Pragmatics at its Interfaces 9781501505089, 9781501514302

Table of contents :
Contents
1. Introduction
I. Pragmatics and philosophy
2. Back to negative particulars: A truth-conditional pragmatic account
3. Colour adjectives, compositionality, and true utterances
4. Talk about the future
5. Illocutionary effects, presupposition, and implicature
II Pragmatics and cognition
6. Is implicit communication a way to escape epistemic vigilance?
7. The structure of utterance meaning: how did it come about?
8. The role of Theory of Mind, grammatical competence and metapragmatic awareness in irony comprehension
III Pragmatics and linguistic analysis
9. Theoretical and methodological issues in the research into implicit arguments in Hungarian
10. The Pragmatics of ‘Yes’ and ‘No’
11. Do not say that for which you lack adequate evidence: On the interpretation of future tense in Modern Greek
12. Redundancy effects in discourse: On the modal particle-combinations ‘halt eben’ and ‘eben halt’ in German
IV. Conversation analysis
13 The pragmatics of intelligible pronunciation: Preemptive and reactive segmental repair in English as a lingua franca interactions in Japan
14 The interactional functions of four repair operations in Hungarian
Index

Citation preview

Stavros Assimakopoulos (Ed.) Pragmatics at its Interfaces

Mouton Series in Pragmatics

Editor Istvan Kecskes Editorial Board Reinhard Blutner (Universiteit van Amsterdam) N.J. Enfield (Max-Planck-Institute for Psycholinguistics) Raymond W. Gibbs (University of California, Santa Cruz) Laurence R. Horn (Yale University) Boaz Keysar (University of Chicago) Ferenc Kiefer (Hungarian Academy of Sciences) Lluís Payrató (University of Barcelona) François Recanati (Institut Jean-Nicod) John Searle (University of California, Berkeley) Deirdre Wilson (University College London)

Volume 17

Pragmatics at its Interfaces Edited by Stavros Assimakopoulos

ISBN 978-1-5015-1430-2 e-ISBN (PDF) 978-1-5015-0508-9 e-ISBN (EPUB) 978-1-5015-0502-7 ISSN 1864-6409 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.dnb.de. 6 2017 Walter de Gruyter Inc., Boston/Berlin Typesetting: RoyalStandard, Hong Kong Printing and binding: CPI books GmbH, Leck ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

Contents

1

Stavros Assimakopoulos 1 Introduction

I

Pragmatics and philosophy

2

Jacques Moeschler Back to negative particulars: A truth-conditional pragmatic account

3

Richard Vallée Colour adjectives, compositionality, and true utterances

4

Kepa Korta Talk about the future

5

Etsuko Oishi Illocutionary effects, presupposition, and implicature

9

33

53

71

II Pragmatics and cognition 6

Anne Reboul Is implicit communication a way to escape epistemic vigilance?

7

Thanh Nyan The structure of utterance meaning: how did it come about?

8

91

113

Márta Szücs and Anna Babarczy The role of Theory of Mind, grammatical competence and metapragmatic 129 awareness in irony comprehension

III Pragmatics and linguistic analysis 9

Enikő Németh T. Theoretical and methodological issues in the research into implicit 151 arguments in Hungarian

vi

Contents

Thorstein Fretheim 10 The Pragmatics of ‘Yes’ and ‘No’

11

175

Michael Chiou Do not say that for which you lack adequate evidence: On the 201 interpretation of future tense in Modern Greek

Sonja Müller 12 Redundancy effects in discourse: On the modal particle-combinations 225 ‘halt eben’ and ‘eben halt’ in German

IV Conversation analysis George O’Neal 13 The pragmatics of intelligible pronunciation: Preemptive and reactive segmental repair in English as a lingua franca interactions in 257 Japan Zsuzsanna Németh 14 The interactional functions of four repair operations in Hungarian Index

311

279

Stavros Assimakopoulos

1 Introduction The present volume is a collection of selected papers from the 6th International Conference on Intercultural Pragmatics and Communication that was held at the University of Malta in June 2014. Forming part of the theoretical session of the conference, the papers selected for this volume report on recent, cutting-edge research within the area of linguistic pragmatics, broadly construed. Apart from the obvious aims of motivating further discussion on topics of central importance for pragmatics, however, another main objective of this volume is to show how research in this field can and does substantially inform research in various related areas of scholarly interest. In order to better understand the diverse array of topics that the study of meaning in context carries implications for, a brief historical overview of the field of pragmatics is in order. The person that is credited with introducing pragmatics in the modern academic plateau is Morris, who, back in the 1930s, defined pragmatics as the “study of the relation of signs to interpreters”, which along with syntactics that studies “the formal relation of signs to one another” and semantics that studies “the relations of signs to the objects to which the signs are applicable” outlines the shape of the theory of signs, i.e. semiotics (1938:6–7). In the years following Morris’ distinction, however, the subjectmatter of pragmatics proved notoriously difficult to practically delineate. A possible reason for this could have been the overwhelming fascination with formal semantic theory, which left pragmatics at a relative standstill, gradually leading it to be considered a “wastebasket” of linguistic information, with linguists resorting to it only when they needed to address phenomena that they could not explain within the remit of semantics. For a while, this left the field inherently unstructured with respect to its specific goals and subject-matter; Acknowledgement: I would like to express my deepest gratitude to all the contributors for entrusting me with their work, for their patience during an unexpectedly eventful reviewing process, and for their friendly attitude and smooth collaboration during the final stages of putting the volume together. I would also like to thank Istvan Kecskes, the general editor of the Mouton Series in Pragmatics, for his sound advice and continuous guidance, as well as Polychronis Chatzoglou and Rebecca Vella Muscat for helping me out with numerous practicalities in the process. Naturally, this volume would never have reached its present form, if it were not for the careful reviewing of all the originally submitted papers by several colleagues, many of whom were willing to jump in and provide reports at very short notice. Stavros Assimakopoulos, University of Malta DOI 10.1515/9781501505089-001

2

Stavros Assimakopoulos

yet, the need for “some order into the contents of this wastebasket” (Bar-Hillel 1971:405) was becoming increasingly apparent. It was not until the 1970s that pragmatics started shaping up as a mainstream discipline in its own right, most notably after the uptake and further development of Austin’s original theory of speech acts (1962) by Searle (1969) and the systematic investigation of the inferential communication of nonnatural meaning in everyday conversation by Grice (1975). From then on, all the more scholars started working in pragmatics, with this surge of interest inevitably bringing a diversification of the topics that pragmatic theory would eventually encompass. Clearly, as most of the papers in this volume demonstrate, the most prolific direction in which research in pragmatics has flourished is the description of meaning-related phenomena to which the study of language use in context can offer novel insights. That said, however, there have also been quite a few occasions when the investigation of meaning in context found a useful application in other directions. Among the numerous such directions, there are, in the context of the present volume, two areas that have provided particularly fruitful ground for the development of pragmatic theory over and above its original aims. The first one is the study of human cognition. The most important contribution in this vein has most probably been that of relevance theorists, and more specifically, Sperber and Wilson, who, in their seminal work (1986), departed from the predominantly philosophical orientation of traditional pragmatics theorising and readdressed inferential reasoning as a cognitive capacity. As some contributions in this volume attest, the (originally relevance-theoretic) requirement of psychological plausibility has had a long-lasting effect in the development of the field, and has also helped pave the way for new explorations in diverse domains, such as experimental cognitive or even evolutionary psychology. Another area that has received a lot of interest among researchers of pragmatics in the past few decades is that of conversation analysis. Following the pioneering work of Sacks, Schegloff and Jefferson (1974), conversation analysts use ethnomethodological tools to study the structure and processes of naturallyoccuring mundane conversation. Given the robust findings about the in situ organisation of talk-in-interaction that this fully data-driven approach has generated through the years, it had to also be represented in this volume. The papers included in this volume have been grouped together on the basis of their orientation with a view to showcasing how research in pragmatics has far-reaching implications for discussions in philosophy, cognitive science, linguistics, as well as conversation analysis. Obviously, given the breadth of the topics that are addressed by this volume’s contributors, there are bound to be cases where a paper will transcend the relative boundaries of the area under

Introduction

3

which it is categorised; yet, this should not be considered a problem. If anything, it shows that research that is informed by pragmatics can cross over disciplinary boundaries, offering new insights to issues that have been traditionally approached from a singular perspective. All in all, as will become evident in the brief volume synopsis that follows, apart from this often interdisciplinary outlook, each one of the papers in this volume makes a novel – and often bold – proposal in relation to the question(s) that it seeks to address. In the first paper of part I of this volume, which comprises contributions dealing with topics of a predominantly philosophical nature, Jacques Moeschler challenges Horn’s classical analysis of logical quantifiers in terms of scalar implicature, arguing instead for a treatment at the level of explicature through a meticulous analysis of the notions’ semantic and pragmatic meaning. Then, Richard Vallée revisits Travis’ famous discussion of colour sentences, offering an alternative account, which does not compromise truth-conditional semantics or semantic compositionality, on the grounds of multipropositionalism. Relying on a comparable notion of content-pluralism in relation to the propositional content of predictions, Kepa Korta reassesses, in his paper, his earlier conviction that, in speech-theoretic terms, assertions about the contingent future have an upward direction of fit. Finally, Etsuko Oishi concludes the section with a novel approach that seeks to connect the notions of conversational implicature and presupposition with Austin’s category of expositive illocutionary acts. Conversational implicatures and presuppositions are also the focus of the first paper of part II of the volume, which includes contributions with more of a cognitive orientation. In it, Anne Reboul puts forth a proposal about the origins of these manifestations of implicit communication, arguing that their adaptive benefit lies in their ability to evade mechanisms that we have developed to avoid deception and manipulation. In keeping with the evolutionary outlook, Thanh Nyan attempts in her contribution to account for the structure of utterance meaning in neuro-cognitive terms, by linking Anscrombe and Ducrot’s constraint on the act of utterance to the structure of intentional action, as approached in Jeannerod’s account of motor cognition. In the final paper of this section, Márta Szücs and Anna Babarczy report on an experimental study which challenges the widespread assumption that performance in false-belief tasks and enhanced grammatical competence are reliable indicators of children’s ability to comprehend irony, while suggesting at the same time that the development of a metapragmatic awareness of the requisite skills for the task can make a dramatic difference in irony comprehension. The third part of the volume comprises papers that deal with the treatment of particular linguistic expressions, focusing on analyses that can be crucially enriched or even completely overturned by taking into account the way in which

4

Stavros Assimakopoulos

the relevant expressions are interpreted in context. In the first paper, Enikö Németh T. provides an overview of her extensive research on implicit arguments in Hungarian, with a view to showing that a full explanation of the said phenomenon can only be reached through the combination of insights from syntactic, lexical-semantic and pragmatic approaches to the phenomenon, as well as through the integration of data from various sources over and above introspection. Then, Thorstein Fretheim develops a pragmatic account of single response word utterances, such as ‘Yes’ or ‘No’, thereby refuting Holmberg’s corresponding analysis on the grounds of the response words’ syntactic complexity. In turn, Michael Chiou extends the standard approach to the future tense in Modern Greek, by suggesting that the preferred future prospective interpretation of the relevant construction, which semantically encodes only epistemic possibility, effectively arises at the level of what is communicated as a generalised conversational implicature. In the last paper of this section, Sonja Müller provides a novel, empirically informed account for the occurrence of ‘halt eben’ and ‘eben halt’ in German, according to which, the two modal particle combinations have a difference in markedness, with ‘halt eben’ being the unmarked, and thus preferred option, as it is the order that respects the non-reinforcement of entailments criterion. The fourth and last part of this volume comprises two papers that fall within the domain of conversation analysis. In the first one, George O’Neil reports on a novel study that combines conversation analytic and IPA transcription with the aim of examining pronunciation miscommunications in the interaction of English as a Lingua Franca speakers at the segmental level. In his analysed corpus, the author identifies two relevant repair strategies, that is, preemptive and reactive segmental repair. In the final paper of the volume, Zsuzsanna Németh explores the interactional functions of recycling, replacement, insertion, and aborting in a sizeable corpus of interactions among native speakers of Hungarian, revealing that different repair operations may not only share the same kinds of interactional functions, but may also be used in combination to fulfil a single interactional task in the same turn. In closing this short introduction, it has to be noted that, even though the contributions in this volume represent four major fields of research with which pragmatics interfaces, there are several other perspectives of central interest for scholars working in pragmatics. Perhaps one such perspective that needs to be mentioned, especially given the venue in which all of the papers in the present volume were originally presented, would be the socio-cultural one, which has motivated a huge amount of research and is increasingly gaining in popularity. Since, however, both this and the equally popular discourse-analytic perspective formed part of a different session of the conference, the publication of papers pertaining to them has been pursued in a separate volume (Kecskes and Assimakopoulos in press).

Introduction

5

References Austin, John L. 1962. How to do things with words. Oxford: Clarendon. Bar-Hillel, Yehoshua. 1971. Out of the pragmatic wastebasket. Linguistic Inquiry 2: 401–407. Grice, H. Paul. 1975. Logic and conversation. In Peter Cole and Jerry L. Morgan (eds.), Syntax and Semantics 3: Speech Acts, 41–58. New York: Academic Press. Kecskes, Istvan & Stavros Assimakopoulos. In press. Current issues in intercultural pragmatics. Amsterdam: John Benjamins. Morris, Charles W. 1938. Foundations of the theory of signs. Chicago: University of Chicago Press. Sacks, Harvey, Emanuel A. Schegloff, & Gail Jefferson. 1974. A simplest systematics for the organization of turn-taking for conversation. Language 50. 696–735. Searle, John R. 1969. Speech acts: An essay in the philosophy of language. Cambridge: Cambridge University Press. Sperber, Dan & Deirdre Wilson. 1986. Relevance: Communication and cognition. Oxford: Blackwell.

I Pragmatics and philosophy

Jacques Moeschler

2 Back to negative particulars: A truth-conditional pragmatic account Abstract: The traditional analysis of the quantifiers some and some . . . not is customarily given via the concepts of scalar implicature, Horn’s scales and the logical relations implied by the logical square. In this contribution, I challenge this view by arguing that the pragmatic relation between these particulars is not an implicature, but rather an explicature, that is, a pragmatic development of an utterance’s propositional form. To this end, I provide a Boolean semantics and pragmatics for the particulars at hand, and discuss the consequences of my account for pragmatic analysis, especially in relation to the properties of calculability and cancellability that implicatures are associated with. In closing, I propose a possible answer to Horn’s conjecture about the well-known issue of the non-lexicalisation of negative particulars (the O vertex in the logical square).

1 Introduction In many of his papers Horn (Horn 2004, Horn 2012, Horn 2014) gives an interpretation of scalar implicatures triggered by logical words (and mainly quantifiers) which is based on the logical square, also known as the Aristotelian square. His analysis crucially involves pragmatic relations between subcontraries (where each subcontrary implicates the other) and the unilateral entailment relation between universals and particulars. In this setting, the recovery of implicatures is explained by a Horn-scale, where the positive (I) and negative particulars (O) are upper-bound with respect to their corresponding universals (A and E), and implicate their negation, which is logically equivalent to their converse particular (O and I respectively).

Acknowledgments: This work is part of my new Swiss National Science Foundation research project LogPrag (Semantics and pragmatics of logical words, project n° 100012_146093, 2014– 2017). Thanks a lot to Joanna Blochowiak and Karoliina Lohiniva. Jacques Moeschler, University of Geneva DOI 10.1515/9781501505089-002

10

Jacques Moeschler

In this paper, I wish to challenge the implicature analysis based on the logical square and Horn-scales (Horn 1976, 1984, 1989), by using a more radical Gricean argument: implicatures are calculable, as well as the result of the enunciation of a sentence (i.e. an utterance), which leads to specific truth conditions. More specifically, the scalar implicatures communicated by the particulars at hand cannot be true unless both particulars are true, which essentially makes their defeasibility property vacuous. The proposed analysis favours a pragmatic enrichment process at the level of explicit meaning, that is, an explicature in relevance-theoretic terms (Sperber and Wilson 1995). To this end, a precise Boolean semantics and pragmatics will be given for the quantifiers some and some . . . not,1 and the consequences for the alternative analysis at the level of explicature will be discussed. Finally, I will offer a positive proposal to one of the most intriguing puzzles of the logic-language interface that I have previously (Moeschler 2007) addressed as ‘Horn’s conjecture’ (for a historical perspective on the puzzle, see Horn 2012). In relation to this, I will argue that the absence of lexicalised expressions of O in natural language is not only a question of complexity, which is Horn’s position, but also a question of calculability at the level of semantics. In this respect, I will demonstrate that the semantics of O quantifiers (some . . . not, not all) is the result of the combination of two Boolean operations (intersection and complement), which leads to a complex and costly computation. The upshot of this analysis is that the partition of domains between particulars at the level of pragmatics ensures a more efficient and less costly processing. Against this backdrop, this paper is organised as follows. Section 2 presents the classical pragmatic analysis of quantifiers within the logical square. Section 3 is devoted to the specific issue of negative particulars, with regards to their implicated meaning. In section 4, I discuss the logical properties of particulars, and present a new approach, based on entailment. Section 5 capitalises on this new approach and gives a detailed analysis of the semantics and pragmatics of positive and negative particulars. Then, in section 6, I present the way a procedural analysis for particulars could lead to a sound account of logical words within the remit of the semantics-pragmatics interface. Section 7 provides a general explanation of the pragmatics of particulars, based on communicative and cognitive principles (à la Grice and Relevance Theory). Finally, section 8 gives a last argument against the implicature analysis, based on the possible

1 I consider some . . . not a quantifier because of its position in the logical square.

Back to negative particulars: A truth-conditional pragmatic account

11

denials of negative universals, while section 9 draws the implications of the proposed analysis for pragmatics.

2 The (augmented) logical square In the logical square, there are four logical relations that are defined between the four vertices: the positive ones (AffIrmo), which are called A and I, and the negative ones (nEgO), which are called E and O. While A and E are universals, I and O are particulars. In this picture, the following definitions arise: Entailment: universals unilaterally entail particulars: A → I, E → O. Contradiction: A and O are contradictory, as are E and I (in the sense that only one of the two in each pair can be true), which is logically signalled by the presence of a negation operator: that is, ¬∃x (E) is contradictory to ∃x (I), and ∀x (A) is contradictory to ¬∀x (O). 3. Contrariety: A and E are contrary, in the sense that they cannot both be true. So, they can either be both false or only one of them can be true. 4. Subcontrariety: I and O are subcontrary, in the sense that they can either be true together, or one can be true and the other false, but they cannot be false together. 1. 2.

As regards quantifiers (all, none, some, not all), it has been long proposed that particulars Q-implicate the negation of their corresponding universals (cf. Gazdar 1979; Horn 1976, 2004). So, whereas their logical meanings include the corresponding universal (at least some if not all, at least some . . . not if not none), their pragmatic meaning is obtained by implicature (e.g. not all for some, not none for some . . . not) arising from the Gricean maxim of quantity – ‘Try to make your contribution one that is true’ – and, more specifically, its first submaxim – ‘Make your contribution as informative as is required.’ From a neo-Gricean perspective, limited to general Q-Principles and I- or R/M-Principles (Horn 1984, Levinson 2000), universals and particulars constitute a semantic scale (, ) where the weaker term is unilaterally entailed by the stronger term, and implicates its negation, as in (1): (1)

a.

A → I, E → O

b.

I +> ¬A, O +> ¬E

In this setting, the logical square can be represented as follows:

12

Jacques Moeschler

Figure 1: The logical square

This representation raises two main questions, namely: 1) Why is there no linguistic realisation of the O vertex and 2) why is the relation between particulars and universals an implicature? The first question is closely linked to what I call Horn’s conjecture: Given that languages tend not to lexicalize complex values that need not be lexicalized, particularly within closed categories like quantifiers, we predict that some . . . not will not be lexicalized, and this is precisely what we find. (Horn 2004: 11)

Horn correctly predicts that *O is not lexicalised in natural language. In English, there are no lexical items such as *nall, *nalways, *noth, *nand., but the relevant meanings are rather communicated using complex syntactic clusters, such as not all, not always, not both, not . . . and. The same asymmetry is found in French, where pas tous (‘not all’), pas toujours (‘not always’), pas les deux (‘not both’), pas . . . et (‘not . . . and’) are used instead of *nitous, *nitoujours, *nideux, *niet. To my mind, a main (empirical) question that arises from Horn’s conjecture is whether the complexity answer is sufficient. As I will argue later on, it is not, since the absence of O is essentially a question of calculability: that is, negative particulars are the result of the combination of (at least) the two Boolean operations of intersection and complement, and can therefore not be translated into a single linguistic marker.2

2 Note that complexity and calculability are not directly visible in logical translation: ¬∃x[P(x)] is not complex, since it means “there is not an x that is a P”, whereas ∃x[¬P(x)] means that you identify an x outside the set defining the x (“there is an x such that it is not a P”).

Back to negative particulars: A truth-conditional pragmatic account

13

Then, the implicature story is also problematic. In order for each particular to implicate the relevant other particular, it just needs to be true, which effectively makes the property of being true the result of the implicature (some meaning not all). This circularity of the implicature analysis opens up the possibility that the restricted interpretation of some and some . . . not is in fact an explicature of the logical form of the utterance containing them, rather than an implicature. If this is correct, it follows that each particular, with its restricted explicature, actually entails the relevant other particular. This analysis is akin to the one pursued by Noveck and Sperber (2012); albeit with a different argumentation: (. . .) linguistic expressions serve to indicate rather than encode the speaker’s meaning, and [. . .] the speaker’s meaning is quite often a narrowing or broadening of the linguistic meaning. Taking ‘some’ to indicate not at least two and possibly all but at least two and fewer than all is a common narrowing of the literal meaning of ‘some’ at the level of the explicature of the utterance. (Noveck and Sperber 2012: 314)

Even though I will provide a different rationale for the proposed explicature reading in Section 4, the output of my analysis will be the same: the pragmatic meaning of particulars is a specification (that is, a narrowing) of their encoded meaning. Against this backdrop, in what follows, I will be using the term semantic meaning to refer to encoded meaning, and the term pragmatic meaning to refer to the relevant explicature output. Before arguing along the aforementioned lines, however, it is necessary to outline the negative particulars issue and its analysis by Horn.

3 The problem of negative particulars Horn’s explanation of scalar implicatures can be found to face two main issues, the first of which lies in the requirement that a Horn-scale should apply to expressions of the same type (cf. the Lexicalization condition, Atlas and Levinson 1981). While positive quantifiers satisfy this condition (all and some are, after all, single lexical items), negative quantifiers do not (since some . . . not and not all are not single lexical items). Also, since, as we have seen, O is never lexicalised, O and E should not in principle belong to a semantic scale. A possible answer to this objection is that the relevant condition is not a condition on scales by itself, since a lot of expressions that belong to scales are not one-word expressions; in French, for instance, the quantifier un peu (‘a little’) is not a one-word expression, and neither is la plupart de (‘most’) or others with a nominal head (une floppée de = ‘a slew’, une myriade de = ‘a myriad’, un tas de

14

Jacques Moeschler

= ‘a lot’, une chiée de = ‘a shitload’, etc.). However, the important issue here is not the one-word condition, but rather the one-meaning condition: if two expressions are connected by a relation of entailment and scalar implicature, they need to be associated to a single meaning. In what follows, I will show that not all, as well as some . . . not, are not associated to a single meaning because their meanings are compositional. The second, more serious, issue has to do with the computability of scalar implicatures with a negative trigger. For example, the scalar implicature (3) is drawn from (2), because it is calculable – a well-known property of conversational implicatures (cf. Grice 1975) – even if it not automatic, as it has been abundantly demonstrated in the relevant literature. Similarly, it can be easily calculated how the true evaluation of (4) is compatible with the all interpretation provided in (5) (cf. Noveck 2001): (2)

Some of my students passed.

(3)

Not all of my students passed.

(4) Some elephants have trunks. (5)

All elephants have trunks.

However, the situation is not the same with negative quantifiers. Consider, for example, the interpretation of (6): (6) Some of my students did not pass. In this case, are we to draw the scalar implicature in (7)? And if yes, is it accessible? (7)

It is not the case that none of my students passed.

Before giving a possible answer à la Horn, let us have a look at the possible contexts in which both (2) and (6) can be uttered. Suppose that one of my colleagues asks me: (8) How was your pragmatics exam? By answering (8) with (2) – i.e. Some of my students passed, I implicate that some students did not pass (= not all). My answer here has a positive orientation: I may, for instance have prepared a very difficult pragmatics exam, and expect only a few students to pass it. So, in this case, the results are consistent with

Back to negative particulars: A truth-conditional pragmatic account

15

my expectation. But suppose that I think that the pragmatics exam was easy enough for all students to pass. In this case, giving answer (6) – i.e. Some of my students did not pass – would be negative, and it could not be contrasted to either the maximal interpretation of some . . . not, given in (9) below, or its positive counterpart; the relevant point being that not all of my students passed, and not that some of them passed. (9) None of my students passed. So, can the identification of my intended implicature in the latter context be addressed from a Hornian perspective? It surely can, since by following Horn’s logical analysis, one can see that the negation of a universal is equal to its opposite particular: I +> not-A = O and O +> not-E = I, so I +> O and O +> I. Thus, by identifying the logical relation of contradiction that holds between the positive universal and the negative particular and between the negative universal and the positive particular, the computational challenge can be met. To put it more simply, the logical interpretation of the scalar implicature analysis leads to a short-circuited process: some directly Q-implicates some . . . not (not all), and some . . . not (not all) Q-implicates some. What follows then, is that that there is a very accessible pragmatic relation which prevents the emergence of an incomputable relation between some . . . not and not . . . none. However, this analysis also has the non-trivial consequence that utterances with different truth conditions implicate their subcontraries. So, returning to our familiar scenario, in uttering (2) I implicate (6), and in uttering (6) I implicate (2). This is obviously problematic, since as we have just shown, the contexts of utterance for (2) and (6) are not the same. Apart from that, if one admits this circular implicature relation between I and O, where I implicates O, and O implicates I, how could one account for the difference in meaning between I and O? Clearly, the relevant quantified utterances (2) and (6) have different truth conditions, as their logical forms in (10) show: (10)

a.

∃x [x is a student and x passed]

b.

∃x [x is a student and x ¬passed]

The only answer to this question that is explicitly given by Horn, refers to what Grice (1975) called conveyed meaning, which consists in what is said + what is implicated and which Horn correspondingly calls communicated meaning. In this picture, I and O may make different contributions to what is said and what is implicated, but they have an identical communicated meaning, consisting in

16

Jacques Moeschler

their addition. In other words, what both subcontraries communicate is their conjunction (11): (11)

a.

I ++> I & O

b.

O ++> O & I

In light of this, a question that immediately arises is the following: Why choose a positive or a negative description if the communicative meaning of both particulars is the same? Obviously, Horn’s analysis cannot provide an answer to this question, nor can it give a reason as to why a scalar implicature is derived or not.

4 The logical meaning of particulars As we have already seen, in the logical square, particulars are subcontraries. By definition, subcontraries can be true together, but cannot be false together. (12) and (13) below show how subcontraries can be true together: (12)

Some of my students passed: Marie, Suzanne and Louise.

(13) Some of my students did not pass: Marc, Luc and Paul. We, therefore, have three logical possibilities with subcontraries: a.

(12) and (13) are true together, as illustrated in (14), where each particular does not only implicate the other, but also communicates their conjunction: (14)

Some of my students passed: Marie, Suzanne and Louise; and some did not pass: Marc, Luc and Paul.

b. If one of the particulars is false, then the meaning of the relevant quantifier includes the universal, i.e. some means all, and some not means none. Yet, in this case, the implicature becomes false and the speaker may be accused of deceiving her addressee by providing too weak a statement, even if it is essentially true. c. If (12) and (13) are both false, it follows that both all students and none of the students passed, which obviously cannot be the case, as (15) shows: (15)

# All of my students passed and none of them passed.

Back to negative particulars: A truth-conditional pragmatic account

17

So, what are the consequences of these three possible situations, if one follows the standard definition of particulars? The answer would be that in order for one particular to implicate the other, the other must be true as well. And if the converse particular cannot be false, the truth conditions for particulars cannot be those of inclusive disjunction (or), but need to be those of conjunction (and), as Table 1 below shows: Table 1: Truth conditions for particulars P

Q

P or Q

P and Q

1 1 0 0

1 0 1 0

1 1 1 0

1 0 0 0

This means that the particulars some and some . . . not are not pragmatically tied by means of implicature, but are rather logically tied by means of entailment; in other words, in order for it to be true, a particular must entail its subcontrary: (16)

a.

If I is true, then I → O

b.

If O is true, then O → I

This, in turn, leads one to wonder whether particulars are still cancellable under this analysis, given that cancellability is a basic property of conversational implicature (cf. Grice 1975; Sadock 1978). In order to test this, one can follow Grice (1989) and use a but-contrast clause. Examples (17) and (18) below are classic examples of implicature cancellation: in unmarked contexts, and triggers a cancellable conversational implicature, as in (17), while but communicates a non-cancellable conventional implicature, as in (18): (17)

(18)

a.

Mary got angry and Peter left, but nobody knows what happened first.

b.

Cancellable conversational implicature: Mary got angry and then Peter left.

a.

# Peter has three castles but only one car, but there is no contrast between these two facts.

b.

Non-cancellable conventional implicature: there is a contrast between having three castles and only one car.

18

Jacques Moeschler

As (19) below shows, the but test cannot be used with the alleged scalar implicatures communicated by particulars, and the only way in which the relevant meanings can be cancelled is by resorting to a connective such as in fact (or en fait in French), as shown in (20); yet, this connective leads to semantic and pragmatic effects which are markedly different from those yielded through standard implicature cancellation: (19)

(20)

a.

# Some of my students passed, but all passed.

b.

# Some of my students did not pass, but none passed.

a.

Some of my students, in fact all, passed.

b.

Some of my students, in fact none, did not pass.

c.

Quelques étudiants, en fait tous, ont échoué.

d.

Quelques étudiants n’ont pas échoué, en fait aucun n’a échoué.

The implicature test is meant to give us “cancellation without contradiction,” in the sense that an assertion (P) which triggers a scalar implicature can be true while its implicature is false. If this assertion is false, however, its implicatum (Q) cannot be true either. Table 2 below presents all possibilities, showing that a true utterance may or may not trigger an implicature, while a false one cannot really trigger an implicature: Table 2: Truth conditions for conversational implicatures (cf. Moeschler 2013a) P (assertion)

Q (implicatum)

P implicates Q

1 1 0 0

1 0 1 0

1 1 0 1

As we saw in (19), Grice’s test for implicature cancellation cannot be used for quantifiers such as some and some . . . not. One could argue that the reason for this is mainly syntactic, but, upon closer inspection, it turns out to be semantic. As (21) below shows, the only possible sequencing on a scalar implicature is a confirmation of the implicature, while cancellation with but is not possible, as (19a), repeated in (22), demonstrates, and a continuation compatible with the lower-bound reading of some, as in (23), is odd:

Back to negative particulars: A truth-conditional pragmatic account

(21)

Some of my students passed, so not all passed.

(22)

# Some of my students passed, but all passed.

19

(23) # Some of my students passed, so all passed. The upshot here is that the not-all meaning communicated by some is effectively not an implicature. So, in the case of the cancellation in (20a), it is actually not the implicature that is cancelled, but rather the subcontrary (I) that is declared to be false. That is why, when no such correction occurs, the subcontrary some still yields the enriched reading only some. (24) Only some of my students passed. In relevance-theoretic terms (Sperber and Wilson 1995; Wilson and Sperber 2012; Noveck and Sperber 2012; Carston 2002), this meaning would qualify as an explicature, that is, as a part of an utterance’s explicitly-expressed content, as defined below: (25) An assumption communicated by an utterance U is explicit if and only if it is a development of a logical form encoded by U. (Sperber and Wilson 1995, 182)

5 The semantic and pragmatic meaning of particulars Given the argumentation up to this point, in order to solve the problem of the meaning of particulars, we have to face two main challenges. The first one concerns the status of the pragmatic meaning of particulars, which I have already addressed in the previous section by giving a general argument in favour of an analysis along explicature rather than implicature lines. This argument will be further supported by the way in which I will now approach the second challenge, that is, the challenge of identifying the semantics and pragmatics of some and some . . . not. To this end, I will explore three hypotheses: 1. There is a strong connection between both particulars, which can be expressed by a complement operation. 2. Their semantics can be approached in terms of their truth-conditional incompatibility: some is semantically (i.e. logically) incompatible with no, as some . . . not is with all, since both pairs are contradictory. 3. Their pragmatics is given in terms of their incompatibility with their corresponding universals.

20

Jacques Moeschler

In my view, the solution to the issue of positive and negative particulars lies in the distinction between the semantics and pragmatics of some and some . . . not. If we can argue for a semantic content consistent with an upper-bound meaning, and a pragmatics accounting for the restriction on truth conditions, I believe that we will be on a good path to solve the problem. So, let’s start by approaching the semantics of some from a set-theoretic perspective: (26)

The semantics of some X are Y: a. the intersection between the sets denoted by X and Y is not empty, and b. some X are Y is semantically incompatible with no X is Y; it is the case that (X ∩ Y) ≠ ∅.

The semantics of some states that, all things equal, some X are Y cannot denote a situation where no X is Y. The simplest way of expressing this in set-theoretic terms is by noting that the intersection between sets X and Y is not empty:

Figure 2: A set-theoretic account for some X are Y

This effectively means that what is excluded at a semantic level is the contradictory meaning no X is Y, which can in turn be represented as follows:

Figure 3: A set-theoretic account for no X is Y

Now, the pragmatics of some X are Y will restrict the intersection condition, deterring some X are Y from denoting that all X are Y, as represented in Figure 4:

Back to negative particulars: A truth-conditional pragmatic account

21

Figure 4: A set-theoretic account for all X are Y

So: (27)

The pragmatics of some X are Y a. X is not included in Y, because there must be a sub-set of X which is not in Y, and b. some X are Y is pragmatically incompatible with all X are Y; it is the case that X ⊄ Y.

As a result, the pragmatic meaning of some X are Y is the explicature only some X are Y. With this in mind, we can now easily tackle the issue of identifying the semantics of some . . . not. What we need is a way of specifying a situation where the complement of the intersection of the sets X and Y is not empty, as represented in Figure 5 below:

Figure 5: A set-theoretical account for some X are not Y

In this setting, the semantics of some . . . not can be approached as follows: (28)

The semantics of some X are not Y a. The complement of the intersection between X and Y (i.e. the sets denoted by X and Y) is not empty, and b. some X are not Y is semantically incompatible with all X are Y; It is the case that ∁ (X ∩ Y) ≠ ∅.

22

Jacques Moeschler

Similarly, the pragmatics of some . . . not can also be easily computed, since, rather than being incompatible with all, as its semantics predicts, it is now incompatible with no: (29) The pragmatics of some X are not Y a. the intersection between X and Y is not empty, and b. some X are not Y is pragmatically incompatible with no X are Y; It is the case that (X ∩ Y) ≠ ∅. As a result, the pragmatic meaning of some X are not Y is the explicature only some X are not Y. All in all, the analysis presented here, can be summarised as follows: Table 3: The semantics and pragmatics of some and some . . . not Semantics

Pragmatics

some X are Y

(X ∩ Y) ≠ ∅

X⊄Y

some X are not Y

∁ (X ∩ Y) ≠ ∅

X∩Y≠∅

From this analysis, the following general conclusions can be drawn: a.

Some X are Y is semantically incompatible with no X is Y (i.e. they are contradictory). b. Some X are not Y is semantically incompatible with all X are Y, i.e. they are also contradictory. c. Some X are Y is pragmatically incompatible with all X are Y. d. Some X are not Y is pragmatically incompatible with no X is Y. e. With some X are Y and some X are not Y, both I and O are true. f. With all X are Y, O is false. g. With no X is Y, I is false. Against this background, some further comments are still in order. For one, the proposed account, in which the semantics of some X are Y is the pragmatics of some X are not Y, and vice-versa, is radically different from the traditional analysis in terms of conversational implicature. In effect, in line with the argument pursued in the previous section, it predicts an enrichment of the propositional form of the utterance containing the particular, rather than an implicature relation. So, if we take the case of some X are Y, all that is required for the quantifier to denote the intended meaning is that the intersection between sets X and Y is not empty, which eradicates the need for an implicature, like not all, to be triggered. Another prediction that this account makes is that,

Back to negative particulars: A truth-conditional pragmatic account

23

the only some reading will be triggered if some is pragmatically incompatible with all. Once again, this output is an explicature, as it determines the conditions under which a relevant truth-conditional meaning can be assessed. Notably, the present analysis does not depend on any maxims of conversation, nor does it appeal to any general principles, like the communicative principle of relevance put forth by relevance theorists. Furthermore, it supports the position that some . . . not is not only syntactically complex, but also – and maybe even to a greater extent – semantically complex, as it requires the simultaneous use of two Boolean operations: complement (∁) and intersection (∩). In this respect, it is not surprising that there is no single lexical item encoding some X are not Y in natural language, as it carries the complex meaning: first, intersect two sets of individuals; then, compute the complement of the restrictor (X). The final issue that needs to be addressed in the context of the present account concerns the nature of the semantic and pragmatic computation required by positive and negative particulars.

6 A truth-conditional pragmatics proposal: the role of explicature in utterance comprehension The proposal pursued so far is that the interpretation of particulars is directly dependent on their truth-conditional meaning. This inevitably means that the choice of a positive or a negative particular is dependent on the situation described by the utterance. On the other hand, as we have already seen in section 4, the meaning of particulars is context-dependent; that is, the not-all and not-none interpretations may or may not be triggered, depending on the utterance context. Nonetheless, there seems to be a general mechanism leading to the appropriate contextual interpretation, which corresponds to the exclusion condition. In a nutshell, at both the semantic and the pragmatic level, the procedure for interpreting some and some . . . not is guided by what is incompatible with (that is, what is excluded by) the quantifier. The exclusion condition can thus be formulated as follows: (30)

Exclusion condition: a. Exclude the incompatible semantic meaning. b.

Exclude the incompatible pragmatic meaning.

c.

Enrich the explicitly-expressed pragmatic meaning.

24

Jacques Moeschler

This procedure leads to a specified reading by narrowing the semantics of the particulars, as illustrated in (31) and (32): (31) Procedure for some

(32) Procedure for some . . . not

At first glance, this explanation seems to be more structural than contextual, but this is not the case. In fact, the first step is truly semantic, since no contextual import is required: some leads to an alternative between all or some, and some . . . not leads to an alternative between no and some . . . not. The context, therefore, intervenes at the pragmatic level, which is exactly what a pragmatic interpretation predicts. But what is additionally required is a semantic step, which, in this picture, is not equivalent to any type of literal interpretation. What is checked, instead, is whether the derived description is on the right track, and whether the encoding allows for all possible interpretations compatible with the logical properties of some and some . . . not to be made. One side question that arises here concerns the status of the type of meaning represented in (31) and (32). Although this issue is crucial in some frameworks of pragmatics, such as Relevance Theory, I am not going to address it at length here, but will rather only give a general description on how it could be handled in this approach. The type of representation given in (31) and (32), resembles what relevance theorists call a procedure (cf. Escandell-Vidal et al. 2011) with input conditions and outputs. In this picture, the targeted meaning is the outcome of a procedure, representing the procedural meaning of positive and negative particulars respectively. From this perspective, quantifiers are prototypical examples of conceptual and procedural information being semantically encoded in tandem, as in the case of tenses and connectives (for tenses, see Grisot and Moeschler 2014; and for connectives, see Moeschler 2016a). From a relevance-theoretic view-

Back to negative particulars: A truth-conditional pragmatic account

25

point, procedural meaning is often contrasted with conceptual meaning mainly because it is not accessible to consciousness, it is paraphrasable with difficulty, and it is not associated to any particular (representational) concept in the Language of Thought (cf. Blakemore 1987; Wilson and Sperber 1993; Wilson 2011). Clearly, quantifiers also encode conceptual content; some, for example, is associated with entailment conditions, which are represented, in the present account, as meaning all or some at the semantic level of the relevant procedure. Quantifiers, such as the particulars at hand,3 are thus examples where conceptual and procedural meaning are mingled together, with conceptual meaning representing input conditions and procedural meaning indicating the path to the target-restricted meaning. And even in this setting, the hypothesis that the output of the procedure is a development of the logical form, that is, an explicature, is not exceptional either, since there are connectives, such as parce que (‘because’ in French), which encode procedural information yielding explicatures too (cf. Moeschler 2016b).

7 A general explanation I am now in a position to give a general explanation of the observed phenomena, that is, the readings that come about through the specification of particulars (only some, only some . . . not) in the way discussed above. This explanation will be two-fold, as it will relate to both the cognitive and the communicative levels. To begin with, a speaker using a sentence with a particular would violate the first maxim of quantity if she were to say some while meaning all. So, when no contextual import allows the lower bound reading, the quantifier would end up being specified in a restricted meaning. Now, the entailment relation from some to some . . . not, as well as from some . . . not to some, makes the procedure given in (31) and (32) efficient enough to obtain the expected interpretation. Interestingly, the restricted meaning of some and some . . . not can receive another explanation from a relevance-theoretic point of view: saying some while meaning all would allow the addressee to make unjustified inferences, leading to false conclusions, which would give rise to unsuccessful communication and therefore weaken the relevance4 of the utterance at hand. So, from a com3 There is no empirical reason that this entrenchment of levels of meaning is only restricted to logical words, like some: other quantifiers in a semantic scale, such as many, a few, most etc., could also be analysed in this way. 4 From a relevance-theoretic perspective, the relevance of an utterance depends on a balance between the positive cognitive effects that its processing offers and the effort expended in processing these effects.

26

Jacques Moeschler

municative point of view, the specification I have provided at the explicature level receives a sound explanation, both from a Gricean and a post-Gricean perspective. Then, the restricted interpretation at the level of explicature can also receive a cognitive explanation: the partition reading for some and some not should allow for an efficient and rapid processing, avoiding useless cognitive processing. In effect, the only some vs. only some . . . not readings yield a straightforward partition of the domain of the restrictor X. What is thus targeted by both quantifiers leads to a defined meaning which explains the accessibility to different communicated meanings. Figure 6 showcases what this partition can be for the examples given in (33): (33)

a.

Some linguists know logic.

b.

Some linguists do not know logic.

Figure 6: Denotation of (33)

What Figure 6 further shows is how the choice between (33a) and (33b) makes sense. For one, what is targeted as a denotation is not the same subset. Moreover, when each subset is obtained, through the procedures (31) or (32), no additional processing, such as the one involved in the computation of costly conversational implicatures, is required. In this respect, the prediction of this

Back to negative particulars: A truth-conditional pragmatic account

27

analysis is that negative particulars are not costlier than positive ones in processing terms.

8 A final argument against the implicature analysis: implicature and denial In an attempt to support the proposed account even further, I would like to provide a further, empirical this time, argument in favour of the explicature analysis, on the basis of the possibility of denial via the negative particular. The point here is that the negative particular (some . . . not) cannot be used to refute the universal none. Let’s imagine the following situation: Mary claims that no linguist knows logic. In this case, the best way to refute Mary’s claim is to use the contradictory of no, which yields the not all reading: (34) Mary: No linguist knows logic. Jacques: No, I know some linguists who know logic. Entailment: it is false that no linguist knows logic. Now suppose that I want to indirectly refute Mary’s claim by using a negative particular assertion containing some . . . not. The prediction of the implicature analysis it that O implicates not-E, that is, the negation of Mary’s claim. But in reality, this is not what happens, as can be seen in (35): (35) Mary: No linguist knows logic. Jacques: No, I know some linguists who do not know logic. Non-inferable conversational implicature: it is false that no linguist knows logic. Non-inferable entailment: I know some linguists who know logic. (35) is clearly not a refutation, since its meaning is not the contradictory statement of E, that is, I. This example shows that the classical analysis makes a false prediction, and cannot indeed explain why an implicature cannot be contradictory to the universal expression that unilaterally entails the triggering sentence. That is because its implicature ‘some linguists know logic’ is not inferable here and as a consequence, the negative particular cannot be a refutation of the negative universal.

28

Jacques Moeschler

In fact, by using some . . . not, Jacques can only confirm Mary’s claim, because what he asserts is already entailed by Mary’s claim, as shown in (34): (36) Mary: No linguist knows logic. Jacques: Indeed, I know some linguists who do not know logic. Entailment: no linguist knows logic → some linguists do not know logic The other possible way of denying Mary’s claim is by using a proposition that is contrary to Mary’s one, that is, a positive universal statement (all), as in (37). So, a denial is not only realised by a contradictory statement, since a contrary one can also do the same: (37) Mary: No linguist knows logic. Jacques: No, all linguists know logic. Entailment: all linguists know logic → it is not the case that no linguist knows logic In conclusion, it can be affirmed that there is no way of connecting a negative particular with the negation of its corresponding universal, as the Horn-scale theory predicts. And, to my mind, this argument alone should be strong enough to vindicate the explicature analysis over the implicature one.

9 Conclusion This paper has dealt with the meaning of negative particulars, and more specifically the implicated meaning of logical quantifiers, such as some and some . . . not. In it, I have argued that the pragmatics of positive and negative particulars is truth-conditional and, therefore, their treatment belongs to the level of explicature rather than that of implicature. To this end, I developed some arguments for the explanation of the computation of the relevant particulars’ pragmatic meaning on the basis of their logical meaning which excludes the particulars’ contradictory counterparts. Along these lines, I also showed that the pragmatic computation of particulars excludes the corresponding universals for reason of relevance. Finally, I suggested that the restricted truth-conditional meanings attributed to subcontraries in fact entail their subcontraries, and indicated that the analysis that relates subcontraries via implicature is questionable, to say the least. If the arguments given in this paper are sound, then some further conclusions, in relation to the pragmatic analysis of quantifiers, and more generally, the analysis of scalar implicatures, can be drawn.

Back to negative particulars: A truth-conditional pragmatic account

29

Firstly, logical quantifiers such as all, no, some have meanings that are largely dependent on precise logical relations (for a general discussion on the relation between logic and natural languages, see Moeschler in press). The discussion of the relation between particulars and universals in terms of implicature does not put into question the very notion of scalar implicature itself, but rather its domain of application alone. As such, and without taking into account other quantifiers and more complex scales, positive and negative alike,5 this paper does not target the notion of scalar implicature such, but only some properties of scalar implicatures that have been associated to logical words. Secondly, although scalar implicatures have given rise to a substantial amount of research, a lot remains to be done. More specifically, the precise semantics and pragmatics of logical and non-logical quantifiers, as well as logical and non-logical connectives, need to be investigated at both the descriptive and the experimental level. For instance, apart Horn’s treatise on negative scales (Horn 1989), a lot less work has been devoted to negative statements than to positive ones in the literature on scalar implicatures. To give a simple example, it is still pretty much unknown what the impact of negation (both descriptive and metalinguistic) on scalar implicature is. Can we say, for instance, that a scalar implicature is computed under both descriptive and metalinguistic negation? A pragmatic theory of scalar implicatures should explain why the scalar implicature of the positive counterpart in (38) is not active under the descriptive negation in (39), when it is targeted under metalinguistic negation, as in (40) (cf. Moeschler 2013b): (38) Anne has three children. Scalar implicature: Anne does not have four children. (39) Anne does not have three children. Non-inferred scalar implicature: Anne does not have four children. Entailments: Anne has two children, Anne has one child. (40) Anne does not have three children, she has four of them. Defeated scalar implicature: not (Anne does not have four children) Entailments: Anne has four children; Anne has three children, etc. Luckily, this is in the agenda of the LogPrag research project that is currently under way. 5 For a complete analysis of positive and negative quantifiers, taking into account both their logical and pragmatic properties, see Lohiniva (2014).

30

Jacques Moeschler

References Atlas, Jay R. & Stephen C. Levinson. 1981. It-Clefts, informativeness, and logical form: Radical pragmatics (revised standard version). In Peter Cole (ed.), Radical pragmatics, 1–61. New York: Academic Press. Blakemore, Diane. 1987. Pragmatic constraints on relevance. Oxford: Basil Blackwell. Carston, Robyn. 2002. Thoughts and utterances: The pragmatics of explicit communication. Oxford: Basil Blackwell. Escandell-Vidal, Victoria, Manuel Leonetti & Aoife Ahern (eds.). 2011. Procedural meaning: Problems and perspectives. Bingley: Emerald Brill. Gazdar, Gerard. 1979. Pragmatics: Implicature, presupposition, and logical form. New York: Academic Press. Grice, H. Paul. 1975. Logic and conversation. In Peter Cole & Jerry L. Morgan (eds.), Syntax and semantics 3: Speech acts, 41–58. New York: Academic Press. Grice, H. Paul. 1989. Studies in the way of words. Cambridge, MA: Harvard University Press. Grisot, Cristina & Jacques Moeschler. 2014. How do empirical methods interact with theoretical pragmatics? The conceptual and procedural contents of the English simple past and its translation in French. In Jèsus Romero-Trillo (ed.), Yearbook of corpus linguistics and pragmatics 2014, 7–33. Cham: Springer. Horn, Laurence R. 1976. On the semantic properties of logical operators in English. Bloomington: IULC. Horn, Laurence R. 1984. Towards a new taxonomy for pragmatic inference: Q-based and R-based implicature. In Deborah Schiffrin (ed.), Meaning, form, and use in context: Linguistic applications (GURT’ 84), 11–42. Washington, D.C.: Georgetown University Press. Horn, Laurence R. 1989. A natural history of negation. Chicago: The University of Chicago Press. Horn, Laurence R. 2004. Implicature. In Lawrence R. Horn & Gregory Ward (eds.), The handbook of pragmatics, 3–28. Oxford: Blackwell. Horn, Laurence R. 2012. Histoire d’*O: Lexical pragmatics and the geometry of opposition. In Jean-Yves Béziau & Gilbert Payette (eds.), New perspectives on the square of opposition, 393–426. Bern: Peter Lang. Horn, Laurence R. 2014. The singular square: Contrariety and double negation from Aristotle to Homer. In Papers dedicated to Jacques Moeschler, Joanna Blochowiak, Cristina Grisot, Stephanie Dürrleman-Tame & Christopher Laenzlinger (eds), Geneva. URL: www.unige.ch/ lettres/linguistique/moeschler/Festschrift/Festschrift.php. Levinson, Stephen C. 2000. Presumptive meanings: The theory of generalized conversational implicature. Cambridge, MA: The MIT Press. Lohiniva, Karoliina. 2014. Squares and scales: A study of opposition and entailment. Geneva: University of Geneva MA thesis. Moeschler, Jacques. 2007. Why are there no negative particulars? Horn’s conjecture revisited. Generative Grammar in Geneva 5. 1–13. Moeschler, Jacques. 2013a. Is a speaker-based pragmatics possible? Or how can a hearer infer a speaker’s commitment?. Journal of Pragmatics 43. 84–97. Moeschler, Jacques. 2013b. How ‘Logical’ are Logical Words? Negation and its Descriptive vs. Metalinguistic Uses. In Maite Taboada & Radoslava Trnavac (eds), Nonveridicality, evaluation and coherence relations, 76–110. Leiden: Brill. Moeschler, Jacques. 2016a. Where is procedural meaning located? Evidence from discourse connectives and tenses. Lingua 175–176. 122–138.

Back to negative particulars: A truth-conditional pragmatic account

31

Moeschler, Jacques. 2016b. Argumentation and connectives: How do discourse connectives constrain argumentation and utterance interpretation?. In Alessandro Capone & Jacob Mey (eds.), Interdisciplinary studies in pragmatics, culture and society, 653–675. Cham: Springer. Moeschler, Jacques. in press. Formal and natural languages: What logic tells us about natural language. In Anne Baron, Gerard Steen & Gu Yueguo (eds.), Routledge handbook of pragmatics. Abington: Routledge. Noveck, Ira. 2001. When children are more logical than adults: Experimental investigations of scalar implicatures. Cognition 78. 175–188. Noveck, Ira & Dan Sperber. 2012. The why and how of experimental pragmatics: the case of ‘scalar inferences’. In Deirdre Wilson & Dan Sperber, Meaning and relevance, 307–330. Cambridge: Cambridge University Press. Sadock, Jerrold M. 1978. On testing for conversational implicature. In Peter Cole (ed.), Syntax and semantics 9: Pragmatics, 281–297. New York: Academic Press. Sperber, Dan & Deirdre Wilson. 1995. Relevance: Communication and cognition, 2nd edn. Oxford: Basil Blackwell. Wilson, Deirdre. 2011. The Conceptual-Procedural distinction: Past, present and future. In Victoria Escandell-Vidal, Manuel Leonetti & Aoife Ahern (eds.), Procedural meaning: Problems and perspectives, 3–31. Bingley: Emerald Brill. Wilson, Deirdre & Dan Sperber. 1993. Linguistic form and relevance. Lingua 90(1/2). 1–25. Wilson, Deirdre & Dan Sperber. 2004. Relevance theory. In Lawrence R. Horn & Gregory Ward (eds.), The handbook of pragmatics, 607–632. Oxford: Basil Blackwell. Wilson, Deirdre & Dan Sperber. 2012. Meaning and relevance. Cambridge: Cambridge University Press.

Richard Vallée

3 Colour adjectives, compositionality, and true utterances Abstract: According to Travis, the meaning of a simple colour sentence, like “The leaf is green”, does not determine its truth conditions because, while both this meaning and the world can stay the same, an utterance of this sentence might be true and another false. In this paper, I offer a different account of colour sentences grounded on multipropositionalism; an approach that distinguishes between meaning and content, but brings utterances to the centre of the stage. In this picture, an utterance of a sentence semantically determines many different contents, among which one encounters semantically determined and official content, with the latter being amenable to alterations that bear no consequence on semantically determined content. I argue that my proposal preserves both truth-conditional semantics and compositionality.

1 Introduction 1.1 Colour adjectives Travis (1997) contends that the truth conditions of an utterance of a colour adjective sentence are not solely a function of its component parts – or, to put it more simply, are not solely determined by the meaning of its component parts. In support of this position, he gives what has by now become the standard example: A story. Pia’s Japanese maple is full of russet leaves. Believing that green is the colour of leaves, she paints them. Returning, she reports, ‘That’s better. The leaves are green now.’ She speaks truth. A botanist friend then phones, seeking green leaves for a study of greenleaf chemistry. ‘The leaves (on my tree) are green,’ Pia says. ‘You can have those.’ But now Pia speaks falsehood. (Travis 1997: 89)

Acknowledgements: I want to thank the audience at the Intercultural Pragmatics Conference in Malta. Eros Corazza made helpful comments on a previous version of this paper. Research for this work has been made possible thanks to the Social Sciences and Humanities Research Council of Canada. Richard Vallée, Université de Moncton DOI 10.1515/9781501505089-003

34

Richard Vallée

For simplification, let’s assume that there is only one leaf and that both utterances that Pia produces are of the very same syntactically complete affirmative sentence: ‘The leaf is green’. Let’s also assume that the world does not change at all during the temporal period that elapses between the two utterances. Evidently, since the sentence contains no semantically ambiguous or indexical terms, it should have the same meaning in both of its utterances, being true in worlds where the leaf is green, and false in worlds where the same leaf is not. So, since the world remains the same in both utterances of the same nonindexical sentence, one could not hold one of them to be true and the other false. In Travis’ example, however, each utterance is truth assessed after considering facts, and, following brute intuitions, with the first one being considered true and the second one being considered false. Prima facie, these utterances should differ accordingly in their truth conditions, in some way or another, but we do not seem to be able to trace any such differences back to meaning. Travis concludes that [. . .] one speaking of it may clearly state what is true, while another state what is false. That can only be so if the semantics of [1] [‘The leaf is green’]1 on some speakings is substantially richer than that fixed for it by the meanings of its constituents, and richer in different ways for different such speakings. So what [1] says on a speaking, of given leaves, etc., is not determined by what it, or its parts, mean. (Travis 1997: 91).

Discussing the relevant argument, Bezuidenhout notes that “the mainstream view in philosophy of language is that sentence meaning determines truthconditions” with “a corollary [. . . being] that the truth or falsity of an utterance depends only on what the words mean and how the world is arranged” (2002: 105). Following the colour argument, however, both the mainstream view and its corollary seem to be mistaken, as they fail to provide an explanation for why the first occurrence of ‘The leaf is green’ is true, as it is about the colour in which the leaf has been painted, and the second is false, on the grounds that it is about the leaf’s natural colour. Effectively, this leads her to the conclusion that “meaning undermines truth-conditions” (Bezuidenhout 2002: 105). Suppose now that Pia writes the first token of the sentence under question on a post-it, and later on, while answering her botanist friend’s request on Skype, she is just plain lazy and simply reuses her post-it rather than utter the sentence. In this setting, it would be the same token that is taken to be respectively true and false. Yet, there is no explanation for how, in this scenario, the same token can be true on one occasion and false in another. Basic linguistic 1 This is my own, simplified version of the example of course.

Colour adjectives, compositionality, and true utterances

35

intuitions call for explanations. Unfortunately, none can be found in Travis’ paper. Along the same lines, one can also question the philosophical principle the example is intended to support, namely that natural language sentences have truth conditions determined by factors beyond and above meaning, since this principle is suspiciously silent about what these factors or what truth conditions for specific token sentences are.2 The colour argument raises many important questions, namely: What is the relationship between language and the world? Are truth conditions of sentences compositionally determined? What is the relationship between the truth conditions of sentences and the assessment of truth? What is assessed as true when considering facts? A token of a sentence? An utterance of a sentence? Does it make any difference? With respect to the last question, it seems that the debate on colour adjective sentences sometimes confuses the underlying distinction between tokens and utterances. This is a distinction that can also be traced in the different ways in which Kaplan (1989) and Perry (2012) approach indexicals, with Kaplan’s theory pertaining to the study of token-oriented semantics and Perry’s one pertaining to that of utterance-oriented semantics. Travis seems to be interested in tokens. For instance, in the introduction to Occasion-sensitivity (2008), he insists on discussing sentences in context, which are effectively tokens from the perspective of this paper. That is because, if he were to be considering utterances, he would be taking into account all of their features, like for example, the fact that they are actions made by speakers in a context; yet, he emphasises only the linguistic inscriptions that they leave, that is, tokens. Now, the argument that I wish to pursue in this paper is that the assessment of the truth of a token of a sentence differs from the assessment of the truth of an utterance of that same sentence. All in all, I contend that accepting that the meaning of a sentence determines its truth conditions does not entail that utterances of sentences have truth conditions that are solely determined by meaning,

2 To be fair, Travis’ argument has prompted some suggestions on the semantics of colour terms; yet none offered by him. According to some philosophers and linguists, colour adjectives are indexical expressions, like ‘I’ (cf. Szabo 2001; Rotschild and Segal 2009), or contain variables that are assigned values in a context (cf. Kennedy and McNally 2010; Hansen 2011; but, for a critique, see Clapp 2012). It also fuels truth-conditional pragmatics, the view that “truth conditional content depends on an indefinite number of unstated background assumptions, not all of which can be made explicit” (Bezuidenhout, 2002: 105) or, alternatively, that “the information available in the context that is not semantically encoded by an uttered sentence is relevant for determining the truth conditions of, or what is said, by the utterance’ (Clapp, 2012: 72). In this paper, however, I set these proposals aside and offer my own suggestion.

36

Richard Vallée

or that utterances are assessed as true solely by considering linguistic meaning and facts. In this, I agree with Travis when he states that “truth depends on what words mean, the way the world is, and further factors: aspects of the circumstances in which the words were produced” (Travis 1996: 96); but still, I see no argument against truth-conditional semantics here. Travis seems to be making his point by relying on a semantically idle, common-sense notion of truth that connects language to facts and depends on, but does not affect meaning. If that is correct, it seems to me that the assessment of the truth of tokens or utterances has no impact on basic semantic principles, such as semantic compositionality and truth-conditional semantics. What is more, Travis neither presupposes a specific position on the semantics of colour adjectives, nor does he draw any conclusions about the semantics of colour terms from his example. He simply assumes that colour adjectives, like many adjectives, determine a condition that an object either satisfies or not. Against this backdrop, I will argue that colour terms have specific features, which can be used to explain the relevant data.

1.2 Truth bearers A token of a sentence is a particular object located in space and time. Written and recorded occurrences can be copied and moved in space and time. One can copy an occurrence of one’s score on an exam, or take the paper it is written on back home, much like one can copy a sound signal on a voice recorder, and play it later on at a different location. In this respect, tokens of sentences are structured, complex linguistic entities composed of lexical items. Different tokens of a sentence can be of the same type, like, for example, three tokens of the sentence ‘Whales are mammals’ found in three different books.3 An utterance of a sentence, on the other hand, is a spatio-temporally located, particular event performed by a speaker. Utterances, in the relevant sense, are individuated by their location, the time of the event and their speaker. Once produced, they cannot change location in space or time, that is, utterances cannot be copied or moved to a different location in space, say, next door, or in time, say, the following day. So, one can clearly copy the written or acoustic trace left by an utterance (i.e. a token), but cannot do so for the utterance itself. One can even read and

3 Resorting to the abstract entity of type captures what is common to all these three tokens. However, since sentence type does not play a major role in the present argumentation, I will not analyse the term any further here.

