Beyond Expressives: Explorations in Use-Conditional Meaning [1 ed.] 9789004183988, 9789004252172

Citation preview

Beyond Expressives: Explorations in Use-Conditional Meaning

Current Research in the Semantics/Pragmatics Interface Series Editors

Klaus von Heusinger Ken Turner Editorial Board Nicholas Asher, Université Paul Sabatier, France Johan Van der Auwera, University of Antwerp, Belgium Betty Birner, Northern Illinois University, USA Claudia Casadio, Universitá degli studi G. d’Annunzio Chieti Pescara, Italy Ariel Cohen, Ben Gurion University, Israel Marcelo Dascal, Tel Aviv University, Israel Paul Dekker, University of Amsterdam, The Netherlands Regine Eckardt, University of Gö ttingen, Germany Markus Egg, Humbolt University Berlin, Germany Donka Farkas, University of California, Santa Cruz, USA Bruce Fraser, Boston University, USA Thorstein Fretheim, Norwegian University of Science and Technology, Norway Brendan Gillon, McGill University, Canada Jeroen Groenendijk, University of Amsterdam, the Netherlands Yueguo Gu, Chinese Academy of Social Sciences, PRC Larry Horn, Yale University, USA Yan Huang, University of Auckland, New Zealand Asa Kasher, Tel Aviv University, Israel Manfred Krifka, Humboldt University, Germany Susumu Kubo, Matsuyama University, Japan

Chungmin Lee, Seoul National University, South Korea Stephen Levinson, Max Planck Institute for Psycholinguistics, the Netherlands Claudia Maienborn, University of Tü bingen, Germany Tony McEnery, Lancaster University, UK Alice ter Meulen, University of Geneva, Switzerland François Nemo, University of Orléans, France Peter Pelyvas, University of Debrecen, Hungary Jaroslav Peregrin, Czech Academy of Sciences and University of Hradec Králové, Czech Republic Allan Ramsay, University of Manchester, UK Rob Van der Sandt, Radboud University Nijmegen, the Netherlands Kjell Johan Sæbo, University of Oslo, Norway Robert Stalnaker, Massachusetts Institute of Technology, USA Martin Stokhof, University of Amsterdam, the Netherlands Gregory Ward, Northwestern University, USA Henk Zeevat, University of Amsterdam, the Netherlands Thomas Ede Zimmermann, University of Frankfurt, Germany

VOLUME 28

The titles published in this series are listed at brill.com/crispi

Beyond Expressives: Explorations in Use-Conditional Meaning Edited by

Daniel Gutzmann Hans-Martin Gärtner

LEIDEN • BOSTON 2013

Library of Congress Cataloging-in-Publication Data Beyond expressives : explorations in use-conditional meaning / Edited by Daniel Gutzmann, Hans-Martin Gärtner. pages cm. – (Current Research in the Semantics/Pragmatics Interface ; 28) (Gärtner Bibliography ; 28) Includes index. “The papers collected in this volume grew out of a workshop on Expressives and other kinds of non-truth-conditional meaning, which was held in March 2009 at University of Osnabruck as part of the 31st annual meeting of the Deutsche Gesellschaft für Sprachwissenschaft (DGfS).” ISBN 978-90-04-25217-2 (hardback : alk. paper) – ISBN 978-90-04-18398-8 (e-book) 1. Semantics. 2. Emotive (Linguistics) 3. Grammar, comparative and general–Conditionals. I. Gutzmann, Daniel, editor of compilation. II. Gärtner, Hans-Martin, editor of compilation. III. Deutsche Gesellschaft für Sprachwissenschaft. Jahrestagung (31st : 2009 : University of Osnabruck) P325.5.E56B49 2013 401'.9–dc23 2013019492

This publication has been typeset in the multilingual “Brill” typeface. With over 5,100 characters covering Latin, IPA, Greek, and Cyrillic, this typeface is especially suitable for use in the humanities. For more information, please see www.brill.com/brill-typeface. ISSN 1472-7870 ISBN 978-90-04-25217-2 (hardback) ISBN 978-90-04-18398-8 (e-book) Copyright 2013 by Koninklijke Brill NV, Leiden, The Netherlands. Koninklijke Brill NV incorporates the imprints Brill, Global Oriental, Hotei Publishing, IDC Publishers and Martinus Nijhoff Publishers. All rights reserved. No part of this publication may be reproduced, translated, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission from the publisher. Authorization to photocopy items for internal or personal use is granted by Koninklijke Brill NV provided that the appropriate fees are paid directly to The Copyright Clearance Center, 222 Rosewood Drive, Suite 910, Danvers, MA 01923, USA. Fees are subject to change. This book is printed on acid-free paper.

CONTENTS

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii List of Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix Expressives and Beyond: An Introduction to Varieties of Use-Conditional Meaning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Daniel Gutzmann

1

German Non-Inflectional Constructions as Separate Performatives . . . . 59 Sebastian Bücking and Jennifer Rau Modal Particles and Context Shift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 Sophia Döring Discourse Particles, Common Ground, and Felicity Conditions . . . . . . . . 125 Markus Egg I Love Me Some Datives: Expressive Meaning, Free Datives, and F-Implicature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151 Laurence R. Horn Good Reasons. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201 Eric McCready and Yohei Takahashi Common Ground Management: Modal Particles, Illocutionary Negation and verum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231 Sophie Repp Biased Polar Questions in English and Japanese . . . . . . . . . . . . . . . . . . . . . . . . 275 Yasutada Sudo Expressing Surprise by Particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297 Henk Zeevat Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321

ACKNOWLEDGEMENTS

The papers collected in this volume grew out of a workshop on Expressives and other kinds of non-truth-conditional meaning, which was held in March 2009 at the University of Osnabrück as part of the 31st annual meeting of the Deutsche Gesellschaft für Sprachwissenschaft (DGfS). We would like to thank the local organizers for their efforts and the DGfS for financial support. Out of the eleven papers presented in Osnabrück, seven are collected in this volume after a double-blind review process; Eric McCready (who had a different paper in Osnabrück) and Yohei Takahashi joined this project with their contribution after the workshop. First of all, we want to thank the many anonymous reviewers, who kindly invested their time to review the contributions and whose detailed comments helped to enhance the quality of the papers. The series editors, Klaus von Heusinger and Ken Turner, as well as an anonymous reviewer provided helpful comments for the volume as a whole. Klaus and Ken were also great advisors for all general questions. It was a pleasure to work with Stephanie Paalvast at Brill, who was always responsive, fast, and supportive in dealing with all technical and editorial matters. We are also thankful to Antonio Fortin for his helpful comments on the introduction, and Elisabeth Ahrends and Jacob Schmidkunz, who helped preparing the final manuscript. Hans-Martin Gärtner gratefully acknowledges support by the German Bundesministerium für Bildung und Forschung (BMBF) (Grant Nr. 01UG0711).

LIST OF CONTRIBUTORS

Sebastian Bücking is a faculty member of the German linguistics department at the University of Tübingen. Among his general research interests are formal approaches to semantics and its interfaces to pragmatics, syntax and morphology. In particular, he has worked on the semantics of modifiers and arguments in nominal structures of German. Sophia Döring has been working as research assistant for English Linguistics at the Humboldt-Universität zu Berlin since 2009. Her main research interests lie in the area of semantics and pragmatics: expressive meaning and its formal description, context dependency and context shift, as well as discourse analysis and the field of information structure. Markus Egg is Professor of Linguistics at the Dept. of English and American Studies of the Humboldt-Universität zu Berlin. His main areas of interest are syntax, semantics, pragmatics, and discourse and the interfaces between them. He works on the theoretical development of these areas as well as on their implementation in NLP systems. Hans-Martin Gärtner is research advisor at the Research Institute for Linguistics of the Hungarian Academy of Sciences. He has mainly worked on grammar formalisms as well as the syntax, semantics, pragmatics, and typology of clause types. Daniel Gutzmann is a faculty member of the Institute of Linguistics at the University of Frankfurt. His research interests are semantics, pragmatics and philosophy of language. He has worked and published on the semantics of various kinds of non-truth-conditional meaning, including expressives, modal particles, personal datives, sentence mood and verum focus, as well as on the pragmatics of quotation. Laurence R. Horn is Professor of Linguistics and Philosophy at Yale University. He is the author of the modern classic A Natural History of Negation (Chicago, 1989/CSLI, 2001) and numerous papers on negation and pragmatics. He is the (co-)editor of Negation and Polarity (OUP, 2000), The Handbook of Pragmatics (Blackwell, 2004), Explorations in Pragmatics (de Gruyter, 2007) and The Expression of Negation (de Gruyter, 2010).

x

list of contributors

Eric McCready is Associate Professor in the Department of English at Aoyama Gakuin University. He holds a PhD from the University of Texas at Austin, Department of Linguistics. His research interests are in semantics, pragmatics, and related areas, and he has published many articles on these topics in various journals and collections. Jennifer Rau wrote an M.A. thesis in German linguistics and wrote her doctoral thesis on the syntax and semantics of complementation in Tübingen. She is currently visiting scholar at the linguistic department of the University of Massachusetss/Amherst. Her research interests are the syntax/semantics interface and the semantics of mood and tense. Sophie Repp is Assistant Professor of Linguistics at the Department of English and American Studies of the Humboldt-Universität zu Berlin. Her main areas of interest are syntax, semantics, pragmatics and the interfaces between them, with a special emphasis on negation, ellipsis, coordination, questions and information structure. Yasutada Sudo is currently a post-doctoral researcher at Institut Jean Nicod. He obtained a Ph.D. in linguistics from the Massachusetts Institute of Technology in 2012. His research interests are in syntax-semanticspragmatics and he has worked on various topics, including presupposition, phi-features, indexical shifting, and the syntax and semantics of degree constructions such as superlatives and comparatives. Yohei Takahashi is a PhD student in the Department of English at Aoyama Gakuin University. He received his MA from that department in 2010 with a thesis on internally headed relative clauses. His research interests are in syntax and semantics, and their interfaces. Henk Zeevat is a senior lecturer at the ILLC, University of Amsterdam, after research jobs in Edinburgh and Stuttgart. He has published on the semantics and pragmatics of natural languages and in computational linguistics, with a focus on discourse semantics, presupposition and the formalization of grammatical and semantical knowledge.

EXPRESSIVES AND BEYOND: AN INTRODUCTION TO VARIETIES OF USE-CONDITIONAL MEANING

Daniel Gutzmann

1. Introduction Formal semanticists, as is well known, have found it useful to make certain simplifying assumptions in approaching the vast field of natural language meaning.1 Thus, following the lead of logicians, they have concentrated on rigidly stating truth conditions for declarative sentences in a quest for clarifying notions like entailment, synonymy, and contradiction (cf., for example, Dowty et al. 1981 or Gamut 1991). However, it was recognized right from the start that much work would remain to be done once the truth-conditional story was told. Again, it is well known that Frege himself directed attention to various lacunae (cf. Horn 2007): (i) the treatment of presupposition, which has oscillated between truth-conditional and “pragmatic” approaches ever since the seminal “Russell-Strawson debate” (Russell 1905; Strawson 1950);2 (ii) the necessity for envisaging something like speech acts (cf. Frege’s Urteilsstrich, which is called assertion sign by Geach 1965), which it took Austin (1962) and Searle (1969) to give theoretical shape to; and (iii) the phenomenon of expressivity, which Frege discussed under the term Färbung “colouring” (see Horn 2013 [this volume] as well as Dummett 1978: 93 and Green & Kortum 2007).3

1 A very good overview over that field is still provided by Lyons (1977). Among the many other sources are von Stechow & Wunderlich 1991, Lappin 1996, and, more recently, Maienborn et al. 2011. 2 See Beaver & Geurts 2011, and references cited there. 3 Many philosophers, disenchanted with the positivist undercurrents of 20th century philosophy of language, have pointed out the relevance of emotive meaning for the study of ethics and aesthetics (Hare 1952; Stevenson 1937). Stevenson (1937: 23) credits Ogden & Richards (1923) with the term emotive meaning. Likewise, Jakobson (1960), who uses the term for a function of language, credits Marty (1908) for it.

