam Butt & Tracy Holloway King
(eds), 440–460. Stanford CA: CSLI.
Rákosi, György & Laczkó, Tibor. In preparation. Spatial particles in dependencies in Hungarian.
Submitted to a special issue of Studia Linguistica.
Surányi, Balázs. 2009a. Locative particle and adverbial incorporation at the interfaces. In Adverbs
and Adverbial Adjuncts at the Interfaces, Katalin É. Kiss (eds), 39–74. Berlin: ­Mouton de
Gruyter.
 György Rákosi

Surányi, Balázs. 2009b. Adpositional preverbs, chain reduction, and phases. In Approaches to
Hungarian 11, Marcel den Dikken & Robert M. Vago (eds), 217–250. Amsterdam: John
Benjamins.
Ürögdi, Barbara. 2003. Feature doubling, aspectual structure, and expletives. In Proceedings of
the North East Linguistic Society 33, Makato Kadowaki & Shigeto Kawahara (eds), 425–444.
Amherst MA: University of Massachusetts.
 chapter 8

A plausibility-based model
of shifted indexicals

Zoltán Vecsey
MTA-DE Research Group for Theoretical Linguistics

In his groundbreaking formal work on context-dependent expressions, Kaplan

(1989a) argued that the conventional meaning of natural language indexicals
cannot be modified by sentential operators. Kaplan’s original theory, however,
has been recently challenged by empirical studies. It turned out that in certain
languages, for example in Amharic and Zazaki, attitude operators are able to
modify the meaning of indexicals. Some of these empirical findings appear to be
clearly inconsistent with the Kaplanian account. From a metatheoretical point of
view, the current debate about the operator-sensitivity of indexicals can also be
interpreted as a debate about the role of linguistic data. In following the p-model
and considering linguistic data as statements with plausibility values, one can
make some progress in this field of research.

1.  Indexicals in the scope of sentential operators

Influenced by David Kaplan’s seminal work (1989a, b), the dominant view on the
semantic interpretation of indexicals is based on the following familiar picture.1
Model-theoretic structures for indexical languages consist of two separable inter-
pretive domains: contexts of occurrence (c) and circumstances of evaluation (i).
Contexts are conceived of as collections of different parameters which are needed
to determine the denotations of indexical expressions. Such collections could be
minimally represented by quintuples of the form 〈s, t, l, w, g〉, where s is an agent,
t is a point of time, l is a location, w is a possible state of the world, and g is a
­(possibly partial) variable assignment function defined on the salient entities of w
at t and l. Circumstances of evaluation are symbolised traditionally with centered

.  In this chapter I will concentrate solely on the behaviour of pure indexicals. For simplicity’s
sake, I will ignore cases where pure indexicals are interpreted metaphorically or analogically.
 Zoltán Vecsey

world-time tuples 〈w, t〉, but if the complexity of the interpreted language fragment
requires this, they may also include a few further parameters. The number and
kind of additional parameters do not influence substantially the model-theoretic
role of circumstances. That role consists in assigning truth values to the contents of
sentences which contain one or more indexical expressions. Within the Kaplanian
framework, sentential contents are treated as structured propositions. If we are
willing to accept this approach, we can say that circumstances settle truth values
for propositions.
The model-theoretic interpretation of indexical languages is then basically a
two-step procedure. In the first step of the interpretation process indexicals are
evaluated with respect to a particular collection of relevant contextual parameters.
The evaluated objects are conceived of traditionally as conventionalised mean-
ings or lexical entries. Kaplan (1989a: 505) calls such objects ‘characters’. As an
example, consider the contextual occurrence of three paradigmatic indexicals in
an intended model M:2
(1) a. [[Ig]]c = AGENTg(c)
b. [[hereg]]c = LOCATIONg(c)
c. [[nowg]]c = TIMEg(c)

In (1)(a), (1)(b), and (1)(c) characters are evaluated with respect to their context
of occurrence. It is commonly assumed that the context of occurrence c and the
assignment function g of the model M together determine the content (denota-
tion) of paradigmatic indexicals. So given an appropriate assignment function g,
the content of the personal pronoun ‘I’ is the agent of c, the content of the locative
adverb ‘here’ is the most salient location of the occurrence of ‘here’ in c, and the
content of the temporal adverb ‘now’ is the most salient moment of time selected
by the occurrence of ‘now’ in c. Other indexicals such as ‘today’ or ‘actual’ are sup-
posed to behave in the same way.
The second step in the interpretation process identifies an actual or counter-
factual circumstance with respect to which it is possible to ask for the truth values
{1, 0} of structured propositions. In intended models, circumstances coincide with
the actual state of the world and the time of this world 〈w@, t@〉, that is to say, with
our actual world. Circumstances may be identified in other admissible models
with arbitrary possible worlds and time points.
Ultimately, then, indexical sentences are evaluated with respect to two differ-
ent collections of parameters. First, contexts determine the content of characters,

.  By ‘an intended model’ I mean simply a model whose contextual domain consists of some
set of objects of our actual world w@.
 A plausibility-based model of shifted indexicals 

thus contributing to the construction of structured propositions. Second, for each

structured proposition, circumstances yield truth values. The emerging picture is
one of double indexing:
(2) Step One:
a. [[Ig]]c = AGENTg(c)
b. [[hereg]]c = LOCATIONg(c)
c. [[nowg]]c = TIMEg(c)
Step Two:
d. [[I am here nowg]]c,i = 1 iff AGENTg(c) is at LOCATIONg(c) and at
TIMEg(c) in wi

Indexicals may denote different objects in different contexts, but it seems that every
particular context determines the denoted objects completely and definitely. The
personal pronoun ‘I’, for example, necessarily denotes in each of its ­occurrences
the agent of the context. If you are the only agent in c, then, of course, ‘I’ cannot
denote me truly in that very context.
This rather simple observation is the basis of the Kaplanian Reference Prin-
ciple, according to which indexicals are directly referential expressions (Kaplan
1989a: 492, 1989b: 569). The principle of direct reference claims that indexical
denotations are directly determined with respect to the relevant parameters of
the actual context of their occurrence. The mechanism of the determination is
called ‘direct’, since the denoted objects become part of the propositions without
the mediation of other propositional constituents. It is irrelevant how characters
find their denoted objects. Although their referential behaviour may be governed
by descriptive rules, these rules are not part of the constituent structure of propo-
sitions. The character of ‘I’ may be identified roughly with the descriptive rule
‘search for the agent who is speaking or writing’. At the ground level of propositions,
however, only the agent denoted by ‘I’ may have any model-theoretic significance.
If the Reference Principle is correct, then it is quite obvious that the content
of indexicals cannot be influenced by counterfactual circumstances. Moreover, it
seems that when an indexical expression denotes a particular object in a given
context, then it must denote the same object in all possible contexts (of its model
M). Therefore, taking into consideration the characteristics of double indexing
displayed above, [[αg]]c,i = DENg(c) may be regarded as a generally valid rule for
determining indexical denotations.
This rule does not mean, in and of itself, that circumstances cannot have an
autonomous function in the process of interpretation. As stated above, circum-
stances might be relevant to assigning truth values to propositions. Consider the
sentence in (2)(d). According to this example, the proposition expressed by ‘I am
 Zoltán Vecsey

here now’ should come out true in such contexts, where the agent of the sentence
is at the location of the utterance at the time of the utterance in the world of the
utterance. It is not hard to see that in intended models every context meets these
requirements. Uttered sentences are bound with necessity to their actual agent,
location and time parameters. So the question arises: could we have any model M*
in which the sentence ‘I am here now’ may be interpreted as false? On Kaplan’s
view, such models do not exist. The reason for this is that a counterfactual circum-
stance in which the agent is not at the location and time of the utterance would
contain an impossible world parameter. If one tried to detach the agent parameter
from the location and time parameters, the resulting picture would be obviously
incoherent. Viewed from this perspective, ‘[[I am here nowg]]c,i = 1’ counts as a
logical truth in Kaplan’s theory.
But now consider the apparent difference between the following two
interpretations:
(3) a. [[I am here nowg]]c,i = 1, via the Reference Principle and the rule ‘[[αg]]
c,i = DEN (c)’
g
b. [[Necessarily, I am here nowg]]c,i = 0, via Necessitation and the
­Reference Principle with the rule ‘[[α ]]c,i = DEN (c)’3
g g

This example illustrates clearly the truth-conditional significance of c­ ircumstances.

In (3)(b) the necessity operator modifies the world and time parameter of the orig-
inal circumstance of evaluation to the effect that the singular proposition about
the agent of (3)(a) should come out true with respect to all possible 〈w, t〉 tuples.
But it is beyond reasonable doubt that an agent’s location in space and time is a
contingent matter. The agent of (3)(b) could have been at other places at the time
of the occurrence of this sentence. This is why in (3)(b) the proposition expressed
by ‘Necessarily, I am here now’ must be false.
It would be an easy task to point out that tense and locational operators can
elicit the same effect as the necessity operator. All these operators seem to be able
to shift the circumstantial parameters of the evaluation process. Although an
example like (3)(b) does not fully reveal the logical and syntactical properties of
the phenomenon, it is now understandable why circumstance-shifting may alter
the truth values of propositions.
The Kaplanian theory becomes a bit more problematic, however, when we
turn to the case of the interaction between sentential operators and contextual
parameters. On the one hand, one might think that there is no way for operators

.  In the present setting, ‘Necessitation’ should be interpreted in the same way as in standard
modal logics.
 A plausibility-based model of shifted indexicals 

to act on contexts, because when operators begin their work the content of the
­characters in the embedded sentences is already secured. In fact, this is Kaplan’s
official view. On the other hand, one might assume that context-shifting is gov-
erned by the same scopal mechanisms as circumstance-shifting. It is sensible to
expect ‒ solely on syntactic grounds ‒ that within their scope operators can manip-
ulate contexts in the same way as circumstances.
As for the first view, Kaplan illustrates the immunity of the content of char-
acters to varying contexts with the locational operator ‘in some contexts it is true
that’ (Kaplan 1989a: 510). According to his interpretation, there is a sharp contrast
between the acceptability conditions of (4)(a) and (4)(b):
(4) a. [[I am tired nowg]]c,i = 1, iff AGENTg(c) is tired at LOCATIONg(c) and
at TIMEg(c) in wi
# [[IN SOME CONTEXTS IT IS TRUE THAT (I am tired now )]]c,i = 1,
b.  g
iff AGENT(c) is tired at LOCATIONg(c*) and at TIMEg(c*) in wi

(4)(b) seems to assume that there is at least one arbitrary context c* significantly
different from the actual context c, and in this context the agent of the sentence is
tired at the time of the utterance. At first blush this assumption is harmless. The
property of being tired may be truly self-attributed by agents in different contexts.
Accordingly, the proposition expressed by ‘In some contexts it is true that I am
tired now’ seems to be true if and only if the agent denoted by ‘I’ is at the appropri-
ate location and time and she is tired. So far, so good.
But on closer examination, (4)(b) is hard to accept as a correct truth condi-
tional structure, because the indexical expressions ‘I’ and ‘now’ in the embedded
sentence cannot denote arbitrary objects. If I am tired now, I am tired in my actual
context c. I could not be at the location of c* at this very moment. Contrary to
this fact, (4)(b) assumes that it may be true that I am now in c* and I am tired
right there. The problem is caused by the operator ‘in some contexts it is true’. The
example displayed in (4)(b) is unacceptable, because the operator in question can-
not interact with the content of ‘I’ and ‘now’.
The upshot of the Kaplanian argumentation is that operators which attempt to
modify the content of indexicals do not exist in natural languages such as English.
The main point of the Kaplanian view concerning the impossibility of context-
shift is aptly summarised by the Fixity Thesis:4
(FT) The denotation or propositional contribution of an indexical expression a is
fixed solely by the relevant parameters of its actual context of occurrence c,
and cannot be influenced by any sentential operators.

.  See Schlenker (2003a, 2003b) and Cable (2005) for other variants of the Fixity Thesis.
 Zoltán Vecsey

As we have seen, Kaplan derives (FT) from two key model-theoretic insights. The
first insight is that indexicals are context sensitive expressions that select their
denotations from their proprietary contextual domain. These domains must be
complete in the sense that they must contain all the parameters which are needed
to determine the content of characters. The second insight teaches us how to think
about the propositional contribution of the contextual occurrences of characters.
According to this insight, (i) characters contribute to propositions only the objects
they denote, and (ii) the denoted objects become part of propositions without
any descriptive material. In fact, (i) and (ii) are the general claims on which the
­Reference Principle is based.
Having outlined these fundamental model-theoretic ideas, the natural ques-
tion one might ask now is whether (FT) is indeed a valid thesis. Remember that
Kaplan explicitly denies the possibility of context-shifting operators only with
respect to English: “I am not saying we could not construct a language with such
operators, just that English is not one” (Kaplan 1989a: 510).
As can be easily observed, there is some ambiguity in Kaplan’s remark just
quoted above. One might emphasise that concerning the fundamentals and work-
ing of its indexical system, English is not an exceptional language. For example, it
would be an astonishing discovery about German or Hungarian, if it turned out
that indexical expressions in these languages do not obey the Reference Principle.
Of course, the German first person pronoun ‘ich’ and the Hungarian temporal
adverb ‘most’ (translated: ‘now’) obey the Reference Principle, and this is hardly
an accidental coincidence. The pivotal role of every indexical system is to help
speakers to locate themselves in physical and mental space and time. Obviously,
the realisation of this role may yield some subsidiary divergences between certain
natural languages that belong to different families. But in any case, a Kaplan-style
theory of indexicals seems called upon to give an account of those divergences that
arise from specific semantic or syntactic features of individual languages.
So I tend to think that the right question to ask is whether (FT) is a crosslin-
guistically valid thesis. For purposes of clarification, it is useful to distinguish two
aspects of the issue we are considering here. First, one might investigate the Kap-
lanian double indexing model with respect to its theoretical structure. Investiga-
tions of this kind concentrate only on the internal properties of the model: what is
analyzed is whether the constituents of the model are appropriately related to each
other, or whether double indexing provides a structurally complete framework
etc. Second, one might investigate how effectively and efficiently the model can be
applied. In this kind of investigation it might be questioned whether the Kaplanian
model is applicable to English and other natural languages. There is a sensible dif-
ference between the level of theoretical structure and the level of its application.
The latter aspect is what interests us here. So the question we should really ask is
 A plausibility-based model of shifted indexicals 

whether (FT) can be applied crosslinguistically. Unfortunately, this question is dif-

ficult to answer definitely. In the following two sections, I will try to explain the
sources of the difficulty.

2.  Introducing plausibility considerations

It is far from evident how to investigate the crosslinguistic applicability of (FT),

because the investigation involves a methodological dimension that is wholly
absent in Kaplan’s work.
First, it is not entirely clear how one can form an opinion about the applica-
bility of (FT) in English. It should not be forgotten that beyond its formal details
Kaplan’s thesis has an empirical angle as well. If (FT) has counterexamples in
­English, then it cannot be a correct thesis (with respect to English). But can one
make unquestionably true statements about the existence or non-existence of the
English counterexamples of (FT)? I do not think so. It is an empirical issue whether
a certain syntactic configuration with such and such semantic effects exists or does
not exist in a natural language, and unquestionable certainty in such matters is
rarely possible.
Second, the prospects are even worse when we try to say something about
the crosslinguistic applicability of (FT). In this case, one should ascertain whether
(FT) is applicable across all existing natural languages. That is, it has to be shown
that context-shifting operators do not exist at all. Again, the performance of this
empirical task seems to be impossible.
All this suggests that in order to investigate the phenomenon of context-shift
with considerable effectiveness we must revise, and to a certain extent modify,
Kaplan’s methodological stance. The following quotation from Rescher (2003)
may serve as a useful guide for orientation:
Presumptions arise in contexts where we have questions and need answers.
And when sufficient evidence for a conclusive answer is lacking, we must, in the
circumstances, settle for a more or less plausible one.
(Rescher 2003: 98f.; emphasis in the original)

The research situation described by Rescher is very similar to ours. We have a ques-
tion pertaining to the crosslinguistic applicability of (FT), but we do not know, at
the moment, how to answer this question conclusively. The method of plausible
argumentation seems to offer a way out of this difficulty. In their contribution to
this volume, Kertész and Rákosi put forward in essence the same idea:
In most cases our hypotheses are not statements the truth of which is guaranteed
by experience, supporting evidence, a proof or whatever. Rather, they are
 Zoltán Vecsey

plausible: we are ready to accept them to some extent on the basis of suitable
sources that support them.
(Kertész & Rákosi this volume: Section 2.2; emphasis in the original)

In Kertész and Rákosi’s framework, the plausibility of statements are signalised by

numerical values in the range from 0 to 1. The role of these numerical values is
comparative: they represent different strengths of plausibility (reliability, support-
edness, acceptance). Or, with other words, numerical values express the degree of
the acceptability of particular statements relative to a given set of their ­supporting
sources. The valuation |s|S = 0 expresses that the statement s has neutral ­plausibility
according to the source S. This means that if s has neutral plausibility, then nei-
ther s nor the negation of s is supported by the source S. The valuation |s|S = 1
expresses that the statement s is certainly true (i.e. it has a maximal p ­ lausibility
value) according to the source S. In empirical research settings |s|S = 1 can be
regarded as a limiting case, which may occur only in exceptional cases. All other
valuations express positive degrees of acceptability. For example, low plausibility
values such as 0.1 or 0.2 do not indicate the implausibility or unacceptability of
statements. Such values reflect merely that the support a statement receives from
its source is relatively weak.
The main motive for introducing plausibility considerations into the debate
around (FT) is the following. In elaborating and defending his thesis, Kaplan
(1989a, b) presented very few empirical examples from English.5 Thus, he left one
part of the supporting source of (FT) unexplicit. In the last few years a number
of empirical surveys have tried to fill in this gap. Surprisingly, these surveys have
come to the conflicting conclusion that there are natural languages in which (FT)
cannot be applied. There seems to be little hope of solving this conflict with the
traditional methodological means of model theory. The best we can do is to com-
pare the plausibility values of the conflicting statements and offer a solution with a
relatively greater plausibility value.

3.  A conflict between different sources of plausibility

As we can learn from the investigations of Kertész and Rákosi (this volume:
­Section 2.2, 2006, 2012: 69ff.), plausibility values come in degrees and are rela-
tive to a given source. We can differentiate between two types of these sources.
In some cases, plausibility values are assigned to statements on the ground of the

.  Curiously enough, Kaplan mentions a potential counterexample to his thesis, but in fact
he says nothing about it. See Kaplan (1989a: 510, Footnote 34).
 A plausibility-based model of shifted indexicals 

r­ eliability of their sources. The members of this type are called direct sources. In
other cases, the plausibility value of a given statement depends on the plausibility
of other statements. In such cases we speak about indirect sources.6 Interestingly,
most counterexamples to the crosslinguistic applicability of (FT) are built on state-
ments the plausibility of which is obtained from direct sources. Let us analyse the
supporting sources of these statements in some details.
The Amharic indexical first person pronoun is perhaps the best documented
counterexample to Kaplan’s thesis. Schlenker (2003b: 31) proposes to analyse the
following speech situation:
(5) Situation to be reported: John says: “I am a hero”.
a. English: John1 says that he1 is a hero.
b. English: # John1 says that I1 am a hero.
c. Amharic: ȷ̌on ȷ̌ǝgna nǝ-ññ yl-all
John hero be.pfv-1sobj 3m.say-aux.3m
‘John1 says that he1 is a hero.’ (lit.: John1 says that I1
am a hero.)

The reported speech in (5)(a) is, of course, unproblematic. In (5)(b), however,

English does not allow for the indexed pronoun to denote the subject of the matrix
sentence, because ‘I’ must denote the agent of the actual context. Therefore, there
arises a discrepancy between the denotation of ‘John’ and the established rule
[[Ig]]c,i = AGENTg(c) (see 1. a. above).
In contrast to English, the Amharic first person pronoun may denote the
agent of the reported speech act rather than of the actual speech act. That is, it can
denote John. Schlenker points out that in (5)(c) the attitude operator modifies the
content of the character of the pronoun in question. Thus, the embedded clause ‘I
am a hero’ expresses the singular proposition that John is a hero.
If Schlenker’s linguistic data were obtained from a reliable direct source, then
it is plausible to claim, as Schlenker does, that in Amharic the operator SAY is able
to modify the content of the Amharic counterpart of ‘I’ in embedded sentences,
which casts doubt on the crosslinguistic applicability of (FT).7

.  For further details see Kertész & Rákosi (this volume: Section 2.2, 2012: 67).
.  In accordance with the usage of Kertész and Rákosi (this volume: Section 3.1, 2012:
Section 13.1), by ‘linguistic data’ I mean statements about some aspect of a language or lan-
guages whose initial plausibility value stems from a direct source. Schlenker’s data cannot be
identified with the Amharic sentence tokens themselves, since in order to be theoretically
useful these sentence tokens must already be evaluated from Schlenker’s own linguistic point
of view. In this sense, linguistic data are strongly theory-dependent. In the remainder of the
text, I will follow this approach to linguistic data.
 Zoltán Vecsey

An alternative interpretation would be that in the example above the ­embedded

clause were quoted. But, according to Schlenker, this interpretation is not pos-
sible in Amharic attitude reports. In this respect Amharic is similar to English. In
spoken English ‘I’ or ‘you’ cannot normally be quoted in indirect discourse. The
presence of a quoted indexical element in attitude reports is not permitted, either
in Amharic or in English. Otherwise, the following sort of interpretation would be
acceptable: “‘Mary didn’t hear what John told her ‘you’ should bring’ should mean
‘Mary didn’t hear what John told her she (= Mary) should bring’, contrary to fact.”
(Schlenker 2003b: 69).
One further set of data strengthens the assumption made by Schlenker.
According to Schlenker, it is reasonable to suppose that in other syntactic posi-
tions the Amharic first person pronoun denotes the agent of its actual context. If
‘I’ occurs in subject position in a matrix sentence, for example, then it invariably
denotes the actual agent of the uttered sentence. This means that in Amharic the
first person pronoun behaves in an appropriate syntactic environment just like a
standard English indexical expression. As regards the direct sources of the plausi-
bility values of his data, Schlenker (2003b: 68) enumerates several authentic infor-
mants in his study. Given that the reliability of the linguistic intuition of native
speakers as a direct source is relatively high in this research setting, Schlenker can
convincingly argue that the reported speech in (5)(c) is a real instance of operator
controlled context-shift.
Some authors have recently claimed that operator controlled shifting is also
present in other natural languages such as Zazaki and Slave. Anand and Nevins
(2004), based on their empirical research, claim that in Zazaki all indexical expres-
sions are in principle shiftable. That is, the Zazaki counterparts to English ‘I’, ‘you’,
‘here’, or ‘yesterday’ all have the option of shifting when they occur within the
scope of the attitude verb ‘vano’ (translated: ‘say’).
Let us take again as an example the indexical ‘I’. According to the interpreta-
tion of Anand and Nevins (2004: 22), the evaluation of the first person pronoun in
the scope of ‘vano’ offers the following option:

(6) Situation to be reported: Hesen said: “I am rich”.

a. English: Hesen1 said that he1 is rich.
b. English: # Hesen1 said that I1 am rich.
c. Zazaki: Hεsenij (mik-ra) va kε εzj/k dεwletia
Hesen.obl (i.obl-to) said that I rich.be-prs
‘Hesen1 said that he1 is/I1 am rich.’

In English, (6)(a) reports correctly the de se attitude of the subject of the matrix
sentence (Hesen). (6)(b), however, represents a misleading report, because the
 A plausibility-based model of shifted indexicals 

English indexical ‘I’ denotes the agent of the reporting context, not the agent
(Hesen) of the original speech situation. Interestingly, the evaluation rules of
Zazaki indexicals allow both forms of reporting. In (6)(c) the embedded clause
mirrors an ambiguity concerning the effects of context-shifting. In the first case,
context-shifting is blocked, and the attitude holder (Hesen) is denoted by the third
person pronoun ‘he’, as is usual in English. In the second case, where shifting is
allowed, the content of the character of the first person pronoun ‘I’ is modified,
and it also denotes the attitude holder. Anand and Nevins (2004: 24) maintain that
the source of this ambiguity lies in the grammar of Zazaki. According to their own
account, there are empirical findings to support this claim. As in the Amharic
example, the crucial point is, again, that shifted indexicals in Zazaki cannot be
quotations. The best evidence for this is that A-extraction is prohibited in direct
quotations, but it is possible out of complements of attitude verbs with shifted
indexicals. Consider the following illustration:
(7) Situation to be reported: Hesen said: “The girl that I kissed is pretty”.
a. English and Zazaki: # The girl that Hesen said, “I kissed t” is pretty.
b. Zazaki: čεnεkε [kε Hεseni va mɨ t paci kεrda] rindεka
girl   that Hesen said I t kiss did pretty.be-prs
‘The girl that Hesen said {Hesen, I} kissed is pretty.’8

This example supports the interpretation according to which the shifting effect in
(6)(c) depends on the grammatical features of the attitude verb ‘vano’ rather than
on the mechanism of quoting. If this is so, then, on the basis of its grammar, the
Zazaki counterpart to English ‘say’ may function as a context-shifting operator.
Roughly the same could be said about the Slave attitude verb ‘hadi’ (translated:
‘say’). There are, however, two minor differences between Zazaki and Slave. In Slave
only the first person pronoun can be shifted in the scope of ‘hadi’, and the shift is
always obligatory. On the other hand, the Slave counterpart to E ­ nglish ‘want’ can
optionally modify the content of the character of the first person p
­ ronoun, and the
counterpart to ‘tell’ can influence optionally all embedded ­pronouns. At the end
of their investigation, Anand and Nevins (2004: 36) draw the tentative ­conclusion
that the analysed cases underpin the existence of Zazaki and Slave context-shifting
operators. It would follow from this that (FT) must be abandoned as an applicable
thesis in these languages.
Let us summarise where we are. On the ground of data whose plausibil-
ity originates from reliable direct sources both Schlenker and Anand & Nevins

.  The example is taken from Anand & Nevins (2004: 24).

 Zoltán Vecsey

c­ hallenge the crosslinguistic applicability of Kaplan’s thesis. We may represent the

theoretical situation with the following simple inference:
|If there are context-shifting languages, then (FT) is not crosslinguistically
applicable.|S = 1
0 < |There are context-shifting languages.|S < 1
0 < |(FT) is not crosslinguistically applicable.|I < 1

The first premise of this inference must be assigned the plausibility value 1 on
conceptual grounds. The plausibility value of the second premise depends on the
degree of the reliability of the arguments provided by Schlenker and Anand and
Nevins. If the reliability of these arguments increases, then the plausibility value of
the consequence increases as well.
One should take into consideration, however, that the plausibility value of the
second premise may be high on the basis of the arguments offered by Schlenker
and Anand and Nevins, but it may become implausible on the ground of other
arguments to be considered. The emergence of such a situation cannot be excluded
in principle, since the plausibility value of a given statement may be evaluated by
different sources differently.
The present case is peculiar in the sense that the plausibility value of the
second premise may be easily increased by involving further introspective data
obtained from other languages, provided that these data indicate the existence
of context-shift in these languages.9 But this premise could only be regarded as
implausible with respect to all available sources, if we were able to show that the
plausibility value of the arguments made by Schlenker and Anand and Nevins
arises from unreliable sources.
Stechow (2003) has recently worked out a syntactic theory which may be use-
ful in this respect. Stechow (2003: 380) rejects the view suggested by Schlenker and
Anand and Nevins, according to which sentential operators are capable of shifting
the content of indexicals. In Stechow’s conception, the alleged facts of shifting may
be explained better through the mechanism of feature deletion under semantic
binding. This explanation treats the symptoms of shifting as an effect of feature
deletion at the logical forms of sentences.
Stechow does not deny that attitude verbs are operators. But according to
­Stechow, attitude verbs such as TELL or SAY can never act on the content of

.  The initial plausibility value of the second premise may be increased, for example, by the
context-shifting phenomena found recently in Navajo. See Speas (2000). On the other hand,
if only such languages were found in which there is no context-shift, then the implausibility
of the statement ‘There are context-shifting languages’ would increase – on the ground of the
investigation of these languages as sources, of course.
 A plausibility-based model of shifted indexicals 

expressions of embedded sentences. If ‘I’ alters its default behaviour as in (8)(b),

then it is due to binding by the matrix verb SAY:
(8) a. Amharic: ȷ̌on ȷ̌ǝgna nǝ-ññ yl-all
b. English translation: John1 says that I1 am a hero.

The binding rule requires that in (8)(b) the attitude holder of the main clause be
co-indexed with the embedded first person pronoun. However, this would involve
an agreement failure both in English and Amharic.10 In order to avoid this failure
the AGENT feature of ‘I’ must be deleted roughly in the following way:11
(8) b′. [[JOHN SAYS THAT (I am a herog)]]c,i = 1, iff (John says that)
XAGENTg(c) is a hero in wi

On this alternative interpretation the acceptability of (8)(a) is ensured by the

deleted AGENT feature of the co-indexed pronoun. The deletion must happen
before the first step of the (semantic) evaluation process takes place, otherwise
‘I’ would unavoidably denote the agent of the context, just as the Kaplanian rule
[[Ig]]c = AGENTg(c) predicts.
Stechow’s solution relies on certain discoveries about embedded pronouns
made earlier by Kratzer (1998) and Heim (2004). So, it is obtained from indirect
sources. In particular, the interpretation in (8)(b’) may be seen as a result derived
from the formal syntax of semantic binding. This is a rather complex and in a
certain sense opaque source. Despite its complexity and opacity, it appears to be
reliable. Thus, given the introduction of this new source, we should modify the
structure of our earlier inference. Since, in contrast to Schlenker and Anand and
Nevins, Stechow argues that there are no context-shifting languages, we have to
apply a new second premise that is the negation of the earlier one: ‘There are no
context-shifting languages’. We arrive accordingly at a new conclusion: ‘(FT) is
crosslinguistically applicable’. The modified inference is as follows:
|If (FT) is crosslinguistically applicable, then there are no context-shifting
languages.|S = 1
0 < |There are no context-shifting languages.|S < 1
0 < |(FT) is crosslinguistically applicable.|I < 1

.  As we have seen, the Amharic first person pronoun does not produce such failures in
other syntactic positions. For further details on semantic binding see Büring (2005) and Ball
(2006).
.  (8)(b’) shows only the semantic results of feature deletion. The syntactic derivations can
be found in Stechow (2003: 397f.) and Anand (2006: 94–98).
 Zoltán Vecsey

In comparing the conclusions of these inferences, we can easily notice that both
‘(FT) is not crosslinguistically applicable’ and ‘(FT) is crosslinguistically applicable’
are plausible to a certain extent. This indicates that the current theoretical situa-
tion concerning the applicability of (FT) is inconsistent. The ground of the incon-
sistency is that both the statement ‘(FT) is crosslinguistically applicable’ and its
negation have been made plausible by some reliable source. This is a special case of
informational overdetermination. If we were to stop here, our analysis of Kaplan’s
thesis would prove to be ineffective, if not useless, for our intended purpose.