Colour adjectives, compositionality, and true utterances

37

understand a token and know nothing about the utterance it is a trace of. Consider, for example, a token of ‘I am sorry for the scratch on your car’ on a piece of paper you find on the street. You do not know who wrote it, when it was written and even who the addressee was; yet, you understand its linguistic meaning. Similarly, the same token can be used to make many different utterances, like when a token of ‘I will be back in five minutes’ on a post-it on your office door is borrowed by the person in the office next-door in order to be put on his/her door. Finally, semantics assumes that sentences have truth conditions – that a competent speaker knows the truth conditions of a sentence, and that sentences can be true or false. Now, semantics is traditionally about the truth conditions of sentences as type, not about whether, given facts, they are true. If we were interested in the truth assessment of tokens, then we would be interested in the way spatio-temporally located linguistic entities, that is sentences, fit the facts. Given knowledge of meaning and facts, speakers are widely presumed to know when a sentence is true and when it is false. All in all, truth assessment gives an epistemic flavour to the actual discussion, of the type that Borg, for example, calls “creeping verificationism” (2004: 238). Travis suggests that correspondence between sentence tokens and facts is not the relevant relationship when it comes to accounting for the data. His argument shows that the same relatas accept different truth assessments: given the same facts, two tokens of the same non-indexical sentence can be assessed, respectively, as true and as false. In the debate raised by Travis, it is first assumed that the truth conditions of any sentence, except indexical ones, are identical with the truth conditions of its token, and vice-versa. It is also assumed that these are the only relevant truth conditions taken into account in truth assessment. And Travis is well-known for arguing that both of these assumptions are false. Suppose now that it is utterances rather than tokens that are truth assessed. Utterances are not linguistic entities, but rather actions performed using such entities. It is commonly assumed that the truth conditions of sentences are the only truth conditions available in the truth assessment of utterances. I wish to reject this assumption. Semantics is interested in the truth conditions of sentences, that is, in the conditions put on the world if a sentence is to be true. In this respect, semantics does not address the truth assessment of tokens and utterances. It is not interested in whether these conditions are satisfied for a token or an utterance of a sentence, or in whether these conditions are all there is to the truth assessment of tokens or utterances. It rather focuses on the truth conditions of sentences as type. As Travis correctly points out, truth assessment is a very complex and clearly under-examined topic. Given the same facts, one

38

Richard Vallée

could argue that a token can be assessed as true, when considering certain aspects of facts, such as the painted colour of the leaf, and as false, when considering different aspects of facts, namely the natural colour of the leaf. The relevant aspect of facts finds no echo in the uttered token. That being said, the intuitive truth conditions of the utterances in Travis’ examples remain unaccounted for.

2 A new perspective My proposal regarding colour adjective sentences focuses on communication and is utterance-oriented, in the vein of Perry (2012). Emphasising utterances rather than the traces they leave seems natural, since it is utterances that speakers are mostly in contact with in their daily lives. Of course, speakers are also routinely exposed to tokens, without access to their underlying utterance, as in the case of ‘Caution, wet floor’ on a sign on the floor, or that of ‘Leave a message and I will call you back as soon as possible’ on an answering machine, but it is utterances that are predominantly used in face to face communication. My proposal is also multipropositionalist. In multipropositionalism, an utterance of a sentence determines many different propositions, contents, or truth conditions,4 depending on different constraints taken into consideration. Many truth conditions are hence made available as the truth conditions of an utterance,5 as they are all relative to different constraints on this utterance. In contrast, Travis’ view depends on his taking tokens, and not utterances, as primary, and on determining one single content. In doing this, he only takes into account meaning or semantically determined constraints on truth; a point to which I will come back later on. Speakers producing utterances, however, have background assumptions and follow certain principles, like, for instance, Grice’s (1975) conversational maxims: ‘Make your contribution as informative as is required (for the current purpose of the exchange)’, ‘Try to make your contribution one that is true’, and ‘Be relevant’. In his approach, Travis does not investigate the truth assessment of utterances, neither does he exploit background assumptions and conversational maxims; he rather emphasises on the traces left by utterances, that is their tokens, and highlights the context bound, ad hoc nature of the truth assessment of tokens.

4 In this picture, there is no difference between proposition, content, and truth conditions. 5 As Perry puts it, “the concept of ‘truth conditions of an utterance’ is a relative concept” (Perry 2012: 93).

Colour adjectives, compositionality, and true utterances

39

2.1 Compositionality Most semantic frameworks endorse the Principle of Compositionality, which applies to complex expressions as type, and, according to which, “the meaning of an expression is a function of, and only of, the meaning of its parts together with the method by which those parts are combined.” (Pelletier 2004: 133). Yet, the principle of compositionality remains vague: “what counts as a part? What is meaning? What kind of function is allowed?” (Pelletier 2004: 599). In addition, as is the case with most basic principles, its status is controversial. Is it a hypothesis (cf. Szabo 2012) or a methodological principle (cf. Janssen 1997: 419)? Is it a metaphysical principle, valid for any human language; yet, one that cannot be proven? Despite these shortcomings, the principle is backed by strong arguments. There is an infinite number of complex expressions in natural languages. Given now that languages are learnable and an infinite number of new complex expressions, including sentences, are to be understood within them, then all complex expressions in a language must be composed from a finite number of learnable lexical items, each having meaning organised according to the language’s syntactic rules (cf. Davidson 1965). Even so, such arguments remain neutral with respect to what meaning is, and put no specific requirement on it. Also, the overall learning argument is grounded in psychology and the finite capacities of speakers. From this perspective, the notions of sentence and truth have no special role to play in defending compositionality, so Travis does not question this general argument for compositionality. On the truth-conditional semantics paradigm, semantics must give the meaning of an infinite number of sentences as type. In this setting, to give the meaning of a sentence is to give the truth conditions of this sentence, and not those of a token or utterance of it. Following the argument for compositionality, the truth conditions of a sentence are to be given by the lexical items in that sentence only. And since semantics is expected to give the truth conditions of an infinite number of sentences, it must assume that these sentences are composed from a finite stock of lexical items. To give the meaning of an expression is thus to give its contribution to the truth conditions of the sentences in which it occurs (Dowty, Wall and Peters 1981: 7). In this respect, compositionality is primarily compositionality of types. Such an argument endorses a specific notion of meaning of lexical items as truth-conditional elements, or as determinants of truth-conditional elements. Therefore, the truth-conditional version of the principle of compositionality is methodological in nature, as it dictates a strategy to develop semantic theory (cf. Janssen, 1997, 2012), and is not offered as an empirical generalisation. In light of this, Travis’ counter-example targets

40

Richard Vallée

the methodological principle and argues that the meaning of sentences does not determine their truth conditions. According to multipropositionalism now, linguistic meaning is a property of lexical items as type and is something we can learn. Setting semantic ambiguity aside, the linguistic meaning of a lexical item is the same for different utterances of that item. In other words, the linguistic meaning of an expression as type is a rule determining the semantic, meaning-fixed contribution of that expression to the content of utterances that comprise it. Such a perspective on meaning fits the learning argument for compositionality and should be favoured if an utterance of a sentence has a manifold of contents or truth conditions, since a view on meaning as a contribution to the truth conditions of sentences with monopropositionalism in mind is too constraining to explain the facts. Just to give a representative example, it cannot even account for the fact that, even though the sentence ‘I am hungry’ has the same linguistic meaning in different utterances, it can still differ in content or truth conditions from utterance to utterance, being true if I utter it, and false if you do (cf. Perry 2012).

2.2 Truth conditions and sentences Many different notions of truth can be found in philosophy. Sometimes truth is treated as a metalinguistic notion, with no connection to facts, as in the case of Tarski (see Burgess and Burgess 2011) who uses this notion in T sentences: ‘s’ in L is true if and only if p,

where s is a sentence of the object-language L, and p is a copy of s in the semantic metalanguage; with truth applying to the metalanguage of semantics, and effectively linking a sentence from the object-language with one from the metalanguage. Also, as Davidson (1967) contends, ‘is true if and only if ’ in this context can be replaced with ‘means that’, and vice-versa. In this respect, T sentences give a theory of meaning for any language, where the sentence s in the object-language is given a structural description, and the sentence p in the metalanguage gives the truth conditions of the sentence s. In this respect, the assignment of truth conditions in T sentences is insensitive to extralinguistic intuitions and facts. Obviously, Tarski does not pay attention to indexical sentences, and offers a theory of sentences as type, so T sentences do not tell you whether a sentence is true after considering facts. Truth conditions are also assigned in semantics to moral evaluations, like ‘Killing is wrong’, and aesthetic judgements, like ‘Bardot is beautiful’, to use Davidson’s (1967) famous example. Again, it is never argued that such sentences

Colour adjectives, compositionality, and true utterances

41

are true or false, when confronted with facts, because contending that they determine truth conditions – realised or not – and are true or false depending on facts is a strong and controversial position, but nonetheless a side issue. All in all, the semantic notion of truth does not connect sentences as type to facts. So, Travis raises an issue concerning the truth assessment of tokens. On the other hand, multipropositionalism is utterance-oriented and gives conditions on the truth of utterances as opposed to conditions on the truth of token sentences.

2.3 Truth conditions and utterances In the multipropositionalist framework, the meaning of a sentence as type is not a truth valuable entity. Rather, it is the contents of utterances that are – and by contents we mean the conditions put on the world for the utterance to be true. From this perspective, semantics is interested in how the linguistic meaning of expressions contributes to the different contents of utterances in which they occur and places constraints on the truth of these utterances. It first emphasises contents determined only by meaning, or semantic constraints on the truth of utterances, and known a priori by competent speakers. Semantically determined contents are mechanically extracted from the lexical items of the sentence as type and faithfully echo sentences. Perry (1988) assigns such contents the role of the cognitive significance of the utterance. A semantically competent speaker understands and can accept as true the cognitive significance of an utterance without any knowledge of the world. For example, given her semantic competence only, a speaker knows that an utterance of ‘Birds have wings’ is true if and only if birds have wings. Every utterance of that sentence has the same mechanically extracted semantic content and, ipso facto, the same cognitive significance. For utterances of some sentences, like, for instance, proper name and indexical sentences, contents in the role of cognitive significance are not constant. Here, the linguistic meaning of ‘I’ as type is philosophically interesting because it is known a priori and allows for a generalisation concerning different utterances of ‘I’; it determines a content constituent: the speaker of the utterance. Such a constituent contains the utterance itself as a constituent, and gives different content constituents for different utterances. As a consequence, it introduces different identifying conditions into the semantically determined content of utterances enabling different utterances of ‘I’ sentences to exhibit different cognitive significance. In this setting, semantics is interested in the way ‘I’ contributes to the truth conditions of any utterance, not in its actual contribution to a specific one. Consider, for example, an utterance u of ‘I am short’, and its semantic, context-independent contribution to content.

42 (1)

Richard Vallée

Given that the sentence is in English, the utterance u of ‘I am short’ is true if and only if the speaker of u is short.

(1) supposes linguistic meaning only, and connects an utterance of a sentence to its semantically determined content. It thus gives semantic conditions or constraints on the truth of the utterance u. A different utterance of the same sentence would have different semantically determined truth conditions or constraints because it is a different utterance. Let’s now turn to an utterance of ‘The leaf is green’ and consider the linguistic constraints on its truth. (2)

Given that the sentence is in English, the utterance u of ‘The leaf is green’ is true if and only if the leaf is green.

In this case, the adjective ‘green’ is context-insensitive. So, all utterances of ‘The leaf is green’ have the same semantically determined content that a speaker knows a priori, as well as the same cognitive significance. Here, what (2) makes explicit is the meaning determined conditions or constraints on the truth of any utterance of ‘The leaf is green’. Strictly speaking, semantics stops at what is known a priori by linguistically competent speakers. Extralinguistic facts do not impact semantics, the compositionality of the meaning of sentences as type, and the semantically determined content or truth conditions of utterances. Since facts are not taken into account, semantically determined content is not the content which is assessed as true. So, Travis is right in thinking that assessing the truth of a token or an utterance require going beyond linguistic meaning and the semantic contents it determines by considering facts. However, assessing a token or an utterance as true when considering facts presupposes semantics, semantically-fixed truth conditions, and compositionality, as this goes well beyond the semantic project.

2.4 Official content Let’s now go back to my utterance of ‘I am short’. As already mentioned, ‘I’ is sensitive to features individuating the utterance. So, my utterance of ‘I’ designates me as the speaker of the utterance. All in all, once facts about the utterance are taken into account, (3) seems to hold. (3)

Given facts about the utterance, the utterance u of ‘I am short’ is true if and only if RV is short,

Colour adjectives, compositionality, and true utterances

43

where RV is the actual speaker of the utterance himself. These truth conditions, i.e. RV is short, result from semantically determined content and utterance individuating features, with the reflexive meaning of ‘I’ automatically selecting the relevant element. Such content, however, is not utterance reflexive; on the contrary, these truth conditions are utterance independent, in that they are realised whether or not there is an utterance – in other words, I am short whether or not I produce an utterance of ‘I am short’. Perry calls this the official content of an utterance; a content that includes constraints on the truth of utterances which are obtained after considering meaning and facts about the utterance. This official content is what is said in perhaps an artificial sense of ‘say’: you and I do not say the same thing when utterring ‘I am short’, as our utterances have different official contents. Of course, what the linguistic meaning of ‘I’ determines when taking into account the facts individuating the utterance is important in linguistic communication, but much less relevant from a semantic point of view. Semantics is first and foremost about the factors that determine ‘what is said’, rather than about ‘what is said’ per se. In this respect, the official content of utterances is a plausible candidate for the truth assessment of utterances, since it is by considering facts that my utterance can be true, and yours false. Let’s say then that in assessing the truth of an utterance, what is considered is the official content of that utterance, and whether or not it fits the facts. In this way, monopropositionalism is abandoned, yet truth conditions are preserved. Indexical utterances introduce at least two different contents that are distinguished by the factors that are taken into account when making these contents explicit. One has the role of the cognitive significance of the utterance and emphasises what is understood by relying on linguistic meaning only; the other has the role of the official content of the utterance, and is obtained once facts about the utterance are taken into account. This picture can generalise to utterances of any sentence, including those of non-indexical sentences. Such indexical motivated arguments support the multipropositionalist view, according to which, an utterance determines a manifold of contents, rather than a unique one. From this perspective, identifying the meaning of a lexical item with its contribution to truth conditions is an oversimplification, since meaning is given by rules, not truth-conditional constituents. Moreover, different contents or truth conditions with different roles take centre stage; focusing on one single content is misleading, as it ignores what is taken into account when making an utterance’s contents explicit, and essentially confuses their different roles. Let’s go back to the utterance of ‘The leaf is green,’ and consider its official content. ‘The leaf ’ is a description, and descriptions are referring terms introducing objects into contents (Perry, 2012, chapter 5). Setting aside the fact that

44

Richard Vallée

this description is incomplete,6 the most plausible referred to object here is THE LEAF. (4) Given facts about the utterance, the utterance u of ‘The leaf is green’ is true if and only if THE LEAF is green. THE LEAF is the object referred to itself. (4) gives us what the speaker said in our artificial sense of ‘said’. Since there are no context sensitive terms in the sentence, a different utterance of the same sentence talking about the same leaf would have the same cognitive significance, as well as the same official content. Following Travis (1997), two utterances of colour sentences can differ in their truth assessment even when they share the same cognitive significance and official content. So, based on Travis’ argument (1997), one could at most conclude that the content as cognitive significance as well as the official content of the utterance is the same in both cases, but that the content assessed as true differs in each utterance. The content as cognitive significance cannot be modified without altering the meaning of an utterance since it only depends on this meaning; thus, arguing that the linguistic meaning of the two utterances differs lacks plausibility. And indeed, Travis does not deny that the tokens or the utterances have the same linguistic meaning and hence, in the present picture, the same semantically determined content. Moreover, as we have already seen, the cognitive significance of an utterance is independent of facts as well as what satisfies our intuitions on the truth of utterances of a sentence having that cognitive significance. Similarly, the official content of an utterance is also meaning dependent, constrained this time by facts about the utterance and truth assessed. Still, it can exist without the utterance. The leaf has the property of being green whether or not one utters ‘The leaf is green’. In the present case, the utterance’s official content, or what is said, is the same for both utterances because the facts about the utterance, and the facts tout court, do not differ. For an uneducated ear, in Travis’ example, Pia said the same thing twice. However, prima facie, official content, or ‘what is said’, does not capture fine-grained intuitions about the truth assessment of each utterance. In this I agree with Travis, who argues that Understanding requires sensitivity. Understanding words consists, in part, in sensitivity to how they fit with the circumstances of their speaking. Part of that is sensitivity to how they need to fit in order to be true. (Travis 1996: 102).

6 Due to space restrictions, I will not address issues related to incomplete descriptions in the present paper.

Colour adjectives, compositionality, and true utterances

45

So, the speaker, Pia, may have said the same thing twice, but, intuitively, this is not what is assessed as true or false. The meaning of a sentence and facts about the utterance do not always give all the relevant conditions or constraints needed to assess whether an utterance of that sentence is true. Yet, we want to preserve both the intuitions that (a) the two utterances have the same cognitive significance and (b) that, on both occasions, the speaker says the same thing or used a sentence that puts identical constraints on the truth of her utterances; neither the cognitive significance of the utterance nor what is said need be modified. The problem, however, remains: what Pia says is not enough to assess the truth of her utterances. So, the difference must be located elsewhere.

2.5 True utterances As Ludlow (1989) discusses, the semantically-determined truth conditions of sentences with relative gradable adjectives, such as ‘tall’ or ‘strong,’ do not, prima facie, give the truth-assessed content of their utterances. An utterance of ‘Peter is tall’ can be assessed as true when considering 5-years olds, while another utterance of the same sentence can be assessed as false when considering basketball players. All in all, an utterance of ‘Peter is tall’ is assessed as true only after considering a potential comparison class, and, more specifically, the comparison class that the speaker had in mind when producing her utterance. So, the relevant comparison class is important as far as truth assessment is concerned. But, in this paper, I do not want to add to the already important literature on relative adjectives. I am mentioning them, because a similar phenomenon is at work as far as the utterances of a colour sentence are concerned. As Lahav notes, expressing a widely shared intuition, when writing about ‘red’, In short, what counts for one type of thing to be red is not what counts for another. Of course, there is a feature that is common to all the things which count as (nonmetaphorically) red, namely that some part of them, or some item related to them, must appear wholly and literally reddish. But that is only a very general necessary condition, and is far from being sufficient for a given object to count as red. (Lahav 1989: 264).

According to common sense, colour adjectives, like ‘green’, semantically determine conditions, such as be green, that apply to objects. Yet, these conditions can apply to part of an object, or the whole object, and are generally very plastic in that respect. This feature does not give rise to problems for the semantically determined truth conditions of utterances: in ‘the leaf is green’, ‘green’ determines the condition of being green. Hence, it does not affect the compositional

46

Richard Vallée

nature of colour adjective sentences.7 But it does raise problems for the truth assessment of colour sentence utterances. Which surface or part of an object is relevant for the utterance at hand? Which surface or part of an object is the speaker talking about? With no additional specification about the relevant object, trying to assess a colour sentence token or utterance as true is a failed project. All semantically competent speakers are familiar with this aspect of colour adjectives: Hearing a speaker say that she is looking for a red house, you might ask which part of the house is red – Is it the bricks? The roof? The inside walls? Colour adjectives are intersective like this. In this respect, ‘The leaf is green’ is equivalent to ‘There is a leaf and it is green,’ with ‘green’ giving a feature of the object at hand. However, it does not give any information on the dimension or aspect of the object that it is a condition of. Following Lahav’s intuition, the utterance of ‘The leaf is green’ does not qualify as true, unless specifications or additional conditions on the relevant object are added. In this, he is right. Colour attribution to an object is sensitive to aspects of the object that the speaker is interested in – Is the speaker talking about part of the object? Which part? About the whole object? About a normal aspect of the object? So, when assessing the truth of a colour sentence utterance, background assumptions about the speaker as well as constraints on the utterance’s relevance, informativity and truth are to be taken into consideration. Suppose now that another speaker of ‘The leaf is green’ is a biologist interested in the biology of leaves. Suppose further that she is talking about the leaf Pia was talking about, in which case, her utterance has the same cognitive significance, the leaf is green, as well as the same official content, THE LEAF is green, as Pia’s earlier utterance(s). The view that all utterances of sentences have their own truth conditions, especially conditions relevant for their truth assessment, fully semantically determined is questionable. That is why I will use the term subsemantic constituent relevant to the truth of the utterance, a term that underlines the fact that such constituents are not part of the truth conditions of the sentence, as they are not determined by linguistic meaning, but are still part of the conditions on the truth of the sentence’s utterances. Conditions, like being a natural colour, can be such subsemantic constituents. Prima facie, there is an extra, not meaning-determined but rather background-and-maxim-constrained condition put on the truth of the biologist’s utterance of ‘The leaf is green;’ given our background assumptions about the speaker and the supposition that she is following conversational maxims, her utterance is most likely about the natural colour of the leaf at hand. In this respect, the official content of this utterance needs to be refocused on the basis of such subsemantic constituents 7 On this point, I disagree with Lahav’s position of colour adjectives.

Colour adjectives, compositionality, and true utterances

47

so as to fit the intuitions about the truth assessment of the utterance. But this refocusing of the official content is meaning independent, ad hoc, and utterance specific, since it concerns truth assessment, filling the slack left by linguistic meaning between utterances and facts, whatever the latter may be. Being solely a feature of the truth assessed content of the utterance, the subsemantic constituent does not depend on the sentence, nor does it depend merely on facts. In my view, it is speaker-background relative and maxim-constrained. With this in mind, let’s go back to the biologist’s utterance. (5)

Given conversational maxims and background assumptions about the speaker of u, the utterance u of ‘The leaf is green’ is true if and only if THE LEAF is green and green is its natural colour.

or ‘if and only if THE LEAF is naturally coloured green.’8 The refocused conditions on the truth of the utterance contain a subsemantic constituent, a condition like being its natural colour, which is provided for the specific utterance on the basis of background assumptions and maxims. In this picture, compositionality is not called into question for the uttered sentence, nor does the supplemented material impose semantically-determined conditions on the truth of the utterance. A different utterance, u’, given by a different speaker and in a different background, or even by the same speaker but in a different background, could be about the painted colour of the leaf, with being painted green being the relevant constituent. A speaker unable to determine any subsemantic constituent for her utterance of ‘The leaf is green’ would not be able to explain why her utterance is true, or even what it is that makes her utterance true. Conversely, hearers unable to identify the refocused conditions on the truth of an utterance of the aforementioned sentence would not be linguistically at fault, but rather unclear about the speaker’s background assumptions. Admittedly, this view can account for the fact that different utterances of the same token of ‘The leaf is green’ can be respectively true and false, since each utterance’s official content could depend on the contribution of different subsemantic constituents. In Pia’s example that was presented at the beginning of this paper then, if both utterances were intended to have the same subsemantic constituent, being its painted colour for instance, both of them would be true, but the second would be irrelevant because the biologist would, obviously, be interested in the natural colour of the leaf. If the second utterance is about the 8 Such a small variation is tolerable and does not affect my main point. I also disregard and put aside other aspects of the colour attribution – such as “Completely green?”, “Partly green?”

48

Richard Vallée

natural colour of the leaf, however, then it is false. There is an utterancegrounded explanation for this difference in the truth assessment of ‘The leaf is green’. Going back to the utterance by the biologist, it is the case that she did not say anything false; yet, the refocused, speaker-dependent, official content of her utterance, which is the relevant content in the utterance’s truth assessment is false: the leaf’s natural colour is not green. Using her own background assumptions and following the conversational maxims, she wrongly assesses her utterance as true, but her mistake has nothing to do with the meaning of the sentence or its semantically-determined truth conditions. It just ensues from her background assumptions concerning the leaf.

3 Concluding remarks In this paper, I have argued that, given background assumptions and conversational principles, the official content of utterances can be refocused. The truth of an utterance now is assessed on the basis of refocused content and facts, but this assessment is made using a non-semantic, common-sense notion of truth that connects utterances and facts. That is because linguistic communication is flexible. Semantics, on the other hand, does not need to assign to utterances linguistic meaning that connects them to fine-grained, structured conditions in order to account for intuitions about their truth. The problem of identifying under which conditions, beyond those semantically provided, the utterance is true is a familiar one. However, it is not a semantic issue, as it has nothing to do with meaning and semantically determined truth conditions. In concluding that meaning undermines truth conditions and in drawing dramatic consequences for the semantic program, Travis talks about the semantically-determined truth conditions, or the cognitive significance, of the utterance. Yet, he is actually addressing the official content of utterances, and this view has quite different consequences. Travis’ point clearly questions the view that assessing the truth of an utterance depends solely on meaning and facts. However, this view is not part of the core of semantics. If I am correct, the truth assessment of utterances may sometimes demand the addition of subsemantic truth-conditional components to the official content of an utterance, but it does not necessitate the addition of meaning-determined truth-conditional material, or the addition of material in a systematic way that justifies a generalisation on sentences and utterances. Semanticists are wrong in adopting a view according to which the meaning of a sentence as type determines truth conditions that perfectly mirror facts and always fit the truth assessment of the sentence’s utterances, as in so doing, they

Colour adjectives, compositionality, and true utterances

49

ignore the complexity of the assessment of the truth of utterances and wrongly reject semantically innocuous alterations. By assuming that the assessment of the truth of utterances depends only on meaning and facts, Travis mislocates the impact of counter-examples for colour-term sentences on semanticallydetermined truth conditions, with disastrous consequences for truth-conditional semantics. Ignoring meaning as a rule, he loses the focus on rules assigned to expressions as type. What is more, emphasising tokens, he neglects utterances and supposes that the truth conditions of a sentence are always identical to the truth conditions of its utterance. Missing the distinction between cognitive significance and official content, he confuses both roles and fails to pinpoint the impact of the problems raised by the truth assessment of colour adjective sentences’ official content, rather than their cognitive significance, and the ad hoc fix available for this official content. Often, the truth of the token left by an utterance is assessed when the utterance is no longer available. What is then considered is the utterance’s official content. Official contents are sometimes stable and unresponsive to utterances, like the official content of tokens of ‘Whales are mammals,’ all utterances of which have the same official content and need no subsemantic constituent to be assessed as true. Sentences like this were once taken as paradigmatic in semantics. Utterances of indexical sentences, like ‘I am short’, on the other hand, do not all have the same official content, but their difference can be traced back to the meaning and utterance features. Such utterances are not truth assessable unless information about the utterance is obtained. The official content of an utterance of a non-indexical sentence is also sometimes insufficient for the truth assessment of the utterance. This content though is responsive to utterances by way of background assumptions and conversational maxims, as in the case of the official content of an utterance of ‘The leaf is green’. In this case, official content and truth-assessed content differ, but now the difference in the truth assessed content of the utterance is meaning independent. All in all, when we ignore the source utterance of a token, its official content is sometimes impossible to assess as true or false. Let me change the example. You find a note in an old book, which says ‘The apple is red’, and you do not know who wrote it or when it was written. Is the speaker talking about the skin or the flesh of the apple? Whether this utterance is true or false depends on background assumptions and facts. Not only are the facts out of your reach, but ignoring the speaker’s background assumptions and relying on the utterance’s official content alone, you cannot identify the assessed, refocused official content of this particular utterance. Suppose now that Anne’s son is selecting apples for cooking, when he says ‘Here is a red apple’ (Bezuidenhout 2002: 107). Is this utterance true? What is the adjective applied to, the skin or flesh of the

50

Richard Vallée

apple? Which condition should be supplemented in this utterance’s refocused content? Given his background assumptions, we surmise that his utterance is about the skin. So, if the apple has red skin, then the utterance is true. In different circumstances, with a different background, however, the utterance’s official content could be about the flesh of the apple. To sum up, my view deploys many different contents and preserves the idea that the semantically-determined contents of colour sentences are not modified to fit the assessment of the truth of utterances. The only altered content, the official content, is refocused on a subsemantic condition on the basis of background assumptions and principles of conversation. In this approach, colour adjectives provide no reason to question compositionality or truth-conditional semantics in a multipropositionalist framework. Some see the supplemented subsemantic constituent as semantically relevant. This is a path that semanticists need not follow.

References Allwood, Jens, Lars-Gunnar Anderson & Osten Dahl. 1977. Logic in linguistics. Cambridge: Cambridge University Press. Bezuidenhout, Anne. 2002. Truth-conditional pragmatics. Language and Mind 16. 105–134. Borg, Emma. 2004. Minimal Semantics. Oxford: Oxford University Press. Burgess, A. G. & John P. Burgess. 2011. Truth. Princeton and Oxford: Princeton University Press. Clapp, Lenny. 2012. Indexical color-predicates: Truth-conditional semantics vs. truth conditional pragmatics. Canadian Journal of Philosophy. 2012. 71–100. Davidson, Donald. 1965. Theories of meaning and learnable languages. In Yehoshua Bar Hillel (ed.), Proceeding of the 1964 International Congress for Logic, Methodology and Philosophy of Science, 383–394. Amsterdam: North-Holland Publications. Davidson, Donald. 1967. Truth and meaning. Synthese 17. 304–323. Dowty, David, Robert E. Wall & Stanley Peters. 1981. Introduction to Montague semantics. Dordrecht: Reidel. Grice, H. Paul. 1975. Logic and conversation. In Peter Cole & Jerry L. Morgan (eds.), Syntax and semantics 3: Speech acts, 41–58. New York: Academic Press. Hansen, Nathaniel. 2011. Color adjectives and radical contextualism. Linguistics and Philosophy 34. 201–221. Janssen, Theo M.V. 1997. Compositionality. In Johan van Benthem & Alice Ter Meulen (eds.), Handbook of logic and language, 417–473. Amsterdam: Elsevier. Janssen, Theo M.V. 2012. Compositionality: Its historic context. In Markus Werning, Wolfram Hinzen & Edouard Machery (eds.), The Oxford handbook of compositionality, 19–46. Oxford: Oxford University Press. Kaplan, David. 1989. Demonstratives. In Joseph Almog, John Perry & Howard Wettstein (eds.), Themes from Kaplan, 481–563. New York: Oxford. University Press. Kennedy, Christopher & Louise McNally. 2010. Color, context and compositionality. Synthese 174. 74–98.

Colour adjectives, compositionality, and true utterances

51

Korta, Kepa & John Perry. 2011. Critical pragmatics. An inquiry into reference and communication. Cambridge: Cambridge University Press. Lahav, Ran. 1989. Against compositionality, the case of adjectives. Philosophical Studies 57. 264–279. Ludlow, Peter. 1989. Implicit comparison classes. Linguistics and Philosophy 12. 519–533. Pelletier, Francis Jeffrey. 1994. The principle of semantic compositionality. Topoi 13. 11–24. Perry, John. 1988. Cognitive significance and new theories of reference. Noûs 22. 1–18. Perry, John. 2012. Reference and reflexivity. Stanford: CSLI Publications. Rothschild, Daniel & Gabriel Segal. Indexical predicates. Mind & Language, 24. 467–493. Sainsbury, Mark. 2001. Two ways to smoke a cigarette. Ratio 14. 386–406. Szabo, Zoltan Gendler. 2001. Adjectives in context. In Robert M. Harnish & Istvan Kenesei (eds.), Perspectives on semantics, pragmatics, and discourse. 119–146. Amsterdam: John Benjamins. Szabo, Zoltan Gendler. 2012. The case for compositionality. In Markus Werning, Wolfram Hinzen & Edouard Machery (eds.), The Oxford Handbook of Compositionality, 64–80. Oxford: Oxford University Press. Travis, Charles. 1994. On constraints of generality. In Proceeding of the Aristotelian Society 94. 165–188. Travis, Charles. 1996. Meaning’s role in truth. Mind 105. 451–466. Also in Travis, Charles. 2008. Occasion-sensitivity, 94–108. Oxford: Oxford University Press. Travis, Charles. 1997. Pragmatics. In Bob Hale & Crispin Wright (eds.), A Companion to the philosophy of language, 87–107. Oxford: Blackwell. Also in Travis, Charles. 2008. Occasion-sensitivity, 109–129. Oxford: Oxford University Press. Vicente, Agustin. 2012. On Travis cases. Linguistics and Philosophy 35. 3–19.

Kepa Korta

4 Talk about the future Abstract: In a recent paper (Korta 2015), I compared assertions about the contingent future with what speech act theory calls commissive and directive speech acts, which are also about contingent acts situated in the future. I concluded that those assertions are closer to commissives and directives than to assertives. Contrary to what standard speech act theory assumes, they should all share a common direction of fit between the propositional content of the utterance and the world: world-to-words or upward (↑). That would explain why the indeterminacy intuition, as MacFarlane (2003) calls it, arises in our estimates of the future. I now think, however, that my point about the upward direction of fit of predictions is not correct. The aim of this paper is to discuss this and to correctly identify what all predictions – assertive, commissive and directive alike – have in common: a propositional content concerning the future. Being about the future presents common specific problems for their propositional contents. I discuss these problems and their solutions. Prediction is very difficult, especially about the future. Niels Bohr, Danish physicist (1885–1962).

1 Introduction Most philosophical papers about future contingents that discuss the issue raised by Aristotle in his Peri Hermeneias, and revived more recently by John MacFarlane in his ‘Future contingents and relative truth’ (2003), focus on predictions about, say, meteorological or nautical events, leaving aside statements about future actions of the speaker or the addressee. Using Anscombe’s (1957) terms, one would say that they discuss estimates of the future rather than expressions of intention; or, using Searle’s (1969, 1975) taxonomy of speech acts, that their Acknowledgements: I am grateful to Josu Acosta, Eros Corazza, María de Ponte, Joana Garmendia, John Perry, and Larraitz Zubeldia for their comments and suggestions. I also thank Stavros Assimakopoulos for his patient help. The work was partially supported by grants of the Spanish MINECO (FFI2012-37726; FFI2015-63719-P) and the Basque Government (IT780-13). Kepa Korta, Institute for Logic, Cognition, Language and Information (ILCLI). University of the Basque Country (UPV-EHU) DOI 10.1515/9781501505089-004

54

Kepa Korta

apparent interest is on assertive speech acts and not on commissives and directives, even if they all involve the contingent future. In a recent paper (Korta 2015), I compared these three kinds of utterances, and against mainstream speech act theory, I claimed that estimates of the contingent future are closer to commissives and directives than to assertives. Thus, all statements about the contingent future, including commissives and directives, share the world-towords (or upward, ↑) direction of fit. That explains why the indeterminacy intuition, as MacFarlane (2003) calls it, is present in our estimates of the future. Unlike assertions about the past or the present, estimates of the future are not just the sort of speech act that is true or false. They are more like promises or requests, which are not true or false, but fulfilled or not fulfilled, accurate or inaccurate. Now I think that the point about the upward direction of fit of predictions is not correct. I incorrectly identified the relation of direction of fit between the propositional content of the speech act and the truth-maker of the speech act – the relation between ‘words’ and ‘world’ – with a temporal relation. As I took it, if the truth-maker of the propositional content lies in the past or the present with respect to the time of the utterance, then the direction of fit is words-toworld (or downward, ↓); and, if the truth-maker of the propositional content lies in the future of the utterance, the direction of fit is world-to-words (↑). But direction of fit, in its original presentation by Anscombe (1957) or in standard speech act theory (Searle 1969, 1975; Searle and Vanderveken 1985; Vanderveken 1990), is not a temporal relation but a causal one. Still, even if estimates of the future, unlike commissives and directives, are like assertions on their direction of fit, they are not true or false at the time of their utterance, and this is because, like commissives and directives, their propositional content is about the future. Being about the future, they present common specific problems regarding the notion of propositional content. The main aim of this paper is to consider this notion, which is in need of revision for two reasons. First, because, as Korta and Perry (2007, 2011) argue, in speech act theory, propositional content is required to play two incompatible roles. On the one hand, it should represent what speech acts with different illocutionary force may have in common; on the other hand, it should meet different propositional conditions that are specific to various illocutionary forces. And second, because, from commonsense views about the existence of objects and events in time, our predictions (including estimates of the future as well as expressions of intention) could lack not only a truth-value (as the indeterminacy intuition suggests), but even a full propositional content. I argue that Korta and Perry’s content-pluralism (2007, 2011, 2013) offers a natural account of these issues, according to which, our predictions lack a truth-value at the time of utterance,

Talk about the future

55

but do have full truth-conditions, even when they concern inexistent events and objects. In those cases, predictions lack referential truth-conditions, but they do have utterance-bound truth-conditions. The structure of the paper is the following. In the next section, I present a summary of the classic dilemma about future contingents. Section 3 sums up the relevant notions of speech act theory, necessary to understand the subsequent discussions about the direction of fit of predictions (section 4), the propositional content of speech acts in general (section 5) and the propositional content of predictions in particular (section 6). I finish with some brief concluding remarks in section 7. Before proceeding, however, it is worth adding various preliminary caveats about the scope of the paper. First, this paper deals with literal statements or utterances of simple declarative sentences about events and objects that lie in the future, and not with the semantics and pragmatics of the future tense – which, incidentally, according to most linguists, is not even properly a tense. Second, their truth (or falsity) must be contingent and not determined by facts existing at the time of the utterance. Thus, only (1) to (5) below fall into the scope of the paper. (1) to (4) need no clarification. (5) is in, because, even if it uses the present tense, it concerns a future and contingent state of affairs. (6) is out if we interpret it not as concerning the future but rather a present and uncertain state of affairs. (7) does not concern a future state of affairs but an eternal truth, so it is clearly out; as is (8), which though contingent, is determined by a past fact, my birthday (and supposing I am still alive next year). (1) (2) (3) (4) (5) (6) (7) (8)

Tomorrow it will rain in Donostia. There will be a sea battle tomorrow. I will leave the room. You will leave the room. Your mother arrives tomorrow. Your son will be alright. Two plus two will equal four. Next year I will be 51 years old.

For convenience, all statements about the future contingent considered throughout the paper, including estimates of the future and expressions of intention, i.e. assertives, commissives and directives alike, will be called “predictions.” Finally, it is worth noting that no underlying strong assumption is made regarding the metaphysics of time. The only assumptions are that time exists, that moments (instants or intervals) are objects that can be referred to and thought about, in the sense that they can be constituents of our propositional contents, and that the future is open, that is to say, that past and present facts

56

Kepa Korta

do not wholly determine what is going to happen in the future. It is not only that we do not know what is going to happen, but that the future is open.

2 The dilemma of future contingents In his celebrated ‘Future contingents and Relative Truth’ (2003),1 John MacFarlane revisits Aristotle’s classic issue regarding the truth of estimates of the future. Consider again the following utterances (renumbered below for convenience): (1a) Tomorrow it will rain in Donostia. (2a) There will be a sea battle tomorrow. Uttered today, we have the intuition that they are neither true nor false. This intuition is associated with a view of the world that takes it to be non-deterministic. Focusing on (2a), if it were true today, then, no matter what we do, a sea battle will happen tomorrow. It is commonly assumed, however, at least with regard to events like sea battles, that the future is “open,” that there is something we can do to alter the course of events, that is, something we can do today to avoid (or to promote) the occurrence of the battle tomorrow.2 This inevitably leads one to consider that today (2a) is neither true nor false. MacFarlane calls this the ‘indeterminacy intuition.’ Now, suppose the following day someone is assessing the facts. In MacFarlane’s words: Sally is hanging onto the mast, deafened by the roar of the cannon. She turns to Jake and says ‘Your assertion yesterday turned out to be true’. Sally’s reasoning seems unimpeachable: Jake asserted yesterday that there would be a sea battle today There is a sea battle today So Jake’s assertion was true. When we take this retrospective view, we are driven to assign a determinate truth-value to Jake’s utterance: this is the determinacy intuition. (MacFarlane 2003: 324–5)

1 Awarded the Philosophical Quarterly Essay Prize 2002, this paper has had a great impact in the literature on future contingents, and relative truth applied to problems of ‘faultless’ disagreements, aesthetic judgments, epistemic modals, etc. 2 With regard to the weather, we might think that there is no room for indeterminacy, but global climate change seems to show that the intervention of humans can decisively affect the future.

Talk about the future

57

We can make the same point in a slightly different way. Suppose Jake utters (2a) on day 1, referring with ‘tomorrow’ to day 2. Then it seems that Sally would say the same thing when uttering (2b) on day 2, or (2c) on day 3: (2b) There is a sea battle today. (2c) There was a sea battle yesterday. Hearing (2b) or (2c) from Sally, we would be naturally inclined to admit that what she said was either true or false. But if what she said is true (or false), and she said the same thing (i.e. expressed the same proposition) as Jake, then what Jake said uttering (2a) must also be true (or correspondingly false). Put this way, the determinacy intuition would be directly about (2b) and (2c), and indirectly, via a different intuition of same-saying, about (2a). Anyhow, we would be compelled to assign a truth-value to (2a), and this, according to MacFarlane, would go against the indeterminacy intuition: If Jake’s utterance is neither true nor false, as the indeterminacy intuition demands, it is not true and it is not false. But the determinacy intuition demands that it must be one or the other. (MacFarlane 2003: 327)

It is interesting that, in this context, estimates of the future like (1a) and (2a) have rarely been compared with other utterances about the future, like (3) and (4), that can be taken to be not only estimates of the future but also expressions of intention, or, using speech act theoretic jargon, a commissive (a promise, for instance) and a directive (e.g. an order), respectively. I contended elsewhere (Korta 2015) that the comparison sheds some light on the competing intuitions on future contingents, and I still think it does. But I think I was wrong in taking estimates of the future to be non-assertions. I reconsider this in section 4, but, before I do so, I need to introduce some basic notions of speech act theory.

3 Basic notions of speech act theory An early lesson from John L. Austin (1961, 1962) was that asserting that a certain state of affairs is real is one among many things we can do with utterances. We can get married, baptise a ship, promise to finish a paper, bet 20 euros that our favorite team will win the match, etc. John Searle (1969, 1975) developed Austin’s insights into what is known as speech act theory, according to which, in a speech act, we should distinguish between its illocutionary force F and its propositional content p. In the right circumstances, the following utterances would share their propositional content but differ in illocutionary force:

58

Kepa Korta

(9)

John left the room.

(10) [=3] I will leave the room [uttered by John]. (11)

John, leave the room!

(12)

I wish John would leave the room.

The propositional content p contains John, a certain room, and the action of leaving the room at a particular time t.3 But each utterance constitutes a speech act with a different illocutionary force F that belongs to a different class of illocutionary act. According to Searle (1975), illocutionary points serve to build the taxonomy of speech acts into five classes: assertives, commissives, directives, declaratives and expressives. Each illocutionary point determines a certain sincerity condition, a direction of fit,4 and certain conditions of satisfaction, that together with some other elements of the speech act like the preparatory conditions, the mode of achievement, or the degree of strength, determine the illocutionary force of the speech act. An assertion like (9) illustrates the assertive illocutionary point, which consists in representing a state of affairs as real. That’s what the speaker minimally intends when making an assertion, according to speech act theory. Besides, the sincerity condition establishes that the speaker believes that the state of affairs is real, i.e. that the propositional content of the speech act is true. That’s the mental state she expresses in producing a speech act with assertive illocutionary point. The direction of fit of an assertion is words-to-world (or downward, ↓), that is to say, the propositional content is supposed to fit reality, which is supposed to be causally independent of the speech act. The conditions of satisfaction of an assertive are just its truth-conditions: an assertive illocutionary act is satisfied if and only if its propositional content is true. A promise like (10) is the standard commissive act, in which the speaker commits to perform a future action. If he’s sincere, he intends to perform the action represented by the propositional content. In this case, the direction of fit is world-to-words (or upward, ↑), i.e. it is the world that is supposed to fit the words; and the change in the world is causally dependent on the propositional 3 At this point, I am assuming that propositional contents are non-relative regarding time, and adopt a fairly orthodox notion of propositional content. In sections 5 and 6 that follow, we will see that this notion needs to be refined and substantially changed to meet the requirements of speech act theory. 4 The term ‘direction of fit’ was coined by Austin (1953) but was made popular by Anscombe (1957) and especially Searle (1969, 1975). For a systematic and formal presentation of these notions, see Searle and Vanderveken (1985) and Vanderveken (1990).

Talk about the future

59

content, as the speech act will be satisfied if and only if the speaker performs the action because he committed to do it. (11) is an order, the paradigm of the directive illocutionary point. Here, the speaker tries to make the addressee do the act represented by the propositional content, and she is sincere if she desires the addressee to do so. The direction of fit is again world-to-words (or upward, ↑), i.e. it is the world that is supposed to fit the words; and the change in the world is causally dependent on the propositional content, because the speech act will be satisfied if and only if the addressee performs the action with the intention to comply with the order. (12) is an expressive speech act, but for the sake of simplicity, I will ignore expressives here, as I will also ignore speech acts with the declarative point, which roughly correspond to explicit performatives. For my purposes in this paper, there is another component of the illocutionary force of a speech act that deserves attention: the conditions on the propositional content. Some illocutionary points put some conditions on the propositional content of the speech act. The commissive point, for instance, requires that the propositional content represent a future action of the speaker, while the directive requires it to represent a future action of the addressee. The assertive illocutionary point, however, does not pose any such condition on the propositional content, which is an important difference between assertives, on the one hand, and commissives and directives, on the other.5

4 Estimates of the future are assertives Turning to predictions, one might ask whether they have any speech act theoretic feature other than being about the contingent future. In Korta (2015), I argued that they do. Contrary to what traditional speech act theory assumes,6 I contended that our estimates of the future have the world-to-words (↑) direction of fit because they are about the future. So, utterances like (1a) and (2a) should be classified together with commissives and directives, and separated from assertives. The notion of direction of fit is central to speech act theory but it is 5 Incidentally, I venture that the propositional content conditions of commissives and directives should also include that the propositional content brings about a contingent event. According to Searle (1969), a promise to do something that is going to be done, anyway, is a defective promise, and the same seems to go for commissives in general and, perhaps, directives also. 6 See, for example, Searle (1975), Bach and Harnish (1979), Searle and Vanderveken (1985), Vanderveken (1990).

60

Kepa Korta

not clearly defined and I suspect that my mistake was to take the notion of direction of fit to involve a temporal relation instead of a causal one.7 As I took it, the difference in direction of fit would amount to this: at the time of utterance, the truth-makers of asserted propositional contents are situated in the present or the past, while the truth-makers of propositional contents of predictions – including commissives and directives – are situated in the future. But I now see that this is quite wrong. Causes typically precede effects, so when the utterance (words) is causally relevant for the occurrence of the event (world), and, hence, there is a worldto-words direction of fit (↑), then the event (world) will typically lie in the future with respect to the utterance (words). Yet, we have no guarantee that the reverse is also true. We cannot establish that whenever the event lies in the future of the speech act there is a world-to-words direction of fit (↑). Furthermore, we have several reasons to say that the suggested correspondence between direction of fit and temporal order of utterance and event is incorrect. For one, there is the fact that we can have an utterance occurring more or less simultaneously with the relevant event with either of the two main directions of fit. Suppose you are typing (world) as you think (words). As pointed out by Anscombe herself (and, apparently, other distinguished scholars some time earlier about other cases), this can be considered a case in which your thoughts represent truly (or falsely) what you are typing. But it can also be taken to be a case in which you type to fit what you think: [. . .] facts are, so to speak, impugned for not being in accordance with the words, rather than vice versa. This is sometimes so when I change my mind; but another case of it occurs when e.g. I write something other than I think I am writing: as Theophrastus says (Magna Moralia, [footnote: Assuming that we are correctly told that Theophrastus was the author.], 1189b 22), the mistake here is one of performance, not of judgement. (Anscombe 1957: 4–5)

Besides, in assertives, which have words-to-world (↓) direction of fit, the speech act is satisfied if and only if the state of affairs represented by the propositional content exists, with no causal and, hence, no temporal relation between propositional content and state of affairs. There is no condition over the present, past or future existence of the state of affairs. Commissives and directives, on the other hand, establish causal conditions on the occurrence of the event represented by the propositional content. And that’s what makes them have a worldto-words (↑) direction of fit. 7 “It would be very elegant if we could build our taxonomy entirely around this distinction in direction of fit, but though it will figure largely in our taxonomy, I am unable to make it the entire basis of the distinctions” (Searle 1975/1979: 4).

Talk about the future

61

When we promise something or issue an order, future events happen because they were promised or ordered; because the speech act was made with a certain illocutionary force and a certain propositional content. In contrast, when we estimate that such-and-such will happen, the occurrence of the suchand-such is independent of the speech act; it would have happened all the same, even if nobody had estimated anything.8 To make things clearer, it can be worth comparing a mere estimate of the future with a promise made using the very same sentence and with the very same propositional content. As Anscombe’s example of the shopping lists (1957: 54–55) shows, a difference in direction of fit results in different conditions of success and failure. Consider a man going shopping using a shopping list and a detective following his steps and writing down everything the shopper buys. Ideally, the contents of the two lists will be identical. But a discrepancy between list and reality in each case will be significantly different. If the shopper’s list reads chicken and he buys turkey, the error is in his actions, not the list. But if the detective’s list says chicken when the shopper buys turkey, the error is on the list, not the actions. The resolutions are, of course, very different. The detective should correct her list. The shopper would be cheating or self-deceiving if he did that. Let’s now turn to see what would count as a discrepancy between propositional content and the relevant event, and what would constitute a resolution in each case. Consider the sentence (13) In 2020, I’ll be married and with children. This can clearly be a promise, i.e., an expression of intention, a speech act with world-to-words (↑) direction of fit, whose fulfilment requires not only that the speaker be married and with children in 2020, but also that the speaker brought about that state of affairs because she wanted to fulfil the promise. But the utterance can also be a mere estimate of the future, whose fulfilment does not require more than the truth of the propositional content. If in 2020 the speaker is not married and with children, it is clear that the discrepancy has different sources in each case. Not fulfilling a promise is a matter of performance; when the time arrives, I do not do what I promised to do (the error is in the shopper’s

8 So-called “self-fulfilling” prophecies present an interesting exception. Prophecies, in general, are a kind of estimate about the future, with words-to-world (↓) direction of fit; self-fulfilling prophecies, however, play a causal role in their own satisfaction, causing the occurrence of the event that is being predicted, so their direction of fit seems to be world-to-words (↑). Or, perhaps, they have double direction of fit like declaratives.

62

Kepa Korta

acts). Making an inaccurate estimate of the future is an error of judgment; as it turns out, when I uttered (13) I had wrong beliefs or expectations (the error is in the detective’s list). If this is so, pace Korta (2015), we should conclude that estimates of the future are assertives and have words-to-world (↓) direction of fit, while commissives and directives have world-to-words (↑) direction of fit. Other than that, the account given in Korta (2015) on the competing intuitions regarding our estimates of the contingent future is left practically intact. The determinacy intuition is just a by-product of an intuition of same-saying – or common propositional content – among utterances with different temporal relations to the relevant state of affairs. The indeterminacy intuition is a natural consequence of our estimates of the future. Their truth-maker is inexistent at the time of the utterance, so they have no truth-value then. The notion of propositional content in speech act theory is problematic, however. That is our next issue.

5 The propositional content of illocutionary acts There are at least two problems related to the notion of propositional content that should be addressed. One concerns all kinds of speech acts and belongs to the roles of the notion within speech act theory. The other is particular to predictions and affects any theory of propositional content that takes at least some constituents of the proposition to depend on the existence of particular objects or events. The solution to both problems will come from the adoption of a view that puts forward content-pluralism. I will start by considering the first general problem. As Korta and Perry (2007) argue, the notion of propositional content is supposed to play two roles in speech act theory. On the one hand, the propositional content represents what utterances like (9)–(12) have in common in the right circumstances: what they say. This is what remains when all components of illocutionary force are left aside. We can represent this propositional content as the proposition constituted by a certain individual, John, a certain room, the act of leaving and some particular time:9

9 Small capitals are used to represent propositions. Roman bold means that it is the objects themselves which are the constituents of the proposition, and not any identifying condition associated with them via the particular expression used. In this case, it is also assumed that the incomplete description ‘the room’ is used referentially and that the verbal tense refers to a particular time.

Talk about the future

63

(14) JOHN LEAVE THE ROOM AT T. Being the propositional content that is common to all these speech acts requires objects themselves to be the constituents of the propositions, and not any identifying condition associated to the particular expressions used. Take (9) and (10), for instance. The speaker of (9) uses the name ‘John’ to refer to John; in (10) he uses the first-person singular pronoun. If there is a common element contributed by these two sub-utterances, it must be their referent. The same seems to be the case for any other referential expression. This means that the notion of propositional content as the common content of utterances with different illocutionary force (in the appropriate circumstances) favours a direct reference view on the contribution of referential terms to the propositional content. It also points to a de-tensed view of the representation of time such that the temporal adverbs in (2a)–(2c) contribute with a particular day to their common propositional content, and not with any inherently perspectival identification of the day. As I have explained in section 3, however, speech act theory assumes that illocutionary points and forces can put some requirements on their propositional contents. The commissive illocutionary point, for instance, requires the propositional content to represent a future act of the speaker, and thus (14) would not be an appropriate formulation of the propositional content of (10). Instead, the correct propositional content respecting the conditions established by the commissive illocutionary point would be the following one:10 (15)

THE SPEAKER OF (10) LEAVES THE ROOM AT A TIME LATER THAN THE TIME OF (10) .

This proposition is determined by the fact that a certain utterance – (10) – has occurred and the meaning of the sentence uttered. It is a general (existential) proposition about the speaker of the utterance and a time which is later than the time of the utterance, and it is singular about the utterance itself (and the room). (15) meets the requirements imposed by the commissive illocutionary point, but the price is to include elements that, following the previous notion of propositional content as the common content of (9)–(12), were required to be left aside in (14). In other words, the two roles that the notion of propositional content is required to play in speech act theory yield two different propositions for an utterance of (10): (14) and (15). (14) represents faithfully the propositional 10 Italic bold indicates that it is the identifying conditions associated to the first-person pronoun and the future tense particle which are constituents of the proposition and not the referents themselves.

64

Kepa Korta

content common to (9)–(12), but it is not able to meet the propositional content conditions of the commissive illocutionary point of (10).11 (15) appropriately represents these conditions but does not work as a representation of the common propositional content of (9)–(12). This dilemma points toward a general assumption in contemporary semantics and pragmatics that Korta (2007) dubs monopropositionalism. Leaving presuppositions and implicatures aside, monopropositionalism associates one single proposition to the utterance; a proposition which is supposed to play various semantic or pragmatic roles such as what is said, the proposition expressed, the literal content, the literal truth-conditions, the subject matter, the thought expressed, the common possible content of speech acts with different illocutionary points, and what not. The problem is that, as in speech act theory, it is impossible for a single proposition to play all these roles. A way out of the dilemma is to adopt a content pluralism like the one proposed by Perry (2012) and Korta and Perry (2011, 2013).12 According to Korta and Perry’s content-pluralism, (14) and (15) correspond to (10)’s referential content and utterance-bound (reflexive) content, respectively. The utterance-bound content is determined by the meaning of the sentence used plus the fact that an utterance of that sentence has been made. This does not constitute what the speaker said in uttering the sentence nor its subject matter, but it is the relevant propositional content when considering the propositional content conditions established by the illocutionary point. To get at the subject matter, we need to consider the referential content of the utterance that gets determined by what determines the utterance-bound-content plus the facts that determine the referents of names, pronouns and other referential expressions. This is the kind of content that gets at what the utterances (9)–(12) have in common. This sort of content-pluralism, which is independently motivated to deal with problems related to reference and communication, is well equipped to tackle the general problem of the notion of the propositional content in speech act theory as well as another problem that has to do with the propositional content of our predictions.

11 The same point arises with a directive like (11), of course. The common propositional content would be aptly represented by (14), while the proposition meeting the propositional content conditions of the illocutionary point would include the condition of being the addressee of the utterance who performs the act at a later time. 12 Bach (1999), Carston (2002) and Neale (1999) are early proposals of different versions of content-pluralism for particular issues. For a content-pluralistic account of tense and time, see Korta & Ponte (2014a, 2014b).

Talk about the future

65

6 The propositional content of predictions As I said in section 1, our only explicit metaphysical assumptions about time are that time exists, that we can refer to instants or intervals as objects, and that the future is open. These assumptions require that the events (states, acts. . .) that our predictions predict do not exist at the time of utterance. As Perry puts it, Until an event happens (. . .), the event is a merely possible event, and not a real event. And by saying it is merely a possible event, I mean to say, basically, that it is not an event at all; there are descriptions and abstract types, that will denote or characterize the event once it exists to be denoted or characterized. But there is no concrete event. (Perry 2006: 15–6)

This sounds pretty plausible to me. It does not affect the propositional content of all predictions, however. Even if events are the truth-makers of the propositional content of predictions, and do not exist at the time of utterance, they do not belong to the propositional contents of all predictions. Take (1a), for instance. (1a) It will rain in Donostia tomorrow. The speaker of (1a) refers to Donostia and tomorrow, given the singular terms used in the sentence and the contextual facts determining a particular town and a particular date, but there is no reference to a particular event. The event is not a constituent of the propositional content of (1b) or (1c) either (uttered on the subsequent two days, respectively): (1b) It is raining in Donostia today. (1c) It rained in Donostia yesterday. Here, the event is not referred to, but existentially quantified over. Similarly, (2a)–(2c) do not refer to any sea-battle, so the inexistence of the sea-battle at the time of (2a) creates no problem. There is no problem with (2a)–(2c) sharing a common propositional content either. This agrees with our diagnosis of the indeterminacy intuition associated to predictions, which is also in tune with Broad’s view: [Predictions] do not refer to any fact, whether positive or negative, at the time when they are made. They are therefore at that time neither true nor false. They will become true or false when there is a fact for them to refer to; and after this they will remain true or false, as the case may be, for ever and ever. (Broad 1923: 73)

66

Kepa Korta

The problem with the propositional content of predictions arises with singular terms that are supposed to refer to future individuals, i.e. individuals that do not exist yet. I am not talking about an utterance like (13). In this case, if uttered by me, the first-person pronoun refers to me at the time of the utterance, and not to my future self in 2020. I am thinking about utterances like (16), in which I could be using the name ‘Kepa Junior’ to talk about the first son I will eventually have: (16) Kepa Junior will have curly hair. Kepa Junior does not exist and, given present circumstances, will also not exist in the immediate future, so the time referred to is an indeterminate and possibly distant future. So, given that ‘Kepa Junior’ is a proper name, i.e. a referential term that normally contributes its referent to the propositional content of the utterance, how can an inexistent individual be referred to and enter as a constituent of the prediction? A solution is the admission of gappy propositions; propositions that have empty slots and have no values (references) for future individuals. This kind of proposition is used for the case of “empty” and fictional names, for example.13 Another option is to bite the bullet and accept that the causal-historical chains that secure reference backwards do also work forward and secure reference to future objects (Jeshion 2010). With these options in mind, let’s explore another way here: the solution put forward by content-pluralism. Once we admit that utterances can have at least two kinds of contents – one referential, the other utterance-bound – the solution comes natural. At the time of the utterance, (16) does not have referential truth-conditions, because ‘Kepa Junior’ does not have a referent, since Kepa Junior does not exist. The name, however, provides an identifying condition for an individual in the future. In other words, at the time of utterance, (16) does not express a singular proposition about my son but a general one about the future existence of such an individual. Like any other prediction, the utterance does not have a truthvalue at the time of the utterance, but it has complete truth-conditions; it is thus a complete proposition, and not a gappy one.14

13 Broad (1923) takes future times to be fictional entities and compares an expression like ‘tomorrow’ with fictional names, as we will shortly see. 14 A consequence of this approach is that we cannot express singular propositions or entertain singular thoughts about future entities. This is in agreement with Gale (1968), Gale and Thalberg (1965), Prior (1959) and Ryle (1954).

Talk about the future

67

The problem of the propositional content of predictions would generalise if we make stronger metaphysical assumptions about the inexistent future. In this case, any prediction would involve a referential gap, because as Broad puts it: Something called tomorrow is not a constituent of judgments which are grammatically about “to-morrow”, any more than the individual called Puck is a constituent of judgments which profess to be about “Puck”. (Broad 1923: 77)

The problem, of course, concerns not only ‘tomorrow,’ but any prediction. If we assume, as I do, that propositions are temporally complete and that they include a temporal parameter (articulated or unarticulated by the sentence uttered), all predictions should include a temporal future element; but what could this parameter be, if future times, dates and years do not exist? How do we refer to inexistent times? Well, we do not; we only describe them. According to the proposed solution, the propositional content of predictions would involve identifying descriptions of days instead of days themselves. Take (1a), at the day of the utterance. Given the inexistence of the day corresponding to ‘tomorrow,’ it would lack referential truth-conditions, but it would have utterance-bound truth-conditions: (16) RAIN ON THE DAY AFTER THE DAY IN WHICH (1A) OCCURRED IN DONOSTIA . (16) is the utterance-bound content of (1a). It is a singular proposition about Donostia and (1a). The utterance itself is also a constituent of the proposition, but, at the time of the utterance, the following day is not. In its place, we have an identifying condition – i.e. the day after the day of the utterance. Given that the utterance occurred on a particular day, say, August 30th, 2015, we can also distinguish its time-bound content: (17)

RAIN ON THE DAY AFTER AUGUST 30TH, 2015 IN DONOSTIA .

or, equivalently: (18) RAIN ON AUGUST 31ST, 2015 IN DONOSTIA. Notice, however, that this is not a singular proposition about that date, neither can it be, given that we assumed that (1a) was uttered on August 30th and future days do not exist. The day itself is not a constituent of (18), but just the identifying condition provided by the date (in bold italic).15 15 This supports Perry’s suggestion that dates are descriptions, designating days in terms of an identifying condition, but quite extraordinary descriptions given their systematicity (see Perry 2013).