2

daniel gutzmann

At the same time, researchers conceiving of language as a tool for communication (in the broadest sense possible) have always taken expressivity to constitute one of language’s major functions (Bühler 1934; Jakobson 1960).4 Although the need for unification was perhaps in the back of the minds of a lot of people, it took until the development of the field of pragmatics, beginning sometime in the late 1960s, that (one of) the missing link(s) between formalist and functionalist approaches to meaning was forged.5 Pragmatics, of course, started out as an extremely heterogeneous enterprise, sometimes considered a “wastebasket” (Bar-Hillel 1971) or whatever “meaning minus truth conditions” (Gazdar 1979) amounts to. However, the conception that semantics focusses only on those aspects that are truth-conditionally relevant, leaving the rest to pragmatics, has been questioned.6 At least, as Kaplan (1999: 42) notes, semantics should also deal with those “nondescriptive features of language that are associated with certain expressions by linguistic convention”. As I hope to show in this survey, the empirical domain of conventional non-truth-conditional meaning proves to be very rich. And even if non-truth-conditional meaning has more or less been neglected or at least excluded from formal studies of natural language meaning from the very beginning, expressions that are associated with non-truthconditional meaning by linguistic convention have increasingly found their way into formally or analytically oriented literature since the “pragmatic turn”, and recent decades have seen considerable steps toward integrating those aspects of meaning into semantic theory. In particular, building on previous efforts by Kaplan (1999), Chris Potts (2005, 2007c) managed to develop the formal tools to analyze expressives like damn, which have been well studied since then. The following survey shall convince the skeptical reader that conventional non-truth-conditional meaning goes beyond expressives and is by no means a marginal phenomenon. Indeed, it can be found across all layers of language, from the word level down to the phonological level and

4 Concern with this aspect of language use can be traced back—at least—to Aristotle’s study of rhetoric. 5 Still one of the most useful introductions to pragmatics is the one by Levinson (1983). For further work, see, among other things, Horn & Ward 2004. 6 The most famous deviation, emerging in the 1980s, is the development of dynamic semantic theories like discourse representation theory (DRT, e.g. see Beaver & Geurts 2007; Kamp & Reyle 1993, Heim’s (1982) file change semantics or dynamic semantics see Groenendijk & Stokhof 1991).

expressives and beyond

3

up to the syntactic one, and even beyond that. That this may seem surprising to many formal semanticists only shows how deeply the focus on truth conditions is rooted in our perspective. In contrast, functionally oriented linguists may be astonished that the diversity of conventional non-truthconditional expressions is presented as something noteworthy. The fact that such content comes in so many varieties should not be any more surprising than the fact that ordinary truth-conditional meaning can be found at all levels of linguistic analysis as well, since there are no a priori reasons why the non-truth-conditional domain should not be as diverse as the truth-conditional one. Furthermore, I will try to highlight the different ways in which conventional non-truth-conditional content interacts with truthconditional meaning and, elaborating on Potts (2007c), discuss a couple of special properties of expressions that conventionally convey non-truthconditional meaning. Before going on, a few terminological remarks are in order. The phenomenon of meaning that is contributed by the conventional meaning of expressions but that nevertheless does not become part of the truthconditional content of an utterance, has been given many different names over the years. For instance, it has been called affective, colored (Frege 1897/1979), expressive (Potts 2007c), connotational, emotive (Jakobson 1960), evaluative, procedural (Bezuidenhout 2004), non-cognitive (Cruse 1986), non-descriptive, non-ideational, subjective, and, of course, also use-conditional meaning (Recanati 2004). Clearly, not all of these terms are synonymous and they are often associated with different theoretical architectures, but the empirical overlap between all these concepts is so great that in many cases these labels can be substituted for each other. However, I take it that term use-conditional meaning perfectly captures the basic essence of the kind of meaning and seems relatively neutral, in contrast to, for instance, procedural meaning, which is directly tied to the ideas of relevance theory. At the same time, it is broad enough to capture the entire range of phenomena that we will investigate in the following, in contrast to, say, emotive or expressive meaning, which seem to be more restrictive in their application. Therefore, I will stick to Recanati’s (2004) suggestion, which is also mentioned by Kaplan (1999). Accordingly, I will speak of use-conditional content and contrast it with the truth-conditional. In the same vein, expressions or constructions that contribute use-conditional content, will be referred to as use-conditional items or, for short, UCIs. I will start the survey with various lexical expressions, as the lexicon is where the most obvious cases of use-conditional meaning can be found. After that, I will present use-conditional expressions below and beyond the

4

daniel gutzmann

word level, ranging from morphological devices over syntactic constructions to intonation patterns. After the predominantly descriptive overview, I will do some analytical work by introducing some binary features that help to categorize the use-conditional phenomena discussed into five classes. First, expressions diverge with respect to the question of whether they carry only use-conditional meaning or whether they simultaneously contribute to the truth conditions of a sentence. I call this the dimensionality of a useconditional expression. The second criterion concerns what I call their functionality. By this I mean the question of whether a use-conditional expression applies to an argument or already comes saturated. The last feature is called resource-sensitivity and concerns the question of whether a useconditional item returns its truth-conditional argument or not. These three distinctions will not only be descriptive observations but will prove to provide good clues to what to look for when one develops formal analyses of use-conditional meaning. 2. The Word Level The majority of conventional non-truth-conditional meanings studied so far come from lexical items, mostly words. The UCIs at the word level can be divided into different subgroups, depending on what kind of non-truthconditional meaning they display. The resulting categorization is, of course, not clear-cut and serves primarily the function of providing structure to the following overview of the varieties of non-truth-conditional expressions. 2.1. Expressives in the Narrow Sense The group of UCIs that has received the most attention in formal semantics are what I refer to as expressives in the narrow sense, i.e. expressions that express some emotional and evaluative attitude with a high degree of affectedness. Thanks to the work of Potts (2005, 2007b,c, 2012), who picks up the theme as laid out by Kaplan (1999), expressives have received a lot of attention during the last few years. Standard examples include pejorative epithets and attributive adjectives.7

7

ples.

Throughout this text, I will use bold face to highlight relevant aspects of the exam-

expressives and beyond (1)

Epithets a. That bastard Kresge is famous. b. That idiot Kresge dropped the bottle again.8

(2)

5

(Potts 2007c: 168)

Expressive attributive adjectives a. I hear your damn dog barking. b. My friggin’ bike tire is flat again.

(Potts 2005: 18)

Both epithets and expressive attributive adjectives contribute use-conditional content and no truth-conditional meaning. From a truth-conditional point of view, such expressions are therefore optional. Adding or omitting them does not alter the truth conditions of a sentence. Accordingly, all the variants of (2a) in (3) are truth-conditionally equivalent.

(3)

⎧ ∅ ⎫ { } { blasted } { } I hear your ⎨ bloody ⎬ dog barking. { damn } { } { } ⎩ fucking ⎭

Following Cruse (2004: 57), I call such expressives and other UCIs that do not contribute anything to the truth-conditional dimension of meaning expletive UCIs.9 Of course, the expressive attitude conveyed by expletive UCIs like damn or bastard is lost if they are omitted; or a different emotion may be displayed when you use awesome instead of damn, but the truth-conditional content remains unaffected. I will write this informally in a fraction-like fashion with the use-conditional content on top of the truth-conditional one.10 (4)

I hear your damn dog barking =

damn dog I hear your dog barking

The entire meaning of (2a) consists of its truth conditions that equal the ones of the variants in (3)—that the speaker hears the addressee’s dog barking, plus whatever use-conditional content is expressed by applying damn to dog, e.g. that the speaker has a negative attitude towards the addressee’s dog. 8 The use of the demonstrative that instead of the plain definite determiner enhances the expressive nature of the epithet, cf. Lakoff 1974; Potts & Schwarz 2010 for the expressive function of demonstratives in English and Section 2.3 for further expressive uses of pronouns. 9 Such semantic expletives must not be confused with syntactic expletives like it in It’s raining, which are semantically empty but are syntactically obligatory. 10 For the purposes of this survey, this kind of representation serves only illustrative purposes and has no theoretical implications, even if the majority of the approaches to use-conditional content take such multidimensionality seriously (Potts 2005).

6

daniel gutzmann

Further examples of expressives in the narrow sense are interjections (Ameka 1992) such as ouch and oops as famously discussed by Kaplan (1999). In addition, some otherwise truth-conditional expressions like man or the already expressively loaded shit can be used as expressive interjections. Expressive adjectives like damn regularly can be used interjectively as well.11 (5)

Interjections a. b. c. d. e. f.

Ouch, I’ve hit my thumb! Oops! Oh, I have another suit. It’s hot, man. Shit, I’ve lost my keys! Damn, I’ve lost my keys!

(Kaplan 1999) (Kaplan 1999) (Ameka 1992) (McCready 2009)

Like expressive adjectives and epithets, interjections like these are expletive UCIs and do not add anything to the truth conditions of the sentence. However, what makes many interjections interesting is that they do not seem to interact with the truth-conditional content at all (man being an exception, see below). In contrast to the expressives discussed above, interjections do not need a truth-conditional argument—they are already saturated and convey an use-conditional attitude without further ado. Therefore, just as interjections can be omitted without effect on truth-conditions, the rest of the sentence can be dropped, leaving the use-conditional attitude intact. (6)

a. b. c. d.

Damn! Ouch! Oops! Oh!

That is, even without any truth-conditional content, the examples in (6) express attitudes of anger, pain, awkwardness, or surprise respectively. Semantically, they are more isolated from the rest of the sentence than expressive adjectives or epithets are. This is mirrored by the syntactic fact that they appear only in peripheral positions. (7)

a. *I’ve lost my oops keys. b. *It’s man hot!

Following Potts (2005: 65), I call such examples isolated UCIs. In contrast to the informal description of the meaning of a sentence containing an argument-seeking expressive like damn given in (4), the composition for

11

There are also interjections like by god that transcend the “word level.”

expressives and beyond

7

(5a) looks like (8). That is, the truth-conditional meaning of a sentence containing an already saturated, isolated interjection consists of the semantic content of the sentence without the interjection, while its use-conditional part is given solely by the use-conditional content of the interjection itself. (8)

Ouch, I’ve hit my thumb =

I’m in pain I’ve hit my thumb

A further group I want to subsume under the caption of expressives in the narrow sense are what could be called expressively coloured expressions (after Frege’s Färbung “colouring”). These kinds of UCIs differ crucially from the expressions discussed so far. They are lexical items that have an ordinary truth-conditional denotation but in addition, have a use-conditional component that displays some (in most cases negative) attitude towards the denotation. A classic example comes from Frege (1897/1979: 140): (9)

Coloured expression a. This dog howled the whole night. b. This cur howled the whole night.

The difference between dog and cur is that while the former is expressively neutral and just refers to the set of dogs, the latter additionally expresses a negative attitude towards members of the set or the set as a whole. A systematic set of expressively coloured expressions are ethnic slurs. By this label, I mean expressions that, besides denoting some kind of nationality or ethnic group, convey a derogative racist attitude.12 As examples, I use the antiquated Boche and Kraut, both being derogative variants of German. (10) Ethnic slurs a. Lessing was a Boche. b. Hitler was a Kraut.

(Williamson 2009: 149) (Saka 2007: 39)

The composition of coloured items is therefore very different from all the other kinds of expressions discussed in this section so far, because in contrast to expletive UCIs, they do make a truth-conditional contribution. In the case of Kraut, the truth-conditionally relevant part equals that of German, and for cur, it equals that of dog. Following McCready (2010), I will call UCIs that conventionally contribute both truth-conditional and non-truthconditional meaning mixed UCIs.

12

Amongst many others, cf. Green & Kortum 2007; Hom 2008, 2010; Hornsby 2001.

8

daniel gutzmann

As shown above, expletive UCIs can be omitted from a sentence without altering its truth-conditional content. Since they also contribute truthconditional content, this does not hold for mixed UCIs. The following example illustrates this point.13 (11)

a. That Kraut Lessing wrote a lot of books. → Lessing was a German. b. Lessing wrote a lot of books. ↛ Lessing was a German.

The sentence (11a) containing that Kraut implies that Lessing was a German. If the mixed UCI is omitted as in (11b), that entailment is lost, since the property of being German is contributed by the truth-conditional dimension of Kraut. If you want to get rid of the negative attitude conveyed by ethnic slurs or another expressively-coloured expression without altering the truth conditions of the sentence, you have to substitute the racist slur by the corresponding neutral expression. (12) a. Lessing was a German. b. Hitler was a German. c. This dog howled the whole night.

What the truth-conditionally equivalent, but expressively neutral, expression for a mixed UCI is, cannot be directly read off from the expression. This has to be encoded in the lexicon.14 The informal schema used above to illustrate the composition of the meaning of a sentence containing UCIs therefore needs to rely on lexical knowledge for coloured expressions. For (10) for instance, the use-conditional part of the sentence consists of the negative attitude expressed by it, while the truth-conditional part corresponds to the same sentence with German substituted for Kraut. (13) Lessing was a Kraut =

Generally, I don’t like Germans Lessing was a German

13 Note that we have to change the example to one in which the mixed UCI is used attributively, because otherwise, omitting it would render the sentence ungrammatical.