4.  A proposal for the resolution of the conflict

In order to eliminate the apparent inconsistency, we have to find an a­ ppropriate

problem-solving strategy. According to Kertész & Rákosi’s proposal, the most
effective way to handle inconsistencies is the heuristic method of successive re-
evaluation. When an inconsistent situation arises, one should carefully compare
the reliability of the plausibility sources of the conflicting statements and elaborate
the best possible solution to the revealed conflict. This evaluation process may
even be repeated several times. The final aim is the detection of all available solu-
tions and the decision as to which of them may be accepted as the resolution of the
given conflict.12 In agreeing with this, I think the natural reaction is to re-evaluate
some parts of our present theoretical situation.13
First, let us return to the theory advocated by Schlenker and Anand and
­Nevins. If their theory explains the phenomenon of context-shift correctly, then
it must be comprehensive at least to such an extent that it leads to precise predic-
tions. The Schlenkerian theory locates the roots of shifting to the featural makeup
of the lexical entries of Amharic indexicals. If an indexical expression has an
AGENT{+} feature in this language, then its content cannot be shifted. On the
other hand, if an indexical has an AGENT{+, –} feature, then its content may be
shifted. But this position leads to some unwelcome results. Even if we accept that
the Amharic first person pronoun has the feature AGENT{+, –}, we will find dif-
ficulties in extending the view to Slave. The Slave counterpart to English ‘want’
can optionally shift the content of the first person pronoun, and the counterpart
to ‘tell’ can shift optionally all embedded pronouns. Now, the problem is that if

.  See Kertész & Rákosi (this volume: Section 2.5–6, 2012: Section 10).
.  Adherents of the extended plausibility theory may apply more than one problem-solving
strategy in this situation. On the constraints of problem solving see Rescher (1976) and
Magnani (2009).
 A plausibility-based model of shifted indexicals 

the roots of shifting lie at the features of the lexical entries, then it is not clear why
there occurs any difference between ‘tell’ and ‘want’.14 If an indexical expression
shifts in the scope of ‘tell’, then it should have to shift in the scope of ‘want’ as well.
But, as we have learned, ‘want’ is able to shift only the content of the first person
pronoun. This is a contradiction.
Second, a closer examination would reveal that Stechow’s theory also has an
implicit problem. At first sight, it appears that the mechanism of feature deletion
offers a comprehensive explanation for the phenomenon of context-shift. What
we have seen in (8)(b’) is, however, only the first half of an explanation. We have
seen that ‘I’ has lost its AGENT feature under the binder SAY. But Stechow’s theory
does not shed light on the relationship between the residual features of ‘I’ and
‘John’. Specifically, it is not made clear how the pronoun in question may denote
the attitude holder:
(9) ?[[IAGENTg]]c,i = ATTITUDE HOLDERg(c) in wi

As long as the underlying mechanisms of (9) are not fully understood, Stechow’s
theory cannot be regarded as satisfactory.
In the remainder of this section, I will outline a theory – or, rather, a the-
ory-element – which has the following advantages: (i) it is able to eliminate the
inconsistency revealed by the re-evaluation of the theoretical situation above, and
(ii) it is able to assign plausibility values to the statements about context-shifting
languages and (FT).
In my view, Stechow’s theory is basically on the right track. But something is
missing from his view. As (9) shows, quantified first person pronouns in Amharic
and Slave require a passage from their contextually determined denotation to
another one. The passage has the following form:
(10) AGENTg(c) in wi → ATTITUDE HOLDERg(c) in wi

If we try to generalise the phenomenon of passage for all shiftable indexicals, then
we get the following form:
(11) DENg(c) in wi → DEN*g(c) in wi

In (11), the second denotation DEN*g(c) functions as a proxy of the first one.
The role of the proxy is to make it possible to evaluate the shifted content. It is
important to see that if DENg(c) is given, then DEN*g(c) cannot be arbitrarily
determined. Not any candidate object can function as a proxy for DENg(c). There

.  Compare Anand (2006: 91). It must be added that the view advocated in Anand & Nevins
(2004) is rejected by Anand (2006). This latter work rejects the view according to which
­indexical shifting is driven by context-shifting operators.
 Zoltán Vecsey

must be a passable passage between DENg(c) and DEN*g(c). And my contention

is that the passage is pragmatic.
The essence of the pragmatic account is this. It should be seen that neither the
character itself nor the parameter structure of the context can explain how shift-
ing may happen. It is also unclear how can the appropriate proxy for DENg(c) be
selected.
The explanation of these unclear features is that the possibility of s­hifting
depends on an explanatory component which lies outside of these formal
­structures. Or, to be more precise, there is a set of explanatory components whose
members make the passage either individually or collectively possible. It is beside
the point to provide a fine-grained identity criterion for these explanatory compo-
nents. There is no need to be too specific here. It is sufficient to say that they belong
to the primary pragmatic processes which take place before the first step of the
semantic interpretation of indexicals begins. Obviously enough, the appearance
of such processes is triggered in large part by extralinguistic factors such as shared
background knowledge, conventionalised norms of reasoning, c­ommunicative
intentions etc.
Let us refer comprehensively to these explanatory components with the sym-
bol p. Then we can say that the correct evaluation of shifted indexicals depends on
the presence or absence of p. If a occurs in a context c, and p is present in c, then
there is a passage from DENg(c) to DEN*g(c) and the denotation of a becomes
determined through this passage. In the absence of p, the passage is closed and
a remains unevaluated. This allows us to formulate a pragmatic constraint on
context-shift:15
(PC) If in a given context c [[OP(αg)]]c,i = DEN*g(c) is the case, then to evaluate
a correctly in c there must be a pragmatic passage p between DENg(c) and
DEN*g(c) in wi
This pragmatic constraint indicates that the modifications of characters are
licensed at the presemantic level of the evaluation process. In this respect, we do
not stray too far from Stechow’s theory. However, the explanatory power of the
theory is improved in that this new version also explains the prerequisites of shift-
ing. For example, (PC) makes understandable how in (8b) the indexical ‘I’ may
denote the attitude holder ‘John’ and not someone else.16

.  In (PC), the symbol OP represents an appropriate operator that acts on the content of a.
.  Moreover, if we delete the OP component from (PC), we can also explain the cases of
free context-shift. Free context-shift may occur, for example, in free indirect discourses or in
historical narratives even in English. On this topic see Sidelle (1991) and Sharvit (2004).
 A plausibility-based model of shifted indexicals 

5.  Conclusion

Remember, we have departed from Kaplan’s original theory about the impossibil-
ity of operator controlled context-shift. As a result of metatheoretical reflection,
we have extended the Kaplanian theory by adding new data sources to it. There
then emerged a conflict because the plausibility of a statement was judged differ-
ently by different sources: we have realised that the current theoretical situation
concerning the crosslinguistic applicability of (FT) is inconsistent. In the following
step, we have partly re-evaluated the extended theory. This has brought to light
that the re-evaluated theory still suffers from some deficiencies. In particular, it
has turned out that the view advocated by Schlenker and Anand and Nevins leads
to contradictory consequences, and Stechow’s account has proved to be essentially
incomplete. In the final step, we have tried to remove these deficiencies by intro-
ducing a new theory-element into the debate.
According to the position we have arrived at, context-shift is a binding
­phenomenon that is governed language-externally by a pragmatic constraint.
From this it can be seen that our final position acknowledges the existence of
context-shifting languages, but denies the possibility of operator controlled con-
text-shift. In the end, what we have done may be seen as an attempt to provide a
consistent solution to the problem raised by the inconsistent approaches to the
crosslinguistic applicability of (FT). It should be noted, however, that (PC) is
no more than a provisional resolution to the current problem. It is by no means
a definitive solution. For it rests on a set of theoretical statements that are not
maximally reliable, but only to a certain extent plausible. For example, the state-
ment according to which a pragmatic component is needed to explain the pos-
sibility of context-shift cannot be regarded as true with certainty. Statements
like this are at best plausible in the light of the certain sources. It should also
be noted that (PC) might lead to further problems and, in turn, the analysis of
these problems might lead to new insights that modify the plausibility status of
(PC). That is, (PC) should be seen as a fallible attempt to formulate a pragmatic
constraint on shifting.
In spite of its provisionality and fallibility, the theoretical position advanced
in this chapter has a notable advantage. If we compare the applied methodolo-
gies of the rival positions in the debate revolving around Kaplan’s theory, then
we can see that (PC) represents the most consistent and wide-ranging approach
to the problem of shifted indexicals. In sum, it should be stressed once more that
without supplementing the methodology of Kaplan’s double index theory with
plausibility considerations, we would not have been able to develop our alterna-
tive account.
 Zoltán Vecsey

Acknowledgements

Work on the present chapter was supported by the MTA-DE Research Group for
Theoretical Linguistics and the TÁMOP-4.2.2/B-10/1-2010-0024 project, which is
co-financed by the European Union and the European Social Fund.

References

Anand, Pranav. 2006. De de se. Ph.D. dissertation, MIT.

Anand, Pranav & Nevins, Andrew. 2004. Shifty operators in changing contexts. In Proceedings
of the 14th Conference on Semantics and Linguistic Theory, Kazuha Watanabe & Robert B.
Young (eds), 20–37. Ithaca NY: CLC Publications.
Ball, Derek. 2006. Bound variables and monstrous semantics. Paper presented at The 34th
Annual Meeting of the Society for Exact Philosophy.
Büring, Daniel. 2005. Binding Theory. Cambridge: CUP.
Cable, Seth. 2005. Binding local person pronouns without semantically empty features. Unpub-
lished manuscript, MIT, http://people.umass.edu/scable/papers/binding-local-pronouns.
pdf.
Heim, Irene. 2004. Puzzling reflexive pronouns in de se reports. Unpublished handout (presented
at Bielefeld), 22 December 2004, 〈http://semanticsarchive.net/Archive/ mVkNjZlY/〉.
Kaplan, David. 1989a. Demonstratives. In Themes from Kaplan, Joseph Almog, John Perry &
Howard Wettstein (eds), 481–563. Oxford: OUP.
Kaplan, David. 1989b. Afterthoughts. In Themes from Kaplan, Joseph Almog, John Perry &
Howard Wettstein (eds), 565–614. Oxford: OUP.
Kertész, András & Rákosi, Csilla. 2006. Inconsistency and plausible reasoning in an analysis
of German affricates. A case study in the philosophy of linguistics. Language Sciences
28: 386–423.
Kertész, András & Rákosi, Csilla. 2012. Data and Evidence in Linguistics. A Plausible Argumenta-
tion Model. Cambridge: CUP.
Kertész, András & Rákosi, Csilla. this volume. The p-model of data and evidence in linguistics.
Kratzer, Angelika. 1998. More structural analogies between pronouns and tenses. In SALT VIII,
Devon Strolovitch & Anna Lawson (eds), 92–110. Amherst MA: GLSA.
Magnani, Lorenzo. 2009. Abductive Cognition. Berlin: Springer.
Rescher, Nicholas. 1976. Plausible Reasoning. Assen: Van Gorcum.
Rescher, Nicholas. 2003. Epistemology: An Introduction to the Theory of Knowledge. Albany
NY: State University of New York Press.
Schlenker, Philippe. 2003a. Indexicality, logophoricity, and plural pronouns. In Research in
Afroasiatic Grammar II. Selected Papers from the Fifth Conferenceon Afroasiatic Languages,
Paris, 2000. [Current Issues in Linguistic Theory 241], Jacqueline Lecarme (ed.), 409–428.
Amsterdam: John Benjamins.
Schlenker, Philippe. 2003b. A plea for monsters. Linguistics and Philosophy 26: 29–120.
 A plausibility-based model of shifted indexicals 

Sharvit, Yael. 2004. Free indirect discourse and de re pronouns. In Proceedings of Semantics and
Linguistic Theory (SALT) 14, Robert B. Young (ed.), 305–322. Ithaca NY: CLC Publications.
Sidelle, Alan. 1991. The answering machine paradox. Canadian Journal of Philosophy 81: 525–539.
Speas, Margaret. 2000. Person and point of view in Navajo. In WCCFL Papers in Honor of Ken
Hale, Eloise Jelinek (ed.), 19–38. Cambridge MA: The MIT Press.
von Stechow, Arnim. 2003. Feature deletion under semantic binding: Tense, person, and mood
under verbal quantifiers. In NELS 33. Proceedings of the Thirty-Third Annual Meeting of
the North East Linguistic Society, Makoto Kadowaki & Shigeto Kawahara (eds), 379–404.
Amherst MA: GLSA.
part iii

Metatheoretical applications
 chapter 9

Thought experiments and real experiments

as converging data sources in pragmatics

András Kertész1, 2 & Csilla Rákosi2

1University of Debrecen, Institute of German Studies,
2MTA-DE Research Group for Theoretical Linguistics

The aim of the present chapter is to reconstruct the relationship between thought
experiments and real experiments in pragmatics. In the first part, Searle’s (1969)
thought experiment on Kennst du das Land, wo die Zitronen blühn? is analysed
with the help of the p-model. The second part reconstructs the argumentation
structure of the real experiments reported on in Holtgraves & Ashley (2001),
and shows that the latter and Searle’s thought experiment closely interact.
Basically, their relationship is characterised by the dynamic process of the cyclic,
prismatic and retrospective re-evalutation of information in the sense of the
p-model. However, they have different strengths and weaknesses. Accordingly,
the generalised finding is that the conscious integration of real and thought
experiments as data sources within some research may considerably enhance the
reliability of pragmatic hypotheses.

1.  Introduction

One of the most important insights obtained in the course of the current discus-
sion on linguistic data and evidence is that in many cases it is not sufficient to
restrict one’s data to just one data type and data source. Rather, the present state
of the discussion seems to suggest the integration of several data sources and data
types. However, this desideratum is highly problematic. In particular, it is any-
thing but clear which data sources and data types can be integrated, how integra-
tion can be achieved and under what conditions. With respect to these questions,
the relationship between thought experiments and real experiments in pragmatics
is especially challenging.
Austin’s, Searle’s, Grice’s and others’ thought experiments shaped the clas-
sical topics of pragmatics; moreover, without further thought experiments later
approaches could not have been elaborated. Psycholinguistic experiments were
 András Kertész & Csilla Rákosi

sparse and did not gain ground for decades (see Noveck & Sperber 2007: 2). By
now they have, however, become increasingly popular. Basically, this development
could proceed in two directions.
The first possible direction is the replacement of thought experiments by real
ones. The rapid development of experimental pragmatics has triggered decisive
changes in the state of the art. The relevance of the current focus on real experi-
ments in pragmatics is beyond doubt: the rise of experimental pragmatics has been
accompanied by the introduction of new methods, the detection of new research
topics, the publication of new findings and the restructuring of the scientific com-
munities involved. Furthermore, there has also been a revision of earlier hypothe-
ses and assumptions, solutions to a series of unsolved problems were proposed and
attempts were made at substantial modifications of methodological rules applied
in pragmatics research. For example, in Noveck and Sperber’s characterisation,
traditional pragmatics relied on pragmatic intuitions which must be regarded as
results of thought experiments and cannot be interpreted as “facts” in the same
sense as data which are normally regarded as objects of linguistic theorising:

It is a mistake to believe that pragmatic intuitions of the kind used in pragmatics

are data of the same kind as semantic intuitions used in semantics. Genuine
pragmatic intuitions are those that addressees have about the intended meaning
of an utterance addressed to them. Quite generally, pragmatic intuitions invoked
in theoretical pragmatics are not about actual utterances addressed to the reader
of a pragmatic article, but about hypothetical cases involving imaginary or generic
interlocutors. Pragmatic intuitions on hypothetical utterances have proved useful
in a variety of ways, but it is important to keep in mind that these are not about
how an utterance is interpreted, but about how an utterance would be interpreted
if it were produced in a specific situation by a speaker addressing a listener,
with referring expressions having actual referents, and so on. These intuitions
are educated guesses – and, no doubt, generally good ones – about hypothetical
pragmatic facts, but are not themselves pragmatic facts and they may well [be]
in error. That is, we may be wrong about how, in fact, we would interpret a given
utterance in a given context.
(Noveck & Sperber 2007: 2f.; bold emphasis as in the original, italics added)

The features mentioned in this quotation – intuition, hypothetical cases, imagi-

nary situation, educated guess – strongly resemble those features which the philo-
sophical literature on thought experiments discusses. Although there are many
different views of the nature of thought experiments, most authors agree that
a thought experiment is an answer to a ‘What would be the case if?’ question.
This answer is hypothetical: it makes use of an imaginary situation and intuition.
Therefore, as a point of departure, the use of pragmatic intuition in the sense men-
tioned in the quotation basically seems to correspond to what is generally meant
 Thought experiments and real experiments as converging data sources in pragmatics 

by thought experiments. If so, then it follows that neither pragmatic intuitions nor
results gained from other kinds of thought experiments can be used for testing
hypotheses, they can only motivate the raising of hypotheses. This is in accord
with the requirements of the standard view of linguistic data and evidence that
thought experiments (and particularly, pragmatic intuitions) cannot play the role
of evidence for or against hypotheses.1 From this Noveck and Sperber conclude
that they have to be replaced by experimental data:
Besides helping compensate for the inherent limits of pragmatic intuitions, an
experimental approach can provide crucial evidence when deciding between
alternative theories that may agree on the content of the interpretations
of utterances, but that have different implications regarding the cognitive
mechanisms through which these interpretations are arrived at.
(Noveck & Sperber 2007: 2f.; emphasis added)

According to this view, there is a kind of linear development in pragmatics as a

result of which the method of thought experimentation should give way to experi-
mental pragmatics.2
The second possibility is to acknowledge that experimental pragmatics has
not been built on a ground which was prepared by thought experiments alone but
that thought experiments and real experiments are intertwined and co-exist:
While the invention of (more or less fanciful) stories still plays a role in standard
pragmatic argumentation, it appears that even in experimental pragmatics stories
are an important part of methodology. (Meibauer 2012: 775)

At present, one cannot decide which of these alternative directions future develop-
ment will take, and whether there are further possibilities as well. Therefore, one of
the most relevant, but still unclarified, methodological aspects of current pragmat-
ics is the relationship between thought experiments and real experiments. Thus, the
present chapter centres on the problem (P) below:
(P) What kind of relationship is there between thought experiments and real
experiments as data sources in pragmatics?

.  For the notion of the standard view of linguistic data and evidence see Kertész & Rákosi
(this volume a, 2012: Section 3.3).
.  In a considerably milder fashion, Noveck and Reboul (2008: 5f.) evaluate the appear-
ance of experiments in pragmatics as the result of a linear development, too, where thought
­experiments are interpreted only as (historical) starting points. Nevertheless, they interpret
the r­ elationship between “traditional” pragmatics and experimental pragmatics not as a con-
flict or opposition but as an example of “how the cognitive sciences can work together in
harmony” (Noveck & Reboul 2008: 5).
 András Kertész & Csilla Rákosi

In order to get an insight into the nature of this relationship, we will apply Kertész
and Rákosi’s (2012) p-model as summarised in Kertész & Rákosi (this volume b).
In Section, 2 we will provide a rough overview of the philosophical literature
on thought experiments. We will highlight two points. First, according to sev-
eral philosophers of science, there is a close relationship between real and thought
experiments. Second, thought experiments are interpreted in a part of the current
literature as a special kind of reasoning, and experimental narratives are supposed
to be constitutive elements of both real and thought experiments. Therefore, the
application of an argumentation theoretical model to real and thought experi-
ments seems to be well-motivated. We have already proposed a model of psycho-
linguistic experiments based on the p-model (cf. Kertész & Rákosi 2012; Rákosi
2011a, b, 2012). Since the p-model characterises real experiments in such a way
that it assumes that they are organised by a plausible argumentation process, we
will try to characterise thought experiments along similar lines (see also Kertész
2010; Kertész & Kiefer 2013).
In order to illustrate how the p-model works and in order to exemplify how
it yields a possible solution to (P), Section 3 will be devoted to a case study. We
will reconstruct the argumentation structure of one of Searle’s (1969) thought
experiments.
Section 4 will focus on another case study: the analysis of real experiments as
discussed in Holtgraves & Ashley (2001) which are related to the thought experi-
ment discussed in Section 3.
On the basis of our analyses in Sections 3 and 4, in Section 5 we will reveal dif-
ferent aspects of the relationship between Searle’s thought experiment and Holt-
graves and Ashley’s real experiments. After generalising our results, we will obtain
the solution to (P).
Finally, in Section 6 we will summarise our findings and relate them to the
contemporary discussion on data and evidence in linguistics.

2.  On the state of the art

2.1  Current views on thought experiments in science

Thought experiments have been highly appreciated and frequently applied tools
in science since Galileo:
There is widespread agreement that thought experiments play a central role
both in philosophy and in the natural sciences and general acceptance of the
importance and enormous influence and value of some of the well-known thought
experiments in the natural sciences, like Maxwell’s demon, Einstein’s elevator or
 Thought experiments and real experiments as converging data sources in pragmatics 

Schrödinger’s cat. The 17th century saw some of its most brilliant practitioners
in Galileo, Descartes, Newton, and Leibniz. And in our own time, the creation
of quantum mechanics and relativity are almost unthinkable without the crucial
role played by thought experiments. (Brown & Fehige 2011)

In contrast, in the philosophy of science – under the influence of the harsh criti-
cism of Pierre Duhem and the standard view of the analytical philosophy of
­science – for a long time they were attributed only a psychological relevance as
being solely a means of discovery but not of the justification of theories.3 Partly on
the basis of the rediscovery of Ernst Mach’s views on the subject, over the last two
decades or so thought experiments have become an attractive topic in the philoso-
phy of science and there has been a radical turn in the way they are evaluated. For
most authors the focus no longer seems to be on deciding whether or not they are
legitimate tools in scientific theorising but rather on metascientific reflection con-
cerning their common features, classification, structure, functioning, limits, etc.
Although there is no widely accepted definition of the concept ‘thought exper-
iment’, most authors agree that argumentation belongs to the constitutive elements
of thought experiments. Nevertheless, there are several markedly different views
about the nature and role of these argumentative aspects.
As we have already mentioned, many authors agree that thought experiments
consider the consequences of a hypothesised state of affairs. However, this assump-
tion gives rise to different lines of reasoning. For example, John Norton assumes
that thought experiments are just “ordinary argumentation that is disguised in a
vivid pictorial or narrative form. […] They can do nothing more epistemically
than can argumentation” (Norton 2002: 1).
Norton’s position did not find general acceptance and became the object of
severe criticism. For example, Nersessian (1993) argues that although argumenta-
tion is one of their constitutive elements, thought experiments are basically of a
non-argumentative character:

While I agree with Norton (1991) that thought experiments can often be
reconstructed as arguments, the modeling function cannot be supplanted by an
argument. As Norton acknowledges, the argument can be constructed only after
the fact. That is, on my account, the argument is not evident until after the thought
experiment has been constructed and executed. […] real tworld experimental
outcomes can be recast in argument form as well, but no one would argue that the
experiment can be replaced by the argument. (Nersessian 1993: 297)

.  For more on the history of the philosophy of thought experiments, see e.g. Nersessian
(1993), Kühne (2005), Moue et al. (2006).
 András Kertész & Csilla Rákosi

As this quotation suggests, Nersessian (1993: 292) distinguishes between the

“original” thought experiment, on the one hand, and the “narrative” account of
it, on the other. The former involves, first of all, the construction and manipula-
tion of a mental model of some imagined situation by the thought experimenter.
The latter makes it possible that members of the scientific community are capa-
ble of reconstructing the mental model and performing the same manipulations
to it.
In the same vein, while emphasising the role of argumentation, Gooding
points out that thought experiments cannot be reduced to deductive inferences
as Norton requires. According to his view, thought experiments – similarly to
real experiments – involve several different kinds of activities and are proce-
dural instead of being simply propositional and static. According to Gooding
(1993: 280), thought experiments and real experiments alike have to be regarded
as forms of experimental reasoning. This means that they go beyond pure theoris-
ing, i.e. generalising and systematising our experiences about the world. Both are
special cases of the activity of “experimenting”, i.e. making experiences by conduct-
ing strictly regulated procedures under special, carefully controlled circumstances.
Similarly to Nersessian, Gooding (1993) differentiates between the process of real/
thought experimentation and its narrative account:
It is still the case that most people – including most scientists – encounter
experiments through narratives. The narrative elements of arguments that invoke
real experiments require the same ability to participate. […] To explain the force
of an experiment it helps to understand it as a process to be worked through,
rather than as a logical structure. (Gooding 1993: 283)

Gooding remarks that his and Nersessian’s standpoints are close to each other.
They share a common general view according to which thought experiments and
real experiments are cognates. Nevertheless, they take different perspectives: while
Gooding’s point of reference seems to be real experiments in physics, Nersessian’s
position can be interpreted as that of a cognitive psychologist.4

2.2  Points of departure for the solution of (P)

We accept Nersessian’s and Gooding’s view that real and thought experiments
are kindred. Nevertheless, we see the role of argumentation in both kinds of
experiments differently from these two authors. In Kertész & Rákosi (2012),

.  “Like Nersessian (1991, 1993) I regard thought experiments as process-narratives to

which visualization is essential. Where she has developed a mental-models account of the
process, I emphasize the role of embodiment in visualization.” (Gooding 1993: 281)
 Thought experiments and real experiments as converging data sources in pragmatics 

Rákosi (2011a, b) and Rákosi (2012) we presented a model of psycholinguistic

experiments and argued that argumentation has a twofold task in real experi-
ments. First, the experimental process is organised and conducted by an argu-
mentation process which remains the private affair of the experimenters. It aims
at the clarification of the relationship among different components and stages
of the experimental procedure in order to obtain a coherent and comprehen-
sive picture of the state of affairs at issue from a multitude of uncertain, fal-
lible and conflicting pieces of information. Second, the main steps and results
of the experimental process are transformed to a piece of public argumentation
as presented in the experimental report. In analogy to the term ‘experimental
report’ used with respect to real experiments, in the case of thought experi-
ments we will call such a presentation a thought experimental report. Both kinds
of reports present those facets of the real/thought experimental process that
might be ­pertinent to the evaluation and acceptance of the results. To this end,
relevant pieces of information are selected from the original, non-public argu-
mentation and arranged into a well-structured chain of arguments leading from
the problems raised to the data which have been obtained by carrying out the
real/thought experiment. Nevertheless, the real/thought experimental report
still contains traces of the original argumentation process organising the real/
thought experimental process.5
This chain of thought motivates the elaboration of a metatheoretical model of
thought experiments which is based on the idea that the similarities of real and
thought experiments as well as the peculiarities of the latter might be grasped
from an argumentation theoretical point of view. We hypothesise that the p-model
meets this desideratum. In the next subsection we will show how it can be applied
to one of the most famous classical thought experiments in pragmatics: Searle’s
Kennst du das Land, wo die Zitronen blühen? 6 With this thought experiment Searle
argues both against Grice’s explication of the notion of ‘non-natural meaning’ and
for his own account of illocutionary acts. In accordance with the p-model, we will
reconstruct the presentation of this thought experiment in Searle (1969) – that is,
his thought experimental report – as a cyclic and prismatic process of the retro-
spective re-evaluation of the information at hand.

.  Similarly to real experiments, thought experiments are not directly accessible to the
reader but they can be partially reconstructed from the paper describing and summarising
them. Therefore, the primary object of an argumentation theoretical reconstruction can be the
paper presenting the results, that is, the ‘(thought) experimental report’.
.  In Goethe’s original poem, one finds blüh’n instead of blühen. In quoting Searle, we will
preserve his way of quoting Goethe.
 András Kertész & Csilla Rákosi

3.  Case study 1: Thought experimental report in Searle (1969)

3.1  Setting up the starting p-context

The initial problem Searle raises is this:
(1) “[W]hat is it for one to mean something by what one says […]”?
(Searle 1969: 42)

The starting p-context includes Grice’s (1989a [1957]) explication of the notion
of ‘non-natural meaning’ which is one of the central theses of his model of
communication:7
(2) “[t]o say that a speaker S meant something by X is to say that S intended
the utterance of X to produce some effect in a hearer H by means of the
­recognition of this intention.” (Searle 1969: 43)

(2) receives a positive plausibility value on the basis of Grice (1989a [1957]) as a
direct source. This source is part of the p-context of Searle’s argumentation, but
its reliability is not sufficient to assign to it a higher plausibility value. Therefore,
Searle seems to be ready to acknowledge Grice’s paper as a reliable source only to
a limited extent and only provisionally. The reason for this decision is the circum-
stance that Searle does not find Grice’s argumentation in support of (2) convinc-
ing, because, according to him,
(3) (2) “fails to account for the extent to which meaning can be a matter of rules
or conventions”. (Searle 1969: 43)

Should one accept (3), then (2) would be implausible; nevertheless, (3) is at this
point not reliable enough to legitimise the rejection of (2). Thus, the starting
p-context is p-inconsistent in the sense of Subsection 2.5 of Kertész & Rákosi (this
volume b). Accordingly, Searle’s aim can be characterised in such a way that he
tries to increase the plausibility value of (3) and to decrease that of (2).