68

Kepa Korta

It can be concluded then, that the issue of the propositional content of predictions, even when generalised by a certain conception of time that takes the future to be inexistent, is satisfactorily tackled by a content-pluralism á la Korta and Perry. Of course, this does not count as evidence for that or any other metaphysical view about time, but it shows that content-pluralism offers a natural way of dealing with a common-sense view about the future and our natural ways of talking and thinking about it.

7 Conclusions In this paper, I have done two things. First, I have amended a previous approach to predictions that held them all to have a world-to-words (↑) direction of fit and that separated estimates of the future from the speech acts with assertive illocutionary point. Second, I have offered a content-pluralistic view of the propositional content of predictions. Regarding the first issue, I have argued that estimates of the future are assertive – they have words-to-world (↓) direction of fit, but still, like commissives and directives, they are not true or false at the time of utterance, so the indeterminacy intuition is justified. The determinacy intuition, on the other hand, stems from intuitions of same-saying that only arise when the circumstances for doing the correspondent assertions about the present and, then, the past are met, that is to say, the event satisfying the speech act has occurred. Regarding the second issue, I have shown, first, that the notion of propositional content is problematic for speech act theory and, second, that, given some assumptions about the future derived from a conviction that it is metaphysically open, predictions can be taken to have incomplete propositional contents. I have demonstrated that a pluralistic view of content à la Korta and Perry (2011, 2013) provides a natural account for both these issues, by suggesting that our predictions lack a truth-value at the time of utterance, but do have full truthconditions, even when they concern inexistent events, objects or times. In those cases, predictions lack referential truth-conditions, but they do have utterancebound or reflexive truth-conditions.

References Anscombe, Gertrude Elizabeth Margaret. 1957. Intention. Oxford: Blackwell. Austin, John L. 1953. How to talk – some simple ways. Proceedings of the Aristotelian Society 53. 227–246. Austin, John L. 1961. Performative utterances. In James Opie Urmson & Geoffrey James Warnock (eds.) Philosophical papers, 233–252. Oxford: Clarendon.

Talk about the future

69

Austin, John L. 1962. How to do things with words. Oxford: Clarendon. Bach, Kent. 1999. The myth of conventional implicature. Linguistics and Philosophy 22. 327– 366. Bach, Kent & Robert M. Harnish. 1979. Linguistic communication and speech acts. Cambridge, MA: The MIT Press. Broad, Charlie Dunbar. 1923. Scientific thought. London: Kegan Paul. Carston, Robyn. 2002. Thoughts and utterances: The pragmatics of explicit communication. Oxford: Blackwell. Gale, Richard M. 1968. The language of time. London: Kegan Paul. Gale, Richard M. & Irving Thalberg. 1965. The generality of predictions. Journal of Philosophy 62. 195–210. Jeshion, Robin. 2010. Singular thought: Acquaintance, semantic instrumentalism, and cognitivism. In Robin Jeshion (ed.) New essays on singular thought, 105–140. Oxford: Oxford University Press. Korta, Kepa. 2007. Acerca del monoproposicionalismo imperante en Semántica y Pragmática. Revista de Filosofía 32. 37–55. Korta, Kepa. 2015. Mañana lloverá (o no): Aserciones, predicciones y verdad. In Eleonora Orlando (ed.), Significados en contexto y verdad relativa: Ensayos sobre semántica y pragmática, 243–264. Buenos Aires: Título. Korta, Kepa & John Perry. 2007. How to say things with words. In Savas L. Tsohatzidis (ed.), John Searle’s philosophy of language: Force, meaning, and thought, 169–189. Cambridge: Cambridge University Press. Korta, Kepa & John Perry. 2011. Critical pragmatics: An inquiry into reference and communication. Cambridge: Cambridge University Press. Korta, Kepa & John Perry. 2013. Highlights of Critical Pragmatics: reference and the contents of the utterance. Intercultural Pragmatics 10. 161–182. Korta, Kepa & María de Ponte. 2014a. Tenses, dates and times. Research in Language 12. 1– 17. Korta, Kepa & María de Ponte. 2014b. On times and contents. In Iwona Witczak-Plisiecka (ed.), Cognitive and Pragmatic Perspectives on Speech Actions, Berlin: Peter Lang. 13– 34. MacFarlane, John. 2003. Future contingents and relative truths. The philosophical quarterly 53. 321–336. Neale, Stephen. 1999. Coloring and composition. In Kumiko Murasugi & Robert Stainton (eds.), Philosophy and linguistics, 35–82. Boulder, CO: Westview. Perry, John. 2006. How real are future events? In Friedrich Stadler & Michael Stöltzner (eds.), Time and History, Proceedings of the 28th International Wittgenstein Symposium, 13–30. Frankfurt: Ontos Verlag. Perry, John. 2012. Reference and reflexivity. 2nd edn. Stanford: CSLI Publications. Perry, John. 2013. Temporal indexicals. In Heather Dyke & Adrian Bardon and (eds.), A companion to the philosophy of time, 486–506. Oxford: Blackwell. Prior, Arthur. 1959. Identifiable individuals. Review of Metaphysics 12. 521–539. [Reprinted in Prior, Arthur. 1968. Papers on time and tense. Oxford: Clarendon Press]. Ryle, Gilbert. 1954. Dilemmas. Cambridge: Cambridge University Press. Searle, John R. 1969. Speech acts: An essay in the philosophy of language. Cambridge: Cambridge University Press.

70

Kepa Korta

Searle, John R. 1975. A taxonomy of illocutionary acts. In Keith Gunderson, (ed.), Language, Mind and Knowledge, 344–369. Minneapolis: University of Minnesota Press. Searle John R. 1979. Expression and meaning. Cambridge: Cambridge University Press. Searle, John R. & Daniel Vanderveken. 1985. Foundations of illocutionary logic. Cambridge: Cambridge University Press. Vanderveken, Daniel. 1990. Meaning and speech acts: Volume I – Principles of language use. Cambridge: Cambridge University Press.

Etsuko Oishi

5 Illocutionary effects, presupposition, and implicature Abstract: The present paper sets out to develop effect-based speech act theory (Austin [1962]1975) with a view to contributing to controversial issues in pragmatics, such as presupposition, implicature and metalinguistic negation. More specifically, it puts forth the following idea about how utterances affect the ongoing discourse: as the conventional effects of expositives, which are one of Austin’s categories of illocutionary acts, utterances can bring about changes in the assumptions shared by discourse participants. This concept of illocutionary acts in discourse draws a connection between illocutionary acts and conversational implicature (Grice 1989), according to which, the latter is meaning which comes from the supposition that the hearer (that is, the subsequent speaker) creates a response or sequel to the preceding illocutionary act, while trying to observe the Cooperative Principle and its accompanying maxims of conversation.

1 Introduction Searle (1969) developed Austin’s ([1962]1975) original idea that to say something is to do something, and, ever since, it is Searle’s ideas, rather than Austin’s, that are usually examined when it comes to understanding and developing speech act theory1. There is, however, a significant difference between the two approaches when it comes to the success of illocutionary acts. For Austin, illocutionary acts are successful when they bring about conventional effects, while, for Searle, they are successful when the speaker’s intention of performing the illocutionary act is understood by the hearer. Against this background, it is this intention of the speaker that has become a key concept among mainstream speech act theorists. As Sbisà characteristically notes, “since Searle (1969: 46–49), the illocutionary act has generally been conceived as the act a speaker successfully

1 Exceptionally, Sbisà (2001, 2002a, 2002b, 2006, 2007, 2013, and 2014) and Oishi (2011, 2013, 2014a, and 2014b) continue to develop Austinian speech act theory. Etsuko Oishi, Tokyo University of Science DOI 10.1515/9781501505089-005

72

Etsuko Oishi

performs when, uttering a sentence with a certain intention in certain circumstances, he or she gets the hearer to understand his or her intention” (2001: 1795). This shift from effect-based to intention-based speech act theory, however, is not without consequence. As I will now turn to show, it carries important implications for (a) illocutionary acts sequences, (b) the felicity conditions and unsuccessful performances of illocutionary acts, as well as (c) the classification of illocutionary acts.

2 Illocutionary act sequences The sequentiality of speech acts, that is, a discourse consisting of a series of speech acts, has not been explained successfully by Searle’s version of speech act theory (Searle 1969, 1979, 1983; Searle and Vanderveken 1985). Even Searle admits his approach’s inability to explain conversation: Now the question naturally arises: Could we get an account of conversations parallel to our account of speech acts? Could we, for example, get an account that gave us constitutive rules for conversations in a way that we have constitutive rules of speech acts? My answer to that question is going to be “No.” (Searle 1992: 7)

In Searle’s version of speech act theory, the speaker’s intention makes the utterance a particular illocutionary act, and this act becomes successful only when the intention behind it is understood by the hearer. This, however does not apply to conversations, since the speaker’s intention cannot make a sequence of illocutionary acts a particular conversation. As Searle himself states: The point I am making is: in the way that, e.g., a commitment to truth is in part constitutive of statement making, and therefore explanatory of statement making, the way that relevance is ‘constitutive’ of conversation is not similarly explanatory of conversation; because what constitutes relevance is relative to something that lies outside the fact that it is a conversation, namely the purposes of the participants. [. . .] The fact that a sequence of utterances is a conversation, by itself, places no constraints whatever on what would count as a relevant continuation of the sequence. (Searle 1992: 13–14)

It is therefore not the speaker’s intention but “the purpose of the participants” that makes the connection among illocutionary acts to form a conversation, and that is why, for Searle, “traditional speech act theory will not go very far in giving us sequencing rules for discourse” (1992: 11). According to Austin (1975: 116–117), on the other hand, the illocutionary act produces three different kinds of effects: (i) securing uptake, (ii) producing a

Illocutionary effects, presupposition, and implicature

73

conventional effect, and (iii) inviting a response or sequel2 (cf. Sbisà 2002a: 73)3. The connection between one utterance and the following is then explained as a sequence created by (i) the speaker’s attempt to produce a conventional illocutionary effect and her invitation for the hearer to understand and respond to it, as well as (ii) the hearer’s actual uptake, and his response to it, which again attempts to produce a conventional effect, and invites uptake and response. This can be illustrated as follows:

Figure 1: The model of producing illocutionary effects

In this setting, an utterance of ‘It wasn’t your fault’ (U2) by Speaker2 (=Hearer1) would react to Speaker1’s utterance of ‘I’m very sorry’ (U1) in the following way: U2 indicates Speaker2’s (=Hearer1) uptake of the illocutionary act of apologising that Speaker1 tried to perform by U1. b. U2 does not indicate that U1 has brought about the illocutionary effect of apologising. c. U2 becomes a response to the illocutionary effect of apologising (by disclaiming Speaker1’s responsibility).

a.

Moreover, U2 , ‘It wasn’t your fault’, invites an utterance which: indicates Hearer2’s (=Speaker1) uptake of Speaker2’s illocutionary act of disclaiming Speaker1’s responsibility; b. indicates that U2 has brought about the illocutionary effect that Speaker1’s responsibility has been disclaimed; and c. becomes a response or sequel to the illocutionary effect of disclaiming Speaker1’s responsibility. a.

2 The difference between responses and sequels seems to lie in directness and immediacy. Responses are direct, and often immediate, reactions to the illocutionary effect that the speaker tried to bring about: they accept or reject the illocutionary effect. Sequels, on the other hand, react to the speaker’s attempt to perform the illocutionary act under the specific circumstances. 3 Sbisà (2002a) develops a sequential model of illocutionary acts within narrative semiotics (see also Greimas 1983).

74

Etsuko Oishi

In this respect, Austin’s speech act theory gives a general framework in which an illocutionary effect produces a sequence of utterances. One utterance is sequenced with another when the latter is a response or sequel to the illocutionary effect of the former, and the sequence is extended when the third utterance is made as a response or sequel to the illocutionary effect of the second utterance.

3 The felicity conditions and unsuccessful performances of illocutionary acts Another major difference between Austin’s version of speech act theory and that of Searle’s concerns failed illocutionary acts. In Searle’s (1969) version, failed illocutionary acts are just failures, since the hearer fails to understand the intention with which the speaker utters the sentence. In Austin’s (1975) theory, however, even though failed illocutionary acts do not bring about conventional illocutionary effects, they are not without effects either; the utterance ‘I’m very sorry’ in the above-mentioned circumstances may not bring about the conventional effect of apologising, but it still secures, as an illocutionary effect, the hearer’s uptake of the utterance as an act of apologising. This brings to the discourse a particular interpretation of the speaker, the hearer, and the circumstances, a point that we can further examine by cross-analysing Searle’s felicity conditions and rules and Austin’s felicity conditions. Searle (1969: 54–55, 57–64) describes the conditions under which an illocutionary force indicating device is used in terms of (i) the state of affairs that the act is about (the propositional content rule); (ii) the preconditions/circumstances under which the act is performed (the preparatory rule); (iii) the feelings/beliefs/ intentions of the speaker (the sincerity rule); and (iv) the point of the illocutionary act (the essential rule). These conditions and their corresponding rules are used to distinguish one class of illocutionary acts from another, as well as one illocutionary act from another. That is, these conditions and rules show the elements of the illocutionary act in terms of which classes of illocutionary acts and individual illocutionary acts are specified. However, they do not explain how illocutionary acts are performed successfully in general. Rather, they explain what it is to perform a particular illocutionary act, the understanding of which makes the act successful; a point that becomes much clearer in Searle (1979). Austin’s felicity conditions, on the other hand, genuinely describe the conditions under which conventional illocutionary effects are successfully brought about:

Illocutionary effects, presupposition, and implicature

75

(A.1) There must exist an accepted conventional procedure having a certain conventional effect, that procedure to include the uttering of certain words by certain persons in certain circumstances, and further, (A.2) the particular persons and circumstances in a given case must be appropriate for the invocation of the particular procedure invoked. (B.1) The procedure must be executed by all participants both correctly and (B.2) completely. (Γ.1) Where, as often, the procedure is designed for use by persons having certain thoughts or feelings, or for the inauguration of certain consequential conduct on the part of any participant, then a person participating in and so invoking the procedure must in fact have those thoughts or feelings, and the participants must intend so to conduct themselves, and further (Γ.2) must actually so conduct themselves subsequently. (Austin 1975: 14–15)4 Elsewhere (Oishi 2011, 2013, 2014a, 2014b), I have proposed to call the performer of an act and the person for whom the act is performed (in (A.1)) the addresser and the addressee, respectively. Following this, from an Austinian perspective, an illocutionary act brings about its conventional effect when (i) the speaker, the hearer, and the circumstances of the speech situation are assumed to be the addresser, the addressee, and the context of the act, respectively (felicity conditions (A.1) and (A.2)); (ii) the speaker executes the procedure correctly (felicity (B.1)); (iii) the hearer completes the procedure (felicity condition (B.2)); (iv) the speaker has the thoughts or feelings, or intentions of the addresser of the act (felicity condition (Γ.1)); and (v) the speaker (or the hearer) conducts herself subsequently as is specified for the addresser (or the addressee) of the act (felicity condition (Γ.2)). In this picture, the successful performance of an illocutionary act brings about not only the illocutionary effect, but also an understanding of (i) the speaker,

4 It should be noted that felicity conditions (A.1) and (A.2) are entirely distinct, in that felicity condition (A.1) specifies the conventional aspect of bringing about an illocutionary act, while felicity condition (A.2) specifies a given case. Accordingly, a particular speaker in a given case who produces an utterance in (A.2) is to be distinguished from a performer, who performs the illocutionary act specified by the accepted conventional procedure in (A.1). Similarly, the hearer that the speaker addresses in (A.2) is to be distinguished from the person for whom the illocutionary act specified by the accepted conventional procedure in (A.1) is performed. Finally, the circumstances of the situation in which an utterance is made in (A.2) are again to be distinguished from the context of the act specified by the accepted conventional procedure in (A.1).

76

Etsuko Oishi

the hearer, and the circumstances, (ii) the action of the speaker and the reaction of the hearer, and (iii) the speaker’s thoughts and feelings, as well as (iv) her (or the hearer’s) future action. For example, in a context where the utterance ‘I am very sorry’ is a successful act of apologising, the following understandings become valid:5 (A.1) There exists an accepted conventional procedure, which has the conventional effect of expressing a regret for a past offence, and the procedure includes uttering ‘I apologise’, ‘I am sorry’ or something equivalent. The addresser who is responsible for the past offence utters it to the addressee who suffered from it in the context where the addresser has given offence to the addressee and remedial measures have not been taken. (A.2) The speaker is appropriate to be the addresser of apologising, the hearer is appropriate to be the addressee of apologising, and the circumstances are appropriate to be the context of apologising. (B.1) The speaker has executed the procedure correctly, and (B.2) the hearer has completed the procedure. (Γ.1) The speaker has the feeling of regret, and commits herself to avoiding the conduct of the offence, and further (Γ.2) must actually avoid the conduct subsequently. So, what happens when an illocutionary act is not performed felicitously and results in failure? When the hearer’s reaction does not complete the act of apologising, as in ‘It wasn’t your fault’, which is the case of a hitch (Austin 1975: 18), some of the aforementioned understandings still continue to be valid. There exists an accepted conventional procedure that produces the illocutionary effect of apologising, and the hearer has suffered from something, as follows from the supposition that the utterance ‘I am very sorry’ has produced the effect of securing the hearer’s uptake. The hearer, who is the speaker of the second utterance, ‘It wasn’t your fault’, understands that the speaker of the utterance ‘I am very sorry’ had tried to bring about the illocutionary effect of apologising, and assumes that the speaker thinks that the felicity conditions (A.1)–(Γ.2) for apologising are applicable in the present case. With this assumption, the original hearer, and now speaker, disclaims the responsibility of the first speaker, which amounts to disclaiming her legitimacy to be the addresser of apologising. Yet, the fact that the second speaker does not disclaim the conventional procedure of apologising itself and his legitimacy to be the addressee of apologising implies that he agrees with those assumptions. This can be illustrated as follows:

5 For a detailed analysis of the act of apologising, see Oishi (2013).

Illocutionary effects, presupposition, and implicature

77

Figure 2: The model of an apology sequence

Figure shows that the speaker says ‘I am very sorry’ as the addresser of apologising to bring about the conventional illocutionary effect of apologising, while indicating (i) the understandings of herself, the hearer, and the circumstances, and (ii) her feelings and commitment for the future. The first speaker expects the hearer to be the addressee of the apologising who accepts the addresser’s expression of the feeling of regret for the past offence. Then, by disclaiming the speaker’s responsibility, the hearer (that is, the second speaker) refuses to be the addressee of apologising, while accepting some of the assumptions of the speaker, himself, and the circumstances. Against this background, it is easy to imagine that the conversation will develop into a discussion about the hitched performance of apologising itself, the existence or the absence of the speaker’s responsibility, or the hearer’s suffering. What this analysis shows is that it is not only the illocutionary effects produced, but also the suppositions on the basis of which the production of the illocutionary effects is attempted that make a sequence of utterances: an act and a response/sequel to it. For this reason, different types of illocutionary act result in different types of connection between utterances: a response/sequel depends on what change the illocutionary act attempts to bring about in the communicative situation, and what suppositions about the speech situation the act is based on. This brings us to the final aspect in which there is a fundamental difference between Austin and Searle: their distinct classifications of illocutionary acts, and, perhaps more importantly, the motivations for them.

78

Etsuko Oishi

4 The classification of illocutionary acts Searle’s goal in classifying illocutionary acts is to specify each illocutionary act and compare it with others on the basis of clear principles. In this respect, he criticises Austin’s classification by observing that, in it, “there is no clear principle of classification”, “there is a great deal of overlap from one category to another and a great deal of heterogeneity within some of the categories”, and “a very large number of verbs find themselves smack in the middle of two competing categories” (Searle 1979: 10). Having already described illocutionary acts by means of the propositional content, the preparatory, the sincerity and the essential rule in his earlier work (Searle 1969), Searle (1979: 2–8) returned to the topic, where he replaced the concept of essential conditions with (a) illocutionary points, and added further description criteria, including (b) directions of fit (words-to-world or world-to-words), (c) expressed psychological states, (d) degrees of strength/commitment, and (e) relations to the rest of the discourse. On the basis of these updated criteria6, Searle classifies illocutionary acts into five categories: (i) assertives, which “commit the speaker to something’s being the case;” (ii) directives, which are “attempts by the speaker to get the hearer to do something;” (iii) commissives, which “commit the speaker to some future course of action;” (iv) expressives, which “express the psychological states [. . .] about a state of affairs;” and (v) declarations, which bring about a “correspondence between the propositional content and reality.” (Searle 1979: 12–18) Austin’s (1975) classification, on the other hand, is based on the different aspects of communication in which illocutionary acts bring about changes as their conventional effects. So, unlike Searle’s goal of comparing illocutionary acts on the basis of clear principles, Austin’s goal is to clarify the types of effect that utterances bring about in communication. In this respect, the first four categories of illocutionary acts that Austin distinguishes between are: (i) verdictives, which bring findings to the discourse on the basis of evidence or reasons, like, for example, juries delivering a verdict or a referee calling a foul;

6 As Owen (1983: 128) notes, Searle specifies 12 ways in which illocutionary acts can differ, but his taxonomy only uses criteria (i) to (iii) above.

Illocutionary effects, presupposition, and implicature

79

(ii) exercitives, which exercise power, rights, or influence, as in the case of a President appointing a member of his/her party as Vice President or the head of a country declaring a war; (iii) commissives, which are used to commit oneself to a certain course of action, like when a person promises that s/he will lend a friend some money, or s/he will vote for a particular political party; and (iv) behabitives which show attitude and social behaviour, like, for instance, when a person apologises for her/his past conduct, sympathises with someone’s misfortune, or greets someone. As is obvious from the definitions above, these acts bring about effects in different realms of communication, namely (i) facts about the world, (ii) the addresser’s social power and rights, (iii) the addresser’s commitment, and (vi) the addresser’s social behavior vis-à-vis the addressee. Turning to expositives, the fifth and last category of illocutionary acts that Austin posits, they seem to bring about a different kind of effect: they bring about changes in the ongoing discourse by clarifying how utterances fit into an argument or conversation, how words are used, and so on. Austin (1975) divides expositives into seven subclasses, which seem to correspond to different ways in which expositives bring about changes in the discourse.7 These are: 1. 2. 3.

affirm, deny, state, describe, class, identify remark, mention, ?interpose 8 inform, apprise, tell, answer, rejoin 3a. ask 4. testify, report, swear, conjecture, ?doubt, ?know, ?believe 5. accept, concede, withdraw, agree, demur to, object to, adhere to, recognize, repudiate 5a. correct, revise 6. postulate, deduce, argue, neglect, ?emphasize 7. begin by, turn to, conclude by 7a. interpret, distinguish, analyse, define 7b. illustrate, explain, formulate 7c. mean, refer, call, understand, regard as (Austin 1975, 162–3) Now, the object of each class of expositive illocutionary acts seems to indicate the way in which the act affects the ongoing discourse. For example, the answer

7 The analysis of the illocutionary effects of expositives here is a development of Oishi (2015). The idea is further developed in Oishi and Fetzer (2016). 8 The question marks here are Austin’s.

80

Etsuko Oishi

to the question “What type of object can the speaker affirm, deny, state, describe, class and identify?” is likely to indicate what this subclass of expositive illocutionary acts refers to and acts on; in this case, things/events/situations in the world. As illocutionary effects, expositives bring into the discourse things/ events/situations in the world with a specific value: affirmed, denied, or stated situations, or described, classed or identified things/events/situations. So, by asking the question of “What type of object can the speaker . . . ?”, we can obtain a rough idea of what each subclass of expositive illocutionary acts refers to and acts on, as well as how it affects the ongoing discourse. The following is a provisional description of the objects or discourse entities on which each subclass of expositive acts brings about effects: 1.

affirm, deny, state, describe, class, identify things/events/situations in the world

2.

remark, mention, ? interpose statements

3.

inform, apprise, tell, answer, rejoin (and 3a. ask) the addresee

4. testify, report, swear, conjecture, ?doubt, ?know, ?believe knowledge about things/events/situations in the world 5.

accept, concede, withdraw, agree, demur to, object to, adhere to, recognize, repudiate (and 5a. correct, revise) statements that have already been brought into the discourse

6.

postulate, deduce, argue, neglect, ?emphasize thoughts

7.

begin by, turn to, conclude by the ongoing discourse 7a. interpret, distinguish, analyse, define words in the ongoing discourse, 7b. illustrate, explain, formulate concepts in the ongoing discourse 7c. mean, refer, call, understand, regard as utterances of the ongoing discourse

When the hearer accepts the speaker’s expositive act and regards the contents of the utterance as the speaker specifies, the relevant illocutionary effect is brought about, as shown in the diagram that follows:

Illocutionary effects, presupposition, and implicature

81

Figure 3: The model of the effects of expositive illocutionary acts (I)

The subclass of expositives in (3) (inform, apprise, tell, answer, rejoin, and ask) take the addresser as an object, and when the act is accepted by the hearer, the hearer is brought into the discourse as the addresser who is informed, apprised, told, answered, rejoined, or asked, as illustrated in the Figure 4:

Figure 4: The model of the effects of expositive illocutionary acts (II)

Finally, the expositives in (7), (7a), (7b), and (7c) take the ongoing discourse itself as their objects, and when the act is accepted by the hearer, the discourse begins, is turned to, or concludes ((7)), a word is interpreted, distinguished, analysed, or defined ((7a)), a concept is illustrated, explained, or formulated ((7b)), or the former utterance is meant, referred to, called, understood, or regarded as something ((7c)). Again, this can be diagrammatically shown as follows:

82

Etsuko Oishi

Figure 5: The model of the effects of expositive illocutionary acts (III)

There are some significant consequences of classifying illocutionary acts as Austin does, and describing illocutionary effects in the suggested way. For one, Searle’s criticism of Austin’s classification, which was mentioned above, is no longer valid. There is a “clear principle of classification,” since illocutionary acts are classified according to how their effects are brought about: they are brought about as (i) delivered facts about the world delivered (verdictives), (ii) exercises of the addresser’s social power and rights (exercitives), (iii) addresser commitments (commissives), (vi) or her social behaviour vis-à-vis the addressee (behabitives), and (v) expounded discursive functions (expositives). And what seems like an “overlap” to Searle is nothing more than the result of two or more illocutionary effects being brought about at the same time; for example, the effect of a jury’s act of delivering a verdict is brought about as a fact delivered upon evidence, with the jury’s judicial power exercised. The most important consequence, however, is the clarification of the effects of illocutionary acts in discourse. A successful illocutionary act which is secured by the action of the speaker and the reaction of the hearer brings to the discourse the understandings of (i) the speaker, the hearer, and the circumstances and (ii) the speaker’s feelings and thoughts, as well as (iii) her (or the hearer’s) future action. Furthermore, expositives bring the content of an utterance to the discourse in a particular way according to the discursive function of this utterance. In the following section, I will turn to show how this developed concept of illocutionary acts in the discourse can bring about a new perspective for concepts that have been central in pragmatics theorising, like presupposition and implicature.

Illocutionary effects, presupposition, and implicature

83

5 Illocutionary acts, presuppositions, and implicature According to Sbisa, the concept of presupposition can be approached in terms of the assumption of felicity as follows: When S utters a declarative sentence p and p (because of some presupposition indicator it contains) presupposes q, if q is not satisfied by the objective context, S’s utterance cannot be a felicitous assertion. So if we take it that S’s utterance is a felicitous assertion, then we also have to assume that q is not merely believed by S (or believed to be shared, etc.), but is satisfied by the objective context. But we do assume by default that S’s utterance of a declarative sentence constitutes a felicitous assertion. (Sbisa 2002b: 429)

Sbisa argues that presuppositions are assumptions about the objective context, or what I have here called the circumstances of the speech situation, which are satisfied when the utterance of a sentence is taken to be a felicitous assertion. A similar argument has been put forth in Subsection 1.2.: we tend to interpret a sentence while also assuming that it is uttered as a felicitous illocutionary act. In this respect, the assumptions of (i) the speaker, the hearer, and the circumstances of the speech situation, (ii) the speaker’s thoughts, feelings, or intentions, and (iii) the speaker’s (or the hearer’s) subsequent conduct are by default assumed to be applicable. In this respect, presuppositions seem to be these default assumptions, including the assumption of the circumstances of the speech situation when a sentence is assumed to be uttered felicitously, as Sbisa (2002b) claims. Austin’s classification of illocutionary acts, however, shows that asserting does not have a privileged status as the illocutionary act performed by uttering a declarative sentence. As we have seen, by uttering a declarative sentence p, an addresser can (i) deliver a fact upon evidence (verdictive), (ii) exercise her social power or rights (exercitive), (iii) commit herself (commissive), (vi) exhibit her social behavior vis-à-vis the addressee (behabitive), or (v) expound her utterance’s discursive function (expositive). Then, in the case where uttering a declarative sentence p performs an expositive act, the utterance can (1) affirm/state/describe a thing/event/situation, (2) mention or remark on a statement, (3) inform or apprise the hearer, (4) testify or report knowledge, (5) accept or withdraw a previous statement, (6) postulate or deduce a thought, (7) begin or conclude the discourse, as well as (7a–c) clarify a word, illustrate a concept, or refer to a former utterance. That is, the supposition of the circumstances of the speech situation depends on the illocutionary act that uttering the declarative sentence p performs, and its felicity conditions. If this is correct, the presuppositions of p should also depend on the type of illocutionary act that uttering p performs.

84

Etsuko Oishi

Let’s examine this further by using the following example: (1)

George hasn’t stopped smoking.

At face value, this utterance describes George and his life style, which makes it a subclass (1) expositive illocutionary act. The felicity conditions are by default assumed to be observed, as are all the accompanying assumptions about (i) the speaker, the hearer, and the circumstances of the speech situation, (ii) the speaker’s thoughts, feelings, or intentions, and (iii) the speaker’s (or the hearer’s) subsequent conduct. These assumptions include George’s having smoked before, since to describe George as not having stopped smoking, George must have smoked before. Now, if the same utterance is used to object to a former utterance that George has stopped smoking, it can be characterised as a subclass (5) expositive illocutionary act, with the negation included in it being what Horn (1983) has called “metalinguistic”. In this case, the assumptions that come by default from accepting that this utterance as a felicitous act of objecting to someone else’s statement that George has stopped smoking do not include George’s having smoked before, since the person who had put forth that statement might have been completely mistaken, which is the reason why the speaker of (1) objects to it. This shows that, following Austin’s classification, presuppositions, as assumptions about the speech situation, are sensitive to the type of illocutionary act that the utterance containing them performs. Conversational implicatures also seem to be sensitive to Austinian types of illocutionary act. Let’s examine some of Grice’s (1989) original examples of conversational implicature: (2)

A: I am out of petrol. B: There is a garage round the corner.

(Grice 1989: 32)

As Grice has argued, when we engage in conversation, we tend to assume by default that a response in a communicative exchange is reasonable, unless otherwise indicated – just as we tend to interpret an utterance while assuming that it is uttered as a felicitous illocutionary act, as argued here. Therefore, B’s utterance in (2) is assumed by default to be a reasonable response to A’s illocutionary act of describing her/his unfortunate situation of being out of petrol, in which B informs A of a way to get petrol. Since it is also assumed by default that B’s illocutionary act of informing A is performed felicitously, the supposition that the garage is open, or at least may be open, which is an implicature, is a part of its assumptions about the circumstances of the speech situation.

Illocutionary effects, presupposition, and implicature

85

Now, the suppositions about the hearer and the circumstances of the speech situation might come from an infelicitous but informative response to the former illocutionary act. Let’s examine the following exchange between A and B, who are planning a holiday in France, where A wants to meet his friend C (Grice 1989: 32): (3)

A: Where does C live? B: Somewhere in the South of France.

In (3), it is assumed by default that B’s utterance should be a felicitous response to A’s illocutionary act of asking B where C lives (for the purpose of visiting C), but B’s utterance is not a felicitous act of answering A’s question: somewhere in the South of France is not informative enough for them to visit C. By virtue of A’s utterance, B is invited to be the addressee of the act of asking where C lives (for the purpose of visiting C), who is expected to provide an answer regarding where C lives (an answer that will be specific enough to enable A to visit C). While being aware of the invitation, by saying ‘Somewhere in the South of France’, B indicates that she does not want to be or cannot be the addressee of A’s act. Unless there is a reason why B does not want to be the addressee, in other words, unless there is any reason why B is uncooperative, the supposition that B does not know exactly where C lives is applicable. This follows from the assumption that B’s utterance is infelicitous and the realisation that the infelicity comes from the fact that B cannot be the addressee of the act, as shown in the following diagram:

Figure 6: The infelicitous act of answering (I)

Let’s now examine another example of Grice (1989: 37): (4) A: What song did Miss X sing? B: She produced a series of sounds that corresponded closely with the scores of “Home Sweet Home.”

86

Etsuko Oishi

Again, it is assumed by default that B’s utterance is a felicitous response to A’s illocutionary act of asking what song Miss X sang, but B’s utterance is not a felicitous act of answering A’s question, since it is far from being perspicuous. While being aware of the invitation to be the addressee of the act of asking, B utters what she does, which indicates that B does not want to be or cannot be the addressee of the act. So, unless there is any reason why B is uncooperative, the supposition that “Miss X’s performance suffered from some hideous defect” (Grice 1989: 37) is applicable. Again, this is derived from the assumption that B’s utterance is infelicitous and the realisation that the infelicity comes from the fact that B could not answer ‘Home Sweet Home’ because of the hideous defect in Miss X’s performance. This can be illustrated as follows:

Figure 7: The infelicitous act of answering (II)

Austin’s felicity conditions specify suppositions about (i) the speaker, the hearer, and the circumstances of the speech situation, (ii) the speaker’s thought, feelings, or intention, or (iii) the speaker’s (or the hearer’s) subsequent conduct, when an illocutionary act is felicitous. Grice’s Cooperative Principle and its maxims, on the other hand, specify suppositions about the hearer and the circumstances of the speech situation when the hearer’s, and second speaker’s, utterance is a reasonable response or sequel to the illocutionary act of the first speaker. In this respect, when the hearer’s, and second speaker’s, utterance deviates from those suppositions, there is an indication of something noteworthy about the hearer, and second speaker, and/or the circumstance of the speech situation, which gives rise to the conversational implicature. Under this interpretation, presuppositions are suppositions that come from the default assumption that an utterance is a felicitous act of a particular illocutionary act, while conversational implicatures are suppositions that come from the assumption that an utterance is a rational response or sequel to the preceding illocutionary act in spite of some apparent deviation. In either way, it is the illocutionary act of the present or preceding utterance that determines both kinds of suppositions.

Illocutionary effects, presupposition, and implicature

87

6 Conclusion The present paper has developed Austin’s (1975) effect-based speech theory in an attempt to show that it is not only individual utterances, but also sequences of utterances that can be analysed by this version of illocutionary act theory. This is based on the premise that expositive illocutionary acts, which are not included in Searle’s (1969) classification, are acts of importing the contents of an utterance to the on-going discourse, and affect the discourse directly. By a series of such illocutionary acts, the discourse participants share understandings or develop a conversation into a particular direction. As we have seen, successful illocutionary acts do not only bring about particular illocutionary effects, but also specific understandings of the speech situation. Using Austin’s felicity conditions, I have shown that when an illocutionary act is successful, particular understandings of (i) the speaker, the hearer, and the circumstances of the speech situation; (ii) the speaker’s feelings, beliefs, and intentions; and (iii) the speaker’s (or the hearer’s) subsequent course of action are brought into the discourse. On the basis of this argument, I suggested that a restricted part of these understandings is typically understood as carrying presuppositions. Apart from presuppositions, however, conversational implicatures have also been shown to be sensitive to illocutionary act types. According to the proposal I put forth, conversational implicatures come from the default supposition that an utterance is a reasonable response to some preceding illocutionary act, and that a deviation from it is customarily interpreted as indication of special circumstances of the speech situation. Because of this, an utterance’s conversational implicatures are also sensitive to the illocutionary act of the preceding utterance to which the utterance at hand offers a response or sequel.

References Austin, John L. 1975. How to do things with words. 2nd edn. Oxford: Oxford University Press. [1st edn. 1962] Greimas, Algirdas J. 1983. Due sense II. Paris: Seuil. Horn, Laurence R. 1985. “Metalinguistic negation and pragmatic ambiguity.” Language 61. 121– 174. Oishi, Etsuko. 2011. How are speech acts situated in context? In Anita Fetzer & Etsuko Oishi (eds.), Context and contexts: Parts meet whole?, 181–204. Amsterdam: John Benjamins. Oishi, Etsuko. 2013. Apologies. In Marina Sbisà & Ken Turner (eds.), Handbooks of pragmatics: Volume 2 – pragmatics of speech actions, 523–543. Berlin: De Gruyter Mouton. Oishi, Etsuko. 2014a. Evidentials in entextualization. Intercultural Pragmatics 11. 437–462.

88

Etsuko Oishi

Oishi, Etsuko. 2014b. Discursive functions of evidentials and epistemic modals. In Sibilla Cantarini, Werner Abraham, & Elisabeth Leiss (eds.), Certainty-uncertainty – and the attitudinal space in between, 239–262. Amsterdam: John Benjamins. Oishi, Etsuko. 2015. Follow-ups as speech acts in mediated political discourse. In Anita Fetzer, Elda Weizman, & Lawrence N. Berlin (eds.), The dynamics of political discourse: forms and functions of follow-ups, 33–58. Amsterdam: John Benjamins. Oishi, Etsuko & Anita Fetzer. 2016. Expositives in discourse. Journal of Pragmatics 96. 49–59. Owen, Marion. 1983. Apologies and remedial Interchanges: A study of language use in social interaction. Berlin: Mouton de Gruyter. Sbisà, Marina. 2001. Illocutionary force and degrees of strength in language use. Journal of Pragmatics 33. 1791–1814. Sbisà, Marina. 2002a. Cognition and narrativity in speech act sequences. In Anita Fetzer & Christiane Meierkord (eds.), Rethinking sequentiality: Linguistics meets conversational interaction, 71–97. Amsterdam: John Benjamins. Sbisà, Marina. 2002b. Speech acts in context. Language and Communication 22. 421–436. Sbisà, Marina. 2006. After Grice: neo- and post-perspectives. Journal of Pragmatics 38. 2223– 2234. Sbisà, Marina. 2007. How to read Austin. Intercultural Pragmatics 17. 461–473. Sbisà, Marina. 2013. Locution, illocution, and perlocution. In Marina Sbisà & Ken Turner (eds.), Handbook of pragmatics: Volume 2 – pragmatics of speech actions, 25–76. Berlin: De Gruyter Mouton. Sbisà, Marina. 2014. Evidentiality and illocution.” Intercultural Pragmatics 11. 463–483. Searle, John R. 1969. Speech acts: An essay in the philosophy of language. Cambridge: Cambridge University Press. Searle, John R. 1979. Expression and meaning: Studies in the theory of speech acts. Cambridge: Cambridge University Press. Searle, John R. 1983. Intentionality: An essay in the philosophy of mind. Cambridge: Cambridge University Press. Searle, John R. 1989. How performatives work. Linguistics and Philosophy 12. 535–558. [Reprinted in Daniel Vanderveken and Susumu Kubo (eds.). 2002. Essays in Speech Act Theory, 85–107. Amsterdam: John Benjamins.] Searle, John R. 1992. “Conventions.” In John R. Searle, Herman Parret & Jef Verschueren (eds.), (On) Searle on Conversation, 7–29. Amsterdam: John Benjamins. Searle, John R. & Daniel Vanderveken. 1985. Foundations of illocutionary logic. Cambridge: Cambridge University Press.

II Pragmatics and cognition

Anne Reboul

6 Is implicit communication a way to escape epistemic vigilance? Abstract: The present paper focuses on conversational implicatures and presuppositions and attempts to provide an answer to the question of why we have them, given that they are costly for both the speaker and the hearer, relative to explicit communication. I examine some possible answers, in terms of Universal Grammar, of pragmatic minimax, of politeness and of plausible denial and find them either wanting or incomplete. My alternative proposal is that this type of implicit communication has all the features that allow a speaker to eschew or bypass epistemic vigilance, that is, the mechanisms that speakers and hearers have developed to avoid deception and manipulation.

1 Introduction I will start with a caveat: this paper is not devoted to all implicit communication. It is, notably, not concerned with pragmatic intrusion (cf. Levinson 2000) or the use of pragmatic inferential processes in the determination of the propositional content of an utterance (cf. Carston 2002). The term implicit communication, in the way I am using it here, is to be understood strictly as referring to presuppositions and, among implicatures, to conversational implicatures. Let me illustrate with two examples: (1)

John has stopped drinking.

(2)

The pianist has played some Mozart sonatas.

Though (1) explicitly communicates that John does not drink, it presupposes that John used to drink. And though (2) explicitly communicates that the pianist played at least some Mozart sonatas (which is compatible with the fact that he played all of them), it conversationally implicates that he played only some Mozart sonatas, and not all of them.

Anne Reboul, Institute for Cognitive Sciences-Marc Jeannerod, CNRS UMR5304, Lyon DOI 10.1515/9781501505089-006

92

Anne Reboul

In 2008, von Fintel and Matthewson published a paper, entitled ‘Universals in semantics,’ in which they point out that most candidates for semantic universals turn out either not to be universal, or to be in need of further empirical research. One exception to that skepticism over the existence of semantic universals,1 however, is implicit communication. Thus, it seems that conversational implicatures and presuppositions (with a few minor variations for presupposition2 that I will ignore here) are universal. So my point of departure is the universality of implicit communication or to put it more accurately, the universality of conversational implicatures and presuppositions. The very existence of implicit communication, in the present restricted sense, is somewhat mysterious, because implicit communication seems to impose costs on both the speaker and the hearer that explicit communication does not. Regarding the speaker, the main cost associated with implicit communication has to do with the production of an utterance which runs greater risks of being misunderstood relative to the explicit communication of the same content. In other words, even if both the speaker and the hearer are competent, uptake is far from guaranteed. Additionally, all experimental work on implicatures has shown a greater cognitive cost for pragmatic interpretation, manifested in children by a lengthy developmental course for the production of pragmatic interpretations, and by greater reaction times, when compared to the production of semantic interpretations, in adults (cf. Bott and Noveck 2004; Bott, Bailey, and Grodner 2012; Noveck 2000; Noveck and Posada 2003). This leads to the major question that I will try to answer in the present paper: Why do we have implicit communication? Here are a few potential answers to it, which I will turn to examine in the following section: a. Implicit communication is part of Universal Grammar; b. Implicit communication is the automatic product of general principles that govern the use of language in communication (e.g. minimax principles); c. Implicit communication is used for social reasons.

1 In von Fintel and Matthewson’s paper, the notion of semantics is understood fairly widely and encompasses most, if not all, linguistic phenomena related to meaning, including some that would usually be categorised as pragmatic in nature. 2 von Fintel and Matthewson detail the single claim that presupposition is universal in three sub-claims: All languages allow their speakers to express aspects of meaning which (a) are not asserted, but somehow taken for granted; (b) impose some constraints on when an utterance is felicitous, and (c) project through certain entailment-cancelling operators. For some languages, the relevant items possess only properties (a) and (c).

Is implicit communication a way to escape epistemic vigilance?

93

2 Potential answers 2.1 Implicit communication as part of Universal Grammar If implicit communication is part of Universal Grammar, one would expect it to be triggered automatically by the same linguistic items (be they words, expressions or constructions) in all languages where these items exist. A first problem with this arises when one looks at conversational implicatures. While some seem strongly linked to specific lexical items (e.g. the so-called scalar implicatures), quite a lot of them, and, more specifically, those dubbed particularised conversational implicatures, are not. Additionally, even for those that are linked to lexical items, a pragmatic interpretation is not mandatory, which seems to raise a problem for any view according to which they are grammatical in nature. Regarding presupposition, it is also unlikely that an appeal to Universal Grammar could be fruitful, given that “not all languages possess exactly the same presuppositional triggers” (von Fintel and Matthewson 2008: 179). For example, Jayez and van Tiel (2012) investigated only in three European languages: French, English and Dutch, demonstrating that it has different projection potentialities and can thus be deemed more strongly presuppositional in French and Dutch than in English. Against this backdrop, the prospects for a UG-based account of the existence of implicit communication seem fragile.

2.2 Minimax accounts Minimax principles are basically economy-related principles based on the assumption that costs should be minimised, while benefits should be maximised. As noted by Horn (2004), a lot of approaches, from classical rhetorical prescriptions to post-Gricean through neo-Gricean accounts, can be understood as relying on minimax reasoning. In this section, I will focus on the neo-Gricean and post-Gricean versions of this reasoning.

2.2.1 Neo-Gricean accounts Horn has been the most explicit neo-Gricean proponent of a minimax account for scalar implicatures, so I will concentrate on his proposal here. All in all, Horn’s account is based on the Gricean Principle of Cooperation and its associated Maxims (see Grice 1989). However – and this is my main interest, Horn

94

Anne Reboul

retains the Quality maxim3 and subsumes all other maxims under two principles, the Q-principle (that recapitulates the Gricean maxim of Quantity) and the R-principle (that recapitulates the Gricean maxim of Relation): Q-principle: Say as much as you can modulo Quality and R. R-principle: Say no more than you must modulo Q.

In his comments about these two principles, Horn explicitly refers to the economics of communication. So, he approaches the Q-principle as “a lowerbounding hearer-based guarantee of sufficiency of informative content” (Horn 2004: 13, emphasis mine), and the R-principle as “an upper-bounding correlate of the Law of Least Effort dictating minimization of form” (Horn 2004: 13, emphasis mine), essentially rendering it speaker-based. In other words, the Qprinciple is hearer-based and aims at maximising interpretive benefit, while minimising interpretive cost and the R-principle is speaker-based and aims at minimising productive cost, while preserving interpretive benefit. What is more, Horn insists on the importance of form over content in the generation of implicit meaning, most notably in his discussion of categorical sentences: A: All/Every F is G. E: No F is G. I: Some F is/are G. O: Not every F is G/Some F is not G.

As is well-known, A implies I, E implies O, while I implicates not-A and O implicates not-E. In relation to this, Horn says – and this is the heart of his brand of Neo-Gricean pragmatics: “Because the basic forms [A/E] are not only more informative, but briefer than their I/O counterparts, the use of the latter will strongly implicate against the former” (Horn 2004: 15, emphasis mine). One could argue that minimax principles oriented toward both the speaker and the hearer, such as the balance of Q- and R-principles, could account for the emergence of scalar implicatures as follows: Satisfying both Q- and R-principles will necessarily lead to the emergence of scalar implicatures as a result of the pressure for making communication profitable (i.e. as a result of the pressure for communicating as much content as possible from as economic a linguistic form as possible). 3 The Quality maxim enjoins the speaker to ‘try to make his/her contribution one that is true’ and is declined into two sub-maxims: ‘Do not say what you believe to be false’ and ‘Do not say that for which you lack evidence.’

Is implicit communication a way to escape epistemic vigilance?

95

There are still problems with this proposal, however. The main one has to do with the fact that Horn considers all the costs of scalar implicatures for the hearer to be linguistic (inference is supposed to be cheap, if not free), but it is far from clear that this is actually the case. Experimental studies on scalar implicatures (cf. Bott and Noveck 2004; Noveck 2000; Noveck and Posada 2003) show that they are indeed quite costly to generate (and the some-implicatures have been extensively investigated), leading to their non-generation in an important proportion of cases (up to 40%, depending on the experimental paradigm). Thus, conveying more content with fewer or shorter words is far from being ensured despite the Q- and R-principles. If this is the case, it is also far from clear that hearer/speaker-based minimax principles, à la Horn, can successfully account for the existence of implicit communication. 2.2.2 Post-Gricean accounts The most prevalent post-Gricean account is Relevance Theory (Sperber and Wilson 1995), which is basically concerned with interpretation rather than production; hence it is hearer-based. It can also be considered a Minimax account, since the very definition of relevance it puts forth is formulated in terms of cost and benefit. Roughly: An utterance is relevant – to the extent that its interpretation cost is low; – to the extent that its cognitive effects4 are important.

The main tenet of Relevance Theory in relation to the present discussion is that any utterance, as an instance of ostensive-inferential communication, conveys the presumption of its own optimal relevance, or, in other words, conveys the presumption that its cognitive effects will balance its interpretive costs. Relative to conversational implicatures, this means that, given the extra interpretive costs,5 extra cognitive effects have to be present to balance the extra interpretive costs. Let’s take an example: (3)

A: Do you want some wine? B: No, thank you. B’: I don’t drink alcohol.

4 The cognitive effects of an utterance would be on the benefit side of the Minimax. 5 In a major departure from neo-Gricean pragmatics, Relevance Theory does not consider inference to be cheap. For it, both linguistic decoding and pragmatic inferential processes are costly.

96

Anne Reboul

B and B’ communicate the same content, to wit: I don’t want any wine; yet, B communicates it explicitly, while B’ does so implicitly. But, in addition, B’ conveys (and B does not convey): (4) B does not drink beer. (5)

B does not drink vodka.

(6) B does not drink whisky. and so on and so forth. This suggests that conversational implicatures might merely be the result of the hearer-based relevance-theoretic minimax account: Because a straight answer (‘No, thanks’) only conveys I don’t want any wine, while an implicit answer (‘I don’t drink any alcohol’) conveys, in addition, the information that the speaker is a tea-totaller and would not want to drink any sort of alcoholic beverage, conversational implicatures occur as a result of the search for optimal relevance. However, one may wonder whether comparing a direct answer with only one piece of information with an indirect answer with two or more pieces of information is the right comparison, as, in each case, both costs and effects differ. I do not think that it is, and I suggest that a more appropriate comparison would be one that keeps cognitive effects constant, such as that between the respective costs of ‘No thanks, I don’t drink alcohol’ and ‘Thank you, I don’t drink alcohol.’ Contrary to relevance-theoretic predictions, recent experimental data (cf. Bott, Bailey and Grodner 2012) obtained while comparing identical contents that have been implicitly and explicitly communicated (e.g. Some elephants have trunks vs. Only some elephants have trunks) in terms of reaction time measurements, indicate that, despite the additional linguistic material (here only), explicit communication is less costly than implicit communication,6 when the cognitive effects are kept constant. It should be noted, of course, that this does not mean that optimising relevance is not the right explanation for how implicatures are accessed. Rather, it means that optimising relevance may not be a satisfying answer to the question of why we have conversational implicatures. A potential counter-objection could be that conversational implicatures are produced because there is no other way to convey the information (as is, for instance, the case for live metaphors). 6 This has to be relativised: the experiment targets conversational implicatures, and more specifically, scalar implicatures. It is far from obvious that the results can be extended to other kinds of conversational implicatures beyond scalars, while they also have nothing to say about the pragmatic processes involved in the recovery of explicatures (or fully propositional explicitly expressed forms).

Is implicit communication a way to escape epistemic vigilance?

97

However, this is clearly not the case for at least some conversational implicatures, i.e. scalars. Hence the relevance-theoretic minimax cannot be the explanation for why we generally make use of conversational implicatures.

2.3 Implicit communication is used for social purposes Social theories of why we have implicit communication are (again) centered on conversational implicatures. Two of the most popular such theories are Brown and Levinson’s (1987) Politeness Theory and Pinker’s Theory of the Strategic Speaker (Lee and Pinker 2010; Pinker 2007; Pinker, Nowak, and Lee 2008).

2.3.1 Politeness Theory Politeness Theory was inspired by Goffman’s (1967) notion of face, which can be defined as a positive social value that a person claims for himself/herself. Some speech acts, such as orders, threats, etc., pose a threat to the interlocutor’s face and politeness is a way to deal with this problem. In this setting, Brown and Levinson propose that there are four ways of managing such face threats, which, from the least to the most efficacious, are: positive politeness (sympathy), negative politeness (deference), indirect speech acts and off-the-record indirect speech acts. While some claims of Politeness Theory have been confirmed, as, for example, the intuition that indirect speech acts sound more polite than their direct counterparts, face threat does not seem to lie on a single scale, with negative repercussions for Brown and Levinson’s proposed scale of politeness: notably, off-the-record indirect speech acts are not perceived as more polite than indirect speech acts, and are also typically perceived as less polite than deferential politeness. Additionally, as noted by Pinker “veiled threats, oblique bribes, sexual come-ons are hardly examples of a speaker being polite” (2007, 443).

2.3.2 The Theory of the Strategic Speaker Pinker’s Theory of the Strategic Speaker is, in turn, based on the following diagnosis of what’s wrong with Politeness Theory (and Gricean cooperation): “Like many good-of-the-group theories in social science, [it] assume[s] that the speaker and the hearer are working in perfect harmony” (Pinker 2007, 443). Based on the assumption that the speaker and the hearer’s interests may indeed diverge, Pinker proposes instead a three-part theory, consisting of the logic of

98

Anne Reboul

plausible deniability, relationship negotiation, and language as a digital medium.7 For my current purposes, it suffices to concentrate on the logic of plausible deniability, which is after all at the heart of the framework. To this end, let’s reproduce Pinker’s (2007) example of a driver who is stopped by a police officer and decides to bribe him in order to avoid a fine. Table 1: Payoffs of explicit bribing in the speeding driver situation Dishonest officer

Honest officer

Don’t bribe

Traffic ticket

Traffic ticket

Bribe

Go free

Arrest for bribery

As Table 1 above shows, the driver can select one of two potential strategies. The first one, Don’t bribe, will give the same (negative) payoff, i.e. a traffic ticket, whether the officer is honest or dishonest. However, when it comes to the second strategy, Bribe, the ethics of the officer are crucial in relation to the payoff: if the officer is dishonest, the driver goes scot-free (a positive payoff), while if he is honest, the driver goes to jail for attempted bribery (a much more negative payoff than a mere traffic ticket). The main problem is then that the driver cannot know beforehand whether the officer is honest or not. Pinker’s solution is to add a third strategy, Implicate bribe, which turns out to be optimal as it changes the payoff if the officer is honest; at worst, the driver will receive a traffic ticket, and will not face going to jail, which makes it worth her while to attempt bribery, as long as she merely implicates it. Table 2: Payoffs of explicitly and implicity bribing the officer Dishonest officer

Honest officer

Don’t bribe

traffic ticket

traffic ticket

Bribe

Go free

arrest for bribery

Implicate bribe

Go free

traffic ticket

Pinker has defended his Theory of the Strategic Speaker through a formal model (Pinker, Nowak and Lee 2008), as well as through a series of psychological experiments (Lee and Pinker 2010). There are, however, two problems with it. First of all, Pinker claims that it is incompatible with Gricean cooperation, 7 This last part is supposed to preserve deniability even in cases where the speaker’s meaning seems fairly transparent.

Is implicit communication a way to escape epistemic vigilance?

99

as the speaker’s interests may diverge from the hearer’s interests. This seems misguided as regards the relevant mechanism of interpretation. Any view about implicit communication has to incorporate some mode of cooperation, whether Gricean or not Gricean, in the general sense that the speaker has to tailor her utterance to what she believes to be her hearer’s abilities and preferences, if she wants to be understood. Equally, unless the hearer has some reason to think that her utterance will not be of interest to him, he should make the effort of trying to recover the speaker’s meaning, again taking into account her own abilities and preferences. At least in this prudential sense, communication has to be mutualistic, i.e. the speaker’s and the hearer’s interests have to converge, if communication is to succeed. This, however, does not mean that, beyond the communicative episode itself, the interests of the speaker and the hearer may not diverge. They both may have further and more or less incompatible intentions. The second problem is that there may be more to implicit communication than just avoidance of institutional or social sanctions.8 In this respect, the Theory of the Strategic Speaker is subject to the same objection of incompleteness that Pinker puts forth in relation to Politeness Theory; even though his account may explain quite a lot about conversational implicature use, it does not explain all of it. At the same time, it entirely ignores presupposition, which is included in the present inquiry.

2.4 Taking stock As already mentioned in the first section of this paper, implicit communication is costly, both for the speaker who runs the risk of being misunderstood, and for the hearer who undergoes additional interpretive costs. This raises the question of why we resort to implicit communication (in the current setting, presuppositions and conversational implicatures) at all, given that, in most, if not all, cases, we could convey what we intend to communicate using explicit means, which would be less costly for both the speaker and the hearer. Having examined a number of answers to this question, we have found them wanting: – –

Structural explanations in terms of Universal Grammar fail in as much as conversational implicatures seem to be at most only weakly linguistic; Explanations in terms of the economics of communication fail because implicit communication is more costly than the proponents of this view claim it to be;

8 As exhibited in Pinker’s discussion of the example of sexual innuendos, mentioned in the previous subsection.

100 –

Anne Reboul

Finally, social explanations, though basically sound, are incomplete in two ways: – they only account for conversational implicatures at best; – they do not account for all uses of implicatures.

That said, going for a social account that might incorporate or complete Pinker’s Strategic Speaker Theory seems like the most promising road. One of the limitations of the Strategic Speaker Theory, however, is that, from a social point of view, it targets specific social phenomena where implicit communication is used to avoid institutional or social costs. More precisely, the speech acts concerned seem to be mainly of the directive sort (e.g. orders or requests), while implicit communication can occur in any variety of speech acts, including, not least of all, constatives (e.g. assertions).

3 Epistemic vigilance and implicit communication The use of implicit communication in constatives may be due to politeness considerations (e.g. the speaker wants to contradict her hearer’s previous utterance without thereby hurting his “face”), but, again, this seems too restricted to account for all the use of implicit communication in constatives. Additionally, as Pinker rightly pointed out, Politeness Theory takes for granted that the speaker’s and the hearer’s interests necessarily converge in communication, which is at best doubtful: as already mentioned above, they converge inside the communication process in a prudentially mutualistic way, but may diverge in the further respective motivations of the participants. This may be the place to note that most accounts of communication are extremely sanguine as to the social attitudes of communicators: these are usually described as “altruistic” (rather than merely – and more reasonably – mutualistic), and it is not always obvious what exactly is meant by that and on what basis this characterisation is given. So, let’s first clarify what is at stake here.

3.1 Social attitudes and varieties of social interactions An important first step is to give a definition of the relevant social attitudes, something that is all too rarely attempted in accounts of communication. Such definitions have been pursued in biology (see, for example, Trivers 2002), in terms of evolutionary fitness. It is, however, still possible to give a definition

Is implicit communication a way to escape epistemic vigilance?

101

in terms of the protagonists’ respective interests, couched in terms of cost and benefit. Basically there are four relevant social attitudes: – – – –

Mutualism, which occurs when a cooperative or collaborative action is directly beneficial to both participants; Altruism, which occurs when a cooperative action is beneficial to the recipient, but costly for the agent; Exploitation, which occurs when a manipulation is beneficial for the agent, but neutral (neither costly nor beneficial) for the recipient; Free-riding, which occurs when a manipulation is beneficial for the agent and costly for the recipient.

As follows from the definitions, these attitudes manifest themselves in different kinds of interactions, which will typically involve communication of some sort, but which are not necessarily communicative in nature. These include: –

– –

Collaboration, which occurs when an activity is performed by several individuals, and where the behaviour of each participant is coordinated with the activity of the other participants, with all participants having a common goal; Cooperation, which occurs when an agent does something beneficial for the recipient; Manipulation, which occurs when an action by an agent toward a recipient leads the recipient to perform an action that the recipient might otherwise not have performed, and which is beneficial for the agent.9

According to these definitions, cooperation may be either altruistic or mutualistic. There is, thus, no reason to deduce from the fact that human communication involves cooperation that it is altruistic. Indeed, as argued in section 2.3. above, there is good reason to think that it is mutualistic. Its being mutualistic, however, does not preclude the communicators from entering the communication process with other, less savory motivations and social attitudes, such as exploitation and free-riding. These attitudes, which are typical of manipulation, may lead humans10 to aim at changing one another’s mental states (e.g. beliefs or desires) in order to induce specific actions. This may, but need not, involve deception. But, in any case, it is to the advantage of the audience to be at least mildly cautious about what it is told, especially given 9 Note that, in manipulation, communication is necessarily involved. 10 In non-human animals the motivation is usually considered to bear directly on the other’s behaviour.

102

Anne Reboul

the possibility of decoupling,11 which makes deception easier (cf. Trivers 2011). As proposed by Sperber et al. (2010), this has probably led to the development of a range of mechanisms dedicated to epistemic vigilance in human cognition.