(i)

*Lessing was a. 14

Diachronically, ethnic slurs may have some connection to what they denote truthconditionally. Arguably, many slurs come into existence by some kind of metaphorical or metonymical process by which something that is in some relation to the intended referent— like the Kraut eaten by Germans—is used to refer to it, thereby conveying some expressive attitude. For the expressive power of metaphorical and metonymic transfers and their role in semantic change, cf. amongst others Claudi & Heine 1986; Traugott & Dasher 2001.

expressives and beyond

9

This already shows some interesting facts about the semantic composition of mixed UCIs, namely that while their truth-conditional component may fall under the scope of some semantic operator—like the past tense in the example—its use-conditional component does not (Potts 2005, 2007c). That is, whereas the truth-conditional part of Lessing was a Kraut means that there is some time prior to the utterance time for which Lessing is a German is true, the negative or jocular attitude against Germans displayed by Kraut is not evaluated with respect to that point in the past but is attributed to the utterance time and speaker. The same holds for other semantic operations like negation or questions (Cruse 2004: 57). Even the negated variant of Lessing was a Kraut and the corresponding question convey the anti-German sentiment. This is shown by the fact that the following discourse continuations are impossible. (14) a. Descartes was not a Kraut. #But I like Germans. b. A: Was Descartes a Kraut? B: #No, Germans are nice.

Using again the informal fraction notation, the negation and question operators only show up at the lower truth-conditional level but not on the uclayer on top of it. (15) Descartes was not a Kraut = (16) Was Descartes a Kraut? =

Generally, I don’t like Germans ¬(Descartes was a German)

Generally, I don’t like Germans ?(Descartes was a German)

This feature may be called scopelessness (Potts 2005: 41) and it is the reason why ethnic slurs cannot be denied by a simple negation or used in a question without unfolding their offending content. This corresponds to a feature that Potts (2007c: 167) calls immediacy, by which he means that “like performatives, expressives achieve their intended act simply by being uttered”. A further observation to be made is that the coloured expressions discussed so far can be regarded as being isolated UCIs as well. On the one hand, they are of course more integrated than expletive UCIs because they also contribute to the truth-conditional tier and hence cannot be omitted without affecting the truth conditions or grammaticality of a sentence. But, on the other hand, their uc-content is isolated because the negative attitude does not apply to a specific argument in the sentence. That is, while the truth-conditional part of Kraut predicates over Lessing in the example, the negative attitude does not apply to Lessing, but to Germans in general (cf. McCready 2010). Using an informal paraphrase of Kraut to make this explicit, we have the following characterization of the sentence’s truth- and use-conditional content.

10

daniel gutzmann

(17) Lessing was a Kraut=

Generally, I don’t like Germans Lessing was a German

A negative attitude towards Lessing is not directly expressed by the ethnic slur. However, it can be inferred if the two levels of meaning are taken together. This is supported by the observation that a negative attitude towards Lessing can be cancelled, whereas this is not possible for the negative evaluation of Germans. (18) a. Lessing was a Kraut, but he was fine guy. b. #Lessing was a Kraut, but generally, I like Germans. c. #Generally, I like Germans, but Lessing was a Kraut.

Before we leave the descriptive survey of expressives in the narrow sense, let me briefly mention a problem that holds for many of those expressions. As we have seen, a UCI like damn can be used as an expressive attributive adjective or as an expressive interjection. In the same vein, many UCIs are multifunctional. For instance, I have characterized fucking and bloody as expressive adjectives, in which case they are expletive UCIs that contribute nothing to the truth-conditional level of meaning. However, many expressive adjectives can also be used to modify or intensify another adjective, as discussed by Geurts (2007) and Morzycki (2011). (19) Rufus is {

fucking } tall. goddamn

In such environments, fucking or goddamn cannot be considered as expletive but as mixed UCIs, because they make a contribution to the truthconditional dimension by grading the adjective they modify. Therefore, fucking cannot be omitted without changing the truth-conditional content. This is illustrated by the following examples. (20) a. Ringo is tall, but Rufus is fucking tall. b. #Ringo is tall, but Rufus is tall.

To study the relations between different functions of expressives and, more generally, UCIs in order to find generalizations regarding their shiftability is an interesting research question that, to my knowledge, has not been addressed so far. 2.2. Particles Besides all the different types of expressives in the narrow sense, there are many classes of expressions in different languages that arguably contribute to the use conditions of an utterance instead of affecting its truthconditional meaning.

expressives and beyond

11

An entire part of speech that seems to have some kind of affinity to the use-conditional domain are particles.15 This is illustrated by the fact that five out of the eight contributions to this volume deal with particles. In general, many of the different kinds of particles found around the world’s languages do not have any influence on the truth conditions of a sentence but, rather, impose appropriateness conditions on their use. For instance, modal particles in German have been regarded as conveying non-truthconditional meaning since the earlier functional studies (Helbig 1977; Weydt 1969) and even in a formal semantics framework, there are some early attempts to relate them to use-conditional meaning (Kratzer 1999). German modal particles are a small, more or less closed set of specific lexical items that convey information about the discourse participants’ beliefs and attitudes towards the propositional content. For instance, in rough approximation, wohl expresses that the speaker merely assumes that the propositional content is true and ja roughly conveys that the hearer may already know the proposition.16 (21) a. Hein ist wohl auf See. Hein is mp at sea ‘(As I assume) Hein is at sea.’

(Zimmermann 2004: 543)

b. Webster schläft ja. Webster sleeps mp ‘(As you may know) Webster sleeps.’

(Kratzer 1999: 4)

Like expressive attributive adjectives or epithets, modal particles are expletive UCIs, as they are optional and leaving them out does not alter a sentence’s truth condition. That is, all the variants in (22) are true iff Webster is sleeping regardless of what kind of attitude is conveyed by the modal particles. ⎧ ∅ ⎫ { } { doch } { } (22) Webster schläft ⎨ halt ⎬. ‘Webster is sleeping ∅/MP.’ { ja } { } { } ⎩ wohl ⎭

15 I use the term in a non-technical sense here. Regarding particles as a formal category is not without problems, cf. Zwicky 1985. 16 In order to reconcile the clause-medial position of German modal particles with their taking scope over the entire proposition, Zimmermann (2004) postulates LF-movement while Bayer & Obenauer (2011) rely on an Agree operation.

12

daniel gutzmann

While all being true if Webster is sleeping, the variants in (22) are, of course, not appropriate in the same contexts. For instance, using ja in a context like in (23) in which the hearer cannot reasonably be assumed to have already known what the speaker is telling her renders sentence (23a) infelicitous. In the same context, the utterance becomes felicitous if the modal particle is left out, as shown in (23b). (23) [Context: A happy father rushes out of the delivery room] a.#Es ist ja ein Mädchen! It is mp a girl ‘It’s a girl!’ b. Es ist ein Mädchen! It is a girl ‘It’s a girl!’

Furthermore, modal particles mirror expressive attributive adjectives and epithets insofar as they are not saturated and hence not as isolated as the interjections discussed in the previous section. What distinguishes modal particles from the attributive use of damn and the like is that they take the entire propositional content of the sentence as their argument.17 (24) Webster schläft ja =

ja(Webster is sleeping) Webster is sleeping

Modal particles exhibit a lot of interesting syntactic and semantic features and, as I argued elsewhere (Gutzmann 2009, 2012), many of them can plausibly be derived from their use-conditional character. The contributions to this volume by Döring (2013), Egg (2013), and Repp (2013) all address various issues connected with this class of particles. Of course, German modal particles are not the only class of use-conditional particles. There are far too many cases to go through all of them here. But for two further examples, consider the following Japanese particles

17 This is, however, arguably also possible for expressive adjectives like damn, even if they still take a DP as their argument in the syntax.

(i)

I’ve spilled that damn bottle again.

The most natural reading of (i) is one in which the speaker has a negative emotion regarding his spilling of the bottle, not regarding the bottle or bottles in general. Note that there are usages of some modal particles that syntactically seem to take a more narrow scope over a DP. (ii)

der wohl größte Bankenskandal aller Zeiten the mp biggest bank.scandal of.all time

expressives and beyond

13

which are studied by Sudo (2013) [this volume] and McCready & Takahashi (2013) [this volume] respectively. (25) Japanese particles a. ima ame futteru no? now rain is.falling part ‘Is it raining now?’ b. A: dooshite ookikunat-tara pairotto-ni naritai no? why get.big-when pilot-acc want.become q ‘Why do you want to be a pilot when you grow up?’ B: datte, kakkoii mono. come.on cool mono ‘Well, because it’s cool.’

In his contribution, Sudo (2013) [this volume] discusses the Japanese question particle no in polar questions. Adding the particle changes the bias involved in a positive polar question entirely, but not its truth-conditional content. While the positive polar question without a particle does not presuppose any positive evidence, using no imposes such a requirement on a felicitous utterance of the question.18 (26) Positive polar questions with -no a. Neutral Context: We’re looking for a left-handed person. I’m wondering about John, who is not around. #John-wa hidarikiki-na no? John-top lefty-cop q ‘Is John lefty?’ b. Negative Context: My friend has just entered our windowless office wearing a dripping wet raincoat. #ima hareteru no? now sunny q ‘Is it sunny now?’ c. Positive Context: Same context as (26b). ima ame futteru no? now rain is.falling q ‘Is it raining now?’

18

This can be related to the bias induced by “rising declaratives” (Gunlogson 2003).

14

daniel gutzmann

Another case of a Japanese particle that conveys non-truth-conditional content is mono, illustrated in (25b) and examined in the paper by McCready & Takahashi (2013) [this volume]. As they argue, mono is a mixed UCI that conventionally contributes truth-conditional as well as use-conditional content. At the truth-conditional tier, it has the meaning of a causal connective, roughly equivalent to English because. In addition, it carries the use-conditional content that the speaker is not neutral about the propositional content expressed by the proposition mono attaches to, but that she is personally affected by it. In this respect, mono resembles Grice’s (1975) decomposition of therefore into a conjunctive and causal part. 2.3. Pronouns UCIs are not only attested in the specialized class of particles. Besides expressive adjectives and nouns, which are almost always expressives in the narrow sense, pronouns can also carry non-truth-conditional meaning in many languages. The prototypical example of pronouns that have a use-conditional function can be found in languages that have a distinction between formal and familiar pronouns. This is, amongst many others, the case in German and French. (27) Formal vs. familiar pronouns a. Ich rufe dich/Sie an. I call you.familiar/formal on ‘I’ll give you a call.’ b. Tu es / Vous êtes soûl. you.familiar are / you.formal are drunk ‘You are drunk.’

(Potts 2007c: 190)

(Horn 2007: 49)

Like expressively-coloured nouns, such pronouns are mixed UCIs. On the truth-conditional layer, their meaning is just their referent, that is, the addressee of the context in this case. Hence, they cannot be dropped entirely as expletive UCIs. The distinction between formal and familiar resides on the use-conditional layer. Choosing the wrong pronoun can never make an otherwise true sentence false, but it may result in a high degree of social infelicity. (28) du =

informal relationship between speaker and hearer the addressee

Another case of personal pronouns contributing use-conditional meaning are free personal dative pronouns that can have a use-conditional function in some languages. For English, Horn (2008; 2013 [this volume]) argues that

expressives and beyond

15

dative pronouns may have such an uc-use like in (29a). Other languages which have a more systematic system of free datives, like German (Gutzmann 2007; Lambert 2007; Wegener 1989) or Hebrew (Borer & Grodzinsky 1986), exhibit a pattern known as ethical dative. (29) Personal datives a. I want me an iPod. b. Dass du mir ja nicht zu spät kommst. that you me.dat mp not too late come ‘Don’t you be late.’

(Horn 2008: 175)

(Lambert 2007: 5)

c. hem kol ha-zman mitxatnim li they all the-time marry to.me ‘They are getting married on me all the time (and it bothers me)’ (Borer & Grodzinsky 1986: 179)

Common to these three free personal datives is that they all express some affection of the speaker towards the fact or event described by the sentence. That is, the speaker of (29a) expresses that she is somehow affected by her wanting an iPod. In a similar vein, the speaker (29b) expresses that she has some personal interest in the hearer not being late (Gutzmann 2007: 277). In contrast, the Hebrew ethical dative in (29c) expresses the speaker’s negative affection with all the marrying. However, all this is expressed solely in a use-conditional way. The presence of the personal datives does not alter the truth-conditional content of the sentence. They are all instances of expletive UCIs and could be dropped without any change in grammaticality or truth-conditional meaning. (30) I want me an iPod =

I am affected by wanting an iPod I want an iPod

Like modal particles, the free datives are not isolated since they take the propositional content of the sentence as their argument. 3. UCIs beyond and below the Word Level Up to this point, every UCI we have presented has been a single word. However, there are also use conditions beyond and below the word level. By this, I mean use-conditional content that stems from intonation, syntactic constructions, morphological operations, the pragmatics of speech acts, or even from orthographic devices.19 All the examples I discuss in the following 19

Another source for use-conditional meaning may be speech-accompanying gestures

16

daniel gutzmann

show that the phenomenon of use-conditional content is widespread and not restricted to some fancy words. In fact, it pervades every layer of natural language. 3.1. Use-Conditional Intonation Intonation may be one of the most prominent and obvious means of conveying use-conditional meaning.20 By intonation, one can express all kinds of emotions and attitudes, ranging from joy to anger, interest and boredom.21 Furthermore, intonation can be used to signal special communicative functions, like irony, sarcasm, or hyperbole, that may guide the pragmatic interpretation of a sentence. Used to indicate emotions or other rhetorical means, intonation is said to be paralinguistic, as it sits on top of the ordinary linguistic signs without being a sign in itself. However, there are cases in which intonation actually is a genuine part of the language system and is reflected in the structure of grammatical constructions and their semantic interpretation. For instance, intonation has linguistic impact in focus-sensitive constructions. A focus particle like only associates with an accented expression yielding different readings for different placements of the focus accent. (31) a. Piet only wears a pink tie at work. b. Piet only wears a pink tie at work.