3.2  The elaboration of the starting p-context

First, Searle applies the Exclusive Strategy (Kertész & Rákosi, this volume b:
­Subsection 2.6.2), that is, he develops a p-context version based on (2) in order to
examine whether it is a tenable alternative. Searle extends it by the consideration of

.  Following Searle, we will speak simply of ‘meaning’, because we are not concerned with
such uses of ‘mean’ as e.g. Clouds mean rain which Grice (1989a [1957]) distinguishes from
‘non-natural meaning’.
 Thought experiments and real experiments as converging data sources in pragmatics 

a new source – namely, his thought experiment. For convenience we quote Searle’s
thought experimental report at length:8

Suppose that I am an American soldier in the Second World War and that
(4) a. 
I am captured by Italian Troops. And suppose also that I wish to get these
troops to believe that I am a German soldier in order to get them to release
me. What I would like to do is to tell them in German or Italian that I am a
German soldier. But let us suppose I don’t know enough German or Italian
to do that. So I, as it were, attempt to put on a show of telling them that I
am a German soldier by reciting those few bits of German I know, ­trusting
that they don’t know enough German to see through my plan. Let us
­suppose I know only one line of German which I remember from a poem
I had to memorise in a high school German course. Therefore, I, a captured
­American, address my Italian captors with the following sentence: Kennst
du das Land, wo die Zitronen blühen?
b. Now, let us describe the situation in Gricean terms. I intend to produce a
certain effect in them, namely, the effect of believing that I am a German
soldier, and I intend to produce this effect by means of their recognition of
my intention. I intend that they should think that what I am trying to tell
them is that I am a German soldier. But does it follow from this account
that when I say, Kennst du das Land…etc. what I mean is ‘I am a German
soldier’?
c. Not only does it not follow, but in this case I find myself disinclined to say
that when I utter the German sentence what I mean is ‘I am a German
soldier’, or even ‘Ich bin ein deutscher Soldat’, because what the words mean
and what I remember that they mean is ‘Knowest thou the land where
the lemon trees bloom?’ Of course, I want my captors to be deceived into
thinking that what I mean is: “I am a German soldier”, but part of what is
involved in that is getting them to think that that is what the words I utter
mean in German.
d. In the Philosophical Investigations, Wittgenstein (discussing a different
problem) writes “Say ‘It’s cold here’ and mean ‘it’s warm here’”.
e. The reason we are unable to do this without further stage setting is that
what we can mean is at least sometimes a function of what we are saying.
f. Meaning is more than a matter of intention, it is also at least sometimes a
matter of convention.

.  We will divide the quotations into smaller units in order to make it easier to refer to parts
of the quotation to be analysed. The units do not necessarily correspond to paragraphs in the
original text.
 András Kertész & Csilla Rákosi

g. One might say that on Grice’s account it would seem that any sentence can
be uttered with any meaning whatever, given that the circumstances make
possible the appropriate intentions. […]
h. Grice’s account can be amended to deal with counter-examples of this kind.
We have here a case where I intend to produce a certain effect by means of
getting the hearer’s recognition of my intention to produce that effect, but
the device I use to produce this effect is one which is conventionally, by the
rules governing the use of that device, used as a means of producing quite
different illocutionary effects, and the stage setting or conditions which
would permit us to say one thing and mean something totally unrelated are
not present.
i. We must […] reformulate the Gricean account of meaning in such a way as
to make it clear that one’s meaning something when one utters a ­sentence
is more than just randomly related to what the sentence means in the
language one is speaking. In our analysis of illocutionary acts, we must
capture both the intentional and the conventional aspects and especially the
relationship between them.”
(Searle 1969: 44f.; italics as in the original, bold emphasis added)
(4)(a) delineates an imaginary situation. In (4)(b)–(c), plausible inferences are
­presented which provide an interpretation of some facets of the given situation
with the help of the terminology of speech act theory. These inferences rely on
the following data introduced into the p-context through the imaginary situation
in (4)(a):
(5) a. 0 < |In the situation depicted in (4)(a), John Searle says, Kennst du das
Land… etc. because he wishes to get Italian soldiers to believe that he is a
German soldier.|S < 1
b. 0 < |In the situation depicted in (4)(a), Italian soldiers do not know what
the utterance Kennst du das Land… etc. means.|S < 1
c. 0 < |In the situation depicted in (4)(a), Italian soldiers can realise that the
utterance Kennst du das Land… etc. is a German sentence.|S < 1
The direct source of the plausibility of the statements in (5) is a compound of sub-
sources, abbreviated as S, convering, among other things, Searle’s everyday (linguis-
tic) intuition. Nevertheless, we have no direct access to it but we can reconstruct it
from the text of Searle (1969) by making use of our own (linguistic) intuition. The
statements (5)(a)–(c) do not contain theoretical terms but everyday concepts.
Based on these data, in (4)(b) the characterisation of Searle’s behaviour as
described in (4)(a) is reformulated, that is, retrospectively re-evaluated in accor-
dance with (2) in Gricean terms. Thereby, ‘wish to get X to believe that Y’ as used
in (5)(a) is explicated as ‘intending that X thinks that what was told him is Y’ in
the sense of (2):
 Thought experiments and real experiments as converging data sources in pragmatics 

(6) 0 < |In the situation depicted in (4)(a), John Searle says, Kennst du das
Land… etc. because he wishes to get Italian soldiers to believe that he is a
German soldier.|S < 1 (=(5)(a))
[0 < |If in the situation depicted in (4)(a), John Searle says, Kennst du das
Land… etc. because he wishes to get Italian soldiers to believe that he is a
German soldier, then he intends that the German soldiers should think that
what he is trying to tell them is ‘I am a German soldier’ when saying Kennst
du das Land… etc.|S < 1]
0 < |In the situation depicted in (4)(a), Searle intends that the German
soldiers should think that what he is trying to tell them is ‘I am a German
soldier’ when saying Kennst du das Land… etc.|I(6) < 19

Since the conclusion of (6) will play an important role in Searle’s argumentation,
we will represent it as hypothesis (7):
(7) 0 < |In the situation depicted in (4)(a), Searle intends that the German
soldiers should think that what he is trying to tell them is ‘I am a German
soldier’ when saying Kennst du das Land… etc.|I(6) < 1

The question in the last sentence of (4)(b) rests on the assumption that the appli-
cation of (2) to the situation described in (4)(a) would lead to the conclusion that
the meaning of Searle’s words has to be identified with the content of his intention:
(8) 0 < |In the situation depicted in (4)(a), Searle intends that the German
soldiers should think that what he is trying to tell them is ‘I am a German
soldier’ when saying Kennst du das Land… etc.|I(6) < 1 (= (7))
0 < |The meaning of an utterance has to be identified with the speaker’s
intention related to this utterance.|G < 1 (= (2))10
0 < |In the situation depicted in (4)(a), Searle means with his words ‘I am a
German soldier’ when saying Kennst du das Land… etc.|I(8) < 1

We will refer to the conclusion of inference (8) as hypothesis (9):

(9) 0 < |In the situation depicted in (4)(a), Searle means with his words ‘I am a
German soldier’ when saying Kennst du das Land… etc.|I(8) < 1
This means that the datum in (5)(a) provides weak evidence for (9): the plausible
inferences (6) and (8) establish a connection between (5)(a) and this hypothesis.

.  In the conclusion of (6), I(6) stands for the inference in (6) as an indirect source of the
p| indicates the plausibility value of statement p; see Kertész & Rákosi (this
conclusion. | _ 
volume b: Section 2.2, Example 3; Section 2.3).
.  G indicates Grice (1989a [1957]) as a direct source.
 András Kertész & Csilla Rákosi

Thus, the analysis of the imagined situation in (4)(a) based on Grice’s theory of
meaning, and especially, on hypothesis (2) leads to the conclusion that the mean-
ing of Searle’s words in the situation in (4)(a) has to be identified with the speaker’s
intention, that is, with the information that he wants to communicate. The current
state of the p-context is as in Figure 1.11

(5)(a) (6) (5)(b) (5)(c)

(7) (8) (2)

(9)

Figure 1.  The state of the p-context after its extension with (4)(a–b)

(9) is, however, questioned immediately in the first sentence of (4)(c). The starting
point of Searle’s argumentation against (2) is the following datum:
(10) 0 < |According to the rules of German as well as what Searle remembers,
the words Kennst du das Land… etc. mean that ‘Knowest thou the land
where the lemon trees bloom?’.|S < 1

The extension of the p-context with (10) makes it possible to draw the following
plausible inference:
(11) 0 < |According to the rules of German as well as what Searle remembers,
the words Kennst du das Land… etc. mean ‘Knowest thou the land where
the lemon trees bloom?’.|S < 1 (= (10))
[0 < |There is no difference between the meaning of these words according
to the rules and conventions of the German language as Searle remembers
and what Searle means with them in the given situation.|S < 1]
0 < |In the situation depicted in (4)(a), Searle means with his words
‘Knowest thou the land where the lemon trees bloom?’.|I(11) < 1

In (11), the latent background assumption identifies what Searle means by his
utterance in this situation with the conventional meaning of his words. Thus, the
argumentation process has led to the following hypothesis:

.  In Figures 1–10, hypotheses are in italics, and inferences are in normal script. C
­ onclusions
of inferences are connected with their premises and the given inference by simple arrows.
Double dotted arrows indicate p-inconsistency.
 Thought experiments and real experiments as converging data sources in pragmatics 

(12) 0 < |In the situation depicted in (4)(a), Searle means with his words
‘­Knowest thou the land where the lemon trees bloom?’.|I(11) < 1

(8) and (11) lead to contradicting conclusions – clearly due to the circumstance
that they use ‘mean’ in two different senses: in (8) in accordance with (2), that is,
by identifying meaning with the speaker’s intention, and in (11) by identifying it
with the conventional meaning of the uttered words. Figure 2 shows the current
state of the p-context.

(5)(a) (6) (10) (5)(b) (5)(c)

(7) (8) (2) (11)

(9) (12)

Figure 2.  The state of the p-context after its extension with the first part of (4)(c)

Searle deems (12) more plausible than (9) and solves this contradiction in
such a way that he gives up (9) and retains (12). Thus, the following indirect source
presents itself:
(13) 0 < |In the situation depicted in (4)(a), Searle means with his words
‘­Knowest thou the land where the lemon trees bloom?’.|I(11) < 1 (= (12))
0 < |In the situation depicted in (4)(a), Searle intends that the German
soldiers should think that what he is trying to tell them is ‘I am a German
soldier’ when saying Kennst du das Land… etc.|I(6) < 1 (= (7))
[0 < |If in the situation depicted in (4)(a), Searle intends that the German
soldiers should think that what he is trying to tell them is ‘I am a G­ erman
soldier’ when saying Kennst du das Land… etc. but he means with his words
‘Knowest thou the land where the lemon trees bloom?’, then the meaning of
this utterance cannot to be identified with the speaker’s intention related to
this utterance.|S < 1]
0 < |In the situation depicted in (4)(a), the meaning of Searle’ utterance
cannot be identified with the speaker’s intention related to this
utterance.|I(13) < 1

Thus, the interpretation of the situation in (4)(a) comes to the conclusion that
(14) 0 < |In the situation depicted in (4)(a), the meaning of Searle’s ­utterance cannot
be identified with the speaker’s intention related to this utterance.|I(13) < 1
 András Kertész & Csilla Rákosi

In the last sentence of (4)(c), Searle provides a possible analysis of the mechanism
of the deceit applied in the story. There he seems to exploit the given situation with
the help of the data in (5)(b) and (c):

(15) 0 < |In the situation depicted in (4)(a), the meaning of Searle’s ­utterance
cannot be identified with the speaker’s intention related to this
utterance.|I(13) < 1 (=(14))
[0 < |If in the situation depicted in (4)(a), the meaning of Searle’s utterance
cannot be identified with the speaker’s intention related to this utterance,
Italian soldiers do not know what the utterance Kennst du das Land… etc.
means, and they can realise that the utterance Kennst du das Land… etc.
is a German sentence, then they will erroneously think that the sentence
Kennst du das Land… etc. means ‘I am a German soldier’ according to the
­conventions of the German language.|S < 1]
0 < |If in the situation depicted in (4)(a), Italian soldiers erroneously think
that the sentence Kennst du das Land… etc. means ‘I am a German soldier’
according to the conventions of German, then they should erroneously
think that with the utterance of the sentence Kennst du das Land… etc.
Searle means ‘I am a German soldier’.|S < 1
0 < |In the situation depicted in (4)(a), Italian soldiers do not know what
the utterance Kennst du das Land… etc. means.|S < 1 (= (5)(b))
0 < |In the situation depicted in (4)(a), Italian soldiers can realise that
the ­utterance Kennst du das Land… etc. is a German sentence.|S < 1
(= (5)(c))
0 < |In the situation depicted in (4)(a), Italian soldiers may erroneously
think that with the utterance of the sentence Kennst du das Land… etc.
Searle means ‘I am a German soldier’.|I(15) < 1

As a result, a hypothesis was obtained that would be p-inconsistent with (9) that
has already been rejected, but is p-consistent with (12):

(16) 0 < |In the situation depicted in (4)(a), Italian soldiers may erroneously
think that with the utterance of the sentence Kennst du das Land… etc.
Searle means ‘I am a German soldier’.|I(15) < 1

After its coordination, the p-context is as in Figure 3.

As Figure 3 illustrates, the Exclusive Strategy applied by Searle in order to
examine the p-context which can be built around hypothesis (2) failed. The rea-
son is that the plausible inferences drawn on the basis of the data available made
hypotheses plausible which are incompatible with (2).
 Thought experiments and real experiments as converging data sources in pragmatics 

(5)(a) (6) (10) (5)(b) (5)(c)

(7) (11) (15)

(12) (13)

(2) (14)

(16)

Figure 3.  The state of the p-context after its extension with the second part of (4)(c)

3.2.1  Subcycle 1: Analysis of Wittgenstein’s thought experiment

Although Searle gave up (9), he does not modify the p-context by also deleting
(2) from the p-context. The reason for this decision might be that at this point of
his argumentation, he is not in the possession of an alternative hypothesis to (2).
Therefore, he starts a new subcycle in order to formulate and test a rival hypoth-
esis and to find further data for a well-founded solution to the starting problem.
­Nevertheless, it is important to remark that Searle’s argumentation is enthyme-
matic here. Of course, this makes the reconstruction of his argumentation more
difficult and disputable.12
In (4)(d), one of Wittgenstein’s thought experiments is referred to. The
­p-context is extended by a new source which is a quotation from §510 of
­Wittgenstein’s Philosophical Investigations.13 The starting point of Searle’s argu-
mentation is the following datum:
(17) 0 < |Without further stage setting, when one says, It’s cold here, one cannot
mean ‘It is warm here’.|S < 1

.  According to Walton et al. (2008: 189), “[o]ne problem with enthymemes is that reason-
able people can have differences of opinion on what the implicit assumptions are supposed
to be. Filling in the missing parts of an enthymeme may depend on interpreting the natural
language text in which the argument was put forward […].”
.  The whole passage is this: “510. Make the following experiment: say ‘It’s cold here’ and
mean ‘It’s warm here’. Can you do it? – And what are you doing as you do it? And is there only
one way of doing it?” (Wittgenstein 1958 [1953]: 140; italics as in the original, bold emphasis
added).
 András Kertész & Csilla Rákosi

At this point, Searle does not clarify what “further stage setting” means; only
later does the reader receive explanation, in (4)(h). Here again, the reliability of
this thought experiment depends crucially on the question of whether the reader
accepts this datum by judging it with the help of his/her own (linguistic) intuition.
Based on (17), the following plausible inference presents itself:
(18) 0 < |Without further stage setting, when one says, It’s cold here, one cannot
mean ‘It is warm here’.|S < 1 (= (17))
0 < |If without further stage setting, when one says, It’s cold here, one cannot
mean ‘It is warm here’, then there are cases where, what one can mean is a
function of what one is saying.|S < 1
0 < |There are cases in which, what one can mean is a function of what one
is saying.|I(18) < 1

(18) is based on a possible explanation of the datum in (17). A decisive element

in the evaluation of Searle’s argumentation is whether one is ready to accept the
second premise, and via this, the conclusion of (18), which, for future reference,
we highlight as (19):
(19) 0 < |There are cases in which, what one can mean is a function of what one
is saying.|I(18) < 1

(19) is clearly p-consistent with (14), which states that the meaning of Searle’s
utterance in the situation depicted in (4)(a) cannot be identified with the speaker’s
intention related to this utterance.
(4)(f) continues the analysis of the Wittgensteinean situation and puts for-
ward a plausible inference based on (19) in order to elucidate and generalise it.
Searle’s argument can be reconstructed in the following way:
(20) 0 < |There are cases in which what one can mean is a function of what one is
saying.|I(18) < 1 (= (19))
[0 < |If there are cases in which what one can mean is a function of what
one is saying, then meaning is more than a matter of intention, it is also at
least sometimes a matter of convention.|S <1]
0 < |Meaning is more than a matter of intention, it is also at least sometimes
a matter of convention.|I(20) < 1

The analysis of the state of affairs referred to in Wittgenstein’s thought experiment

leads to the hypothesis that the meaning of an utterance cannot be identified with
the speaker’s intention in these cases but involves the conventional meaning of
the sentence at issue as well. Thus, a highly important result of Searle’s thought
experiment provided in the conclusion of inference (20) is a hypothesis which is
a rival to (2):
 Thought experiments and real experiments as converging data sources in pragmatics 

(21) 0 < |Meaning is more than a matter of intention, it is also at least sometimes
a matter of convention.|I(20) < 1

At this point, Searle’s argumentation turns back to the point where it had been
suspended before the subcycle was started, and the new information obtained is
added to the p-context. The state of the p-context as depicted in Figure 3 has been
retrospectively re-evaluated. Data originating from Wittgenstein’s thought experi-
ment contributed to the increase of the plausibility value of ~(2), since besides (14)
and (16), (21) is also p-inconsistent with (2). Thereby, (21) is a reformulated and
re-evaluated version of ~(2). Since (19) covers (14) as a special case, it would be
possible to construct a plausible inference with (14) as one of its premises and (21)
as its conclusion. After its coordination, the p-context is as in Figure 4.

(5)(a) (6) (10) (5)(b) (5)(c)

(7) (11) (15)

(12) (13)

(2) (14) (16)

(21)

(19) (20)

(17) (18)

Figure 4.  The state of the p-context after its extension with (4)(d–f)

3.2.2  Th
 e continuation of the argumentation cycle: Further elaboration
of a rival hypothesis
After the coordination of the p-context, Searle turns to the application of the Con-
trastive Strategy in order to arrive at a decision between (2) and its rival, (21). In
(4)(g), he generalises the conclusion which emerges from the application of the
Gricean hypothesis in (2) to the imaginary situations delineated in (4)(a) and (4)(d):
(22) 0 < |If the meaning of an utterance is identified with the speaker’s intention
related to this utterance, then any sentence can be uttered with any meaning
whatever, given that the circumstances make possible the appropriate
intentions.|S < 1
 András Kertész & Csilla Rákosi

0 < |In the situation depicted in (4)(a), the meaning of Searle’s ­utterance
cannot be identified with the speaker’s intention related to this
utterance.|I(13) < 1 (= (14))
0 < |In the situation depicted in (4)(d), without further stage setting, when
one says, It’s cold here, one cannot mean that ‘It is warm here’.|S < 1 (= (17))
0 < |The meaning of an utterance cannot be identified with the speaker’s
intention related to this utterance.|I(22) < 1 (= ~(2))

That is, the plausible inference presented in (22) makes (2) implausible, because it
shows that (2) overgeneralises. The p-context is as in Figure 5.

p-context version of (2) p-context version of (21)

central hypothesis: (2) central hypothesis: (21)

(7), (9) (7), (12), (14), (16), (19)

Grice’s theory of meaning Searle’s theory of meaning

(5)(a), (b), (c)

(10), (17)

Figure 5.  The p-context resulting from the application of (2) to the imaginary situations
­delineated in (4)(a) and (4)(d)

While the data stemming from the two thought experiments provide weak
evidence for (21), they also provide weak evidence against (2) in the sense of Sub-
section 3.2 in Kertész & Rákosi (this volume b). Therefore, since the p-context
version of (2) is burdened with a high amount of p-inconsistency, Searle gives it
up and preserves only its rival and elaborates further on it. He turns back to the
Exclusive Strategy in order to find out whether the p-context version of (21) pro-
vides a satisfactory solution of the starting p-problem.
(17), (19), and (21) do not exclude the possibility that sometimes a situation can
be established in which what one can mean is independent of what one is saying. As
we have remarked with respect to (17) in Subsection 3.2.1, Searle does not provide
any explanation of what “further case setting” means. Thus, it remains unclarified
how cases in which utterance meaning can be identical with a completely different
speaker’s intention can be distinguished from cases in which this is not possible. (4)
(h) deals with this question and identifies the factors which are supposed to consti-
tute the difference between the two types of cases. Since it is not clear from Searle’s
formulation which of the two thought experiments – i.e. Searle’s or Wittgenstein’s – is
referred to in this passage, our reconstruction will pertain to both of them:
 Thought experiments and real experiments as converging data sources in pragmatics 

(23) 0 < |In the situation depicted in (4)(a), Searle intends that the hearer should
think that what he is trying to tell him is ‘I am a German soldier’ when
saying, Kennst du das Land… etc.|(I)(6) < 1 (= (7))
0 < |According to the rules of German as well as what Searle remembers,
the words Kennst du das Land… etc. mean ‘Knowest thou the land where
the lemon trees bloom?’.|S < 1 (= (10))
[0 < |If according to the rules of German, the words Kennst du das Land…
etc. mean ‘Knowest thou the land where the lemon trees bloom?’, then these
words are a device which is conventionally used as a means of producing
quite different illocutionary effects than ‘I am a German soldier’.|S < 1]
0 < |In the situation depicted in (4)(a), Searle intends to produce a certain
effect by means of getting the hearer’s recognition of his intention to
produce that effect, but the device he uses to produce this effect is one
which is conventionally used as a means of producing quite different
illocutionary effects.|I(23) < 1
(24) 0 < |In the situation depicted in (4)(d), one intends that the hearer should
think that what one is trying to tell him is ‘It is warm here’ when saying,
It’s cold here.|S < 1
[0 < |According to the rules of English, the words It’s cold here mean ‘It’s
cold here’.|S < 1]
[0 < |If according to the rules of English, the words It’s cold here mean ‘It’s cold
here’, then these words are a device which is conventionally used as a means of
producing quite different illocutionary effects than ‘It is warm here’.|S < 1]
0 < |In the situation depicted in (4)(d), one intends to produce a certain
effect by means of getting the hearer’s recognition of his intention to
produce that effect, but the device he uses to produce this effect is one
which is conventionally used as a means of producing quite different
illocutionary effects.|I(24) < 1
As a next step, in (4)(h), relying, among other things, on the conclusion of (23) and
(24), Searle generalises the explanation which could be applied to these two cases
with the help of the following inductive inference:
(25) 0 < |In the situation depicted in (4)(a), Searle intends to produce a certain
effect by means of getting the hearer’s recognition of his intention to
produce that effect, but the device he uses to produce this effect is one
which is conventionally used as a means of producing quite different
illocutionary effects.|I(23) < 1
0 < |In the situation depicted in (4)(d), one intends to produce a certain
effect by means of getting the hearer’s recognition of his intention to
produce that effect, but the device he uses to produce this effect is one
which is conventionally used as a means of producing quite different
illocutionary effects.|I(24) < 1
 András Kertész & Csilla Rákosi

0 < |In the situation depicted in (4)(a) and (4)(d), the stage setting or
conditions which would permit one to say one thing and mean something
totally unrelated are not present.|S < 1
0 < |In the situation depicted in (4)(a) and (4)(d), the sentences mentioned
cannot be uttered with a meaning identical to the speaker’s
intention.|(14) & (17) < 1
[0 < |In every other case, too, if the circumstances do not make possible the
appropriate intentions or (i) the sentence cannot be conventionally, by the
rules governing the use of this sentence, used as a means of producing this
illocutionary effect, and (ii) the stage setting or conditions which would
permit one to say one thing and mean something totally unrelated are not
present, then this sentence cannot be uttered with a meaning identical to the
speaker’s intention.|S < 1]
0 < |A sentence can be uttered with a meaning only if the circumstances
make possible the appropriate intentions and (i) the sentence can be
conventionally, by the rules governing the use of this sentence, used as
a means of producing this illocutionary effect, or (ii) the stage setting or
conditions which would permit one to say one thing and mean something
totally unrelated are present.|I(25) < 1

From this it seems that (4)(a) and (4)(d) refer to one’s linguistic behaviour under
normal circumstances, that is, to a situation in which the speaker and the hearer
know the language used well enough so that he/she cannot be mistaken or deceived
about the conventional meaning of the words used, and both the speaker and the
hearer are supposed to use language in a normal way, without any special modifi-
cation of the meaning of words.
The conclusion of (25) is a retrospectively re-evaluated version of (21), that is,
a further refinement of ~(2):

(26) A sentence can be uttered with a meaning if the circumstances

make ­possible the appropriate intentions and (i) the sentence can be
conventionally, by the rules governing the use of this sentence, used as
a means of ­producing this illocutionary effect, or (ii) the stage setting or
conditions which would permit one to say one thing and mean something
totally ­unrelated are present.

3.3  The comparison of the solutions to the problem raised in (1)

On the basis of the above considerations we can state that a better solution can
be built around (26) rather around (2), since the latter does not receive any sup-
port from other direct or indirect sources in this p-context and its presence in the
p-context leads to p-inconsistencies.
 Thought experiments and real experiments as converging data sources in pragmatics 

3.4  The final p-context

Thus, the p-context is modified: (2) is given up and replaced by a rival hypothesis
proposed by Searle, because only the latter is p-consistent with the data obtained in
the thought experiment. Therefore, the final p-context does not contain (2) because
it has been rejected and replaced by (21) and further elaborated as (26). The recon-
struction of the inference in (4)(i) which presents the methodological consequences
emerging from the two thought experiments analysed by Searle is as follows:
(27) 0 < |Meaning is more than a matter of intention, it is also at least sometimes
a matter of convention.|I(20) < 1 (= (21))
0 < |If meaning is more than a matter of intention, and it is also at least
sometimes a matter of convention, then the analysis of illocutionary acts
must capture both intentional and conventional aspects of meaning and the
relationship between them.|S < 1
0 < |The analysis of illocutionary acts must capture both intentional and
conventional aspects of meaning and the relationship between them.|I(27) < 1

That is, the results of the two thought experiments motivate the following meth-
odological principle:
(28) The analysis of illocutionary acts must capture both intentional and
conventional aspects of meaning and the relationship between them.

Figure 6 shows the final p-context.

(5)(a) (6) (10) (5)(b) (5)(c)

(7) (11) (15)

(12) (13)

(14) (16)

(28)

(21) (26) (27)

(19) (20)
(23) (24) (25)

(17) (18)

Figure 6.  The final p-context

 András Kertész & Csilla Rákosi

This solution to (1) is provisional and does not count as a resolution in the
sense of Subsection 2.5 in Kertész & Rákosi (this volume b). It is exposed to fur-
ther revision for at least three reasons.
First, the argumentation process continues with Searle’s (1969: 43) consider-
ation of the problem of “what is it for something to have a meaning in analogy
to the process we have just exemplified”. This problem is discussed on pp. 49f. by
contrasting Grice’s solution with Searle’s. The final p-context of this continuation
includes the retrospective re-evaluation of (21)/(26) in the light of Searle’s solu-
tion to the above problem. We will not reconstruct this process here, because it is
analogous to the one we have just analysed.
Second, as is well known, Searle’s account of meaning supported by the
thought experiment in (4) turned out to be problematic in several respects again.
Grice (1989b), Meggle & Ulkan (1979) and other contributions argued against
(21)/(26); i.e. these authors re-evaluated the results of the plausible argumentation
which we have reconstructed above.
Third, the counter-arguments against (21)/(26) motivated, among others, real
experiments. In the next section we will analyse an example which illustrates how
(21)/(26) can be retrospectively re-evaluated by real experiments.
We choose Holtgraves & Ashley (2001) as the subject matter of the case study
that scrutinises a problem with the help of a real experiment which is directly con-
nected to the thought experiment presented by Searle and analysed in the previous
section. For lack of space, we will dispense with the reconstruction of Holtgraves
and Ashley’s (2001) whole argumentation. We will focus on those aspects that are
instructive with respect to our problem (P).

4.  Case study 2: Real experimental reports in Holtgraves & Ashley (2001)

4.1  Setting up the starting p-context

In Holtgraves & Ashley (2001), the starting p-context includes Searle (1969), Aus-
tin (1962) and Searle & Vanderveken (1985) as direct sources. It also includes
the result of Searle’s thought experiment as summarised in (21)/(26) which the
authors reformulate in the following way:
(29) “[…] utterances have both a propositional content and an illocutionary
force (the speech act performed with the utterance).”
(Holtgraves & Ashley 2001: 83)

The starting p-context is problematic in the sense of Subsection 2.5 in Kertész &
Rákosi (this volume b):
 Thought experiments and real experiments as converging data sources in pragmatics 

(30) a. “[…] it is not clear whether comprehension even involves recognition

of the action that is performed with an utterance. That is, when we
­comprehend a speaker’s remark, does that comprehension involve
recognition, at some level, of the action that the speaker is undertaking?
Does comprehension of ‘Don’t forget to stop at the store’ involve
recognition that the speaker is performing the act of reminding?
b Surprisingly, with the exception of research on recognition of indirect
speech acts (e.g. Gibbs 1983), very little empirical research has
­addressed this issue.
c. The purpose of this research was to undertake some initial analyses of
the extent to which comprehension involves recognition of the action – or
speech act – that is performed with an utterance.”
(Holtgraves & ­Ashley 2001: 83; emphasis added)

As (30)(a) indicates, the p-context is informationally underdetermined. Searle’s

thought experiment (and speech act theory in general) simply presupposes in
(4)(b) that speaker-intentions are recognised by the hearer without providing evi-
dence for this assumption. Holtgraves and Ashley are not sufficiently content with
the low plausibility value of this hypothesis to allow it to exist as a pure assump-
tion. Moreover, Holtgraves and Ashley (2001: 84) refer to rival approaches, namely,
“speech act theory for which illocutionary force recognition is not required” (see,
e.g. Cohen & Levesque 1990). Thus, they raise the following problem:
(31) “[…] is it the case that addressees actually recognise the performance of a
specific speech act?” (Holtgraves & Ashley 2001: 84)

Accordingly, the argumentation process should find out whether it is possible to

assign a plausibility value to the following hypothesis:
(32) Comprehension involves recognition of the action that is performed with
an utterance,

or to assign a plausibility value to its rival, namely, its negation, – i.e. whether (32)
is plausible or implausible.