3.2 Epistemic vigilance According to Sperber et al. (2010), there are two grounds for epistemic vigilance. First, as mentioned above, the interests of the communicators may diverge rather than converge (beyond the prudential mutualism and cooperation entailed by successful communication), leading to manipulation and, occasionally, to deception. Second, as Mercier and Sperber (see Mercier 2009; Mercier and Sperber 2011) have convincingly argued, human reasoning is, first and foremost, dedicated to argumentation, i.e. to persuading one’s audience that one’s own opinion is better than the audience’s. This implies cognitive mechanisms for both the production of reasons and the evaluation of reasons produced by others. It also suggests that human reasoning and language co-evolved. In this respect, it seems that one of the two components of human reasoning, the evaluation of others’ reasons, is already a tool for epistemic vigilance. Two additional detection mechanisms of epistemic vigilance target respectively: – –

the source: is it reliable and competent? and the communicated content: is it compatible with previously held assumptions?

Thus, though epistemic vigilance is, up to a point, costly, some of its costs are mandatory, given that the mechanisms concerned are part of the interpretation process. In most cases, this will be true of the verification of compatibility, as far as the propositions in the context are concerned, since interpretation is always context-based (regardless of whether or not the utterance includes an implicature or a presupposition). This may be further extended to propositions beyond the context in cases where the source is of dubious credibility. This maximal sort of vigilance, however, is unlikely to be permanent, given its costs. It will vary with the relevance of the communicated content: the more relevant it is, the more effort the hearer should invest in processing it and the more vigilance he should exercise. Nonetheless, some minimal epistemic vigilance is always present, which probably explains why young children, despite being 11 Decoupling has to do with the possibility of referring to an object or describing a situation in the absence of the object referred to or the situation described.

Is implicit communication a way to escape epistemic vigilance?

103

generally trustful, can detect and reject some misinformation (see, for example, Harris 2012). This quick sketch of epistemic vigilance suggests that, in a classical arms race, speakers should have developed mechanisms to overcome it. In what follows, I will suggest that implicit communication has developed, at least in part, to eschew epistemic vigilance. Before I do so, however, I want to first point out that there is at least one other relevant mechanism that, though not strictly speaking part of epistemic vigilance, complements it all the same.

3.3 The egocentric bias Even though manipulation may be deceptive in the information communicated, it arguably need not be: one may give true, but incomplete information, leading one’s addressee to a misguided decision or an incorrect belief. Now, since the information at hand is true, this manipulation will not be detected by compatibility checking. In such cases, it may be in the addressee’s interest to stick as far as possible to his own opinion. My hypothesis is that this is the root of the egocentric bias described by Mercier (2009), that is, a preference for one’s own over communicated beliefs. In other words, the egocentric bias steps in, as a supplementary mechanism, where epistemic vigilance might be non-operant. Even though it has to be limited, as one’s own opinion might turn out to be misguided, in the sense that it can be overturned by strong enough evidence to the contrary, it is nevertheless extremely widespread. In this respect, the egocentric bias makes sense as a defense mechanism not so much against deception, as against non-deceptive manipulation.

3.4 Implicit communication as a mechanism to bypass epistemic vigilance and associated mechanisms Taking stock, there are a few interesting consequences of both epistemic vigilance and the egocentric bias for our inquiry of why we have implicit communication: – –

people will tend to hold more strongly to conclusions that they have reached themselves; people will tend to be less epistemically vigilant toward information on the truth of which speakers do not seem to commit themselves, that is, information that is not asserted;

104 –

Anne Reboul

people will tend to be less epistemically vigilant toward information that is not presented to them as a reason to change their own beliefs or decisions, that is, information which is not perceived as communicated with a manipulative or argumentative intention.

In what follows, I wish to argue that varieties of implicit communication have all or most of these characteristics. This leads me to my main hypothesis, which I will call the Manipulation Hypothesis: Implicit communication (i.e. conversational implicatures and presuppositions) has emerged in order to facilitate manipulation and argumentation, by allowing the speakers to hide their intentions, avoid epistemic vigilance and bypass the egocentric bias.

Though the hypothesis covers both conversational implicatures and presuppositions, the mechanisms involved in each are rather different, as we will now see.

3.4.1 Commitment Let’s first look at an example of conversational implicature: (7)

A: Do you know where Anne lives? B: Somewhere in Burgundy I believe.

The conversational implicature of B’s answer is that B does not know where exactly Anne lives. Note that B may know this perfectly well, but may not want to share this information so that A does not write to Anne or visit her. By using a conversational implicature, B can hide her intention to not give A this piece of information, or even later deny having had that intention in the first place. Furthermore, B can deny having meant that she did not know where Anne lives as well. All of this comes from the fact that implicatures are defeasible. In fact, B could even cancel her implicature in the very same utterance, as shown in (8), without rendering it contradictory or problematic in any way: (8) Anne lives somewhere in Burgundy, in Cluny as a matter of fact. This non-contradictory defeasibility of conversational implicatures has a consequence for commitment: The speaker of an utterance that communicates a conversational implicature does not commit herself to the truth of this implicature, but to the truth of the utterance’s explicit content.

Is implicit communication a way to escape epistemic vigilance?

105

Let me quickly outline what I mean by commitment. In order to do so, I will go back to Pinker’s example of the driver bribing the police officer, presented in section 2.3.2. above. In deciding to bribe the officer, she can do so either explicitly, as in (9), or implicitly, as in (10): (9)

Here, Officer, how much would you want to forget about my exceeding the speed limit?

(10) I am awfully sorry, Officer! Couldn’t we solve this little misunderstanding in a civilised way? The difference between the two options is that the speaker of (9) makes it undeniable and mutually manifest to both herself and the officer that she is offering a bribe to him. So to speak, she puts herself on record as having done so, or, to put it more simply, commits herself to having offered a bribe. On the other hand, in uttering (10), she does no such thing.12 This has a few consequences. For one, if you conversationally implicate something you know to be false, you are not thereby strictly speaking lying. Then, your hearer will be much more vigilant toward what you have explicitly communicated, which may well be true, than toward the content that you have conversationally implicated because: – –

you haven’t committed yourself to the truth of that content; it is a content that he has reached by himself (in other words, his own conclusion) and it is thus favoured by his egocentric bias.

Turning to presuppositions, things are rather different. Consider, for example, the exchange in (11): (11)

A: I have decided to give John the manager job. B: That’s an excellent idea, especially now that he has stopped drinking.

Here, the presupposition is obviously that John used to drink. Again, B can claim that she has no hostile feelings toward John and no intention to change A’s decision; on the contrary, she can even point out that she has praised A’s choice. What she cannot do, however, is deny having meant that John drank.

12 This is the heart of the logic of plausible deniability. It also explains why one cannot have Moore’s paradox (e.g. It rains and/but I don’t believe it rains) over an implicature. That is why Anne lives somewhere in Burgundy and/but I know exactly where may be long-winded, but there is nothing pragmatically weird about it.

106

Anne Reboul

Indeed, one major difference between conversational implicatures and presuppositions is that presuppositions are not defeasible. That is why an utterance like (12) is undeniably strange:13 (12)

?John has stopped drinking and he never drank in his life.

So, clearly, presuppositions do not work in the same way as conversational implicatures do. Nevertheless, the presupposed content is still not on a par with the explicitly communicated content. In other words, John has stopped drinking is not equivalent to (13): (13) John does not drink and John used to drink. 3.4.2 The layering of information One should note that, in an utterance with a presupposition, the presupposed content is not available for continuation with a causative (e.g. because), while the explicitly communicated content is (cf. Ducrot 1972; Jayez 2010, 2015). Thus, while (14), where the continuation with because bears on the explicit content ‘John doesn’t drink’, is fine, (15), where the continuation with because bears on the presupposed content ‘John drank’, is not: (14) John has stopped drinking because his doctor told him to do so. (15)

?John has stopped drinking because he liked the feeling of euphoria alcohol gave him.

Admittedly, this is a common feature of both presuppositions and conversational implicatures. For instance, (16), where the relevant continuation with because bears on the explicitly communicated content (Anne lives somewhere in Burgundy), is fine, whereas (17), where it bears on the implicature (I don’t know exactly where), is not: (16) Anne lives somewhere in Burgundy, because she likes that part of France. (17)

?Anne lives somewhere in Burgundy, because I was never told exactly where.

13 In connection to the previous footnote, it is clear that in contrast with conversational implicatures, presuppositions can be the target of a Moorean paradox. So, unless interpreted metalinguistically, as a denial of a content previously communicated by someone else, John has stopped drinking and/but I don’t believe that he ever drank is clearly strange.

Is implicit communication a way to escape epistemic vigilance?

107

In other words, in both presuppositions and conversational implicatures, there is a layering of information to the effect that the implicitly communicated information (be it a presupposition or an implicature) is not salient and available to the same degree. Thus, the hearer may not be aware to the same degree of the two sorts of contents, which effectively allows them to escape the radar of epistemic vigilance. One, however, could object to this argument, by suggesting that continuations on the implicit content are possible with other connectives, such as but,14 as there is nothing wrong with either (18) or (19): (18) John has stopped drinking, but he liked the feeling of euphoria alcohol gave him. (19) Anne lives somewhere in Burgundy, I believe, but I was never told exactly where. So, it might be that the weirdness of (15) and (17) is linked to the semantics of because rather than to the implicit or explicit way in which the content it picks up on is communicated. To my mind, however, this is not an argument against the layering of information, since some connectives may have wider access to different layers of content than others. If this is correct, accounting for their behaviour can still not eschew an appeal to layered contents. 3.4.3 Implicit content as the hearer’s conclusion Indeed, the possibility of a continuation with but bearing on the implicitly communicated content does not only sustain the layering of information proposed above, but also opens up another perspective on the specificity of implicit communication and the opportunity it offers for bypassing epistemic vigilance and the egocentric bias. According to Ducrot (1972, 1980), there is a contrast between two uses of but15: the rectificative use (usually following a negative first component), as in ‘Jean is not Belgian, but French’, and the use illustrated in (18) and (19) above. With respect to the latter (p but q), Ducrot provides an analysis that is directly relevant here: q contradicts not p itself, but a certain conclusion, r, that can be derived from p. Additionally, q is a stronger argument against r than p is for r. This elegant account is full of interesting implications for 14 Thanks to Didier Maillat for having pointed this out to me in the discussion following my talk at the Intercultural Pragmatics Conference in Malta (30 May–1 June 2014). 15 Ducrot also noted that in some languages such as German and Spanish (though not in French or English), these two uses correspond to two different words: sondern and aber in German, sino and pero in Spanish (1980: 11).

108

Anne Reboul

the analysis of implicit communication, and notably of presupposition. So, let’s come back to examples (18) and (19), reproduced here for convenience: (18) John has stopped drinking, but he liked the feeling of euphoria alcohol gave him. (19) Anne lives somewhere in Burgundy, I believe, but I was never told exactly where. (18) conforms entirely to Ducrot’s analysis, since but picks up on the presupposition John used to drink and not on the explicitly communicated content John doesn’t drink now. Notably, a continuation with but on the explicit content would be unacceptable: (20) ?John has stopped drinking, but his doctor told him to do so. Interestingly, on the only interpretation that would make (20) acceptable, but would not pick up on the explicit content (John doesn’t drink now), but rather on another potential conclusion, i.e. that John decided on his own to stop drinking. What this suggests, given Ducrot’s analysis, is that it is because of the presupposition being a conclusion that can be drawn from ‘John has stopped drinking’ that (20) makes perfect sense in this context. Conversely, it is for a rather different reason that (19) is also fine: it is not because but picks up on the implicature I don’t know exactly where, but rather because it picks up another conclusion that can be drawn from ‘Anne lives somewhere in Burgundy, I believe’. This alternative conclusion is that B knows precisely where Anne lives. Now, at face value, this might seem mysterious, since the conclusion I know exactly where seems less accessible than the implicature I don’t know exactly where; yet, this is most likely where defeasibility, a major factor that differentiates between conversational implicatures and presuppositions, comes in. It is therefore defeasibility, which concerns exclusively the implicature and not the alternative conclusion, that seems to make the difference as far as but is concerned. This is further corroborated if one looks at scalar implicatures16 too: 16 Here, it should be noted that a continuation with because, picking up on the implicature, is possible for scalars, as (i) below shows: (i) The pianist has played some Mozart sonatas, because he had little time. This may have to do either with the distinction between strongly communicated and weakly communicated implicatures proposed by Sperber and Wilson (1995), or with the fact that scalar implicatures may belong to the level of explicature (see Carston 2002), in the sense that they belong to the pragmatically enriched proposition that an the utterance explicitly expresses.

Is implicit communication a way to escape epistemic vigilance?

(21)

109

The pianist played some Mozart sonatas, but he had little time.

Again, in (21) but seems to not pick up on the implicature he did not play all of them, but on an alternative conclusion directly derived from the utterance’s explicit content (in which some is interpreted as at least some), i.e. He played all Mozart sonatas. If this is right, Ducrot’s analysis has to be complemented: the conclusion r, which q contradicts, cannot be defeasible, i.e. it cannot be an implicature. Of course, this does not mean that implicatures are not conclusions that the hearer reaches by himself, but rather that they are not among the conclusions that the hearer reaches by himself and that are available for continuation with but. In this respect, presuppositions, like implicatures, are also conclusions that the hearer reaches by himself.

3.4.4 Summing up In conclusion, apart from the three different features which were indicated in the opening part of section 3.4 as those allowing communication to eschew epistemic vigilance and bypass the egocentric bias, differential accessibility should also be added to the list. This gives us the following list: a.

The fact that the information is perceived by the hearer as his own conclusion rather than as information coming from someone else; b. The absence of speaker’s commitment relative to the truth of the information; c. The accessibility of the information (due to the layering described in section 3.4.2); d. Hiding one’s manipulative or argumentative intentions. As I have argued in this paper, both presuppositions and conversational implicatures allow the speaker to hide her argumentative or manipulative intentions (condition d), and both of them are less accessible than the explicit content (condition c), as well as perceived by the hearer as his own conclusion rather than as information coming from the speaker (condition a). It is only conversational implicatures and not presuppositions, however, which allow the speaker to withdraw her commitment to the truth of the implicitly communicated content (condition b). Thus, implicit communication seems to be calibrated for eschewing epistemic vigilance and the egocentric bias, which in turn strongly supports the Manipulation Hypothesis.

110

Anne Reboul

4 Conclusion Implicit communication is obviously more costly than explicit communication for both the speaker and the addressee, since it increases the risk of misunderstanding, making it less sure for the speaker to secure the hearer’s uptake, and demanding greater efforts on the hearer’s part to attain uptake too. On the other hand, explicit communication is interpretively cheaper and offers a greater guarantee of success. Finally, though it might be argued that implicit communication might be the only way the speaker has to communicate her intended content, this seems to hold only in relation to rather specific cases, such as metaphor, and can hardly be the case for scalar implicatures and presuppositions. Against this background, the question of why we have implicit communication in the form of conversational implicatures and presuppositions arises. In this paper, I examined and found wanting a number of potential answers to this question: that implicit communication is part of Universal Grammar, that it is a result of the economics of communication, or that it emerged to allow politeness. Among these, only Pinker’s Strategic Speaker Theory seems to be on the right track, but lacks generality. In an attempt to provide a more plausible answer to the question at hand, I broadened the putative function of implicit communication beyond the avoidance of social sanction, to include a wider function of eschewing epistemic vigilance and bypassing the epistemic bias. After listing a number of features that allow speakers to do this, I showed that all of them are found in one form or another of implicit communication. While this on its own may not confirm the hypothesis that implicit communication emerged to allow speakers to bypass vigilance, it still strongly supports it. In closing, I would like to add a final argument, even though, due to space restrictions, I can hardly discuss it as extensively as it deserves to be discussed here. If the Manipulation Hypothesis is right, implicit communication emerged as part of an arms race, which in turn raises the further question of why a counter-mechanism restoring vigilance was not also developed. A tentative answer to this question is that such a mechanism does exist and is actually responsible for the relatively low rate of pragmatic answers in experiments targeting scalar implicatures (where speakers give up to 40% of semantic interpretations depending on the paradigm). Though no one would expect hearers to be at ceiling for other, non-scalar types of conversational implicatures, one would expect them to be at ceiling for scalar implicatures, given that they are, after all, dependent on particular lexical items. Yet, the best rate obtained never seems to go past 80% (see, for example, Bott and Noveck 2004; Bott, Bailey and

Is implicit communication a way to escape epistemic vigilance?

111

Grodner 2012; Noveck 2000; Noveck and Posada 2003), even when there is a context that strongly encourages the hearer to draw the implicature. A possible explanation is that this is due to a defense mechanism along these lines. Rather than extending their epistemic vigilance, and increasing even more the interpretive cost for conversational implicatures, hearers just ignore them in a number of cases. While this is only a speculative explanation at this stage, I believe that it can provide an answer to what, on the face of it, seems to be an enigma, while at the same time going one step further in support of the manipulation hypothesis.

References Bott, Lewis, Todd M. Bailey & Daniel Grodner. 2012. Distinguishing speed from accuracy in scalar implicatures. Journal of Memory and Language 66. 123–142. Bott, Lewis & Ira Noveck. 2004. Some utterances are underinformative: The onset and time course of scalar inferences. Journal of Memory and Language 51. 437–457. Brown, Penelope & Stephen C. Levinson. 1987. Politeness: Some universals in language use. Cambridge: Cambridge University Press. Carston, Robyn. 2002. Thoughts and utterances: The pragmatics of explicit communication. Oxford: Blackwell. Ducrot, Oswald. 1972. Dire et ne pas dire. Paris: Hermann. Ducrot, Oswald. 1980. Analyses pragmatiques. Communications 32. 11–60. Goffman, Erwin. 1967. Interaction ritual: Essays on face-to-face behavior. New York: Anchor Books. Grice, H. Paul. 1989. Studies in the way of words. Cambridge, MA: Harvard University Press. Harris, Paul L. 2012. Trusting what you’re told: How children learn from others. Cambridge, MA: Harvard University Press. Horn, Laurence R. 2004. Implicature. In Lawrence R. Horn & Gregory Ward (eds.), The handbook of pragmatics, 3–28. Oxford: Blackwell. Horn, Laurence R. 2006. The border wars: A neo-Gricean perspective. In Laurence R. Horn (ed.), Where semantics meets pragmatics: The Michigan papers, 21–48. Oxford: Elsevier. Horn, Laurence R. 2007. Neo-Gricean pragmatics: a Manichean manifesto. In Noel BurtonRoberts (ed.), Pragmatics, 158–183. Basingstoke: Palgrave Macmillan. Horn, Laurence R. 2009. WJ-40: Implicature, truth and meaning. International Review of Pragmatics 1: 3–34. Jayez, Jacques. 2010. Projective meaning and attachment. In Maria Aloni, Harald Bastiaanse, Tikitu de Jager & Katrin Schulz (eds.), Logic, language and meaning, 325–334. Berlin: Springer. Jayez, Jacques. 2015. Orthogonality and presupposition: A Bayesian perspective. In Henk Zeevat & Hans-Christian Schmitz (eds.), Bayesian natural language semantics and pragmatics, 145–178. Berlin: Springer. Jayez, Jacques & Bob van Tiel. 2012. Only only? An experimental window on exclusiveness. In Maria Aloni, Vadim Kimmelman, Floris Roelofsen, Galit W. Sassoon, Katrin Schulz & Matthijs Westera (eds.), Logic, language and meaning: Revised selected papers of the 18th Amsterdam colloquium, 391–400. Berlin: Springer.

112

Anne Reboul

Lee, James L. & Steven Pinker. 2010. Rationales for indirect speech: The theory of the strategic speaker. Psychological Review 117. 785–807. Levinson, Stephen. 2000. Presumptive meanings: The theory of generalized conversational implicatures. Cambridge, MA: The MIT Press. Mercier, Hugo. 2009. La théorie argumentative du raisonnement. Paris: E.H.E.S.S. PhD thesis. Mercier, Hugo & Dan Sperber. 2011. Why do humans reason? Arguments for an argumentative theory. Behavioral and Brain Sciences 34. 57–111. Noveck, Ira. 2000. When children are more logical than adults: Experimental investigations of scalar implicature. Cognition 78. 165–188. Noveck, Ira & Andres Posada. 2003. Characterizing the time course of an implicature: An evoked potentials study. Brain and Language 85. 203–210. Pinker, Steven. 2007. The evolutionary social psychology of off-record indirect speech acts. Intercultural pragmatics 4. 437–461. Pinker, Steven, Martin A. Nowak & James L. Lee. 2008. The logic of indirect speech. Proceedings of the National Academy of Sciences 105. 833–838. Sperber, Dan, Fabrice Clément, Christophe Heintz, Olivier Mascaro, Hugo Mercier, Gloria Origgi & Deirdre Wilson. 2010. Epistemic vigilance. Mind & Language 25. 359–393. Sperber, Dan & Deirdre Wilson. 1995. Relevance: Communication and cognition, 2nd edn. Oxford: Basil Blackwell. Trivers, Robert. 2002. Natural selection and social theory: Selected papers of Robert Trivers. Oxford and New York: Oxford University Press. Trivers, Robert. 2011. Deceit and self-deception: Fooling yourself the better to fool others. London: Penguin. von Fintel, Kay & Lisa Matthewson. 2008. Universals in semantics. The Linguistic Review 25. 139–201.

Thanh Nyan

7 The structure of utterance meaning: how did it come about? Abstract: The meaning of utterance types, as defined by Anscombre and Ducrot’s argumentation theory (TAL), includes a constraint on the context of utterance, and more specifically on the act of utterance. Building on previous work on linguistic constraints on context, this article is an attempt to link the above constraint to the structure of intentional action, a concept arising from Jeannerod’s work on motor cognition (1994; 2006). Such a linkage can be envisaged from an adaptive perspective, which allows for the possibility of language availing itself of pre-existing solutions.

1 Introduction According to the ‘Theory of argumentation in language’ (henceforth TAL,1 Anscombre and Ducrot 1983; Anscombre 1995), the meaning of utterance-types is first and foremost to be construed in terms of constraints, which relate to what precedes and what follows utterances, in the sense of the act of utterance and prior interactions, on the one hand, and continuations, on the other.

Figure 1: The structure of utterance meaning as construed by the TAL.2

Consider, for example, (1): (1)

It’s not raining outside.

Prior to any contextual input, the meaning of (1), as analysed by TAL, contains specifications relating to the act of utterance, which, in this instance, is the 1 TAL stands for ‘théorie de l’argumentation dans la langue’. 2 The arrow pointing to the right, symbolises a constraint on continuations; the one pointing to the left, constraints on the act of utterance and on prior interactions. Thanh Nyan, SALC University of Manchester DOI 10.1515/9781501505089-007

114

Thanh Nyan

intention to bring about an assertive act, prior interactions, which include a viewpoint, attributable to the hearer, that it is raining, and appropriate continuations, which are more likely to be along the lines of (2), rather than (3), since it is (2) that depicts a stance more likely to be taken in response to the state of affairs described in the propositional content of (1): (2)

Let’s go for a walk.

(3)

Let’s check out that Klee exhibit.

Insofar as this conception of utterance meaning diverges from well-established paradigms by virtually excluding informational contents from consideration3 and by taking an integrated view of pragmatics, it has been the subject of much controversy. Rather than evaluating it in relation to theoretical assumptions arising from various linguistic frameworks, however, I would like, in this paper, to ask a different type of question; one that might further our understanding of why utterance meaning has the structure that it does, should TAL prove to be basically correct in its construal. This question is: how did utterance meaning come to have the type of structure under scrutiny, and, more specifically, how did it come to include a constraint on the act of utterance? Such a question, which presupposes an adaptive perspective, is amenable to an answer in neuro-cognitive terms, rather than one constrained by how one construes the interface between semantics and pragmatics, or the relationship between informational content and discourse dynamism. In this vein, the type of answer I have in mind runs as follows: i)

The concept underlying utterance meaning is (at least in part) associated with a type of intentional action (as seen from a 1st person perspective), namely the act of utterance. ii) In addition to a physical component, the act of utterance has a mental component, or motor representation (or motor intention), as construed by Jeannerod (1994, 2006) iii) It is this motor representation that makes its way into the conceptual structure of utterance meaning to give rise to the constraint in question, which respectively corresponds to a component of the motor representation. iv) This comes about because the motor representation at hand has an adaptive function at the interpretive level. 3 While TAL makes the decision to rule out informational contents from utterance meaning, Nølke’s ScapoLine (1984), which is inspired by the TAL, chooses not to, on the basis of the same type of data.

The structure of utterance meaning: how did it come about?

115

Against this background, this paper is organised as follows. Section 2 briefly outlines the adaptive perspective that the present discussion adopts. Section 3 seeks to analyse the utterance act in terms of Jeannerod’s model of intentional action. This involves sketching out Searle’s (1983) conception of intentional action, which pertains to the personal level of description, before turning to Jeannerod (1994, 2006), for what might be looked upon as a sub-personal level analogue of Searle’s construal of the structure of intentional action. Finally, Section 4 puts forward a tentative explanation as to how the structure of intentional action would have made its way into utterance meaning.

2 Adaptive assumptions The adaptive perspective this paper relies on is the neo-Darwinian theory of evolution, which arises from the Evolutionary Synthesis of the 1930s and 1940s (cf. Futuyma 1998: 17–30; Mayr 1980: 1–46). Also known as the “synthetic theory,” it reconciles Darwin’s theory of evolution with facts of genetics and its adoption commits us to a set of basic assumptions, of which the key ones, for our purposes, are: a) Evolution tends to recycle existing solutions when faced with the same type of problems (Damasio 1994: 190). b) Language and the brain have co-evolved with language being the latecomer in the evolutionary process (cf. Deacon 1987). c) The language system is rooted in our neurobiology and functions depending largely on structures that are already in place and on pre-existing systems, such as the motor and perceptual as well as the concept formation one (cf. Edelman 1989: 109–148). Another set of relevant adaptive assumptions that I adopt arise from Damasio’s (2010) theory of brain functions and specifically concern representations, conceptual knowledge and concepts. According to Damasio, the brain continuously generates representations, or brain maps as he calls them,4 of external and internal entities and activities, such as action production, for monitoring purposes (2010: 17–18 & 132). These maps are also continuously discarded, unless they, directly or indirectly, serve an adaptive function, such as recall, learning

4 Brain maps are patterns of activity of circuits of neurons “that represent things and event located outside the brain, either in the body or in the external world, [with] some patterns also represent[-ing] the brain’s own processing of other patterns.” (Damasio 2010: 17–18)

116

Thanh Nyan

or action selection. In this picture, conceptual knowledge is information about “nonlanguage interactions between the body and its environment, as mediated by varied sensory and motor systems” (Damasio and Damasio 1992: 89)5 and the “basis for a concept is a collection of [. . .] sensory motor representations” corresponding to “many separate neural population ensembles distributed in various cortical regions” (Damasio 1989: 24–25). In this respect, this neural basis gets activated each time a concept is evoked by means of a stimulus, be it verbal or nonverbal.

3 The structure of intentional action Since Jeannerod builds on a basic distinction introduced by Searle (1983), I will begin with an overview of Searle’s approach, before moving on to present Jeannerod’s model of intentional action.

3.1 Searle’s conception of intentional action According to Searle, intentional actions are caused by intentions,6 which contain a representation of the intended action and a self-referential causality condition. The need for this condition arises because actions that match the representation within a given intention are not necessarily caused by it. For an action to be intentional in relation to an intention, it has to be performed by way of carrying out that intention. Suppose, for example, you wake up one morning intent on breaking a certain cup you have no other way of getting rid of. Suppose further that later in the day you completely forget about that intent, but accidentally drop the cup on the floor. In Searle’s analysis, this event, no matter how welcome, 5 This view of linguistic concepts arises from the model of language processing proposed by Damasio and his colleagues (Damasio and Damasio 1992; H. Damasio et al. 1996; H. Damasio et al. 2004; Tranel et al. 2001). According to this model, language processing involves a number of neural sites besides the Broca and Wernicke areas. The first supports conceptual knowledge (as just described); the second, “the implementation of word forms in eventual vocalizations” (Tranel et al. 2001: 656); the third consists of intermediary structures that broker between the other two (H. Damasio et al. 2004: 221). 6 Intentions are intentional states, of which further examples include belief, hope and fear. Intentional states are mental states “directed at or about objects and states of affairs in the world” and “can be said to be satisfied or not satisfied depending on whether the representation they include of the entities they are directed at actually matches or represents anything in reality” (Searle 1980: 48)

The structure of utterance meaning: how did it come about?

117

will not be intentional in relation to your (initial) intention, as this intention had no causal role in its occurrence. In relation to an action such as raising one’s arm, Searle distinguishes two types of intention: a prior intention, which is formed in advance of the action, and is not necessary for an action to be intentional,7 and an intention-in-action,8 the mental component of the action, which causes its physical component, in this case the bodily movement. Unlike the prior intention, the intention-in-action cannot be separated from the action.

Figure 2: Intentional action (based on Searle 1983: 94)

In this picture, in terms of representative content, the prior intention, which, in Searle’s own terms, can be expressed as “I perform the action of raising my arm by way of carrying out this intention” (1983: 92), has the whole action (i.e., the event whose occurrence would carry out the intention) as its intentional object, and includes a self-referential causality condition that specifies that it is only satisfied if the action is caused by it. In turn, the intention-in-action has the bodily movement as its intentional object, and a self-referential causality condition to the effect that it is only satisfied if the movement is caused by it.9 Of these two intentions, it is the second – the intention-in-action – which appears to be better suited to our purposes. Like many everyday intentional actions, acts of utterance can occur quite spontaneously, without any prior intention being formed. Furthermore, insofar as intention-in action is a component of the act of utterance, it presents an obvious isomorphism with the constraint under consideration, which would indicate that we are on the right track. Having decided in favour of intention-in-action as the first term of the causality relation between intention and action, we can now turn to the way in which this intention might cause the intended action. In his account, Searle places great emphasis on causality by having a self-referential causality condi7 As Searle points out (1983: 84), “. . . many of the actions one performs, one performs quite spontaneously, without forming, consciously or unconsciously, any prior intention to do those things.” An example he gives is that of suddenly getting up from his chair and starting to pace about. These actions are “clearly intentional actions”, yet one does not need to “form an intention to do them prior to doing them”, in the sense that one does not “have to have a plan to get up and pace about.” 8 Searle also uses the expression ‘experience of action’ to refer to intention-in-action when it is conscious. 9 Searle points out that “in any real-life situation the intention in action will be much more determinate” and will indicate how the arm is to go up, and at what speed, etc. (1983: 93).

118

Thanh Nyan

tion as part of the intentional content. However, he is not concerned with how an intention can cause the action that is its intentional object, even though, to be fair, his primary goal is to integrate intentional action to his theory of intentionality rather than specify how action execution is carried out. In light of this, I propose that we turn to Jeannerod (1994, 2006), who specifically focuses on the causal link between intention and action (or at least the earlier part of it), and takes a sub-personal view of intentional action, while relying on distinctions similar to Searle’s.

3.2 Jeannerod’s conception of intentional action In Jeannerod’s account, motor intentions are a type of action representation which is “closely linked to action,” and, as such, contributes to “the stream of processing [which] goes from the early part of the representation [of action] to action execution, when it occurs” (Jeannerod 2006: 59). For Jeannerod, representing an action and executing it are on the same continuum: the way motor intention contributes to action execution is by “simulating” it, which entails activating “mechanisms which prepare for the forthcoming action” (Jeannerod 2006: 165), thereby rehearsing the pathways involved in potential execution. The idea of a functional continuum between representation and action, which presupposes a common neural basis, capable of different levels of activation arises from the following findings: a) Motor intention shares the “same neural mechanisms” as motor imagery10 (i.e. the result of imagining a movement) (Jeannerod 1994: 192), which also influences the performance of a movement, as reported by numerous studies in sport psychology. (Jeannerod 1994: 191) b) Mental imagery is “accompanied by sub-threshold motor activation and even small movements.” (Jeannerod 1994: 191) In other words, if motor intention is functionally equivalent to motor imagery11 in relation to action execution, and shares with it a common neural basis, motor 10 Unlike motor intention, motor imagery is accessible to consciousness, hence the possibility of identifying its contents, which then can be assumed to be similar to those of motor intention, the non-conscious counterpart of motor imagery. 11 The relatedness of motor preparation and imagery has an important methodological consequence. If motor preparation and imagery have the same goals, “they should contain the same information” (Jeannerod 1994: 193). In other words, in order to determine the parameters represented during motor preparation, one can avail oneself of the fact that the contents of imagery are accessible to consciousness. This methodological stance, to which Jeannerod commits, may go some way towards explaining why he endorsed Searle’s distinctions: the latter he may well have regarded as the result of introspection of the kind associated with mental imagery.

The structure of utterance meaning: how did it come about?

119

intention, like motor imagery, can also be assumed to integrate a low level of motor activation, which can be stepped up during the action execution phase.

Figure 3: From motor intention to action execution.

As shown in Figure 3, which on the whole represents the action production process, the simulation process (symbolised by the thick arrow) occurs, regardless of whether action execution (symbolised, in turn, by the thin arrow) follows. Now, given that motor intention contributes to action execution by simulating it, the contents involved in that simulation are assumed to include representations of 12 (a) the “amount of force required to produce the desired motor effect” (Jeannerod 1994: 194); (b) “rules and constraints inherent to executed actions” (Jeannerod 2006: 41), such as “kinematic rules for overt movements” or “biomechanical constraints of the represented movements” (Jeannerod 1994: 194–196)); and (c) goals.

12 Though non-conscious, these contents are assumed, on the basis of considerable experimental evidence, to be similar to those of another type of motor representation that are accessible to consciousness, namely, motor images, with which they “share common mechanisms and are functionally equivalent.” (Jeannerod 1994: 190). Examples of motor images include imitating someone’s movements, anticipating the effect of some action, feeling kinaesthetic or bodily sensations (muscle contraction, heart beats) (Jeannerod 1994: 189)

120

Thanh Nyan

In cases where the action is directed towards an object, which are the only cases whose goal Jeannerod discusses – like, for example, grasping a glass – the relevant goal is encoded by a representation of the object’s attributes, and a representation of the “final state of the organism when [the goal] has been reached” (Jeannerod 1994: 197). These object attributes are represented in the ‘pragmatic mode’,13 which describes them in a way that allows appropriate movements to be selected (Jeannerod 1994: 198); in other words, these “attributes are represented as affordances, that is, as the extent to which they afford specific motor patterns” (Jeannerod 1994: 199, see also Jeannerod 2006: 110). So, in such cases, motor intentions also contain motor patterns that are predetermined by the pragmatic mode of representation.14

Figure 4: Goal of actions directed towards an object.

Insofar as the motor effects under discussion pertain to a lower level of analysis and can be intermediate steps towards a more complex final goal, their motor intentions are likely to be influenced by higher level action representations which control “the selection, the activation or the inhibition” of processes at the lower level (Jeannerod 1994: 200). Moving on to higher level representations, Jeannerod suggests that they encode “the serial order of requisite movements”, as well as “final configurations (of the environment, of the body, of the moving segments, etc.) as they would arise at the end of the action” (Jeannerod 1994: 201). But still, the neurons which encode these configurations “remain active until the requisite configuration has been attained.” (Jeannerod 1994: 200–201)15

13 This pragmatic mode is to be distinguished from the semantic mode, whereby all the perceptual properties of the object are listed. 14 The possibility of ‘a neurophysical substrate’ for pragmatic representations lies in premotor neurons in area F5 which are “selective for a certain type of objects and the corresponding actions on these objects”, and can “discharge on observation of the object alone”, which entails that “they encode a represented action, i.e., an action that remains covert according to the instructions that the monkey has received” (Jeannerod 2006:110). 15 The neurons which encode these configurations would “remain active until the requisite configuration has been attained.” (Jeannerod 1994: 200–201)

The structure of utterance meaning: how did it come about?

121

Having outlined Jeannerod’s account of how motor intentions contribute to action execution, it is now possible to analyse utterance acts in these terms.

3.3 Motor intention and utterance act Assuming that utterance act is a form of intentional action, it can be expected to share the same production process as other types of intentional action (as shown in Figure 3 above). However, in attempting to describe the causal relation between motor intention and utterance act, one is faced with an obvious difficulty: the type of utterance act under scrutiny is not an end in itself, but an intermediate step in the performance of an illocutionary act. Given that illocutionary acts are not, physically, separate acts, even though we are again dealing with a complex goal, this rules out viewing them in terms of a sequence of (physical) actions, such as grasping a glass and raising it. A way around this problem might lie in the fact that the type of illocutionary act which is brought about depends (at least in the first instance) on the linguistic elements that inform the utterance act. These elements can then be specified in the motor intention as the linguistic means (e.g. type of syntax) to be selected in the production of the utterance act. The higher level of determinacy of intention in action that Searle mentions (1983: 93)16 would relate to appropriate linguistic choices. Now, if there is a further goal beyond the illocutionary act, like for instance, the goal of getting the hearer to engage in non-linguistic behaviour in response to the state of affairs depicted in the propositional content of the utterance at hand, then the motor intention would reflect this by encoding that state of affairs in the pragmatic mode. But how would such a mode manifest itself in the depiction of the state of affairs? Since at the object level, such a mode entails describing object attributes in a way “that can be used for the selection of appropriate movements” (Jeannerod 1994: 198), I wish to propose that, at the state of affairs level, it would be inherent in the choice of the objects to be represented and the aspect under which they are represented, both of which would be indicative of the intended type of action. This proposal is based on what Damasio (2010: 71–72) proposes in relation to perception (see also Edelman 1989: 97–98): during the perception of a situation, objects that have been singled out by adaptive values (arising from the genome) for their relevance to survival apparently have added saliency imparted 16 According to Searle “in any real-life situation the intention in action will be much more determinate” and will indicate how the arm is to go up, and at what speed, etc (1983: 93)

122

Thanh Nyan

to some of their features. This saliency, in turn, leads to processing resources being recruited to deal with whatever problem the objects represent. In this respect, the aspects highlighted by the pragmatic mode of representation would correspond to those rendered more salient by adaptive values, and the intended types of action would be appropriate responses selected by experience and made possible by the processing resources recruited by the added saliency. Figure 5 below shows how adaptive values impart added saliency to object attributes during perception, thereby causing processing resources to be recruited.

Figure 5: Processing resources recruitment.

It has been argued that, if utterance act is a type of intentional action, it can be expected to present the same structure as other intentional actions; a structure comprising a motor intention (or motor representation), whose contents appear to be (at least partially) isomorphic with utterance meaning. In light of this, there is a distinct possibility that the constraint on the act of utterance, which is part of utterance meaning, is derived from a set of rules represented in this motor intention. However from an evolutionary perspective, if the contents of motor intention came to be associated with utterance meaning, one would expect there to be an adaptive reason for it, for, as already mentioned in Section 2, the records of internal and external entities and activities that the brain produces are discarded unless they serve an adaptive function. So, if the representation of the activities generated by the motor did not (directly or indirectly) contribute to an adaptive function, it would not have been retained, let alone made its way into a linguistic concept. It is therefore this adaptive function that we now need to identify, while, of course, keeping track of the fact that the constraint under consideration maps onto the simulation, rather than the execution, phase of the utterance production process.

The structure of utterance meaning: how did it come about?

123

4 From motor representation to utterance meaning What is therefore the kind of adaptive function that the contents under scrutiny serve at the linguistic level that might lead to their integration to utterance meaning? Building on a widely accepted assumption that language production and comprehension are two functions of the same bi-directional device (cf. Tomasello et al. 2005: 683; Jeannerod 2006: 139), I suggest the following: the contents we are concerned with have a key role in a process whereby, during interpretation, utterance production is replicated in the interlocutor’s mind in an attempt to specify the speaker’s motor intention. So, while production is a matter of generating an utterance act that meets specifications contained in the (speaker’s) motor intention, interpretation consists in attempting to recover the specifications which have constrained the (completed) utterance act under consideration. Production and interpretation can be assumed to pertain to the same bidirectional device in the sense that they rely on the same system, comprising a representation phase followed by an execution phase, and are supported by the same pathways. Where they diverge is in the way in which they utilise the common neural basis. Action generation would make use of motor representation and motor execution, in their logical sequence and normal function, as dictated by the action production process.

Figure 6: Producing an act of utterance.17

Action imitation, on the other hand, would proceed differently: on the basis of a given utterance act, hypotheses would be formed about the content of the original (i.e. the speaker’s) motor intention. These hypotheses would then be matched to the completed utterance act (which is already available), at which stage, contextual information would also have been brought to bear.

17 Producing an utterance act is a matter of generating an utterance act that meets specifications contained in the (speaker’s) motor intention.

124

Thanh Nyan

This view of utterance interpretation as including a simulation of utterance production is notably germane to the model that Jeannerod and Anquetil rely on to account for how one “can understand the actions others perform” (2008: 356). According to this model, an observer achieves such understanding by simulating the action of the agent. Based on empirical evidence, “the simulation network in the observer’s brain” would appear to “overlap with the execution network in the agent’s brain” (Jeannerod and Anquetil 2008: 356; see also Rizzolatti and Craighero 2004).

Figure 7: Interpreting an utterance act18

A further argument in support of the idea that the structure of motor intention is involved in interpretation resides in the fact that some of its contents are accessible to consciousness; those very contents that coincide with the ones found in utterance meaning. As Jeannerod observes, there are clinical observations to the effect that “motor representations (intentions) become known to the subject when they are not followed by execution,” either because they fail or because they are “cancelled at the last moment” (1994: 190). In the case under consideration, the action is not carried out, since, from the standpoint of interpretation, there is no need for it to be. Granted this, the content of the motor representation, which normally “would not reach consciousness because it would be cancelled as soon as the corresponding movement was executed”, would, instead, become “accessible to conscious processing” (Jeannerod 1994: 190) as illustrated in Figure 8 below. Following this, I am now in a position to provide an overall picture of what the relevant phase of interpretation would involve. What triggers interpretation is an utterance act produced by the speaker. As a completed intentional action, this utterance act possesses a physical component and a mental one, corresponding to its motor intention. The contents of this intention would be, in Searle’s terminology, its conditions of satisfaction, in that they specify (among 18 Interpretation (symbolised by the thick left-pointing arrow) is an attempt to recover the specifications which have constrained the (completed) utterance act. These specifications, which correspond to the contents of the speaker’s motor intention, are not accessible to the hearer, hence the question mark. Taking the utterance act as its point of departure, interpretation involves simulating (part of) the production process.

The structure of utterance meaning: how did it come about?

125

Figure 8: Motor intention becoming accessible to consciousness

other things) what kind of utterance act would satisfy the intention. From the standpoint of action production, these contents would be fully specified and participate in the action generation process. From that of interpretation, as replicated in the interlocutor’s mind, they would need to be specified if an interpretation of the utterance act is to be reached. How the interpretative process goes about achieving this involves making use of an un-instantiated structure of motor intention to produce hypotheses about the (specific) contents in order to simulate the production process. Repeated use of this structure over time would cause its representation (corresponding to a non-linguistic concept) to make its way into a linguistic concept (representing an internal entity). The constraints being attributed by TAL to utterance meaning would, therefore, correspond to the conditions of satisfaction of the motor intention, which is the intentional component of the utterance act responsible for getting the process under way. If the interpretive process succeeds in matching the conditions of the motor intention to the completed utterance act, this gives rise to a reconstruction of the original context of utterance.

5 Conclusion By way of conclusion, I would like to sketch out what an alternative account, which gives a more central role to brain maps (Damasio 2010: 17–18), could look like. Damasio takes the view that the records (or brain maps) the brain generates of entities (Map 2) include a representation (Map 1) of the apparatus

126

Thanh Nyan

and sensory-motor activities that gave rise to the identification of their perceptual attributes. In this picture, Map1 is deemed necessary because the information it contains allows a more accurate recall of the entity under consideration.

Figure 9: Composite brain map of an object19

Based on the above model, a composite brain map for an action, as seen from a first person perspective, could be schematised as follows:

Figure 10: Composite brain map of an action.20

Now, the difficulty with this alternative (but possibly complementary) account is that it requires viewing interpretation as a form of simulated recall. But whether this take on interpretation is plausible depends on whether the processes that mediate memory encoding and recall21 could underlie interpretation.

19 The arrow in this figure symbolises a causal relation, the terms of which are the perceptual act and the object’s perceptual attributes. 20 The arrow here symbolises the causal relation between the sensory-motor activities in Map1 and the action in Map2. 21 Interpretation, like recall, is about retrieving information that has been encoded in memory. The key difference is that the initial encoding was carried out by someone else. In terms of the mechanism involved, this entails that interpretation, or rather the relevant part of it would require: (a) a simulation of the initial memory encoding, which occurred when the utterance was produced; and (b) a recall-like process serving to “retrieve” the meaning the speaker would have intended. Memory encoding and recall in Damasio are mediated by “convergencedivergence zones”, which are nodes within a “neural architecture of cortical connections [with] convergent and divergent signaling properties” (2010: 141–144).

The structure of utterance meaning: how did it come about?

127

References Anscombre, Jean-Claude (ed.). 1995. Théorie des topoï. Paris: Editions Kimé. Anscombre, Jean-Claude & Oswald Ducrot. 1983. L’argumentation dans la langue. Brussels: Pierre Mardaga. Damasio, Antonio R. 1989. Concepts in the brain. Mind and Language 4. 24–28. Damasio, Antonio R. 2010. Self comes to mind. New York: Pantheon. Damasio, Antonio R. & Hanna Damasio. 1992. Brain and language. Scientific American, Sept. 1992. 89–95. Damasio, Hanna, Thomas J. Grabowski, Daniel Tranel, Richard D. Hichwa & Antonio Damasio. 1996. A neural basis for lexical retrieval. Nature 380. 499–505. Damasio, Hanna, Daniel Tranel, Thomas Grabowski, Ralph Adolphs & Antonio Damasio. 2004. Neural systems behind word and concept retrieval. Cognition 92. 179–229. Darwin, Charles. 1859. On the origin of species by means of natural selection or the preservation of favored races in the struggle for life. John Murray. London. Deacon, Terrence W. 1997. The symbolic species. New York: Norton. Edelman, Gerald, M. 1989. The remembered present. New York: Basic Books. Futuyma, Douglas, J. 1998. Evolutionary biology, 3rd edn. Sunderland, MA: Sinaer Associates. Jeannerod, Marc. 1994. The representing brain: neural correlates of motor intention and imagery. Behavioral and Brain Sciences 17. 187–145. Jeannerod Marc. 2006. Motor cognition. Oxford: Oxford University Press. Jeannerod, Marc & Thierry Anquetil. 2008. Putting oneself in the perspective of the other: A framework for self-other differentiation. Social neuroscience 3. 356–367. Mayr, Ernst. 1976. Evolution and the diversity of life. Cambridge: Harvard University Press. Mayr, Ernst. 1983. How to carry out the adaptationist program. The American Naturalist 121. 324–334. Nølke, Henning. 1984. Linguistique modulaire: de la forme au sens. Louvain-Paris. Editions Peters. Searle, John. 1980. The intentionality of intention and action. Cognitive Science 4. 47–70. Searle, John. 1983. Intentionality. Cambridge: Cambridge University Press. Tomasello, Michael, Malinda Carpenter, Josep Call, Tanya Behne & Henrike Moll. 2005. Understanding and sharing intentions: The origins of cultural cognition. Behavioral and Brain Sciences 28. 675–735.

Márta Szücs and Anna Babarczy

8 The role of Theory of Mind, grammatical competence and metapragmatic awareness in irony comprehension Abstract: The paper investigates the roles of Theory of Mind, receptive language skills, and metapragmatic awareness in irony comprehension. In Experiment 1, 5–8 year-old typically developing Hungarian-speaking children completed a series of false belief tasks, receptive syntax and vocabulary tests and an irony comprehension test, where they were asked to attribute intentions to an actor uttering an ironic utterance as part of a story. In Experiment 2, the children who had shown no evidence of irony comprehension in Experiment 1 were divided into two groups. One group underwent a metapragmatic awareness instruction programme, where the use and purpose of everyday irony were discussed. The other group did not participate in this programme. An irony comprehension test was then administered to both groups of children. We found that while neither syntax and vocabulary scores nor false belief task performance reliably correlated with irony comprehension, metapragmatic abilities made a significant and pronounced contribution.

1 Introduction Verbal irony can be informally defined as an act of communication where a speaker is saying one thing, but means the exact opposite of what she is saying. In Gricean terms, irony is treated as an implicature triggered by the flouting of the maxim of Quality, that is, “Do not say what you believe to be false” (Grice 1989: 34). The post-Gricean relevance-theoretic framework reinterprets irony and argues that it is an instance of interpretive echoic language use, through which the speaker expresses disagreement with a thought attributed to another person (Sperber and Wilson 1986/1995; Wilson and Sperber 2004; Wilson 2013). According to Relevance Theory, to understand an ironic expression, the hearer has to understand not only the basic proposition expressed by an utterance, Márta Szücs, Department of Hungarian Language and Literature, The University of Szeged Anna Babarczy, Department of Cognitive Science, Budapest University of Technology and Economics DOI 10.1515/9781501505089-008

130

Márta Szücs and Anna Babarczy

but also the fact that it is being interpretively used. In other words, for a typical ironic statement, the listener needs to realise that the speaker is expressing an attitude towards a thought of someone else, rather than a belief of her own. In this respect, from a relevance-theoretic perspective, the ability to attribute intentional states to others, commonly known as Theory of Mind, is believed to be a fundamental component of pragmatic competence, since it is this ability that allows the hearer to realise that the speaker’s communicative intentions are not explicitly expressed by the encoded semantic meaning of her utterance. Perner and Wimmer (1985) describe two levels of belief attribution that play a crucial role in children’s understanding of social interactions and appear at different stages of their development: first-order beliefs involve the representation of another person’s thoughts about the real world and appear around the 4th year of age in typically developing children, while second-order beliefs involve the representation of another person’s beliefs about someone else’s thoughts and appear a couple of years after children have developed first-order Theory of Mind abilities (cf. Miller 2009; Frith and Frith 2010). The relevance-theoretic echoic interpretation of irony therefore predicts that second-order Theory of Mind abilities, in Perner and Wimmer’s sense, are a prerequisite to irony comprehension. This prediction is supported by empirical evidence collected by Happé (1993), who found that autistic subjects who passed first-order false belief tests correctly interpreted metaphors but failed to understand ironic utterances, while those who passed second-order false belief tests tended to comprehend irony as well. The performance of a small sample of typically-developing young children in this study also reflected this pattern: again, only those participants in the control group who passed both first-order and second-order false belief tests understood irony. Given the results obtained, Happé’s conclusion was that performance in Theory of Mind tasks is a very good predictor of irony comprehension. Other studies, however, have failed to support Happé’s interpretation of her results. For instance, Sullivan, Winner and Hopfield (1995) and Sullivan, Winner and Tager-Flusberg (2003) studied typically-developing children and children with William’s syndrome, and found that second-order mental state attribution ability precedes ironic joke comprehension by approximately two years suggesting that while second-order false belief attribution may be a necessary condition of irony comprehension, it is not a sufficient one. Other factors also appear to be at play, one of which may be grammatical ability. The relationship between pragmatic competence and syntactic or lexical development has also been hotly debated. Some experimental evidence suggests that children’s pragmatic development is dissociated from their grammatical competence, since there appears to be a stage of language development when

The role of Theory of Mind

131

they have difficulty interpreting non-logical or non-literal meanings – a difficulty that cannot be attributed to lexical or syntactic deficits (cf. Papafragou and Musolino 2003; Bernicot and Laval 2004; Noveck 2004). Yet, other studies have revealed correlations between receptive language skills and the comprehension of non-literal language, especially in relation to speech acts and metaphors (cf. Leinonen et al’s 2003; Norbury 2005; De Mulder 2011). Similarly, the relationship between linguistic ability in general and irony comprehension is far from straightforward. Filippova and Astington (2008) and Angeleri and Airenti (2014), for instance, found receptive vocabulary to have an effect on irony comprehension in 5–9 and 3–6 year-old typically-developing children. In both cases, however, the children’s performance on the vocabulary task showed a strong positive correlation with their false-belief understanding, and regression analyses revealed little additional effect of the vocabulary skills. On the other hand, in Cutica, Bucciarelli and Bara’s (2006) non-verbal implicature task where a picture representing an actor’s communicative intentions had to be selected, patients with left hemisphere brain damage showed poor performance if the story involved irony, such as clapping when someone was clumsy, even though they did not display impaired performance in Theory of Mind tasks. A meta-analysis by Milligant, Astington and Dack (2007) suggests that there tends to be a correlation between language ability and false-belief understanding, although the strength of the relationship varies greatly, which might be a part of the explanation for the contradictory results concerning false belief task performance and non-literal language comprehension. Even so, the metaanalysis strongly suggests that the direction of the relationship is from language ability to false-belief performance rather than vice versa. Turning to a far less intensively studied explanatory factor, little attention has been placed, in the developmental literature, on the relationship between non-literal language comprehension and metapragmatic awareness. Metapragmatic awareness, which is defined as the ability to reflect on language use in a conscious way (cf. Verschueren 2000; Wilkinson and Milosky 1987) can be seen as a kind of interface between the linguistic and the extralinguistic level of language use (cf. Caffi 1994), which presupposes not only linguistic but also contextual and world knowledge (cf. Collins 2013). In one of the few studies in the area, Bernicot, Laval and Chaminaud (2007) investigated the development of metapragmatic awareness and non-literal language comprehension in 6, 8 and 10 year-old typically-developing children. In this study, the children’s comprehension of non-literal language comprising semantic-inference implicatures, indirect requests, idioms and sarcasm was measured in a phrase-picture matching task, followed by a short interview where the children were asked to explain

132

Márta Szücs and Anna Babarczy

their initial responses. The results obtained showed that different types of nonliteral language exhibit different developmental patterns. Notably, the comprehension of ironic-sarcastic implicatures appeared relatively late, and, in contrast with the other types of non-literal language studied, was accompanied by metapragmatic knowledge when it was present. That is, there might be a connection between the development of irony comprehension and metapragmatic awareness, even though the nature of Bernicot, Laval and Chaminaud’s study does not allow us to directly establish such a cause-and-effect relationship. The effects of metapragmatic awareness, albeit not in relation to non-literal language comprehension, were also studied by Robinson and Robinson (1982), who used metapragmatic training to improve awareness of instruction clarity in 4–5 year-old children. The training significantly improved the children’s ability to both give clear instructions and ask for clarification when given ambiguous instructions. This, in turn, suggests that explicit metapragmatic training can boost specific pragmatic abilities of preschool children. Against this background, the investigation of the relationship between false belief attribution, grammatical skills and metapragmatic awareness, on the one hand, and irony comprehension, on the other, seems like a worthwhile venture. The present study further explores this relationship, with a view to (a) shedding additional light on the relative roles of false belief attribution and receptive grammatical ability in irony comprehension, and (b) finding evidence for a cause-and-effect relationship between metapragmatic awareness and irony comprehension that may be independent of false belief attribution and grammatical abilities.

2 Experiment 1 A non-literal language comprehension test suitable for Hungarian-speaking preschool children was devised. The test contained not only ironic but also metaphorical and indirect speech items, but the present analysis focuses only on irony. Apart from the non-literal language comprehension test, the group of children participating in the study were also tested on a series of false belief tasks, a receptive grammar test and a receptive vocabulary test. A subgroup of these children also participated in a metapragmatic awareness instruction programme for the purposes of Experiment 2, which will be described in Section 3.

The role of Theory of Mind

133

2.1 Participants Seventy-one typically-developing monolingual Hungarian preschool children, 32 girls and 39 boys (age range 4;0 to 7;2, mean age 5;7) from two kindergartens in two university towns in eastern Hungary participated in the experiment. Five of the children could not be reached for the vocabulary test session and were excluded from analyses involving vocabulary scores. (see Table 1 for details). Table 1: Participant details N (Boys, Girls)

Mean Age

Age Range

Irony comprehension test, False belief tasks, TROG

71 (39, 32)

5;8

4;0−7;2

PPVT

66 (35, 31)

5;7

4;0−7;2

2.2 Materials and Procedures 2.2.1 The false belief tasks Four traditional false belief tasks were used to assess the children’s metarepresentational skills: two tasks testing first-order and two tasks testing secondorder false belief attribution. Each task involved the description of a story or a situation and ended in a question regarding the intentions or beliefs of one of the story characters. To pass the test, the child needed to answer this question correctly. The first-order false belief tasks assessed whether a child can attribute a false belief to a story character or to another person. To give a correct answer, the child must be able to look beyond (or inhibit) his/her own knowledge of reality and appreciate the false belief of the other person. In this experiment the Hungarian translations of two well-known tests were used: the Sally-Anne test (Baron-Cohen, Leslie & Frith 1985, 1986), which uses the unexpected transfer paradigm, and the Smarties test (Hogrefe, Wimmer and Perner 1986), which uses the unexpected content/deceptive box paradigm. Second-order false belief tasks are more complex, and require participants to attribute a false belief about another person’s belief to a story character. In order to reduce the effects of test complexity, we used two relatively short and simple tasks that had been specifically designed to reduce processing demands: the Birthday test (Herold 2005) and the Robot test (Coull, Leekam and Bennett 2006), both of which are based on the unexpected transfer paradigm.

134

Márta Szücs and Anna Babarczy

2.2.2 The irony comprehension test The non-literal language comprehension test we administered was similar to Happé’s test (1993), but the general structure was modified, firstly, to make the context stories less explicit in order to avoid echoing effects with young children, and secondly, to increase the number of forced choice answers from two to three in order to reduce the probability of selecting correct answers by chance and allow for an analysis of error types. The test consisted of five short stories, each illustrated by a set of four pictures to help the children follow and remember the events in the story. Each story included an ironic statement uttered by one of the characters in the story. As the test sentences had to be novel non-idiomatic phrases with the individual words of the sentences familiar to preschool children, the frequencies of the ironic phrases and the vocabulary they contained were checked against the Hungarian National Corpus of about 180 million words (http://corpus.nytud. hu/mnsz; cf. Váradi 2002) and those satisfying our criteria of fewer than 15 occurrences as a phrase but more than 1500 occurrences of the individual words were selected. Three of the ironic phrases included in the study were unrepresented in the corpus and two occurred with very low frequencies (9 and 13 occurrences, all with literal meaning). After listening to each story, the children were asked what the characters meant by their ironic utterances and were given three possible answers to choose from. The three options were the correct ironic interpretation, a literal interpretation, and an interpretation of deception. The children were not given any feedback on their answers. An example is given below: (1)

Story: Katie was helping her mother make cookies. After making the dough they put it in the oven, and went out to the garden to play. Unfortunately, the cookies stayed in the oven for too long, and were burnt. Later Katie’s father came home, saw the cookies and said: − What soft cookies! (ironic ending) Question: Why did the father say that? Answers to choose from: a. He thought that the cookies are soft (literal) b. He wanted to trick the mother (deception) c. He expressed in a funny way that the cookies are hard (ironic)

2.2.3 The Test for Reception of Grammar The standardised Test for Reception of Grammar test (TROG, Bishop 1983) is an individually administered, multiple-choice test designed to assess the com-

The role of Theory of Mind

135

prehension of grammatical contrasts marked by inflection, function words and word order, among other grammatical phenomena. The Hungarian version of the test (Lukács, Győri and Rózsa. 2013) consists of 72 items, split up into 18 blocks of 4 items each, in order of increasing difficulty. Each block assesses the child’s comprehension of a specific type of grammatical contrast, involving nouns, verbs, negation, number and person agreement, and relative clauses. In each item, the subject is shown a set of pictures and asked to point to the one that matches the phrase or sentence produced by the tester. A block is passed only if the child responds correctly to all 4 items and the child’s final score is the number of blocks passed. 2.2.4 The Peabody Picture Vocabulary Test The Peabody Picture Vocabulary Test (PPVT, Dunn and Dunn 1959) is one of the most commonly used tests to assess receptive vocabulary. In our study, we used the Hungarian version of the test (Csányi 1974), which consists of 150 vocabulary items of gradually increasing difficulty. To administer the PPVT, the investigator presents a series of plates with four black-and-white pictures on each. For each plate, the investigator utters a word and the child is asked to select the picture that best represents the word’s meaning. The test ends when the child has answered six consecutive items incorrectly. The child’s score is the number of correct answers preceding the failed block. 2.2.5 General Procedure The children were tested individually in a quiet room on the premises of their regular kindergartens. Participation was voluntary and participants received small gifts in way of thanks. The tests were administered in one or two sessions depending on the needs of the child; if two sessions were needed, these were separated by at most one week. The tests were administered in the following order: false-belief tasks, non-literal language comprehension task, the Test for Reception of Grammar and the Peabody Picture Vocabulary Task.

2.3 Results 2.3.1 The role of false belief attribution ability The children were classed into three different groups based on their performance in the first- and second-order false belief tasks.

136

Márta Szücs and Anna Babarczy

Table 2: Number and ages of children in each ToM group ToM group

NoToM

1stToM

2ndToM

Number Age (mean) Age (range)

29 5;2 4;2–6;11

22 5;11 4;0–7;2

20 5;11 4;10–6;11

The children in the No-ToM group failed either one or both of the first-order false-belief tasks. Those who passed both first-order false belief tests, but failed one or both of the second-order tests were placed in the 1st-ToM group. Finally, the children in the 2nd-ToM group passed all four false-belief tasks. The results of the irony comprehension task for the children grouped by performance on the false belief tests are shown in Figure 1.