While (31a) is falsified if Piet wears a non-pink tie to work, (31b) allows Piet to wear any tie at work as long as he does not wear a pink tie anywhere else than at work. Different accents lead to different semantic interpretations. Focussensitive construction like these are, by far, the best-studied phenomena in which intonation cannot be merely paralinguistic but must be reflected in the syntactic or semantic representation of a sentence. However, there is still some dispute on which meaning-level the difference between (31a) and (31b) is located.22 Some of the various approaches to the meaning of (Ebert et al. 2011). Interestingly, many UCIs share important properties of gestures, for instance the nondisplaceability which is typical of gestural communication; cf. the discussion below in Section 5.2. 20 Cf. for instance Ladd (1990), who reviews earlier work by Bolinger. “[T]he unifying idea of B’s work is […] the general claim that intonational features, including accent placement, are beyond grammar and are directly linked to emotion.” (Ladd 1990: 806) 21 For an overview over the linguistic encoding of emotions, cf. Fries 2007, 2009. For general considerations regarding the role of emotion in language, cf. Jay & Janschewitz’s (2007) reply to Potts 2007c. 22 For approaches to the syntax and semantics of focus sensitive particles and various

expressives and beyond

17

focus are couched in use-conditional rather than truth-conditional terms. For instance, Kratzer (2004) proposes that what seems like the presupposition of the backgrounded material may in fact be rendered more adequately as use-conditional meaning. In his contribution on the expression of surprise, Zeevat (2013) [this volume] argues for an analysis that combines elements from presuppositional and use-conditional approaches, and tries to derive the main properties of focus particles like only from their emotive component. In any case, it is clear that, in the absence of any focus-sensitive expression, the focus accent does not contribute truth-conditional content but, rather, influences the conditions on the felicity of an utterance. This is most obvious in the case of question-answer pairs. (32) A: Who likes Bruce? B: #Rachel likes Bruce. Bʹ: Rachel likes Bruce.

In the context of the question in (32), an utterance of (32) is infelicitous since it bears the focus accent at a given constituent (Schwarzschild 1999). If the accent pattern is changed to the expected one in (32Bʹ), the sentence can be felicitously uttered. Besides the use conditions imposed by focus accentuation, there are some other cases of intonation that—unlike the patterns signaling delight or sarcasm, for instance—show some reflections in the grammar. Take, for instance, the intonation pattern that can be called exclamative or unexpectedness intonation (Castroviejo Miró 2008) that is typical for exclamative sentences across different languages (d’Avis 2002; Rett 2008; Zanuttini & Portner 2003). (33) Exclamative or unexpectedness intonation a. How tall Michael is! b. Wie groß Michael ist!

Exclamative sentences like these express that the speaker is surprised or astonished about the degree of Michael’s height. That unexpectedness intonation (“ui” henceforth) is really part of the language system of English and German is not only shown by the fact that its realization is very constant from a phonological point of view (Oppenrieder 1989), but that it is

focus sensitive constructions, cf. amongst many others, Altmann 1976; Beaver & Clark 2008; Büring & Hartmann 2001; Jacobs 1983; Horn 1969, 1996; König 1991; Kratzer 2004; Rooth 1985, 1992; Schmitz 2008; Sudhoff 2010.

18

daniel gutzmann

also directly reflected in their grammars. They are introduced by a word or phrase which usually introduces a question and therefore licenses the inversion of the subject and the verb in (standard) main clauses. (34) a. How tall is Michael? b. Wie groß ist Michael?

In how-exclamatives however, no inversion takes place. Rather they look like embedded questions. Without unexpectedness intonation, the sentences become ungrammatical as they are either exclamatives lacking the needed intonation or questions lacking the needed inversion.23 (35) a. *How tall Michael is. (without ui) b. *Wie groß Michael ist. (without ui)

What is special about how-exclamatives is that they only have use-conditional meaning but no truth-conditional content. Utterances like (33) are not a statement about or an assertion of Michael’s height. For instance, they cannot serve as an answer to a query about Michael’s height. In contrast, an assertion that Michael has a surprisingly high degree of tallness is perfectly fine. (36) A: How tall is Michael? B1: #How tall Michael is! B2: Surprisingly tall.

Furthermore, the unexpectedness is also not part of the truth-conditional content. This is shown by the fact that it can neither be denied nor even affirmed. (37) A: How tall Michael is! B1: #That’s not true. I don’t think this is unexpected at all. B2: #You’re completely right. That is unexpected.

In how-exclamatives, no truth-conditional content seems to be left behind. If this is the case, there would be an empty lower level in the informal fraction notation. (38) How tall Michael is! =

It is unexpected how tall Michael is ∅

23 Of course, both sentences are grammatical, if they are interpreted as a special kind of question, for instance as a kind of deliberative or reflective question or as an echoic clarification question. Furthermore, both sentences are possible in book titles or the like, in which case they are akin to embedded questions.

expressives and beyond

19

This raises the question of whether ui or the corresponding semantic operator it realizes is different from all the other UCIs presented so far. If it is, then ui is a new kind of expressive insofar as it affects the truth-conditional content. Informally, it takes its truth-conditional argument with it to the use-conditional layer and does not leave it, unlike all the other UCIs. In a paper on various kinds of use-conditional content, McCready (2010) calls such items shunting UCIs. However, it might be the case that the impression that how-exclamatives have no truth-conditional content is only superficial. An argument for this comes from exclamatives that have a more declarative structure instead of being introduced by how (Castroviejo Miró 2008). (39) Obama won the Nobel Prize!

In contrast to how-exclamatives, declarative exclamatives can be used to assert their propositional content. (40) A: What happened? B: Obama won the Nobel Prize!

Like in how-exclamatives, ui is not part of the truth-conditional content of a declarative exclamative either. That is, they do not make an assertion about the speaker’s attitude regarding the propositional content. (41) A: How do you feel about Obama’s current situation? B1: #He won the Nobel Prize! B2: I’m surprised that he won the Nobel Prize.

The behavior of ui in declarative exclamatives thus parallels that of other expletives like modal particles, and accordingly, there are two levels of meaning involved in declarative exclamatives. (42) Obama won the Nobel Prize! = It is unexpected that Obama won the Nobel Prize Obama won the Nobel Prize

Given this analysis of declarative exclamatives, it can be argued that ui does makes the same contribution in how-exclamatives. That the latter cannot be used to assert anything must then follow from the fact that their truth-conditional content corresponds to a question meaning and is thus not suitable for being asserted. For an elaboration of this argument and a formalization, see Castroviejo Miró (2008), from whom I have borrowed the arguments in this short discussion.24 24 From the conclusion that ui is not a shunting UCI, it does not, of course, follow that there are no such items, cf. Section 4 below.

20

daniel gutzmann

Another kind of intonation that does not alter the truth-conditional content of a sentence and which could be analyzed as contributing useconditional content is verum focus. As coined by Höhle (1992), the term verum focus refers to a special kind of non-contrastive focus in German that is realized on the finite verb or a complementizer, both located in C in German. Informally, the contribution of verum focus is that it puts emphasis on the propositional content it scopes over. (43) Verum focus A: Peter is supposed to have written a book. B: Peter hat ein Buch geschrieben. Peter has a book written ‘Peter has indeed written a book.’

(Höhle 1992)

Höhle (1992: 112) assumes that verum focus is a means to realize a semantic operator he calls verum. The working paraphrase he uses as the meaning of the verum operator in his paper (Höhle 1992: 112) is simply that of a matrix sentence that states that the embedded proposition is true. verum is not restricted to German. Cross-linguistically, it can be realized in many different ways. As we have seen, verum is realized by verum focus in German. In contrast, both in English and in Spanish, we find special kinds of lexical insertions to instantiate the verum operator (Gutzmann & Castroviejo Miró 2011). (44) A: B1: B2: B3:

I wonder whether Carl has finished his book. Karl hat sein Buch beendet. (German → verum focus in C) Carl did finish his book. (English → do insertion) Carlos sí acabó su libro. (Spanish → sí insertion)

The verum operator realized by these different constructions behaves like other expletive UCIs as it does not alter the truth-conditional content of the sentence. All the answers in (44) are true if the person in question has finished his book. The contribution of verum is in the use-conditional layer. Its argument is the propositional content of the sentence. (45) Carl did finish his book ≈

It is true that Carl finished his book Carl finished his book

Things are not that straightforward with verum since it is also possible in a variety of non-declarative sentence types (cf. Höhle 1992 for an overview). This brings up the interesting question of how verum interacts with the sentence mood of the different sentence types in which it is allowed. For instance, in a polar question with verum (Romero 2005; Romero & Han 2004), the use-conditional meaning of the interrogative cannot simply be that the content of the question is true, as in (47).

expressives and beyond (46) Verum focus in polar question

21 (Höhle 1992: 112)

A: I have heard that Carl kicked the dog. B: HAT er den Hund denn getreten? has he the dog mp kicked? ‘Has he kicked the dog?’

(47) Has he kicked the dog? ≠

Is true that he kicked the dog? Has he kicked the dog?

Verum focus and verum are therefore an interesting subject to study the interaction between UCIs and sentence mood.25 Furthermore, as a means of information structure, verum focus is governed by the discourse structure and can therefore also provide insights into the dynamics of use-conditional meaning. In her detailed study on the discourse function of verum focus, Repp (2013) [this volume] shows that verum interacts in delicate ways not only with the information structure of a sentence, but also with modal operators, negation, and modal particles. I have used unexpectedness intonation and verum focus as two phenomena of intonation that conventionally express use-conditional content. However, there are certainly many others. Arguably, the alternative invoking function of ordinary information focus could be understood as being use-conditional. Only if there are focus sensitive operators like only present in the sentence, does it have an impact on the truth-conditional content of a sentence. See Zeevat’s (2013) [this volume] contribution for elaboration. 3.2. Use-Conditional Syntax Besides intonation, another way to alter the use conditions of a sentence without affecting its truth-conditional content is by using certain, often non-canonical syntactic structures. Many of those are associated with specific functions of information structuring which commonly have no influence on the truth conditions of a sentence. Of the syntactic constructions that do not affect the truth conditions of a sentence, non-restrictive relative clauses as in (48a) and other supplements like as-appositives (48b), nominal appositives (48c), and parentheticals (48d) are the most prominent ones in the literature on multidimensionality in semantics (Jayez & Rossari 2004; Nouwen 2007; Potts 2002, 2005).

25 Sentence mood itself has been argued to contribute use-conditional content. See Stenius 1967 for an early outline, and Portner 2007 or Gutzmann 2012 for more recent implementations.

22

daniel gutzmann

(48) Appositives & parentheticals a. b. c. d.

Ames, who was a successful spy, is now behind bars. (Potts 2005: 90) Ames was, as the press reported, a successful spy. Ames, a former spy, is now behind bars. Ames—and you will never believe this—is now behind bars.

All these kinds of unintegrated syntactic supplements are semantically independent of the rest of the sentence. Using them leads to a very intuitive form of semantic multidimensionality as there are two sentences presented as one. The contents of these two sentences are transparently distinct from each other, except for the anchor of the supplement. (49) Ames, a former spy, is now behind bars =

Ames is a former spy Ames is now behind bars

However, even if all the phenomena in (48) involve straight multidimensionality, they differ crucially from other UCIs. In contrast to them, it is not obvious that supplements convey conventional non-truth-conditional content. Instead, it is reasonable to assume that they primarily have truthconditional content. While it may be mistaken in the first place to ask what the truth conditions for the interjection ouch or the modal particle ja are, this is an easy question when it comes to appositives and their kin. For a supplement, we can easily give the truth conditions if it is applied to its anchor. For instance, the content of the nominal appositive a former spy combined with its anchor Ames is given by the top layer in (49). Obviously, this is true, if Ames is a former spy. In the same vein, the as-appositive in (48b) is true, if the press reported that Ames was a successful spy. That is, what we have on the top layer in (49) is truth-conditional rather than use-conditional. This is also shown by the possibility of denying the content of an appositive, even if doing this is not as straightforward as for the primary, asserted content. (50) A: Ames, a former spy, is now behind bars. B: Yes, however, he was not a spy but a corrupted politician.