4.2  The elaboration of the starting p-context

The authors extend the p-context by new sources which are based on what they call
“empirical investigations”. Although they emphasise that the problem in (31) has
not yet been addressed by ‘empirical’ investigations, they refer to a series of similar
problems which have been tackled with the help of psycholinguistic experiments.
Thereby, they accept the following latent background assumption:
(33) [0 < |Real experiments are highly reliable data sources.|HA < 1]
 András Kertész & Csilla Rákosi

(33) exemplifies that the p-context may include methodological assumptions

which significantly influence the plausibility of the experimental results. Since,
however, the experiments referred to by Holtgraves and Ashley (2001: 84) bear
only an indirect relationship to illocutionary force activation, they provide only
weak evidence for (32). Their results are either coherent with (32) or they sup-
port (32) on the basis of some analogy between illocutionary force activation and,
for example, drawing spontaneous trait inferences during the comprehension of
a person’s behaviour. The analogy is established on the assumption that both are
based on some kind of categorisation. Clearly, these pieces of evidence do not
allow the authors to prefer (32) to its negation and make a decision between them.
Accordingly, the task they set themselves is “to investigate whether or not illo-
cutionary force is activated during comprehension” (ibid.) by carrying out real
experiments. Thus, the authors carry out four real experiments.

4.2.1  Experiment 1
Regarding the experimental design,14 participants were presented scenarios fol-
lowed by a dialogue. For example:
(34) “Jenny and Emily had been close friends since grade school. Now they were
rooming together at college.
Emily was very forgetful.
Today, Jenny was sure Emily didn’t remember (had forgotten) her dentist
appointment.
Don’t forget (I’ll bet you forgot) to go to your dentist appointment
Jenny: 
today.15
Probe: remind.” (Holtgraves & Ashley 2001: 85)

The task of the participants was to decide whether the probe word – in (34) the
speech act verb remind – had been literally mentioned. Holtgraves and Ashley put
forward the following conjectures about the possible outcome of the experimental
procedure:
(35) a. “If comprehension of a remark involves activation of its illocutionary
force, then, when the probe represents the speech act just performed,
participants’ ability to verify that the probe had not been literally
present should be slowed; activation of the speech act term should

.  We do not reconstruct the experimental designs completely in either case. For further
details of these, see the relevant passages in Holtgraves & Ashley (2001).
.  “The speech act version contained the italicized material; the control version was created
by replacing the italicized material with the material in parentheses” (op. cit.).
 Thought experiments and real experiments as converging data sources in pragmatics 

interfere with performance on this task. For example, participants

should be slower at verifying that the word remind had not literally
been present in the remark Don’t forget to go to your dentist
appointment today relative to a control remark I’ll bet you forgot to go to
your dentist appointment today.
b. On the basis of speech act theory and the empirical research cited
above, we expected reaction times to be slowed, and errors to
be increased, when the probe word characterized the speech act
performed with the final remark than when it did not.”
(Holtgraves & Ashley 2001: 85)

The participants of the experimental procedure were 55 students of psychol-

ogy, all native speakers of English. They received a set of scenarios consisting
of six speech act scenarios and six control scenarios. In the case of the speech
act versions, the final sentence performed the particular speech act referred
to by the probe word but this was not the case with the control scenarios. The
correspondence between the intended speech act and the probe word was con-
trolled with the help of a pretest. In the pretest, 75% of participants named the
expected term correctly. The probe word never appeared in the text of the final
remark of the scenarios. Nevertheless, in order to prevent such expectations,
12 filler scenarios were used which contained the probe word. Several details
of the experimental situation such as the software used, the instructions, the
presence of practice trials, the timing of the different stages of the completion
of the required tasks etc. were controlled for, and described in the experimental
report.
The perceptual data obtained had the structure “The time between the
appearance of the probe word W on the screen and the keystroke by person No.
P in item No. I was x ms.” Then, the perceptual data underwent authentication
and interpretation. For example, perceptual data related to an incorrect decision
about the presence/absence of the probe word were excluded from the analyses.
Statistical analyses were conducted with both participants and stimuli as ran-
dom variables. As a result, the following experimental data were obtained and
made public:

(36) a. 0 < |Mean probe reaction time (and error rate) in speech act scenarios
was 1,034 ms (6.36%) in Experiment 1.|Exp1 < 1
b. 0 < |Mean probe reaction time (and error rate) in control scenarios was
960 ms (1.52 %) in Experiment 1.|Exp1 < 1
c. 0 < |“Participants were significantly slower at verifying that the probe
had not been literally present when it represented the speech act
performed with the final remark than when it did not.”|Exp1 < 1
 András Kertész & Csilla Rákosi

It is very important to realise that these data are not true with certainty. The experi-
mental procedure, the interpretation and the authentication of the perceptual data
may be faulty and be in need of modification. They cannot be totally controlled;
it is always possible that relevant factors which influence the outcome of the
­experimental process to some extent remain unrevealed. From this it follows that
experimental data can be only plausible.
As a next step, Holtgraves and Ashley confronted these data with one of the
rival hypotheses, namely, with (32). With this aim in view, first they referred to the
conjecture formulated in (35):
(37) 0 < |“If comprehension of a remark involves activation of its illocutionary
force, then, when the probe represents the speech act just performed,
participants’ ability to verify that the probe had not been literally present
should be slowed.” |HA < 1 (= (35))
0 < |When the probe represented the speech act just performed,
participants’ ability to verify that the probe had not been literally present
was slowed.|Exp1 < 1 (= (36))
0 < |Comprehension of a remark involves activation of its illocutionary
force.|I(37) < 1

Although this inference is not deductively valid, it is capable of making its conclu-
sion plausible, since it can be shown that it belongs to the third group of plausible
inferences (see Kertész & Rákosi, this volume b: Section 2.3):
(38) 0 < |Comprehension of a remark involves activation of its illocutionary
force.|I(37) < 1

As (39) indicates, (38) supports (32):

(39) 0 < |If comprehension of a remark involves activation of its illocutionary
force, then comprehension involves recognition of the action that is
performed with an utterance. |HA < 1
0 < |Comprehension of a remark involves activation of its illocutionary
force. |I(37) < 1 (= (38))
0 < |Comprehension involves recognition of the action that is performed
with an utterance. |I(39) < 1 (= (32))

As for the coordination and modification of the p-context, experimental data origi-
nating from Experiment 1 as a data source provide support for (32). This means
that the hypothesis that comprehension involves recognition of the action that
is performed with an utterance has become plausible. The current state of the
­p-context is as in Figure 7.
 Thought experiments and real experiments as converging data sources in pragmatics 

(35) (36) (37)

(38) (39)

(32)

Figure 7.  The state of the p-context after conducting Experiment 1

Nevertheless, the authentication of perceptual data raised some concerns

about the reliability of the results. By analysing the data set, the authors revealed a
possible systematic error which may have distorted the results:

(40) “A potential problem with this experiment is that participants took
longer to read the final remark when it performed the relevant speech
act […] than when it did not […]. Hence, slowed judgement speeds
following the relevant speech act utterances could have been due to the
greater comprehension difficulty for these remarks, relative to the control
remarks.”  (Holtgraves & Ashley 2001: 86)

In (37), the authors supposed that slowed reaction times indicate the presence of
illocutionary force activation. After extending the p-context by (40), however, (37)
had to be transformed into another inference which takes into consideration the
presence of a second possible explanation of the experimental data as well. Thus,
the following inference presents itself:

(41) 0 < |If the probe represents the speech act just performed, and participants’
ability to verify that the probe had not been literally present slows, while
there is no difference in the comprehension difficulty for the final sentences
of the speech act scenarios and the control scenarios, respectively, then
comprehension of a remark involves activation of its illocutionary
force.|HA < 1 (= (40))
0 < |When the probe represented the speech act just performed,
participants’ ability to verify that the probe had not been literally present
was slowed.|Exp1 < 1 (= (36))
[|There was no difference in the comprehension difficulty for the
final sentences of the speech act scenarios and the control scenarios,
respectively.| = ?]
|Comprehension of a remark involves activation of its illocutionary
force.|I(41) = ?
 András Kertész & Csilla Rákosi

The latent background assumption of (41), which was put between ‘[’ and ‘]’
above, has been retrospectively and prismatically re-evaluated in the sense of Sub-
section 2.6.1 of Kertész & Rákosi (this volume b). It had neutral plausibility during
Experiment 1. Subsequent analysis of the experimental data, as (40) shows, how-
ever, questioned the assumption and showed that it was implausible. As a conse-
quence, the premises of (41) are no longer capable of supporting the conclusion
in the given state of the p-context, even though their plausibility did not change.
Therefore, (38) loses its earlier plausibility value. This does not mean, of course,
that the conclusion became implausible.
In order to find out whether the data originating from Experiment 1 can
be used to support (32), a control experiment was proposed and a new subcycle
initiated.

4.2.2  Subcycle 1: Experiment 2

Subcycle 1 of the argumentation process corresponds to Experiment 2. It repli-
cates the first experiment with two modifications in the experimental design. First,
the text of the final remarks of the control scenarios was lengthened. As a con-
sequence, the latent background assumption of (41) was no longer implausible.
Second, participants had a further task: from 24 words they had to choose those
which described the remarks they read during the experiment. 6 words referred to
speech acts which were performed with them in one of the scenarios, 6 words were
irrelevant speech act terms, while 12 words were fillers.
There were two sets of perceptual data. The first set consisted of data with a
similar structure as had been the case with Experiment 1: “The time between the
appearance of the probe word W on the screen and the keystroke by person No. P
in item No. I was x ms.” The second set of perceptual data had the structure “Per-
son No. P hit the button Y/N in item No. I.” The following experimental data were
obtained after the interpretation and authentication of the perceptual data:
(42) a. 0 < |Mean probe reaction time (and error rate) in speech act scenarios
was 1,099 ms (9.42%) in Experiment 2.|Exp2 < 1
b. 0 < |Mean probe reaction time (and error rate) in control scenarios was
1,047 ms (5.44 %) in Experiment 2.|Exp2 < 1
c. 0 < |“First, the participants made more errors when the probe
represented a performed speech act than when it did not […]. And
second, the participants were again significantly slower at making this
judgement when the probe represented a prior speech act than when it
did not […].”|Exp2 < 1
(42) provides support for (38) in a similar manner as (36) did:
 Thought experiments and real experiments as converging data sources in pragmatics 

(43) 0 < |If the probe represents the speech act just performed, and participants’
ability to verify that the probe had not been literally present slows, while
there is no difference in the comprehension difficulty for the final sentences
of the speech act scenarios and the control scenarios, respectively, further
participants make more errors when the probe represented a performed
speech act than when it did not, then comprehension of a remark involves
activation of its illocutionary force.|HA < 1
0 <|When the probe represented the speech act just performed, participants’
ability to verify that the probe had not been literally present was
slowed.|Exp2 < 1 (= (42))
0 < |The participants made more errors when the probe represented a
performed speech act than when it did not.|Exp2 < 1 (= (42))
0 < |There was no difference in the comprehension difficulty for the
final sentences of the speech act scenarios and the control scenarios,
respectively.|Exp2 < 1
0 < |Comprehension involves recognition of the action that is performed
with an utterance. |I(43) < 1 (= (32))

Moreover, (42) indicates that the concern in (40) is unfounded. Thus, (41) can be
rejected and (37) can be added to the p-context again. Figure 8 shows the current
shape of the p-context.

(35) (36) (37) (42) (43)

(38) (39)

(32)

Figure 8.  The state of the p-context after conducting Experiment 2

The authentication of the perceptual data, however, indicated another pos-

sible error source. Namely, the authors found that the differences in reaction
times were significant only over participants but not over items in both experi-
ments. Therefore, a retrospectiove re-evaluation of the results was needed and a
second subcycle was initiated in order to find out whether there is a relevant, but
unrealised factor which biases the results and makes the proposed interpretation
faulty.
 András Kertész & Csilla Rákosi

4.2.3  Subcycle 2: Experiment 3

The purpose of Experiment 3 was to check whether the results of the first two
experiments were influenced by factors different from those which have been
assumed. More precisely, it might have been the case that the results which sup-
ported (32)
[…] were due to some type of context checking procedure that occurred when
the participants verified the probes. That is, reaction times might have been
slowed for the speech act probes, not because these words had been activated
during comprehension, but because when the judgments were made, the speech
act probes were more plausible than the control words.
(Holtgraves & Ashley 2001: 87)

To filter out such factors, Experiment 3 was carried out. Although the stimulus
material remained unchanged, the participants had to solve another task. Instead
of indicating whether or not the probe word literally occurred in the conversation
they were presented, they had to check whether a certain letter string was a word.
Holtgraves and Ashley put forward the following conjecture about the possible
outcome of the experimental procedure:
(44) 0 < |“If comprehension of the utterance involved speech act activation,
the participants should be significantly faster at this task when the target
represented the performed speech act than when it did not.” |HA < 1

The interpretation and authentication of the perceptual data led to the following
set of experimental data:
(45) a. 0 < |Mean lexical decision reaction time in speech act scenarios was
1,002 ms in Experiment 3.|Exp3 < 1
b. 0 < |Mean lexical decision reaction time in control scenarios was
1,095 ms in Experiment 3.|Exp3 < 1
c. 0 < |In error-free trials, the overall error rate was 9.6%, and this rate did
not vary across conditions in Experiment 3.|Exp3 < 1
d. 0 < |“[…] the participants were significantly faster at the lexicon
decision task when the target represented the prior speech act than
when it did not […]” in Experiment 3.|Exp3 < 1

It this case, the results were significant both over items and participants. Thus,
the following plausible inference indicates the confrontation of these experimental
data with the previous conjectures:
(46) 0 < |“If comprehension of the utterance involved speech act activation,
the participants should be significantly faster at this task when the target
represented the performed speech act than when it did not.” |HA < 1 (= (44))
 Thought experiments and real experiments as converging data sources in pragmatics 

0 < |“[…] the participants were significantly faster at this task when
the target represented the performed speech act than when it
did not”.|Exp3 < 1 (= (45))
0 < |Comprehension of the utterance involved speech act activation.|I(46) < 1

The conclusion of (46) provides support for (32) again. Changes in the p-context
after extending it with the results of Experiment 3 are as in Figure 9.

(35) (36) (37) (42) (43)

(38) (39) (44) (45) (46)

(32)

Figure 9.  The state of the p-context after conducting Experiment 3

Nevertheless, the reliability of Experiment 3 as a data source can be ques-

tioned according to Holtgraves & Ashley (2001: 87), since it is possible that in
Experiment 3 the reason for faster lexical decisions was not the participants’ rec-
ognising the speech act at issue, but the priming of semantic associates of words.
Thus, we are in a similar situation as in the case of Experiment 1: (46) has to be
retrospectively re-evaluated. In the light of new information, (46) is transformed
into an inference which is not capable of providing support for the hypothesis
about speech act activation. Namely, since no plausibility value can be assigned to
the latent background assumption of (47), it is not capable of providing support
to its conclusion:
(47) 0 < |If the participants are significantly faster when the target represents
the performed speech act than when it does not, and there is no semantic
priming effect, then comprehension of the utterance involves speech act
activation.|HA < 1
[|There was no semantic priming effect.| = ?]
0 < |”[…] the participants were significantly faster at this task when the
target represented the performed speech act than when it
did not”.|Exp3 < 1 (= (45))
|Comprehension of the utterance involved speech act activation. |I(47) = ?

Therefore, in order to eliminate possible priming effects, the experimental design

was modified again.
 András Kertész & Csilla Rákosi

4.2.4  Subcycle 4: Experiment 4

In a new subcycle, Experiment 4 replicated Experiment 3, the only difference
being that the sentences performed were equated with their semantic associates.
That is, the text of the control sentences and the corresponding speech act sen-
tences has been made as similar as possible. The authors obtained the following
experimental data:
(48) a. 0 < |Mean lexical decision reaction times in speech act scenarios was
887 ms in Experiment 4.|Exp4 < 1
b. 0 < |Mean lexical decision reaction times in control scenarios was 999
ms in Experiment 4.|Exp4 < 1
c. 0 < |In error-free trials, the overall error rate was 9%, and this rate did
not vary over conditions in Experiment 4.|Exp4 < 1
d. 0 < |“[…] the participants were significantly faster at the lexicon
decision task when the target represented the prior speech act than
when it did not […]” in Experiment 4.|Exp4 < 1
(Holtgraves & Ashley 2001: 87f.)

Since the results of Experiment 4 are in harmony with those of Experiment 3,

these data not only provide further support for (32) but they also motivate adding
(46) to the p-context again. Therefore, the outcome of the former cannot be due
to any priming effect. This means that (47) can be retrospectively re-evaluated and
transformed into the following plausible inference:
(49) 0 < |If the participants were significantly faster at the lexicon decision task
when the target represented the prior speech act than when it did not, and
there is no semantic priming effect, then comprehension of the utterance
involves speech act activation.|HA < 1
0 < |There was no semantic priming effect.|Exp4 < 1
0 < |“[…] the participants were significantly faster at the lexicon decision
task when the target represented the prior speech act than when it did not
[…]” in Experiment 4.|Exp4 < 1 (= (48))
0 < |Comprehension of the utterance involved speech act
activation. |I(49) < 1

After adding the data obtained in Experiment 4, the p-context is as in Figure 10.

4.3  The comparison with other solutions

The p-context includes the methodological norm which attributes high reliabil-
ity to real experiments as the source of the plausibility of hypotheses. Therefore,
Experiments 1–4 as sources make it possible to assign a very high plausibility
 Thought experiments and real experiments as converging data sources in pragmatics 

(35) (36) (37) (42) (43)

(38) (39) (44) (45) (46)

(32)

(48) (49)

Figure 10.  The state of the p-context after conducting Experiment 4

value to (32): “[…] we have converging evidence from four separate experiments
and three different methods that illocutionary force is activated” (Holtgraves &
Ashley 2001: 87f.).
The authors explicitly compare their result – i.e. (32) along with its increased
plausibility value gained from the four real experiments – to two rival hypotheses
and, accordingly, apply the Contrastive Strategy. The first is Cohen and Levesque’s
(1990) claim according to which
(50) “[s]peech act recognition is not required, although it may be entailed in the
recognition process.” (Holtgraves & Ashley 2001: 87f.)

The second hypothesis is the stance adopted by conversation analysis as put for-
ward for example in Schegloff (1991) which says that
(51) “[…] an interactant’s understanding of what has been accomplished with a
remark is revealed in that person’s subsequent turn at talk; there is no need
to consider a dimension of illocutionary force.” (ibid.)

Experimental data from Experiments 1–4 are not only weak evidence for (32)
but also against its rivals. Thus, (36), (42), (44) and (48) provide strong evidence
for (32) and against its rivals, (50) and (51), as well (see Kertész & Rákosi, this
­volume b: Section 3.2).

4.4  The final p-context

The final p-context includes (32) as the resolution of the problem raised in (31)
on the basis of the information available. It is the resolution of (31) in the sense of
Section 2.5 of Kertész & Rákosi (this volume b), because among the solutions (50),
(51) and (32), it is (32) that has been assigned the highest plausibility value and
 András Kertész & Csilla Rákosi

its acceptance leads to the least p-problematic p-context – even if this is merely
provisional and can be retrospectively re-evaluated in later cycles of the argumen-
tation process.
However, with respect to the relationship between thought experiments and real
experiments in pragmatics, Holtgraves and Ashley’s experiments work as a prism
through which the outcome of Searle’s thought experiment can be re-­evaluated
retrospectively:
(52) 0 < |Comprehension involves recognition of the action that is performed
with an utterance.|HA & Experiments 1–4 < 1 (= (32))
0 < |If comprehension involves recognition of the action that is performed
with an utterance, then “[…] utterances have both a propositional
content and an illocutionary force (the speech act performed with the
utterance).”|HA < 1
[0 < |If “[…] utterances have both a propositional content and an
illocutionary force (the speech act performed with the utterance)”, then
the analysis of illocutionary acts must capture both the intentional and
conventional aspects of meaning and especially the relationship between
them.|HA < 1]
0 < |The analysis of illocutionary acts must capture both the intentional and
conventional aspects of meaning and especially the relationship between
them.|I(52) < 1 (= (28))

This also means that experimental data originating from the four experiments
conducted by Holtgraves and Ashley considerably increased the plausibility value
of (21) as well.

5.  The relationship between thought experiments and real experiments

5.1  Introductory remarks

On the basis of the analyses presented in Sections 3 and 4, we have reached the
point where different aspects of the multifaceted relationship between Searle’s
thought experiment and the experiments presented by Holtgraves and Ashley
can be revealed. These findings can be generalised in order to obtain an answer to
problem (P). We will do this in three steps:
Step 1: The analyses of the thought/real experimental reports suggest the gener-
alisation that thought experiments and real experiments may be continuous in
several respects.
 Thought experiments and real experiments as converging data sources in pragmatics 

Step 2: On the basis of our analyses in Sections 3 and 4, we can realise that there
is a striking similarity between the argumentation structures of Searle’s thought
experimental report and Holtgraves and Ashley’s experimental report. It seems to
be instructive to examine possible analogies between the argumentation processes
presented in thought experimental reports and in reports on real experiments.
Step 3: Analogies between the argumentation structure of thought experimental
reports and real experimental reports might indicate a deeper similarity between
real and thought experiments themselves. That is, it is possible that there is a basic
similarity between the functioning of both experimental processes, too. Of course,
the question is whether and to what extent the experimental processes can be
reconstructed from the experimental reports.

5.2  S tep 1: The continuity between thought experiments

and real experiments
On the basis of the analyses we presented in Sections 3 and 4, there seem to be
three kinds of continuity between thought experiments and real experiments in the
cases we examined.
First, the starting p-context of Experiment 1 also includes the result of Searle’s
thought experiment as summarised in (28):
(28) The analysis of illocutionary acts must capture both the intentional and
conventional aspects of meaning and especially the relationship between
them.

Moreover, as we have noticed in Section 4.1, the starting p-context of Holtgraves &

Ashley (2001) includes hypotheses whose direct sources are, alongside Searle’s
thought experiment, further seminal contributions to speech act theory. We have
also seen that the experimental designs of Experiments 1–4 were built upon the
application of speech act theory. Namely, in the text of the scenarios, probe words
which characterised the given speech act were chosen on the basis of the applica-
tion of speech act theory.16 Since the classical contributions to speech act theory
made use of thought experiments (cf. Section 3), thought experiments are  –
although not solely – sources of the plausibility of certain premises used in real
experiments. Accordingly, we are entitled to put forward the generalisation that

.  The decisions were checked in a pretest. Despite this, the primary source of the plau-
sibility of several hypotheses on which the experimental design was based, is speech act
theory – the pretest only increased these values.
 András Kertész & Csilla Rákosi

the results of thought experiments also contributed to the plausibility of the result of
Experiments 1–4 in an indirect way.
Second, Searle’s thought experiment has taken the existence of the
­illocutionary aspects of meaning for granted and focused solely on the pres-
ence of the conventional aspects.17 Conversely, Holtgraves and Ashley tried to
find experimental evidence for the presence of the illocutionary force of utter-
ances but did not investigate their conventional aspect. Since all four experi-
ments they conducted provided evidence for the existence and recognition
of speaker intentions, they considerably increased the plausibility of that part
of (28) which is related to the intentional aspects of utterances. In this way,
the results of Experiments 1–4 also contribute to the plausibility of the result of
­Searle’s thought experiment.
To sum up, Holtgraves and Ashley’s experiments work as a prism through
which the outcome of Searle’s thought experiment can be re-evaluated retrospec-
tively. At the same time, however, Experiments 1–4 relied on hypotheses which
were partially and indirectly supported by thought experiments. It is important
to emphasise that this process, as a result of which the finding of Holtgraves and
Ashley’s real experiments feeds back to that of Searle’s thought experiment and
increases its plausibility, is not circular. Thus, the generalisation is that the result of
real experiments may increase the plausibility of the result of thought experiments
in the course of the cyclic, retrospective and prismatic re-evaluation of informa-
tion: new pieces of information were made use of, and the thought experiment
and the real experiment were conducive to the creation of a more comprehensive
picture by unifying their virtues.
There is also a third kind of relationship between Holtgraves and Ashley’s
experiments and thought experiments. Namely, the experimental designs of Experi-
ments 1–4 (and in general, every experimental design) can be interpreted as a special
kind of thought experiment. This interpretation is based on the circumstance that
experimenters go through the experimental process in their mind in order to test
whether the procedure proposed will provide relevant and reliable data. Of course,
the results of such thought experiments are considerably vaguer and less reliable
than those of the real experimental procedure. Despite this, the experimental
design is an indispensable part of real experiments. It delineates and motivates the
details of the experimental procedure to be carried out, and it also provides sev-
eral points of reference against which the results and all facets of the experimental
process can be interpreted and checked in order to gain reliable data.

.  See also Section 5.2.

 Thought experiments and real experiments as converging data sources in pragmatics 

5.3  S tep 2: Analogies between the thought experimental reports

and experimental reports
In Sections 3 and 4, we reconstructed one of Searle’s thought experimental reports
and Holtgraves and Ashley’s experimental report about four real experiments with
the help of the p-model. In both cases, we have found that the presentation of
the thought experiment and the real experiments can be interpreted as plausible
argumentation processes. These argumentation processes show a strong analogy at
several points:

–– The starting p-contexts included some theory of meaning. Nevertheless, the

plausibility of some basic hypotheses was questioned.
–– In order to find out whether the questioned hypotheses are tenable, they were
tested under controlled circumstances: in the case of Searle’s thought experiment,
an imagined situation was described and analysed, while with ­Holtgraves and
Ashley’s real experiments, a group of native speakers was put into an experi-
mental setting in order to monitor their linguistic behaviour.
–– Controlled circumstances were, in contrast to naturally emerging situations,
carefully designed and created with the aim of gaining relevant, reliable and
unbiased information.
–– Conjectures were put forward summarising the expectations of the researcher(s)
about possible consequences of the created/imagined state of affairs.
–– A chain of plausible inferences was made in both cases. These inferences con-
nected the directly emerging pieces of information from the given situation
with higher level hypotheses which summarised the results and could be
confronted with the hypotheses of the theory (i.e. compared to the previous
conjectures).
–– The hypotheses proposed were reformulated during the (thought) experimen-
tal process and were gradually replaced by refined, re-evaluated versions.
–– Nevertheless, the outcome of the experimental procedure and the picture
emerging from the imagination of the described situation, respectively, were
not totally reliable but only plausible to a certain extent. In the case of real
experiments, control experiments were conducted to rule out possible sys-
tematic errors, while Searle’s thought experiment made use of an analogous
situation as well as different possible interpretations of the same situation.
–– In both cases, the argumentation process was cyclic, not linear, in order to
make some sources more reliable.
–– Therefore, one of the most substantial characteristics of the argumentation
processes was that they operated with uncertain, fallible and often conflicting
pieces of information and have led not to certainly true, but only plausible results.
 András Kertész & Csilla Rákosi

–– Several details of the experimentation process/of the imagination of the

described situation were missing from the experimental report. A concise and
edited version replaced the original, real event and provided an information-
ally reduced and/or modified picture of the experimental procedure.
–– The evaluation of experimental results, however, could only start from the
experimental report which contained elements or traces of the original
argumentation process, as well as argumentative tools needed to make this
reduced set of information coherent, comprehensible and persuasive for the
reader.

Of course, there are also differences between the argumentation structures of

Searle’s thought experimental report and Holtgraves and Ashley’s real experi-
mental report, since analogy means solely (structural) similarity but not iden-
tity. The most important difference is perhaps that Holtgraves and Ashley’s
argumentation is of a more cyclic character than Searle’s, since the four experi-
ments presented are built on each other and provide revised versions of the
previous one, while in Searle’s thought experimental report, the central hypoth-
esis is elaborated gradually, that is, it is re-evaluated during the argumentation
process.
Although there is a strong analogy between the argumentation structures of
Searle’s thought experimental report as well as Holtgraves and Ashley’s experi-
mental report, two concerns emerge as a result of the above survey. First, from the
analogy between the argumentation structures, one must not automatically con-
clude that there is also an analogy between the thought experimental process and
the real experimental process itself. It may be the case, for example, that thought
experiments only imitate the style of real experiments by applying the rhetorical
tools of their experimental reports, while there are considerable and deep differ-
ences between the two kinds of experimental processes themselves. Second, it is
questionable whether, and to what extent, these experimental processes can be
reconstructed from the reports about them. A sufficiently complete reconstruc-
tion can only be achieved if reports contain all information which is relevant for
the identification and evaluation of the steps of the experimental process. This
means that the reader should be in a position to reconstruct and continue the
non-public argumentation process. In this way, the reader could become a virtual
participant in the creation, analysis and evaluation of the thought/real experimental
data. Therefore, in the next subsection we will examine whether this is the case
and try to reconstruct and compare the thought experimental and the real experi-
mental process.
 Thought experiments and real experiments as converging data sources in pragmatics 

5.4  Step 3: Analogies between real and thought experiments

5.4.1  The structure of real experiments18
The experimental design of a real experiment is a comprehensive preliminary
description of all facets of the process of experimentation allowing for a delin-
eation of the events which are supposed to occur and a rough estimation of the
results. It is therefore nothing else but a thought experiment which is, of course,
less informative than the real experiment. The outcome of the experimental proce-
dure always contains new pieces of information which may strengthen (and make
more precise) but also weaken or refute this preliminary guess.
The experimental designs of the four experiments were detailed and basically
well-planned. Nevertheless, there were a few problems. The most important con-
cern is typical of psycholinguistic experiments: there is a considerable amount
of information loss because the experimental report presents the text of only
one scenario. Therefore, the appropriateness of the stimulus material used in the
experiments cannot be checked by the reader of the experimental report. Second,
the change of the text of the control scenarios in Experiment 4 in comparison to
Experiment 3 is not clear (cf. Holtgraves & Ashley 2001: 87). Third, measuring the
reaction times during the task of deciding whether a letter string is a word or a
non-word counts as a method for avoiding lexical priming effects (cf. Keysar et al.
2000: 589). It was, however, precisely lexical priming effects which Holtgraves and
Ashley referred to as possible sources of error in Experiment 3 which made use of
this method.
The experimental procedure is a material procedure where an experimental
apparatus is set up, its operation is monitored and recorded under controlled
circumstances. During the experimental procedure perceptual data19 are gained
which undergo authentication and interpretation. Authentication means that the
experimenter has to check whether sources of noise do not distort the results, that
is, whether they have been ruled out successfully or, if this is not possible, whether
their effect can be eliminated with the help of statistical methods. As a result of

.  We have already proposed a model of psycholinguistic experiments based on the p-model
(cf. Kertész & Rákosi 2012; Rákosi 2011a,b, 2012), by relying on insights in the current litera-
ture on scientific experiments. Therefore, we will apply this model to the reconstruction of the
structure of Experiments 1–4.
.  In accordance with our definition of the notion of ‘datum’ (Kertész & Rákosi, this
volume b: Section 3.1), by ‘perceptual data’ we mean plausible statements the plausibility
value of which stems from perception as a direct source.
 András Kertész & Csilla Rákosi

this process, experimental data20 are obtained which are then confronted with the
given theory or with some rival theories. Nevertheless, it may happen that the
interpretation and authentication of the perceptual data indicates shortcomings
in the experimental procedure, in the experimental design, or in the theoretical
model of the phenomena or of the apparatus. If things do not run smoothly, one
turns back to some earlier stage of the experimentation process and modifies a
component until there is mutual support among the constituents. All facets of the
experiment have to be re-examined.
The experimental report does not contain the set of perceptual data, as is the
case with most psycholinguistic experiments on cognitive metaphors. Thus, their
interpretation and authentication cannot be properly judged by the reader. This
circumstance – together with the absence of the stimulus material – decreases the
plausibility of the experimental data. On the other hand, Holtgraves and Ashley
revealed and checked several possible systematic errors with the help of control
experiments. This makes their results more plausible, because in this way the reli-
ability of their experiments increased. Nevertheless, the authentication can never
be complete. In this case, for example, the following problematic points can be
mentioned:

–– In contrast to most psycholinguistic experiments, where feedback is used only

in practice trials, participants received feedback about the correctness of their
decision and their reaction times. This might have influenced the participants’
decisions.
–– The description of the statistical tools applied is rather sparse.
–– Experiment 2 was intended to be a control experiment for Experiment 1.
Nevertheless, the experimental procedure was not only changed at the point
which was deemed to be relevant for preventing a possible systematic error
(lengthening the final remarks). It was also indicated that the participants will
receive a second task as well. This might have influenced the focus of attention
and the behaviour of participants.
–– The authors did not pay attention to the great, and probably significant differ-
ence between error rates in Experiment 1 and Experiment 2.
–– Error rates were high (above 9%) in all experiments.
–– There was a significant difference between decision times in Experiments 3
and 4 both in speech act and in control scenarios. This is surprising because
the tasks in the speech act scenarios were completely identical in the two cases.