Figure 1: Percentage of correct answers to irony items in each ToM group

As Figure 1 shows, the percentage of correct irony responses was similarly low in each ToM group (25%, 40% and 37%) and a Kruskal-Wallis test comparing the three groups showed no statistically significant difference between them. One-Sample Wilcoxon Signed Rank tests revealed that while the No-ToM group performed significantly below the 33% chance level (p = .023), the other two groups’ performance did not differ from chance in either direction. An analysis of incorrect answers in the irony task might shed some light on the strategies used by the children. Figure 2 displays the distribution of errors in the irony task.

The role of Theory of Mind

137

Figure 2: Distribution of literal and deception answers in the irony task in each ToM group (as a percentage of the total number of incorrect answers)

As Figure 2 reveals, the most frequent error for children in all groups was by far the interpretation of the ironic statement as an act of deception, gathering 83% of errors in the no-ToM group, 69% in the 1st-ToM group and 74% in the 2nd-ToM group.

2.3.2 Results: The role of language ability As mentioned above, the Peabody Picture Vocabulary Test scores could be obtained for only 66 of the 71 children, as five of the children could not be reached for testing within the allocated one-week time frame. Statistical tests involving Peabody Picture Vocabulary Test scores were therefore only run on these 66 children, while those related to the Test for Reception of Grammar included all 71 participants. The children’s mean scores and minimum and maximum values for the two tests are shown in Table 3. Table 3: Mean scores, ranges and standard deviations for the TROG and PPTV

TROG PPTV

Range

Mean

StDev

0–17 62–130

11.48 87.36

3.83 19.43

138

Márta Szücs and Anna Babarczy

In order to explore the relative roles of Theory of Mind and language ability in irony comprehension, hierarchical linear regression models were built for the 66 children that had completed all tests. In these models, Theory of Mind performance was represented by the number of false belief tests passed (0 to 4), while the Test for Reception of Grammar and Peabody Picture Vocabulary Test scores represented language ability. Besides these three variables (ToM, TROG and PPVT), the age of the children (in months) was also included in the model, as a variable standing for unknown factors of maturity, such as unmeasured language ability, working memory, cognitive flexibility, and so on. As shown in Table 4, however, the contribution of the variables was very low and barely approaching statistical significance. Table 4: Regression analysis predicting irony comprehension from language abilities (TROG, PPVT), ToM and age (p value is significant if p < 0.05) Model

R2

1 ToM TROG PPVT Age

0.123

Standardised ß

0.179 –0.247 0.191 0.252

t

p

1.337 –1.808 –1.345 1.691

0.186 0.076 0.184 0.096

2.4 Discussion Experiment 1 considered the development of irony comprehension in four to seven-year-old typically-developing Hungarian-speaking children. The two questions at the heart of this investigation were: (a) What role do the children’s false belief attribution abilities play in the comprehension of this pragmatic phenomenon, and (b) what is the relationship between the general language abilities of the children and their comprehension of irony? As regards the first question, there was no difference in irony comprehension between children at different stages of Theory of Mind ability as measured through the false belief attribution tasks. We should note that the irony scores in the present study were much lower than the indirect speech and metaphor comprehension scores (Babarczy and Szücs, in preparation) or the irony scores in Happé’s (1993) test, and, notably, around chance level. These results therefore have nothing to say about whether second-order false belief attribution ability is a necessary condition for irony comprehension, but they do seem to suggest that it is certainly not a sufficient condition.

The role of Theory of Mind

139

One possible explanation for this result could be that the children had to understand not only the ironic meaning but also the ironic attitude as a whole in the task. Preschool children may realise that literal/semantic meaning is not appropriate in a given context, but they may still not be able to recognise the speaker’s ironic attitude behind the utterance. This much is also suggested by the error pattern: children are more prepared to believe that the speaker is lying than that they themselves have misunderstood the literal meaning of the utterance or perhaps the story context of the utterance. Notably, this error pattern is consistent with the results of previous studies. Sullivan et al. (1995, 2003), for example, reported that children who were unable to recognise the ironic attitude conveyed by ironic utterances all made the same kind of error, namely, they considered all the ironic jokes lies. Experiment 1 also looked at the relationship between general receptive language development and irony comprehension. Here, we found that the variance in irony comprehension could not be explained either by the reception of grammar or by receptive vocabulary. This could of course be due to the low irony scores observed throughout our sample. As some of the children reached very high scores on the language tests (17/18 in TROG and 130/150 in PPVT), however, it is again quite clear that good receptive grammar and vocabulary skills do not seem to be sufficient for irony comprehension.

3 Experiment 2 Building on the results of Experiment 1, Experiment 2 was designed to investigate the role of metapragmatic awareness in irony comprehension over and above ToM and grammatical ability.

3.1 Participants Thirty-nine children, whose irony comprehension scores on the non-literal test in Experiment 1 were around chance level (with at most 2 out of 5 items answered correctly), were selected as participants for the metapragmatic awareness (MPA) instruction experiment. These children were randomly allocated into two groups: 20 children in the metapragmatic awareness instruction group and 19 children in the control group, as follows:

140

Márta Szücs and Anna Babarczy

Table 5: Number and age (mean and range) of the children in the MPA group and in the control group

MPA instruction group Control group Total

N (boys, girls)

Mean score on irony test (%)

Age (mean)

Age (range)

20 (12, 8) 19 (9, 10) 39 (21, 18)

18 15 16

5;9 5;4 5;7

4;5−6;11 4;2−7;2 4;2−7;2

3.2 Materials and procedures The children in the metapragmatic awareness instruction group participated in three metapragmatic awareness development sessions, where they were explicitly taught about ironic statements and intentions. Three new irony stories similar in content and structure to the irony test administered in Experiment 1 were constructed for the programme. Each story represented a natural dialogue between a child and an adult and included an ironic utterance. The stories were illustrated by pictures to aid comprehension and memory. As in Experiment 1, the children’s task was to select the best interpretation of the ironic utterance from three options: an ironic interpretation, a literal interpretation, and an interpretation of deception. Again, the children participated in the sessions individually. All sessions were led by one of the investigators, who gradually reduced the amount of help given to the child: during the first session, the three stories were fully explained and interpreted together with the children, in the second session no explanation was given but leading questions were asked to help the children, and in the third one, each child was asked to provide his/her answer without help, but feedback was given after each answer. The details of the three sessions were as follows:

3.2.1 Instruction session 1 (2)

Sample story in the metapragmatic awareness instruction programme: Toby and his father went on holiday to Lake Balaton. Before leaving town they had to pick up Toby’s cousin called Pete, who had been waiting for them at the gate of his house. His father went to help him put his bag into the car. When he noticed Pete’s bag, he said:

The role of Theory of Mind

141

Father: − What a big bag! (Ironic statement with ironic intonation) Toby: − Daddy, why did you say that? His rucksack is so small. Father: − Yes, I know. I was just joking about his small bag. (explanation of the ironic statement, MPA knowledge) During the first session, the interpretation of the ironic statement was assisted with leading questions that allowed children to follow and understand each step of the interpretation of the ironic statement. These questions related to (i) knowledge of the actual state of affairs; (ii) knowledge of the difference between the semantic meaning of the utterance and the speaker’s knowledge of the actual state of affairs; (iii) the recognition of the inappropriateness of the deceptive interpretation; and (iv) the consideration of the ironic use of language: i) Q: How big do you think the rucksack is in reality? A: Small. ii) Q: How big did Toby’s father say the luggage was? A: Big. Q: How big is the rucksack according to his father in reality? A: Small. iii) Q: Why did his father say that the luggage was big? Did his father want to trick his son? A: No, he did not want to trick him. iv) Q: What did Toby’s father want to say when he said that? A: (repeating the father’s response) He wanted to say in a funny way that he thought the luggage was small. Apart from the leading questions, the role of ironic intonation and contextual clues were also mentioned during this session and the children’s answers to the leading questions were further discussed. For instance, in Hungarian, there is a special sentence structure (Ez aztán. . .), similar to the English construction What a. . . , which can signal the use of irony, even though it does not necessarily introduce an ironic statement. 3.2.2 Instruction session 2 The second sessions of the metapragmatic awareness instruction programme were consistently held 4–5 days after the first sessions. The children heard the same three stories as during their first session, only this time the investigator played a more passive role. She activated the relevant metapragmatic knowledge with the leading questions but the children were expected to answer the ques-

142

Márta Szücs and Anna Babarczy

tions on their own and were only given help if they had difficulty providing the correct answers. 3.2.3 Instruction session 3 The third session was again held a few days after the second. During this session the children’s interpretation of the ironic utterances was tested using the metapragmatic awareness instruction material consisting of the three stories but no leading questions were asked. After listening to each story, the children were only asked to answer the interpretation question, i.e., to choose the correct answer to the question about the intentions of the speaker of the ironic utterance. The children’s correct irony answers were counted, while their incorrect answers were also recorded. The children were then given feedback. 3.2.4 Second irony comprehension test After the instruction programme was concluded, both the metapragmatic awareness instruction group and the control group were tested again, using the irony comprehension test used in Experiment 1, which contained five short stories ending in an ironic utterance and three options to choose from as an answer to the question about the interpretation of the ironic utterance (ironic, literal, deception). Again like in Experiment 1, the number of correct answers was noted and the children’s incorrect answers were also recorded.

3.3 Results Table 5 below shows the mean percentages of correct answers in the two different irony tests: the results of the first test administered in Experiment 1, the scores achieved by the metapragmatic awareness instruction group at the third session of their programme, and the results of the test administered after the instruction programme. Table 6: Mean percentages of correct answers in irony comprehension before, at the end of, and after MPA instruction MPA group Mean % (StDev)

Control group Mean % (StDev)

Irony comprehension test 1 (Experiment 1)

18 (17)

15 (16)

Irony comprehension in MPA 3rd session

88 (19)

−

Irony comprehension test 2 (after MPA development procedure)

71 (35)

18 (20)

The role of Theory of Mind

143

The results of the Metapragmatic Awareness 3rd session test (88%) clearly indicate that the children in this group had no difficulty choosing the correct ironic interpretation in a well-known context. This could, however, be the result of simple associative learning and does not provide evidence that the children have indeed developed metapragmatic awareness. The children would show evidence of metapragmatic knowledge, only if they could transfer what they had learnt to new situations. This is exactly what happened in the irony comprehension test following the metapragmatic awareness instruction sessions. The results of this test reveal an enormous improvement in irony interpretation in this group when compared to the control group: the metapragmatic awareness instruction group answered correctly to 71% of the questions on average, which is an improvement of 53 percentage points relative to the pre-instruction test, while the corresponding figure for the control group is 18%, an improvement of only 3 percentage points relative to the first irony test. The difference between the two groups seems to be reliable as shown by statistical analysis: independent samples t-tests showed no difference between the groups prior to the training sessions (t(37) = –.614, n.s.) and a highly significant difference in favour of the metapragmatic awareness instruction group after the training sessions (t(37) = –5.832; p < 0.001). The next question we addressed was whether metapragmatic instruction had an effect independently of the children’s false belief attribution ability. The 20 children who participated in the instruction programme were divided into three groups based on the their Theory of Mind performance as in Experiment 1. A One-Way ANOVA with these three groups revealed no difference between them (F(2, 19) = 0.24, n.s.) in terms of their performance in the second irony test. As shown in Table 7, the means of the three groups are almost identical, which suggests that metapragmatic training improves performance regardless of Theory of Mind ability. Table 7: Mean percentage of correct answers in Irony Test 2 in the MPA instruction group by ToM ability

No-ToM 1st ToM 2nd ToM

N

Mean score on Irony Test 2 (%)

StDev

8 5 7

72 68 71

30 46 36

Similar comparisons were made in relation to receptive grammar and vocabulary skills. After combining their Peabody Picture Vocabulary Test and Test for Reception of Grammar scores, we divided the children into two groups: those

144

Márta Szücs and Anna Babarczy

scoring below average on the combined receptive language tasks and those scoring at or above average. As one of the children in the metapragmatic awareness instruction group had not completed the PPVT task, the analysis was run on the remaining 19 children. Table 8 shows the average scores of the two groups: Table 8: Mean percentage of correct answers in Irony Test 2 in the MPA instruction group by receptive grammar and vocabulary ability

Below average TROG-PPVT Above average TROG-PPVT

N

Mean score on Irony Test 2 (%)

StDev

10 9

66 73

27 43

Although the children with above-average receptive grammar and vocabulary skills performed slightly better on the second irony test than the children with below-average language ability, the difference was again not statistically significant (t(17) = .447, n.s.).

3.4 Discussion Experiment 2 considered the effects of enhancing children’s metapragmatic awareness on irony comprehension performance in four to seven-year-old typicallydeveloping Hungarian-speaking children. The goal of the study was to gain insight into the role of metapragmatic awareness training in the development of children’s irony comprehension. The findings demonstrated that metapragmatic awareness can indeed play an important role in the development of irony comprehension in preschool children and that its benefits do not depend on either false belief attribution or linguistic ability – at least to the extent that these abilities are captured through performance in the tasks undertaken by our participants. All in all, our results clearly show that, if given explicit instruction, preschool children have the ability to integrate their knowledge of linguistic (linguistic and speaker meaning of ironic utterances) with metalinguistic rules (contextual and intonation clues) and use this knowledge to recognise irony.

4 Conclusion Summing up, the findings of Experiment 1 reported here have shown a loose relationship between Theory of Mind ability and irony comprehension in typically developing children, since a developed second-order Theory of Mind

The role of Theory of Mind

145

ability did not appear to be sufficient to ensure better irony comprehension. Grammar reception and receptive vocabulary were also found to be relatively unrelated to irony comprehension in our group. Yet, metapragmatic training involving drawing children’s attention to the fact that a literally false statement may be uttered with an intention other than deception, made a dramatic difference to our group’s irony comprehension performance regardless of their ToM or general language ability. Our Experiment 2 showed the relationship between metapragmatic awareness and irony comprehension to be clearly causal. It could thus be the case that the puzzlingly slow development of irony comprehension relative to lexical and syntactic development is simply a matter of insufficient experience with ironic language use. Given our results, it is further possible that the direction of the relationship between false belief understanding and irony comprehension is not quite as simple as we had assumed; while having a Theory of Mind may aid irony comprehension, another explanation for the previously observed co-appearance of the two is that metapragmatic knowledge of the kind needed for irony comprehension, for instance, also aids the development of a Theory of Mind. We therefore suggest that ironic interpretation is a social-pragmatic phenomenon and – under non-laboratory circumstances – children learn to comprehend ironic statements from natural discourses, in which they meet them in a context where the ironic interpretation is either explained or is the only cognitively plausible interpretation.

References Angeleri, Romina & Gabriella Airenti. 2014. The development of joke and irony understanding: A study with 3- to 6-year-old children. Canadian Journal of Experimental Psychology/Revue canadienne de psychologie expérimentale 68. 133–146. Babarczy, Anna and Márta Szücs. In preparation. The comprehension of hint and metaphor in preschool children: The contribution of metarepresentational abilities and language skills. Baron-Cohen, Simon, Alan M. Leslie & Uta Frith. 1985. Does the autistic child have a “theory of mind”? Cognition 21. 37–46. Baron-Cohen, Simon, Alan M. Leslie & Uta Frith. 1986. Mechanical, behavioral and intentional understanding of picture stories in autistic children. British Journal of Developmental Psychology 4. 113–125. Bernicot, Josie and Virginie Laval. 2004. Speech acts in children: The example of promises. In Ira A. Noveck & Dan Sperber (eds.), Experimental pragmatics, 207–228. London: Palgrave Macmillan. Bernicot, Josie, Virginie Laval & Stephanie Chaminaud. 2007. Nonliteral language forms in children: In what order are they acquired in pragmatics and metapragmatics? Journal of Pragmatics 39. 2115–2132.

146

Márta Szücs and Anna Babarczy

Bishop, Dorothy. 1983. Test for the Reception of Grammar (TROG). Age and Cognitive Performance Research Centre: University of Manchester. Caffi, Claudia. 1994. Metapragmatics. In Ronald E. Asher & James M.Y. Simpson (eds.), Encyclopedia of Language and Linguistics, 2461–2466. Oxford: Pergamon Press. Collins, Anna. 2013. Metapragmatic awareness in children with typical language development, pragmatic language impairment and specific language impairment. Manchester: University of Manchester PhD thesis. Coull, Greig J., Susan R. Leekam & Mark Bennett. 2006. Simplifying second-order belief attribution: What facilitates children’s performance on measures of conceptual understanding? Social Development 15. 548–563. Csányi, F. Ivonne. 1974. Peabody Szókincs-Teszt [Peabody Vocabulary Test]. Budapest: Bárczi Gusztáv Gyógypedagógiai Fõiskola. Cutica, Ilaria, Monica Bucciarelli & Bruno G. Bara. 2006. Neuropragmatics: Extralinguistic pragmatic ability is better preserved in left-hemisphere-damaged patients than in righthemisphere-damaged patients. Brain and Language 98. 12–25. De Mulder, Hanna L.N.M. 2011. Putting the pieces together: The development of theory of mind and (mental) language. Utrecht: Utrecht University PhD dissertation. Dunn, Lloyd M. & Leota M. Dunn. 1959. Peabody picture vocabulary test. Circle Pines, MN: American Guidance Service. Filippova, Eva & Janet W. Astington. 2008. Further development in social reasoning revealed in discourse irony understanding. Child Development 79. 126–138. Frith, Uta & Chris Frith. 2010. The social brain: Allowing humans to boldly go where no other species has been. Philosophical Transactions of the Royal Society B: Biological Sciences 365.1537. 165–176. Grice, H. Paul. 1989. Studies in the way of words. Cambridge, MA: Harvard University Press. Happé, Francesca G.E. 1993. Communicative competence and theory of mind in autism: A test of relevance theory. Cognition 48. 101–119. Herold, Róbert. 2005. Mentalizációs deficit szkizofréniában [Mentalising deficit in schizophrenia] Pécs: University of Pécs PhD dissertation. Hogrefe, Juergen G., Heinz Wimmer & Josef Perner. 1986. Ignorance versus false belief: A developmental lag in attribution of epistemic states. Child Development 57. 567–582. Leinonen, Eeva, Nuala Ryder, Margaret Ellis & Claire Hammon. 2003. Use of context in pragmatic comprehension by specifically language-impaired and control children. Linguistics 41. 407–423. Lukács Á., Győri M. and Rózsa S. 2013. TROG-H: új sztenderdizált módszer a nyelvtani megértés fejlődésének vizsgálatára [TROG-H: A new standardised method of testing the development of receptive grammar]. Gyógypedagógiai Szemle 2013/1. Miller, Scott A. 2009. Children’s understanding of second-order mental states. Psychological Bulletin 135. 749–73. Milligant, Karen, Janet Wilde Astington & Lisa Ain Dack. 2007. Language and theory of mind: Meta‑analysis of the relation between language ability and false-belief understanding. Child Development 78. 622–646. Norbury, Courtenay Fraser. 2005. The relationship between theory of mind and metaphor: Evidence from children with language impairment and autistic spectrum disorder. British Journal of Developmental Psychology 23. 383–399. Noveck, Ira A. 2004. Pragmatic inferences related to logical terms”. In Ira Noveck & Dan Sperber (eds.), Experimental pragmatics, 301–322. London: Palgrave Macmillan.

The role of Theory of Mind

147

Papafragou, Anna & Julien Musolino. 2003. Scalar implicatures: Experiments at the semanticspragmatics interface. Cognition 86. 253–282. Perner, Josef & Heinz Wimmer. 1985. ‘John thinks that Mary thinks that. . .’ Attribution of second-order beliefs by 5–10 years old children. Journal of Experimental Child Psychology 39. 437–471. Robinson, Elizabeth J. & Peter W. Robinson. 1981. Ways of reacting to communication failure in relation to the development of the child’s understanding about verbal communication. European Journal of Social Psychology 11. 189–208. Robinson, Elizabeth J. & Peter W. Robinson. 1982. The advancement of children’s verbal referential communication skills: The role of metacognitive guidance. International Journal of Behavioural Development 5. 329–355. Sperber, Dan & Deirdre Wilson. 1995. Relevance: Communication and cognition. 2nd edn. Oxford: Blackwell. Sullivan, Kate, Ellen Winner & Natalie Hopfield. 1995. How children tell a lie from a joke: The role of second-order mental state attributions. British Journal of Developmental Psychology 13. 191–204. Sullivan, Kate, Ellen Winner & Helen Tager-Flusberg. 2003. Can adolescents with Williams Syndrome tell the difference between lies and jokes? Developmental Neuropsychology 23. 85–103. Váradi, Tamás. 2002. The Hungarian National Corpus. In Proceedings of the 3rd LREC Conference, 385–389. Las Palmas: University of Las Palmas de Gran Canaria. http://www. lrec-conf.org/proceedings/lrec2002/pdf/217.pdf (accessed 20 May 2016) Verschueren, Jef. 2000. Notes on the role of metapragmatic awareness in language use. Pragmatics 10. 439–456. Wilkinson, Louise Cherry & Linda M. Milosky. 1987. School-aged children’s metapragmatic knowledge of requests and responses in the classroom. Topics in language disorders 7. 61–70. Wilson, Deirdre. 2013. Irony comprehension: A developmental perspective. Journal of Pragmatics 59. 40–56. Wilson, Deirdre & Dan Sperber. 2004. Relevance theory. In Laurence R. Horn & Gregory Ward (eds.), The handbook of pragmatics, 607–632. Oxford: Blackwell.

III Pragmatics and linguistic analysis

Enikő Németh T.

9 Theoretical and methodological issues in the research into implicit arguments in Hungarian Abstract: Based on my research in the area, the present paper intends to argue for a complex approach which can provide an adequate theoretical background for the explanation of the occurrence and identification of verbal implicit arguments in Hungarian. This complex approach analyses implicit arguments in utterances of language use, taking into consideration the interaction between grammatical and contextual factors. So, it makes available those implicit arguments which are usually excluded from the description in sentence-oriented accounts. Apart from this, the proposed complex approach integrates data from various data sources as well. All in all, it is argued that the assumption of an interaction between grammar and pragmatics, the investigation of implicit arguments in utterances rather than just sentences, and the use of data from various sources can result in a more complete and plausible account of implicit arguments in Hungarian.

1 Introduction Critical evaluation of the literature on implicit arguments in different languages has attested that purely syntactic, pragmatic and lexical-semantic explanations are inadequate to account for the occurrence of verbs with implicit arguments and the interpretation possibilities of lexically unrealised arguments (cf. Cote 1996; Goldberg 2005a, 2005b; Németh T. 2010). The criticism put forth in these

Acknowledgements: The research reported on in the present paper was supported by the Hungarian Scientific Research Fund OTKA NK 100804 Comprehensive grammar resources: Hungarian, as well as by the MTA-DE Research Group for Theoretical Linguistics. I wish to thank Csilla Rákosi for her remarks and suggestions which helped me clarify some issues. I also want to express my gratitude to Anna Fenyvesi and Stavros Assimakopoulos for correcting my English. Enikő Németh T., University of Szeged, Hungary DOI 10.1515/9781501505089-009

152

Enikő Németh T.

explanations has led researchers to develop complex approaches, which consider both grammatical and contextual information (cf. García Velasco and Portero Muñoz 2002; Goldberg 2005a, 2005b; Németh T. 2010, 2012, 2014a, 2014b). Arguably, the examination of implicit arguments in such complex approaches makes it possible to explain various phenomena which cannot be accounted for in purely syntactic, pragmatic and lexical-semantic approaches, as complex approaches can provide an adequate theoretical background for the description of the interaction between grammatical and contextual factors, by combining contextual factors and grammatical requirements. Furthermore, they also make it possible to extend the scope of the relevant investigation beyond the boundaries of the sentence, considering implicit arguments in utterances of language use, which are typically excluded from purely syntactic and lexicalsemantic explanations. So, a significant amount of previously non-supporting evidence can become supporting data for theoretical purposes. In the present paper, I aim to discuss three issues related to research into implicit arguments in Hungarian. First, if contextual factors influence the occurrence and identification of implicit arguments, we have to accept that there is an intensive interaction between grammar and pragmatics in the course of the use and interpretation of implicit arguments. Second, assuming that such an interaction exists, implicit arguments cannot be accounted for in sentential environments. So, various occurrences of implicit arguments in Hungarian, which are judged ungrammatical in sentence-oriented accounts, can be easily interpreted in utterance contexts. Moreover, even if the utterance context cannot provide the necessary information for the identification of implicit arguments, it can be extended by incorporating information from previous discourse, physical, and/or encyclopedic context.1 And, third, if implicit arguments are investigated in the context of their utterance, it seems appropriate to research them on the basis of data collected from various sources, such as intuition, spoken and written corpora, as well as thought experiments. Against this background, the organisation of my paper is as follows. In Section 2, I present the terminological diversity that is characteristic of the literature on implicit arguments and critically discuss some aspects of purely syntactic, pragmatic and lexical-semantic explanations of verb occurrences with implicit arguments with a view to showing that complex approaches are necessary. In Section 3, I discuss the three aforementioned theoretical and methodological reasons for opting for a complex approach and exemplify them 1 Németh T. (2014b) also discusses these two questions in detail. The present paper develops the research further and situates it in a more comprehensive perspective, adding to it a third important methodological question regarding the data on which it can be built.

Theoretical and methodological issues in the research into implicit arguments

153

on the basis of my research into implicit arguments in Hungarian. Finally, in Section 4, I summarise my conclusions.

2 Purely syntactic, pragmatic and lexical-semantic explanations 2.1 Terminological diversity Since the well-known seminal papers by Fodor and Fodor (1980), Dowty (1981), and Fillmore (1986), a wide range of contributions have focused on the behaviour of verbs with lexically unrealised, syntactically missing complements/arguments in English. The term verbal complement is mostly applied in syntactically-oriented approaches to indicate syntactically required verbal constituents such as subject, direct object, and indirect object arguments (see, for example, Haegeman and Guéron 1999). In this picture, the syntactic valence of a verb determines what syntactic complements are needed in order for it to form a grammatical sentence (Cornish 2005: 21−23). For instance, in a grammatical sentence, the English verb lock requires two syntactic constituents, a subject and a direct object complement, as in (1), where Mary is the subject and the noun phrase the door is the direct object. (1)

Mary locked the door.

That said, the term complement cannot only be used in the sense of a syntactic constituent but also in that of a semantic one (see, for example, Fillmore 1986; Cornish 2005). In the semantic sense, a verbal complement is a lexical-semantic entity which is determined by the semantic valence of the verb as a predicate (Cornish 2005: 21−23). The semantic valence of a verb, now, refers to the number and role of the participants of the event or situation described by it. Naturally, it is not necessary to have a one-to-one correspondence between the syntactic and semantic valences of a verb. For example, from the point of view of semantic valence, the verb eat is a two-place predicate which can be syntactically instantiated as a verb with a subject and a direct object. (2)

John is eating an apple.

In the occurrence of the verb eat in (2), there is a one-to-one correspondence between semantic and syntactic valences, since both participants of the event

154

Enikő Németh T.

that eat denotes are syntactically realised and lexically instantiated. However, eat can also be instantiated syntactically with a subject and no direct object, as in (3): (3)

John is eating.

Here, the semantic and the syntactic valences of the verb eat diverge. The syntactic valence is reduced to one, while the semantic one remains at two, as the meaning of the verb eat always involves that participant of the eating event which is the object of eating, even when this is not syntactically realised. Similar to the twofold interpretation of the term verbal complement, the term verbal argument – very often used as a synonym to verbal complement – is also ambiguously used in linguistics. The verb meanings of a language are represented in the lexicon of that particular language, and this lexical-semantic representation of verbs also contains information regarding their argument structure. Again, this verbal argument structure has two sides, according to some grammatical traditions. On the semantic side, it involves the main participants in the events (states or processes) designated by the verb predicate, and on the syntactic side, it represents the syntactic dependents of the argumenttaking verbal head (Bresnan 1995: 1). In other words, the argument structure of a verb encodes lexical-semantic information about the number and roles of arguments, their syntactic types, and their hierarchical organisation necessary for the mapping to syntactic structure (cf. Rappoport Hovav and Levin 1995; Bresnan 1995: 1, 2001: 30). In this respect, verbal argument structure can be considered an interface between the semantics and syntax of verbs. In the present paper, I will be using the term (verbal) argument to refer to the lexically realisable elements of the lexical-semantic representation of a verb that make the verb meaning complete, in consonance with general custom in linguistics over the last two decades (cf. Gillon 2012: 314). As far as the syntactic appearance of a verbal argument is concerned, I will be using the adjective syntactic before the term argument, while when the semantic role of a verbal argument is referred to, the adjective semantic will precede it. Whenever the term argument is used without the adjectives syntactic or semantic, it will mean that there is a one-to-one correspondence between the semantic and syntactic arguments and it is, therefore, not important to refer separately to one of the two categories. When it comes to the occurrence of verbal arguments in English, even though there are cases where a syntactic argument is optional, and can thus be omitted with the resulting sentence remaining acceptable (Gillon 2012: 314), there are also cases where an argument requires syntactic realisation. For

Theoretical and methodological issues in the research into implicit arguments

155

instance, in (4) the occurrence of the verb lock with a lexically unrealised and syntactically missing direct object argument is unacceptable, since the verb lock requires that its direct object argument be also syntactically present and lexically expressed. In contrast, in (5), the direct object argument of the verb eat can be left lexically unrealised without affecting the sentence’s acceptability, as we have, after all, already seen in (3) above.2 (4) *Are you locking [the door]? (5)

Mary is eating [sandwiches].

Since the direct object in (4) cannot be omitted, it follows that the verb lock does not have an optional direct object slot associated with it, which the verb eat clearly has. Now, if such an optional argument of a verb is present in a sentence, it is said to be explicit, whereas if it is omitted, it is dubbed implicit (Gillon 2012: 314). In the relevant literature, one can find several terms for indicating implicit arguments, such as lexically unrealised argument, missing argument, omitted argument, zero argument, null argument, zero complement, null complement, etc. Although these terms are used in various approaches, they can be considered synonymous and are employed as terminological variants of the category of implicit arguments, since they roughly refer to the same phenomenon: a verbal argument which is not explicitly present in the syntactic realisation of the verb at hand.

2.2 Purely syntactic approaches to implicit arguments The implicit arguments that are discussed in the literature can be of various subtypes, like, for example, indefinite implicit object arguments, reflexive, reciprocal, and contextual (definite) implicit complements, as examples (6)−(9), taken from Gillon (2012: 341), respectively illustrate (see also Cote 1996; Németh T. and Bibok 2010) (6) I am cooking [Ø]. (7)

Peter shaved [himself].

(8) Mary and Peter divorced [from each other]. (9) Bill stayed in the parki all day. He left [Øi ] at sunset. 2 In both examples, the lexically unrealised arguments are provided in square brackets.

156

Enikő Németh T.

Purely syntactic accounts analyse only those types of implicit arguments which can be explained on the basis of the syntactic structure of sentences. They do not only underestimate the role of other factors such as constructions, lexicalsemantic and discourse constraints, but some of them even completely ignore them (for a relevant discussion, see Goldberg 2005a, 2005b). If one considers, for example, the indefinite implicit argument in (6) and the definite, contextual implicit argument in the second sentence in (9), neither of them can be explained by a purely syntactic account, since, in both cases, their occurrence is not a purely syntactic phenomenon. For one, in English, not all verbs can be used with indefinite and/or definite contextual implicit arguments (Fillmore 1986: 99). So, while the verbs cook in (6) and leave in (9) can be used with implicit arguments, the verbs await in (10) and vacate in the second sentence in (11) cannot. (10) *I am awaiting. (11)

(Fillmore 1986: 99)

The protesters stayed in the park all day. *They vacated [the park] at sunset. (Gillon 2012: 341)

Then, it is not easy for purely syntactic approaches to explain the behaviour of verbs, such as cook, eat, drink, iron, read, write, etc., that can be used both with indefinite implicit objects and with overt objects. When these are used without explicit objects, their object arguments are not projected into the syntax, in the sense that the objectless uses of these verbs are considered intransitive from the syntactic point of view. That is, purely syntactic explanations can only cope with the objectless uses of these verbs by assuming that they are intransitive verbs. However, since they can be used with explicit objects as well, as if they are transitive verbs, an obvious problem arises: one and the same verb can sometimes be seen to behave as intransitive, and as transitive in others. Clearly, this behaviour cannot be explained purely syntactically. And even if we also take into account semantic/pragmatic factors, the act of cooking, eating, drinking, ironing, reading, writing, etc. necessarily involve the act of cooking something, drinking something, ironing something, reading something, writing something, etc. It is therefore reasonable to assume that the verbs cook, eat, drink, iron, read, write, etc. always have two arguments with the second, direct object arguments having the option of being left lexically unrealised with indefinite interpretations based on the information concerning their type involved in the verbs’ lexical-semantic representations (cf. Cote, 1996; Groefsema 1995; Németh T. 2001).

Theoretical and methodological issues in the research into implicit arguments

157

What is more, purely syntactic approaches to implicit arguments rely on a latent background assumption3 that the occurrence of lexically unrealised arguments is a sentence-oriented phenomenon. Consequently, they are not sensitive to contextual analyses, neither do they consider implicit arguments in utterances. Consider the following Hungarian examples. (12)

a.

Mari-Ø vasal-Ø Mari-NOM iron-PRS . INDEF.3SG 4 ‘Mari is ironing [Ø].’

[Ø].

b.

Áron-Ø tologat-Ø [Ø]. Áron-NOM push-PRS . INDEF.3SG ‘Áron is pushing [Ø] back and forth.’

According to the sentence-oriented Hungarian grammatical tradition (see, for example, Komlósy 1992, 1994), while Mari vasal [Ø] (‘Mari is ironing [Ø]’) in (12a) is a grammatical sentence, Áron tologat [Ø] (‘Áron is pushing [Ø] back and forth’) in (12b) is unacceptable as a Hungarian sentence. However, this judgment of unacceptability is valid only if one considers Áron tologat [Ø] (‘Áron is pushing’) out of context, since it may still be acceptable as a Hungarian utterance, given an appropriate context, as in (13): (13)

(A mother is walking with her children. The baby is sitting in the stroller, and the elder brother, whose name is Áron, is walking next to it. Suddenly the mother notices the nurse and she wants to talk to her, but the baby begins to cry.) Ne no

sír-j! cry-IMP. INDEF.2SG

Áron-Ø Áron-NOM

tologat-Ø push-PRS . INDEF.3SG

[téged]. [you]

‘Don’t cry! Áron is going to push [you] back and forth.’ (Németh T. 2001: 117) In (13), the direct object argument of the verb tologat (‘push back and forth’) is lexically unrealised, but can still be identified if one considers information

3 For a detailed discussion of the role of latent background assumptions in the course of linguistic argumentation, see Kertész and Rákosi (2012: 85−128). 4 The abbreviation used in the glosses throughout this paper are the following: PRS − present tense, P ST − past tense, 2 − second person, 3 − third person, SG − singular, IMP − imperative, INDEF − indefinite (conjugation), DEF − definite (conjugation), NOM − nominative, ACC − accusative, and ALL − allative, SFX − suffix, POSS − possessive suffix.

158

Enikő Németh T.

from the observable physical context. But, since purely syntactic approaches do not take into account such contextual information, they are bound to evaluate (12b) as unacceptable. (For a more thorough criticism of purely syntactic approaches, see Cote 1996; Goldberg 2005a, 2005b; Németh T. 2010.)

2.3 Purely pragmatic approaches to implicit arguments Turning to purely pragmatic accounts, they also attempt to describe all types of implicit arguments identified in the relevant literature, but do not distinguish between them, and do not rely on any grammatical information either when drawing their conclusions. In this respect, purely pragmatic approaches lead to analyses, according to which, every argument can be omitted if it is inferable. All in all, there are two main types of pragmatic accounts of implicit arguments, which differ at the level of meaning to which they assign implicit arguments and their inferential procedures. According to the first type, the missing content can be inferred as a conversational implicature through Gricean maxims (cf. Rice 1988). However, as Elbourne (2008) points out, this missing content is most likely part of the literal content of utterances rather than a Gricean conversational implicature. Indeed, this seems to be the case, if one considers a range of cases, such as the direct object argument of the verb eat in (14), or other kinds of unarticulated constituents – ‘somewhere’ in (15) (cf. Recanati 2007) or ‘at the dinner party’ in the reply in (16) (cf. Neale 1990: 94−95). (14) I haven’t eaten. (15)

It’s raining.

(16) A: How did your dinner party go last night? B: Everyone was sick. The other type of pragmatic approach, most notably favoured by relevance theorists, suggests that free pragmatic enrichment is responsible for the interpretation of implicit contents in these utterances (Sperber and Wilson 1995: 189; Carston 2002). According to Relevance Theory, the decoded meanings of the utterances in (14)−(16) are linguistically underspecified and can only be developed into full-fledged conceptual representations of literal utterance meanings by means of free enrichment, which contains general-purpose inference rules (Sperber and Wilson 1995: 176). Now, although the relevance-theoretic explanation considers the missing contents part of the literal meaning of the relevant

Theoretical and methodological issues in the research into implicit arguments

159

utterances rather than conversational implicatures, their recovery is still entirely based on pragmatic inference and does not involve any decoding grammatical procedures. Thus, the missing bits of content in (14)−(16), are purely pragmatically inferred both in the Gricean and the relevance-theoretic proposals, with their main difference being that the former considers them conversational implicatures while the latter considers them part of the utterances’ explicitly expressed meaning. The description of implicit arguments in purely pragmatic terms cannot be supported for at least three reasons. The first one, which I have already mentioned at the beginning of this section, is that purely pragmatic accounts, of either the Gricean or the relevance-theoretic kind, cannot distinguish between different types of implicit arguments, as those identified for example in (6)−(9) of Section 2.2. That is because they focus mainly on the inferential mechanisms that underlie the identification of implicit arguments, and these mechanisms by themselves are not adequate to differentiate between the various types of implicit arguments in the first place. Then, if it were possible to clearly account for the occurrences of zero objects purely pragmatically, there would not be any evidence for the presence of null objects in the syntax, and the occurrences of such zero objects would never be lexically determined, since all the information concerning them would just be pragmatically inferred (cf. Németh T. 2010). So, if a reading is not available for a particular utterance, purely pragmatic approaches cannot refer to syntactic constraints to explain its absence, and if it is pragmatically plausible, there are no tools available to pragmatic approaches to account for why the reading in question is missing. Elbourne, for example, argues against free enrichment approaches on the basis of an argument from binding (2008: 293−294). His analysis revolves around sentences containing definite descriptions made with Saxon genitives and contrasts sentences with a bound reading, like (17) and (18), with sentences that lack a bound reading, like (19) (17)

John fed no cat of Mary’s before it was bathed.

(18) John fed no cat of Mary’s before the cat of Mary’s was bathed. (19) John fed no cat of Mary’s before Mary’s cat was bathed. What he notes is that free enrichment approaches cannot prevent the pragmatically plausible but unavailable bound reading (19), since they do not have any syntactic tools to account for it (much like Fillmore’s example in (12) where the use of an implicit direct object argument is also correspondingly blocked).

160

Enikő Németh T.

Finally, as Fillmore (1986: 98) argues, giving a purely pragmatic account of the occurrence of null objects is not possible since there are arguments that cannot be left lexically unrealised even when there is an immediately retrievable antecedent which would enable them to be easily inferred from the context. Consider Fillmore’s example in (20) (see, also, example 4 in Section 2.1). (20) *Did you lock? Although the door that the speaker intends to refer to in (20) would be quite straightforward to infer in a particular context, (20) is still ungrammatical, since the verb lock cannot be used with a lexically unrealised direct object argument. Furthermore, verbs with indefinite and definite implicit objects show certain lexical alternations independently of pragmatic factors, allowing reference to their zero objects with overt forms that are grammatically constrained and not characteristic of pragmatic inferences (cf. Cote 1996).

2.4 Purely lexical-semantic explanations Finally, for purely lexical-semantic explanations, it is the verbs’ lexical-semantic representation that fully determines whether a verb can occur with implicit arguments, and if yes, with what type and how. For example, in Fillmore’s (1986) approach, information about which arguments can be left implicit, as well as which arguments can be used as indefinite null complements (INC) and which as definite null complements (DNC) is marked in the lexical-semantic representation of a verb.5 In this approach, indefinite null complements are obligatorily disjoint in reference with anything saliently present in the context, while definite ones are unambiguously identifiable with a saliently present entity in the context (cf. Németh T. 2001: 119). Similarly, Gillon (2012) may have included elements of syntax and modeltheoretic semantics in his account of verbs’ behaviour with implicit complements, but he still shares Fillmore’s opinion that the occurrence of such complements is lexically determined. As he discusses, syntax and model theoretic semantics require that the possibility of the occurrence of an implicit complement be signaled by the lexical item itself, which also determines what kind of interpretation should be assigned to the implicit complement in question (Gillon 2012: 352). In this respect, verbs admitting implicit arguments can be divided into two groups:6 verbs that can be used with indefinite implicit complements (cf. 5 In his discussion, Fillmore uses the term implicit complement. 6 Gillon (2012: 351) convincingly argues that it is not only verbs which can occur with implicit complements, but also binary relational words in every lexical class.

Theoretical and methodological issues in the research into implicit arguments

161

Fillmore’s INC-verbs), and verbs that can only be used with lexically unrealised arguments if the reference of these arguments can be identified in context (cf. Fillmore’s DNC-verbs). Then, if an implicit complement can be identified with the help of another expression (typically an antecedent) in the context, it is endophoric, while if this identification requires knowledge of the circumstances of the implicit argument’s use, it is called exophoric, and belongs to the category of indexicals. In his approach, Gillon (2012) deals with indefinite implicit complements in detail and proposes a formal, model theoretic account for the behaviour of verbs that take them, while emphasising that the licensing of an occurrence with an indefinite null complement is an idiosyncratic property of the relevant verbs and not the result of what they denote or how they are used. In this respect, he follows Fillmore (1986) and Saeboe (1995) in assuming that the licensing of implicit indefinite complements by some verbs is a purely lexical matter, rather than a matter of pragmatics – as Groefsema (1995) suggests, or a matter of metaphysics – as Iten et al. (2005) propose. Still, purely lexical-semantic approaches cannot provide a sufficient account for the occurrences of verbs with implicit arguments in pro-drop languages. Pro-drop languages, such as Japanese, Chinese, Korean, Spanish, and Hungarian, allow null arguments far more freely than English does. Yet, this freer occurrence does not mean that an argument can be arbitrarily left implicit in every utterance and context, as purely lexical-semantic approaches would predict. Its behaviour is rather moderated by various lexical-semantic, morpho-syntactic, semantic, and contextual factors (cf. (24) in Section 3.1.). But even looking at English more closely, one can still identify problems with an explanation on lexical-semantic grounds. In Fillmore’s (1986) approach, for example, the markedness regarding the occurrence of verbs with indefinite or definite complements in their lexical-semantic representations does not always make the right predictions about the behaviour of verbs in different contexts (cf. Groefsema 1995: 141−142). What is more, the description of the behaviour of DNC-phenomena does not always make the right predictions, since definite implicit arguments do not necessarily have discourse antecedents. For example, analysing the behaviour of win, Fillmore (1986: 100) argues that it can only occur with direct object arguments if there is a particular competition in the context of which the subject is the winner or the loser. Groefsema (1995: 144) argues that there is no need to mention a particular competition in the context in the case of famous sports people, like, for example, Martina Navratilova has won again (Németh T. 2001: 124). Also, it can happen that the same verb behaves differently with regard to INC-DNC-phenomena in its different occurrences. While the use of the verb put with an implicit locative argument in the example

162

Enikő Németh T.

Mary walked to the table with her hand. *She put the book is unacceptable, the occurrence of ‘put’ in the second turn of the following conversation is acceptable: Ann: I don’t know how to finish this letter. B: Why don’t you put ‘yours sincerely’? (Groefsema 1995: 141−142; Németh T. 2001: 124). The same criticisms also hold for Gillon’s (2012) account, since his classification of verbs regarding occurrence with indefinite implicit complements and context-sensitive implicit arguments are the same as Fillmore’s categories of indefinite and definite null complements, respectively. On top of this, however, there is a further important issue to be noted in relation to Gillon’s approach. If we consider the context sensitivity of implicit complement occurrences, it needs to hold that definite null complements should be analysed in context-sensitive utterances of language use rather than in context-independent sentences, which crucially contradicts the very definition of implicit complements that Gillon puts forth (2012: 314). In Gillon’s conception, a word can have a null complement if the omission of the word’s complement from an acceptable simple declarative sentence leaves the resulting sentence acceptable as well. In view of this definition, instances like (12b) above would not be acceptable, even though, as we have seen, they are in everyday use.7 When it comes to indefinite null complements now, they are not context sensitive, which means that a latent background assumption can be recovered again. They can be examined at the level of the sentence. This leads to the odd realisation that definite/contextual implicit arguments and indefinite ones belong to two different levels of analysis, i.e. the level of the utterance and that of the sentence, respectively. This is a conclusion that directly contradicts Gillon’s contention that “the licensing of implicit indefinite complements by some verbs is a purely lexical matter” (2012: 314). This becomes even more obvious, if one looks at verbs, like contribute and give (in the sense of ‘donate’), which can be used with both definite and indefinite implicit complements. Consider Fillmore’s (1986: 97−98) examples in (21): (21)

a.

I contributed to the movement.

b.

I contributed five dollars.

c.

I’ve already contributed.

7 And indeed, they are also deemed unacceptable in the sentence-oriented Hungarian grammatical tradition (see, for example, Komlósy 1992, 1994), even though they would be acceptable in particular contexts.

Theoretical and methodological issues in the research into implicit arguments

163

The direct object argument of the verb contribute in (21a) is an indefinite null complement which is left lexically unrealised. For Fillmore, when this direct object argument is not explicitly expressed, the nature or quantity of donation is a matter of indifference.8 In (21b), the direct object argument of the verb contribute is lexically expressed, but the receiver complement is not. In this case, Fillmore argues that the implicit receiver complement must be a definite null complement with an identity recoverable from context. Finally, in (21c), both the arguments of the verb contribute are omitted. Since this omission involved both an indefinite and a definite implicit complement, it follows, that contribute’s occurrence with implicit complements needs to be examined both at the level of the utterance (vis-a-vis the definite implicit complement) and at that of the sentence (vis-a-vis the indefinite implicit complement).

2.5 The necessity of complex approaches To avoid the insufficiency of purely syntactic, pragmatic or lexical-semantic explanations, recent proposals assume an interaction between lexical-semantic, grammatical and pragmatic factors in both the licensing of the occurrence of verbs with implicit arguments and their interpretation processes (see, for example, Groefsema 1995; Cote 1996; Cummins and Roberge 2005; Goldberg 2005a, 2005b; Iten et al. 2005; Scott 2006, 2013; Bibok 2008; Pethő and Kardos 2009; Németh T. 2008, 2012, 2014; Németh T. and Bibok 2010). And even though these different proposals emphasise the role of each of the three factors to a different extent, they share the view that implicit arguments can be adequately accounted for only by assuming an interaction between them. From this viewpoint, I will define implicit arguments as follows: (22)

Implicit arguments: arguments in lexical-semantic representations of verbs which are lexically unrealised, and whose implicit presence in utterances is attested by lexical-semantic, grammatical (phonological, morphological, syntactic and semantic), and/or pragmatic (contextual) evidence (cf. Németh T. 2012, 2014a, 2014b).

8 In relation to this, Goldberg (2005b) argues that the verbs contribute, donate and give (=‘donate’) are fused with the implicit theme construction, and are thus used with indefinite implicit direct object arguments for pragmatic reasons, since it is culturally preferred not to make public what kind and amount of donation has been given to a cause. There is therefore no need to have a shared advance understanding of the identity or nature of the donation, which enables the lexically unexpressed direct object argument to have an indefinite interpretation.

164

Enikő Németh T.

3 Theoretical and methodological consequences 3.1 Interaction between grammar and pragmatics Defining implicit arguments from a complex perspective, which takes into account both grammatical and contextual factors, has some essential theoretical and methodological consequences. First, if grammatical and contextual factors license the occurrence of implicit arguments in an intensive interaction, it would be plausible to assume that grammar and pragmatics are not independent of each other. Moreover, the relationship between grammar and pragmatics cannot be considered one-sided, as it traditionally has, with pragmatic inference being assumed to follow grammatical parsing. Instead, pragmatic information has to be licensed to interact with grammatical information. In the relevant literature, there are various theories with significantly different conceptions of grammar and pragmatics,9 which makes it necessary to clarify the relevant notions that the present analysis takes on board. To fulfil this theoretical and metatheoretical requirement, I therefore define grammar as the explicit model of the knowledge of language, i.e. grammatical competence, which is a component of the theory of language, and crucially not independent of pragmatics. Pragmatics, on the other hand, can be characterised as the model of the faculty of language use, i.e. pragmatic competence, which is another component of the theory of language, again not independent of the grammar (cf. Németh T. and Bibok 2010: 504−505). In this respect, I consider the interaction between grammar and pragmatics a co-operation of two separate, but not independent components of the theory of language. To illustrate this interaction in the context of implicit arguments in Hungarian, let us take an example and its analysis from Németh T. and Bibok’s (2010: 510) work. elkísér-t-e a feleség-é-tj az (23) A férj-Øi the husband-NOM accompany-PAST- DEF.3SG the wife-POSS -ACC the orvos-hozz, mert [Øi/j/z ] nagyon izgul-t. very.much was.nervous-PAST- INDEF.3SG doctor-ALL because ‘The husband accompanied his wife to the doctor, because [he/she] was very nervous.’

9 For a detailed reflection on different conceptions of grammar and pragmatics, see Ariel (2008, 2010), as well as Németh T. and Bibok (2010).

Theoretical and methodological issues in the research into implicit arguments

165

According to what has been dubbed the subject continuation principle, the subject of the second clause in (23) can be realised as a zero anaphor only if it is co-referential with the subject of the first clause in Hungarian (cf. Pléh 1994: 297−298, 1998: 174; Németh T. and Bibok 2010: 510). In other words, the basic syntactic rule is that when subjects are repeated in this circumstance, they can be dropped. But if a noun other than the subject in the first clause is selected as the antecedent of the null subject in the second clause, then the latter would need to be pronominalised by the demonstrative pronoun az (‘that’) (cf. Pléh and Radics 1978; Pléh and McWhinney 1987; Németh T. 2012: 455). Consequently, according to traditional Hungarian grammar, the zero subject of the second clause in (23) has to be coreferential with a férj (‘the husband.NOM ’). Otherwise, if a speaker wants to refer in the second clause to the noun phrase a feleségét (‘the wife.ACC ’) or az orvoshoz (‘the doctor.ALL ’), then (s)he must use the demonstrative pronoun az (‘that’), as in (24). (24) A férj-Øi elkísér-t-e a feleség-é-tj az the husband-NOM accompany-PAST- DEF.3SG the wife-POSS -ACC the orvos-hozz, mert az nagyon izgul-t. doctor-ALL because shej/hez very.much was.nervous-PAST- INDEF.3SG ‘The husband accompanied his wife to the doctor, because [he/she] was very nervous.’ Even so, pragmatic factors, such as the speech situation and background knowledge, can override this principle, allowing the speaker to use a zero pronoun in the second clause to also refer to the noun phrases a feleségét (‘the wife.ACC ’) and az orvoshoz (‘the doctor.ALL ’).10 So, even though the grammar-induced interpretation is supported in the case that the husband is nervous and his wife is not, if it is known that it is the wife that is nervous instead, the zero anaphor is naturally taken to refer to the direct object phrase a feleségét (‘his wife.ACC ’). Crucially, if both the husband and the wife are nervous, one needs to find out who is 10 I would like to thank the audience of my talk at the 6th International Conference on Intercultural Pragmatics and Communication (30 May−1 June 2014, Malta) for drawing my attention to this interpretation possibility too. In this case, much like in the case of the zero anaphor referring to a feleségét (‘the wife.ACC ’), the interpretation becomes available if it is obvious that it is the doctor who is nervous (as, for example, in a situation where the doctor has diagnosed the wife with a serious disease and asks the husband to accompany his wife, so that he breaks the news to her). However, it should be noted that, unlike the reference of the zero anaphor to a feleségét (‘the wife.ACC ’), there are Hungarian native speakers who evaluate this interpretation possibility as rather strange, while other Hungarian native speakers − including myself − can accept it easily without any difficulties.

166

Enikő Németh T.

(more) nervous in order to resolve the reference of the zero anaphor; otherwise, (24) remains ambiguous. As this analysis shows, pro-drop requirements (belonging to the grammar) and contextual information (belonging to pragmatics) interact intensively, and perhaps in multiple ways, in the licensing and recovery of implicit arguments in Hungarian. If that is correct, it could be argued that the use or interpretation predicted by the grammar alone is nothing more than just the most typical one that emerges in the absence of any specific context.

3.2 Focusing on the utterance From the definition of implicit arguments in (22), as well as the theoretical and methodological issues discussed in the previous section, it follows that implicit arguments cannot be described and explained only in sentential environments, i.e. in sentences as units of grammatical competence strictly determined by the grammar of a particular language (cf. Chomsky 1986: 3; Németh T. 1995: 393). Instead, when researching implicit arguments, one should take into account the utterance environment, and, in most cases, even extend this with information from the wider context, i.e. from previous discourse, the directly observable physical environment, and encyclopaedic knowledge (cf. Sperber and Wilson 1995: 137−142; Németh T. and Bibok 2010: 511). In this setting, utterances are units of language use, which have both grammatical and pragmatic properties and can be defined as follows (cf. Németh T. 1995: 394, 1996: 17−40): (25) u = (ins (pu, c, p, t) In (25), an utterance u is an inscription ins that a person p relates to a pragmatic unit pu at a time t in a context c. The term inscription refers to a realisation through a physical medium. Notably, the definition in (25) is neutral with respect to the production and interpretation of ins. If the person p is the speaker, then she produces an inscription ins making use of a pragmatic unit pu at a time t and in a context c. If this person is the hearer, then she interprets the inscription ins produced by the speaker as a pragmatic unit pu at a time t in a context c.   ls ; pfÞ (26) pu ¼ ð leint As (26) suggests, since utterances are considered units of language use, they are by definition characterised from both a grammatical and a pragmatic point of view.

Theoretical and methodological issues in the research into implicit arguments

167

The grammatical description of utterances can be given in relation to their corresponding well-formed sentences. In this respect, there are two basic classes of utterances; those that have linguistic structure ls, and those that do not. The members of the first class can have either a complete or an elliptical sentence structure. It can also happen that they only consist of a constituent which can be integrated into the syntactic structure of a sentence whose missing parts can be recovered from the context directly, rather than through an elliptical or anaphoric relation. The members of the second class consist only of one lexical entry leint, namely, an interjection or an idiom-like interjection which cannot be integrated in the structure of a well-formed sentence, cf. for example the English interjection Wow! and the Hungarian idiom-like interjection Az ördögbe! (‘Hell’). Implicit arguments can only occur in the first type of utterance which has linguistic structure ls. Turning to the pragmatic description of utterances, this can be made by defining the pragmatic functions pf that these utterances have in language use with respect to context c. As I suggested previously, this context c consists of two parts, namely, the physically observable context cphys and the cognitive context ccog (see, also, Németh T. 1995: 18, 1996: 394). In accordance with the results of my particular analyses of occurrences of implicit arguments as well as findings in the pragmatic literature, it seems best to treat encyclopaedic pieces of information and information from the preceding discourse separately when it comes to the cognitive context, since each of them affords different linguistic markers. Thus, context contains pieces of information from the immediately observable physical environment (cphys), encyclopaedic knowledge (cenc), as well as preceding discourse (cdisc) (after Sperber and Wilson 1995: 137−142; see also, Németh T. & Bibok 2010: 511; Németh T. 2012: 459): (27) c = (cphys, cenc, cdisc) Against this backdrop, pragmatic functions have appropriate values in particular utterances in discourse: the function of literalness ( flit), as well as the interpersonal ( fip), the illocutionary ( fill), and the attitudinal function ( fatt): (28) pf = (flit, fip, fill, fatt) According to the function of literalness ( flit), an utterance can be used literally or non-literally, with non-literal uses resulting in implicitly conveyed pieces of information. The interpersonal function ( fip) concerns the starting, maintaining and finishing of communicative interaction, while the illocutionary function ( fill) indicates what kind of illocutionary acts can be performed by utterances.

168

Enikő Németh T.

Finally, the attitudinal function ( fatt) refers to the speaker’s attitude that can be assigned to utterances. All in all, the pragmatic functions of utterances belong to the general pragmatic knowledge and can be distinguished from the particular contextual information, which either coincides with the predictions of general pragmatic knowledge or differs from it. In this latter case, the particular contextual information can override the information from the general pragmatic knowledge. This differentiation has also an important role in the licensing and interpretation of implicit arguments, especially in the course of use and recovery of zero anaphors (cf. Németh T. and Bibok 2010: 511−512).

3.3 Integration of various data sources From the two theoretical issues discussed above, it follows that verbal argument structure must be considered as belonging not only to the interface between semantics and syntax, as it is widely accepted in the literature (cf. Rappoport Hovav & Levin 1995, 2001; Bresnan 1995, 2001), but also to the interface between grammar (semantics, syntax) and pragmatics. What is more, however, the application of a complex approach to the research of implicit arguments also influences the spectrum of data sources which can be used therein. Sentence-oriented approaches to implicit arguments in Hungarian have relied mostly on data from intuition and introspection. But, once one decides to study the verbs’ occurrence with implicit arguments in utterances and not in sentences, those occurrences of implicit arguments which are excluded by sentence-oriented approaches because of their strong context dependence can now also be integrated in the relevant research. Recall, for example Áron tologat (‘Áron is pushing’) which is not acceptable as a sentence in (12b), but becomes acceptable as an utterance in (13), given an appropriate context. In this respect, the use and interpretation of verbs with implicit arguments in context can be investigated, for instance, by searching occurrences in various written and spoken corpora (for example, with respect to Hungarian, the Hungarian National Corpus, at mnsz.nytud.hu,11 or various Hungarian spoken corpora, including my own 310 minute long Hungarian spoken corpus) or by conducting different thought experiments in which one can imagine and assign appropriate situations to implicit arguments.

11 The second, updated version of Hungarian National Corpus, which is a spoken language corpus, is also available at http://clara.nytud.hu/mnsz2-dev/.

Theoretical and methodological issues in the research into implicit arguments

169

However, it should be emphasised that researchers also have to rely on their own intuition when searching a corpus, so that they exclude utterances which contain performance errors rather than implicit arguments. Similarly, in thought experiments, the researcher needs to start out from her/his own intuitions while building or selecting a context for an utterance. Furthermore, during the design of relevant experimental studies, and the evaluation of their results, the researcher also needs to use her/his own intuition. It can therefore be generalised that when a researcher collects data s/he always has to rely on her/his intuition, independently of the kind of data sources s/he takes into consideration. The data I have used in my own research has originated from various kinds of direct sources. When I talk about direct data sources, I follow Kertész and Rákosi’s Plausible Argumentation Model (2012, 2014), which, based Rescher’s (1976: 6) approach to plausible reasoning, takes such sources to include perceptual experiences, experiments, theories, method applications, historical documents, etc. This conception of direct data sources is notably different from Penke and Rosenbach’s (2004: 487−488) conception, according to which data from direct sources should directly mirror the knowledge of language, excluding thus data from psycholinguistic experiments from the relevant analysis. Taking into account data from various direct sources eliminates the disadvantages of using only a particular data source and increases the plausibility of conclusions made on the basis of multiple analyses (cf. Németh T. 2010: 339−340). Similarly, Kertész and Rákosi (2012: 169−185) propose a new interpretation of the term datum which I have also applied in my research: (29) A datum is a statement with a positive plausibility value originating from some direct source. (Kertész and Rákosi 2012: 169, emphasis in original) As it can be seen in their definition, Kertész and Rákosi (2012: 169−187) assign a structure to data, which can be taken to consist of an informational content and a plausibility value. In other words, data are plausible statements and not only pure occurrences of a linguistic phenomenon in a corpus or a sentence coming from one’s intuition. The plausibility of data is initially determined by their source, or more precisely, by the reliability of their source (Kertész & Rákosi 2012: 170). So, the more reliable the source is, the higher plausibility value can be assigned to the data. So, the initial plausibility values of data can change in the course of research, like, for instance, if there are other supporting or nonsupporting sources. Let’s take an example. In traditional grammars of Hungarian (see, for example, Keszler 2000: 85, 409), verbs for natural phenomena such as hajnalodik (‘[for day to] break’) are considered subjectless, i.e. they cannot occur with an

170

Enikő Németh T.

explicit subject. Relying on the subjectless occurrences of hajnalodik cited in traditional grammars of Hungarian, and considering Kertész and Rákosi’s (2012: 169) definition of data in (29), the structure of data in traditional grammars of Hungarian can be reconstructed as follows: (30) The verb hajnalodik is a verb of natural phenomena and it cannot occur with explicit subjects. However, in the Hungarian National Corpus (mnsz.nytud.hu), one can find a considerable number of occurrences of hajnalodik (and other verbs of natural phenomena) with explicit subject, such as (31): (31)

Hajnal-od-ik az idő-Ø. [for day] break-SUFF- PRS . INDEF.3SG the time-NOM ‘The day is breaking. [lit. The time is coming to dawn.]’