However, the truth-conditional content of the supplement is independent of the truth-conditional content at the lower level. We can judge whether it is true that Ames was a former spy independently of the question of whether he is now behind bars, and vice versa. Informally, we can thus give the following truth conditions for (48c). (51)

Ames was a former spy true, iff Ames was a former spy = Ames is now behind bars true, iff Ames is now behind bars

That is, even if appositives and other syntactic supplements express truthconditional instead of use-conditional content, they mirror expletive UCIs

expressives and beyond

23

insofar as they convey their content independently of the content of the main clause, thereby leading to multidimensional content. There are other syntactic constructions that, unlike the supplements in (48), are more likely to contribute use-conditional instead of truth-conditional content. Take, for instance, topicalization (Birner & Ward 1998; Frey 2010). (52) Topicalization John, Mary loves.

Whatever the exact use conditions imposed by topicalization may be,26 it is clear that they do not have any influence on the truth conditions of the sentence, which are the same as the truth conditions of the corresponding sentence with canonical word order. That is, both (53a) and (53b) are true if Mary loves John. (53) a. John, Mary loves. b. Mary loves John.

The use-conditional contribution of topicalization has to be given in information structural terms. Of course, giving a paraphrase for such functions means leaving aside many of the details of the use conditions for topics. But for the sake of illustration, let us adopt Portner’s (2007: 418) formulation that the speaker’s “mental representation” of the topical element is active. We thus have the following truth and use conditions for (52): (54) John, Mary loves =

The speaker’s mental representation of John is active Mary loves John

That the contribution of topicalization puts such a requirement on the discourse context can be illustrated by the following examples. (55)

[Rushing into the room:] #John, Mary loves.

(56) A: What’s up with Mary? B: #John, Mary loves.

In (55), there is a neutral context in which there is nothing active in the mental representation. In (56), A asks a question about Mary and thereby activates B’s mental representation of Mary but not of John. Therefore, topicalizing John is infelicitous in such contexts. 26 Büring (1997, 2003), for instance, provides accounts of topics in general, relating them to the management of the question under discussion. For various use-conditional ways to manage the common ground, cf. Repp 2013 [this volume].

24

daniel gutzmann

That the semantics of topicalization behave very differently from that of appositives can be shown by trying to deny them, as in (50) for appositives. As expected, this is impossible as it is for the other UCIs. (57) A: John, Mary loves. B: #Yes, but I haven’t thought about John.

Apart from topicalization, there are many more non-canonical syntactic constructions that may be used to expressively convey something about the discourse role of a moved constituent, like right and left dislocation or “freetopic” constructions in German (cf. e.g., Altmann 1981). The so-called preprefield in German, i.e., the position before the ordinary first position before the main verb in non-subordinated clauses, seems to be systematically associated with the function of framing and integrating the truth-conditional content of the sentence into the discourse structure, thereby affecting the use conditions of the utterance. 3.3. Use-Conditional Morphology The UCIs discussed so far are all lexical words or—in case of intonation of syntactic constructions conventionally associated with UC-meaning— something beyond or above the word level. But use conditions can also be imposed below the word level on a regular basis, as many languages employ a subsystem that may be called expressive morphology (Bauer 1997; Stump 1993; Zwicky & Pullum 1987).27 Even if use-conditional morphology, as I prefer to call it, is “associated with an expressive, playful, poetic, or simply ostentatious effect of some kind” (Zwicky & Pullum 1987), it is still part of the language system, as it is governed by regularities, even if the rules for use-conditional morphology may differ from what Zwicky & Pullum (1987) call plain morphology. In the following, I will present a small selection of the phenomena that can be found in this huge terrain. To keep the presentation focused, I will concentrate on use-conditional phenomena in the morphology of German (cf. Dressler & Merlini-Barbaresi 1994 for an overview). One kind of phenomenon of use-conditional morphology that is common across many languages is expressive derivation. In particular, diminutive suffixes are capable of conveying affective, use-conditional meaning when used with expressions referring to persons. For instance, there is systematic usage of the suffix -i in German to derive affective nicknames from ordinary proper names. 27

Alternative notions are evaluative or affective morphology.

expressives and beyond

25

(58) Expressive diminutives Guten Morgen, Hans-i good morning Hans-dim.familiar ‘Good morning, Hans.’

Like the familiar second-person pronoun du discussed in (27), the suffix -i expresses a familiar relationship. In contrast with the pronoun, this relationship does not necessarily involve the speaker and the addressee but the speaker and the referent of the stem to which the suffix attaches.28 The expression of familiarity does not alter the truth conditions of a sentence. But, unlike the familiar pronoun du, expressive -i arguably has no truthconditional content at all. Accordingly, we can give the following representation for -i suffixed to a proper name like Hans. (59) Hans-i =

familiar relationship between the speaker and Hans Hans

Another instance of use-conditional morphology as discussed in the paper by Bücking & Rau (2013) [this volume] is a morpho-syntactic construction involving non-inflected verbal stems. These verb forms can be used— primarily in electronic chat or forum texts—to transform the meaning of the verbal stem into a kind of performative. For instance, the verb stem grins-‘smile’ in (60a) counts as a substitute for performing the act of smiling. (60) Non-inflected verbs a. grins smile.stem b. dich in den Arm nehm you.acc in the arm take.stem

As in (60b), such non-inflected verbs need not to be bare forms, but can also take further arguments. That shows that it is not a purely morphological phenomenon. Interestingly, it can take the same arguments as its inflected forms.29 That is, both constructions in (60) are used by the speaker to substitute for the action expressed by the construction in a remote conversation, 28 Of course, this still may be the addressee as in (58), where -i is affixed to the vocatively used proper name Hans. 29 The subject is an exception, as it is hardly ever realized, which Bücking & Rau (2013) [this volume] show in their paper. In this respect, the non-inflected constructions mirror infinitives, to which the authors relate them. For a thorough treatment of (adult) “root infinitives,” and speculations on why adding a “subject” to them leads to directive interpretations only, see Reis 1995, 2003; Truckenbrodt 2006a, 2006b; or Gärtner (2013).

26

daniel gutzmann

namely, that the speaker is smiling or hugging the addressee. Using the terminology introduced so far, the constructions are shunting UCIs like the unexpectedness intonation discussed above, since they do no leave behind any truth-conditional content. The following representation illustrates the dimensions of meaning involved in these non-inflected verbal constructions. (61) dich in den Arm nehm = the speaker is acting as if performing the action of hugging the addressee ∅

Of course, use-conditional morphology encompasses much more than the two phenomena just sketched. For an overview of more and different useconditional elements in morphology, see Dressler & Merlini-Barbaresi 1994; Fortin 2011; Zwicky & Pullum 1987. 4. Types of Use-Conditional Items In the preceding survey of UCIs from different categories, I introduced some distinctions regarding which levels of meaning a UCI contributes to and how it interacts with the truth-conditional content. In this section, I will summarize these findings and establish some terminology for talking about these distinctions. As we saw in the overview, there are at least two binary dimensions with respect to which UCIs may differ and therefore, we can distinguish at least five different types of UCIs. The first distinction concerns the question of whether a UCI has only use-conditional content or whether it carries truthconditional meaning as well. I call this criterion dimensionality. A UCI that only conveys UC-meaning is said to be one-dimensional, whereas an expression that contributes both kinds of meaning is two-dimensional. I will render this as the binary feature [±2-dimensional], or [±2d] for short. As I have done in the previous subsections, I call UCIs that are [−2-dimensional], i.e. UCIs that contribute only use-conditional but no truth-conditional content, expletive UCIs, following Cruse (2004: 57). In order to denote UCIs that contribute content to both dimensions of meaning, I adopt the term mixed UCIs from McCready (2010). These are specified as [+2-dimensional]. Amongst others, examples of expletive UCIs have been expressive adjective attributes, (cf. (2)), modal particles (cf. (21)), or the discourse structuring effect of topicalization (cf. (52)). Removing or adding these to a sentence does not affect its truth-conditional content, since they convey only useconditional meaning. Mixed UCIs, on the other hand, also express content

expressives and beyond

27

that is truth-contentionally relevant. They include ethnic slurs (cf. (10)), which contribute truth-conditional content that equals that of their neutral, non-racist counterparts. The familiar/formal pronouns du/tous vs. Sie/vous (cf. (27)) are a further case of mixed UCIs. They refer to the addressee at the truth-conditional tier, while expressing a familiar or formal relationship between speaker and addressee at the use-conditional one. The feature of dimensionality is, however, not sufficient to account for all differences that can be attested between various UCIs. For instance, we have seen that the use-conditional component of interjections behaves differently than that of modal particles for instance, even if both are expletive UCIs. Likewise, amongst mixed UCIs, the use-conditional content of ethnic slurs behaves differently from the content conveyed by the mixed UCI man when used with a gradable adjective inside the sentence. These differences do not concern what content a UCI delivers but what it needs. Interjections and ethnic slurs have use-conditional content that comes already saturated. By this, I mean that they do not need any further argument to unfold their meaning; that is, they come with complete use conditions and are not any kind of function from (an) argument(s) to use conditions. As illustrated in (8), the interjection ouch directly expresses the emotion of pain without needing an argument. The same holds for the slur Kraut in (17). The negative attitude it expresses towards Germans does not depend on any argument, even if it needs an argument in the truth-conditional dimension. By contrast, other UCIs seek an argument to which their content can apply. For example, modal particles like ja display an attitude towards a propositional argument, and expressive adjectival attributes like damn express an attitude towards a nominal argument. This is informally described in (4) and (24). This is the second aspect with respect to which UCIs can differ, which I call their functionality. I use the binary feature [±functional], or [±f], to distinguish what I label isolated UCIs (after Potts 2005: 65) from functional UCIs. Given the two binary features of dimensionality and functionality, we can distinguish at least four different types of UCIs. However, this is a simplification since I have left out of the picture those expressions that I have called shunting UCIs (McCready 2010). Those are functional UCIs that do not leave their truth-conditional argument unmodified in the truthconditional dimension like functional expletive UCIs do. Instead, they shunt their argument over to the use-conditional dimension. Unexpectedness intonation, as discussed in (33), could be understood as an example of a shunting UCI. McCready (2010) discusses the Japanese adverbial yokumo, which transfers an assertion into a use-conditional speech act of negative

28

daniel gutzmann

attitude and unexpectedness. Finally, as argued by Bücking & Rau (2013), [this volume] the non-inflected constructions discussed above only contribute use-conditional meaning, too. Since both functional expletives and shunting UCIs need an argument and contribute only to the use-conditional dimension, they have the same features. Therefore, we need an additional feature to distinguish between functional UCIs that shunt and those that do not. Because, as McCready (2010) points out, shunting UCIs consume their argument whereas functional expletives do not, which means that it can be reused in a semantic derivation, we can use resource-sensitivity [±r-sensitive], or [±rs] as a distinguishing feature.30 However, this feature only makes sense for one-dimensional functional UCIs. On the one hand, isolated UCIs do not take any argument and therefore, the question whether an application is resource-sensitive or not does not arise. On the other hand, functional mixed UCIs always map their argument to both dimensions of meaning and hence, we cannot meaningfully distinguish between two variants of functional mixed UCIs.31 Taking the distinction between expletive and shunting functional UCIs, we end up with five different types of UCI. Table 1 gives the matrix for the different kinds of UCIs and provides a label for each type. Table 1: Types of use-conditional items f 2d isolated expletive UCIs – – isolated mixed UCIs – + functional UCIs, expletive + – shunting + – functional mixed UCIs + +

rs

– +

Depending on their feature structure, these five types of UCIs can be ordered hierarchically. The simplest type are isolated expletive UCIs as they have a negative specification for dimensionality and functionality. The next level of complexity consists of both functional UCIs and isolated mixed UCIs. These two structures cannot be related to each other, since they are derived from isolated expletives by the addition of a different feature. In the case of

30 Resource-sensitivity figures prominently in linear logic (Girard 1987) and has been adopted in certain varieties of natural language semantics (Asudeh 2005; Barker 2010). 31 In one sense, functional mixed UCIs are [+r-sensitive], because they do not give back their argument unmodified. In another sense, they are [–r-sensitive], since their argument appears in both dimensions.

expressives and beyond

29

Fig. 1. Relations between the five types of use-conditional items

functional UCIs, there is a positive specification for being functional, in contrast to isolated expletives. Here, we can distinguish between shunting and expletive functional UCIs, depending on the specification for resourcesensitivity. When a positive specification for dimensionality is added to the simple case of isolated expletives, we arrive at isolated mixed UCIs. Functional mixed UCIs, the fifth and most feature-rich type, share one feature with each functional and isolated mixed UCIs. We can arrive from the latter two types at functional mixed UCIs if we add the feature that they are missing. Figure 1 illustrates the relations between the five types of UCIs with respect to their feature structure. When comparing the informal fraction representations I have used throughout this chapter to illustrate the truth- and use-conditional content of sentences containing UCIs, we can abstract a schematic representation for each of the five types of UCIs. The simplest schema is that for isolated expletives, since they do not interact with the rest of the sentence in any non-trivial way. When an interjection like ouch is used together with a sentence, the truth-conditional content of the sentence is not affected by the presence of that interjection, whereas its use-conditional content is given by the emotion conveyed by ouch itself. To be a bit more precise, I make use of the following notational conventions. The variable S ranges over sentences, and ε over useconditional expressions. I use brackets to indicate that something is included in a sentence. That is, S[…ε…] denotes a sentence that includes a UCI in an unspecified position. For isolated expletives, we thus arrive at the following schema.