.  Likewise, by ‘experimental data’ we mean plausible statements the plausibility value of
which stems from an experiment as a direct source.
 Thought experiments and real experiments as converging data sources in pragmatics 

From these considerations it is clear that the experimental data gained are
not true with certainty but only plausible on the basis of these experiments as
sources. Their confrontation with the hypotheses of the theory, with previous
conjectures, involves factors which indicate that experimental results are not
decisive. They cannot prove or falsify hypotheses but are capable of increasing
their plausibility or making them implausible as weak or strong evidence for/
against them. For example, the reconstruction of the argumentation structure
of Holtgraves and Ashley’s experimental report contains the following plausible
inference (cf. Section 4.2):
(37) 0 < |“If comprehension of a remark involves activation of its illocutionary
force, then, when the probe represents the speech act just performed,
participants’ ability to verify that the probe had not been literally present
should be slowed.” |HA < 1 (= (35))
0 < |When the probe represented the speech act just performed,
participants’ ability to verify that the probe had not been literally present
was slowed.|Exp1 < 1 (= (36))
0 < |Comprehension of a remark involves activation of its illocutionary
force.|I(37) < 1

After the (partial) reconstruction of the experimentation process, it might become

more evident that the first premise of this inference is not true with certainty; it is
a long way from the antecedent to the consequent. Similarly, our reconstruction
has shown that there are many factors which might influence the plausibility of the
latent background assumptions which would be needed to transform this infer-
ence into a deductively valid one (cf. Kertész & Rákosi, this volume b: Section 2.3).
From this it follows that the experiments could only provide partial support for
the acceptance or rejection of the hypotheses of the given theory.
To sum up our analyses, the p-model interprets experiments as cyclic pro-
cesses. This cyclic process is organised and conducted by a plausible argumentation
process. This process governs the relationship among hypotheses of the experi-
mental design, the theoretical model of phenomena, the theoretical model of the
experimental apparatus, the theory being tested and its rivals as well as statements
describing the events of the experimental procedure, which capture the results of
the interpretation and authentication of perceptual data etc. See Figure 11.21

.  In Figure 11 and 12, simple arrows indicate successive stages of the experimental process;
dotted arrows signify the non-public argumentation process which organises the experi-
mental process.
 András Kertész & Csilla Rákosi

Experimental
design;
Theoretical
model of the Experimental
Perceptual data
phenomena; procedure
Theoretical
model of the
apparatus

Plausible
Revision
argumentation

Confrontation
Authentication
of the
and
experimental Experimental
interpretation
data with the data
of perceptual
theory/rival
data
theories

Transformation

Experimental
report

Figure 11.  The structure of real experiments

The effectiveness and comprehensiveness of the plausible argumentation process

organising the conduct and control of real experiments largely determines the reli-
ability of the obtained results.
This argumentation process is, however, not public but remains a private affair of
the experimenters. It is transformed into an experimental report/paper summarising
the results of the experiment and making them available to the scientific community.
Clearly, this transformation can be regarded as acceptable if it does not change the
plausibility value of the hypotheses of the original argumentation. This is of utmost
importance because the researcher runs the risk of eliminating relevant information
from the published report so that important decisions remain outside public control,
and she/he overestimates the plausibility of the results. From this it follows that the
acceptability of the experimental report is also influenced by the reliability of the
transformation of the non-public argumentation process into its public version.
 Thought experiments and real experiments as converging data sources in pragmatics 

5.4.2  The structure of thought experiments

Our next task is to examine whether this model of real experiments can be applied
to thought experiments smoothly, in such a way that besides certain differences,
we find a deep analogy between thought experiments and real experiments. We
cannot, of course, perform this task comprehensively and in general. Therefore,
we will try to find analogies between the stages of the model of real experiments
presented in the previous subsection and the stages of Searle’s thought experiment.
If this is possible, then we can generalise our results step by step and the outcome
of our considerations will be a model of thought experiments developed by anal-
ogy to our model of real experiments.
As we have seen in the previous subsection, with real experiments the starting
point is the elaboration of the experimental design. This includes a comprehensive
description of the equipment to be used and the circumstances under which the
experimental procedure will take place, as well as a rough depiction of the events
which are supposed to occur in this situation. The first step of Searle’s thought
experiment seems to be similar but somewhat simpler: it is the delineation of a
state of affairs (situation) which has to be detailed enough to be easily imagined
(visualised or simulated) by the members of the scientific community. Our recon-
struction of Searle’s thought experimental report in Section 3.1 interpreted the
passage cited in (4)(a) as the description of a situation in the above sense. Thus,
an important common feature is that both real and thought experiments intend
to create a situation from which relevant and reliable information can be gained.
The second component of real experiments is a material procedure, i.e. the
operation and monitoring of the functioning of an experimental apparatus. With
thought experiments, the imagination (visualisation, simulation) of the situation at
issue seems to correspond to this part of the experimental procedure. It is a cogni-
tive procedure which has to be performed both by the author and the reader. Of
course, there is an important difference: while the material procedure takes place
in the physical world, partly independently of the researcher, the cognitive proce-
dure is performed in the head of the researcher. Nevertheless, the material pro-
cedure is not totally independent of the researcher, either, because it has been set
up and controlled by him/her; and the theoretical commitments of the researcher
may influence the way the situation is imagined. Moreover, a thought experiment
cannot work without the active participation of the reader: he/she has in fact to
imagine the situation described. Therefore, a simple reading and acceptance of
Searle’s narrative is not enough. The reader should “process” the situation by rely-
ing on his/her linguistic intuition and everyday knowledge. Similarly, the reader of
an experimental report has to have some understanding of the functioning of the
 András Kertész & Csilla Rákosi

equipment used – although a thorough knowledge of it is usually not required.22

Therefore, the reader’s control over the thought experimental procedure usually
seems to be stronger than is the case with real experiments.
The third step of real experiments is the collection of the perceptual data
obtained during the experimental procedure. In most cases this means reading
the dial on the equipment and formulating the corresponding plausible statement.
In Searle’s thought experiment, this constituent is missing: it does not contain data
whose plausibility would originate directly from contact with the physical world.
Nevertheless, our everyday experience and linguistic intuition may allow us to
make highly plausible statements about the events which should occur in the given
situation on the basis of some analogy to similar real-world situations, that is, on
our experiences. These will be termed the experiential data emerging from the sit-
uation investigated. For example, the data presented in (5) or in (10) in ­Section 3.2
are experiential data in this sense because they provide a description of an imagi-
nary situation based on everyday linguistic intuition. That is, thought experiments
rely on data which result from an indirect contact with the empirical world.
The next move in real experiments is the authentication and interpretation of the
perceptual data. The experimenter checks whether the perceptual data are reliable
enough, and describes them with the help of the concepts of some theory. Since the
authentication involves the use of statistical tools, the obtained experimental data
lose a direct connection to the physical world. As a consequence, they are not true
with certainty. One cannot be sure that all possible sources of random and systematic
errors have been ruled out: every check of the correct functioning of the experimen-
tal apparatus requires new measurements, which requires new authentication, and so
on – the experimenter’s regress (cf. Collins 1985) can be broken only in a more or less
arbitrary manner. In thought experiments, the plausibility value of the experiential
data depends, among other things, on the degree of the typicality, transparency and
clear-cut character of the components of this state of affairs.23 The more determined
and unanimous the everyday experiences, the cultural, linguistic or other kinds of
intuitions related to this state of affairs are, the higher the plausibility value which can
be assigned to the experiential data. It may be the case, of course, that the reader does
not share the thought experimenter’s intentions and judges the experiential data less
plausible or even implausible; it is also possible that the reader has other intuitions
in connection with the depicted situation. Then, on the basis of the experiential
data, the theoretical consequences of the events which are supposed to occur in the

.  Thus, for example, theoretical physicists’ and experimental physicists’ knowledge of the
working of an experimental apparatus may differ in depth.
.  Of course, the whole of the situation need not be typical – situations depicted in thought
experiments are often fantastic or unrealistic, but contain realistic and well-known elements.
 Thought experiments and real experiments as converging data sources in pragmatics 

s­ ituation have to be identified. That is, the experiential data have to be described with
the help of the conceptual apparatus of the theory. As a result, thought experimental
data24 are obtained. With Searle’s thought experiment, hypothesis (7) is, for instance,
a consequence of the situation analysed in (4)(a)-(c) which has been obtained on
the basis of the experiential datum (5)(a). Of course, our reconstruction of Searle’s
argumentation process could have interpreted (7) as a thought experimental datum
as well; i.e. if we analyse the relationship between (5)(a) and (7) not as a connection
between a premise and a conclusion of the plausible inference in (6), but deem (7)
to be the re-evaluation of (5)(a) with the help of the terminology of Gricean theory
of meaning, then (7) will originate from a direct source and could be regarded as a
datum. Similarly, (12) could also be interpreted as a thought experimental datum
which is the re-evaluation of the experiential datum (10).
Following this, in both real and thought experiments, experimental data are
confronted with the hypotheses of a theory or rival theories. If the results seem to
be reliable, then the real/thought experimental process is transformed into a real/
thought experimental report. Akin to real experimental reports, thought experi-
mental reports are informationally impoverished in comparison to the thought
experimental process conducted by the experimenter:
By the time the thought experiment is communicated it is in a polished form.
That is, we are not shown the tinkering that went into setting it up and refining it.
[…] When conveyed, a thought experiment is as packaged and polished as a real-
world experiment is when it is published. (Nersessian 1993: 296)

The above characterisation of thought experiments supports the hypothesis that

there is a strong parallelism between the stages of real and thought experiments.
This correlation can also be found between the argumentative aspects of real
and thought experiments. A further component of the analogy between real and
thought experiments is that thought experiments are akin to real experiments
insofar as the experimental process is guided by a plausible argumentation process.
This argumentation process establishes the connections among hypotheses related
to the components of the real/thought experiment and makes it possible to decide
un which cases some component has to be revised because it is in conflict with
other components or under what circumstances no revision is needed because
there is a fit among the different elements. Moreover, with both real and thought
experiments, this argumentation process is not made public. One can reconstruct
it to some extent from the experimental report/paper which contains several traces
of it. This means that the real/thought experimental process can be e­ valuated only

.  In analogy to the notion’experimental data’, ‘thought experimental data’ should be un-
derstood as plausible statements the plausibility value of which stems from a thought experti-
ment as a direct source.
 András Kertész & Csilla Rákosi

through the prism of its informationally reduced and edited version. Reproduc-
ibility is in both cases a very important means of judging the reliability of the out-
come of the experimental process: repetitions may reveal suppressed but relevant
factors which indicate that the results are less reliable or even unreliable.
In summary, the p-model interprets thought experiments by analogy with
real experiments, that is, as cyclic processes organised and conducted by a plausible
argumentation process.25 Both real and thought experiments are characterised by
a double argumentative structure: they are organised and guided by a non-public
argumentation process and they are transformed into a public argumentation pro-
cess (thought/real experimental report). See Figure 12.

Delineation of a
situation;
Theoretical
model of the
involved Imagination of Experiential
phenomena; the situation data
Construction of
a mental model
of the situation

Plausible
Revision
argumentation

Determination
Confrontation
of the theoretical
of the Thought consequences
experimental experimental of the events
data with the data occurring in the
theory/rival
depicted
theories
situation

Transformation

Thought experimental
report

Figure 12.  The structure of thought experiments

.  It is worth remarking that in Holtgraves and Ashley’s experimental report the sources
were used cumulatively, because the experiments presented a revised version of the previous
experiment. This need not always be so; in other experimental reports the sources may be
independent of each other and may independently increase or decrease the plausibility of a
certain hypothesis (see Kertész & Rákosi 2012).
 Thought experiments and real experiments as converging data sources in pragmatics 

6.  The solution to (P)

We can now summarise our solution to the problem (P) as follows:

(SP) a. The result of a thought experiment may be included in the starting
p-context of a real experiment, and, accordingly, there may be an
indirect continuity between them. Such thought experiments may
serve as sources of the plausibility of the premises which certain
plausible inferences constituting the argumentation structure of the
experimental report are built on.
b. Through the cycles of the retrospective re-evaluation of information
the outcome of real experiments may modify the plausibility of the
result of the thought experiment included in the starting p-context.
c. Thought experiments are constitutive parts of real experiments:
the experimental design summarises the researcher’s previous
considerations about the workability of the experiment. Its task is
to check whether the experimental procedure will be capable of
producing reliable results.
d. Thought experiments and real experiments are closely related and have
a similar structure.
e. Both real and thought experiments provide plausible, but not certainly
true, experimental/thought experimental data.
f. There are also important differences between thought experiments and
real experiments.
g. Therefore, they have different strengths and weak points. From this it
follows that the integration of real and thought experiments as data
sources within some research may considerably enhance the reliability
of the hypotheses obtained.

The present chapter started with the remark that, while on the one hand, in con-
temporary linguistics there is a need to integrate different data sources, on the
other hand, the methodology which should govern such an integration is missing.
By obtaining (SP) with the help of the p-model, we have made an attempt to fill
this gap with respect to the relationship between thought experiments and real
experiments in pragmatics. Nevertheless, to reveal additional aspects of this rela-
tionship, numerous further case studies should be carried out.

Acknowledgements

Work on this chapter was supported by the MTA-DE Research Group for
Theoretical Linguistics, the project OTKA K 77823 as well as the TÁMOP-
4.2.2/B-10/1-2010-0024 project, which is co-financed by the European Union and
 András Kertész & Csilla Rákosi

the European Social Fund. Csilla Rákosi’s work was also supported by the János
Bolyai Research Scholarship of the Hungarian Academy of Sciences.

References

Austin, John L. 1962. How to do Things with Words. Oxford: OUP.

Brown, James Robert & Fehige, Yiftach. 2011. Thought experiments. Stanford Encyclopedia of
Philosophy. 〈http://plato.stanford.edu/entries/ thought-experiment/〉 (23 September 2013).
Cohen, Philip R. & Levesque, Hector J. 1990. Rational interaction as the basis for communica-
tion. In Intentions in communication, Philip R. Cohen, Jerry L. Morgan & Martha E. Pollack
(eds), 221–256. Cambridge MA: The MIT Press.
Collins, Harry M. 1985. Changing Order: Replication and Induction in Scientific Practice. Beverly
Hills CA: Sage.
Gibbs, Raymond W. 1983. Do people always process the literal meaning of indirect requests?
Journal of Experimental Psychology: Learning, Memory & Cognition 9: 524–533.
Gooding, David. 1993. What is experimental about thought experiment? In PSA: Proceedings
of the Biennial Meeting of the Philosophy of Science Association, 1992, Vol. 2: Symposia and
Invited Papers, 280–290. East Lansing MI: Philosophy of Science Association.
Grice, Paul. 1989a. [1957]. Meaning. In Grice 1989c, 213–223.
Grice, Paul. 1989b. Utterer’s meaning and intentions. In Grice 1989c, 86–116.
Grice, Paul. 1989c. Studies in the Way of Words. Cambridge MA.: Harvard University Press.
Holtgraves, Thomas & Ashley, Aaron. 2001. Comprehending illocutionary force. Memory &
Cognition 29: 83–90.
Kertész, András. 2010. On stage setting thought experiments in cognitive linguistics: A case
study. In Von Katastrophen, Zeichen und vom Ursprung der menschlichen Sprache, Cornelia
Stroh (ed.), 7–30. Bochum: Brockmeyer.
Kertész, András & Kiefer, Ferenc. 2013. From thought experiments to real experiments in
pragmatics. In Perspectives on Philosophy and Pragmatics, Alessandro Capone, Franco Lo
­Piparo & Marco Carapezza (eds). Berlin: Springer.
Kertész, András & Rákosi, Csilla. 2012. Data and Evidence in Linguistics. A Plausible Argumenta-
tion Model. Cambridge: CUP.
Kertész, András & Rákosi, Csilla. This volume a. Introduction: The state of the art and the struc-
ture of the book.
Kertész, András & Rákosi, Csilla. This volume b. The p-model of data and evidence in linguistics.
Keysar, Boaz, Shen, Yeshayahu, Glucksberg, Sam & Horton, William S. 2000. Conventional
­language: How metaphorical is it? Journal of Memory and Language 43: 576–593.
Kühne, Ulrich. 2005. Die Methode des Gedankenexperiments. Frankfurt: Suhrkamp.
Meggle, Georg & Ulkan, Maria. 1979. Kennst du das Land, wo die Zitronen blühn? (Zu Searles
Kritik an Grice). Papiere zur Linguistik 20: 75–87.
Meibauer, Jörg. 2012. Pragmatic evidence, context, and story design. An essay on recent develop-
ments in experimental pragmatics. Language Sciences 34(6): 768–776. (= Kertész, András,
Schwarz-Friesel, Monika & Consten, Manfred (eds), Converging Data Types in Cognitive
Linguistics. Amsterdam: Elsevier).
 Thought experiments and real experiments as converging data sources in pragmatics 

Moue, Aspasia S., Masavetas, Kyriakos A. & Karayianni, Haido. 2006. Tracing the develop-
ments of thought experiments in the philosophy of natural sciences. Journal for the General
­Philosophy of Science 37: 61–75.
Nersessian, Nancy J. 1993. In the theoretician’s laboratory: Thought experimenting as mental
modelling. In Proceedings of the 1992. Biennial Meeting of the Philosophy of Science Associa-
tion, Vol. 2, 291–301. East Landing MI: Philosophy of Science Association.
Nersessian, Nancy J. 1991. The cognitive sciences and the history of science. In Proceedings, joint
HSS-SHOT conference on Critical Problems and Research Frontiers in History of Science and
Technology, October 1991, 92–116. Madison WI.
Norton, John D. 1991. Thought experiments in Einstein’s work. In Thought Experiments in
­Science and Philosophy, Tamara Horowitz & Gerard J. Massey (eds), 129–48. Savage MD:
Rowman and Littlefield.
Norton, John D. 2002. On thought experiments: Is there more to the argument? 〈http://philsci-
archive.pitt.edu/960/1/Norton.doc〉
Noveck, Ira A. & Sperber, Dan. 2007. The why and how of experimental pragmatics. In Pragmat-
ics, Noel Burton-Roberts (ed.), 184–212. Basingstoke: Palgrave Macmillan.
Noveck, Ira A. & Reboul, Anne. 2008. Experimental Pragmatics: a Gricean turn in the study of
language. Trends in Cognitive Sciences 12(11): 425–431.
Rákosi, Csilla. 2011a. Metatheoretical reconstruction of psycholinguistic experiments. Part 1.
Sprachtheorie und germanistische Linguistik 21(1): 55–93. (Csatár, Péter (ed.), Metapher,
Metonymie und Argumentation/Metaphor, Metonymy and Argumentation 1. Münster:
Nodus).
Rákosi, Csilla. 2011b. Metatheoretical reconstruction of psycholinguistic experiments. Part 2.
Sprachtheorie und germanistische Linguistik, 21.2 [= Csatár, Péter (ed), Metapher, Metony-
mie und Argumentation/Metaphor, Metonymy and Argumentation 2. Münster: Nodus],
159–187.
Rákosi, Csilla. 2012. The fabulous engine: Strengths and flaws of psycholinguistic experiments.
Language Sciences 34(6): 682–702. (= Kertész, András, Schwarz-Friesel, Monika & ­Consten,
Manfred (eds), Converging Data Types in Cognitive Linguistics. Amsterdam: Elsevier).
Schegloff, Emanuel A. 1991. Conversational analysis and socially shared cognition. In Per-
spectives on Socially Shared Cognition, Lauren B. Resnick, John M. Levine & Stephanie D.
­Teasley (eds), 150–171. Washington DC: American Psychological Association.
Searle, John R. 1969. Speech Acts. An Essay in the Philosophy of Language. Cambridge: CUP.
Searle, John R. & Vanderveken, Daniel. 1985. Foundations of Illocutionary Logic. Cambridge:
CUP.
Walton, Douglas, Reed, Chris & Macagno, Fabrizio. 2008. Argumentation Schemes. Cambridge:
CUP.
Wittgenstein, Ludwig. 1958[1953]. Philosophical Investigations. Oxford: Blackwell.
 chapter 10

Data and the resolution of inconsistency

in Optimality Theory

Csilla Rákosi
MTA-DE Research Group for Theoretical Linguistics

This chapter provides a survey about the most frequent methods of inconsistency
resolution in Optimality Theory. With the help of the p-model, inconsistencies
in OT are divided into two main groups. The first group includes conflicts
that are deemed to be fatal and are solved usually by the modification of the
theory: namely, clashes between “linguistic data” (acceptability judgements) and
applications of the model (results of the evaluation procedure). The second group
consists of conflicts that are, in contrast, tolerable in the view of OT theorists:
inconsistencies between constraints and the best output candidates. The author’s
analyses result in the conclusion that there is no radical break between OT and
“standard” Generative Grammars with respect to the treatment of inconsistencies.

1.  Introduction

Diana Archangeli characterises the development which the role and treatment of
inconsistencies has undergone in the history of generative linguistics as follows:
[…] the original idea of a generative grammar (as in Chomsky 1957) […] accounts
directly for (i.e. generates) all and only all the expressions of a given language
with no auxiliary devices to remove ungrammatical expressions. Because of the
enormous complexity of the grammar which results from trying to put that idea
into practice, many linguists chose to drop the only all proviso for the generative
mechanism (the technical description of this state of affairs is that the grammar
overgenerates), and to add devices, called filters, to eliminate the ungrammatical
expressions that the generative mechanism allows […]. This situation, in turn,
has been viewed as unsatisfactory: why have two components of the grammar
responsible for separating out the expressions which are ungrammatical in a
particular language? […] Optimality Theory was introduced in response to
this situation. Optimality Theory opts for the ‘ideal separator’: a very simple
 Csilla Rákosi

generative mechanism (GEN) that allows ungrammatical expressions to be

created essentially without restriction, leaving all the work of separating out the
ungrammatical ones to filtering devices (EVAL).
(Archangeli 1997: viii–ix; emphasis as in the original)

She also emphasises a second significant difference between generative grammars

and OT and argues that OT ushered in a new era in theoretical linguistics:
Optimality Theory redefines the role of constraints and in so doing redefines
the research focus. All constraints are violable. Grammars define the relative
significance of violating specific constraints.
(Archangeli 1997: 27; emphasis added)

The quotation suggests that Optimality Theory has brought a radical change in the
treatment of inconsistencies. The scope of this chapter does not allow for a system-
atic comprehension of the strategies applied in different stages of the development
of generative linguistics and OT. Instead, we will focus on the role and treatment
of inconsistencies in OT. Thus, we will raise the following problem:
(P) (a) Which are the basic types of inconsistency in Optimality Theory?
(b) Which strategies of inconsistency resolution are applied in OT?

Section 2 provides a brief overview of the basic ideas of OT. In Section 3, I will try
to provide a partial answer (P)(a) with the help of an instructive case study. Some
passages of Rene Kager’s argumentation in connection with the introduction of
two correspondence constraints will be presented which are quite simple but seem
to be especially illuminating concerning the emergence and resolution of incon-
sistencies in OT. Section 4 will focus on (P)(b) and analyse the strategies applied
by Kager in order to resolve the conflicts with the help of the p-model. Section 5
will summarise our findings.

2.  Basic ideas of Optimality Theory

In OT the relation between the input and output forms is established by two for-
mal mechanisms, GEN (Generator) and EVAL (Evaluator). GEN produces all pos-
sible output-candidates from the input (a compound of elements of the linguistic
vocabulary). EVAL has the task of choosing the best candidate with the help of the
constraint hierarchy of the given language. The constraints are supposed to belong
to the Universal Grammar but their ranking is language-specific. The constraints
are violable; all output candidates infringe at least one constraint. It is the weight
of the conflicts with the constraints that is different and decisive. This weight is
basically determined not by the number of constraint violations but the ranking
 Data and the resolution of inconsistency in Optimality Theory 

of the violated constraints in comparison to the other candidates. Optimal is that

output-form which satisfies the constraint hierarchy of the given language to the
greatest extent: “Violation of constraints is tolerated in a very limited context. A
constraint may be violated successfully in order to satisfy a higher ranked con-
straint.” (Archangeli 1997: 11)
Constraints can be divided into two main groups, which embody conflicting
forces:
The first is markedness, which we use here as a general denominator for the
grammatical factors that exert pressure toward unmarked types of structure. This
force is counterbalanced by faithfulness, understood here as the combined
grammatical factors preserving lexical contrasts.
(Kager 1999: 4f.; emphasis as in the original)

Markedness and faithfulness are inherently conflicting. Whenever some lexical

contrast is being preserved, there will be some cost associated in terms of
markedness since in every opposition one member is marked.
(Kager 1999: 6; emphasis as in the original)

The optimal output form is chosen along the following lines:

(1) a. From a set of output-candidates the candidate which infringes the
­language specific hierarchy of constraints to the slightest degree is the
first one for which a constraint can be found in the hierarchy (from
the top towards the bottom of the constraint ranking) that is violated
to a lesser extent (or not at all) by this candidate than it is by the other
candidates.
b. From all possible output-candidates for a given input the most optimal
candidate is the one which infringes the rank order of constraints in the
sense of (1)(a) to a lesser extent than all other output candidates.
c. The optimal output-form as surface form may be with some constraints
in conflict.

3.  O
 n (P)(a): Emergence of inconsistencies in Kager (1999)’s
argumentation

As I already mentioned in Section 1, I will pick out certain passages of Kager

(1999)’s argumentation which provide a good overview of different types of incon-
sistencies and their treatment in OT. The investigated examples are relevant from
a metatheoretical (methodological) point of view. As object-theoretical problems,
however, most of them are simple and relatively small but this makes it easier for
us to concentrate on the treatment of inconsistencies in Optimality Theory. This
 Csilla Rákosi

section aims at revealing the most frequent and basic types of inconsistency in OT,
whose treatment will be analysed in Section 4 with the help of the p-model.
After some general typological considerations, Kager (1999: 92ff.) character-
ises Axininca Campa as one of the languages which do not allow onset-less sylla-
bles. As the data in (2) show, consonant epenthesis is applied when two vowels are
adjacent in the input; in contrast, there is no epenthesis when the input contains
an onset in every syllable:
(2) a. In Axininca Campa, it is the output-form [noŋkomati] that belongs to
the input /no-N-koma-i/.
b. In Axininca Campa, it is the output-form [noɲčhiki] that belongs to the
input /no-N-čhik-i/ and not the form *[noɲčhikti].

From typological insights and data such as (2), Kager (1999: 93) concludes that the
following structural well-formedness constraint has to be supposed to be undomi-
nated by other constraints in Axininca Campa:
(3) Onset: Syllables must have an onset.

3.1  I nconsistency No. 1: Satisfying a constraint leads to conflicts

with another constraint
Nevertheless, as Kager (1999: 100f.) remarks, epenthesis leads to the violation of
faithfulness, because a segment appears in the output-form which was not present
in the input. One of the faithfulness constraints, Dep-IO forbids such situations:
(4) Dep-IO: Output segments must have input correspondents.

Several data are in conflict with Dep-IO in Axininca Campa, because they show
that in order to satisfy Onset, this language employs epenthesis. This means that
the output forms have a consonant inserted between two vowels and this conso-
nant does not have any input correspondence (cf. Kager 1999: 101):
(5) a. In Axininca Campa, it is the output-form [noŋkomati] that belongs to
the input /no-N-koma-i/.
b. In Axininca Campa, it is the output-form [noŋkomataati] that belongs
to the input /no-N-koma-aa-i/.
c. In Axininca Campa, it is the output-form [noŋkomatakoti] that be-
longs to the input /no-N-koma-ako-i/.
d. In Axininca Campa, it is the output-form [noŋkomatakotaatiro] that
belongs to the input /no-N-koma-ako-aa-i-ro/.
e. In Axininca Campa, it is the output-form [noɲčhiki] that belongs to the
input /no-N-čhik-i/.
 Data and the resolution of inconsistency in Optimality Theory 

f. In Axininca Campa, it is the output-form [noɲčhikaati] that belongs to

the input /no-N-čhik-aa-i/.
g. In Axininca Campa, it is the output-form [noɲčhikakoti] that belongs
to the input /no-N-čhik-ako-i/.
h. In Axininca Campa, it is the output-form [noɲčhikakotaatiro] that
belongs to the input /no-N-čhik-ako-aa-i-ro/.

According to Kager (1999), this conflict arises from a contest of (prosodic)

markedness and faithfulness (correspondence). It is resolved by OT by supposing
the following ranking of Onset and Dep-IO:
(6) Axininca Campa is characterised by the constraint hierarchy
Onset >> Dep-IO.