On the basis of the occurrence of the verb hajnalodik with explicit subject az idő (‘the time’) in (31), the data structure in (32) can be constructed, in turn: (32) The verb hajnalodik is a verb of natural phenomena and it can occur with explicit subjects. Naturally, the contradiction between (30) and (32) is easy to notice. This contradiction can be eliminated if we compare the plausibility values of (30) and (32). According to the traditional grammars of Hungarian, (31) is not acceptable, but on the basis of the Hungarian National Corpus, my own intuition and grammaticality judgments of other native speakers of Hungarian, (31) can be evaluated as acceptable. Therefore, since the data structure in (32), built on the occurrences with explicit subjects, is supported by more direct data sources, it is much more plausible than (30), i.e. (30) loses its plausibility.12

4 Concluding remarks On the basis of the two theoretical considerations discussed so far, I have studied how Hungarian verbs can occur with various types of implicit arguments and how language users can interpret lexically unrealised verbal arguments in utterances of language use (cf. Németh T. 2010, 2012, 2014). 12 For a more detailed analysis of the occurrences of Hungarian verbs of natural phenomena with explicit subject arguments, see Németh T. (2014a).

Theoretical and methodological issues in the research into implicit arguments

171

One of the main sources of data on which I have based my research is spoken and written corpora. For one, I have collected a 310 minute long Hungarian spoken corpus, which consists of 8 interviews, of 30-45 minutes each, with 8 different people who differ in gender, age, education, profession and place of residence. After transcription and segmentation of the interviews into utterances, on the basis of the definition of utterances in (22) (cf. Németh T. 1996), the written version of my spoken corpus is 189 pages long (with 22 lines per page, 57−60 characters per line). The written corpus I have used in my research has been the Hungarian National Corpus (http://corpus.nytud.hu/mnsz/), whose relevant characteristic is its detailed morphosyntactic annotation. Every word form in it is annotated with stem, part of speech and inflectional information; information which is especially useful for the study of verbs with implicit subject and direct object arguments. In addition to my spoken corpus and the Hungarian National Corpus, I also used definitions of various verbs from Hungarian monolingual dictionaries, analyses from my previous papers (see, for example, Németh T. 2000, 2001, 2008, 2010, 2012, 2014a, 2014b; Bibok & Németh T. 2001; Németh T. & Bibok 2010), minimal pair and – especially – thought experiments, as well as comparisons with other languages treated in the literature as data sources. Apart from those, however, intuition – both my own and that of other informants – remains one of the most useful sources of data for my research. Both the theoretical and the methodological considerations discussed in this paper have served as latent background assumptions for my research into implicit arguments in Hungarian. In the evaluation of rival hypotheses from the literature as well as different interpretations of various occurrences of verbs with implicit and explicit arguments, I have also relied on the latent background assumptions of the relevant scholars; a necessary step in the assessment of rival solutions. All in all, I firmly believe that the assumption of an interaction between grammar and pragmatics in the investigation of implicit arguments in utterances and use of data from the integration of various data sources have resulted in a more complete and plausible account of implicit arguments in Hungarian.

References Ariel, Mira. 2008. Pragmatics and grammar. Cambridge: Cambridge University Press. Ariel, Mira. 2010. Defining pragmatics. Cambridge: Cambridge University Press. Bibok, Károly. 2008. Az igék szemantikája és a szintaktikai alternáció [The semantics of verbs and syntactic alternation]. In Ferenc Kiefer (ed.), Strukturális magyar nyelvtan 4. A szótár szerkezete [Structural Hungarian grammar 4: Structure of the lexicon], 23−70. Budapest: Akadémiai Kiadó.

172

Enikő Németh T.

Bresnan, Joan. 1995. Lexicality and argument structure. Paris Syntax and Semantics Conference, October 12, 1995. http://www.stanford.edu/~bresnan/paris.pdf. (accessed 10 May 2013) Bresnan, Joan. 2001. Lexical-functional syntax. Oxford: Blackwell. Carston, Robyn. 2002. Thoughts and utterances: The pragmatics of explicit communication. Oxford: Blackwell. Chomsky, Noam. 1986. Knowledge of language: Its nature, origin and use. New York: Praeger. Cornish, Francis. 2005. Null complements, event structure, predication and anaphora: A functional discourse grammar account. In John Lachlan Mackenzie & Maria de los Ángeles Gómez-González (eds.), Studies in functional discourse grammar, 21−47. Bern: Peter Lang. Cote, Sharon A. 1996. Grammatical and discourse properties of null arguments in English. Philadelphia: University of Pennsylvania PhD dissertation. Cummins, Sarah & Yves Roberge. 2005. A modular account of null objects in French. Syntax 8. 44−64. Dowty, David. 1981. Quantification and the lexicon: A reply to Fodor and Fodor. In Teun Hoekstra, Michael Moortgrat & Harry van der Hulst (eds.), The scope of lexical rules, 79−106. Dordrecht: Foris. Elbourne, Paul. 2008. Implicit content and the argument from binding. In Tova Friedman & Satoshi Ito (eds.), Proceedings of semantics and linguistic theory XVIII (SALT 18), 284−301. Ithaca: Cornell University, CLC Publication. Fillmore, Charles J. 1986. Pragmatically controlled zero anaphora. Berkeley Linguistic Series 12: 95−107. Fodor, Jerry & Janet Dean Fodor. 1980. Functional structure, quantifiers and meaning postulates. Linguistic Inquiry 11. 759−769. García-Velasco, Daniel & Carmen Portero-Muñoz. 2002. Understood objects in functional grammar. (Working Papers in Functional Grammar 76) Amsterdam: University of Amsterdam. Gillon, Brendan S. 2012. Implicit complements: A dilemma for model theoretic semantics. Linguistics and Philosophy 35. 313−359. Goldberg, Adele. 2005a. Argument realization: The role of constructions, lexical semantics and discourse factors. In Jan-Ola Östman & Mirjam Fried (eds.), Construction grammars: Cognitive grounding and theoretical extensions, 17−43. Amsterdam: John Benjamins. Goldberg, Adele. 2005b. Constructions, lexical semantics and the correspondence principle: Accounting for generalizations and subregularities in the realization of arguments. In Nomi Erteschik-Shir & Tove Rapoport (eds.), The syntax of aspect, 215−236. Oxford: Oxford University Press. Groefsema, Marjolein. 1995. Understood arguments: A semantic/pragmatic approach. Lingua 96. 139−161. Haegeman, Liliane & Jacqueline Guéron. 1999. English grammar: A generative perspective. Oxford: Blackwell. Iten, Corinne, Marie-Odile Junker, Aryn Pyke, Robert Stainton & Catherine Wearing. 2005. Null complements: Licensed by syntax or by semantics-pragmatics? Actes du congrés de l’ Association canadienne de linguistique 2004. Proceedings of the 2004 Annual Conference of the Canadian Linguistic Association. http://eastcree.org/mojunker/ACL-CLA/pdf./ItenJunker-Pyke-CLA-2004.pdf. (accessed 8 October 2010) Kertész, András & Csilla Rákosi. 2012. Data and evidence in linguistics: A plausible argumentation model. Cambridge: Cambridge University Press. Keszler, Borbála. 2000. Magyar grammatika [Hungarian Grammar]. Budapest: Nemzeti Tankönyvkiadó.

Theoretical and methodological issues in the research into implicit arguments

173

Komlósy, András. 1992. Régensek és vonzatok [Predicates and complements]. In Ferenc Kiefer (ed.), Strukturális magyar nyelvtan: Mondattan [A structural grammar of Hungarian: Syntax], 299−527. Budapest: Akadémiai Kiadó. Komlósy, András. 1994. Complements and adjuncts. In Ferenc Kiefer & Katalin É. Kiss (eds.), Syntax and semantics 27, 91−178. New York: Academic Press. Komlósy, András. 2001. A lexikai-funkcionális grammatika mondattanának alapfogalmai [Basic notions in the syntax of lexical-functional grammar]. Budapest: Tinta Könyvkiadó. Neale, Stephen. 1990. Descriptions. Cambridge, MA: MIT Press. Németh T., Enikő. 1995. On the role of pragmatic connectives in Hungarian spoken discourse. In Brita Wårvik, Sanna-Kaisa Tanskanen & Risto Hiltunen (eds.), Organization in discourse. Proceedings from the Turku Conference, 393−402. Turku: University of Turku (Anglicana Turkuensia 14). Németh T., Enikő. 1996. A szóbeli diskurzusok megnyilatkozáspéldányokra tagolása [Segmentation of spoken discourse into utterance-tokens]. Budapest: Akadémiai Kiadó (Nyelvtudományi Értekezések 142) 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: Manners of their occurrence and possibilities of their identification. In István Kenesei (ed.), Argument structure in Hungarian, 113−156. Budapest: Akadémiai Kiadó. Németh T., Enikő. 2008. Az implicit alanyi és tárgyi argumentumok előfordulásának lexikaiszemantikai jellemzői [The lexical-semantic properties of the occurrence of implicit subject and direct object arguments]. In Ferenc Kiefer (ed.), Strukturális magyar nyelvtan 4. A szótár szerkezete [A structural grammar of Hungarian 4: The structure of the lexicon], 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 Enikő Németh T. and Károly Bibok (eds.), The role of data at the semantics–pragmatics interface, 305−348. Berlin: Mouton de Gruyter. Németh T., Enikő. 2012. Lexical-semantic properties and contextual factors in the use of verbs of work with implicit subject arguments in Hungarian. Intercultural Pragmatics 9. 453−477. Németh T., Enikő. 2014a. Hungarian verbs of natural phenomena with explicit and implicit subject arguments: their use and occurrence in the light of data. In András Kertész & Csilla Rákosi (eds.), The evidential basis of linguistic argumentation, 103−132. Amsterdam: John Benjamins. Németh T., Enikő. 2014b. Implicit arguments at the grammar-pragmatics interface: some methodological considerations. Argumentum 10. 679−694. Németh T., Enikő & Károly Bibok (eds.). 2001. Pragmatics and the flexibility of word meaning. Amsterdam: Elsevier. Németh T., Enikő & Bibok, Károly. 2010. Interaction between grammar and pragmatics: The case of implicit arguments, implicit predicates and co-composition in Hungarian. Journal of Pragmatics 42. 501−524. Penke, Martina & Annette Rosenbach (eds.). 2004. What counts as evidence in linguistics? [Special issue]. Studies in Language 28 (3). Pethő, Gergely & Éva Kardos. 2009. Cross-linguistic evidence and the licensing of implicit arguments. In Bergljot Behrens & Catherine Fabricius-Hansen (eds.), Structuring information in discourse: The explicit-implicit dimension. Oslo Studies in Language 1. 33−61.

174

Enikő Németh T.

Pléh, Csaba. 1994. Mondatközi viszonyok feldolgozása: Az anaphora megértése a magyarban [Processing intersentential relations: Interpreting anaphors in Hungarian]. Magyar Pszichológiai Szemle 50. 287−320. Pléh, Csaba. 1998. A mondatmegértés a magyar nyelvben [Sentence comprehension in Hungarian]. Budapest: Osiris Kiadó. Pléh, Csaba & Katalin Radics. 1978. Truncated sentence, pronominalization and the text. Acta Linguistica Academiae Scientiarum Hungaricae 28. 91−113. Pléh, Csaba & Brian McWhinney. 1987. Anaphora resolution in Hungarian. Acta Linguistica Academiae Scientiarum Hungaricae 37. 103−124. Rappoport-Hovav, Malka & Beth Levin. 1995. The elasticity of verb meaning. In Myra Hovav & Anita Mittwoch (eds.), IATL 2: Proceedings of the Tenth Annual Conference of the Israel Association for Theoretical Linguistics and the Workshop on the Syntax-Semantics Interface. Haifa: The Israel Association for Theoretical Linguistics. 153−171. Rappoport-Hovav, Malka & Beth Levin. 2001. An event structure account of English resultatives. Language 77. 766−797. Recanati, François. 2007. It is raining (somewhere). Linguistics and Philosophy 30. 123−146. Rescher, Nicholas. 1976. Plausible reasoning. Assen/Amsterdam: Van Gorcum. Rice, Sally. 1988. Unlikely lexical entries. Berkeley Linguistic Series 14. 202−212. Saeboe, Kjell. 1996. Anaphoric presupposition and zero anaphora. Linguistics and Philosophy 19. 187−207. Scott, Kate. 2006. When less is more: Implicit arguments and relevance theory. UCL Working Papers in Linguistics 18. 139−170. Scott, Kate. 2013. Pragmatically motivated null subjects in English: A relevance theory perspective. Journal of Pragmatics 53. 68−83. Sperber, Dan & Deirdre Wilson. 1995. Relevance: Communication and cognition. 2nd edn. Oxford: Blackwell. Tóth, Ildikó. 2001. Impersonal constructions and null expletives. In István Kenesei (ed.), Argument structure in Hungarian, 51−78. Budapest: Akadémiai Kiadó.

Thorstein Fretheim

10 The Pragmatics of ‘Yes’ and ‘No’ Abstract: Building on Fretheim (2014), which analysed response words as anaphors that represent propositional content, this paper offers a pragmatic analysis of response word utterances, with a focus on certain implications of the fact that a response word does more than just encode affirmative or negative polarity. In this respect, it refutes Holmberg’s (2013a) claim that a response word is syntactically complex; an item obligatorily followed by a sentence that may be deleted if and only if its proposition is equal to the interlocutor’s most recently expressed proposition. Differences between the pragmatic consequences of a response word system with two members, like English, and one with three members, here Scandinavian, are highlighted, and a slight modification of the relevancetheoretic concept of explicature is suggested.

1 Introduction In a talk exchange, where the first speaker (I call her Anne in this paper) expresses a proposition whose truth the second speaker (called Ben) affirms or denies, the response word used by Ben is an anaphor whose semantic value, i.e. propositional content, is resolved through a process of saturation (cf. Recanati 2010). It is a linguistic variable in need of contextual instantiation, and its resolution depends, above all, on information derived from its antecedent, which may be a sub-clausal constituent of Anne’s sentence or even her entire sentence (cf. Fretheim 2014). In the data examined in this paper, Anne’s sentence contains a predicate phrase with an adverb followed by a negative-polarity complement. My focus here, is on types of anaphora relations between Anne’s and Ben’s stimuli; I am therefore not concerned with the use of response words as back-channel signals and the like, but rather only with their function as the linguistic vehicle of a

Acknowledgements: I would like to thank Stavros Assimakopoulos for his indefatigable editorial work with my paper. Thorstein Fretheim, Norwegian University of Science and Technology DOI 10.1515/9781501505089-010

176

Thorstein Fretheim

communicated proposition – an explicature in the relevance-theoretic sense (Sperber and Wilson 1995).1 According to Anders Holmberg (2012, 2013a, 2013b), a speaker uses a response word to assign a truth value to a proposition which is either expressed in a follow-up sentence or deleted under identity with the proposition expressed in the addressee’s utterance which the response word is a verbal reaction to. In his account, he analyses answers to yes/no questions as sentence constructions in which the propositional part is often deleted, i.e. not pronounced, under the condition of identity to the proposition expressed by the question (Holmberg 2012: 5). A more technical, though brief statement geared to a syntactically sophisticated public is found in the abstract of Holmberg (2013a: 51): It is proposed that bare yes and no-answers to yes/no questions are sentential expressions with the structure [yes/no [Foc [IP . . . [Pol X] . . .]], where the answer particle is merged in the spec of Focus in the CP-domain, and assigns a value, either affirmative or negative, to the polarity variable in IP. The IP has a polarity variable because it is inherited from the question. For the same reason the IP is typically elided, being identical to the IP of the question.

Holmberg believes that the optionally present IP accompanying the response word encodes a proposition which is fixed, requires no pragmatic computation of any sort and may be omitted from surface syntax because it is identical to the proposition expressed in the first speaker’s sentence. This proposition is grammatically represented with a polarity variable, because the whole point of asking a question is to request the interlocutor’s confirmation or denial of the proposition expressed. Holmberg reasons like this, under the assumption that the first speaker’s speech act is invariably a question in the syntactic form of a polar interrogative; yet, a response word whose function is to assent to or dissent from a proposition may be a linguistic reaction to a variety of speech act types, including, besides yes/no questions, assertions performed by means of a declarative, and commands or requests performed by means of an imperative. In this paper, I am going to demonstrate the inadequacy of the assumption that the identity of the proposition affirmed or denied by a response word can be established without pragmatic inference (see also Fretheim 2014). A response word is a truth-conditional expression whose reference is determined in the same way as the reference of any indexical term. In other words, the response word represents a truth-evaluable proposition, but it is an anaphor that must be saturated by a set of truth conditions pragmatically transferred from the first speaker’s utterance. The correct match between the anaphoric response word 1 See, however, section 6 below, in which I express a concern about the notion of explicature as currently defined, if it is to be applied to the propositional content of response words.

The Pragmatics of ‘Yes’ and ‘No’

177

and the proposition targeted by it depends on pre-semantic pragmatic inference in conjunction with the lexical semantics of the response word. For Holmberg, a response word is an encoder of a truth value, while the proposition to which the positive or negative truth value is assigned is expressed in a follow-up sentence. For me there is no follow-up sentence, unless it is realised physically, that is, either acoustically or in printed form.

2 Polarity and the antecedent – anaphor relation 2.1 The data The crucial type of data examined in this paper are talk exchanges in which Anne’s utterance is an interrogative consisting of a finite verb, a subject, an adverb and a complement with negative polarity, and Ben’s answer is a response word. My examples are mostly adapted versions of illustrations appearing in Holmberg’s papers on the syntax of response words, offered in support of various aspects of his syntactic analysis. The same kind of data is used here in an attempt to refute Holmberg’s claim that a response word is a component in a complex syntactic construction. In his account, the response word is void of content, since the proposition to which the truth value is assigned is expressed in a sentence whose surface-structural appearance is considered dispensable, on condition that it is identical with the proposition expressed by the first speaker: a classic case of recoverability of syntactic deletion. In every instance, I shall demonstrate the superiority of my own analysis of response words, as anaphors whose truth-conditional content is determined through a pragmatic process of saturation that draws on the meaning of a linguistic antecedent.

2.2 Sloppy identity Although Holmberg’s analysis of response word constructions does not address talk exchanges in which the first speaker’s sentence is a declarative, I cannot imagine that he might want to postulate different analyses of response words depending on the sentence type in the first speaker’s turn. Consider, for example, (1) where Anne’s sentence is not a polar interrogative, but a negative declarative. (1)

Anne:

I didn’t like Peter’s performance much.

Ben1:

No.

Ben2:

No. I see.

Ben3:

No. I agree.

178

Thorstein Fretheim

How do we process the bare response word in (1) Ben1? This depends a lot on the intonation imposed on the monosyllabic answer. An unemotional, flatpitched No is hardly more than a back-channel signal, an indication that the speaker has comprehended the interlocutor’s message. The response No produced with an energetic falling pitch contour through the syllable sounds more like an announcement of the fact that Ben too disliked Peter’s performance. The follow-up utterance ‘I see’ in (1) Ben2 conveys something that could be paraphrased as ‘Right, you didn’t like Peter’s performance much’. In contrast, the token of No in (1) Ben3 confirms a negative proposition which is not the proposition expressed by Anne. A consequence of Holmberg’s analysis is that the deletion of the sentence postulated as an obligatory sequel to a response word should be impermissible if the identity relation between the first speaker’s proposition and the second speaker’s proposition is one of sloppy rather than absolute identity. As the bare response word No in (1) Ben1 is consistent with the same truth-evaluable meaning as No in (1) Ben3, Holmberg’s deletion under identity requirement for acceptable use of a bare response word is evidently violated in Ben1, contrary to the general public acceptance of sloppy-identity relations between the first speaker’s proposition and the second speaker’s use of a response word to express agreement with it.

2.3 The “polarity-based” and the “truth-based” answering strategies As Holmberg correctly points out (Holmberg 2012, 2013a, 2013b), there are two competing strategies for production and comprehension of English response words. On the one hand there is a polarity-based answering strategy characterised by the fact that a positive response word, like the English Yes, is used to assert the truth of a proposition with positive polarity, while a negative response word, like the English No, asserts the truth of a proposition with negative polarity. While this is the basic answering strategy employed by native speakers of English, there is a competing strategy, which Holmberg calls truth-based. This alternative implies that a positive response word affirms the proposition which the interlocutor has expressed and may seem to endorse in the absence of adverse information, while a negative response word denies it. This is typical of all language codes that use independent words, particles or bound morphemes as response markers which indicate the speaker’s agreement or disagreement with the interlocutor’s proposition, regardless of its polarity, with Chinese and Japanese being prominent examples.

The Pragmatics of ‘Yes’ and ‘No’

179

Resolving the truth-evaluable meaning of a token of a bare Yes or No given in response to a yes/no interrogative with negative polarity, however, may turn out to be rather tricky for the addressee. Does No in (2) Ben1 confirm or deny the negative proposition expressed by Anne? Does Yes in (2) Ben2 confirm or deny Anne’s negative proposition? Are No and Yes interchangeable here? (2)

Anne:

Do you not remember where you put your keys?

Ben1:

No.

Ben2:

Yes.

In English talk, the utilisation of the truth-based strategy is most typically found in answers to questions with negative polarity. When Anne’s sentence is a negative interrogative, there may be a reason to doubt that the bare response word No confirms her negative proposition and that the bare response word Yes rejects it as false and conveys that its positive counterpart is true. Holmberg suggests that speakers of English are sometimes, subject to grammar-internal constraints, liable to adopt the truth-based answering strategy in their distribution of Yes and No, so that the token of No in (2) Ben1 above might mean that Ben is protesting against the thought that he possibly does not remember where he put his keys. By the same token, Yes in (2) Ben2 could express agreement with Anne’s negative proposition. The proposition expressed by Anne in (2) is a singular proposition, and Ben’s negative answer No in (2) Ben1 and affirmative alternative Yes in (2) Ben2 likewise express singular propositions. No in (2) Ben1 may strike the addressee as being ambiguous between two mutually incompatible interpretations. On the assumption that Ben is acting in accordance with the polarity-based strategy, Anne will conclude that Ben1 confirms Anne’s negative proposition, but it is conceivable that Ben is instead using No as a sign of disagreement with the proposition expressed by Anne. The former tack is in line with the polarity-based answering strategy, while the latter conforms to the truth-based answering strategy. Similarly, the positive response word Yes in (2) Ben2 either communicates that Ben does remember where he put his keys, or that he does not remember, because Yes may be understood either as a sign of confirmation of the nonnegative counterpart of the negative proposition expressed by Anne (the polaritybased strategy) or as an indication that her negative proposition is true (the truthbased strategy). The talk exchange in (3) below contrasts with (2) in a way that bears directly on the main objective of the present paper, which is to demonstrate that a response word stands for a propositional content to be identified through a

180

Thorstein Fretheim

pragmatic process of bottom-up truth-conditional saturation, that is, a mandatory resolution of a proposition communicated by way of an indexical. Both speakers Anne and Ben express general propositions in (3). The focus of Anne’s question is whether there have been occasions when Ben failed to remember where he put his keys. In uttering Ben1, Ben denies that this ever happens to him, whereas the utterance of Ben2 confirms it. (3)

Anne:

Do you occasionally not remember where you put your keys?

Ben1:

No.

Ben2:

Yes.

The negative complement of the adverb occasionally in Anne’s interrogative does not by itself express a truth-evaluable negative proposition, because occasionally scopes over it and informs Ben that the proposition should be processed as general, not singular. The adverb has the semantic role of a temporal predicate, as in the paraphrase of the affirmative answer Yes in (3) Ben2: ‘It happens occasionally that I do not know where I put my keys’. Holmberg does not share my view of what causes Ben to say Yes in (3) Ben2. For him, the answer Yes affirms the truth of what he takes to be a negative proposition expressed by Anne. What he says about the truth-conditional adverb occasionally is that it belongs to a subcategory of adverbs whose presence in a yes/no question with negative polarity makes it very natural for speakers of English to resort to a truth-based answering strategy (more on this in section 4). By so doing, they follow a more general trend, he says, which gradually pushes the English use of response words a bit more in the direction of a truthbased answering system, as opposed to the traditional polarity-based system that is still dominant in English dyadic conversation. Holmberg understands No in (3) Ben1 as rejecting a negative proposition expressed by Anne, and Yes in (3) Ben2 as confirming this negative proposition. This pattern, he says, is an indication of the likelihood of an on-going drift towards a truth-based strategy that determines the semantic value of the English response words Yes and No, as suggested by the fact that this usage pattern is on the increase. My approach implies that the added adverb occasionally in the exchange of (3) actually makes the so-called truth-based response analysis less amenable in (3) than in (2), where there is no adverb, in contradistinction to Holmberg’s claim that the added adverb is conducive to a treatment of English response words as indicators of agreement or disagreement with what the speaker takes to be the interlocutor’s epistemic stance. Anne will construe the content of the anaphor No in (3) Ben1 as a denial of the assumption that it happens occasionally that Ben forgets where his keys are, much like she will construe the

The Pragmatics of ‘Yes’ and ‘No’

181

positive alternative Yes in (3) Ben2 as confirmation of the assumption that Ben occasionally does forget that. Since the negative complement of occasionally is out of focus, Ben’s distribution of the two response words No and Yes is not affected by the negative polarity of the infinitival phrase in Anne’s sentence. The hypothesis that English is a language code whose response-word system exhibits an exploitation of two distinct principles that govern the distribution of Yes and No seems to be supported by a type of data that Holmberg does not give much attention to. English negation may be expressed by means of a negative prefix like un-, dis-, mis- or de-, but can also be incorporated in the lexical verb, as in fail or forget. In such cases Yes does mean that the speaker asserts the truth of the negative proposition expressed in the antecedent structure. In this respect, the formal difference between Anne’s use of ‘not remember’ in (4) and her use of ‘forget’ in (5) affects Ben’s choice of response word. (4)

(5)

Anne:

Did he not remember to lock the door?

Ben1:

No. ‘No, he did not remember’ or, if produced with emphasis: ‘No, it is not so that he did not remember.’

Ben2:

Yes. ‘Yes, that’s exactly what happened.’ (the truth-based strategy) or possibly: ‘Oh yes, he did remember.’ (the polarity-based strategy)

Anne:

Did he forget to lock the door?

Ben1:

No. (‘He didn’t forget.’/‘He remembered to lock it.’)

Ben2:

Yes. (‘He forgot.’)

In (4), both No and Yes may in principle be interpreted in accordance with either the polarity-based or the truth-based strategy. The polarity-based strategy would yield the content ‘He did not lock the door’ if Ben says No, as in (4) Ben1. If Ben’s response is Yes, as in (4) Ben2, the truth-based strategy might lead Anne to the same content. Alternatively, an utterance of No produced with an emphatic rise-fall contour could mean that Ben rejects Anne’s negative proposition (external negation), and Yes could possibly mean that Ben did lock the door, though this is probably a less accessible interpretation. In turn, in (5),

182

Thorstein Fretheim

No and Yes are unambiguous and mutually exclusive answers. The negative response No denies the assumption that the man referred to forgot to lock the door, while the affirmative response Yes confirms that he forgot to lock it. There is thus no meaningful opposition between a polarity-based and a truth-based answering strategy when the negation is not expressed with the free form not (or enclitic n’t) but is instead incorporated in the verb stem, as in the case of the verb forget.

2.4 Presupposed infinitival complements One characteristic feature of several of Holmberg’s examples where Yes allegedly assigns the truth value ‘True’ to a negative proposition is that the negative complement of the adverb in the first speaker’s interrogative is presupposed to be true. In this respect, the infinitival complement does not contain any information that the speaker asks the addressee to confirm or disconfirm. This fact is overlooked by Holmberg, who links the affirmative response word to what he believes to be the speaker’s expression of agreement with and endorsement of a negative proposition, an act made possible by the speaker’s adherence to the truth-based answering strategy. Yet, the affirmative Yes was directed at the focus of the first speaker’s question in example (3) above, and the focus was the adverb occasionally. The illustrations in the present subsection show even more convincingly that the negative complement is out of focus for Anne and therefore not targeted by Ben’s response words. Information focus (cf. Gundel and Fretheim 2004) is a notion that applies equally to the performance of acts of requesting information and acts of offering information. In the talk exchanges presented in this paper, the speaker requesting information employs a polar interrogative in nearly all instances, with a sentence that contains both a focused part and a contextually given part. The questions asked in the crucial examples concern whether or not the concept designated by the focused part of the linguistic stimulus contributes what the hearer would accept as a truth condition. So, the hearer answers with a response word, which invariably targets the focus of the first speaker’s question. A focal pitch accent on the adverbial information focus and a post-focal prosodic handling of the following complement is a natural way for Anne to indicate where her focus is located. In (6), one of Holmberg’s examples, Anne presupposes the truth of the assumption that the man referred to had failed to return the books on time at least once before. Once more is the focus of Anne’s question; that is, she wants to know if the man’s failure to return the books on time happened once again. No informs Anne that it did not happen again, while Yes informs her that it did.

The Pragmatics of ‘Yes’ and ‘No’

(6)

Anne:

Did he once more not return the books on time?

Ben1:

No.

Ben2:

Yes.

183

Similarly, in (7), Anne presupposes that Ben did not dress up for the occasion referred to and asks whether this was deliberate on his part. No in (7) Ben1 tells Anne that Ben just happened to wear casual clothes, whereas Yes in (7) Ben2 confirms the thought expressed by Anne, that Ben’s failure to dress up was his deliberate choice. (7)

Anne:

Did you deliberately not dress up for the occasion?

Ben1:

No.

Ben2:

Yes.

So, both No in Ben1 and Yes in Ben2 preserve the presupposition that Ben did not dress up for the occasion.

2.5 Tag question constructions An English tag question appended to a declarative host sentence is invariably directed at the information focus of the host sentence and not at any presupposed information expressed in it. The unmarked types of tag question construction in English consist of a negative declarative followed by a positive-polarity tag question or a positive declarative followed by a negative-polarity tag question. This is illustrated in (8b) and (8c). (8a), a negative tag after a negative declarative, is ungrammatical. (8d), on the other hand, is a same-polarity construction with positive polarity in the host as well as the tag, which is grammatical but pragmatically marked compared to the opposite-polarity constructions in (8b–c). In this way, an utterance like (8d) places a special constraint on the context required for its acceptability, a phenomenon that will be described below. (8)

a. *You didn’t dress up for the occasion, didn’t you? b.

You didn’t dress up for the occasion, did you?

c.

You dressed up for the occasion, didn’t you?

d.

You dressed up for the occasion, did you?

184

Thorstein Fretheim

So, what happens when the adverb deliberately in Anne’s interrogative in (7) is added in front of didn’t in (8a), where the declarative host of the appended tag question is negative? How does the relative linear order of the adverb and the negation marker affect the acceptability and the semantics of such data? I asked twenty-five native English informants to judge the acceptability of the four written tag question constructions in (9), two with deliberately outside, and two with deliberately inside the scope of negation. (9)

a.

You deliberately didn’t dress up for the occasion, didn’t you?

b.

You deliberately didn’t dress up for the occasion, did you?

c.

You didn’t deliberately dress up for the occasion, didn’t you?

d.

You didn’t deliberately dress up for occasion, did you?

It turned out that every single informant ruled out (9b) and (9c), and everyone accepted (9d). As for (9a), the judgments were divided; sixteen accepted it, while nine rejected it. Having found that the syntactic forms of (9b) and (9c) were rejected as illformed by everyone, I asked my informants what they felt that an utterance of Yes and an utterance of No given in response to the remaining stimuli (9a) and to (9d) meant for them. Even those nine people who were not willing to call (9a) well-formed were asked how they might interpret an answer to it in the form of a response word Yes or No, if they could for a moment disregard the fact that they had judged its syntactic form as bad. In what follows, I shall first discuss at length the results from that part of the test which concerned the acceptability of (9a–d), before commenting briefly on the second part, which concerned the meaning of Yes and No produced in response to the two stimuli that were not uniformly rejected: (9a), where deliberately precedes the negation marker and the tag has negative polarity, and (9d), where the negation marker precedes deliberately and the tag has positive polarity. In contrast to (9a) and (9b), the linear order of elements in (9c) and (9d) causes us to perceive deliberately as an item that falls within the scope of negation. Unlike the positive-polarity tag question in the well-formed structure of (9d), which indicates that the speaker wants the hearer to confirm the belief that the hearer did not deliberately choose to dress casually, a negative-polarity tag question may not be added to the negative-polarity declarative in (9c), hence its ill-formedness. With (9c) and (9d) the situation is black-and-white: no acceptance of (9c), no rejection of (9d).

The Pragmatics of ‘Yes’ and ‘No’

185

The situation is much less clear when we turn to (9a) and (9b). (9a) looks superficially like a tag question construction of the same sort as the ill-formed structure of (8a). Why, then, was it judged to be acceptable by sixteen out of twenty-five native informants? The negative polarity in the tag should match a host sentence with non-negative polarity. Since it is presupposed by the speaker of (9a) that the hearer did not dress up, the tag question didn’t you? matches the positive-polarity adverb deliberately in the host sentence, an adverb that scopes over its negative complement and is the focus of the speaker’s question. Thus, the majority of informants processed (9a) as an opposite-polarity tag question construction made up of a positive-polarity host sentence and a negative-polarity tag, as in the adverb-free structure of (8c). My conjecture about the nine informants who felt that (9a) is syntactically ill-formed is that they may have ignored the presence of the adverb deliberately due to the conspicuity of the double appearance of didn’t. For them, (9a) was presumably a construction with negative polarity (didn’t) in host sentence and tag alike, just like (8a), which is a syntagmatic combination which speakers of English are generally not comfortable with. As far as (9b) is concerned, all informants were sensitive to the mismatch between host sentence and tag in it, even though it is deceptively similar to an opposite-polarity construction. The syntactic form of (9a) may have been somewhat easier for the majority of informants to perceive as an opposite-polarity construction than (9b), despite structural appearances. Allow me to digress slightly at this point and pinpoint, as I said before, that positive polarity in host as well as tag is not eo ipso bad. The two same-polarity sentences in (10) may be marked compared to their opposite-polarity counterparts but they are not ill-formed. They are just pragmatically different from opposite-polarity constructions. In fact, (10b), which starts with the inference connective so, seems to me to be more natural than the opposite-polarity sentence in (11), where the positive polarity of the declarative matches the negative polarity of the tag question. (10)

a.

You deliberately dressed up for the occasion, did you?

b.

So you deliberately dressed up for the occasion, did you?

(11) ?So you deliberately dressed up for the occasion, didn’t you?

186

Thorstein Fretheim

An opposite-polarity tag question construction allows the speaker to ask for confirmation of a thought that the speaker may have entertained prior to the current conversation, but it is possible to use a same-polarity tag question construction to ask for confirmation of a thought made manifest to the speaker by way of a stimulus produced by the interlocutor immediately before the request for clarification in (10a) or (10b).2 Why, then, did all my informants disapprove of (9b) above, even though same-polarity sentences like (8d) and (10a–b) are well-formed? I believe it must have been difficult for them to reconcile the positive tag ‘did you?’ with a host declarative that contains not just the positive-polarity item deliberately, but also a negative-polarity complement that gives the combination of host sentence and tag the superficial look of a regular opposite-polarity construction. If one perceives a link between the positive tag in (9b) and the negation in the host sentence, one makes oneself guilty of eschewing the information focus deliberately, an item responsible for the impression that the speaker presupposes that the hearer did not dress up. It is thus possible that the positive tag after the host sentence in (9b) was uniformly rejected because it was impossible to reconcile it, on the one hand, with deliberately and, on the other, with the negative complement didn’t dress up for the occasion. Paradoxically, (9a) probably fared better than (9b) because a same-polarity structure like *You didn’t dress up for the occasion, didn’t you? with negation in host as well as tag is so glaringly bad, which may have made it somewhat easier for the interpreter confronted with (9a) to spot a rescuing link between the negative tag question and the adverb deliberately right away. I admit that what I have presented as an attempt to justify my informants’ pattern of responses to the written stimuli (9a) and (9b) is a bit speculative, and it may also be objected that my group of informants was too limited. Still, I would like to mention yet another plausible reason for the acceptance of (9a) by a majority of informants and for the fact that (9b) was rejected wholesale. Although the four stimuli in (9) were written sentences, the informants probably tried out a silent pronunciation of each one of them before submitting their answers. It is very likely that they activated a mental image of a spoken utterance with a phonetic realisation that included a focal accent on the adverb deliberately in the form of a fall from a high to a low pitch-level on the second, accented syllable of this word form, and a post-focal low level of pitch through its complement: the string didn’t dress up for the occasion in (9a–b), and the string 2 I might add at this point that a tag question construction with positive polarity in host as well as tag may sound a bit arrogant, especially if the syntactic form is accompanied by a special tone-of-voice indicating that the speaker is disenchanted or sarcastic.

The Pragmatics of ‘Yes’ and ‘No’

187

dress up for the occasion in (9c–d). This intonation pattern cues an interpretation of the negative tag in (9a) as an item directed solely at the question of whether the hearer had deliberately not bothered to dress up, a parallel to a question like What you did was deliberate, wasn’t it?. In the case of (9b), however, a focal accent on deliberately cues recognition of the negative complement of deliberately as the linguistic vehicle of an assumption that is presupposed to be true. An internalised intonation pattern of the sort described here presumably makes it extremely hard for the interpreter to link the positive tag did you? to the postfocal negative complement of deliberately. In this respect, my informants must have failed to spot any connection between host and tag, which could have rescued (9b) as a grammatically well-formed sentence. The second step in my investigation of how a group of native English speakers reacted to the written stimuli in (9) concerned their comprehension of the propositional content of the answers No and Yes to (9a) and (9d), the two sentences in the set which were not unanimously rejected. (12)

(13)

Anne:

You deliberately didn’t dress up for the occasion, didn’t you?

Ben1:

No.

Ben2:

Yes.

Anne:

You didn’t deliberately dress up for the occasion, did you?

Ben1:

No.

Ben2:

Yes.

The resolution of the explicature expressed by means of No and Yes in (13) caused no problem. Ben1 meant ‘No, I didn’t’ for everyone, and Ben2 meant ‘Yes, I did’. As for the question – answer pairs in (12), three of the twenty-five informants felt that No in the utterance of (12) Ben1 was incoherent as an answer to Anne’s question, with the remaining twenty-two saying that it was a denial of the thought that Ben deliberately chose not to dress up. Twenty-one informants understood Yes in (12) Ben2 to confirm the belief that Ben’s behaviour was deliberate, but the remaining four people said the meaning was that Ben did dress up. No one said that Ben’s Yes in (12) communicates the truth of the belief that Ben did not dress up. Those four people who did not follow the majority in their semantic analyses of the question – answer pairs in (12) all belonged to the group who judged (9a) to be unacceptable or unnatural in the first part of the test.

188

Thorstein Fretheim

To sum up, on the basis of the results of my investigation of tag question constructions with the adverb deliberately in the host sentence, I can now conclude that the acceptance of (9a) by a majority of informants shows that they construed deliberately as the speaker’s focus. Moreover, the intuition that the negative infinitival complement is presupposed and therefore not addressed by the negative-polarity tag in (9a) is strongly supported by the informants’ comprehension of the truth-conditional content of the response words in (12). Nearly all informants saw an antecedent-anaphor relation between Anne’s adverb deliberately and the indexicals No in (12) Ben1 and Yes in (12) Ben2.

3 Three-valued response word systems: Swedish and Norwegian The Scandinavian languages differ from English in the number of response words available to the speaker. As in numerous other languages both in and outside of Europe, there is one negative response word and two affirmative ones, with Ja and Jo being the relevant forms in Swedish, Danish and Norwegian.3 The glide that is common to the forms Ja and Jo reflects the functional similarity between them in an iconic way, while the distinct vowels reflect what separates them semantically. All in all, they encode distinct procedures for the addressee to follow in the inferential phase of their truth-theoretic interpretation (for an explication of the relevance-theoretic notion of procedural semantics, see Blakemore 1987; Wilson and Sperber 1993; Escandell-Vidal, Leonetti, and Ahern 2011, among many others). In this picture, Ja encodes the procedural information that the hearer must seek out a linguistic antecedent with positive polarity and construe a positive-polarity proposition whose truth Ja affirms. Jo instructs the hearer to seek out a linguistic antecedent with negative polarity and affirm the truth of a positive-polarity proposition, which contradicts the meaning of the negative antecedent. The negative response word Nei affirms a negative proposition, either by denying a positive-polarity proposition expressed by the interlocutor or by confirming a negative-polarity proposition expressed by the interlocutor. Anne’s Swedish interrogative in (14) is a negative-polarity sentence, so answering Ja is no option for Ben. Jo, on the other hand, is an adequate answer if Ben intends to deny Anne’s negative proposition.

3 In certain varieties of spoken Norwegian, Jau replaces Jo.

The Pragmatics of ‘Yes’ and ‘No’

(14)

Anne:

Drack du inte upp ditt kaffe? drank you not up your coffee ‘Did you not drink up your coffee?”

Ben1:

Nej. ‘No.’ (‘I didn’t.’)

Ben2:

#Ja.

Ben3:

Jo. ‘Yes.’ (‘I did.’)

189

The Swedish adverb avsiktligt expresses much the same concept as the English ‘deliberately’. Avsiktligt may be targeted and confirmed by means of the response word Ja that requires a positive-polarity antecedent, but the question–answer pairs in (15), adapted from Holmberg (2013b), demonstrate that Ja and Jo are not mutually exclusive in Swedish conversation. (15)

Anne:

Drack du avsiktligt inte upp ditt kaffe? drank you deliberately not up your coffee ‘Did you deliberately not drink up your coffee?’

Ben1:

Nej. ‘No.’

Ben2:

Ja. ‘Yes.’ (‘It was deliberate on my part.’)

Ben3:

Jo. ‘Yes.’ (‘I did drink it up.’)

For the hearer, Ben, avsiktligt counts as a positive-polarity antecedent; an antecedent that licenses the use of Ja in an act of confirming the belief that Ben did not empty his cup of coffee on purpose. Jo, however, tells the hearer to identify a negative-polarity proposition derived from Anne’s utterance and affirm the content of ‘Ben did drink up his coffee’. Thus, the utterance of (15) Ben3 Jo rejects Anne’s presupposition and displays the irrelevance of Anne’s information focus, that is the adverb avsiktligt. The Norwegian Nei, Ja and Jo pattern like their Swedish counterparts. The position of the quantified adverbial phrase hele uka (‘all week’) between the subject NP and the negation marker in (16) is less natural than its position at the end of the sentence, shown in (17). Still, of the two interrogatives, it is only

190

Thorstein Fretheim

(16) which licenses an answer Ja that is meant to target the universally quantified phrase that scopes over negation. Norwegian informants uniformly frowned upon the use of Ja as a response word intended to target hele uka in (17), in an act of affirming the assumption that no milk was bought that week. (16)

Hadde Jan hele uka ikke kjøpt had Jan whole week-DEF not bought ‘Had Jan all week not bought any milk?’

noe any

melk? milk

(17)

Hadde Jan ikke kjøpt noe melk HELE UKA? had Jan not bought any milk whole week-DEF ‘Had Jan not bought any milk the whole week?’

It would be inaccurate to claim that the contrastive prosody indicated by the caps in (17) cannot compensate for what the relative order of quantifier and negation operator tells us about scope relations. ‘Had Jan all week not bought any milk?’ and ‘Had Jan not bought any milk all week?’ express the same content: the universal quantifier scopes over the negation. Nevertheless, it is only the linear order in (16) that permits, and even favours, the use of Ja to convey the quantifier-first meaning in a response to the interrogative. In the case of (17), confirmation of the quantifier-first meaning requires the negative response word Nei, while to use the third response word, Jo, would be to assert that some milk was bought that week. The existence of an affirmative response word like Scandinavian Jo alongside the other affirmative response word Ja undermines Holmberg’s belief that a bare response word is legitimate just in case the proposition expressed by the speaker is the same as the proposition expressed in the interlocutor’s question. Ja instructs the hearer to locate a non-negative discourse antecedent and inferentially derive an affirmed, non-negative proposition, while Jo instructs the hearer to locate a negative antecedent part of the interlocutor’s syntactic construction and construe the speaker’s proposition as one which conforms to the formula ~(~p), equaling p. The major lexical difference between Ja and Jo is that their respective procedural meanings direct the hearer’s attention to distinct antecedents in the sentence that the response word addresses. Holmberg’s response words just fix the polarity of the propositional form expressed in a follow-up sentence, but that fails to explain how, in the Scandinavian languages, the two affirmative response words Ja and Jo differ semantically and communicate distinct propositions in (15). Fretheim (2014: 208) illustrates how the reference of the Norwegian demonstrative, abstract-entity pronoun det (‘that’) in Ben’s alternative answers in (18)

The Pragmatics of ‘Yes’ and ‘No’

191

depends on the propositional content assigned to Ben’s preceding response word, a content which is in turn resolved partly by the lexical meanings of the respective response words. The three indexicals Nei in (18) Ben1, Ja in (18) Ben2 and Jo in (18) Ben3 give Anne procedural information sufficient to delimit the intended antecedent (in Anne’s own interrogative) and to identify the distinct propositions targeted by Nei, Ja and Jo. (18) Anne: Syns du at pengene ikke skal deles likt? think you that money-DEF not shall share-PASS equally ‘Do you think that the money should not be shared equally?’ Ben1: Nei. Det syns jeg ikke. no that think I not ‘No. I don’t.’ (denial of Anne’s matrix-clause proposition) Ben2: Ja. Det syns jeg. yes that think I ‘Yes. That’s what I think.’ (confirmation of Anne’s matrix-clause proposition) Ben3: Jo. Det syns jeg. yes that think I ‘Yes. I think it should.’ (affirmation of the positive counterpart of Anne’s lower-clause proposition) It is particularly interesting to compare the disjunctive responses Ben2 and Ben3 in (18), two formally identical answers except for the speaker’s choice of affirmative response word, Ja in Ben2 and Jo in Ben3. But Ben’s alternative turns in (18) are not bare response words, since Ben adds something to the response word. What Anne crucially recognises as the reference of the demonstrative pronoun det (‘that’) that introduces the affirmative sentence ‘Det syns jeg’ in (18) Ben2 and (18) Ben3 and the negated sentence ‘Det syns jeg ikke’ in (18) Ben1, however, is wholly determined by Anne’s interpretation of the truth-theoretic content of the preceding response word. Also, while the sentence after the response word in (18) is dispensable, an omission of the response word would neutralise the truth-theoretic distinction between (18) Ben2 and (18) Ben3. Ben’s answer in (19), for example, cannot mean what (18) Ben3 with Jo means, but rather only what (18) Ben2 with Ja means.

192

Thorstein Fretheim

(19) Anne: Syns du at pengene ikke skal deles likt? think you that money-DEF not shall share-PASS equally ‘Do you think that the money should not be shared equally?’ Ben:

Det syns jeg. that think I ‘That’s what I think.’

That is because the indexical det here matches the negative ‘that’-complement in Anne’s complex interrogative, licensing this as its only accessible antecedent, and making the content of (18) Ben3 inaccessible in the absence of Jo as a response word. A consequence of Holmberg’s syntactic analysis of response word constructions is that the truth-theoretic proposition affirmed by the response word can be read directly off the syntax and semantics of the follow-up sentence when the response word does not exhaust the speaker’s turn. What I have shown, however, is that, because the lexical items Ja and Jo encode distinct procedural meanings, Ja in (18) Ben2 and Jo in (18) Ben3 target different segments of Anne’s interrogative and therefore communicate distinct propositions. In this vein, Anne identifies the respective propositions affirmed by Ja in (18) Ben2 and Jo in (18) Ben3 independently of the propositional content of the added declarative ‘Det syns jeg’. This suggests that it is the explicated content of Jo in (18) Ben3 that determines the truth-theoretic content of the abstract-entity pronoun det in the follow-up declarative there. Therefore, we need to assume the existence of a pre-semantic pragmatic process of saturation of the anaphora relation between Jo and its antecedent in Anne’s question. While the response word assumes a communicative role analogous to that of a discourse indexical, the pronoun in the follow-up sentence is an indexical whose referential resolution depends entirely on the truth-theoretic interpretation of the earlier response word. In other words, Ja in (18) Ben2 and Jo in (18) Ben3 place different constraints on the hearer’s selection of their linguistic antecedents and lead Anne to the inference that Ja affirms Anne’s matrix clause proposition and Jo affirms the positive counterpart of Anne’s negative embedded clause proposition.

4 An adverbial antecedent of a response word must express a truth condition So far, I have argued that certain English adverbs placed in front of a negative infinitival complement are the foci of Anne’s polar questions and therefore the targets of Ben’s answers, and demonstrated how an additional Scandinavian

The Pragmatics of ‘Yes’ and ‘No’

193

response word of the affirmative sort – Jo – directs the hearer to antecedentanaphor relations that would have been less accessible, or even totally inaccessible, if the language possessed just one affirmative response word. In this section, I will focus on the lexical properties of those adverbs which on Holmberg’s account license the use of an affirmative response word in an act of communicating a negative proposition. Trying to delimit the class of adverbs that allow a speaker to use Yes in what he describes as a truth-based affirmation of a negative proposition, Holmberg (2013a, 2013b) identifies a continuum ranging from those adverbs that definitely possess this property to those that definitely lack it. More specifically, he says that the former adverbs offer the producer of the answer an ability to employ Yes, or Swedish Ja, even if the polarity of the communicated proposition is negative. Adverbs which license a truth-based distribution of English and Swedish response words are low in Holmberg’s syntactic hierarchy, and adverbs that block such a distribution of response words are high in the syntactic tree. In this picture, the syntactic height of the adverb is meant to reflect its semantic scope: wide for high adverbs, narrower for low ones. Due to space restrictions, I will not present an exposition of Holmberg’s syntactic analysis here. After all, I have already demonstrated that there are serious flaws in his account of the distribution and meaning of response words. In my view, what regulates a speaker’s choice of response word in adjacency pairs of the type discussed in this paper can be defined in terms of the distinction in Anne’s interrogative between adverbs which encode a concept and contribute a truth condition, and adverbs which do not. Adverbs like deliberately, willfully, often, occasionally, once more, preferably all belong to the former category, while therefore, consequently, really, in fact, absolutely are all non-truthconditional. What Holmberg calls high adverbs with a wide scope are actually adverbs with no semantic scope at all, since an adverb or adverbial phrase cannot be targeted by a response word unless it expresses a truth condition. One adverb whose grammatical properties Holmberg (2013b) discusses is the Swedish modal adverb kanske (‘maybe’, ‘perhaps’), which he calls a high adverb. So, Ja is an unacceptable response to Anne’s interrogative in (20), where kanske precedes the Swedish negation marker inte. (20)

Anne:

Vill du kanske inte ha kaffe? will you perhaps not have coffee ‘Do you perhaps not want (any) coffee?’

Ben1:

Nej. (polarity-based use: ‘I don’t want coffee.’) ‘No.’

194

Thorstein Fretheim

Ben2: #Ja. (truth-based use of Ja is not possible here) #‘Yes.’ Ben3:

Jo. ‘Yes.’ (‘I’d like a cup of coffee.’)

The modal adverb kanske in Anne’s question in (20) has certain lexical properties worth considering in regard to Holmberg’s “adverb effect”. In the following, I shall explore how these properties manifest themselves in my native Norwegian. The Norwegian questions in (21) and (22), where kanskje (‘maybe’, ‘perhaps’) is sentence-initial and right-detached, respectively, may not be answered with Ja, in keeping with Holmberg’s observations about the Swedish kanske.4 (21)

Anne:

Ben: (22)

Anne:

Ben:

Kanskje du ikke vil ha maybe you not will have ‘Maybe you don’t want coffee?’

kaffe? coffee

#Ja. Vil du ikke ha kaffe, kanskje? will you not have coffee maybe ‘Don’t you want coffee, maybe?’ #Ja.

The right-detached position of kanskje in (22) offers the hearer the procedural information that the word should be processed and interpreted as a non-truthconditional epistemic downtoner particle, and this is equally true of kanskje when it appears in the same medial position as Swedish kanske in (20). In contrast, Anne’s sentence-initial placement of kanskje in (21) could in principle have directed Ben to a truth-conditional interpretation of this word as a weak modal operator of the epistemic sort. Even so, that assumption would not make Ben’s 4 Of the two questions performed by Anne in (21)–(22), only the one in (22) is an interrogative on syntactic grounds. Anne’s sentence in (21) is a declarative, since it is only declaratives that can start with an adverb. The fact that the subject NP appears immediately after kanskje in (21), as well as later on in (25) and (26), is an exception to the V2 rule, which demands that the second syntactic position in a Scandinavian matrix clause is occupied by a finite verb. The V2 order of Kanskje vil du ikke . . . (‘maybe will you not . . .’) is also grammatical but Kanskje du ikke vil . . . (‘maybe you not will . . .’) apparently sounds more idiomatic for many native speakers. The order of syntactic elements in (21) reflects the fact that the historical origin is a matrix clause plus an embedded complement clause: Det kan skje at . . . (‘It may happen that . . .’).

The Pragmatics of ‘Yes’ and ‘No’

195

answer Ja in (21) relevant. It would have rather indicated, contrary to people’s normal beliefs, that Ben is not in a position to give Anne a straight answer to the question of whether or not he wants a cup of coffee. A token of kanskje in the sentence-initial position is ambiguous; it is either a downtoner particle or a modal operator. In (21) the former alternative surely makes more sense than the latter, which is why Ja is misplaced there, since there is no accessible positivepolarity antecedent with which Ja can be matched. (22) differs from (21) in that the right-detached placement of kanskje constrains the interpretation of this item by eliminating the truth-conditional alternative altogether. So, its syntactic position does not license the response Ja in any conceivable context.5 The set of data in (23)–(25) shows that in 3rd person constructions kanskje is interpreted differently than it is in the 2nd person constructions presented above. The sentence-initial kanskje in (25) may be interpreted as a truthconditional modal operator. So, Ja can in this case affirm the possibility that some of them may not be interested in coffee. On the contrary, the medial kanskje in (23) and the right-detached kanskje in (24) are non-truth-conditional particles that may reflect Anne’s earlier impulse to brew coffee for everyone present without having enquired in advance whether everyone wants coffee. (23)

Anne:

Ben: (24)

Anne:

Ben: (25)

Vil noen av dem kanskje ikke ha will some of them maybe not have ‘Do some of them maybe not want coffee?’

kaffe? coffee

#Ja. Vil noen av dem ikke ha kaffe, will some of them not have coffee ‘Do some of them not want coffee, maybe?’

kanskje? maybe

#Ja.

Anne:

Kanskje noen av dem ikke vil ha kaffe? maybe some of them not will have coffee ‘Maybe the fact is that some of them don’t want coffee?’

Ben:

Ja. ‘Yes.’

5 The relevance of the Swedish particle kanske in (20) and the equivalent Norwegian kanskje in (21) and (22) stems above all from the fact that it gives the hearer the procedural information that the speech act is a request for confirmation of Anne’s negative proposition; it cannot be interpreted as an act of offering the hearer some coffee (‘Won’t you have a cup of coffee?’), which is arguably the most likely interpretation of Anne’s question if kanske/kanskje is left out.

196

Thorstein Fretheim

Kanskje is no truth-conditional item in (23) and (24). The information focus of Anne’s interrogatives in (23) and (24) is the negative polarity, typically signalled by a focal accent on the infinitive ha (‘have’) and a post-focal prosodic handling of the noun kaffe (‘coffee’), which indicates the discourse-given status of the general concept encoded by that noun. Ja is unavailable in (23) and (24), because Anne’s utterance contains no linguistic material that gives Ben access to a positive-polarity proposition which he could then affirm with an utterance of Ja. Ja in (25), on the other hand, affirms the modally modified proposition expressed by Anne, as, in this case, Anne asks Ben if he believes it may be the case that some of them do not want to drink coffee, and Ben seconds that thought. The existential quantifier noen (‘some’) scopes over the negation marker in (23) and (24), while in (25) it is kanskje (‘maybe’) that scopes over the existential quantifier, which in turn scopes over the negation. Although the sentence-initial kanskje in (21) could not get a truth-conditional interpretation as modal operator on the grounds that the 2nd person subject pronoun in Anne’s question makes it pragmatically weird, sentence-initial kanskje in (26) is more likely to be a modal operator than a non-truth-conditional particle for pragmatic reasons, in spite of the 2nd person subject. Here, Ben’s answer Ja shows that he indeed construes kanskje as a modal operator. (26) Anne: Kanskje du ikke tåler kaffe så sent på dagen? maybe you not tolerate coffee so late on day-DEF ‘Maybe you don’t tolerate coffee that late in the day?’ Ben:

Ja. (confirmation of the modally modified proposition) ‘Yes.’

This acceptability of Ja is markedly reduced in (27), where kanskje is in the medial position. (27)

Anne:

Du tåler kanskje ikke kaffe så sent på you tolerate maybe not coffee so late on ‘You maybe don’t tolerate coffee that late in the day?’

Ben1: #Ja. (no accessible non-negative antecedent) Ben2:

Nei. (confirmation of the negative proposition)

dagen? day-DEF

The Pragmatics of ‘Yes’ and ‘No’

197

5 English not vs. n’t The claim that the English free form not and the contracted form n’t have distinct logical properties, the former being a low adverb and the latter a high adverb, is central in Holmberg’s syntax-based account of what regulates the distribution of Yes and No in an English speaker’s affirmation of a negative proposition (Holmberg 2013a). In this account, Yes may therefore be used in an act of affirming a negative proposition expressed by means of a sentence whose negation marker is not, whereas the alternative negation marker n’t forces a “high negation reading”, which precludes the use of Yes in an act of affirming a negative proposition. When added to a finite verb of an English polar interrogative which happens to be sentence-initial, the contracted form n’t mandatorily cliticises to it. In (28) below, because the cliticised negator n’t precedes once more, Ben is given the procedural information that, in Anne’s interrogative, the negation marker takes scope over once more. (28)

Anne:

Didn’t he once more return the books on time?

Ben:

Yes. (Scandinavian Jo)

In (6), however, reproduced below as (29) for convenience, we understand the earlier adverb once more to scope over not. So, the Yes in (28) and the Yes in (29) have no truth-conditional interpretation in common. (29)

Anne:

Did he once more not return the books on time?

Ben:

Yes. (Scandinavian Ja)

The distinct readings of the questions in (28) and (29) cannot be pinned on a difference between inherent lexical properties of n’t compared to not. Rather, Ben’s pragmatic processing is guided by the surface-syntactic difference in the linear order of the two negation markers n’t and not relative to the position of once more. It is useful to compare the pair (28)–(29) to (30)–(31), where the speech act is a request for confirmation but the sentence type is declarative. Here, once more precedes the negation marker in both talk exchanges, making n’t and not semantically equivalent in (30) and (31), and Ben’s Yes confirms that it happened once more in both (30) and (31).6 6 It should be noted that it is marginally possible that Ben’s Yes in (30)–(31) is meant as a denial of the thought that the man’s failure to return the books on time happened once again, in which case Yes would be equal to Scandinavian Jo, but this is licensed by a different antecedent than in (28), hence a different content.

198 (30)

(31)

Thorstein Fretheim

Anne:

He once more didn’t return the books on time, then?

Ben:

Yes. (Scandinavian Ja)

Anne:

He once more did not return the books on time, then?

Ben:

Yes. (Scandinavian Ja)

6 Concluding remarks An anaphoric response word targets the information focus of the conversational partner’s speech act. So, when a polar question is performed, this focus is what the person asking the question wants to be informed about, most typically whether the addressee believes the proposition expressed by the producer of the question to be true or false. I have examined English, Norwegian and Swedish data where the first speaker’s predicate phrase consists of an adverb followed by a negative complement, a linear order that places the adverb outside the scope of negation, and have shown that a response word affirms or denies a thought derived from the pragmatic processing of linguistic material that constitutes the interlocutor’s information focus, which may well be a subclausal syntactic unit, like an adverb. In this way, I believe to have refuted Holmberg’s claim that the adverbs in the data examined are open for affirmation of a negative proposition by means of an affirmative response word like English Yes or Scandinavian Ja. The matching positive polarity in antecedent and anaphor is rather between the focused adverb and the response word, while the negative complement of the adverb is outside the first speaker’s focus. In this respect, whatever semantic representation may be associated with this complement, it is something incapable of being affirmed or denied – unless, of course, the response word has semantic characteristics similar to Scandinavian Jo, which may be used to reject a presupposed proposition with negative polarity. Response words are used for communicating a negative-polarity or positivepolarity proposition, and a communicated proposition is what relevance theorists call an explicature. In closing, I would like to briefly discuss an issue that has to be addressed if one hopes to reconcile the relevance-theoretic notion of explicature as currently conceived with my analysis of response words as indexicals. A relevance-theoretic explicature is defined as a communicated proposition inferentially developed from one of the incomplete conceptual representations (logical forms) encoded by an utterance (cf. Sperber and Wilson 1995; Carston 2002). As we have seen, a response word communicates a proposition in an

The Pragmatics of ‘Yes’ and ‘No’

199

explicit way, but one may legitimately ask what the conceptual representation of a response word could possibly be. As it is hardly meaningful to say that a response word encodes a logical form, how can we claim that such words are enriched in context by means of some inferred conceptual meaning added to a decoded logical form? My proposal is that response words encode procedural information which directs the addressee’s attention to conceptual information expressed in a discourse antecedent, information that is essential for the resolution of their context-dependent reference. But procedural information is information about how to manipulate truth-conditionally incomplete representations encoded by utterances (Wilson and Sperber 1993: 2) and a response word does not encode any information about truth conditions. It is rather the input to what comes out of the hearer’s pragmatic processing as a mentally represented proposition that is a wholly inferred result of this rather special type of indexical truthconditional saturation. Against this backdrop, it is certainly tempting to make a proposal that Relevance Theory loosens the concept of explicature so as to include in it the explicitly communicated propositional content of response words, even though this content is derived pragmatically in a way that does not involve the development of a logical form. If this proposal is on the right track, a response word would be recognised as an indexical in need of mandatory, bottom-up saturation, and the current rule that an explicature is necessarily a pragmatic development of an encoded logical form would be lifted. All in all, there are many kinds of indexical expressions in the lexicon of a language whose truth-theoretic interpretation must be based on information derived inferentially from a sentence-external source. A personal pronoun like the English she or her tells the hearer to identify a unique singular referent of female gender. In most cases, a token of one of these pronouns will cause the hearer to look for a nearby discourse antecedent that denotes the general concept FEMALE and to use this selected antecedent as a source of information that constrains the hearer’s search for the referent of the pronoun. In much the same way that a personal pronoun offers vital procedural information about how to identify its intended referent, a response word is a kind of indexical that constrains its truth-conditional content by virtue of the fact that the lexical meaning it encodes enables the hearer to establish the intended relation between the response word as anaphor and an antecedent located in the hearer’s most recent utterance.