30

daniel gutzmann

(62) Schema for isolated expletives

S[…ε…] =

ε S

From isolated expletives we get to the type of functional expletives by adding the feature [+functional]. In contrast to the former, the meaning of the latter is a use-conditional function which, as such, needs an argument, for which I use the variable 𝛼. A functional expletive UCI, like the adjective damn for instance, needs a nominal argument to which the negative attitude conveyed can apply. Not all functional UCIs modify their argument at the truth-conditional layer. However, depending on whether the application is [−resource-sensitive] or not, we have to distinguish between two schemas for functional UCIs. Functional expletives like damn pass their argument back to the truth-conditional layer. Truth-conditionally, the damn dog makes the same contribution as the dog. Therefore, the argument 𝛼 has to show up at the lower level of the fraction representation in addition to being the argument for the UCI at the use-conditional layer on top of it. (63) Schema for expletive functional UCIs

S[…ε(𝛼)…] =

ε(𝛼) S[…𝛼…]

This contrasts with the schema for shunting functional UCIs, for which the application is [+resource-sensitive], because their argument is not reused at the truth-conditional level. Therefore, the shunting application removes the argument 𝛼 from the truth-conditional tier. (64) Schema for shunting functional UCIs

S[…ε(𝛼)…] =

ε(𝛼) S

The difference between functional UCIs and isolated mixed UCIs is that the former contribute only to the use-conditional meaning component, whereas isolated mixed UCIs carry also truth-conditionally relevant meaning; they are [+2-dimensional]. However, the use-conditional component of isolated mixed UCIs is [−functional] since it does not depend on an argument. An ethnic slur like Kraut, for instance, descriptively predicates German to a nominal argument 𝛼 while expressing a negative attitude towards Germans in the use-conditional dimension. To indicate the 2-dimensionality of a UCI, I make use of subscripted t and u to label the truth- and use-conditional content of a mixed expressive. Since the negative attitude conveyed by Kraut does not need an argument at the truth-conditional tier, the argument 𝛼 does not show up at the top layer of the following representational schema for isolated mixed UCIs:

expressives and beyond (65) Schema for isolated mixed UCIs

S[…ε(𝛼)…] =

31

εu S[…εt (𝛼)…]

The last type, functional mixed UCIs, differs from the last two types insofar as it is both [+functional] and [+2-dimensional]. The use-conditional component of a functional mixed UCI is a use-conditional function and therefore needs an argument 𝛼. And since it has mixed content, such an item contributes something to the truth-conditional dimension as well. An example of a functional mixed UCI is given by the interjectional use of English man when combined with a sentence that contains a gradable adjective. On the truth-conditional layer, man intensifies the property predicated in the propositional content, i.e., Man, it’s hot! means that it is very hot (McCready 2009). On the use-conditional layer, man expresses that the speaker is somehow affected by that high degree of heat. The schema for functional mixed UCIs hence involves an argument 𝛼 that is present in both dimensions of meaning. In the truth-conditional tier, it is modified by εt , the truth-conditional meaning component of the mixed expressive, while on the use-conditional level on top of it, 𝛼 serves as the argument for the useconditional content εu . (66) Schema for functional mixed UCIs

S[…ε(𝛼)…] =

εu (𝛼) S[…εt (𝛼)…]

These five schemas are useful for illustrating the differences and similarities between the four types of UCIs. They could serve as a good starting point for anyone aiming to develop a formal model for dealing with UCIs. 5. Characteristics of Use-Conditional Meaning So far, I have given an overview of the richness of expressions and constructions that convey use-conditional meaning, and distinguished between four different types of UCIs. However, I have not yet given a description of the main characteristics and properties of expressions that convey useconditional meaning. That will be the topic of this section. One feature that ties all UCIs together is trivial insofar as it follows from their definition. UCIs convey meaning that does not contribute to the truth conditions of a sentence, but instead, they affect the conditions in which the sentence can felicitously be uttered. However, even if this is what defines UCIs, one may ask whether there are further properties

32

daniel gutzmann

common to all UCIs. If there are no additional similarities between the different expressions and constructions sketched in this chapter, would not the notion of UCIs just be an artificial cover term to lump together various phenomena that are not really related to each other? This question becomes especially prominent when considering the broad range of phenomena that I have characterized as being use-conditional. Although I will show that there are certain characteristics that are shared by all UCIs, I should briefly address this worry directly. I think that even if there were no properties common to all UCIs (of course, besides being use-conditional), this should not be considered as problematic. First, when speaking of UCIs, I do not want to presuppose that all expressions that contribute use-conditional instead of truth-conditional content form a natural class of expressions. Rather, I would say that useconditional meaning is a class of meaning, just as truth-conditional meaning is another. Secondly, finding common properties in addition to useconditionality should be considered as a bonus and not as a necessary condition, as it should not even be expected. When considering all expressions that contribute truth-conditional content, it is very hard to find properties that are shared by all of them. For instance, what is a property that verbs, quantifiers like many, and conjunctions share? Truth-conditional content is contributed by expressions of a large variety of parts of speech or constructions, yet we speak of truth-conditional expressions. What unites them is that they help to determine the conditions under which a sentence is true. In the same vein, I think that we should not be worried if expressions that convey use-conditional meaning exhibit the same diversity as truth-conditional expressions. This worry therefore may not be as problematic as it seems, at first sight. And it becomes even less worrying, since there actually are some characteristic properties that are common to UCIs. These, however, follow more or less from the fact that UCIs are not part of the truth conditions of the sentence in which they are used. But this does not make these criteria less useful; quite the contrary. They can be used as a tool to check whether an alleged UCI does indeed contribute use-conditional content rather than just rely on our intuition that it does not affect the truth conditions of an utterance. In the following, I present and discuss some properties that have been ascribed to UCIs in the literature. I then check how these can be used to test the non-truth-conditionality of an expression that is supposed to contribute use-conditional meaning. As a starting point, I will take the following list of properties that Potts (2007c) suggests as being essential to expressives (in the narrow sense) in his influential paper.

expressives and beyond (67) Potts’s (2007c) properties of expressives a. b. c. d. e. f.

33 (Potts 2007c: 166 f.)

Independence Nondisplaceability Perspective dependence Descriptive ineffability Immediacy Repeatability

I will now sketch out Potts’s (2007c) aims with these properties, and whether they apply to all of the five types of UCIs. 5.1. Independence The most important property on Potts’s (2007c) list is that use-conditional content is independent of the truth-conditional content. Here is how Potts (2007c: 166) formulates this property: (68) Independence—Expressive content contributes a dimension of meaning that is separate from the regular truth-conditional content.

The independence of use-conditional from truth-conditional content should directly follow from this distinction. Use-conditional content affects the conditions under which a sentence can be uttered felicitously, not the conditions that have to be fulfilled in order to make a sentence true. Hence, using a UCI that is not licensed by the utterance context leads to infelicity, but does not render a sentence false that would be true if the UCI were left out or—in the case of mixed UCIs—were substituted by a non-UCI counterpart (see below). We are dealing with two separate criteria to evaluate a sentence. The independence from the regular truth-conditional content is also what distinguishes use-conditional content from presuppositions, because presupposed content can affect truth-conditional content, at least if it is not satisfied. Since independence follows directly from the fact that use-conditional content is not truth-conditional content, it holds for all kinds of UCIs. The first way to test independence is by giving the truth conditions of a sentence that contains an alleged UCI and comparing it with the same sentence without it, or with a sentence containing a different UCI. If the truth conditions are the same, then the content of the UCI in question does not contribute truth-conditional content. I have already employed such a test in the preceding survey to illustrate that an expression does not contribute to the truth conditions of a sentence. In (22) for instance, I gave the truth conditions for an utterance of the sentence Webster schläft {∅/doch/halt/ja/wohl}, all the variants of which are true, if Webster is sleeping, regardless of whether or

34

daniel gutzmann

which modal particle is used. As we saw in (23b), modal particles can also be used to illustrate that the use-conditional content of an utterance can lead to infelicity even if its truth-conditional content is true. Since the modal particle ja expressively ascribes potential previous knowledge of the truth-conditional content to the addressee, using ja in the context of (23) in an utterance of Es ist ein Mädchen! “It’s a girl!” makes the utterance infelicitous, even if the sentence is still true. Before I provide some more procedures to use the independence property to derive further characteristics of UCIs, let me add a qualification to this property and Potts’s (2007c) formulation of it. Even if Potts (2007c) is right by saying that use-conditional content is independent of the truthconditional content, it is important to keep in mind that it neither generally holds that the content of a UCI is independent of the truth-conditional content nor that the truth-conditional content is independent of the content of an expressive. This is so because, as we have seen, there are what I have called mixed UCIs that contribute both, use- as well as truth-conditional content. This is illustrated schematically in (65) and (66). Mixed UCIs like Kraut or the German familiar pronoun du cannot be omitted from a sentence without also removing their truth-conditional contribution.32 Besides giving the truth conditions of a sentence, there are other tests that can be used to argue for the non-truth-conditional status of the content of an expression. The first one involves negation. If an expression contributes truth-conditional content, it should be possible to target that content by ordinary negation. For instance, an ordinary adverb can be negated directly, as (69) shows. (69) a. Peter is running fast. b. Peter is not running fast.

For the negated sentence to be true, it should not be the case that Peter runs fast. However, that is still compatible with a situation in which Peter runs and therefore, a continuation that states that the rest of the sentence holds, without the negated adverb, is coherent. (70) Peter is not running fast, even if he is running.

This contrasts with use-conditional content. Take modal particles again. Even if negation can be used in a sentence containing ja, it cannot target

32 However, they can be viewed as multidimensional lexical items whose two dimensions of meaning are nonetheless independent.

expressives and beyond

35

the modal particle. Consider the following example, which is a negation of (21b). (71) Webster schläft ja nicht. Webster sleeps mp not ‘(As you may know) Webster is not sleeping.’

In this example, the negation cannot target ja—it has to target the entire proposition. But crucially ja is still not part of the negated proposition but itself scopes over the entire negated proposition. Therefore, a discourse continuation that tries to establish the truth of the corresponding sentence without the negation and the modal particle is not possible. In fact, trying to do this just leads to a plain contradiction. (72) #Webster schläft ja nicht, auch wenn er schläft. Webster sleeps mp not even if he sleeps #‘Webster is not ja sleeping, even if he is sleeping.’ (intended: ‘As you may not have known, Webster is sleeping.’)

A related way to employ negation to test the non-truth-conditionality of an expression is by denial in dialogue (cf. Jayez & Rossari 2004). Instead of trying to negate use-conditional content in the same sentence, one can try to deny it by a reacting utterance in a discourse. (73) A:

Webster schläft ja. ‘Webster is ja sleeping.’ B1: No, he isn’t. B2: #No, I didn’t know that. B3: #No, even if he is sleeping.

B’s answer in (73B1) shows that it is perfectly fine to deny the truth-conditional content that Webster is sleeping. However, (73)-B1 does not say anything about the contribution of ja. The acceptability of (73B1) contrasts with B’s possible reactions in (73B2) and (73B3), which are both infelicitous. In(73B2), the speaker is trying to deny the use-conditional contribution of ja in A’s utterance by negating what A attributed to her by the use of ja. This is infelicitous. The same holds if B tries to negate the use-conditional content while asserting that the truth-conditional content holds, which is possible for truth-conditional expressions as shown in (70). Maybe the infelicity of (73B2) and (73B3) is related to the presence of the answer particle no that seems to inevitably target the truth-conditional content. When no is omitted and B asserts that she did not know what A has asserted, we get a felicitous utterance in which B, at first sight, seems to target the knowledge-ascribing content of ja.

36

daniel gutzmann

(74) A: Webster schläft ja. ‘Webster is ja sleeping.’ B: I didn’t know that.