Table 1 illustrates the resolution of the conflict.

Table 1.  Resolution of the conflict between Onset and Dep-IO

/no-N-koma-i/ Onset Dep-IO
 noŋ.ko.ma.ti *
  noŋ.ko.ma.i *!

Since the output-candidate [noŋkomati] violates only the lower ranked con-
straint Dep-IO, while [noŋkomai] infringes the higher ranked Onset, [noŋkomati]
counts as more optimal according to (1). This means that the conflict between Dep-
IO and the optimal output-form [noŋ.ko.ma.ti] is tolerated, while the inconsistency
between Onset and the output-candidate [noŋ.ko.ma.i] is deemed to be fatal.

3.2  I nconsistency No. 2: A constraint ranking fails to rule out certain

incorrect output forms belonging to the given input form
The next problem discussed by Kager (1999: 102) is that output forms which
result from the deletion of a vowel – such as the ungrammatical output candidate
*[noŋkoma] – are not ruled out by the constraints supposed so far.
Kager (1999: 102) resolves this conflict by the introduction of a second faith-
fulness constraint:
(7) Max-IO: Input segments must have output correspondents.

Max-IO is supposed to be ranked above Dep-IO but having the same rank as
Onset:
(8) Axininca Campa is characterised by the constraint hierarchy
Onset, Max-IO >> Dep-IO.
 Csilla Rákosi

The application of (8) yields the following evaluation. See Table 2.

Table 2.  Evaluation of the possible output-forms belonging to /no-N-koma-i/

/no-N-koma-i/ Onset Max-IO Dep-IO
 noŋ.ko.ma.ti *
  noŋ.ko.ma *!
  noŋ.ko.ma.i *!

While the infringement of the two higher ranked constraints Onset and
Max-IO is treated as fatal failure, the violation of Dep-IO by the optimal output-
form is t­ olerated. The reason for this decision is the circumstance that this tolera-
tion makes it possible to avoid the violation of higher ranked constraints.

3.3  I nconsistency No. 3: A constraint ranking fails to produce the correct

output form in the case of input forms having a certain characteristic
Although the constraint hierarchy in (8) seems to be working properly for /no-
N-koma-i/, there are inputs in the case of which (8) would produce false results.
The reason for this lies in the circumstance that vowels in word-initial position are
possible and no epenthesis is required, as the following data show:
(9) a. In Axininca Campa, it is the output-form [osampi] that belongs to the
input /osampi/ not the form *[tosampi].
b. In Axininca Campa, it is the output-form [iŋkomati] that belongs to
the input /i-N-koma-i/ not the form *[tiŋkomati].

Kager (1999: 110) first proposes modifying Onset as follows:

(10) Onset (naïve version): Syllables must have onsets, except word-initially.

He argues, however, that (10) would be problematic from two points of view. First,
in some languages all syllables must have onsets, that is, word-initial epenthesis
is not required. Thus, (10) cannot function as a linguistic universal. Second, the
left edge of words is “neglected” by a series of phonological changes. From this
Kager (1999: 110f.) concludes that raising the following alignment constraint is
well-motivated:
(11) 
Align-L: The left edge of the Grammatical Word coincides with the left
edge of the Prosodic Word.

In order to capture the data in (9), Kager (1999: 111) suggests giving up (8) and
replacing it with an extended version of it:
 Data and the resolution of inconsistency in Optimality Theory 

(12) Axininca Campa is characterised by the constraint hierarchy

Align-L, Max-IO >> Onset >> Dep-IO.

As Table 3 shows, the application of (12) leads to the correct results.

Table 3.  Application of the ranking Align-L, Max-IO >> Onset >> Dep-IO
/osampi/ Align-L Max-IO Onset Dep-IO
 o.sam.pi *
  to.sam.pi *! *
  
sam.pi *! *

Here, again, the violation of Onset by the optimal output-candidate is toler-

ated, while the infringement of the highest ranked constraint is deemed fatal.

3.4  I nconsistency No. 4: Constraint hierarchies contradict each

other within a language
Kager presents several unsolved problems in Chapter 9 of his book. One of them
is related to the description of the opacity of approximant devoicing in Isthmus
Nahuat (cf. Kager 1999: 383ff.).
Approximant devoicing means that in Isthmus Nahuat, approximants /l, w, j/
are devoiced at the end of a word; see, for example, (13):
(13) In Isthmus Nahuat, it is the output form [tajoːl ̥] that belongs to the
­underlying form /tajoːl/.

Approximant devoicing is interpreted by OT as the result of the constraint ranking

in (14):
(14) Isthmus Nahuat is characterised by the constraint hierarchy
*Voiced-Coda >> Ident-IO (voice).

The two constraints involved in (14) can be formulated as follows:

(15) *Voiced-Coda: Obstruents must not be voiced in coda position.
(16) 
Ident-IO (voice): The specification for the feature [voice] of an input
­segment must be preserved in its output correspondent.

Approximant devoicing becomes “underapplied” when it is engaged in interaction

with apocope. The process of apocope means the (optional) deletion of a word-
final unstressed vowel as in (17):
(17) In Isthmus Nahuat, it is the output form [tám] that belongs to the
­underlying form /támi/.
 Csilla Rákosi

OT tries to capture this process with the help of the following constraint hierarchy
(cf. Kager 1999: 377):
(18) Isthmus Nahuat is characterised by the constraint hierarchy
Final-C >> Max-IO,
where the dominating constraint is a well-formedness constraint:
(19) Final-C: Stem ends in a consonant.
The interaction of the two processes takes place as follows:
(20) In Isthmus Nahuat, it is the output form [ʃikɑkíl] that belongs to the
­underlying form /ʃikɑkíli/.
The problem is, however, that by taking into consideration the rankings in (14) and
(18), the four relevant constraints cannot be ranked in such a way that they pro-
vide the correct result by the evaluation of the possible output-forms of /ʃikɑkíli/.
For example, the hierarchy
(21) Isthmus Nahuat is characterised by the constraint hierarchy
Final-C >> Max-IO, *Voiced-Coda >> Ident-IO (voice).
evaluates an ungrammatical form as optimal. See Table 4.

Table 4.  Incorrect prediction produced by the constraint hierarchy Final-C >> Max-IO,
*Voiced-Coda >> Ident-IO (voice)
/ʃikɑkíli/ Final-C Max-IO *Voiced-Coda Ident-IO (voice)

  ʃikɑkíli *!
  ʃikɑkíl * *!
 ʃikɑkíl̥ * *

In order to obtain the correct form as optimal output form, the ranking of
*Voiced-Coda and Ident-IO should be reversed:
(22) Isthmus Nahuat is characterised by the constraint hierarchy
Final-C >> Max-IO >> Ident-IO (voice) >> *Voiced-Coda.
This constraint ranking leads to the evaluation as shown in Table 5.

Table 5.  Correct prediction produced by the constraint hierarchy Final-C >> Max-IO
>> Ident-IO (voice) >> *Voiced-Coda
/ʃikɑkíli/ Final-C Max-IO Ident-IO (voice) *Voiced-Coda

  ʃikɑkíli *!
 ʃikɑkíl * *
  ʃikɑkíl̥ * *!
 Data and the resolution of inconsistency in Optimality Theory 

(22) would lead to the correct outcome but it is inconsistent with (14) which is
motivated by data such as (13). This situation is labelled by Kager (1999: 377) as
a “ranking paradox”. Kager mentions some attempts to grasp this problem and
others in connection with opacity. He is of the opinion that they cannot be solved
simply by extending the constraint set with some new elements because these
phenomena
[…] seem to motivate strongly abstract intermediate levels of representation
between input and output. OT recognizes no such intermediate levels, in line with
standard assumptions […]. In standard OT, the mapping from input (‘Lexical
Representation’) to output (‘Phonetic Representation’) is direct. Moreover, well-
formedness constraints state requirements on output forms only, ignoring the
input. This assumption is another cornerstone of OT, essential to explanation of
segment inventories, conspiracies, and the Duplication Problem […].
(Kager 1999: 376; emphasis as in the original)

3.5  The solution to (P)(a)

To sum up, on the basis of our considerations presented in this Section, the follow-
ing answer to (P)(a) presents itself:
(S) (a) Four basic types of inconsistency in OT have been identified:
– Satisfying of a constraint leads to conflicts with another constraint.
– A constraint ranking fails to rule out certain incorrect output
forms to a given input form.
– A constraint ranking fails to produce the correct output form in
the case of input forms having a certain characteristic.
– Constraint hierarchies contradict each other within a language.
That is, a constraint hierarchy motivated by a subset of data is
­incompatible with a constraint ranking supported by another
­subset of data.

4.  O
 n (P)(b): Analysis and evaluation of the strategies applied
by Kager (1999)

In this Section, I will reconstruct and evaluate the strategies of inconsistency reso-
lution which were applied by Kager (1999) with the help of the p-model.

4.1  T
 reatment of Inconsistency No. 1: Joint application of the Contrastive
and the Combinative Strategies
In order to evaluate the treatment of inconsistencies by Kager, one has to recon-
struct the p-contexts in which the inconsistencies emerged. As we have seen in
 Csilla Rákosi

Section 2.4 of Kertész & Rákosi (this volume),1 the p-context involves, among
other things, the statements which can be judged to be plausible on the basis of
some source, as well as these sources. First of all, we have to reveal and examine
the sources of the plausibility of the conflicting statements. In Section 3.1 we have
shown that Inconsistency No. 1 results from a conflict between the satisfaction of
the constraints Onset and Dep-IO.
Let us first examine the source of the plausibility of the hypothesis regard-
ing Onset and hypotheses related to it. Statements in (2) can be interpreted as
“linguistic data”, that is, as plausible statements which capture some characteristic
of Axininca Campa and whose plausibility value originates from a direct source
in the sense of Section 2.2 of Kertész & Rákosi (this volume). (23) shows the data
structure of the statements in (2):2
(23) a. |In Axininca Campa, it is the output-form [noŋkomati] that belongs to
the input /no-N-koma-i/.|K = 0.93
b. |In Axininca Campa, it is the output-form [noɲčhiki] that belongs to
the input /no-N-čhik-i/ not the form *[noɲčhikti].|K = 0.9

These two statements seem to be assigned a maximal or at least very high plau-
sibility value by Kager, since he never indicates that statements of this type could
turn out to be false and be in need of revision. The evaluation of the reliability of
these hypotheses is partly based on grammaticality judgments of native speakers.
Nevertheless, we should not forget that both statements contain linguistic analy-
ses, too. Therefore, their plausibility does not stem solely from the linguistic intu-
ition of native speakers but is influenced by the reliability of the linguistic analyses
conducted by Kager, too. Moreover, acceptability judgments cannot be regarded
as totally reliable sources (cf. Kertész & Rákosi 2012: 20f., 68, 170ff.). Therefore,
instead of maximal plausibility (truth with certainty), a very high plausibility value
seems to be more appropriate.
Onset, as introduced in (3), belongs to the structural well-formedness
constraints:
(3) Onset: Syllables must have an onset.

.  See also Kertész & Rákosi (2012: 121ff.).

.  See also Kertész & Rákosi (2012: 67f.).
.  |S|K = 0.9 indicates that statement S obtains a very high plausibility value on the basis of
source K, that is, Kager’s work. See Section 2.2 of Kertész & Rákosi (this volume) as well as
Kertész & Rákosi (2012: 70ff.) on the assignment of plausibility values.
 Data and the resolution of inconsistency in Optimality Theory 

Its plausibility value originates from two sources (cf. Kager 1999: 93f.): it is
supported by typological considerations (onsets are allowed in every language,
and there are languages in which they are obligatory) and by phonetic reasons (the
articulation and perception of vowels is easier after a preceding consonant).
The “linguistic datum” in (23)(a) as well as the typological insight that in most
languages syllables usually start with a consonant serve as premises of the follow-
ing plausible inference:

(24) |In Axininca Campa, it is the output-form [noŋkomati] that belongs to the
input /no-N-koma-i/.|K = 0.9 (= (23)(a))
|In most languages syllables usually start with a consonant.|K = 0.9
|If it is the output-form [noŋkomati] that belongs to the input
/no-N-koma-i/ in Axininca Campa, and in most languages syllables usually
start with a consonant, then output-forms in which all syllables have an
onset are optimal while output-forms containing syllables without an onset
are not optimal in Axininca Campa.|K = 0.3
|If output-forms in which all syllables have an onset are optimal while
­output-forms containing syllables without onset are not optimal in
­Axininca Campa, then the constraint Onset is undominated in this
language.|K = 1
|The constraint Onset is undominated in Axininca Campa.|I(24) = 0.74

This inference belongs to the first type of plausible inferences (Kertész & Rákosi,
this volume: Section 2.3, 2012: 86ff.), because there is a logical consequence rela-
tion between its premises and conclusion but some premises are only plausible
instead of being true with certainty. The indices K and I(24) refer to the sources
Kager (1999) and this inference, respectively. The premises of the inference in (24)
make the hypothesis that Onset is undominated in this language highly plausible
but not totally certain. This means that its later revision cannot be ruled out but
requires strong counter-arguments.
As a second step, we turn to the other originator of Inconsistency No. 1, the
constraint Dep-IO:

(4) Dep-IO: Output segments must have input correspondents.

.  For the assignment of plausibility values to the conclusion of inferences, see Kertész &
Rákosi (2012: Section 9.5).
 Csilla Rákosi

Dep-IO results from general linguistic considerations. It belongs to the faithful-

ness constraints, which are of central importance in OT and belong to its basic
assumptions:
From a functional viewpoint, faithfulness constraints protect the lexical items of
a language against the ‘eroding’ powers of markedness constraints, and thereby
serve two major communicative functions. First, they preserve lexical contrasts,
making it possible for languages to have sets of formally distinct lexical items to
express different meanings. […] Second, by limiting the distance between input
and output, faithfulness constraints restrict the shape variability of lexical items.
Faithfulness thus keeps the contextual realizations of a single morpheme (called
its alternants) from drifting too far apart. This enhances the one-to-one relations
of meaning and form. (Kager 1999: 10; emphasis as in the original)

This means that Dep-IO has a high plausibility value in OT. It is, however, incom-
patible with data which show the application of epenthesis. The linguistic data in
(5)(a)–(h) belong to this group and are regarded as having a very high plausibility
value by Kager. From the joint presence of Dep-IO and (5)(a)–(h) in the p-context,
a series of p-inconsistencies5 emerge. For instance, (5)(a) is weak evidence against
Dep-IO, because it is one of the premises of a plausible inference the conclusion of
which is a special case of the negation of Dep-IO:
(25) |In Axininca Campa, it is the output-form [noŋkomati] that belongs to the
input /no-N-koma-i/.|K = 0.9 (= (23)(a))
|If in Axininca Campa, it is the output-form [noŋkomati] that belongs to
the input /no-N-koma-i/, then in the input form /no-N-koma-i/ an onset ‘t’
is supplied.|K = 0.9
|If in the input form /no-N-koma-i/ an onset ‘t’ is supplied, then the out-
put form [noŋkomati] contains a segment which does not have an input
correspondent.|K = 1
|The output form [noŋkomati] contains a segment which does not have an
input correspondent.|I(25) = 0.9

On the basis of similar plausible inferences relying on the data (5)(b)-(h), the fol-
lowing inductive generalisation can be achieved:
(26) There are output segments in Axininca Campa which do not have input
correspondents.

.  P-inconsistency means that both members of an inconsistent pair of statements are plau-
sible on the basis of some (but not the same) source; see Kertész & Rákosi (this volume:
Section 2.5, 2012: 130).
 Data and the resolution of inconsistency in Optimality Theory 

To sum up, this conflict can be characterised as follows:

(27) a. The hypothesis Dep-IO receives high plausibility value in OT because
it belongs to a very important type of constraints, the faithfulness
­constraints.
b. “Linguistic data” in (5) are deemed very plausible or even true with
certainty by Kager; this means that they are never given up.
c. The hypothesis Dep-IO is p-inconsistent with data in (5).
d. 
Dep-IO cannot be rejected, because it rules out a series of possible
output-forms which must not be optimal. Thus, without the application
of this constraint, the model would provide results which are in conflict
with “linguistic data”.
e. The data in (5) and further pieces of evidence mentioned earlier make
the hypothesis about Onset highly plausible, because this constraint is
assumed to provide explanation for them (that is, for the phenomenon
of t-epenthesis).
From (27) it follows that this conflict requires a solution which makes it possible
to keep both constraints. With this aim in view, the ranking Onset >> Dep-IO is
proposed, and the following thought experiment is conducted in order to test this
hypothesis:

Thought Experiment 1
Step 1: Simulation of functioning of GEN
The plausibility values of hypotheses about output-candidates are ignored and they
receive the same low initial plausibility value which conjectures usually obtain.6
This means in this case that output-candidates employing t-epenthesis as well as
output-candidates mirroring the input are examined as equally plausible possible
output-forms:
|In Axininca Campa, it is the output-form [noŋkomati] that belongs to the
input /no-N-koma-i/.|TE = 0.2
1
|In Axininca Campa, it is the output-form [noŋkomai] that belongs to the
input /no-N-koma-i/.|TE = 0.2
1

Step 2: Simulation of functioning of EVAL 1: Confronting constraints

and ­output-candidates
Hypotheses about the output-candidates are regarded as rivals and checked
to see whether they are in conflict with the hypotheses capturing the supposed

.  That is, we “forget” the plausibility value of, for example, the statements in (2) and (5).
 Csilla Rákosi

c­ onstraints. The ranking of the constraints is not yet taken into consideration at
this step. Two conflicts can be detected:
i. |onset: Syllables must have an onset.|TE = 0.8
1
vs.
|The last syllable of the output-form [noŋkomai] does not have an
onset.|TE = 0.9
1
ii. |dep-io: Output segments must have input correspondents.|TE = 0.8
1
vs.
|Segment ‘t’ in the last syllable of the output-form [noŋkomati] does not
have an input correspondent.|TE = 0.9
1

Figure 1 shows the current state of the p-context. It boils down to two p-context
versions which contain the two constraints as well as one of the rival hypotheses.

p-context version1 p-context version2

central hypothesis: central hypothesis:
the output is [noŋkomai] the output is [noŋkomati]

Onset Dep-IO

Figure 1.  The p-context after the application of the Contrastive Strategy in
Thought ­Experiment 1

At his point, an attempt is made to apply the Contrastive Strategy in the sense
of Subsection 2.6.2 of Kertész & Rákosi (this volume)7 insofar as it is checked
whether one of the rival p-context versions can be deemed to be definitely better
than the other. This, however, does not lead to a satisfactory result (resolution of
the p-inconsistency), because both p-context versions are p-inconsistent.

Step 3: Simulation of functioning of EVAL 2: The decision procedure

As a next step, hypothesis (6) on ranking of the constraints and hypothesis (1)
which describes the process of the choice of the optimal output-form are taken
into account as well:
(6) Axininca Campa is characterised by the constraint hierarchy
onset >> dep-io.

.  For a more detailed characterisation of the problem solving strategies, see also Kertész &
Rákosi (2012: Section 10.6).
 Data and the resolution of inconsistency in Optimality Theory 

(6) receives a moderate plausibility value which is higher than the plausibility of
the hypotheses about the possible output forms but lower than that of the con-
straints. A further important characteristic of the treatment of (6) is that it is inter-
preted in such a way that it states that in Axininca Campa, there are only these two
constraints; that is, the constraint hierarchy in this language is as described in (6).
The reason for this is the circumstance that constraint hierarchies cannot simply
be extended by adding new constraint(s) to them, because the new hierarchy can
evaluate certain output-forms differently from the old hierarchy.8 Thus, the addi-
tion of a new constraint has to be regarded as a rather serious modification of the
original ranking.
During the simulation of the decision procedure, the following plausible
hypotheses are obtained:
(28) a. Since Dep-IO is dominated by Onset according to (6), and the
output-candidate [noŋkomati] does not infringe Onset while its rival,
[noŋkomai] does, the former form violates the constraint hierarchy in
(6) to a lesser extent according to (1)(a).
b. Therefore, the output-candidate [noŋkomati] is more optimal in the
sense of (1)(b) than the output-candidate [noŋkomai].
c. Syllables must generally have an onset in Axininca Campa, but the
optimal output-candidate ([noŋkomati]) as a surface form may contain
onset-less syllables along the lines of (1)(c).9
d. Input segments must generally have output correspondents in
­Axininca Campa, but the optimal output-candidate ([noŋkomati]) as
a ­surface form may contain output segments which do not have input
­correspondents along the lines of (1)(c).
e. Since [noŋkomai] is not the optimal output form, therefore the
­p-inconsistency between Onset and the hypothesis stating that the
output is [noŋkomai] cannot be tolerated but could be resolved by giv-
ing up the latter.

(28)(e) means that the p-context version around the hypothesis that the output is
[noŋkomai] remains p-inconsistent. As opposed to this, the rival p-context version
resolves the p-inconsistency with the help of the Combinative Strategy, that is, by
keeping both conflicting hypotheses systematically separated from each other (cf.
Kertész & Rákosi, this volume: Subsection 2.6.2, 2012: 158). Namely, (28)(c) and

.  Cf. Sections 4.2–4.4 on this.

.  That is, the constraint dep-io can be applied only to the possible output-forms during the
evaluation process, but not to the surface forms at the surface representation.
 Csilla Rákosi

(d) mean that this p-context version can be divided into two co-existing p-context
versions: into the p-context representing the final state of the evaluation process
conducted by EVAL on the one hand, and the surface representation on the other
hand. The former p-context includes the constraints as well as a hypothesis about
the ranking and optimality of the output-candidates. The latter involves a hypoth-
esis about the output form which appears at the surface but the constraints do
not form a part of it. The reason for this finding is that hypotheses capturing the
constraints cannot be regarded as linguistic rules without exceptions but rather
as tendencies which – at least in the case of highly ranked ones – are satisfied by
most linguistic forms within a given language. In this way it is possible to separate
the constraints and the surface forms in the case of the optimal output-candidate.
This strategy, however, does not function with output-candidates which are not
optimal, because they do not reach the surface representation but remain within a
p-context where the constraints are also present.
From this it follows that on the basis of the information taken into consider-
ation so far it is reasonable to give up the hypothesis that the output is [noŋkomai]
and the p-context version around it and retain its rival. Thus, the p-context
depicted in Figure 1 is transformed into the p-context in Figure 2.

Onset
Dep-IO
Onset >> Dep-IO
(1)
the output is [noŋkomati] the output is the output form is
[noŋkomai] [noŋkomati]
the form [noŋkomati] is
the most optimal form
Final state of the decision Surface
procedure by EVAL representation

Figure 2.  The p-context after application of the Combinative Strategy in

Thought Experiment 1

Step 4: Confrontation the outcome of the evaluation procedure

with “linguistic data”
As the last step of the thought experiment, the outcome of the evaluation proce-
dure – namely, the hypothesis about the most optimal output-candidate – is com-
pared with “linguistic data”, that is, with data whose plausibility value originates
(at least partly) from the linguistic intuition of native speakers such as the state-
ments in (5). If they are p-inconsistent, then the constraint hierarchy is modified:
 Data and the resolution of inconsistency in Optimality Theory 

the hypothesis about their ranking is given up and altered or new constraints are
proposed etc. In Thought Experiment 1, there is no p-inconsistency between the
hypothesis that it is the output-candidate [noŋkomati] which is the most optimal
form and the datum in (23)(a).
After conducting the thought experiment, the plausibility value of the hypoth-
eses formulating the two constraints in (3) and (4), as well as their ranking in (6)
increased, because inference (29) as an indirect source makes it possible to assing
a positive plausibility value to (6), too:

(29) |If Axininca Campa is characterised by the constraint hierarchy Onset >>
Dep-IO (= (6)), then the output-candidate [noŋkomati] is more optimal
than the output-candidate [noŋkomai].|TE = 1
1
[|If the output-candidate [noŋkomati] is more optimal than the output-
candidate [noŋkomai], then it is the output-candidate [noŋkomati] that is
the most optimal among the output-forms belonging to the input /no-N-
koma-i/.|TE = 0.3]
1
|If it is the output-candidate [noŋkomati] that is the most optimal among
the output-forms belonging to the input /no-N-koma-i/, then it is the
form [noŋkomati] that belongs to the input /no-N-koma-i/ at the surface
representation.|TE = 0.9
1
|It is the form [noŋkomati] that belongs to the input /no-N-koma-i/ at the
surface representation.|K = 0.9 (= (23)(a))
|Axininca Campa is characterised by the constraint hierarchy
Onset >> Dep-IO.|I(29) = 0.6 (= (6))

Thus, the plausibility of (6) on the basis of its direct source and (29) as an indirect
source together is higher than its initial value. To sum up, with the help of the
p-model, Kager’s treatment of the conflict between the optimal output-candidate
[noŋkomati] and the two constraints can be evaluated as follows:
(30) a. Step 2 of the thought experiment makes it possible to reveal
which constraints the single output-candidates violate. In this way,
­p-inconsistencies between constraints and output-candidates are
­identified.
b. These p-inconsistencies are divided into two groups. Conflicts between
non-optimal output-forms and constraints are not tolerated but they
are regarded as fatal. As a consequence, these output-forms are given
up and they are supposed to not reach the surface representation.
That is, with the help of the Contrastive Strategy, hypotheses about
these output-forms as well as the p-context versions around them are
rejected.
 Csilla Rákosi

c. As opposed to this, conflicts between optimal output-candidates and

constraints are resolved with the help of the Combinative Strategy.
They are ­represented as weak inconsistencies. This means that they
can be tolerated because the conflicting statements became separated
systematically: constraints belong to the p-context gained as the final
state of the evaluation procedure, while the optimal form is added to
the surface representation.
d. Nevertheless, there is a far-reaching difference between this ­application
of the Combinative Strategy and its use in the traditional ­generative
grammars. In GB Theory, for example, the members of the weak
­p-inconsistency are separated into different p-context ­versions on
the basis of p-context-independent criteria. For instance, certain
­statements are related to the deep structure, others to the ­surface
structure etc. (see Kertész & Rákosi 2013). In OT, however, a
­context-sensitive ­criterion is applied. Only hypotheses about the optimal
output-candidate can be transformed into the surface representation
and separated from the p-context containing the constraints. Which
form can be ­regarded as optimal depends, however, on the evaluation
of all possible ­output-candidates.

To sum up, the joint application of the Contrastive and the Combinative Strategies
resulted in a solution of the inconsistency between the hypothesis Dep-IO and lin-
guistic datum in (23)(a). Nevertheless, it would be premature to claim that this is
the resolution of this p-inconsistency (cf. Kertész & Rákosi this volume: ­Section 2.5,
2012: 133), since, for example, the above thought experiment took into consideration
only two output-candidates associated with an input form. There are, however, fur-
ther possible forms as well. That is, the proposed solution is not satisfactory, since it
does not take into consideration all information available. Therefore, the emergence
of further inconsistencies cannot be ruled out. As we have seen in Section 3.2, the
extension of the p-context by further available data leads in fact to new conflicts.

4.2  T
 reatment of Inconsistency No. 2: Modification of the hypothesis about
the constraint hierarchy
As I have already mentioned, the weak point of inference (29) is its first premise
and latent background assumption. On the basis of the comparison of only two
possible output-candidates it is stated that it is [noŋkomati] which is the most
optimal form among the output-forms belonging to the input /no-N-koma-i/.
Thus, the thought experiment made use of only a small subset of data inasmuch
as only two output-candidates were taken into consideration by the simulation of
the evaluation process. It is therefore possible that on the basis of a wider range
of data, this hypothesis loses its plausibility, and, as a consequence, the conclusion
can no longer be regarded as plausible, either.
 Data and the resolution of inconsistency in Optimality Theory 

Actually, if the range of the output-candidates is extended by noŋkoma, then

the thought experiment will not run smoothly and will lead to an incorrect out-
come, as we have seen in Section 3.2. Thus, the next task will be to compare the
conduct of this thought experiment with Thought Experiment 1.

Thought Experiment 2
Step 1: Simulation of the functioning of GEN
A third datum is added to those in Thought Experiment 1, which receives the
same low initial plausibility value as the other two:
|In Axininca Campa, it is the output-form [noŋkoma] that belongs to the
input /no-N-koma-i/.|TE = 0.2
2

Step 2: Simulation of the functioning of EVAL 1: Confronting constraints and

output-candidates
Since the output-candidate [noŋkoma] does not infringe the two constraints, there
is no further p-inconsistency between constraints and candidates. Figure 3 shows
the current state of the p-context.

p-context version1 p-context version2

central hypothesis: central hypothesis:
The output is [noŋkomai] The output is [noŋkomati]

Onset Dep-IO

p-context version3
central hypothesis:
The output is [noŋkoma]

Figure 3.  The p-context after the application of the Contrastive Strategy in
Thought Experiment 2

In this case, contrary to Thought Experiment 1, the application of the

­ ontrastive Strategy seems to lead to the resolution of the p-inconsistency, since
C
the third p-context version is p-consistent. That is, the p-context version around
[noŋkoma] seems to be the best solution on the basis of the information taken into
consideration so far.
Step 3: Simulation of the functioning of EVAL 2: The decision procedure
The optimal output-candidate along the lines of (1) is in this case clearly the form
noŋkoma, because it is the only possible output form that is compatible with both
constraints.
 Csilla Rákosi

Step 4: Confrontation of the outcome of the evaluation procedure with

“linguistic data”
The evaluation process resulted in the hypothesis that [noŋkoma] is the most opti-
mal output form. Therefore, this form should appear at the surface representation.
This is, however, cannot be the case, because if we add the relevant “linguistic
datum” to the p-context version of the surface representation, it becomes p-incon-
sistent. See Figure 4.