200

Thorstein Fretheim

References Blakemore, Diane. 1987. Semantic constraints on relevance. Oxford: Blackwell. Carston, Robyn. 2002. Thoughts and utterances: The pragmatics of explicit communication. Oxford: Blackwell. Escandell-Vidal, Victoria, Manuel Leonetti & Aoife Ahern (eds.). 2011. Procedural meaning: Problems and perspectives. Bingley: Emerald. Fretheim, Thorstein. 2014. Response words are anaphors. In Kate Burridge & Réka Benczes (eds.), Wrestling with words and meaning: Essays in honour of Keith Allan, 198–218. Clayton, Victoria: Monash University Publishing. Gundel, Jeanette & Thorstein Fretheim. 2004. Topic and focus. In Lawrence R. Horn & Gregory Ward (eds.), The handbook of pragmatics, 175–196. Oxford: Blackwell. Holmberg, Anders. 2012. Answering yes/no-questions in English and other languages. Krishnaswamy Endowment Lecture at the EFL University in Hyderabad, 31 January. Holmberg, Anders. 2013a. The syntax of answers to polar questions in English and Swedish. Lingua 28. 31–50. Holmberg, Anders. 2013b. Adverbs and the scope of negation in questions and answers. Paper presented at the 25th Scandinavian Conference of Linguistics, Reykjavík, 13–15 May. Recanati, François. 2010. Truth-conditional pragmatics. Oxford: Oxford University Press. Sperber, Dan & Deirdre Wilson. 1995. Relevance: Communication and cognition, 2nd edn. Oxford: Blackwell. Wilson, Deirdre & Dan Sperber. 1993 Linguistic form and relevance. Lingua 93. 1–25.

Michael Chiou

11 Do not say that for which you lack adequate evidence: On the interpretation of future tense in Modern Greek Abstract: Even though the future is unknown, speakers can use the future tense to make predictions or simple statements of fact about the future. In Modern Greek, the future tense is formed with the particle tha (‘θα’ = ‘will’) followed by the [+perfective], [-past] verb form. Current research has shown that: a) the particle tha is an epistemic modal operator with present (now) perspective and b) the Greek [+perfective], [-past] verb form cannot function as an independent tense and is best treated as a non-deictic time. Moreover, it has been argued that, at the level of sentence meaning, future constructions have the semantics of inquisitive assertions and convey epistemic possibility (p/~p); yet, what is actually communicated is an informative statement conveying an apparently homogenous epistemic state, namely, a future prospective reading (p only). In this paper, I address the question of how a non-homogenous modal interpretation at the level of sentence meaning turns out to be a future prospective reading at the level of what is communicated. Building upon the theoretical findings so far, I argue that the preferred prospective reading arises as a generalised conversational implicature, in the spirit of the Gricean theory. In the proposed account, the future prospective reading is implicated by virtue of background assumptions about language use, interacting closely with the form of what has been said.

1 Setting the scene It is well known that futurity and modality are interrelated in such a way across languages that there is an ongoing debate concerning the status of future tense as a marker of tense or of modality (see, among others, Jespersen 1924; Prior 1957, 1967; Haegeman 1983; Thomason 1984; Comrie 1985; Palmer 1987; Enç 1996; Sarkar 1998; Ludlow 1999; Copley 2002; Condoravdi 2002; Squartini 2004; Jaszczolt 2006, 2009, 2011, 2013; Kissine 2008; Mari 2009; Giannakidou and Mari

Michael Chiou, Metropolitan College, Athens, Greece DOI 10.1515/9781501505089-011

202

Michael Chiou

2012, 2013a, 2013b). Modern Greek is no different in this respect and, therefore, the question of whether the so-called future tense has a temporal or a modal basis is still pretty much under discussion (see, for example, Condoravdi 2003; Giannakidou 2009, 2012, 2014; Giannakidou and Mari 2012, 2013a, 2013b; Tsangalidis 1999, 2001). Modern Greek forms a periphrastic future tense by employing the particle tha (‘will’), usually referred to as the future marker (Philippaki-Warburton 1994; Rivero 1994) followed by the [+/-perfective], [-past] verb form.1 Nevertheless, as will be shown, only the combination of tha and the [+perfective], [-past] provides a “pure” future interpretation, while tha combined with the [-perfective], [-past] systematically conveys epistemic non-future modal readings. What is more, the future tense seems to also express what I will call future prospective 2 readings. All expressions that encode futurity place an event e at a time te, which follows the time of the utterance tu, so that {te > tu}. In this vein, a future prospective reading can be glossed as follows: a proposition p expressed at tu will be true at te, that is, when the event described in p will actually happen. More specifically, given a set of possible worlds in the future, p will be identified with those possible worlds that will become actual worlds in the future. For instance, consider sentences (1a) and (1b): (1)

a.

John might go to school.

b.

John will go to school.

In both (1a) and (1b) we get a future-time interpretation in the sense that the event e described in the proposition p will take place after the utterance time (e > Ut ). This much is communicated by both the modal sentence in (1a) and the future tense in (1b). The crucial difference here is that while sentence (1a) is open for both p and ¬p (i.e. it also allows for those possible worlds in the future in which John’s going to school will not take place), sentence (1b) communicates that John’s going to school will be actually true in the future and is not just a possibility (i.e. p only). On this basis, it can be argued that even though the future is epistemically and metaphysically unknown, speakers can communicatively express certainty

1 As we will also see later on, in traditional Modern Greek grammar, the combination [+perfective], [-past] is termed “simple future” while the combination [-perfective], [-past] is termed “future continuous”. 2 The term is adapted from Escandell-Vidal (2014).

Do not say that for which you lack adequate evidence

203

about future events through the use of the future prospective interpretation which, in the case of Modern Greek, is associated with the use of tha with [+perfective], [-past]. Future prospective readings can include predictions, commissives (promises and threats), and directives (suggestions and commands) (for a similar distinction, see Escandell-Vidal 2014), and can also be identified with future tense. In current literature (cf. Giannakidou 2009, 2012, 2014; Giannakidou and Mari 2012, 2013a, 2013b; Tsangalidis 1999), it has been argued that the particle tha is not a typical future tense marker. In particular, Giannakidou and Mari (2012, 2013a, 2013b) propose, among other things, that the particle tha is an epistemic modal operator with present (now) perspective. Moreover, since the Greek [+perfective], [-past] verb form cannot function as an independent tense (Holton et al. 1997: 220), it is treated as non-deictic time. If this line of analysis is correct, and there is nothing in the semantics of tha constructions that functions as a future tense marker, we need an account for the future prospective reading conveyed by the combination of tha and the [+perfective], [-past]. In this paper, I argue that future tense in Modern Greek has a modal semantic base conveying epistemic modality, in the spirit of Giannakidou (2009, 2012, 2013b) and Giannakidou and Mari (2012, 2013a, 2013b), and that the preferred future prospective reading is a pragmatic development of the semantic modal base. In other words, future prospective readings in Modern Greek are not part of what is expressed literally in a sentence, but rather fall within what is conversationally implicated by an utterance. This line of analysis suggests that they are calculated upon what is said and are determined by underlying principles of efficient and effective language use. Against this backdrop, the structure of the paper is as follows. In section 2, I focus on the typology and properties of tha constructions. In section 3, I outline the current literature focusing on the study of future constructions and future-time reference, before turning to present the Gricean pragmatic theory of communication in section 4, and further develop the pragmatic analysis of future tense in Modern Greek in section 5.

2 Facts on tha-constructions: Epistemic and future uses 2.1 Tense, aspect and verb morphology In Modern Greek the verb form is inflected for the grammatical categories of tense and aspect. In this vein, there is a morphological distinction between past and non-past in terms of tense, and perfective and imperfective in terms of

204

Michael Chiou

aspect. The importance of aspect in Modern Greek is supported by the fact that all tenses, moods and voices are marked for perfective or imperfective aspect (cf. Joseph 1983; Holton et al. 1997; Tsangalidis 1999); the imperfective aspect is used to mark a progressive, habitual or repeated action, whereas the perfective one marks an action which is perceived as a completed whole (Xidopoulos 1996; Holton et al. 1997: 220). The combinations of tense and aspect give us four morphologically distinct verb forms which are illustrated in the following examples: (2)

graf -o [-perfective], [-past] write IMP-1SG.NON-PAST ‘I am writing (right now).’ ‘I write (generally).’

(4)

e-graf-a [-perfective], [+past] PAST-write-IMP.1SG.PAST ‘I used to write.’ ‘I was writing.’

(3) *grap -s -o [+perfective], [-past] write-PERF.1SG.NON-PAST (no English equivalent)

(5)

e-grap-s-a [+perfective], [+past] PAST-write-PERF.1SG.PAST ‘I wrote.’

As one can see, there are two past forms, namely, the imperfective past (IP) in (4) and the perfective past (PP) in (5), and two non-past forms, namely, the imperfective non-past (INP) in (2) and the perfective non-past (PNP) in (3). Traditional grammar treats the verb form in (2) as present, yet Giannakidou (2009) argues that non-past verb forms are not equivalent to the present. The verb form in (3), i.e. the perfective non-past, is not possible without the presence of certain particles (such as na, as, tha and an) as illustrated in examples (6)–(8) below. This is the reason why it is typically referred to as a “dependent” form (Holton et al. 1997: 220). (6) *Ο Νikos grapsi the Nikos write PNP ‘Nikos write a letter.’

ena a

(7)

As grapsi o Νikos Let write PNP the Nikos ‘Let Nikos write a letter.’

(8)

Ο Νikos tha grapsi the Nikos will write PNP ‘Nikos will write a letter.’

grama. letter

ena a

ena a

grama. letter

grama. letter

Do not say that for which you lack adequate evidence

205

In this paper, I am going to focus on the perfective non-past and imperfective non-past verb forms since these two forms are used in the formation of the traditionally called future constructions.

2.2 The typology of future tense As already mentioned, in Modern Greek, future tense is not morphologically marked in the verb form. According to traditional grammar, it is formed with the particle tha followed by the perfective non-past (as in 9) or the imperfective non-past verb forms (as in 10). (9)

(10)

Ο Νikos tha petaksi the Nikos will fly PNP ‘Nikos will fly to London.’

gia for

to the

Londino. London

Ο Νikos tha petai gia to Londino the Nikos will fly INP for the London ‘Nikos will be flying to London tomorrow.’

avrio. tomorrow

The combination of thα with perfective non-past verb forms, as in (9), is used to “express an action which will take place and be completed at a future point in time” (Holton et al. 1997: 227). In traditional terms, this type of future is dubbed the simple future. Alternatively, when thα is combined with the imperfective nonpast verb form, as in (10), “it describes an action which will be taking place in the future either as a habitual event, or as a continuous, progressive one” (Holton et al. 1997: 226), forming the future continuous. What has to be noted though, is that, in examples like (10), future-time reference seems to arise from the use of the time adverbial avrio (‘tomorrow’) rather than the tha construction. If we remove the adverb, and in the absence of any specific context, the future interpretation does not survive, as in (11): (11)

Ο Νikos tha petai gia to Londino. the Nikos will fly INP for the London ‘Nikos will be flying to London.’ (now)

In such cases, the combination of tha with imperfective non-past verb forms is mostly interpreted as epistemic present (cf. Giannakidou 2012), expressing a highly strong possibility and an inference about the state of affairs at the time of utterance. In these contexts, tha constructions do not have the force of a

206

Michael Chiou

pure future tense, but are rather glossed like ‘most probably/possibly’ and, more importantly, make reference to the time of utterance. Giannakidou (2012) and Giannakidou and Mari (2012) also argue that, when combined with imperfective non-past verb forms, tha exhibits evidential behaviour and is very similar to the evidential modal prepi (‘must’). In this respect, tha with an imperfective nonpast verb form is co-operatively used when the speaker lacks direct evidence about the situation at hand. For instance, in example (11), the speaker communicates that she does not have direct evidence about the truth of the proposition at hand, but is just making an inference based on indirect evidence. This epistemic present reading can be further reinforced by the use of high probability adverbs such as malon (‘probably’) (cf. Holton et al. 1997; Giannakidou and Mari 2012): (12)

Ο Νikos malon tha petai gia to the Nikos probably will fly INP for the ‘Nikos will be probably flying to London.’ (now)

Londino. London

Nevertheless, the epistemic present interpretation is not the preferred one when the verb is in the 1st or 2nd person. Consider the following examples: (13)

?Tha petao gia to Londino Will fly INP, 1SIN for the London ?‘I will be flying to London now.’

tora. now

(14)

?Tha petas gia to Londino Will fly INP, 2SIN for the London ?‘You will be flying to London now.’

tora. now

(15)

Tha petas gia to Londino avrio. Will fly INP, 2SIN for the London tomorrow ‘You will be flying to London tomorrow.’

The addition of an adverb referring to the time of utterance, such as tora (‘now’), renders examples (13) and (14) odd, if not unacceptable. Yet, there is no problem at all with the future time adverb avrio (‘tomorrow’) as in (15). The fact that the 1st and 2nd person are more likely to express futurity than the third person, which favours epistemic present readings, is also observed in Tsangalidis (1999: 212), following Heine (1995). Turning now to the tha with perfective non-past combination, it stereotypically receives a future prospective interpretation. In other words, the meaning

Do not say that for which you lack adequate evidence

207

conveyed by the use of tha with imperfective non-past verb forms is that an event e will certainly occur at a time t and t is temporally located after the utterance time (t > Ut). By contrast, the epistemic reading in this case is very weak, yet not impossible in certain contexts (cf. Giannakidou 2012). (16)

O Nikos tha petaksi gia to Londino. The Nikos will fly PNP for the London ‘Nikos will fly to London.’ (in the future)

The preferred future prospective interpretation of tha and perfective non-past combinations is not affected by the presence or absence of an adverb of time, as it is the case with tha and imperfective non-past constructions. The time adverbial merely anchors the event described by the verb at a specific point in the future, making the sentence more informative but not more predictive, as in (17): (17) O Nikos tha petaksi gia to Londino avrio / se tris meres. The Nikos will fly PNP for the London tomorrow / in three days. ‘Nikos will fly to London tomorrow/in three days. . .’ The only case in which the tha and perfective non-past combination does not convey a temporal (future) interpretation is when it is used to describe habitual timeless actions or indicate obligation. Consider examples (18) and (19) respectively (from Holton et al. 1997: 227): (18)

Kathe proi tha sikothi, tha pji to kafedaki tu, tha djavasi tin efimerida tu ke kata tis 8.30 tha figi gia to grafio tu. ‘Every morning he will get up, drink his coffee, read his newspaper and at approximately 8.30 he will leave for the office.’

(19)

Oxi tha mu to epistrepsis afto to grama amesos! ‘No, you will [must] give this letter back to me immediately!’

In examples like (17), adverbs of time like tora (‘now’) or avrio (‘tomorrow’), which are always possible with tha and perfective non-past constructions, are not acceptable, which shows that a temporal reading cannot be intended.

2.3 Other ways of expressing futurity As Escandell-Vidal notes, “futurity can be expressed not only by the grammatically marked future tense, but also by other, competing expressions” (2010: 219).

208

Michael Chiou

Modern Greek is not an exception here either, since futurity can be conveyed by a range of expressions3 apart from the future tense i.e. the combination of tha and [+perfective], [-past]. One of the most typical ways is the use of the imperfective non-past verb form followed by an adverb of time which indicates future time. (20)

i Eleni erxete avrio. the Eleni comes-INP tomorrow ‘Helen is coming tomorrow.

The use of the unmodalised sentence with the present tense encodes a rather strong assertion that the event will take place at the time indicated by the adverb. Sentences like these are used to denote scheduled events. It has to be noted though that, in cases like this, future-time reference is clearly assigned by and is completely dependent on the linguistic context, and more specifically the temporal adverb. Therefore, when this adverb is not present, the future-time reference evaporates. Subjunctives and optatives can also be used to express futurity, even though this is not their primary meaning: (21)

i Eleni theli na pai the Eleni wants SUBJ go to ‘Helen wants to go to Cambridge.’

sto the

(22)

As erthi i Eleni opote Let-OPT come the Eleni whenever ‘Let Helen come whenever she wants.’

Cambridge. Cambridge

theli. wants

Finally, the modal adverb isos (‘maybe’) can refer to the future by expressing weak possibility. Consider the example: (23)

Ο Νikos isos grapsi the Nikos maybe write PNP ‘Nikos might write a letter.’

ena a

grama. letter

Here, the forward shifting is probably due to the presence of the perfective nonpast.

3 Futurity can be also expressed by the present tense, the subjunctive, modal expressions and conditionals, among others.

Do not say that for which you lack adequate evidence

209

To sum up, it has been shown that tha is neither a necessary nor a sufficient condition of future-time reference.4 On the contrary, tha sentences systematically receive a range of interpretations which are typically associated with epistemic modality,5 and can also refer to the time of utterance or the past. Moreover, we can have future-time reference without a tha construction. As a result, it seems natural to argue that describing tha as a purely temporal operator, i.e. a future tense marker, cannot account for all its uses (see also Giannakidou and Mari 2012, 2013a, 2013b; Roussou and Tsangalidis 2010). Future-time reference is available in a wide range of expressions and seems to be the result of the division of labour between tense, aspect, and lexical semantics. By contrast, future prospective readings are restricted to the combination of tha followed by the perfective non-past and are essentially context-free, as they are assigned neither by the semantics of the construction nor by the linguistic context.

3 Future in Modern Greek: The literature Even though the issue of time, and future-time reference in particular, has been in the focus of research in linguistics and philosophy of language for a substantially long time, it has received relatively little attention in the literature on Modern Greek. At this point I shall discuss the work of Tsangalidis (1999), Giannakidou (2012, 2013, 2013b) and Giannakidou and Mari (2012, 2013a) which are important for the present analysis.

3.1 The category future: Tsangalidis (1999) Within the framework of a comparative study between will constructions in English and tha constructions in Modern Greek, as means by which future-time reference is expressed, Tsangalidis (1999) proposes a theory which permits future to be defined independently of the core categories of tense and modality. More precisely, the analysis in Tsangalidis (1999) is built upon the notion of grammatical morpheme or gram-type introduced in Dahl (1985) and Bybee and Dahl (1989). The term gram-type refers to cross-linguistic categories which are 4 Comrie (1985) argues the same for the use of the auxiliary will in English. 5 Apart from tha with INP constructions, which receive epistemic readings, the combination of tha with the PP also gives rise to epistemic readings referring to an event prior to the time of utterance, while the combination of tha with the IP is a conditional construction, which mainly gives rise to counterfactual readings, i.e. alternative worlds in the past (cf. Iatridou 2000).

210

Michael Chiou

instantiated in each language by a specific type of gram. Gram, in its turn, refers to the notion of grammatical morpheme. As Bybee, Perkins and Pagliuca discuss, “grammatical morphemes are closed-class elements whose class membership is determined by some unique grammatical behaviour, such as position of occurrence, co-occurrence restrictions, or other distinctive interactions with other linguistic elements” (1994: 2). Moreover, these grammatical morphemes, or grams, may appear in various types in terms of their form, ranging from affixes up to complex constructions, and are “identifiable by their semantic foci” (Bybee and Dahl 1989: 52). In this sense, the notion of gram-type cuts across traditional categories, such as tense, aspect, mood and so on, which, in this picture, are considered rather vague and undetermined. According to Tsangalidis (1999), tha is not a typical future tense marker, neither is it a modal operator; tha is best described as an instance of the “future gram-type”. The semantic content of future gram-types indicates that “the speaker predicts a situation will occur subsequent to the speech event” (Bybee and Dahl 1989: 55). As a result, the category of future is approached in an alternative way, thus rendering the debate over its temporal or modal status irrelevant. As Tsangalidis himself observes, “future-grams [. . .] are seen as autonomous entities [. . .] and their definition in terms of polysemic associations of diachronically related senses does not exclude their use as markers of temporal, aspectual and modal notions” (1999: 255). The main proposal in Tsangalidis (1999) is that the particle tha does not qualify as either a prototypical modal or as a pure future tense marker. Accordingly, a future-gram type status of tha is proposed, which allows for a description independent of membership to either a temporal or a modal category, in an analysis that is assumed to have “advantages over any attempt to decide on the centrality of either tense or modality in the semantics of future markers or any attempt to recognize distinct underlying elements which only happen to be homophonous in these languages” (Tsangalidis 1999: xi). Concerning the interpretation of tha constructions, Tsangalidis concludes that: a) tha combined with perfective past gives pure epistemic past readings, b) tha combined with the perfective non-past gives pure non-epistemic future readings, and c) all other combinations are open to both temporal and modal readings. In addition, it is also put forward that “the default interpretation [of tha] is dependent on the form of the lexical verb – and crucially on its aspectual and temporal characteristics” (Tsangalidis 1999: 253). Beginning with the dependent form, Tsangalidis assumes the now widely accepted view that it is a typical perfective non-past form and not subjunctive.6 6 The view that dependent or ‘grapso’ forms are not subjunctives has also been put forward in Veloudis and Philippaki-Warburton (1983) and Philippaki-Warburton (1992).

Do not say that for which you lack adequate evidence

211

What is more, following literature on aspect,7 according to which, when combined with non-past, perfectives generate a contradiction, since perfective events cannot occur at the same time as perfective speech events and therefore cannot practically refer to the utterance time, Tsangalidis (1999) explains why the perfective non-past can only receive pure future or habitual interpretations. Finally, in this picture, the fact that perfective non-past forms are non-specific and nonpast can account for their dependent status. Coming now to tha with imperfective non-past constructions, Tsangalidis (1999) contends that the epistemic present reading should not be considered the default one, since epistemically modalised statements about the speaker and the addressee do not normally hold. As a result, it is proposed that the combination of tha with imperfective non-past is underspecified for tense, which effectively means that tha with imperfective non-past “does not force a future time reference as such – but rather ‘prediction’” (Tsangalidis 1999: 212). As it is explained, this prediction normally refers to the future (hence the potential future time reference), but in certain contexts, as in the case of statives, progressives and imperfectives, this prediction can equally refer to the present. In these latter cases, the contribution of the particle tha is related to epistemic modality rather than futurity. This suggests that there is an ambiguity between future and epistemic present time interpretations, which, for Tsangalidis, can be best accounted for by the analysis of tha as a “future gram-type.”

3.2 Tha as an epistemic modal More recently, Giannakidou (2012, 2013, 2014) and Giannakidou and Mari (2012, 2013a, 2013b) put forward an alternative analysis for the particle tha, mainly based on Giannakidou’s (2009) seminal work. The main proposal is that tha is not a future tense marker but a modality operator; a proposal that is based on the non-future readings of tha when combined mainly with the imperfective non-past and imperfective past forms, but is also related to the interaction of tha with modal adverbials (Giannakidou 2012). Giannakidou (2009) observes that non-past forms in Modern Greek are not equivalent to the present. More specifically, the imperfective non-past is not a present tense, but is rather “used for habitual and generic statements, as well as to denote progressive and ongoing events” (Giannakidou 2009: 1896). Now, in cases where the imperfective is used for the progressive, it denotes the function PROG. By contrast, when used generically, the imperfective contributes 7 See Dahl (1985), Comrie (1976, 1985) and Smith (1991) among others.

212

Michael Chiou

GEN. The output is an interval during which generic quantification takes place (Giannakidou 2009). It is therefore assumed that the imperfective non-past does not make reference to a specific time. In this approach, the perfective non-past cannot function as a present tense either. Giannakidou (2009, 2012, 2014) suggests that the dependent nature of the perfective non-past is attributed to its inability to make reference to the utterance time. As proposed, the perfective non-past contains a time interval (t, ∞) whose left boundary t is a non-deictic variable. Following this, non-past in Modern Greek has the following semantic representation, introduced in Abusch (2004)8 (Giannakidou 2009: 1899): (24) [non-past] = λP λt P((t, ∞)) For Abusch (2004), since t is a non-deictic variable, it must be bound by n (PRES, utterance time) in order to be licensed. In a real present tense this variable receives its n ‘now’ value (referring to the utterance time) from a PRES feature. Based on this, the Greek perfective non-past is similar to Abusch’s WOLL, even though “the Greek non-past contains no higher temporal information, that is, no PRES and it will thus require some other element to supply n” (Giannakidou 2009: 1899). As a result, the perfective non-past is treated as a temporal polarity item which will need a particular licensing context in order to receive the missing n or PRES feature and, therefore, acquire a time value – unlike ‘will’ where PRES is triggered. In the case of the imperfective non-past, which conveys either a generic or a progressive interpretation, it is the time adverbials that provide the relevant time interval which replaces (t, ∞). By way of illustration consider the following example: (25)

O Jianis grafi sixna the John writes often ‘John writes often.’ OFTENt [t ∈ C ∧ t ⊆ i: write (j, t) ∧ i = (t, ∞)]

“In this sense, the problematic interval (t, ∞) is replaced by the generic interval i [. . .] and the result is a statement with no direct reference to the utterance time” (Giannakidou 2009: 1900). In the same fashion, adverbials or temporal 8 According to Abusch,“in the substitution operator, t is a bound variable that corresponds to the tense argument of will. For a top-level occurrence of will, the effect is to substitute (n, ∞) for n” (2004: 39).

Do not say that for which you lack adequate evidence

213

expressions like tora (‘now’) or olo to proi (‘all morning’) provide a time interval which binds the variable t and therefore gives the progressive readings. The explanation described above, however does not hold for the perfective non-past. Consider the example: (26) *O Jianis grapsi sixna / tora / olo to proi the John write PNP often / now / the whole morning ‘John writes often / now / all morning.’ The addition of adverbials or temporal expressions in (26) does not improve the illicit perfective non-past. Therefore, the need for n to be introduced still remains. Giannakidou (2009, 2012) suggests that the perfective non-past receives n from particles such as tha. This motivates the introduction of a Now-TP into the syntactic structure with tha being its head (Now-T), as illustrated below: (27)

Now-TP: pe [win (j,e) ∧ e ⊆ (n, ∞) ]

Now-T: tha: n TP: λt∃e [win (j, e) ∧ e ⊆ (t, ∞) ] “kerδisi o Janis” It is therefore suggested that the particle tha semantically functions as the present tense which is missing from the perfective non-past verb form, providing in this way a satisfying explanation for the dependency of the perfective nonpast to particles such as tha.9 In a more up to date version, which specifically deals with the “predictive future” (i.e. tha + PNP), Giannakidou and Mari (2013b) propose that the assessment made by the speaker contains a truth-conditional component that carves out metaphysical branches into reasonable and non-reasonable futures, with the reasonable ones (cf. Mari 2013) being such that p is true in them.10 With the future being epistemically unknown, and therefore non-veridical, this analysis treats future sentences like inquisitive assertions. Still, the domain of possibilities 9 According to Giannakidou (2009), apart from the so-called future particle tha, the semantic function of the present tense can also given by the subjunctive na, the optative as and conditional an. 10 Following Mari (2013), a reasonable future determined at time t is such that the rules that hold at t are maintained. Reasonable futures are such that also habits and behaviours do not change there.

214

Michael Chiou

is restricted, since speakers use their knowledge in order to determine the metaphysical alternatives; in other words, speakers can calculate what would count as a course of events such that p will be true. According to Giannakidou and Mari, this is a ‘positive bias’ towards p and it is exactly this bias that is responsible for the strength that predictions typically have, even though it is also made clear that “a bias towards p for the future does not imply commitment of the speaker to p, as is the case of veridical unmodalized past and present assertions” (2013b: 116).

3.3 The prospective readings puzzle In sum, both accounts presented above agree that the particle tha is not a pure future tense marker. Yet, in contrast to Tsangalidis (1999), who suggests that tha does not qualify as a modal either, Giannakidou (2012) and Giannakidou and Mari (2012, 2013a, 2013b) argue that tha is an epistemic modal operator which is temporally anchored at the utterance time. What is also crucial to recall is that the [+perfective], [-past] cannot function as an independent tense (Holton et al. 1997: 221) and is therefore treated as signaling non-deictic time. As Giannakidou (2013) notes, the perfective non-past verb form in Greek does not shift to the future and cannot function as present tense either (hence its ungrammaticality). When combined with particles (among which is the particle tha), [+perfective], [-past] becomes licensed and denotes a forward shifted time interval, which, however, is still not a future tense, in the sense described in the beginning of this paper, at least at the level of what is coded. Based on the discussion so far, it is still not clear how speakers communicate prospective readings (i.e. the future tense) using tha and [+perfective], [-past], since these readings are neither semantically encoded nor context-free (i.e. they are not assigned by the linguistic context). So, even though the current accounts in the literature give an insight into the nature of the particle tha, they do not address the major question of how a modal interpretation at the level of sentence meaning becomes a future prospective reading at the level of what is communicated. In the remainder of the paper, I will attempt to provide an answer to this question by resorting to pragmatic notions such as intentional communication, mutual knowledge, as well as knowledge of pragmatic principles of language use. More specifically, I will maintain that while at the level of what is coded, tha with [+perfective], [-past] constructions have a non-homogeneous modal semantic base (in the spirit of Giannakidou 2012) which allows for both p and ¬p, the strong preference for p for the future (i.e. the future prospective reading)

Do not say that for which you lack adequate evidence

215

is not part of what is said, but rather part of what is inferred on the basis of some fundamental assumptions about co-operative communication in the Gricean sense. In this respect, I will propose that the future prospective reading of tha and perfective non-past sentences is a generalised conversational implicature calculated upon the modal semantic base and triggered by the maxim of informativeness.

4 Gricean pragmatic theory The distinction between what is said by a sentence and what is implicated by an utterance of that sentence, which was introduced and developed by the Oxford philosopher H.P. Grice, is very well established in the literature. The Gricean construal of meaning and communication is founded upon two basic theories, namely the theory of meaningnn (non-natural meaning) and the theory of conversational implicature, in both of which, Grice attempts to not only show the importance of non-conventional means in communication, but to also draw the line between what is said and what is actually communicated in a communicative event based on verbal means. In his theory of meaningnn,11 Grice draws the distinction between what a speaker can mean naturally and non-naturally when producing an utterance. As he himself states, a speaker means something non-naturally by uttering x, if and only if the speaker “intended the utterance of x to produce some effect in an audience by means of the recognition of this intention” (1957: 385): Gricean theory of meaningnn S meansnn p by uttering U to A if and only if S intends: (i) A to think p (ii) A to recognize that S intends (i) (iii) A’s recognition of S’s intending (i) to be the primary reason for A thinking p

In this respect, by considering what lies over and above the natural meaning of utterances, i.e. their conventional meaning, Grice distinguishes between what is conventionally stated and what is conversationally inferred. Grice then develops his theory of conversational implicature, based on the idea that speakers can intend to communicate meanings, which are not formally (linguistically) coded in their utterances. As Levinson states, the theory of conversational implicature provides “some explicit account of how it is possible 11 For more see also Grice (1957, 1969, 1989), Levinson (1983), Marmaridou (2000), among others.

216

Michael Chiou

to mean . . . more than what is actually said (i.e. more than what is literally expressed by the conventional sense of the linguistic expressions uttered)” (1983: 97). This theory is built on the supposition that conversation moves along certain guidelines and assumptions. As Grice himself states, “our talk exchanges do not normally consist of a succession of disconnected remarks [. . .] they are, to some degree at least, co-operative efforts” (1975: 45). This co-operativeness is assumed for both the speaker and the hearer in a reciprocal manner, and is taken by Grice to lie behind our ability to communicate implicatures, over and above what we actually say; hence, the stipulation of the co-operative principle, and its further explanation through maxims and submaxims which, for Grice, regulate efficient language use in communication: The co-operative principle Make your conversational contribution such as is required at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged. (Grice 1989: 26) The Gricean maxims of conversation Quality: Try to make your contribution on that is true. (i) Do not say what you believe to be false. (ii) Do not say that for which you lack adequate evidence. Quantity: (i) Make your contribution as informative as is required (for the current purposes of the exchange). (ii) Do not make your contribution more informative than is required. Relation: Be relevant. Manner: Be perspicuous. (i) Avoid obscurity of expression. (ii) Avoid ambiguity. (iii) Be brief (avoid unnecessary prolixity). (iv) Be orderly. (Grice 1989: 26–27)

In relation to these maxims, a speaker S has four alternatives. On the one hand, the speaker can observe the maxims, i.e. she can follow the dictum described by them. On the other hand, she could violate them. In addition, the speaker can opt out of a maxim and finally, she can ostentatiously flout or exploit a maxim. For the last three cases, Grice argues that even when the maxims are violated by the speaker, the hearer assumes that this violation is rather superficial and that the speaker is still co-operative. A conversational implicature may then be generated either when the maxims are observed or when they are flouted. In either case, the hearer employs inferential mechanisms in order to work out in which way the speaker has been co-operative. These inferences, which have

Do not say that for which you lack adequate evidence

217

to be drawn by the hearer, do not follow from the conventional content of utterances but they are non-logical pragmatic inferences, namely conversational implicatures (for more information, see Levinson 1983; Marmaridou 2000; Huang 2007).

5 Future prospective readings: a pragmatic account At the risk of redundancy, it is important to recall that tha with perfective nonpast sentences: (a) never indicate an event overlapping with the time of the utterance, despite containing a [–past]12 verb form, and (b), with the exception of the timeless or atemporal constructions examined earlier in the discussion, systematically encode futurity. What we have also seen is that they can also express what I have called prospective readings, in which it is communicated that all possible worlds in the future will be p worlds. As I understand it, this is what the future tense is primarily used for, namely to locate an eventuality on the timeline while not expressing uncertainty. What is more, this future prospective interpretation is context-free, in the sense that it does not depend upon the presence of an adverb of time or any other kind of linguistic context. In other words, it seems to be a generalised interpretation. Now, consider the following example: (28)

I Eleni tha erthi. The Eleni will come-PNP ‘Helen will come.’

At the level of sentence meaning, (28) conveys an epistemic necessity reading, which can be glossed as: ‘for all that I know, Helen will come’. Since the modal base is a non-homogenous space (Giannakidou 2012), there is at least one possible world in the future in which p will not be the case; yet, it is intuitively clear that this reading does not correspond to the speaker’s intended meaning. What is actually communicated, instead, is a future prospective reading, i.e. a homogenous space containing only p worlds (or what Giannakidou and Mari 2013b call “reasonable future,” after Mari 2013). In this setting, what needs to 12 Interestingly, this also shows that, in Modern Greek, the category [-past], which does not make reference to the utterance time, is distinct from the category [+present], which typically does.

218

Michael Chiou

be accounted for is essentially the fact that there is a gap between the sentence (p/¬p) and the utterance meaning (only p). Now, since the future is metaphysically and epistemically uncertain, the speaker cannot claim knowledge of eventualities that will occur after the time of the utterance. So, even if the speaker in (28) possesses evidence in order to be confident enough that ‘Helen will come’ at a future time, she does not actually know whether p will be true in the future. She therefore uses a modal construction (tha with PNP), which codes as much information as she is in a position to provide, namely (p or ¬p), making in this way the strongest statement she can while adhering to Grice’s second submaxim of Quantity (Q2). Yet, the intention of the speaker when using this particular expression is to communicate not a mere possibility but a certainty concerning the eventuality in the future. Assuming that the speaker is co-operative and that she has made the strongest possible statement (given Q2), the hearer will go on to interpret the sentence as an informative one. Since it is clear that the meaning provided by ‘Helen will come’ at the sentence level, i.e. (p/¬p), is not relevant or informative enough, and given the Co-operative Principle, the hearer will engage in reparatory inferences in order to recover the speaker-intended meaning. More precisely, a generalised conversational implicature will be generated inducing the richer, more informative interpretation provided by the future prospective reading. Therefore, the gap between the non-homogenous space of the modal base (p and non ¬p) and the homogenous space (only p) at the level of communication can be calculated in terms of the knowledge of the literal meaning and the assumption that the speaker is obeying the Co-operative Principle and its accompanying conversational maxims. According to this pragmatic analysis, sentence (28) encodes (29a) and conversationally implicates (29b): (29)

a.

Sentence meaning: Helen will come / will not come (p/¬p)

b.

Utterance meaning +>: Helen will come (p only)

In this picture, the speaker produces the minimal linguistic information that is sufficient to achieve her communicational ends, namely the epistemic modal content, In turn, the addressee enriches this modal content, by finding the most specific interpretation, up to what she judges to be the speaker’s m-intended point, namely, the future prospective interpretation. Since the implicated meaning in (29b) is more specific than the coded one in (29a), the future prospective reading enriches what is coded by reshaping the range of possible states of

Do not say that for which you lack adequate evidence

219

affairs to a narrower range of possible states of affairs and therefore, it is a subset of the extension of what is coded.13 Assuming now that the future prospective reading is a conversational implicature, it should be subject to cancellations by the addition of further premises to the original one. More accurately, the speaker should be able to opt for an addition of an expression to the original sentence with a view to cancelling the future prospective reading which would have been induced otherwise. By way of illustration, consider the following example: (30)

O Nikos malon tha petaksi The Nikos probably will fly-PNP ‘Nikos will probably fly to London.’

jia for

to the

Londino. London

In (30), the use of the modal adverb malon (‘possibly’) is bound to communicate that the future prospective interpretation associated with the use of tha with perfective non-past is not intended. In particular, by saying p (tha with PNP), the speaker knows that, assuming the Co-operative Principle, the hearer will infer q (the future prospective reading). So, since q is not intended, the speaker uses a modal adverb to avert the hearer from inferring it, cancelling in this way the future prospective interpretation, and allowing the hearer to induce the modal epistemic reading (p/¬p) with no apparent contradiction. What is more, like all conversational implicatures, the future prospective reading can also be explicitly stated without redundancy. By way of illustration consider the examples: (31)

O Nikos sigura tha petaksi The Nikos certainly will fly-PNP ‘Nikos will certainly fly to London.’

(32)

I Eleni tha erthi The Eleni will come-PNP ‘Helen will come 100%.’

jia for

to the

Londino. London

100%. 100%

In both cases, the speaker overtly states that only p is bound to happen, which is the implicated meaning, with no apparent contradiction. All in all, the future prospective interpretation can be derived directly on the basis of the linguistically coded content, on the one hand, and the maxims and 13 Such an analysis also echoes Zipf’s opposing pressures exerted by the speaker’s and the hearer’s economies of effort, the one oriented toward minimal linguistic articulation, the other toward maximal explicitness (cf. Levinson 2000: 112–115).

220

Michael Chiou

the Co-operative Principle, on the other. Echoing the Gricean rationale, let us name the sentence meaning, i.e. the modal reading, p and the future prospective reading q. In our case, the speaker has said that p, and there is no reason to think that she is not being co-operative. So, by saying that p the speaker must think that q. What is more, the speaker knows that the hearer knows that q must be inferred since co-operativeness is mutually assumed, and she has done nothing to stop the hearer from inferring that q. As a result, the speaker intends the hearer to think that q, and in saying that p has implicated q.

6 Conclusion and further implications In this paper, I have considered a pragmatic analysis of the interpretation of future tense in Modern Greek. For its purposes, I mainly examined the basic facts and the literature concerning tha with perfective non-past and tha with imperfective non-past, since it is these two constructions that have temporal functions and appear in the tense system of Modern Greek as the future tenses. Based on the theoretical benefits of the proposed literature so far, I argued for a pragmatic account for future tense in Modern Greek in terms of the Gricean pragmatic theoretical framework. More specifically, while assuming that tha constructions contribute an epistemic modal semantic base, following Giannakidou (2009, 2012) and Giannakidou and Mari (2012), I claimed that future prospective interpretation associated with tha and perfective non-past combinations arises as an inference at the level of communication. In essence, what I proposed is that future prospective interpretations are calculated upon the semantic meaning of the sentence containing tha and perfective non-past constructions and arise as a conversational implicature as a result of the lack of further specification or the lack of need for it. If it is correct, this proposal has two major implications for current thinking on future-time reference. In the first place, at least for Modern Greek, future tense can be considered to be a special case of modality with its future reading arising since it is more informative than the epistemic modal one. The idea that future time reference can be modal is not novel; Giannakidou (2009, 2012) and Giannakidou and Mari (2012) also argue that future is a kind of epistemic modality and more specifically that it is related to evidentiality. What is more, in an analysis of the English will within the Default Semantics framework, Jaszczolt (2006) also suggests that the different readings of will can be better explained by a scale of epistemic modality,14 showing thus that future-time reference can be modal. 14 For more arguments for temporality as epistemic modality, see Jaszczolt (2009, 2013).

Do not say that for which you lack adequate evidence

221

Secondly, it appears that the interpretation of future tense is regulated by the division of labour between semantics and pragmatics. The coded content of the traditionally called “future tense” is non-past, epistemic and makes reference to possible worlds, leaving also the possibility of a future reading open. Nevertheless, what is actually communicated is a non-past, non-present meaning, which is a subset of the meaning of the semantic base. The future interpretation arises as a more specific, temporal interpretation based on the semantic content of what is coded when it is consistent with what the speaker intends to communicate.

References Abusch, Dorit. 2004. On the temporal composition of infinitives. In Jacqueline Guéron & Alexander Lecarme (eds.), The syntax of time, 1–34. Cambridge, MA: The MIT Press. Bybee, Joan L. & Östen Dahl. 1989. The creation of tense and aspect systems in the languages of the world. Studies in Language 13. 51–103. Bybee, Joan L., Revere D. Perkins & William Pagliuca. 1994. The evolution of grammar: Tense, aspect and modality in the languages of the world. Chicago & London: University of Chicago Press. Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press. Comrie, Bernard. 1985. Tense. Cambridge: Cambridge University Press. Condoravdi, Clio. 2002. Temporal interpretation of modals: Modals for the present and for the past. In David Beaver, Stefan Kaufmann, Brady Clark & Luis Casillas (eds.), The construction of meaning, 59–88. Stanford, CA: CSLI. Condoravdi, Clio. 2003. Moods and modalities for ‘will’ and ‘would’. Paper presented at the Amsterdam Colloquium, University of Amsterdam, 19–21 December. Copley, L. Bridget. 2002. The semantics of the future. Cambridge, MA: MIT PhD dissertation. Dahl, Östen. 1985. Tense and aspect systems. Oxford: Blackwell. Enç, Murvet. 1996. Tense and modality. In Shalom Lappin (ed.), Handbook of contemporary semantic theory, 345–358. Oxford: Blackwell. Escandell-Vidal, Victoria. 2010. Review of Bert Cornillie’s ‘Evidentiality and epistemic modality in Spanish (semi-) auxiliaries: A cognitive-functional approach’. Spanish in Context 7. 314– 318. Giannakidou, Anastasia. 2009. The dependency of the subjunctive revisited: Temporal semantics and polarity. Lingua 120. 1883–1908. Giannakidou, Anastasia. 2012. The Greek future particle as an epistemic modal. In Zoe Gavriilidou, Angeliki Efthimiou, Evangelia Thomadaki and Penelope Kambakis-Vougiouklis (eds.), Proceedings of 10thInternational Conference of Greek Linguistics, 48–61. http://www.icgl.gr/ files/Selected_papers.zip. (accessed 20 April 2016.) Giannakidou, Anastasia. 2014. The futurity of the present and the modality of the future: A commentary on Broekhuis and Verkuyl. Natural Language and Linguistic Theory 32. 1011– 1032.

222

Michael Chiou

Giannakidou, Anastasia & Alda Mari. 2012. The evidential nature of reasoning with the future. Ms. University of Chicago & ENS Paris. Giannakidou, Anastasia & Alda Mari. 2013a. The future of Greek and Italian: An epistemic analysis. In Emmanuel Chemla, Vincent Homer, & Grégoire Winterstein (eds.), Proceedings of Sinn und Bedeutung 17. 255–270. http://semanticsarchive.net/sub2012/ (20 April 2016.) Giannakidou, Anastasia & Alda Mari. 2014. A two dimensional analysis of the future: Modal adverbs and speaker’s bias. In Maria Aloni, Michael Franke & Floris Roelofsen (eds.), Proceedings of the 19th Amsterdam Colloquium, 115–122. Amsterdam: ILLC. Grice, H. Paul. 1957. Meaning. Philosophical Review 67. 377–388. Grice, H. Paul. 1969. Utterer’s meaning and intentions. Philosophical Review 78. 147–177. Grice, H. Paul. 1975. Logic and conversation. In Peter Cole and Jerry L. Morgan (eds.), Syntax and Semantics 3: Speech Acts, 41–58. New York: Academic Press. Grice, H. Paul. 1989. Studies in the way of words: Cambridge, MA: Harvard University Press. Haegeman, M. Lilliane. 1983. The semantics of will in present-day British English: A unified account. Brussels: Paleis der Academien. Heine, Bernd. 1995. Agent-oriented vs. epistemic modality: Some observations on English modals. In Joan Bybee & Suzanne Fleischman (eds.), Modality in grammar in discourse (Typological Studies in Language 32), 17–53. Amsterdam / Philadelphia: John Benjamins. Holton, David, Peter Mackridge & Irene Philippaki-Warburton. 1997. Greek: A comprehensive grammar of the modern language. London: Routledge. Huang, Yan. 2007. Pragmatics. Oxford: Oxford University Press. Iatridou, Sabine. 2000. The grammatical ingredients of counterfactuality. Linguistic Inquiry 31. 231–270. Jaszczolt, M. Kasia. 2006. Futurity in default semantics. In Klaus von Heusinger & Ken Turner (eds.), Where semantics meets pragmatics, 471–492. Oxford: Elsevier. Jaszczolt, M. Kasia. 2009. Representing time: An essay on temporality as modality. Oxford: Oxford University Press. Jaszczolt, M. Kasia. 2011. Time as degrees of epistemic commitment. In Piotr Stalmaszczyk (ed.), Turning points in the philosophy of language and linguistics, 19–34. Frankfurt: Peter Lang. Jaszczolt, M. Kasia. 2013. Temporality and epistemic commitment: An unresolved question. In Kasia M. Jaszczolt & Louis de Saussure (eds.), Time: Language, cognition and reality, 193– 209. Oxford: Oxford University Press. Jespersen, Otto. 1924. Philosophy of grammar. New York: W.W. Norton & Co. Joseph, D. Brian. 1983. The synchrony and diachrony of the Balkan infinitive: A study in aerial, general and historical linguistics. Cambridge: Cambridge University Press. Kissine, Mikhail. 2008. From predictions to promises. Pragmatics & Cognition 16. 169–189. Levinson, C. Stephen. 1983. Pragmatics. Cambridge: Cambridge University Press. Ludlow, Peter. 1999. Semantics, tense and time. Cambridge, MA: The MIT Press. Mari, Alda. 2009. Disambiguating the Italian future. In GL2009: Proceedings of the 5th International Conference on Generative Approaches to the Lexicon, 209–216. http://www.gl2009. org (20 April 2016.) Mari, Alda. 2013. Each other, asymmetry and reasonable futures. Journal of Semantics 31. 209– 261 Marmaridou, A. Sofia. 2000. Pragmatic meaning and cognition. Amsterdam/Philadelphia: John Benjamins. Palmer, R. Frank. 1987. Mood and modality. Cambridge: Cambridge University Press.

Do not say that for which you lack adequate evidence

223

Philippaki-Warburton, Irene. 1992. Η συντακτική ταυτότητα του να [The syntactic identity of na]. Studies in Greek Linguistics 12. 255–274. Philippaki-Warburton, Irene. 1994. Verb movement and clitics in Modern Greek. In Irene Philippaki-Warburton, Katerina Nikolaides & Maria Sifianou, (eds.) Themes in Greek Linguistics (Current Issues in Linguistic Theory 117), 53–60. Amsterdam/Philadelphia: John Benjamins. Prior, N. Arthur. 1957. Time and modality. Oxford: Oxford University Press. Prior, N. Arthur. 1967. Past, present and future. Oxford: Oxford University Press. Rivero, Maria-Luisa. 1994. Clause structure and V-movement in the languages of the Balkans. Natural Language and Linguistic Theory 12. 63–120. Roussou, Anna & Anastasios Tsangalidis. 2010. Reconsidering the ‘modal particles’ in Modern Greek. Journal of Greek Linguistics 10. 45–73. Sarkar, Anoop. 1998. The conflict between future tense and modality: The case of will in English. University of Pennsylvania Working Papers in Linguistics 5. 91–117. Smith, Carlota. 1991. The parameter of aspect (Studies on Linguistics and Philosophy 43). Dordrecht / Boston: Kluwer Academic Publishers. Squartini, Mario. 2004. Disentangling evidentiality and epistemic modality in Romance. Lingua 114. 873–895. Thomason, H. Richmond (1984). Combination of tense and modality. In Dov Gabbay & Franz Guenthner (eds.), Handbook of philosophical logic: Extensions of classical logic, 135–165. Dordrecht: Reidel. Tsangalidis, Anastasios. 1999. Will and tha: A comparative study of the category of the future. Thessaloniki: University Studio Press. Tsangalidis, Anastasios. 2001. Criteria of modalhood: The case of Modern Greek modal particles. Studies in Greek Linguistics 21. 759–770. Xidopoulos, Georgios. 1996. Tense, aspect and adverbials in Modern Greek. London: UCL PhD dissertation.

Sonja Müller

12 Redundancy effects in discourse: On the modal particle-combinations ‘halt eben’ and ‘eben halt’ in German Abstract: This article investigates the combination of the modal particles eben and halt in German. Given the common view in the literature that the orders of modal particles are fixed when they occur in sequence, halt and eben are exceptional because they can be found in both orders. After empirically testing hypotheses from the literature (e.g. performance errors, sentence mood, dialectal variation) on how to account for this variation through a corpus study and rating experiments, I will argue for a difference in markedness based on the interpretation of the particles’ combination. On the basis of the description of the modal particles’ isolated occurrence, which reveals that the meanings of halt and eben are connected by a relation of entailment (eben → halt), I will argue that the unmarked halt eben is the preferred order because it mirrors a sequencing which is independently considered unmarked when presenting an entailment and its trigger within the same utterance.

1 Modal particle combinations and combinations of halt and eben Modal particles, like the ones highlighted in (1) and (2) below1, constitute a part of speech whose members are typically attributed the following properties (for an overview, see Diewald 2007; Thurmair 2013: 628–630; Müller 2014b: chapter 2):

1 It is noted by most German authors who have written about modal particles that it is very hard (if possible at all) to provide appropriate translations or even paraphrases for them in English. It is even harder to come up with such renderings for sequences of modal particles. Therefore, in this paper, I will not provide translations/paraphrases for the modal particles in the example utterances. If necessary for my argumentation, I will describe the modal particles’ contributions in the text. Sonja Müller, Bergische Universität Wuppertal DOI 10.1515/9781501505089-012

226 – – – – –

Sonja Müller

They do not inflect. They cannot be accentuated. They are restricted to the middle field. They are synsemantic words. They do not contribute to the truth conditions of the sentence.

(1)

Kanada besteht ja aus zehn Provinzen und drei Territorien. Canada consists MP of ten provinces and three territories ‘As we both know, Canada consists of ten provinces and three territories.’

(2)

A:

Sabrina ist ganz aus dem Sabrina is completely out the ‘Sabrina is giddy with excitement!’

B:

Sie hat auch gerade einen Wal gesehen. She has MP just a whale seen ‘That doesn’t surprise me as she has just seen a whale.’

Häuschen! house

Another typical characteristic of modal particles is that they can combine. Along certain restrictions (see below), they can occur in sequences, as shown in (3) and (4). (3) A: Jetzt tut mir der Rücken weh. Now does me the back sore ‘My back is aching now.’ B: Warum musstest du denn auch unbedingt Quad fahren? Why had you MP MP at all costs quad drive ‘Why did you have to go quad driving at all costs?’ (4) Können wir die Route eigentlich einfach ändern, wenn wir Lust haben? Can we the route MP MP change if we desire have ‘Can we change the route, if we feel like that?’ Although modal particles can in principle combine, it is well-known that such sequencing is subject to restrictions. Two types of restrictions seem to apply here. The first one has to do with the fact that not all modal particles can combine with each other. It seems to be generally accepted that syntactic and semantic/pragmatic (in)compatibilities play a role here (see, for example, Dahl 1985: 218 & 222–223; Thurmair 1989: 205, 1991: 20 & 25–27). According to Thurmair (1989: 204–205, 1991: 20), for example, two modal particles can only be combined if the set of sentence moods in which both particles can occur

Redundancy effects in discourse

227

independently is not empty. In this respect, the combined use of eigentlich and einfach, as in (4), is acceptable because these two particles can also occur in yes/no-interrogatives in isolation. Yet, they cannot occur together in an imperative, as in (5), because eigentlich is excluded from this domain. (5)

a.

Guck einfach aus dem Fenster! Look MP out of-the window ‘Just look out of the window!’

b. *Guck eigentlich aus dem Fenster! c. *Guck eigentlich einfach aus dem Fenster! If two modal particles can principally combine, a second restriction that comes into play concerns the relative ordering of the particles at hand. For instance, when comparing (3) and (4) above with (6) and (7) that follow, the first two utterances are clearly preferred. That means the orders of the modal particles cannot be readily reversed. (6) ??Warum musstest du auch denn unbedingt Quad fahren? (7)

??Können wir die Route einfach eigentlich ändern, wenn wir Lust haben?

Diverse proposals have been made in order to account for this observation. Approaches range from mere classifications (Helbig and Kötz 1981), semantic/ pragmatic criteria (e.g. Doherty 1985; Abraham 1995), syntactic conditions (e.g. Doherty 1985; Ormelius-Sandblom 1997; Rinas 2007) and information structural criteria (de Vriendt, Vandeweghe and van de Craen 1991), to phonological (Linder 1991) and historical argumentations (Abraham 1995). The combination that the present paper focuses on is that of the modal particles halt and eben. When occurring in isolation, the two particles can be used in declarative and imperative sentences or, in illocutionary terms, in assertive and directive utterances. And in perfect accordance with the condition on intersection mentioned above, they can combine in exactly these sentential contexts, too, as (8) and (9) below show (after Karagjosova 2004: 340/215). (8)

A:

Peter sieht sehr schlecht Peter looks very bad ‘Peter looks very bad.’

aus. PRT

B:

Er war halt/eben / halt eben He was MP/MP / MP MP ‘He had been ill for a long time.’

lange long

krank ill

gewesen. been

228 (9)

Sonja Müller

A:

Ich schaffe es nicht bis morgen! I manage it not until tomorrow ‘I cannot make it until tomorrow.’

B:

Arbeite halt/eben Work MP/MP ‘Work faster!’

/ /

halt eben MP MP

schneller! faster

Given the common assumption that the order of modal particles in combinations is restricted, the combined occurrence of halt and eben has to be considered an exception because, in this case, both orders are attested; for both sequences, individual examples can be found in the literature and it is also very easy to find evidence for halt eben as well as eben halt in corpora. For example: (10)

2-1 in lead against nine men and a striker Jan Koller as a goalkeeper, the industry leader preferred to keep the ball in their own ranks, rather than set the record straight.2 In der Krise geht Vorsicht halt eben über alles. about everything in the crisis goes caution MP1 MP2 ‘During the crisis, caution comes first.’ (RHZ02/NOV.07350 Rhein-Zeitung, 11/11/2002)

(11)

Approximately eight times I’ve moved in Braunschweig. The spice jars always moved as well. They belong to the standard arsenal of a joyless single life. So was hat man eben halt. MP1 Something has one MP2 ‘One has something like this.’ (BRZ10/FEB.11489 Braunschweiger Zeitung, 23/02/2010)

Against this background, there are several questions that arise with respect to this state of affairs: Why are both sequences used and how do they relate to each other? Are they equivalent? Is there a preference for one over the other? If yes, what does it depend on? In what follows, I will argue that, in present-day German, halt eben is the unmarked and eben halt the marked sequencing. In order to establish this empirical generalisation, I will present, in section 3, the results of a large-scale corpus study (section 3.1) as well as two controlled rating experiments involving 2 All translations from German into English provided in examples are my own.

Redundancy effects in discourse

229

paired-comparisons (section 3.2). As I will have discussed in section 2, the motivation for these studies is that the literature on the combination of modal particles contains contradicting and untested claims, like that dialectal variation is at issue here, that the order depends on sentence mood and that the combination does not exist (in any order) at all. Then, in section 4, I will propose an explanation for the difference in markedness between the two orders by arguing that iconicity is at play here. The main idea is that the meaning of eben entails the meaning of halt (section 4.2.1), so a reinforcement of entailments arises in the sequence eben halt – a configuration which will have independently been shown to be marked (section 4.2.2). A precondition for this analysis is the assumption that both modal particles relate to the same proposition. By weighing different views on how to interpret the combinations of the two modal particles, my conclusion is that their meanings add up (section 4.2). Reinforcing entailed information involves a type of discourse structural redundancy and although this does not lead to complete ungrammaticality/unacceptability, the order which presents the information non-redundantly is usually preferred. Section 4.2.3 provides evidence for the assumption that speakers are indeed sensitive to such ways of presenting information from the experiments described in section 3, which also contained items of reinforced entailments coded by other linguistic means. Section 5 summarises the findings and discusses potential directions for future research.

2 Views on the combination of halt and eben in the literature As mentioned in section 1, in principle, both perceivable combinations (halt eben and eben halt) are easy to attest in corpora. In the literature, examples for both orders are mentioned as well. However, there is no consensus on how to cope with the variation. One explanation is that sentence mood is decisive for which sequence occurs. Zifonun, Hoffmann and Strecker, for example, claim that in the declarative mood (Aussagemodus in their terminology) the order is eben halt, and in the imperative mood (Aufforderungsmodus) it is halt eben (1997: 1542–1543), while they also note elsewhere that halt eben is the order in both the declarative mood and the imperative mood (1997: 908–909). Dahl states that in constative utterances the order is eben halt, and that in imperative utterances it is halt eben or eben halt (1985: 227, 230 & 234).

230

Sonja Müller

Only few authors have worked with authentic data: Thurmair has only hits for halt eben in her corpus and goes on to say that the reversed order is not familiar to her (1989: 257, fn 32). Yet, she refers to Hentschel (1986: 256) who has identified eben halt in spoken data. There are single examples for eben halt in Hartog and Rüttenauer (1982) (2x) and Rost-Roth (1998) (1x), while Dittmar (2000) predominantly finds eben halt, but halt eben is also attested in his corpus. Because of the fact that Thurmair (1989) uses corpora of Bavarian and has only hits for halt eben and Dittmar (2000) predominantly finds eben halt in his corpus which contains interviews with inhabitants of Western and Eastern Berlin in the 1990s, Elspass states in a side remark that this diverse distribution is indicative of a regional difference (2005: 17, fn 41). To understand Elspass’ assumption, a short digression to studies which claim that halt and eben used to be regionally distributed in the German speaking world is necessary. To this end, consider the map in Figure 1 below, which exhibits the results of a study (mostly concerned with lexemes) aiming at capturing regional differences in language use across Germany, Austria and Switzerland. Halt and eben were part of this survey by means of the question in (12). (12)

Question 113 Does one put halt or eben or a different expression in a sentence such as: “Der The ich I

Zug train ______ ______

fährt leaves

erst not before

so lange so long

in in

einer one

Stunde, hour

da then

muß must

warten?” wait

‘The train does not leave before an hour, I will have to wait that long then.’ If yes, please insert. As one can see in Figure 1, the survey revealed a relatively clear separation with eben being used in the top part of the country and halt in the lower part. But this survey by Eichhoff (1978) took place at the beginning of the 70s. Elspass (2005) reports on a subsequent study of the same style in the form of an online-survey which took place in 2002. In turn, this study produced the map in Figure 2. What Figure 2 shows is that the use of halt has spread north and that eben also occurs in the south. On the basis of such studies, it has been argued that there used to be a north-south- and a west-east-divide (on the west-east-axis, see Protze 1997) which, however, seems to have been abolished: In the south and west, halt and eben have co-existed for a longer time, but, in the north and

Redundancy effects in discourse

231

Figure 1: Distribution of halt and eben in the German speaking countries in the 1970s (Eichhoff 1978: 103)

east, only eben used to be known and halt has found its way into language use. That said, in present-day German, it can be safely assumed that both particles are used across the entire area where German is spoken.3 3 Interestingly, this development does not seem to be as recent as Elspass’ (2005) survey might suggest because authors who looked at the first survey by Eichhoff (1978) in the 1980s (cf. Hartog and Rüttenauer 1982: 78; Dahl 1985: 97; Thurmair 1989: 124) already assumed that both

232

Sonja Müller

Figure 2: Distribution of halt and eben in the German speaking countries in 2002 (Elspass 2005: 51)

forms are known and used in the German of that time. Consequently, the use of halt in the upper part of the country seems to have had already been established in the 80s. For more details and subsequent studies concerned with regional differences as well as criticism on the conclusions drawn, see Müller (2016a: chapter IV, section 1).