In this form, B’s denial is felicitous. However, it only seems as if B’s response is denying the content of ja, but in fact, it is not. This is shown by the following dialogue in which B can also assert that she did not know that Webster is sleeping, but A has not even used ja. (75) A: Webster is sleeping. B: I didn’t know that.

That shows that B’s reaction in (74) does not deny the use-conditional content contributed by ja but just makes explicit how A’s utterance fits into her discourse model, even if her utterance contradicts what is displayed by A’s usage of the modal particle. Besides negation and denial, there are other tests to show that the meaning of UCIs does not contribute to the truth-conditional content of an utterance. For instance, use-conditional content does not become part of the content of a question. Consider first an example that shows that a truthconditional expression like fast is indeed part of a question. (76) A: Is Peter running fast? B1: No, but he is running. B2: #Yes. However, he isn’t fast.

As (76)-B1 illustrates, giving a negative answer to A’s question is still compatible with Peter running. Furthermore, one cannot give a positive answer to the question without committing to the contribution of fast, as shown in (76)-B2. For use-conditional content, it is the other way round. In the following example, B cannot give a negative answer to A’s question on the grounds that she rejects only the contribution of the UCI but not the truthconditional content. (77) A: Have you ever met that bastard Kresge? B1: #No. But I have met him. B2: Yes. However, I like him.

The answer in (77)-B2 furthermore shows that committing to the truthconditional content of a sentence does not commit oneself to the useconditional content of the preceding question, at least if one makes oneself sufficiently clear. That use-conditional meaning is not part of what is questioned by an interrogative and that it can neither be the target of negation nor of denial

expressives and beyond

37

in dialogue follows directly from the fact that it contributes a dimension of meaning that is independent from the regular truth-conditional dimension. Hence, I add them as subproperties to the property of (my reformulation of) independence. (78) Independence—Expressive content contributes a dimension of meaning that is separate from the regular truth-conditional content. a. Use-conditional content cannot be negated by ordinary negation. b. Use-conditional content cannot be denied directly in dialogue. c. Use-conditional content is not part of what is questioned by an interrogative. d. Use-conditional content does not affect the truth-conditional content if not fulfilled.

These properties can therefore function as diagnostics to establish the independence of the content of an expression. However, they alone are not sufficient to show that an expression contributes use-conditional content since other kinds of meaning, like presuppositions, show the same behavior with respect to negation and questions. The distinction between presupposed and use-conditional content can be motivated in more detail,33 but as we saw above, the independence holds in both directions for UCIs. While a main characteristic of presuppositions is that the truth-conditional content depends on the satisfaction of its presuppositions, the truth-conditional content is not affected if a UCI is used infelicitously. Therefore, I have added (78d) as a further subproperty of independence. 5.2. Nondisplaceability By saying that use-conditional content is nondisplaceable, Potts (2007c: 169) means that even inside linguistic contexts like speech or attitude reports, modal or conditionalized statements, or reports of past events, the useconditional content cannot be evaluated in that semantic context but always conveys something about the context of the utterance itself (Potts 2007c: 166): (79) Nondisplaceability—Expressives predicate something of the utterance situation.

33 For arguments for this distinction, cf. Potts 2005: 32–36 and Gutzmann 2012: 72–82. For an opposing position, cf. Schlenker 2007.

38

daniel gutzmann

The nondisplaceability of the content of UCIs is a further property that sets them apart from ordinary truth-conditional elements. The truth-conditional meaning of regular non-use-conditional language does not have to be about the utterance situation but may easily be shifted in order to refer to things, events, or attitudes that are not present in the actual utterance situation. This is mostly impossible for UCIs. Displacement, as Hockett (1958: 579) calls it, is often assumed to be one of the “design features” of human languages, setting them apart from other forms of communication (Coleman 2006; Hockett & Altmann 1968). According to Cruse (1986), use-conditional meaning therefore shows similarities to non-verbal communication like gestures or mimic.34 Another characteristic distinguishing expressive meaning from propositional meaning is that it is valid only for the utterer, at the time and place of utterance. This limitation it shares with, for instance, a smile, a frown, a gesture of impatience […]. (Cruse 1986: 272)

Since UCIs lack the possibility of being displaced—at least to some extent, as we will see—they have in this respect more in common with non-verbal communication or the direct signaling nature of non-human communicative systems found in animal communication, which are lacking the property of displacement. Animal communication seems to be designed exclusively for this moment, here and now. It cannot effectively be used to relate events that are far removed in time and place. When your dog says Grrr, it means Grrr, right now, because dogs don’t seem to be capable of communicating Grrr, last night, over in the park. In contrast, human language users are normally capable of producing messages equivalent to Grrr, last night, over in the park, and then going on to say In fact, I’ll be going back tomorrow for some more. (Yule 2006: 9)

Besides the past tense that Yule (2006) uses in his example, common linguistic means to displace the interpretation of an expression include speech or attitude reporting verbs, modal elements, or conditional clauses, amongst many others. In the following examples (80b)–(80f), the propositional content that Peter’s Porsche is pink, expressed by (80a), is displaced to the semantic context created by these constructions. (80) a. Peter’s Porsche is pink. b. Peter’s Porsche was pink. c. Penny says that Peter’s Porsche is pink.

34

Also cited by Potts (2007c: 169).

expressives and beyond

39

d. Penny believes that Peter’s Porsche is pink. e. Peter’s Porsche may be pink. f. If Penny loves Peter, Peter’s Porsche is pink.

In all derived variants of (80a), the proposition that Peter’s Porsche is pink does not have to hold in the actual situation, but in the context to which it is displaced.35 When trying to do the same with use-conditional content, we see that this is indeed hardly possible. Building on Yule’s (2006) examples from the quotation above, example (81) shows that an expressive interjection like ouch behaves more like the Grrr of a dog with respect to displacement than its truth-conditional counterpart in (82). (81) a. Ouch! b. #Ouch, last night, over in the kitchen! (82) a. I feel pain! b. I felt pain, last night, over in the kitchen.

Interjections are an extreme example since they are very isolated from the linguistic context. However, nondisplaceability holds also for use-conditional content that is more integrated into the syntax of a sentence than interjections are. The expressive attitude expressed by bastard, for instance, escapes all attempts to displace it to a non-actual context as shown by the infelicity of continuing the utterance with a sentence that contradicts the negative attitude of bastard (cf. Potts 2007c: 170f.). (83) a. That bastard Kresge is late for work. b. That bastard Kresge was late for work yesterday. #But he’s no bastard today, because today he was on time. c. Sue says that that bastard Kresge should be fired. #I think he’s a good guy. d. Sue believes that that bastard Kresge should be fired. #I think he’s a good guy. e. Maybe that bastard Kresge will be late again. #But if not, he’s a good guy. f. #If that bastard Kresge arrives on time, he should be fired for being so mean.

The negative attitude towards Kresge that is expressed by bastard in these examples is interpreted with respect to the context of the entire utterance. That is, even if bastard is syntactically embedded, semantically it nevertheless conveys that the speaker bears a negative attitude towards Kresge in the actual utterance context. 35

pink.

Semantically, each of (80b)–(80f) is, of course, compatible with Peter’s Porsche being

40

daniel gutzmann

However, nondisplaceability seems to be not as strong a constraint on use-conditional meaning as Potts’s (2007c) formulation in (79) suggests. At least for speech and attitude reports, there are instances in which a UCI is interpreted with respect to the context of the reported speech or attitude— at least, if the right conditions apply. In particular, the use-conditional content can be ascribed to the subject of the reported context and not the speaker of the utterance context, contrary to what is demanded by (79). Potts (2007c: 172) provides an example which he adopts from Kratzer (1999). (84) My father screamed that he would never allow me to marry that bastard Webster.

The negative attitude towards Webster conveyed by bastard is obviously not attributed to the speaker of (84), but to her father. This can be made more obvious by the following two continuations. (85) a. My father screamed that he would never allow me to marry that bastard Webster. But I love him so much that I don’t care about my father’s opinion. b. My father screamed that he would never allow me to marry that bastard Webster. #He thinks that I do not deserve such a decent guy.

For further examples of displaced expressives, see amongst others Amaral et al. (2007) or Anand (2007). Interestingly, not all UCIs seem to behave the same way. Some are relatively easy to shift while others may never be displaced. In her paper on modal particles and contexts shift Döring (2013) [this volume] discusses the displacement potential of different German modal particles and shows that even in this special class of UCIs, variation with respect to shiftability is attested. However, even if there are cases in which a UCI can receive a shifted interpretation, Potts (2007c: 173) argues that they do not challenge the nondisplaceability argument. Instead they point to the perspective dependence of UCIs. 5.3. Perspective Dependence The function of using expressives and use-conditional language in general is not to make objective statements about facts in the world nor to convince the hearer that some state of affairs holds. Instead, they are used to display emotions, feeling or attitudes in a more direct way than any truthconditional statement would do. By default, UCIs display the emotions or attitudes of the speaker. However, this does not need to be the case, as shown by example (84) above, in which the emotional attitude is ascribed to the

expressives and beyond

41

reported subject. Potts (2007c: 173) argues that this shifted interpretation does not show that UCIs can be displaced after all, but that they generally depend on a perspective from which they are evaluated (Potts 2007c: 166). (86) Perspective dependence—Expressive content is evaluated from a particular perspective. In general, the perspective is the speaker’s, but there can be deviations if conditions are right.

The idea is that the interpretation of UCIs is not shifted to the context introduced by an attitude or speech report predicate like scream in (84), but that they are nonetheless interpreted at the utterance level. However, they are evaluated with respect to the perspective of a salient individual. Following Lasersohn’s (2005) influential account of predicates of personal taste, Potts (2007c: 173) calls this individual the contextual judge. Like the speaker, the judge appears as a parameter of the utterance context. In many if not most contexts, the two parameters are set to the same individual, that is, in most contexts, the speaker is also the judge, according to whom perspectival expressions are evaluated. However, there are cases in which an individual other than the speaker is so salient that (s)he counts as the judge of the utterance, even if—for obvious reasons—(s)he cannot count as the speaker.36 In (84), “my father picks out an agent that is so salient and so powerful in the context of the sentence that he becomes not only the attitudinal and deontic judge but also the contextual one” (Potts 2007c: 175). According to this view on the interpretation of use-conditional content, UCIs are not actually shifted, even if they are not interpreted with respect to the speaker. Instead, they are always interpreted at root level, but can be evaluated with respect to an individual other than the speaker if the contextual conditions are such that the judge parameter is set to a different individual. The intuition that, in cases like (84), a UCI is interpreted in a shifted context can however be explained as well by this view. Since the subject of the matrix speech or attitude report can be salient enough to become the contextual judge—like in (84)—one may get the impression that a UCI is interpreted in the shifted context. A prediction that this approach to seemingly-shifted UCIs makes is that a non-speaker evaluation of a UCI should not depend on the presence of any linguistic context-shifter like propositional attitude predicates or reported speech. Instead, they should be freely shiftable, as long as the right

36

Direct quotations being an obvious exception, of course.

42

daniel gutzmann

circumstances apply to establish a non-speaker judge. This prediction is indeed confirmed by examples like the following—borrowed from Potts 2007c—in which there is already enough sarcasm involved to set the judge parameter to the authors of the CPJ report instead of to the blogger. (87) A CPJ report on Venezuela tells us how problems have “escalated” in Venezuela under Chavez, i.e. the physical attacks against journalists under previous presidents have “escalated” to Chavez calling the opposition, which includes the media, names. This is very, very serious, but I don’t think another coup attempt is called for until Chavez resorts to dramatic irony or sarcasm. But if that vicious bastard uses litotes, then there’s no other rational choice than an immediate invasion. (Potts 2007c: 175f., my emphasis, dg)

In two recent papers, Harris & Potts (2009a,b) try to pin-point the circumstances under which UCIs can be interpreted in a shifted context by empirical experiments. If the view on the “shiftablity” of use-conditional content just sketched is sound, then examples like (84) or (87) do not challenge the nondisplaceability of use-conditional content but instead highlight its perspective dependence. This perspectivity is what makes them shiftable in a nontechnical sense. However, they convey something about the utterance situation, namely about the contextual judge, which does not need to equal the speaker under the right conditions. 5.4. Descriptive Ineffability By descriptive ineffability, Potts (2007c: 166) means that it is hard, if not impossible, to completely translate use-conditional content into purely truth-conditional language. (88) Descriptive ineffability—Speakers are never fully satisfied when they paraphrase expressive content using descriptive, i.e., nonexpressive, terms.