Onset
Dep-IO the output is
[noŋkomati] the output form is
Onset >> Dep-IO
[noŋkoma]
(1)
[noŋkoma] is the
most optimal form the output is
[noŋkomai] (2)
the output is [noŋkoma]

Final state of the decision Surface

procedure by EVAL representation

Figure 4.  The p-context after application of the Combinative Strategy in

Thought Experiment 2

This means that Thought Experiment 2 as a source makes the hypothesis that
Axininca Campa is characterised by the constraint hierarchy Onset >> Dep-IO
implausible:
(31) |If Axininca Campa is characterised by the constraint hierarchy Onset >>
Dep-IO (= (6)), then the output-candidate [noŋkoma] is more optimal than
the output-candidates [noŋkomati] or [noŋkomai].|TE = 1
2
[|If the output-candidate [noŋkoma] is more optimal than the output-
candidates [noŋkomati] or [noŋkomai], then it is the output-candidate
[noŋkoma] that is the most optimal among the output-forms belonging to
the input /no-N-koma-i/.|TE = 0.4]
2
|If it is the output-candidate [noŋkoma] that is the most optimal among
the output-forms belonging to the input /no-N-koma-i/, then it is the
form [noŋkoma] that belongs to the input /no-N-koma-i/ at the surface
representation.|TE = 0.9
2
|~ It is the output-form [noŋkoma] that belongs to the input /no-N-koma-i/
at the surface representation.|K = 0.9
|~ Axininca Campa is characterised by the constraint hierarchy Onset >>
Dep-IO.|I(31) = 0.8 (= ~ (6))
 Data and the resolution of inconsistency in Optimality Theory 

In contrast to (29), the first premise and the latent background assumption of
(31) take into consideration three possible output-candidates. The problem is,
however, that the evaluation should not result in our judging [noŋkoma] more
optimal than [noŋkomati]. Therefore, Kager introduces a third constraint that
forbids the deletion of input segments. That is, he modifies the hypothesis about
the supposed constraint hierarchy in Axininca Campa in such a way that it also
contains a constraint which is in conflict with the output-form which has led to
p-inconsistency. Thus, the p-context is extended by (7), further (6) is given up
and replaced by (8):
(7) Max-IO: Input segments must have output correspondents.
(8) Axininca Campa is characterised by the constraint hierarchy
Onset, Max-IO >> Dep-IO
In order to come to a decision between the p-context version built around (6) and
a new solution based on (8), the latter has first to be elaborated. Therefore, the ten-
ability of this constellation is tested again with the help of a thought experiment.

Thought Experiment 3
Step 1: Simulation of the functioning of GEN
The set of “linguistic data” is the same as it was in Step 1 of the previous thought
experiment, because the range of output-candidates has not changed.
Step 2: Simulation of the functioning of EVAL 1: Confronting constraints and
output-candidates
The set of constraints has been extended by Max-IO, which receives the same high
initial plausibility value as Onset and Dep-IO.
Figure 5 shows the result of the confrontation of the extended constraint set
and the output-candidates.

p-context version1 p-context version2

central hypothesis: central hypothesis:
The output is [noŋkomai] The output is [noŋkomati]

Onset, Max-IO Dep-IO

p-context version3
central hypothesis:
The output is [noŋkoma]

Figure 5.  The p-context after the application of the Contrastive Strategy in
Thought Experiment 3
 Csilla Rákosi

All three rival p-context versions are p-inconsistent because all possible output-
forms violate a constraint.
Step 3: Simulation of the functioning of EVAL 2: The decision procedure
When we also take into consideration (8), that is, the hierarchy of the three
­constraints as well, then we obtain that the output-candidate noŋkomati is more
optimal than its two rivals according to (1). Therefore, it is reasonable to give
up the p-context-versions built around the less optimal forms and keep only
­p-context version No. 2 by making use of the Contrastive Strategy. Then, the
Exclusive ­Strategy (cf. Kertész & Rákosi this volume: Subsection 2.6.2, 2012: 158)
has to be applied in order to check whether we obtain a satisfactory solution of
the starting problem. As we have seen, the best p-context-version is p-inconsis-
tent. Therefore, the Combinative Strategy is used in order to prevent a conflict
between the constraints and the surface form as shown in Figure 6.

Onset, Max-IO
Dep-IO the output is
[noŋkoma] the output form is
Onset, Max-IO >> Dep-IO
[noŋkomati]
(1)
[noŋkomati] is the
most optimal form
the output is
the output is [noŋkomati] [noŋkomai]

Final state of the decision Surface

procedure by EVAL representation

Figure 6.  The p-context after application of the Combinative Strategy in

Thought Experiment 3

Step 4: Confrontation of the outcome of the evaluation procedure with

“linguistic data”
Since the outcome of the third thought experiment is fully compatible with the
“linguistic datum” in (23)(a), the plausibility value of the hypotheses in (7) and (8)
increases, as (32) also indicates:

(32) |If Axininca Campa is characterised by the constraint hierarchy Onset,

Max-IO >> Dep-IO (= (8)), then the output-candidate [noŋkomati] is
more optimal than output-candidates [noŋkomai] and [noŋkoma].|TE = 1
3
[|If the output-candidate [noŋkomati] is more optimal than output-
candidates [noŋkomai] and [noŋkoma], then it is the output-candidate
 Data and the resolution of inconsistency in Optimality Theory 

[noŋkomati] that is the most optimal among the output-forms belonging to

the input /no-N-koma-i/.|TE = 0.4]
3
|If it is the output-candidate [noŋkomati] that is the most optimal among
the output-forms belonging to the input /no-N-koma-i/, then it is the
form [noŋkomati] that belongs to the input /no-N-koma-i/ at the surface
representation.|TE = 0.9
3
|It is the form [noŋkomati] that belongs to the input /no-N-koma-i/ at the
surface representation.|K = 0.9
|Axininca Campa is characterised by the constraint hierarchy Onset, Max-
IO >> Dep-IO.|I(32) = 0.6 (= (8))

From these considerations it follows that (8) – that is, the hypothesis stating that
Axininca Campa has three constraints of which Dep-IO has a lower rank than
the other two – should be preferred over (6) on the basis of all the data contained
by the p-context. This means that the p-inconsistency caused by the failure of the
constraint ranking in (6) could be solved with the help of the modification of the
hypothesis about the constraint hierarchy. Nevertheless, the indirect source in (32)
makes (8) (and, of course, (7)) only plausible but not true with certainty. One of its
deficiencies is the circumstance that the first premise and the latent background
assumption relate to only three possible output-forms and to only one input form
of Axininca Campa.

4.3  T
 reatment of Inconsistency No. 3: Modification of the hypothesis
regarding the constraint hierarchy
Inference (32) uses the outcome of Thought Experiment 3 as weak evidence for
the current version of the hypothesised constraint hierarchy in Axininca Campa.10
It is an inference belonging to the third type of plausible inferences (cf. Kertész &
Rákosi this volume: Section 2.3, 2012: Subsection 9.5.4): it is not deductively valid
and most of its premises are only plausible but not true with certainty. It concludes
from the plausibility of a necessary condition of (8) – namely, that the evaluation
procedure based on (8) chooses the output-form belonging to a certain input form
correctly – that this hypothesis regarding the constraint hierarchy in Axininca
Campa is plausible. Nevertheless, Kager’s argumentation presented in Section 3.3

.  For the concept of weak evidence, see Kertész & Rákosi (this volume: Section 3.2,
2012: 178f.).
 Csilla Rákosi

shows that if further input forms are also taken into consideration, p-inconsistency
emerges again.

Thought Experiment 4
Step 1: Simulation of the functioning of GEN
This thought experiment aims at simulating the application of (8) to possible
output-forms related to the input form /osampi/. Therefore, the following three
rival hypotheses are assigned the same low initial plausibility value (cf. (9)(a) in
Section 3.3):

|In Axininca Campa, it is the output-form [osampi] that belongs to the

input /osampi/.|TE = 0.2
4
|In Axininca Campa, it is the output-form [tosampi] that belongs to the
input /osampi/.|TE = 0.2
4
|In Axininca Campa, it is the output-form [sampi] that belongs to the input
/osampi/.|TE = 0.2
4

Step 2: Simulation of the functioning of EVAL 1: Confronting constraints

and candidates
All output-candidates infringe a constraint; thus, every p-context version is
p-inconsistent, as Figure 7 shows.

p-context version1 p-context version2

central hypothesis: central hypothesis:
The output is [osampi] The output is [tosampi]

Onset, Max-IO Dep-IO

p-context version3
central hypothesis:
The output is [sampi]

Figure 7.  The p-context after the application of the Contrastive Strategy in
Thought Experiment 4

Step 3: Simulation of the functioning of EVAL 2: The decision procedure

Since the output-candidate [tosampi] only conflicts with the lowest ranked con-
straint Dep-IO, while [osampi] and [sampi] violate higher ranked c­ onstraints,
 Data and the resolution of inconsistency in Optimality Theory 

the former is the most optimal form among the three rivals examined accord-
ing to (1).

Step 4: Confrontation of the outcome of the evaluation procedure with

“linguistic data”
After the comparison of the outcome of the evaluation procedure with the relevant
data, the p-context in Figure 8 is obtained.

Onset, Max-IO
Dep-IO the output is
[osampi] the output form is
Onset, Max-IO >> Dep-IO
[tosampi]
(1)
[tosampi] is the
most optimal form the output is
the output is [tosampi] [sampi]
(9)(a)
Final state of the decision Surface
procedure by EVAL representation

Figure 8.  The p-context after application of the Combinative Strategy in

Thought Experiment 4

We are in the same situation as with (23)(a) and the output-form [noŋkoma]
in Section 4.2: the p-context version corresponding to the surface representation
is p-inconsistent.
Therefore, as (32) shows, Thought Experiment 4 can be regarded as weak
­evidence11 against the hypothesis that Axininca Campa is characterised by the
constraint hierarchy Onset, Max-IO >> Dep-IO:

(32) |If Axininca Campa is characterised by the constraint hierarchy Onset,

Max-IO >> Dep-IO (= (8)), then the output-candidate [tosampi] is more
optimal than the output-candidates [osampi] and [sampi].|TE = 1
4
[|If the output-candidate [tosampi] is more optimal than the
­output-candidates [osampi] and [sampi], then it is the output-candidate
[tosampi] that is the most optimal among the output-forms belonging to
the input /osampi/.|TE = 0.5]
4

.  For this concept, see in Kertész & Rákosi (this volume: Section 3.2, 2012: 178f.).
 Csilla Rákosi

|If it is the output-candidate [tosampi] that is the most optimal among the
output-forms belonging to the input /osampi/, then it is the form [tosampi]
that belongs to the input /osampi/ at the surface representation.|TE = 0.9
4
|~ It is the output-form [tosampi] that belongs to the input /osampi/ at the
surface representation.|K = 0.9 (cf. (9)(a))
|~ Axininca Campa is characterised by the constraint hierarchy Onset,
Max-IO >> Dep-IO.|I(32) = 0.8 (= ~ (8))

In contrast to this finding, as the inference (31) in Section 4.2 indicates, Thought
Experiment 3 provides weak evidence for (8). As a result, a situation emerges
where one thought experiment provides evidence for the hypothesis that the
constraint hierarchy Onset, Max-IO >> Dep-IO is at work in Axininca Campa,
while another offers evidence against it. This p-inconsistency emerged because
the two thought experiments relied on different “linguistic data”. Since the
­constraint hierarchy should be applicable to all input forms of this language, the
supposed constraint hierarchy should be in conflict with neither data set. In order
to overcome this difficulty, Kager proposes giving up (8). That is, the Contras-
tive Strategy is applied again in order to resolve the conflict between a proposed
constraint hierarchy and the broadened set of “linguistic data”. Accordingly, the
p-context is extended by (11), that is, by a hypothesis of a new constraint, and (8)
is replaced by (12):

(11) 
Align-L: The left edge of the Grammatical Word coincides with the left
edge of the Prosodic Word.

(12) Axininca Campa is characterised by the constraint hierarchy

Align-L, Max-IO >> Onset >> Dep-IO.

This constellation is tested with the help of Thought Experiment 5.

Thought Experiment 5

Step 1: Simulation of the functioning of GEN

Step 1 is identical to Step 1 of Thought Experiment 4; the task is to provide an
explanation for the “linguistic data” for which (8) failed. This occurs with the
help of a new constraint as well as a modified hypothesis regarding the constraint
­hierarchy in Axininca Campa in (12).

Step 2: Simulation of functioning of EVAL 1: Confronting constraints

and candidates
All output-candidates infringe at least one constraint; thus, every p-context ­version
is p-inconsistent, two of them even doubly so, as Figure 9 shows:
 Data and the resolution of inconsistency in Optimality Theory 

p-context version1 p-context version2

central hypothesis: central hypothesis:
The output is [osampi] The output is [tosampi]

Align-L, Max-IO Onset Dep-IO

p-context version3
central hypothesis:
The output is [sampi]

Figure 9.  The p-context after the application of the Contrastive Strategy
in Thought Experiment 5

Step 3: Simulation of the functioning of EVAL 2: The decision procedure

Although the central hypothesis of p-context version 2 that the output is the form
[osampi] is in conflict with only one constraint, from this finding alone it does not
follow that this is the optimal form and the best of the three p-context versions.
(1) requires us to take into consideration the ranking of the constraints, too. Since
the output-candidates [tosampi] and [sampi] violate the highest ranked constraint
Align-IO but [osampi] only conflicts with Dep-IO, the latter is the most optimal
form among the three rivals examined according to (1). Therefore, it will be part
of the surface representation and can be separated from the constraints with the
help of the Combinative Strategy.

Step 4: Confrontation of the outcome of the evaluation procedure with

“linguistic data”
After the comparison of the outcome of the evaluation procedure with the relevant
data, a p-consistent p-context is obtained. See Figure 10.

Align-L, Max-IO
Onset the output is
Dep-IO [tosampi]
Align-L, Max-IO >> the output form is
Onset >> Dep-IO [osampi]
(1)
[osampi] is the
most optimal form the output is
(9)(a)
the output is [osampi] [sampi]

Final state of the decision Surface

procedure by EVAL representation

Figure 10.  The p-context after application of the Combinative Strategy

in Thought Experiment 5
 Csilla Rákosi

The upshot of our considerations in this Section is that Thought Experiment 5

serves as weak evidence for the hypothesis that the constraint hierarchy Align-L,
Max-IO >> Onset >> Dep-IO has to be hypothesised for Axininca Campa:

(33) |If Axininca Campa is characterised by the constraint hierarchy Align-

L, Max-IO >> Onset >> Dep-IO (= (12)), then the output-candidate
­[osampi] is more optimal than the output-candidates [tosampi] and
[sampi].|TE = 1
5
[|If the output-candidate [osampi] is more optimal than the
­output-candidates [tosampi] and [sampi], then it is the output-candidate
[osampi] that is the most optimal among the output-forms belonging to the
input /osampi/.|TE = 0.5]
5
|If it is the output-candidate [osampi] that is the most optimal among the
output-forms belonging to the input /osampi/, then it is the form [osampi]
that belongs to the input /osampi/ at the surface representation.|TE = 0.9
5
|It is the output-form [osampi] that belongs to the input /osampi/ at the
surface representation.|K = 0.9
|Axininca Campa is characterised by the constraint hierarchy Align-L,
Max-IO >> Onset >> Dep-IO.|I(33) = 0.6 (= (12))

The caveat that Thought Experiment 5 made (12) only more plausible but did not
prove its truth is substantial. This indirect source is not perfectly reliable because,
among others, it relies on a limited data set, too, insofar as it confronts the proposed
constraint ranking solely with output-candidates relating to one input form.

4.4  T
 reatment of Inconsistency No. 4: Necessity for the substantial revision
of the theory
The starting p-context contains, among others, the following “linguistic data”:

|In Isthmus Nahuat, it is the output form [tajoːl ]̥ that belongs to the
­underlying form /tajoːl/.|K = 0.9 (= (13))
|In Isthmus Nahuat, it is the output form [tam] that belongs to the
­underlying form /tami/.|K = 0.9 (= (17))
|In Isthmus Nahuat, it is the output form [ʃikɑkíl] that belongs to the
­underlying form /ʃikɑkíli/.|K = 0.9 (= (20))

Datum (13) is weak evidence for hypothesis (14) regarding the ranking of the
­constraints *Voiced-Coda and Ident-IO (voice):

(34) |In Isthmus Nahuat, it is the output form [tajoːl ̥] that belongs to the
­underlying form /tajoːl/.|K = 0.9 (= (13))
 Data and the resolution of inconsistency in Optimality Theory 

|If it is the output form [tajoːl ]̥ that belongs to the underlying form /tajoːl/,
then Isthmus Nahuat is characterised by the constraint hierarchy
*Voiced-Coda >> Ident-IO (voice).|K = 0.3
|Isthmus Nahuat is characterised by the constraint hierarchy
*Voiced-Coda >> Ident-IO (voice).|I(34) = 0.6 (= (14))

Similarly, datum (17) is weak evidence for the constraint hierarchy Final-C >>
Max-IO in (18):
(35) |In Isthmus Nahuat, it is the output form [tam] that belongs to the
­underlying form /tami/.|K = 0.9 (= (17))
|If it is the output form [tam] that belongs to the underlying form /tami/
in Isthmus Nahuat, then Isthmus Nahuat is characterised by the constraint
hierarchy Final-C >> Max-IO.|K = 0.3
|Isthmus Nahuat is characterised by the constraint hierarchy
Final-C >> Max-IO.|I(35) = 0.6 (= (18))

Nevertheless, both inferences (34) and (35) take into consideration only a small
subset of the available data. It is also clear that neither hypothesis (14) nor (18)
are capable of providing an explanation of the data that can be explained by the
other hypothesis. Therefore, they have to be replaced by a hypothesis which unifies
the two constraint hierarchies. In Section 3.4, Kager (1999) was cited who states
that the four relevant constraints cannot be ranked in such a way that the evalua-
tion procedure relying on them would produce the correct outcome. Kager (1999)
illustrates this claim with two thought experiments, based on the rankings in (21)
and (22), respectively:

(21) Isthmus Nahuat is characterised by the constraint hierarchy

Final-C >> Max-IO, *Voiced-Coda >> Ident-IO (voice).
(22) Isthmus Nahuat is characterised by the constraint hierarchy
Final-C >> Max-IO >> Ident-IO (voice) >> *Voiced-Coda.

Thought Experiment 6 tests the applicability of (21).

Thought Experiment 6
Step 1: Simulation of the functioning of GEN
At the beginning of the thought experiment, the following statements are assigned
a low initial plausibility value:
|In Isthmus Nahuat, it is the output form [tam] that belongs to the
­underlying form /tami/.|TE = 0.2
6
|In Isthmus Nahuat, it is the output form [tami] that belongs to the
­underlying form /tami/.|TE = 0.2
6
 Csilla Rákosi

|In Isthmus Nahuat, it is the output form [tajoːl ] that belongs to the
­underlying form /tajoːl/.|TE = 0.2
6
|In Isthmus Nahuat, it is the output form [tajoːl] that belongs to the
­underlying form /tajoːl/.|TE = 0.2
6
|In Isthmus Nahuat, it is the output form [ʃikɑkíli] that belongs to the
­underlying form /ʃikɑkíli/.|TE = 0.2
6
|In Isthmus Nahuat, it is the output form [ʃikɑkíl] that belongs to the
­underlying form /ʃikɑkíli/.|TE = 0.2
6
|In Isthmus Nahuat, it is the output form [ʃikɑkíl ̥] that belongs to the
­underlying form /ʃikɑkíli/.|TE = 0.2
6

Step 2: Simulation of the functioning of EVAL 1: Confronting constraints

and candidates
The four hypotheses capturing constraints result from typological considerations.
Thus, they obtain a relatively high plausibility value.
The evaluation process confronts the supposed constraints with the possible
output forms in the case of each input form. Figure 11a shows the conflicts between
the four constraints and possible output-forms related to the input form /ʃikɑkíli/.

p-context version1 p-context version2

central hypothesis: central hypothesis:
The output is [ʃikkíli] The output is [ʃikkíl]

Final-C Max-IO, *Voiced-Coda Ident-IO(v)

p-context version3
central hypothesis:
The output is [ʃikkíll]

Figure 11a.  The p-context after the application of the Contrastive Strategy in
Thought Experiment 6 I

Figure 11b summarises the conflicts between the constraints and possible
output-forms related to the input form /tami/. Figure 11c displays the conflicts
between the constraints and possible output-forms related to the input form
/tajoːl/.
In all three cases, every p-context version is p-inconsistent because they are in
conflict with at least one constraint.
 Data and the resolution of inconsistency in Optimality Theory 

p-context version1
central hypothesis:
The output is [tam]

Final-C Max-IO, *Voiced-Coda Ident-IO(v)

p-context version2
central hypothesis:
The output is [tami]

Figure 11b.  The p-context after the application of the Contrastive Strategy in
Thought Experiment 6 II

p-context version1
central hypothesis:
The output is [tajoll]

Final-C Max-IO, *Voiced-Coda Ident-IO(v)

p-context version2
central hypothesis:
The output is [tajol]

Figure 11c.  The p-context after the application of the Contrastive Strategy in
Thought Experiment 6 III

Step 3: Simulation of the functioning of EVAL 2: The decision procedure

It is easy to see that the output-candidates [ʃikɑkíl ̥], [tam] and [tajoːl ̥] are the opti-
mal forms along the lines of (1) on the basis of the constraint hierarchy in (21).
Therefore, they will be part of the surface representation and can be separated
from the constraints with the help of the Combinative Strategy.

Step 4: Confrontation of the outcome of the evaluation procedure with

“linguistic data”
The comparison of the outcome of the evaluation procedure with the relevant data
shows that the p-context is burdened with p-inconsistency, because according to
(20), it is the output form [ʃikɑkíl] that belongs to the underlying form /ʃikɑkíli/,
not the form [ʃikɑkíl ̥]. See Figure 12.
 Csilla Rákosi

Final-C, Max-IO
*Voiced-Coda
the output is [ʃikkíl]
Ident-IO (voice)
the output is [tami]
Final-C >> Max-IO, the output forms are
*Voiced-Coda >> Ident-IO (v) [ʃikkíll], [tam] and [tajoll]
(1)
[ʃikkíll], [tam] and [tajoll] are
the most optimal forms
the output is [ʃikkíli]
the output are [ʃikkíll], the output is [tajol]
[tam] and [tajoll] (20)

Final state of the decision Surface

procedure by EVAL representation

Figure 12.  The p-context after application of the Combinative Strategy in

Thought Experiment 6

The thought experiment leads to a p-inconsistent p-context, since the result

of the evaluation procedure is not the correct output form. Therefore, it provides
weak evidence against the proposed constraint hierarchy in the sense of Kertész
and Rákosi (this volume: Section 3.2; 2012: 178):
(36) |If Isthmus Nahuat is characterised by the constraint hierarchy Final-C >>
Max-IO >> Ident-IO (voice) >> *Voiced-Coda (= (21)), then the output-
candidate [ʃikɑkíl ̥] is more optimal than the output-candidates [ʃikɑkíl] and
[ʃikɑkíli].|TE = 1
6
[|If the output-candidate [ʃikɑkíl ̥] is more optimal than the output-­
candidates [ʃikɑkíl] and [ʃikɑkíli], then it is the output-candidate [ʃikɑkíl ̥]
that is the most optimal among the output-forms belonging to the input
/ʃikɑkíli/.|TE = 0.5]
6
|If it is the output-candidate [ʃikɑkíl̥] that is the most optimal among the
output-forms belonging to the input /ʃikɑkíli/, then it is the form [ʃikɑkíl̥]
that belongs to the input /ʃikɑkíli/ at the surface representation.|TE = 0.9
6
|~ It is the output-form [ʃikɑkíl ̥] that belongs to the input /ʃikɑkíli/ at the
surface representation.|K = 0.9 (cf. (20))
|~ Isthmus Nahuat is characterised by the constraint hierarchy Final-C >>
Max-IO >> Ident-IO (voice) >> *Voiced-Coda.|I(36) = 0.8 (= ~ (21))
The next task consists of testing the rival hypothesis (22), that is, finding out
whether the application of the constraint hierarchy Final-C >> Max-IO >>
Ident-IO (voice) >> *Voiced-Coda leads to a consistent solution.
 Data and the resolution of inconsistency in Optimality Theory 

Thought Experiment 7

Step 1: Simulation of the functioning of GEN

This step is identical to Step 1 of Thought Experiment 6.

Step 2: Simulation of the functioning of EVAL 1: Confronting constraints

and candidates
Figures 13a–c show the conflicts between hypotheses on constraints and those
regarding possible output-forms belonging to the input forms /ʃikɑkíli/, /tami/
and /tajoːl/, respectively.

p-context version1 p-context version2

central hypothesis: central hypothesis:
The output is [ʃikkíli] The output is [ʃikkíl]

Final-C Max-IO Ident-IO(v) *Voiced-Coda

p-context version3
central hypothesis:
The output is [ʃikkíll]

Figure 13a.  The p-context after the application of the Contrastive Strategy in
Thought Experiment 7 I

p-context version1
central hypothesis:
The output is [tam]

Final-C Max-IO Ident-IO(v) *Voiced-Coda

p-context version2
central hypothesis:
The output is [tami]

Figure 13b.  The p-context after the application of the Contrastive Strategy in
Thought Experiment 7 II
 Csilla Rákosi

p-context version1
central hypothesis:
The output is [tajoll]

Final-C Max-IO Ident-IO(v) *Voiced-Coda

p-context version2
central hypothesis:
The output is [tajol]

Figure 13c.  The p-context after the application of the Contrastive Strategy in
Thought Experiment 7 III

Step 3: Simulation of the functioning of EVAL 2: The decision procedure

In the sense of (1), it is the output-candidates [ʃikɑkíl], [tam] and [tajoːl] that are
the optimal forms on the basis of the constraint ranking in (22). Therefore, they
are added to the surface representation and become separated from the constraints
with the help of the Combinative Strategy.

Step 4: Confrontation of the outcome of the evaluation procedure with

“linguistic data”
In spite of the change of (21) against (22), the p-context remains p-inconsistent. In
this case, it is the “linguistic datum” (13) which is incompatible with the outcome
of the thought experiment, because it states that it is the output form [tajoːl ̥] that
belongs to the underlying form /tajoːl/. See Figure 14.

Final-C, Max-IO
*Voiced-Coda
the output is [ʃikkíll] the output forms
Ident-IO (voice)
the output is [tami] at the surface are
Final-C >> Max-IO,
[ʃikkíl], [tam] and [tajol]
*Voiced-Coda >> Ident-IO (v)
(1)
[ʃikkíl], [tam] and [tajol] are
the most optimal forms
the output is [ʃikkíli]
the outputs are [ʃikkíl], (13)
the output is [tajoll]
[tam] and [tajoll]

Final state of the decision Surface

procedure by EVAL representation

Figure 14.  The p-context after application of the Combinative Strategy in

Thought Experiment 7
 Data and the resolution of inconsistency in Optimality Theory 

From this it follows that Thought Experiment 7 yields weak evidence against the
proposed constraint hierarchy:
(37) |If Isthmus Nahuat is characterised by the constraint hierarchy Final-C >>
Max-IO >> Ident-IO (voice) >> *Voiced-Coda (= (22)), then the
output-candidate [tajoːl] is more optimal than the output-candidate
[tajoːl ̥]|TE = 1
7
[|If the output-candidate [tajoːl] is more optimal than the output-candidate
[tajoːl ̥], then it is the output-candidate [tajoːl] that is the most optimal
among the output-forms belonging to the input /tajoːl/.|TE = 0.3]
7
|If it is the output-candidate [tajoːl] that is the most optimal among the
output-forms belonging to the input /tajoːl/, then it is the form [tajoːl] that
belongs to the input /tajoːl/ at the surface representation.|TE = 0.9
7
|~ It is the output-form [tajoːl] that belongs to the input /tajoːl/ at the
surface representation.|K = 0.9 (cf. (13))
|~ Isthmus Nahuat is characterised by the constraint hierarchy Final-C >>
Max-IO >> Ident-IO (voice) >> *Voiced-Coda.|I(37) = 0.7
At this point Kager (1999: 377) comes to the conclusion that there is no resolution
of this problem. There is no order of the four relevant constraints that makes pos-
sible an evaluation which is consistent with the “linguistic data”. The main concern
is that the strategy of modifying the hypothesis regarding the constraint hierarchy
and extending it with new elements – which was applied successfully in the ear-
lier cases – is useless in this situation, because constraints work on output forms,
while the difference between (13) and (20) might require the involvement of fur-
ther representation levels, direct access to the input forms, or other types of major
modifications of the theory.

4.5  The solution to (P)(b)

The reconstructions presented in this section motivate the following answer to (P)
(b):
(S) b. The following strategies of inconsistency resolution in OT have been
identified:
– P-inconsistencies emerging from conflicts between constraints are
usually solved by extending the p-context with a hypothesis about
the ranking of the constraints. Such hypotheses are tested with the
help of a thought experiment simulating the evaluation procedure
based on this ranking. Thought experiments of this kind involve the
identification of all conflicts between constraints and possible out-
put-forms and the division of these conflicts into fatal ones on the
 Csilla Rákosi

one hand (conflicts between non-optimal forms and ­constraints)

and tolerable ones on the other hand (conflicts between optimal
forms and constraints). Fatal p-inconsistencies are treated with the
help of the Contrastive and Exclusive Strategies (that is, hypotheses
about the optimal form are given up against linguistic data), while
tolerable ones with the help of the Combinative Strategy (optimal
forms are moved into the surface representation, while constraints
remain part of the evaluation process).
– P-inconsistencies between the outcome of the evaluation ­procedure
applied to a given input form and “linguistic data” are treated
­typically in such a way that the accepted constraint hierarchy is
given up and replaced by an extended version. That is, conflicts
between “linguistic data” and hypotheses about constraint ­rankings
are regarded as fatal. They are not tolerated but the Contrastive
and Exclusive Strategies are applied. The hypothesis about the
constraint hierarchy in the languages examined is given up and is
replaced by a modified (extended) version. The new hypothesis
is then tested on a wider data set containing also those ­“linguistic
data” which were inconsistent with the rejected constraint
­hierarchy.
– P-inconsistencies caused by the limited applicability of a constraint
hierarchy to different input forms are solved in a similar manner:
the constraint hierarchy is modified. If the result of the thought
experiment conducted on the basis of the new constraint ­ranking is
consistent with the “linguistic data”, then a solution of the ­starting
p-inconsistency is achieved. Nevertheless, it is not possible to
­perform a complete test of applicability of ranking hypotheses
because every language contains an infinite number of input forms.
– P-inconsistencies resulting from conflicting constraint ­hierarchies
motivated by different data sets can often be treated with the
extension of the constraint set, too. Nevertheless, it may happen
that there is no appropriate constraint among the set of universal
constraints that could fulfill this task. In such cases, the revision of
some basic assumptions of OT theory may be necessary.