Redundancy effects in discourse

233

Even though Elspass made the relevant remark about the two combinations being distributed regionally, he did not say in which way he thinks that the occurrence of the single particles (which used to be distributed regionally) is connected to the distribution of the combinations. However, if one is to make an educated guess, it would seem that halt eben can be found where halt used to prevail – that is, in the South, hence Thurmair’s (1989) Bavarian data – and that eben halt occurs where only eben was known – in the East, hence, Dittmar’s (2000) Berlin data. In this vein, it would also be plausible that eben halt is used more in the North as well. This overview of the assumptions which are made in the literature on the combinations of halt and eben shows, on the one hand, that there are several statements which contradict each other (e.g. distribution in particular sentence moods/illocutionary types, relevance of hits vs. errors) and, on the other, that a permanent empirical generalisation on the relation between halt eben and eben halt has still to be found, as existing corpus studies have generated opposing results.

3 Solving the puzzle empirically As I will now turn to discuss, I conducted corpus queries (section 3.1) and gathered speaker judgements (section 3.2) with the aim of identifying an empirically valid generalisation concerning the two combinations of halt and eben.

3.1 Frequencies in corpora Starting off with my corpus analysis, I determined the ratio of halt eben- and eben halt-utterances, with a view to finding out whether the two combinations are used in equal measure or whether any differences can be observed in their frequencies. To this end, I ran searches for the structures in the largest corpora of written (Deutsches Referenzkorpus [DeReKo])4 and oral German (Datenbank für gesprochenes Deutsch [DGD2]),5 as well as in a webcorpus (DECOW2012; Schäfer and Bildhauer 2012)6 in which I had access to 8m tokens. DeReKo and DGD2 capture more traditional data, whereas DECOW additionally introduces a novel

4 https://cosmas2.ids-mannheim.de/cosmas2-web/ 5 http://dgd.ids-mannheim.de:8080/dgd/pragdb.dgd_extern.sys_desc 6 http://hpsg.fu-berlin.de/cow/

234

Sonja Müller

type of data into the survey. As the figures in Table 1 below show, halt eben is attested much more frequently than eben halt in all three corpora.7 Table 1: Frequencies halt eben and eben halt in the corpora

DeReKo DGD2 DECOW2012

halt eben

eben halt

715 63 7328

117 10 2291

In cases where information on the speakers’ origins was accessible (this applies to parts of the oral data), I was able to further break down the data. In the same manner, I classified the newspapers which are part of DeReKo, according to the German-speaking country of their origin (Germany, Austria, Switzerland), and within Germany, I opened up subgroups for newspapers from the North, South, East and West.8 Although there were principally more hits for the combination of the two particles in the southern data, both halt eben and eben halt are used by southern, northern, western and eastern speakers alike, and at no point was it the case that the ratio of halt eben and eben halt gets reversed. In almost all cases (I determined the ratio for 30 newspapers and 8 oral sub-corpora.), halt eben outnumbers eben halt. The results of the differences in frequencies, thus, do not provide a reason to assume that there are regional differences in the distribution.9 Still, generalising across this very large amount of data, I was able to deduce that, as far as its use is concerned, the modal particle-combination eben halt is more marked than halt eben. In this respect, I firmly reject Autenrieth’s (2002) assumption that both combinations are due to planning errors, since they occur very frequently in all three corpora. From my point of view, this cannot be regarded as a sporadic use which one could associate with planning errors in spoken language. Furthermore, since the orderings do not only occur in medially and/or conceptually oral language, the planning error explanation would beg the question of why halt eben is used in a markedly higher frequency 7 The differences between the frequencies are highly significant (p < 0.001) and strong effects are on hand in all three cases. DeReKo: χ2 (1, n = 832) = 429, p < 0.001, V = 0.719; DGD2: χ2(1, n = 73) = 38.48, p < 0.001, V = 0.726; DECOW2012: χ2 (1, n = 9619) = 2637.63, p < 0.001, V = 0.524. 8 While it should probably be assumed that all German newspapers represent standard German, it is also likely that, if a certain expression is entirely unknown and unused in certain areas, it would not occur in this area’s press. 9 Due to lack of space, I cannot present the figures for the detailed breakdown here, but see Müller 2016a: chapter IV, section 3.2; 2016b: 150-154).

Redundancy effects in discourse

235

than eben halt accross the board. Put differently, one would have to wonder why the planning errors get preferably realised by the sequence halt eben. I, therefore, consider it more plausible that the combinations of halt and eben are forms which have to be taken seriously within the language system and which speakers employ for very specific communicative purposes in a conversation.

3.2 Speaker judgements As we have seen so far, the figures obtained from the corpus analyses support the rejection of the hypothesis that both combinations of halt and eben should be treated as planning errors. Also, as far as pure frequencies are concerned, halt eben clearly outnumbers eben halt in the three conceptually and medially different types of corpora. Therefore, it can be considered the unmarked sequence in the relevant sense. However, as I could not find positive evidence for the relevance of regional differences, and as the corpus data are not suited to answer the question of whether the sequence depends on sentence mood, since imperative utterances occur only infrequently in the corpora I have examined, more empirical investigation was deemed necessary. Therefore, I conducted three acceptability judgement studies.10 In each of the two tests reported here, the participants judged 48 sentences at the end of small contexts. The tasks involved paired comparisons, with the decision to be made being whether ‘sentence (a) is better than (b)’ or ‘sentence (b) is better than (a)’. 12 of the 48 items were test items (see below on the constitution and nature of the filler items). Each test contained six declarative sentences ([-wh], verb-second sentences) or (respectively) six imperatives in which the orders halt eben and eben halt were contrasted as well as six further items which aimed at investigating a phenomenon that I will come back to present in section 4.2.3. The target sentences were all structured in the same way, as far as both their form and introductory contexts are concerned, following the pattern in (13): (13) Er + ist halt eben / eben halt + bare NP (two syllables, accent on first syllable) he is MP1 MP2 / MP2 MP1 In all the preceding contexts, the discourse partner asks for the reason for a certain characteristic trait. The test sentence is the answer to this question. For example: 10 Due to lack of space, I can only report on two of them here. On the third rating study, see Müller (2016a: chapter IV, section 5.4; 2016b: 161–165).

236 (14)

Sonja Müller

meanness Kathrin:

Why is Andreas always so frugal?

Daniel:

Er Er He ‘He

ist halt eben ist eben halt is MP MP is Swabian.’

Schwabe. Schwabe. Swabian

The directive test items all followed the structure in (15): (15) Dann + finite imperative verb (one syllable) + halt eben/eben halt + Then MP1 MP2/MP2 MP1 adverb (two syllables, ends in {-er}) + verbal particle (one syllable)! In all six directive items, the discourse partner expresses a long-standing problem in the turn immediately preceding the modal particle-utterance. With the modal particle-utterance, the speaker makes a suggestion about what the hearer should do in order to solve the problem. (16) shows a test item for a directive. (16)

rehearsal Ferdinand:

I’m always late for the orchestra rehearsal.

Wolfgang:

Dann Dann Then ‘Then

lauf halt eben lauf eben halt walk MP MP set off earlier!’

früher früher earlier

los! los! off

By choosing the respective non-linguistic contexts and appropriate propositional contents for the test sentences, I tried to guarantee that neither of the single particles would be preferred. If these two factors were not controlled for, they could have an impact on the choice of the modal particle combination. By designing a scenario in which the problem introduced is one which has been going on for a long time for directives, and by choosing inherent characteristics such as nationality, origin or profession for assertions, I excluded the possibility of understanding eben as a temporal adverb, meaning ‘just (now)’ or ‘currently’, which would have skewed the judgements. After all, since almost all expressions that qualify as modal particles also correspond to different parts of speech, one has to guarantee that the rating decisions are indeed based on the modal particle-interpretation. The filler items were chosen as follows: Half of them represented phenomena from the grammar (for example, the selection of different complement clauses,

Redundancy effects in discourse

237

correlates, different tenses in imperatives, word orders in adverbial clauses) and half of them involved coherence relations, such as temporal or causal relations between two sentences which should have an impact on the order of the sentences. The reason for inserting 18 filler items which do not test for grammatical relations in the narrow sense will become obvious in the forthcoming sections 4.2.2 and 4.2.3, where I will introduce in the discussion the phenomenon which the remaining six test items related to. Overall, there were eight different versions of each test in which the same 12 test items had to be judged. In addition, four of these eight versions were presented in two different randomisations. In all the 12 versions of the test, the answers (a) and (b) were counterbalanced across the tests. This applied to both the test sentences as well as to the filler items. Twenty-nine (test 1) and thirty-two (test 2) German native speakers participated in Test 1 and Test 2 respectively.11 Given that a discussion on dialectal differences seems to be relevant to this topic, the demographic information is also worth mentioning in this case. As all testants were students, they were relatively young: On average, they were 26 years (test 1) and 23 years (test 2) of age. So, given their age, my subjects notably belong to a different generation than the participants of the studies described in section 2. In addition to their age, I asked for their place of birth, their place of residence during their school days, and their present place of residence. I finally took the place they had been living at during their school days in order to decide on their origin, as this is the place where they would have lived in for the longest period of their lives, given their young age.12 In relation to the data points from Eichhoff (1978) (see the map in Figure 1), the participants fell almost exclusively in the area C (test 1: 86%, test 2: 80%), with some coming from the middle and northern part of D (test 1: 14%, test 2: 10%).13 Interestingly, the subjects of this study came from the area which is assumed to be an original eben-area and in which halt was considered to be unused or even unknown a while ago. So, if, as Elspass 2005 speculates, the two possible orders of halt and eben in combinations are distributed regionally, these participants should prefer the sequencing eben halt. As far as the frequencies for choosing halt eben or eben halt are concerned, however, both tests showed a clear and also statistically significant preference

11 Thanks to Sandra Pappert and Jens Michaelis for letting me carry out this study in their classes as well as to my own students during the winter term 2013/2014. 12 In fact, the three statements do not even deviate in the large majority of the cases. 13 In test 2, the three remaining speakers would have to be assigned to areas A, H and E.

238

Sonja Müller

for the order halt eben. The overall results obtained are presented in Table 2 below.14,15 Table 2: Frequencies of overall results16

assertions directives

halt eben

eben halt

138 177

29 11

As the figures reveal, the results of both experiments are in perfect accordance with the corpus data: The order halt eben is clearly preferred in assertions as well as in directives. As both sentence moods (declarative and imperative sentences) or illocutionary types (assertions and directives) in which halt and eben can be combined in principle have been tested in relation to the preferred order of the two modal particles and both tests produced the same results, I did not find evidence for the assumption that the illocutionary context or sentence mood controls which order is preferred. Also, even though I did not explicitly intend for these two experiments to form part of a dialectal study, I could not find evidence for the assumption that the choice for the one or the other order is determined by the speakers’ origins either.

4 Deriving the orders by interpretation On the basis of the corpus study as well as the following two surveys, it seems safe to assume that both orders of halt and eben in sequence exist and are used in present-day German. However, in my empirical investigation, I could not find evidence for the influence of any of the factors mentioned in the literature for the distribution of halt eben and eben halt. Factoring out speaker errors, neither sentence mood nor dialectal variation were confirmed to play a role. However, as the frequencies for the sequence halt eben clearly exceed the occurrences of 14 I am grateful to Sandra Pappert for her help with the statistical analysis of the results of the two experiments. 15 The results also turned out to be very clear for all test items (for a detailed presentation, see Müller 2016a: chapter IV, section 3.3/5.4). 16 A Generalised Linear Model with log-link function (Baayen 2008) was computed with order as dependant variable and participants and items as random variables (N = 167, log-likelihood = –64.83/N = 188, log-likelihood = –39.97). The significant intercept (β = 2.750, SE = 0.543, Wald z = 5.07, p < 0.001/β = 3.931, SE = 0.561, Wald z = 7.00, p < 0.001) indicates that significantly more halt eben- than eben halt-responses were produced.

Redundancy effects in discourse

239

eben halt and speakers clearly prefer the order halt eben in a direct comparison with the sequence eben halt in assertions as well as directives, I think it is safe to assume that there is a difference in markedness between the two orders, according to which, halt eben is the unmarked and eben halt the marked sequencing. In this respect, the question I intend to answer in the following sections is: How is this difference in markedness motivated and what makes halt eben the unmarked and eben halt the marked order?17 To this end, the argument I will pursue is that the difference in markedness can be derived from the interpretation of the two modal particle-combinations.

4.1 Interpreting the single particles To begin with, there are several scholars who argue that although the meaning of halt and eben is similar, differences can still be made out. One commonality between utterances with only eben or halt is that they are always reactions to a previous utterance or situation, exhibiting a backward orientation (cf. Dahl 1985: 98 & 224; Thurmair 1989: 120; Karagjosova 2003: 340). Thus, such modal particle utterances cannot be used discourse-initially and are not suited for topic-changes. More specifically, the modal particle utterance and another utterance will always be in an explanatory or a condition-consequence relation (cf. Dahl 1985: 101, 125 & 288, fn. 60; König 1997: 67). In (17), for example, being ill is the explanation for looking bad. (17)

reason/explanation A:

Peter looks very bad. (= q)

B:

Er war eben/halt lange krank He was MP MP long ill ‘He had been ill for a long time.’

gewesen. (= p) been Karagjosova (2003: 340)

17 A potential objection one could raise against my conclusion of looking into interpretation in order to explain the difference in markedness is that, even though the two sentences people compared in the tests are in fact structured in parallel, they are not identical as far as their rhythm is concerned. (I am grateful to Ralf Vogel for pointing this out to me.) It is of course impossible to fulfill rhythmic and segmental identity as the two modal particles display a different number of syllables and the first syllable of eben additionally carries word stress. The phenomenon at issue here is the so-called stress clash which occurs in my eben halt-items and which is generally claimed to be dispreferred (cf. Wagner & Fischenbeck 2002: 1; Gussenhoven 2004: 141; Schlüter 2005: 18). In a third experiment, I am accounting for this factor. Due to lack of space, I can only refer the reader to Müller (2016a: chapter IV, section 5.4) and Müller (2016b: 161–165) for an elaboration on these aspects and the presentation of the study. All in all, however, rhythm does not turn out to influence speakers’ judgements, and speakers also prefer halt eben in cases where eben halt does not occur in a stress clash-environment.

240

Sonja Müller

In this case, the utterance with the modal particle provides the reason for the propositional content of the previous utterance: (18) Because he had been ill (= p), Peter looks bad (= q). The condition-consequence-relation is illustrated in (19), with (20) explicitly formulating this relation between the two propositions involved. (19)

Condition-consequence-relation (assertion) Evi:

That does not suit me today. I have incredibly much to do (= p).

Pit:

Gut, komm ich halt/eben Ok come I MP MP ‘Ok, I’ll come tomorrow then.’

morgen (= q). tomorrow (Thurmair 1989: 122)

(20) If you have so much to do (= p), I’ll come tomorrow (= q). So, in this case, the decision to come the next day is a consequence of the information that the first speaker has a lot to do on the day of the utterance. Similarly, another condition-consequence relation is exhibited in (21), and explicitly formulated in (22). (21)

(22)

Condition-consequence-relation (directive) A:

I cannot make it until tomorrow (= p).

B:

Arbeite eben/halt Work MP MP ‘Work faster then!’

schneller! (= q) faster

If you cannot make it until tomorrow (= p), you have to work faster (= q).

As the examples illustrate, directives always represent the consequence in these relations, while assertions can be either reasons/explanations or consequences. Another feature referred to when characterising eben is categoricity. This property is claimed to apply to the proposition contained in the modal particle utterance as well as to its relation with the previous utterance. Authors who refer to this feature (cf. Thurmair 1989: 120; Autenrieth 2002: 80 & 83; Karagjosova 2003: 340) claim that the eben-proposition expresses that an issue is unalterable, categorical or evident. Other impressions are that the proposition is generally valid, presents a fact or seems to be axiomatic, so that the hearer cannot question it, or that the proposition is already given and currently repeated. (23)

Redundancy effects in discourse

241

shows a dialogue which has been used to argue that categoricity does not only apply to the proposition, but also to the relation between the modal particle proposition and the proposition expressed by the preceding utterance. (23)

A:

Our neighbour is banging about today again.

B:

Er ist eben He is MP ‘He is choleric’

ein a

Choleriker. choleric (Dahl 1985: 98)

In this example, it is not just the proposition that the neighbour is a choleric that is taken to be categorical/axiomatic, but it is also the relation between the two propositions expressed that is categorical, signalling a connection that is known: If you are a choleric, you are sometimes loud. Dahl also assumes that the eben-utterance has a disqualifying effect in the sense that B implies that the preceding utterance is redundant, as its contents can be easily inferred and is, therefore, not worth conveying (1985: 99). A similar situation can be seen to arise with directives, as in (24). (24)

A:

My tooth is aching.

B:

Dann geh eben zum Then go MP to-the ‘Go to the dentist then!’

Zahnarzt! dentist (Dahl 1985: 101)

In this dialogue, A should arrive at the conclusion of the eben-utterance by himself as the connection in (25) is known: (25) If your tooth is aching, you have to go to the dentist. In this sense, there is no reason for A to lament, and A’s information is irrelevant for opening up a problem-solving debate; in other words, B denies the relevance of the previous speech act in this case, too. Such examples have led authors to link categoricity to directives, as the action that is specified is apparent and the only possible solution to the problem at hand (cf. Hentschel 1986: 169; Thurmair 1989: 122). Although halt and eben are often treated on a par in the literature, there are also authors who hold the view that despite having a similar meaning, the two modal particles are not synonymous (cf. Dahl 1985: 254; Thurmair 1989: 124). One argument for this assumption is that halt and eben are not always interchangeable. Thurmair (1989: 124) refers to the dialogues in (26) and (27) to

242

Sonja Müller

illustrate that there are contexts in which eben is not appropriate, whereas halt is perfectly acceptable. Such contexts can be found for assertions as well as directives. (26)

You CAN bring your friends. a.

Wir

halt

haben

kein

Bier

mehr.

b. *Wir haben eben kein Bier mehr. We have MP no beer more ‘However, there is no beer anymore.’ (27)

Monika wants to ask Hans a favour, but she hesitates to call him. After some time, her friend says: a.

Jetzt

ruf

den

Hans

b. *Jetzt ruf den Hans Now call the Hans ‘Well, call Hans!’

halt

an!

eben MP

an! 18 on (Thurmair 1989: 124)

Conversely, according to Thurmair, it is more difficult to find cases in which eben is appropriate and halt seems inappropriate. (28) is an example she refers to: (28)

a.

Der

Wal

ist

eben

ein

Säugetier.

b. ?Der Wal ist halt ein The whale is MP a ‘The whale is a mammal.’

Säugetier. mammal (Thurmair 1989: 124)

Even if both particles can principally occur in the same context (cf. (29)), she assumes that their respective occurrence goes along with a difference in interpretation. (29)

sind

eben

a.

Männer

b.

Männer sind halt Men are MP ‘Men are like that.’

so. so. so (Thurmair 1989: 124)

18 Even though Thurmair marks the eben-utterances as ungrammatical in her discussion, considering them inappropriate seems better suited in my opinion. Either way, the point pursued here is the same under both views.

Redundancy effects in discourse

243

All in all, halt-utterances are generally weaker and less apodictic in comparison to eben-utterances (cf. Schlieben-Lange 1979: 309; Thurmair 1989: 125). As explained above, halt and eben show the same behaviour with respect to the criterion of backward orientation (i.e. causal or condition-consequence relation to previous utterance). The difference is that in the case of halt, the relation is not evident/categorical/the only possibility, but just plausible. As far as haltreasons/explanations (assertions) are concerned, the explanation offered is not presented as the only one possible. Even though the speaker is committed to the reason/explanation she presents, there are still different reasons/explanations conceivable. Halt-consequences in directives are also considered to be weaker than eben-directives, as they again do not present the only possible solution, but rather just a plausible one to the problem mentioned in the previous turn (cf. Hartog & Rüttenauer 1982: 74–75; Thurmair 1989: 125–126; Autenrieth 2002: 98). Based on this difference in the interpretation of eben- and halt-utterances, Thurmair (1989) suggests that the unacceptability of eben in (26) and (27) above, is due to the fact that, in both contexts, there are obvious alternative explanations and consequences conceivable. For (26), the speaker may not want to invite guests, if there is no beer left, but the lack of beer does not render the original suggestion to bring them over redundant. In (27), the action advised is plausible, but it is not the only self-evident solution, as going to the dentist when having a toothache is in (24).

4.2 Interpreting the particle combinations Based on the description of the isolated occurrence of halt and eben, it is now possible to look at the interpretation of their combination. One obvious way of doing so would be to look into how the scopes that each of the two particles takes over the propositions involved interact with each other. Assuming the single particles scope over the proposition, as in (30), there are in principle four possibilities concerning conceivable scope relations for an utterance, as in (31). (30)

Im Winter ist es halt/eben in-the winter is it MP /MP ‘It gets dark early in winter.’ halt (p) eben(p)

früh early

dunkel. dark

244 (31)

Sonja Müller

Im Winter ist es halt eben/eben halt früh dunkel. a.

Different scope – halt(eben(p)) – eben(halt(p))

b.

Same scope – halt(p) & eben(p) – eben(p) & halt(p)

As (31) illustrates, the particles can take different scopes – halt over eben or eben over halt – or take the same scope, in which case, they both refer to the same proposition. Even though there are researchers who suggest that the fixed order of the modal particles mirrors asymmetric scope relations (e.g. Ormelius-Sandblom 1997; Rinas 2007), I believe that an explanation based on scope-interpretation cannot be sustained, since, as we have seen, both orders are attested in everyday use and can often be even interchanged. If one wanted to adopt a scope reading for the interpretation of the combination and wanted to assume that the meaning of the modal particle-combination is related to the order of the modal particles in it, one would have to argue that the two orders of halt and eben go hand in hand with a different scope interpretation. Building on this characterisation, under the scope reading, one would either get ‘it is plausible that it is evident’ (halt eben) or ‘it is evident that it is plausible’ (eben halt).19 However, one cannot make out differences in meaning along those lines in the relevant cases, as we have already seen in (10) and (11) above. Therefore, in my opinion, the particles do not scope over each other, but it is rather the case that their meaning adds up, along the lines of (31b). This much seems to be also confirmed by descriptions authors give on how they understand the combinations; Dittmar, for example, writes on eben halt: “halt attached to eben apparently weakens the hardness of eben” (2000: 226, translation my own).

4.2.1 Eben and halt’s entailment relation A decisive finding by Thurmair mentioned in section 4.1 is that one can find examples in which halt is appropriate, but eben is not equally well-suited, as in (26) and (27). However, cases in which the occurrence of halt is inappropriate 19 For a more concrete modelling of the scope readings within a formal discourse model and a more detailed discussion, see Müller (2016a: chapter IV, section 4) and Müller (to appear).

Redundancy effects in discourse

245

(in contrast to an appropriate eben) are very rare (cf. 28). Thus, even though eben can usually be substituted by halt, the reverse replacement is not always possible (cf. Thurmair 1989: 128; Ickler 1994: 392). If eben indicates that an issue is evident and halt that it is plausible, the relevant relations can be explained in the following way: an evident issue is also plausible, a plausible issue, however, is not necessarily evident (cf. Thurmair 1989: 128). In a sense then, the meaning of eben includes the meaning of halt, but the contribution by halt does not include the contribution by eben. Put differently, eben and halt have an entailment relation, according to which the meaning of eben entails the meaning of halt, as in (32).20 (32) eben → halt Relying on the presence of this entailment relation, the observation that there are few contexts in which eben cannot be substituted by halt also finds a natural explanation. As eben implies halt, one can argue that the two particles open up a scale, with eben being the stronger element. (33)

eben | halt

Because of the entailment relation, halt can usually occur in each context in which eben can come up. However, the use of halt can be argued to give rise to a conversational (scalar) implicature: choosing the weaker element on the scale implicates the negation of the stronger one. As (34) and (35) illustrate, scalar implicatures are commonly based on such entailment scales. (34) Some people left early. +> Not all people left early. (35) all → some So, using halt in contexts such as (28), leads to the inappropriate reading that it is only plausible, rather than a self-evident fact, that the whale is a mammal.

20 On how this relation also comes about when modelling the modal particles’ contributions within a formal discourse model, see Müller (2016a: chapter IV, sections 5.1 and 5.2) and Müller (to appear).

246

Sonja Müller

4.2.2 Reinforcement of entailments Having identified the aforementioned entailment relation between eben and halt, I wish to now turn to a phenomenon that seems to bear upon these particle combinations, and, more specifically, the way in which their relative order in a sequence affects entailment triggering. The relevant phenomenon is the reinforcement of entailments, and according to it, entailed (as well as presupposed) contents cannot be reinforced, in the sense that an overt statement of the entailment (or presupposition) at hand cannot be overtly stated after introducing the element which triggers it, as in (36) to (38): (36) Johni ??managed to leave, and (indeed) hei left.

(Horn 1976: 64)

(37) John ??killed Alvin, and Alvin died.

(Horn 1976: 64)

(38) ??Johni left too, and hei did.

(Horn 1976: 66)

As (39) to (41) below show, the expressions ‘manage’, ‘kill’ and ‘too’ are responsible for the implied contents in these cases: (39) John managed to leave. → John left. (40) John killed Alvin. → Alvin died. (41) John left, too. → John left. This observation has found its way into a restriction formulated by Horn (1976), which states that the second conjunct in a coordination should not be redundant: (42) The second conjunct Q of a conjunction P and Q must assert some propositional content which does not logically follow from the first conjunct P (i.e. P & Q is anomalous if P ⊢ Q or a fortiori, if P >> Q). (Horn 1976: 65) This means that entailments cannot be reinforced and entailed information cannot be re-introduced after an entailment has been triggered. As I have suggested above, the meanings of halt and eben are interpretively coordinated. Furthermore, as I have shown in section 4.2.1, the meaning of eben entails the meaning of halt. Given this, it follows that the order eben halt is marked because, in using it, the speaker conveys the information she wants

Redundancy effects in discourse

247

to communicate in a redundant way. Much like in the examples by Horn, this does not lead to complete ungrammaticality, but the non-redundant version is undoubtedly preferred. Now, as (43) to (45) illustrate, the reversed sequencing of implied contents and entailment triggers is pretty much unproblematic. (43) John left, John managed to leave. (44) Alvin died, John killed Alvin. (45) John left, he left, too. These utterances are not marked in any way because the information is presented non-redundantly; put differently, an increase in information, vs. the decrease in (38) to (40), is on hand. As we are dealing with an increase in information in the case of the order entailed contents > entailment trigger, the order of the modal particles can also be halt eben without any problems. Halt contributes the entailed contents, while eben introduces the entailment into the structure. Therefore, I wish to claim that the order of the two particles can be motivated by the interpretation of the modal particle-sequence. The underlying assumption here is that form and function are connected (for the parallel argumentation when discussing the order of ja and doch, see Müller 2014a). In the case of halt and eben, the criterion is the increase or decrease in information, of which the non-redundant version, namely the increase, is preferred. To my mind, the preference for presenting the two contributions in this way corresponds to the preferences which speakers have in the case of coherence relations. In examples such as (46) and (47), speakers prefer the order of events which mirror the temporal and causal relations even though there is no connector to explicitly introduce this meaning. (46)

Maren opens the window and the flower pot falls down.

vs.

#The flower pot falls down and Maren opens the window. (47)

Paula takes the child to the kindergarden at eight in the morning, she picks her up at four o’clock in the afternoon. vs. #Paula picks the child up at four o’clock in the afternoon, she takes her to the kindergarden at eight in the morning.

All in all, the criterion increase in information is general enough, and can be applied to assertions and directives alike: regardless of whether a demand or a statement is made, the non-redundant version of this linguistic action will

248

Sonja Müller

be preferred. It is, thus, also true of (48), where an entailment relation between cold beer and beer is present: the version in which the information increases from beer to cold beer (48a) is perfectly acceptable, whereas (48b), which involves reinforcing the implication, has to be considered marked. (48)

a.

Get the beer from the cellar, get the cold beer from the cellar!

b. #Get the cold beer from the cellar, get the beer from the cellar! 4.2.3 Increase vs. decrease in information in the experiments As described in section 3.2, in both directives and assertions, speakers prefer halt eben to eben halt. One further aim of the two experiments that I conducted, however, was to find out whether Horn’s assumption, as formulated in (42), is also observed in speakers’ judgements. After all, if speakers would not notice differences between sentences which adhere to or violate (42), one could also doubt my explanation for the preference for halt eben over eben halt. To this end, apart from the six test items on halt eben-/eben halt-assertions or halt eben-/eben halt-directives, both experiments presented in section 3.2. contained six further test items each, in which the increase or decrease in information had to be judged. For these items, I modelled the entailment relation via hyponyms and hypernyms, with the hyponym entailing the hypernym, such as ‘hamster’ entailing ‘pet’, ‘hammer’ entailing ‘tool’, ‘teddy’ entailing ‘toy’, and so on. The test sentences were all embedded in a parallel context, where an objection by the first speaker gets rejected by the second one. All test sentences corresponded to the following formal pattern: (49) name (one syllable) + finite verb (one syllable) + NP (ein (a) + compound) [first noun accentuated], er (he) + the same finite verb + NP (einen (a) + noun) [two syllables, first syllable accentuated] Here is one relevant test item: (50) renovations Lisa:

Dirk is gone again. And we have to do the work.

Sabine: Nein, nein. Dirk holt ein Werkzeug, er holt einen Hammer. he fetches a hammer No no Dirk fetches a tool ‘No, no. Dirk is fetching tools, he is fetching a hammer.’ Nein, nein. Dirk holt einen Hammer, er holt ein Werkzeug. No no Dirk fetches a hammer he fetches a tool ‘No, no. Dirk is fetching a hammer, he is fetching tools.’

Redundancy effects in discourse

249

As half of the test items in each experiment concerned this phenomenon, which does not relate to a linguistic problem in the narrow sense, half of the filler items were also about coherence relations, of the sort illustrated in (46) and (47). Looking at Table 3 below, one can see that the results for both experiments are very clear: in direct comparisons, the cases that violate Horn’s restriction received considerably lower ratings than those that adhere to it. This suggests that the redundancy condition is indeed reflected in speaker judgements, regardless of whether the dispreferred reinforcement of entailments manifests itself in the context of hyponym – hypernym or modal particle sequences. Table 3: Results increase vs. decrease in information21

items

vehicle/tractor tool/hammer toy/teddy pet/hamster document/passport sports boat/eight

results (increase: decrease) exp 1

exp 2

28:0 28:1 26:3 25:1 24:5 21:7

32:0 26:4 30:2 31:1 27:5 28:4

5 Conclusion and future research questions Starting off with the observation that the combination of the modal particles halt and eben is exceptional since both orderings are attested, this paper aimed at suggesting an answer to the question of how this variation comes about. None of the hypotheses presented in the relevant literature, most of which have not addressed the question employing empirical methods, could be confirmed by a corpus study and two rating studies that I conducted. For one, there are too many examples of both orders in corpora to allow for an explanation in terms of “performance errors” as Autenrieth has suggested (2002: 96). Also, I could not find evidence for Elspass’ (2005: 17, fn 41) rather speculative assumption

21 In both cases, a Generalised Linear Mixed Model with log-link function (Baayen 2008) was computed with increase vs. decrease in information as dependant variable and participants and items as random variables (N = 169, log-likelihood = –53.25/N = 190, log-likelihood = –51.15). The significant intercept (β = 2.659, SE = 0.483, Wald z = 5.50, p < 0.001/β = 3.433, SE = 0.518, Wald z = 6.63, p < 0.001) indicates that significantly more increase- than decrease-responses were produced.

250

Sonja Müller

that the two sequences are attested because of dialectal differences, since eben halt and halt eben are used by southern, northern, western and eastern speakers alike, with halt eben being the way more frequent version across the board. Finally, I was also unable to confirm the influence of sentence mood, which has been suggested to play a role in Dahl (1985) and Zifonun, Hoffmann and Strecker (1997), as my experimental subjects showed preference for halt eben over eben halt in both assertions ([-wh], verb second-declarative clauses) and directives (imperative clauses). My own approach was to pursue the idea that the difference in markedness that is at issue here can be explained on the basis of the interpretation of each modal particle combination. The investigation of the contribution of halt and eben in isolation revealed that the meaning of eben entails the meaning of halt. Consequently, when combining the two particles, and depending on their ordering, the entailment trigger precedes the entailment or the entailed information precedes its trigger. The dispreference for eben halt can, therefore, be traced back to the independent pragmatic principle that goes against the reinforcement of entailments. This was further supported by the results obtained in relation to further items used in my experimental surveys, which showed that the participants’ judgements were sensitive to reinforced entailments, since utterances which presented the same information non-redundantly were preferred in a direct comparison. The overall idea is, thus, that the preference in modal particle orders is due to a form of (discourse structural) iconicity. Form and function are connected in the way that the unmarked modal particle sequence mirrors the unmarked way to present information when an entailment and its trigger occur within one utterance. By approaching the phenomenon of (relatively) fixed orders in modal particle-combinations in this way, a number of further research questions opens up. For example, as far as this particular combination is concerned, one can ask oneself whether there are contexts which are reserved for the order halt eben. If my modelling is correct, and we are dealing with a markedness phenomenon here, halt eben should have a wider distribution than eben halt and, therefore, eben halt should also be excluded from (a) certain context(s) (cf. Müller 2016a: chapter IV, section 6; Müller to appear). Depending on the results of such studies, it might also be the case that halt eben and eben halt are each used in particular contexts, in the sense that they both have their areas of use and that the halt eben-contexts do not include the eben halt-ones. Another question is how the combinations relate to the single particles. My impression is that halt eben corresponds to halt, while eben halt corresponds to eben. But then the same question as above arises: Does halt really have a wider distribution than

Redundancy effects in discourse

251

eben or do both particles have their own domains of use in which the other particle is not used? Apart from such concrete issues related to this particular combination, of course, more general questions open up as well. Is the order in modal particlesequences really as fixed as it is argued in the literature? More cases of reversed sequences can in fact be encountered: (51)

a.

hört stop

langsam MP1

Und And

sorry, sorry

mit with

„glänzendund hell“ shiny and bright

zu to

mal MP2

auf up

alle all

Namen names

übersetzen. translate

‘And sorry, stop translating all names with “shiny and bright”.’ (wikipedia-discussion-Arash (mythology)) Kommen Sie mal langsam zum Haushalt! to-the budget Come you MP2 MP1 ‘Come to the budget!’ (protocol of the session of the state parliament Niedersachsen, 22/12/2010) b.

(52) a. Auwei was ist bloß nur aus den Ruhrbaronen geworden? Oh oh what is MP1 MP2 of the Ruhr barons become ‘Oh, oh, what has happened to the Ruhr barons?’ (DECOW2012–06X: 709617545)

(53)

b.

was könnte das nur bloß gewesen sein? *grübel, grübel* has *think, think* what could it MP2 MP1 been ‘What could it have been? *think, think*’ (DECOW2012–03X: 622063131)

a.

Ja Yes

genau exactly

ja MP1

halt MP2

das that in in

meinte meant

eher rather

ich, I

kleinen small

die the

letzte last

Tour tour

war was

Clubs [. . .]. clubs

‘Yes, that’s exactly what I meant, the last tour was in rather small clubs.’ (DECOW2012–C02X3M: 5519850) b.

Aber But

ich I

hatte had

diesbezüglich in this regard

es it

halt MP2

etwas something

ja MP1

versprochen, promised

dass that

ich I

bringe, [. . .] provide

‘But I had promised to provide something in this regard, [. . .].’ (DECOW2012–05X704: 143085)

252

Sonja Müller

In each of these cases, the same issues mentioned in section 2 would need to be addressed: Are the reversals performance errors and coincidental hits? Are they restricted to certain contexts? Is there a preference or are the two sequences interchangeable? Are we generally dealing with markedness phenomena and preferences and differences in frequencies rather than with categorical exclusions and, therefore, sequencing impossibilities? All in all, I believe that such questions can only be answered when going through the maybe tedious way of investigating this phenomenon empirically. Depending on the results obtained through empirical research, one can then go on to answer the question of how to account for the variations at hand. Are there further cases for which explanations in terms of interpretation are appropriate or do approaches have to refer to other factors such as historical developments or phonological issues, too? Regardless of how these analyses work, however, they should be based on thorough empirical investigations. From my point of view, this aspect has not been taken into consideration enough by authors who have already worked on the phenomenon of modal particlecombinations, with notable exceptions being Thurmair (1989), Vismans (1994) and Dittmar (2000).

References Abraham, Werner. 1995. Wieso stehen nicht alle Modalpartikel in allen Satzformen? Die Nullhypothese. Deutsche Sprache 23. 124–146. Autenrieth, Tanja. 2002. Heterosemie und Grammatikalisierung bei Modalpartikeln: eine synchrone und diachrone Studie anhand von “eben”, “halt”, “e(cher)t”, “einfach”, “schlicht” und “glatt”. Tübingen: Niemeyer. Baayen, Rolf Harald. 2008. Analyzing linguistic data. A practical introduction to statistics using R. Cambridge: Cambridge University Press. Dahl, Johannes. 1985. Die Abtönungspartikeln im Deutschen. Ausdrucksmittel für Sprechereinstellungen mit einem kontrastiven Teil deutsch-serbokroatisch. Heidelberg: Groos. Diewald, Gabriele. 2007. Abtönungspartikel. In Ludger Hoffmann (ed.), Handbuch der deutschen Wortarten, 117–141. Berlin: De Gruyter. Dittmar, Norbert. 2000. Sozialer Umbruch und Sprachwandel am Beispiel der Modalpartikeln halt und eben in der Berliner Kommunikationsgemeinschaft nach der ‘Wende’. In Peter Auer & Heiko Haussendorff (eds.), Kommunikation in gesellschaftlichen Umbruchsituationen. Mikroanalytische Arbeiten zum sprachlichen und gesellschaftlichen Wandel in den Neuen Bundesländern, 199–234. Tübingen: Niemeyer. Doherty, Monika. 1985. Epistemische Bedeutung. Berlin: Akademie Verlag. Eichhoff, Jürgen. 1978. Wortatlas der deutschen Umgangssprachen. Vol. 2. Bern: Francke. Elspass, Stephan. 2005. Zum Wandel im Gebrauch regionalsprachlicher Lexik. Ergebnisse einer Neuerhebung. Zeitschrift für Dialektologie und Linguistik 72. 1–51.

Redundancy effects in discourse

253

Gussenhoven, Carlos. 2004. The phonology of tone and intonation. Cambridge: Cambridge University Press. Hartog, Jennifer & Martin Rüttenauer. 1982. Über die Partikel eben. Deutsche Sprache 10. 69– 82. Helbig, Gerhard & Werner Kötz. 1981. Die Partikeln. Leipzig: Verlag Enzyklopädie. Hentschel, Elke. 1986. Funktion und Geschichte deutscher Partikeln. “Ja”, “doch”, “halt” und “eben”. Tübingen: Niemeyer. Horn, Laurence. 1976. On the semantic properties of logical operators in English. Bloomington, Indiana: Indiana University Linguistics Club. Karagjosova, Elena. 2003. Modal particles and the common ground: Meaning and functions of German ja, doch, eben/halt and auch. In Peter Kühnlein, Hannes Rieser & Henk Zeevat (eds.), Perspectives on dialogue in the new millenium, 335–350. Amsterdam: John Benjamins. Karagjosova, Elena. 2004. The meaning and function of German modal particles. Saarbrücken Dissertations in Computational Linguistics and Language Technology. http://folk.uio.no/ elenakar/Diss-submitted-08-12-03.pdf. (accessed 6 January 2014) König, Ekkehard. 1997. Zur Bedeutung von Modalpartikeln im Deutschen: Ein Neuansatz im Rahmen der Relevanztheorie. Germanistische Linguistik 136. 57–75. Lindner, Katrin. 1991. Wir sind ja doch alte Bekannte. In Werner Abraham (ed.), Discourse particles: Descriptive and theoretical investigations on the logical, syntactic and pragmatic properties of discourse particles in German, 168–201. Amsterdam: John Benjamins. Lütten, Jutta. 1977. Untersuchungen zur Leistung der Partikeln in der gesprochenen deutschen Sprache. Göppingen: Kümmerle. Müller, Sonja. 2014a. Zur Anordnung der Modalpartikeln ja und doch: (In)stabile Kontexte und (non)kanonische Assertionen. Linguistische Berichte 238. 165–208. Müller, Sonja. 2014b. Modalpartikeln. Heidelberg: Winter. Müller, Sonja. 2016a. Distribution und Interpretation von Modalpartikel-Kombinationen. Submitted habilitation thesis, Universität Bielefeld. Müller, Sonja. 2016b. Halt eben vs. eben halt – Dialekt, Satzmodus, Rhythmus oder Interpretation?, Sprachwissenschaft 41. 139–184. Müller, Sonja. Forthcoming. Implikationsverstärkung in (halt eben-/eben halt-)Assertionen. Linguistische Berichte 22 (Special Issue Normalität in der Sprache, ed. by Franz d’Avis and Horst Lohnstein). Ormelius-Sandblom, Elisabet. 1997. Die Modalpartikeln ‘ja’, ‘doch’ und ‘schon’. Zur ihrer Syntax, Semantik und Pragmatik. Stockholm: Almqvist & Wiksell International. Protze, Helmut. 1997. Wortatlas der städtischen Umgangssprache. Zur territorialen Differenzierung der Sprache in Mecklenburg-Vorpommern, Brandenburg, Berlin, Sachsen-Anhalt, Sachsen und Thüringen. Köln: Böhlau. Rinas, Karsten. 2007. Abtönungspartikel-Kombinationen und Skopus. Sprachwissenschaft 32. 407–452. Rost-Roth, Martina. 1998. Modalpartikeln in Argumentationen und Handlungsvorschlägen. In Theo Harden & Elke Hentschel (eds.), Particulae Particularum. Festschrift für Harald Weydt, 293–324. Tübingen: Stauffenburg. Schäfer, Roland & Felix Bildhauer. 2012. Building large corpora from the web using a new efficient tool chain. In Proceedings from LREC 2012. http://rolandschaefer.net/de/schaferroland-bildhauer-felix-2012-building-large-corpora-from-the-web-using-a-new-efficienttool-chain/ (accessed 19 October 2014).

254

Sonja Müller

Schlieben-Lange, Brigitte. 1979. Bairisch eh – halt – fei. In Harald Weydt (ed.), Die Partikeln der deutschen Sprache, 307–313. Berlin: De Gruyter. Schlüter, Julia. 2005. Rhythmic grammar. Berlin: Mouton de Gruyter. Thurmair, Maria. 1989. Modalpartikeln und ihre Kombinationen. Tübingen: Niemeyer. Thurmair, Maria. 1991. Kombinieren Sie doch nur ruhig auch mal Modalpartikeln! Combinatorial regularities for modal particles and their use as an instrument of analysis. Multilingua 10. 19–42. Thurmair, Maria. 2013. Satztyp und Modalpartikeln. In Jörg Meibauer, Markus Steinbach & Hans Altmann (eds.), Satztypen des Deutschen, 627–651. Berlin: de Gruyter. Vismans, Roel. 1994. Modal particles in Dutch directives: A study in functional grammar. Amsterdam: IFOTT. de Vriendt, Sera, Willy Vandeweghe & Piet van de Craen. 1991. Combinatorial aspects of modal particles in Dutch. Multilingua 10. 43–59. Wagner, Petra & Eva Fischenbeck. 2002. Stress perception and production in German stress clash environments. In Proceedings of Speech Prosody 2002. Aix en Provence, France. Zifonun, Gisela, Ludger Hoffmann & Bruno Strecker. 1997. Grammatik der deutschen Sprache. Band 2. Berlin & New York: De Gruyter.

IV Conversation analysis

George O’Neal

13 The pragmatics of intelligible pronunciation: Preemptive and reactive segmental repair in English as a lingua franca interactions in Japan Abstract: The conversation analytic study reported in this paper examines the interface between pragmatics and phonology in English as a Lingua Franca interactions (Schegloff 2007; Seidlhofer 2011). Previous research demonstrates that one strategy with which English as a Lingua Franca speakers can overcome miscommunication is to modify pronunciation (Matsumoto 2011). Examining a corpus of pronunciation miscommunications, which are defined as repair sequences in which pronunciation is oriented to as the trouble source, this study claims that the repair of pronunciation takes two primary forms: 1) preemptive segmental repair, which takes the form of self repair and forestalls the possibility of other repair, and 2) reactive segmental repair, which takes the form of both self and other repair and modifies unintelligible pronunciation to intelligible pronunciation.

1 Introduction The present paper reports on a conversation analytic study of how phonology serves pragmatics, or, to put it more aptly, how phonology makes pragmatics possible. In this paper, the term pragmatics is used in a limited sense, to refer to the ways in which speakers use English in order to be understood (cf. Schegloff 2007; Ariel 2010). Phonology is, without a doubt, an aspect of that endeavour (cf. Szczepek Reed 2012). In this respect, the study reported here examines the Acknowledgements: I would like to thank Kazuo Fukuda, Yoshio Otake, Yuka Shigemitsu, Wan Jumani Fauzi, Maaike Naerssen, and Zsuszanna Németh for thorough proof-readings of an earlier draft of this paper and for helpful suggestions. I would also like to extend a special thanks to Donna Fujimoto, Yosuke Ogawa, John Campbell-Larsen, and Tim Greer of the Conversation Analysis Network Kansai group, without whom this would not have been possible.

George O’Neal, Niigata University DOI 10.1515/9781501505089-013

258

George O’Neal

relationship between the repair of segmental pronunciation and the intelligibility of pronunciation in English as a Lingua Franca (henceforth ELF) interactions in Japan. A study of pronunciation repair strategies in ELF is warranted because despite ELF being the most common use of English today, its pronunciation is understudied (cf. Jenkins 2000, 2007, 2014; Ostler 2010; Matsumoto 2011). ELF is defined as any interaction among people from different first language backgrounds for whom English is the communicative medium of choice (Seidlhofer 2011). This is not to claim, however, that ELF is a stable variety of English in the typical sense of the term variety – far from it. It is the inherent vicissitudes and variability of ELF that distinguish it from other types of English (cf. Mauranen 2006; Matsumoto 2011; Canagarajah 2013). Indeed, the negotiation of form and function during interaction is the hallmark of ELF, and as a contact language, this could only be expected. The negotiation of form and function that is inherent in ELF applies to pronunciation as well. For ELF speakers, mutually intelligible pronunciation cannot necessarily be assumed from the beginning of an interaction. After all, the inchoate nature of ELF interactions can necessitate phonological negotiation to achieve mutual intelligibility (Matsumoto 2011). Intelligibility refers to the extent to which an interlocutor understands the speaker’s speech (cf. Jenkins 2000; Munro, Derwing, and Morton 2006; Nelson 2011). But intelligibility is different from comprehensibility, which refers to how subjectively easy or difficult it is for the interlocutor to understand the speaker’s speech (Munro et al. 2006). Comprehensibility is an index of how pleasant a pronunciation sounds to an interlocutor, rather than how effective it is in the communication setting (Isaacs and Trofimovich 2012). Accordingly, under the definitions in use in this study, ELF speakers can be both intelligible and incomprehensible if the interlocutor understands the message but feels that the pronunciation used to convey the message was odd or strange. That said, the distinction between intelligibility and comprehensibility is critical for this study, because it only focuses on the intelligibility of pronunciation and the repair strategies that make it possible. In particular, this paper focuses on two strategies that ELF speakers use to overcome miscommunications in which segmental pronunciation is oriented to as a trouble source: preemptive segmental repair and reactive segmental repair. In this setting, segments refer to the phonemes of pronunciation. For instance, the standard pronunciation of the word cat is /kæt/, which is composed of three segments: /k/, /æ/, and /t/ (cf. Roach 2009). In both preemptive and reactive segmental repair, segments are adjusted to make pronunciation more intelligible. Against this background, the aim of this study is to examine the pragmatics of intelligible ELF pronunciation in Japan. As such, it analyses the way in which

The pragmatics of intelligible pronunciation

259

ELF speakers modify the segments of their pronunciations in order to make mutual intelligibility possible. All in all, this research seeks to answer the following questions: –

–

How do ELF speakers repair their pronunciation after an instance of miscommunication in which segmental pronunciation is oriented to as a trouble source? Are there different kinds of strategies with which ELF speakers can repair segmental pronunciation?

2 Previous studies In this section, I will review previous research on the repair of the intelligibility of pronunciation in ELF, as well as describe the methods through which ELF scholars have assessed intelligibility. In light of this review, it will be argued that previous methods of assessing intelligibility are inadequate, and that a conversation analytic method in which intelligibility is equated to intersubjectivity is warranted.

2.1 The repair of the intelligibility of pronunciation in ELF Although one might assume that miscommunications caused by pronunciation would be common in ELF, this is not true; indeed, many scholars have found that overt manifestations of miscommunication, whether caused by pronunciation or not, are rare in ELF (cf. Mauranen 2006; Kuar 2009; Matsumoto 2011, 2014). There are many potential reasons for this. One reason is that ELF speakers resort to what Mauranen (2006) refers to as “proactive strategies” or what Kuar (2009) refers to as “preventative repair;” that is, they often seek to prevent miscommunications before they are overtly manifested by means of self repair, clarification checks, and confirmation checks (cf. Mauranen 2006; Kuar 2009; Németh 2012). Another reason is that ELF speakers might be more concerned with communication than they are with approximating a linguistic standard, and therefore allow unintelligible utterances to pass without comment in the hopes that further information will appear later in the interaction that will explain the covert misunderstanding (Firth 1996). None of this is to say, of course, that miscommunication never occurs in ELF. But it needs to be emphasised that miscommunication in ELF does not seem to be any more prevalent than it is in native speaker only interactions (cf.

260

George O’Neal

Mauranen 2006; Kuar 2009). However, once a miscommunication has occurred in an ELF interaction, ELF speakers can manage and overcome it in many different ways. One such way is to modify the pronunciation of a word oriented to as a trouble source into a more intelligible form (Matsumoto 2011). This adjustment of pronunciation to the necessities of the interaction is a manifestation of an aspect of the phenomenon known as accommodation, in which ELF speakers modify the form and function of their language to maintain mutual understanding (Giles, Coupland and Coupland 1991). Indeed, accommodation is a common feature of ELF interactions. Some ELF scholars claim that the ability to accommodate to the linguistic needs of the interaction is one of the most important pronunciation abilities in an ELF speaker’s repertoire (see, for example, Jenkins 2000, 2002). This claim flies in the face of traditional notions of pronunciation efficacy. Many scholars claim that approximation to a native speaker pronunciation standard is the most important pronunciation ability an ELF speaker can have (Dauer 2005; Trudgill 2008; Scheuer 2008; Sobkowiak 2008). This paper rejects such a view and asserts that the intelligibility of pronunciation is in fact a property of the ability of an ELF speaker to adjust his or her pronunciation as the interaction requires, and that there are two primary means to do so: preemptive segmental repair and reactive segmental repair.

2.2 Assessing the intelligibility of pronunciation in ELF Any assessment of intelligibility has to determine if an interlocutor understands the speaker’s message, and scholars have developed various methods to do so. The most common one is to use a dictation task in which a recorded message is played to another group of people who attempt to write the message in standard orthography (cf. Brodkey 1972; Gass and Varonis 1984; Bent and Bradlow 2003; Munro et al. 2006; Osimk 2009). In this setting, the amount of the message that ends up being transcribed in standard orthography is assumed to index the intelligibility of the message; in other words, if the listener can only transcribe half of the message, then only half of the message is assumed to be intelligible. The problem with dictation tasks, however, is that the intelligibility of the message is assessed by people for whom the message was not intended. In other words, dictation tasks do not tell us if the intended recipients of the message found the message intelligible. Deterding (2013) worked around this problem, by having a group of ELF speakers dictate their own interactions in the recordings. Through this method, the dictation task can tell if the intended recipient of the message found the message intelligible, or at least dictatable. However, the

The pragmatics of intelligible pronunciation

261

problem with all variants of the dictation assessments of intelligibility persists: the ability to reproduce a dictated message does not necessarily manifest the degree to which it is intelligible. Other scholars have used comprehension questions, summaries, and cloze tests to assess intelligibility, claiming that intelligibility can be equated to the quality of the answer or the summary (cf. Smith and Rafiqzad 1979; Smith and Bisazza 1982; Anderson-Hsieh and Koehler 1988; Perlmutter 1989). However, these methods are very subjective in nature, since judging the quality of an answer can never be an objective task. Recognising this, Jenkins (2000) used observations of student interactions during a classroom activity followed by interviews with the same students. Still, even this method is not without its weaknesses, since it involves direct interaction with the researcher, which can intimidate the student subjects. As one can see, methods of intelligibility assessment have critical flaws. Apart from that, they are all premised on the notion that intelligibility cannot be assessed during an actual interaction. Against this backdrop, the method adopted for this study does exactly that: it assesses the intelligibility of pronunciation during an ELF interaction, using Conversation Analysis (henceforth CA) as a methodology. Using CA for phonology studies may be a recent phenomenon, but both Matsumoto (2011) and Szczepek Reed (2012) argue convincingly that it can be utilised for such purposes. Matsumoto (2011) demonstrated that an unintelligible pronunciation is manifest in an interaction when an interlocutor orients to a word as unintelligible and subjects it to repair, as well as when one party proffers a candidate repair pronunciation to the other party, who, in turn, affirms the candidate repair pronunciation. This paper adopts a similar stance, accepting that pronunciation is unintelligible only when one of the participants orients to the pronunciation as unintelligible. In the same vein, pronunciation is intelligible only when the participants orient to it as intelligible at the culmination of a repair sequence in which pronunciation is oriented to as a trouble source. In this respect, this paper equates the notion of intelligibility that is so often used in phonology studies to the CA notion of intersubjectivity (cf. Schegloff 2007; Matsumoto 2011), basing the determination of pronunciation intelligibility on the orientation of the interlocutor to the pronunciation as such.

3 Data and methodology The data amassed for this study form a corpus of ELF pronunciation miscommunications, which are defined as repair sequences in which pronunciation is oriented to as a trouble source (cf. Schegloff, Jefferson, and Sacks 1977; Schegloff

262

George O’Neal

1992, 1997, 2000; Kuar 2009; Matsumoto 2011). This corpus was gathered from recorded interactions that were part of compulsory homework assignments of an introductory level English communication course at a large public Japanese university between 2011 and 2014. Although the recordings are the result of compulsory homework assignments, students still exhibit interactional competence and require intelligible pronunciation to succeed; therefore, the recordings at hand are a legitimate data source (cf. Hauser 2013). Both Japanese university students and non-Japanese exchange students attended the course. The homework assignments required students to pair up with another student after class and record a conversation in which they exchanged opinions on a topic covered in class, although, in many cases, student conversations digressed into ancillary topics. It should be noted that the teacher was not present during the conversation recordings. Only recordings from students of different first language backgrounds were retained for the target corpus, with the permission of both students. These recordings range from short six-minute interactions to longer thirty-minute ones. All names provided in the transcripts in the results section are pseudonyms. All recordings are audio only, but since some students used Skype, FaceTime or Line for their homework assignments, they might also have used video during their interactions. Whether the addition of a visual element affected the intelligibility of their speech is unknown, but it should be acknowledged that it may have (Kerzel & Bekkering 2000); still, if the additional visual information affected pronunciation intelligibility in some way, the analysis that follows will not have detected it. Last, although the gender, nationality, and university major of each participant are stated in the introduction to the excerpts, this does not mean that any participant oriented to either gender, national origin, or academic specialty as a relevant feature of the interaction. Indeed, most of the time, they did not. As already mentioned above, this paper uses CA methodology to assess the intelligibility of pronunciation in ELF. But since CA transcription is not equipped to handle minute phonetic details, some modifications to the standard CA transcription system have been in order. In this respect, words that are subjected to segmental repair, i.e. to repair of the segments of a pronunciation, have been transcribed into the International Phonetic Alphabet (IPA). However, since CA transcription uses brackets to indicate simultaneous speech and IPA transcription also uses brackets to signify phonetic notation, some further modifications to avoid transcription overlap were necessary. To this end, braces have been used to indicate simultaneous speech, and brackets have been used to mark phonetic notation in IPA. So, simultaneous speech in the IPA transcription has been marked with braces and brackets (See the appendix for the transcription symbols utilised in this study).

The pragmatics of intelligible pronunciation

263

4 Results In this section, I will turn to examine representative samples of preemptive and reactive segmental repair sequences. First, I will identify the words to which the ELF speakers orient as problematic and subject to repair. Then, I will examine the progression of the repair sequence in order to determine which segmental adjustments ELF speakers make in order to change unintelligible pronunciations into intelligible ones.

4.1 Preemptive segmental repair In preemptive segmental repair, one ELF speaker self repairs his or her own segmental pronunciation before the interlocutor can even orient to the pronunciation as problematic, thus forestalling the manifestation of any intelligibility issues. Of course, because of it, it is unknown if the pronunciation would have been oriented to as unintelligible by the other interlocutor. However, the fact that one of the ELF speakers repairs his/her own pronunciation alone is evidence enough to claim that this speaker oriented to his/her own pronunciation as potentially problematic or unintelligible. As we shall see below, all examples of preemptive segmental repair manifest this. In the first example of preemptive segmental repair, the articulation of the word match is self-repaired. Yuya, a male Japanese student who majors in legal studies, and Bai, a female Chinese exchange student who majors in Japanese, are talking about Yuya’s participation in the university’s Japanese fencing club tournament, which leads to the following exchange: (1)

1 2 3 4 5 6 7 8 9 10

Yuya: Bai: Yuya: Bai: Yuya: Bai: Yuya:

yeah and uh two months ago I have tournament. tournament? what is- what is tournament? I have a game. a game? yeah. I have a [mʌt ʃ͡ ]. a [mæt ʃ͡ ] in some prefecture. yeah. yeah. and uhm I won the tourname. I won the game. I- I got first place in division.

In lines one and two, Yuya gives some information to Bai, but Bai orients to the word ‘tournament’ as unintelligible in line three, which triggers a repair

264

George O’Neal

sequence. In line four, Yuya attempts self repair, substituting ‘tournament’ for ‘game’. However, in line five, Bai does not indicate that intelligibility has been restored; in fact, Yuya orients to Bai’s contribution in line five as a request for further repair. In line six, Yuya again substitutes ‘game’ for ‘match’, which is articulated as [mʌt ʃ͡ ], but immediately subjects his own articulation of ‘match’ to preemptive segmental repair and rearticulates it as [mæt ʃ͡ ]. This reveals that Yuya oriented to his own articulation as potentially unintelligible and adjusted his pronunciation. In line eight, Bai accepts what Yuya says, completing the telling-acceptance sequence begun in line one. In line nine and ten, Yuya first deploys a sequence closing third, and then produces the first pair part of another sequence, which demonstrates that Yuya orients to the repair sequence and the superordinate telling-acceptance sequence as complete. The phonological changes in this preemptive segmental repair show exactly which segments were oriented to as unintelligible and intelligible. In line six, Yuya substitutes the segment /ʌ/ with the segment /æ/, which demonstrates that Yuya oriented to the vowel /ʌ/ as potentially unintelligible within the articulation of [mʌt ʃ͡ ]. Furthermore, this also demonstrates that Yuya orients to the vowel /æ/ as intelligible within the articulation of [mæt ʃ͡ ]. Whether Bai would have oriented to [mʌt ʃ͡ ] as an intelligible variant of ‘match’ cannot be known, but the details of the interaction reveal that Yuya oriented to [mʌt ʃ͡ ] as an unintelligible variant of ‘match’ and to [mæt ʃ͡ ] as an intelligible variant of it. The second example further demonstrates that ELF speakers can preemptively preclude potential intelligibility problems through the modification of their pronunciations. In the second example, the articulation of best will be self-repaired. Pan, a female Chinese exchange student who majors in Japanese, and Shun, a male Japanese exchange student who majors in physics, are just about to begin the topic mandated by the homework assignment, which leads to the following exchange: (2)

1 2 3 4 5 6

Pan: Shun: Pan: Shun: Pan: Shun:

I will d- do my [bɛstə]. {hm?} {>do} my [↑bɛst]{[mei.dʒɛɹ]}