5.  Conclusions

Very important lessons can be drawn from the above solution to (P)(b). First,
there is a connection among the different types of inconsistencies in OT: their
emergence can be traced back to similar causes (overgeneralisation) and they
can in most cases be solved with the help of similar tools, namely, with the joint
 Data and the resolution of inconsistency in Optimality Theory 

a­ pplication of the Combinative and Contrastive Strategies. Second, attempts to

solve an inconsistency belonging to a certain type often leads to the emergence
of inconsistencies of some other type. Thus, for example, it may happen that the
modification of the constraint hierarchy works well for some input forms but fails
with others (which could, in contrast, be grasped smoothly with the help of the
rejected constraint ranking). Third, only a subgroup of inconsistencies is tolerable
in OT: those between optimal output forms and constraints. All other kinds of
conflicts are deemed to be fatal and are usually solved with the help of the minor
or major modification of the theory. Therefore, there seems to be no substantial
difference between “traditional” generative theories and OT in connection with
the treatment of inconsistencies.

Acknowledgements

Work on the present chapter was supported by the MTA-DE Research Group for
Theoretical Linguistics, by the project OTKA K 77823, the TÁMOP-4.2.2/B-10
/1-2010-0024 project, which is co-financed by the European Union and the
­European Social Fund, as well as by the János Bolyai Research Scholarship of the
Hungarian Academy of Sciences.

References

Archangeli, Diana. 1997. Optimality Theory: An introduction to linguistics in the 1990s. In

Optimality Theory. An Overview, Diana Archangeli & D. Terence Langendoen (eds), 1–33.
Oxford: Blackwell.
Kager, René. 1999. Optimality Theory. Cambrige: CUP.
Kertész, András & Rákosi, Csilla. 2012. Data and Evidence in Linguistics. A Plausible Argumenta-
tion Model. Cambridge: CUP.
Kertész, András & Rákosi, Csilla. 2013. Paraconsistency and plausible argumentation in genera-
tive grammar: A case study. Journal of Logic, Language and Information 22(2): 195–230.
Kertész, András & Rákosi, Csilla. This volume. The p-model of data and evidence in linguistics.
 chapter 11

Conclusions

András Kertész1,2 & Csilla Rákosi2

1Universityof Debrecen, Institute of German Studies,
2MTA-DE Research Group for Theoretical Linguistics

1.  Introductory remark

In Chapter 1, Section 2, we mentioned five key features of the p-model that may
be decisive for a workable metascientific solution to the problem of linguistic data
and evidence. Accordingly, the task of this concluding chapter is to highlight some
aspects of the case studies that seem to be especially instructive with respect to
these features, without, however, striving for a comprehensive summary of their
findings. We will put special emphasis on the question of whether, and if so, in
what way and to what extent, the application of the p-model can make linguistic
problem solving more effective.

2.  The uncertainty of linguistic data

The acknowledgement of the uncertainty of data played a key role in all case stud-
ies. The conscious reflection on the plausibility of the data meant above all that the
sources of the plausibility value of the data were revealed and in most cases, their
reliability was re-evaluated. A second aspect of this issue was that the authors were
aware of the circumstance that they can build only plausible inferences on the data,
which were plausible but not certainly true, whenever they elaborated on their own
approaches. As a result of this attitude, they were more open to possible counter-
arguments and more critical against the weak points of their own theories.
The significance of the reconstruction of the data structure is best exemplified
in Chapter 5, Section 2 by Enikő Németh T. She has shown that a careful analysis of
the data structure by means of the p-model may provide convincing arguments for
the rejection of theories relying on the data types at issue. First, the re-evaluation
of the reliability of the sources supporting the data of previous theories revealed
that a subset of the data set lost their data status because their sources turned out
 András Kertész & Csilla Rákosi

to be unreliable or at least not reliable enough. Second, in certain cases the exami-
nation of the data structure witnessed that the basic concepts of the theory were
obscure and could not be applied to the description of the relevant data. She also
presented an instructive example of how fruitful it is to rely on as many data types
as possible because this strategy considerably reduced the uncertainty and the fal-
libility of the individual data types (such as the linguist’s own linguistic intuition)
and led to more plausible findings.

3.  Plausible inferences

Similarly to the use of the concept of ‘plausible statement’, the application of the
concept of ‘plausible inference’ resulted in a twofold benefit. First, as a tool of
metascientific analysis of linguistic argumentation, the reliance on plausible infer-
ences made it possible to reconstruct the inferences used in linguistic theories. The
most significant result was in this respect that a considerable amount of inferences
in linguistic theories is deductively invalid but they can be interpreted as plausible
inferences capable of providing a partial support to their conclusions. Accord-
ingly, in several cases the analyses led to the conclusion that the hypotheses of the
investigated theories were neither proven nor falsified by linguistic evidence but
they were made plausible or implausible to some extent. Moreover, the examina-
tion of the structure of the plausible inferences shed light on the latent background
assumptions that often turned out to be implausible on closer examination.
A second gain of the use of this notion appeared at the level of object-­scientific
argumentation. That is, the authors of the volume presented and scrutinized the
plausible inferences they made use of when proposing a new approach to the
investigated topic.
In this respect, Zoltán Vecsey’s contribution in Chapter 8 may be especially
instructive. Although, as we mentioned in Chapter 2, plausible inferences are of
informal nature, he has shown the usefulness of the p-model by its application to a
subfield of linguistics in which formal methods are preferred. Namely, he showed
how plausibility considerations enable us to solve a problem in formal semantics
that could not be solved with the help of logic alone.

4.  The cyclic and prismatic nature of linguistic theorising

The idea of the cyclic and prismatic nature of linguistic theorising seems to be
highly effective in linguistic problem solving because it impelled the authors of the
case studies to more thorough, more comprehensive investigations. If one does
 Conclusions 

not think that linguistic theorising is a linear process but supposes that it is cyclic,
then it is more likely and completely natural that one turns back to the investigated
problems many times from different points of view and considers many possible
solutions.
For example, Katalin Nagy C.’s contribution in Chapter 4 exemplifies the effec-
tiveness of the re-evaluation of the argumentation applied by rival theories. She
re-evaluated several components of the investigated theories: the reliability of the
sources supporting their data, the plausibility of their hypotheses, the acceptability
of the applied methods, the consistency between data and hypotheses. Since she
focused on different components of the p-context, her argumentation was pris-
matic: the revision of the data set or the implementation of methods that are well-
established but were not applied to this research topic shed new light again and
again on the investigated phenomena. The cyclic re-evaluation also comprised the
introduction of further data types and methods into the p-context.
Helga Vanda Koczogh applied in Chapter 6 the techniques of plausible argu-
mentation in the early phase of the linguistic theorising process. Namely, she tried to
solve a problem that is more basic than p-inconsistency or p-incompleteness, which
are related to the plausibility values of hypotheses. She showed how the category
system of a theory can be developed with the help of a cyclic and prismatic argu-
mentation process. She continuously re-evaluated the proposed categorisations and
tried to elaborate a more refined, more clearly-shaped and more comprehensive sys-
tem. To this end, she applied several problem-solving strategies that are similar to
those connected to the resolution of p-inconsistency. For example, the application of
the Diagnostic Perspective may mean in this context that one examines the linguis-
tic phenomenon at issue, seeking for instances whose treatment lead to problems
with the existing category systems. The Contrastive Strategy/Perspective compares
rival category systems in order to come to a decision among them. The Combinative
Strategy/Perspective, in contrast, aims at the integration of several category systems.
This goes beyond putting them simply together. Rather, it requires the selection of
their useable elements and the construction of a novel, more comprehensive model
from them. The evaluative perspective means that the rival category systems are
compared with each other on the basis of a series of criteria.

5.  The treatment of inconsistencies

The p-model provides a differentiated set of tools that facilitate the treatment of
inconsistencies, allowing for the rejection, the temporal or permanent tolerance of
contradictions. The authors of the volume revealed several instantiations of every
strategy and witnessed a high variety of their applications.
 András Kertész & Csilla Rákosi

In Chapter 3, Károly Bibok discussed two rival approaches to syntactic alterna-

tions neither of which was satisfactory because there were linguistic data providing
weak/strong evidence against them. Since, however, both lexical and construc-
tional factors seem to play a role in syntactic alternations, the author concluded
that the p-inconsistencies emerging from the conflicts between the hypotheses of
the two approaches might be best solved with the help of the ­Combinative Strategy.
That is, instead of striving for a decision between the two rivals, he proposed an
approach that systematically unites them. Thereby, the explanatory power of both
approaches was also accumulated by the author’s theory. Thus, Bibok’s approach is
considerably more complex and more comprehensive than the two rivals.
In Chapter 5, Enikő Németh T. applied the Contrastive Strategy and rejected
all existing rivals because they were all p-inconsistent with a series of data that had
been neglected earlier. Nevertheless, the use of this strategy was not destructive
because during the argumentation cycles pieces of information came to light that
motivated the basic idea of a more sophisticated theory. Namely, besides being
p-consistent with the data set, the proposed lexical-semantic representation also
made it possible to describe occurrences of the verbs of natural phenomena both
with and without a subject in a unified way. This is in harmony with the method-
ological principle of simplicity; this finding makes the results even robuster.
In Chapter 7, György Rákosi addressed a particularly interesting and instruc-
tive case of p-inconsistency. He investigated a situation in which there were con-
flicting grammaticality judgements. Usually such cases are dealt with in such a way
that either one group of the judgements is declared to be faulty or both groups of
judgements are kept and deemed to be instances of dialectal or idiolectal variation.
There is, however, a third possibility: the construction at issue may be interpreted
in certain cases as marked. In this way, one obtains a paraconsistent solution of
the p-inconsistency. Nevertheless, there are strict prerequisites for doing so: indi-
vidual speakers either have to accept or reject the given construction and, at the
same time, their acceptance or rejection, respectively, has to be less definite than
in “clear” cases. This means that they may hesitate between outright acceptance
and finding the given linguistic items acceptable but somewhat unusual, or, in the
negative case, between dispreference and strict rejection.
Csilla Rákosi’s Chapter 10 revealed the most typical techniques of incon-
sistency resolution in Optimality Theory. The most important lesson of her
reconstructions is that OT has not brought a radical change in the treatment of
inconsistencies. The reason for this finding is that besides tolerable inconsisten-
cies, there is a huge amount of contradictions in OT that are similar to those in
other areas in linguistics such as in “traditional” generative grammar. The author’s
analyses identified the causes of the emergence of the inconsistencies in OT, the
most frequent being overgeneralisation.
 Conclusions 

6.  The pluralism of linguistic theorising

The contributions of the volume provide convincing arguments for a pluralistic

view in linguistics. First, they do not straight reject the approaches they criticised
but they strive to keep their useable elements and explanatory power. In several
cases, the new proposal results from the combination and development of two or
more former rivals. Second, they do not deem the theories they propose to be true
with certainty but only a more plausible alternative on the basis of the information
available which also may be succeeded by a better theory in future.
With the help of the analysis of the argumentation structure of a thought exper-
imental report and a real experimental report, in Chapter 9, Kertész and Rákosi
provided arguments for the re-evaluation of the relationship between thought
experiments and real experiments in pragmatics. Instead of giving up the former
as outdated and unreliable data sources, the authors argued that thought experi-
ments and real experiments are kindred. They share several features, have a similar
structure but there are also significant differences between them. This finding may
pave the way for their joint application in pragmatics at a new, more sophisticated
level, contributing to the pluralism of research methods in linguistics.

Acknowledgements

Work on the present chapter was supported by the MTA-DE Research Group
for Theoretical Linguistics, the project OTKA K 77823, the TÁMOP-4.2.2/B-10
/1-2010-0024 project, which is co-financed by the European Union and the E
­ uropean
Social Fund, as well as the János Bolyai Research Scholarship of the ­Hungarian
Academy of Sciences.
 Author index
A D
Abraham, Werner  181
Ackerman, Farrell  180, 190
Aijmer, Karin  147
Alvar, Manuel  77
Anand, Pranav  208–213, 215
Archangeli, Diana  271–273
Ashley, Aaron  242–262, 266
Austin, John L.  221, 242
B É. Kiss, Katalin  118, 180, 184,
Badia i Margarit, Antoni M.  80
Ball, Derek  211
Baltin, Mark R.  25–26
Bándli, Judit  138–140, 142,
145–146, 151, 157–158
Bárczi, Géza  55
Bartos, Huba  118
Bibok, Károly  64–66, 69, 115,
117–118, 121, 124
Bierwisch, Manfred  64, 124
Blum-Kulka, Shoshana  167
Borsley, Robert D.  2
Bousfield, Derek  154
Brandom, Robert  29–30
Braun, Angelika  154
Brown, James Robert  225
Brown, Penelope  136–137, 140,
144–146, 153–154, 167
Büring, Daniel  211
Bybee, Joan  74–76, 96–98
Bynon, Theodora  76
C Groot, Casper de  117
Cable, Seth  203
Cardinaletti, Anna  193–194
Cohen, Philip R.  243, 253
Collins, Harry M.  264
Colon, Germà  80–84,
90–93, 98
Cote, Sharon A.  117
Culpeper, Jonathan  153–154,
167
Dér, Csilla Ilona  74
Desclot, Bernat  84, 94–95
Detges, Ulrich  80, 83–86, 90,
93, 98
Dik, Simon C.  117
Du Bois, John W.  144
Duhem, Pierre  225
E Jeffers, Robert J.  76
186–190, 192, 194–195
F Kager, René  272–305
Fehige, Yiftach  225
Fischer, Olga  73, 99
Fitzmaurice, Susan  74, 99
Fludernik, Monika  91–92
Fónagy, Iván  154
G Kepser, Stephan  2–3
García Velasco, Daniel  117,
119, 124
Gibbs, Raymond W.  243
Goldberg, Adele E.  59–61, 64,
118–119
Gooding, David  226
Gougenheim, Georges  86
Grice, Herbert Paul  151,
154, 221, 227–232, 237–238,
242, 265
Gries, Stefan Th.  2
Groefsema, Marjolein  117
H Komlósi, Boglárka  154
Hegedűs, Veronika  180, 185,
188
Heim, Irene  211
Heine, Bernd  96
Hock, Hans Henrich  76, 89,
94, 96
Holmes, Janet  148
Holtgraves, Thomas  145,
242–262, 266
I
Iwata, Seizi  65
J
Jackendoff, Ray  124
Jaszczolt, Katarzyna M.  69
Juge, Matthew L.  86–94, 97–99
K
Kallmeyer, Werner  2
Kálmán, László  180, 184
Kaplan, David  199–207,
210–212, 215
Katz, Albert N.  154
Kenesei, István  180, 184, 188
Kertész, András  1–5, 16–17, 22,
25, 34, 51, 53–54, 56, 58, 67–69,
71–73, 86, 90, 96, 103–105,
107–108, 111, 113, 116, 118, 124,
130, 140, 181, 185, 191, 196,
206–207, 212, 223–224, 226,
259, 266, 280–282, 284–285,
288, 292–293, 295, 302
Keszler, Borbála  104–105,
107–108, 113–115, 122
Keysar, Boaz  259
Kiefer, Ferenc  224
Kocsány, Piroska  118
Koczogh, Helga Vanda  134, 141
Komlósy, András  63, 104–105,
109–111, 113–114, 116, 118–119
Kotthoff, Helga  173, 175
Kratzer, Angelika  211
Kristiansen, Gitte  2
Kühne, Ulrich  225
Kuteva, Tania  96
 Author index
L P
Laczkó, Tibor  180, 190
Ladányi, Mária  74
Leech, Geoffrey  154
Lehiste, Ilse  76
Lehmann, Christian  2
LeSourd, Philip  180
Levesque, Hector J.  243, 253
Levinson, Stephen  136–137,
140, 144–146, 153–154, 167
Locher, Miriam  138, 142,
145–149, 151–152, 167
M Rákosi, Csilla  1–5, 16–17, 22,
Magdics, Klára  154
Magnani, Lorenzo  212
Maleczki, Márta  121
Markó, Alexandra  149
Martí, Luisa  118, 121
Mászlainé Nagy, Judit  140, 151
Meggle, Georg  242
Meibauer, Jörg  223
Moue, Aspasia S.  225
Muntaner, Ramón  84, 92,
94–95
Muntigl, Peter  135–137, 142,
157, 167
N 22, 29–30, 32, 37–38, 104,
Nagy C., Katalin  98, 104
Nauke, Astrid  154
Navarro, Federico  99
Németh T., Enikő  104, 113,
117–118, 124, 142, 181
Nersessian, Nancy J.  225–226,
265
Nevins, Andrew  208–213, 215
Noonan, Máire  181
Norton, John D.  225–226
Noveck, Ira A.  222–223
O 254–258, 263–265
Országh, László  55
Penke, Martina  2–3
Pinker, Steven  55–56, 58–59
Pinkster, Harm  92
Polya, George  16, 22
Portero Muñoz, Carmen  117,
119, 124
Pottier, Bernard  77
Pustejovsky, James  124
Pusztai, Ferenc  55, 106, 113
R 186–190
Radford, Andrew  118
25, 53–54, 56, 58, 67–69,
71–73, 86, 96, 103–105,
107–108, 111, 113, 116, 118, 124,
130, 140, 181, 185, 191, 196,
206–207, 223–224, 226–227,
259, 280–282, 284–285, 288,
292–293, 295, 302
Rákosi, György  180, 190
Reboul, Anne  223
Recanati, Francois  118, 121
Rees-Miller, Janie  135–138, 142,
145–152, 155, 157, 168, 172
Reis, Marga  2–3
Rescher, Nicholas  16–18,
205, 212
Rice, Sally  119
Rosenbach, Anette  2–3
S W
Sárik, Pál  112
Schegloff, Emanuel A.  253
Schiffrin, Deborah  91–92
Schlenker, Philippe  203,
207–212, 215
Schütze, Carson T.  2
Searle, John R.  221, 227– 243,
Seuren, Pieter A.M.  1
Sharvit, Yael  214
Sidelle, Alan  214
Silva-Corvalán, Carmen  92
Speas, Margaret  210
Sperber, Dan  124, 161, 222–223
Starke, Michal  193–194
Stechow, Arnim von  210–211,
213–215
Stefanowitsch, Anatol  2, 57
Sternefeld, Wolfgang  2
Surányi, Balázs  180, 184,
Svindt, Veronika  154
T
Taavitsainen, Irma  74, 99
Tannen, Deborah  144
Taylor, Kenneth  118
Toplak, Maggie  154
Tóth, Ildikó  104–105, 109,
111–114, 116
Trón, Viktor  180, 184
Turnbull, William  135–137,
142, 157, 167
U
Ulkan, Maria  242
Ürögdi, Barbara  180, 190
V
Vanderveken, Daniel  242
Varga, László  142
Walton, Douglas  235
Webelhuth, Gert  180
Wilson, Deirdre  124, 161
Wittgenstein, Ludwig  229,
235–238
Wolfson, Nessa  91
Z
Zifonun, Gisela  2
 Subject index
A 255, 257–258, 261–262,
agreeing RPVC see
reduplicating particle
construction (RPVC),
agreeing
ALIGN-L  276–277, 296, 298
alternation
locative  51, 54–55, 58–59, 61,
65–69
syntactic  51–56, 60–64, 67,
69, 118, 121, 129
present-past  80, 83–84,
91–92, 98
present-preterit  79, 92–93
analogy  21, 73, 76–79, 81,
86–89, 93, 95–99, 183,
257–258, 263–266
analogical change  76, 89,
94, 96, 98
analogical processes  character  200–204, 207, 209,
76–78, 90
analogy-based
argument(ation)
see argumentation,
analogy-based
“anar ‘go’ + infinitive”
construction  78–99
appositive  189–190, 194–196
argumentation
analogy-based  78, 88, 99
cycle  32–36, 46, 105, 118, 124,
134, 141, 160, 175, 237,
254, 267
cyclic nature of  32, 35–45,
161, 176, 182, 221, 227,
256–258, 261, 266,
310–311
frequency-based  78, 88
in historical linguistics  contextual parameters see
71–73
plausible  32–34, 45–46, 54,
71–73, 78, 95–96, 104,
133–134, 176, 181–182,
195, 205, 224–228,
300–301, 303–304,
265–267, 310–313 306–307, 311–312
prismatic nature of  32, 34, corpus  40, 45, 73, 75–76, 87,
36, 45–46, 256, 310–311 89–90, 99, 115, 133–134,
attitude 141–142, 166
holder  209, 211, 213–214 corpus data see data, corpus
verb  208–210
Axininca Campa  274–277, D
280–285, 287, 289–296, data 
298 concept of  37–40, 44–46,
53–54, 72–76, 79–80,
B 82–83, 85–90, 94–96,
booster  138, 148, 155 99, 107–116, 128–130,
207–210, 215, 221–224,
C 245–254, 256, 258–262,
Catalan  72–73, 78–83, 86–87, 264–267, 274, 279–283,
91, 97–99 286–293, 296, 298, 306,
medieval  80, 83, 92, 94–95 309–313
Old  82, 86–88 corpus  73, 82, 87
experiential  264–266
214 experimental  223, 245–248,
circumstantial parameters 250, 252–254, 258,
see parameter, 260–262, 264–265, 267
circumstantial gained from historical
Combinative Strategy/ corpora  73
Perspective  35–36, 141, perceptual  245–250,
196, 285–286, 288, 290, 259–261, 264
292, 295, 297, 301–302, authentication of 
304–307, 311–312 245–250, 259–262, 264
constructional meaning interpretation of 
see meaning, 245–246, 248–250,
constructional 259–262, 264
context of occurrence  200, 203 qualitative  73–74
context-shift  203–205, quantitative  74–76, 99
208–215 thought experimental  258,
context-shifting operator see 265, 267
operator, context- DEP-IO  274–277, 280–285,
shifting 287–288, 290–298
direct source see source, direct
parameter, contextual disagreement strategies 
Contrastive Strategy/ 138–140, 142–149,
Perspective  35–36, 135, 151–160, 164, 167–170,
237, 253, 284, 287, 289, 172–176
291–292, 294, 296–297, double indexing  201, 204
 Subject index
E H
EVAL  272, 283–284, 286, 289,
291–292, 294, 296–297,
300–301, 303–304
evidence  41–46, 67–69, 85, 93,
115, 117, 191–192, 194,
205, 223, 256, 283, 310
relative  42–43, 67, 69
strong  43–44, 67, 111–112,
130, 194, 253, 261, 312
weak  41–44, 67, 231, 238,
244, 253, 261, 282, 293,
295–296, 298–299, 302,
305, 310, 312
Exclusive Strategy  35–36, 228,
234, 238, 292, 306
experiential data see data,
experiential
experimental data see data,
experimental
experimental design  244, 248,
251, 255–256, 259–263,
267
experimental procedure  227,
244–246, 250, 256–264,
267
experimental report  227–229,
245, 255, 257–263,
265–267
explicit
argument  118
subject argument  104, 108,
121, 127, 129
subject  104–111, 113–116,
119–120, 122–123, 125,
128–129
F 159–160, 175, 243
FINAL-C  278, 299, 302, 305
French, Old  86
frequency  72–76, 78, 83, 85,
87–90, 94, 98–99, 161
analysis  73, 76, 85, 87
counting  74–76, 99
token  75–76  84, 90, 190, 232, 243,
type  75
G lexical meaning see meaning,
GEN  272, 283, 289, 291, 294,
296, 299, 303
GO-periphrasis  80–81,
83–84, 98
grammaticalisation  72, 74–75,
78, 82, 86, 92, 94–95
locative alternation see
hedge  138, 143–144, 147–148, alternation, locative
162, 164, 167, 170–171,
174 M
historical documents  73, 83, manner of motion  52–53,
85, 93 63, 69
historical linguistics  71–76, 78, MAX-IO  275–278, 291–293,
85, 90, 96–99 295–296, 298–299,
historical present see present, 302, 305
historical meaning
Hungarian  51–52, 56, 58–59, constructional  64–67, 69
68–69, 103–119, 127, lexical  90–91, 94–95
133–134, 138–141, 149, non-natural  227–228
179–183, 188–189, underspecified  65–66
195–196, 204 metalinguistic sources see
source, metalinguistic
I
IDENT-IO (voice)  277–278, N
298–299, 302, 305 narrative present see present,
implausible see statement, narrative
implausible negative politeness see
implicit politeness, negative
argument  105, 116–118 neutral plausibility see
subject  120, 125, 129 statement of neutral
subject argument  104, plausibility
118–123, 125–130 non-agreeing RPVC see
indexicals  199–201, 203–204, reduplicating particle
207–210, 212–215 construction (RPVC),
indirect source see source, non-agreeing
indirect non-natural meaning see
informational meaning, non-natural
overdetermination 
29–30, 93, 104–105, 134, O
140–141, 181, 212 oblique  179–180, 182, 187–192
informational ONSET  274–277, 280–285,
underdetermination  287, 290–293, 295–296,
29–30, 140–141, 298
operator 
Isthmus Nahuat  277–278, context-shifting  204–205,
298–300, 302, 305 208–209, 213
sentential  199, 202–203,
L 210
latent background
assumption  22–29, 82, P
paraconsistent solution
248, 251, 261, 288, 291, see p-problem,
293, 310 paraconsistent
solution of
lexical partial agreement  136–139,
lexical-semantic representation 143–145, 162, 168, 170,
see representation, 174–175
lexical-semantic particle  76–77, 135, 179–187,
location parameter see 189–196
parameter, location verbal  179–180, 192
 Subject index 

p-context  plausible argumentation quantitative data see data,

concept of  26–36, 38–40, see argumentation, quantitative
51, 56–60, 72–73, 78, plausible
80, 104–105, 134–135, plausible inference  16, 20–29, R
182–183, 187–189, 228, 45, 56, 59, 68–69, 96, real experiment  221–224,
232, 242–244, 267, 103, 185, 187, 210,–212, 226–227, 242–244,
279–280, 284–288 230–232, 236–239, 252–259, 267, 313
coordination of  32–34, 93, 246–247, 250–252, 257, reduplicating particle
105, 116, 166, 182, 234, 261, 267, 281–282, 293, construction
246 309–310 (RPVC)  179, 185,
elaboration of  32–36, 58, plausible statement see 195–196
141, 161, 182–183, 190, statement, plausible agreeing  180–182, 185–190,
195, 212, 228, 241, 243 politeness  119, 136–138, 140, 194–196
extension of  32–34, 56–57, 146, 148, 154 non-agreeing  180, 190,
63, 80, 86, 90, 93, 105, positive  137, 146 194–196
113–114, 116, 134, 161, negative  137, 148 relative evidence see evidence,
182, 215, 228, 232–233, p-problem  relative
235, 243, 291, 305 concept of  30–36, 45, 105, representation
modification of  58, 95, 119, 128, 130, 181, 195, 238 lexical-semantic  61, 64,
246 resolution of  30–35, 105, 68, 104–105, 109–111,
p-problematic  30, 32, 35, 130, 134, 175–176, 196, 116–117, 119–121,
104, 113, 254 212, 215, 253, 284, 288, 123–126, 128–129
p-incomplete  29–30, 113, 305 semantically
129, 311 solution of  30–36, 95, 124, underspecified  65,
p-inconsistent  29–33, 35, 39, 138, 140–142, 159, 167, 67, 69
57, 71–73, 80, 83–85, 93, 173, 175, 188, 193, 195, resolution of a p-problem
95, 97–98, 105, 111–112, 212, 238, 240, 242, 288, see p-problem,
116, 129, 181, 188–189, 306, 311 resolution of
195, 228, 232, 234, consistent  30, 215, 302 retrospective re-evaluation  6,
237–238, 282–296, paraconsistent  30, 36, 191, 32, 45, 161, 227, 242, 267
300–302, 304, 306, 196, 312
311–312 pragmatic constraint see S
perceptual data see data, constraint, pragmatic selection restriction  119–126,
perceptual pragmatic force modifier 128–129
perfective past  72, 77–81, 98 (PFM)  133–134, 167, sentential operator see
PFM see pragmatic force 170–173, 175–176 operator, sentential
modifier present, solution of a p-problem see
p-incompleteness see historical  80–84, 86, 89–95, p-problem, solution of
p-context, 99 source concept of  17–23,
p-incomplete  narrative  80–81 25–34, 38–41, 44–46,
p-inconsistency see p-context, preterit  79–81, 83–89, 92–95, 53–54, 56–60, 62–63,
p-inconsistent 97, 99 72–77, 85–86, 99, 107,
plausibility (value)  17–23, periphrastic  87 109, 114–115, 185–186,
25–33, 35, 37–43, 45–46, problem-solving strategies  35, 189, 191, 206–212,
54, 58, 61, 68, 72–73, 311 215, 221, 228, 235, 243,
75, 78, 80, 82–87, pro-drop  183, 194 246, 251–252, 255, 261,
89–91, 93, 95, 97–99, pronoun  266–267, 280–281, 290,
109, 114–115, 140, 175, strong  193–194 309, 311, 313
185–186, 188, 194, weak  193–194 direct  20, 23, 27, 37, 45,
206–208, 210, 215, 228, Provençal  81 53–54, 81, 83–86,
237, 243–244, 248, 104, 110–112, 114, 141,
251–257, 260, 264–267, Q 207–208, 228, 230–231,
280–283, 285–289, qualitative data see data, 242, 255, 259–260,
309–311 qualitative 265, 280
 Subject index

indirect  20, 22, 25–27, 29, 195, 259–260, 264–265, time parameter see parameter,
41, 90, 231, 233, 287, 280, 310 time
293, 298 strong evidence see evidence, token frequency see frequency,
metalinguistic  74 strong token
typological  74 subject argument type frequency see frequency,
Spanish  76–77, 92 position  107–108, 116, type
medieval  76–77 119, 128–129
standard view subjectless verbs  104–116, U
of linguistic data and 119, 128 universal
evidence  2–5, 44, 223 suitable sources see source, linguistic  88, 96, 276, 306
of the analytical philosophy suitable tendencies  73, 88,
of science  2–3, 16, 225 syntactic alternation see 95–97, 99
statement alternation, syntactic
implausible  17–18, 20, V
22–23, 26–27, 29–30, T verbal disagreement  133–134,
38, 41, 54, 85, 93, 98, thought experiment  221–229, 141, 176
111–112, 114, 228, 238, 235–238, 241–243, verbal particle see particle,
248, 261, 290, 310 254–259, 263–267, verbal
of neutral plausibility  18–19, 283–284, 286–306, 313 *VOICED-CODA  277–278,
22–23, 25, 28, 30, 38, thought experimental data 298–299,
206, 248 see data, thought 302, 305
plausible  17, 21–22, 29, experimental
32, 38, 40–41, 46, thought experimental W
53–56, 67–68, 73–74, report  227–229, 255, weak evidence see evidence,
78, 85, 106, 110, 112, 258, 263, 265, 313 weak