The theoretical status of descriptive ineffability as a sound property of use-conditional content is not clear. Geurts (2007: 211) criticizes it as a special property since it does not “draw the line between descriptive and expressive language”. It can be found all over the lexicon, not just with UCIs. Clear examples are provided by grammatical expressions or function words like articles, conjunctions, or even prepositions, for which speakers, if “pressed for definitions, they resort to illustrating where the words would be appropriately used” (Potts 2007c: 176). But also for adjectives like languid, green, or pretty, many speakers would have difficulties giving a satisfying paraphrase without resorting to contexts of application (Geurts 2007: 211). Finally, even for simple nouns like house or dog it is not easy to give a precise

expressives and beyond

43

definition that covers all aspects of their use, not to mention complex expressions like knowledge or truth for which literally thousands of pages have been spent to develop a definition that accounts for all of their uses. In this respect, I agree with Geurts (2007). The impossibility to be paraphrased in a satisfactory way is not something that is unique to use-conditional content alone. However, I think that Potts (2007c) is right about descriptive ineffability in another respect. Employing use-conditional language is really different from using truth-conditional expressions. That is, even if we can find a perfect paraphrase for a UCI in purely truth-conditional terms and even if they mean (in a non-technical sense) the same, they do not express the same. So, let us follow Kaplan (1999: 17) and assume that the use-conditional meaning of the interjection oops corresponds to the truth-conditional meaning of I just observed a minor mishap. And, for the sake of the argument, let us suppose that this is a perfect paraphrase. (89) a. Oops! b. I just observed a minor mishap.

Then, even if they both mean the same, they express this differently. (89a) expresses that meaning in a use-conditional way, i.e., it is felicitously used in a context in which the speaker just observed a minor mishap. It is a direct expression of that observation, rather than a statement about it. This contrasts with an utterance of (89b) which is an actual statement about a fact in the world, which could therefore be challenged.37 (90) A: Oops! B: #That’s not true. That was intentional! (91) A: I just observed a minor mishap. B: That’s not true. That was intentional!

Hence, even if (90A) and (91A) depict the same situation, namely one in which the speaker has just observed a minor mishap, they express this content in a fundamentally different way. In the words of Kaplan (1999: 8), use-conditional content expresses or displays, while truth-conditional content decribes. These differences in the modus of expressing are of course the reason why use-conditional and truth-conditional content are independent from each other and cannot be merged into a single dimension of meaning,

37 That does not mean however, that UCIs are error-proof and cannot characterize a situation incorrectly. However, infelicitously used UCIs are not apt to be judged as false.

44

daniel gutzmann

not even a conjoined one.38 And it is also the rationale behind Potts’s (2007c) descriptive ineffability. Hence, I suggest to modify Potts’s (2007c) original statement of descriptive ineffability to focus more on this aspect than on the fact that it is hard to come up with a satisfying paraphrase. (92) Descriptive ineffability (modified)—It is impossible to paraphrase use-conditional content using only truth-conditional expressions without changing the modus of expressing.

Thinking about descriptive ineffability in this way captures the intuition that use-conditional content functions differently from regular truth-conditional content, and that the two kinds of meaning cannot be exchanged freely for each other. 5.5. Immediacy With the property of immediacy, Potts (2007c: 167) draws a parallel between UCIs and performative speech acts. (93) Immediacy—Like performatives, expressives achieve their intended act simply by being uttered; they do not offer content so much as inflict it.

Potts (2007c: 180) provides promises as an example. By just uttering a performative construction like (94), one performs the named speech act and expresses that one is doing so. (94) I promise that I will be back tonight.

Hence, it is not intelligible to say that one is not going to fulfill what one has just promised, as shown in (95a). Furthermore, (95b) shows that one can also not take back that one has just made a promise. (95) I promise that I will be back for dinner. a. #But I plan to be back tomorrow. b. #But I made no promises that I will.

UCIs behave in the same way, as we have already seen from some examples. For instance, (83b) shows that it is infelicitous to take back the negative attitude conveyed by bastard by continuing with But he’s no bastard today (Potts 2007c: 180). Moreover, as for the performative promise, it is also not possible to deny that one has just expressed a negative attitude.

38 In some sense, then, the distinction between displayed and described content mirrors Strawson’s (1950) disentanglement of asserted and presupposed content, which were merged into a single (conjoined) dimension by Russell (1905).

expressives and beyond

45

(96) That bastard Kresge was late for work yesterday. a. #But he’s no bastard today, because today he was on time. b. #But I expressed no negative feelings towards Kaplan by saying this.

Interestingly, the restriction that one cannot deny one’s own speech act holds also for speech acts that are not explicitly performative, like the promise in (94), but also for more structurally determined speech acts like assertions or questions.39 (97) a. David is a zombie. #But I did not assert that he is a zombie. b. Is David a zombie? #But I did not ask you whether he is a zombie.

Potts (2007c: 180) argues that performatives and other direct speech acts also mirror the behavior of UCIs with respect to their inability to be denied in discourse. Just as one cannot deny the negative attitude of an expressive UCI, one cannot deny that your interlocutor has just performed an assertion, for instance. (98) A: That bastard Kresge was late for work yesterday. B: #No, you like him. (99) A: David is a zombie. B: #No, you didn’t just assert that.

The parallelism between use-conditional content and speech acts can be pushed even further. In fact, it could be argued that sentence mood (that is, the structurally encoded speech act potential in, e.g., a declarative or interrogative) is a UCI—a view that is suggested by Portner (2007), and that I have developed in Gutzmann 2008, 2012. Instead of conveying something truth-conditionally, sentence mood imposes use conditions on an utterance. Therefore, an analysis that accounts for UCIs and denial may also work for assertions or questions. The property of immediacy is also a consequence of the fact that useconditional content behaves differently from truth-conditional content. That UCIs achieve their intended effect just by being uttered suggests that they are closely related to the utterance context. This also fits their

39 This is in line with what Austin (1962) already had in mind. Once we think of utterances as speech acts, it seems to be a matter of (virtual) conceptual necessity that one cannot “take back” those actions. This not only holds for the illocutionary act but obviously also for the phatic and rhetic act.

(i) (ii)

David is a zombie. #But I did not say “David is a zombie”. David is a zombie. #But I did not say that he is a zombie.

46

daniel gutzmann

nondisplaceability. To account for this, the context-relatedness must somehow be built directly into the semantics of UCIs. 5.6. Repeatability The last property Potts (2007c: 167) lists is that UCIs can be repeated without redundancy. Instead, the repetition of an expressive intensifies the expressed emotion. (100) Repeatability—If a speaker repeatedly uses an expressive item, the effect is generally one of strengthening the emotive content, rather than one of redundancy.

To illustrate this, Potts (2007c: 182) uses the following list of examples, which contain more and more repetitions of the same expressive. (101) a. Damn, I left my keys in the car. b. Damn, I left my damn keys in the car. c. Damn, I left my damn keys in the damn car.

The more instances of damn are used, the stronger (the expression of the) emotional attitude of the speaker or judge gets, with no redundancy at all. According to Potts (2007c), this contrasts with truth-conditional content that cannot be repeated in the same way without, becoming redundant. Even if Potts (2007c) notes that, due to the descriptive ineffability of useconditional content, it is hard to find a minimal contrasting example that only involves truth-conditional language, he illustrates this with the following example (Potts 2007c: 182): (102) #I’m angry! I forgot my keys. I’m angry! They are in the car. I’m angry!

However, as already observed by Geurts (2007), repeatability is not a good candidate for a characteristic property of use-conditional content. First, there are also truth-conditional expressions that do not lead to redundancy but whose regular semantic effect is strengthened if used repetitively. Potts (2007c) himself mentioned the case of (103a), but dismisses it as a counterexample since such examples can be analyzed compositionally. He argues that the first big can be analyzed as modifying big big apple, while the second modifies big apple and the last one just apple. Therefore, we are not repetitively applying the adjective to the same argument but to an already modified argument. This reasoning should also apply to truth-conditional intensifiers like very if used more than once. (103) a. Peter ate a big big big apple. b. David is very very very large.

expressives and beyond

47

However, Potts’s (2007c) aim is to finally give a compositional analysis for the intensifying effect of the repetition of expressive UCIs like damn in (101b) and (101c) and, therefore, I do not see any reason for not counting examples like the ones in (103a) as instances of truth-conditional repetition without redundancy.40 Given this data, it can already be concluded that repeatability is not a distinctive property of use-conditional content. Moreover, this generalization does not even carry over to other kinds of UCI besides expressives in the narrow sense. For instance, interjections and other isolated UCIs are not repeatable without sounding odd. (104) a. #Oops! I forgot my keys. Oops! They are in the car. Oops! b. #Ouch! I’ve hit my thumb. Ouch! It was the hammer. Ouch!

Contra Potts (2007c), I think that repeatability is neither a necessary nor sufficient condition for use-conditional content and therefore, it cannot be used to differentiate between truth-conditional and use-conditional content. 6. Summary Use-conditional content comes in many different varieties and from many empirical domains of language. This is what the survey of this chapter has shown, even if it could only touch upon a small selection of all kinds of use-conditional meaning that can be found across natural languages. While in a majority of cases, use-conditional content is bound to words, there are also constructions below and beyond the word level that are conventionally associated with non-truth-conditional meaning. The description of the different items by which use-conditional content can be conveyed has shown that not every UCI behaves in the same way with respect to how it interacts with the truth-conditional dimension of meaning. I distinguish these different types of UCIs by means of some binary features, which allows us to differentiate between five types of UCIs, as summarized in Table 1. The purest form of use-conditional content is

40 Geurts (2007: 214) mentions a further problem for Potts’s (2007c) line of reasoning. According to Potts’s approach, the instances of damn in (101c) apply to different arguments and hence express emotional attitudes towards different objects, which means there should be no intensification effect at all. Another way of analyzing (101c) would be to say that all instances of damn apply to the entire event. Then, however, (101c) is not very different from (103b) or (103a).

48

daniel gutzmann

provided by isolated expletives. These are interjections like Kaplan’s (1999) ouch and oops that contribute use conditions by themselves, without needing any argument (–functional), and which do not contribute anything to the truth-conditional dimension of meaning (–2-dimensional). They contrast with functional expletives like damn in that damn dog, which also convey no truth-conditional content (–2-dimensional) but which need an argument in order to express a proper use condition (+functional). Functional expletives also pass back their argument unmodified, what makes (–resource-sensitive), in contrast to functional shunting UCIs, which do not. What I have called isolated mixed UCIs differ from isolated expletives in the other feature. Examples are ethnic slurs like Kraut or other coloured expressions like cur. They are also isolated, because their use-conditional content needs no argument (–functional), but in contrast to isolated expletives, they have an additional truth-conditional meaning component (+2-dimensional). The fourth type—functional mixed UCIs—is the most complex one from an interactional point of view. These UCIs need an argument for their use-conditional component (+functional) and contribute to the truthconditional tier (+2-dimensional) at the same time. In (62)–(65), I have given informal composition schemas for the five types of UCI that illustrate how they interact with the truth-conditional content. Despite the different types of use-conditional content and the variety of linguistic means to express it, there are some characteristic properties that can be attributed to UCIs, which can function as a heuristic to detect them. I have reviewed the list of features developed by Potts (2007c), in order to check whether they are characteristic of use-conditional content. The most important and identifying feature of UCIs is, of course, that they do not affect the truth conditions of a sentence, but instead, impose use conditions on the felicity of the utterance of that sentence. This different modus of expressing content is responsible for the majority of the distinctive properties of UCIs. It is what Potts (2007c) calls independence, by which he means that the two modi of expressing contribute to two different dimensions of meaning. The truth conditions of a sentence do not depend on its use-conditional content, and vice versa. The nondisplaceability of UCIs is also closely connected to their useconditional nature. Use conditions are mostly imposed on the actual utterance and evaluated in the current context, and not on a semantic context created by the linguistic environment. The nondisplaceability of UCIs renders them similar to non-verbal forms of communication like gestures or to animal communication, both of which lack the ability to be interpreted in a non-actual context. And even if UCIs are (of course) arbitrarily and by

expressives and beyond

49

convention connected with the content they express, their nondisplaceability makes it reasonable to think of them as being more like indices/signals instead of symbols in semiotic terms. There are obvious counterexamples to nondisplaceability, but according to Potts (2007c) they stem from the perspective dependence of UCIs. That is, the evaluation of use-conditional content always depends on a contextual judge which, by default, is the speaker but who could be a different individual if it is salient enough in the actual utterance context. Another aspect of the observation that use-conditional content is closely tied to the utterance context, is what Potts (2007c) calls immediacy. Like performatives, UCIs alter the context directly when uttered. One of the properties from Potts’s (2007c) list, that does not prove to be very helpful in distinguishing use-conditional from truth-conditional content, is descriptive ineffability. Finding satisfying and correct paraphrases for use-conditional content is very difficult, but so it is for many truthconditional terms. However, if understood more broadly, descriptive ineffability means that one cannot substitute use-conditional content with truthconditional content—even if has the same content—without changing the modus of expressing that content. Repeatability is also not a good candidate for a characteristic of use-conditional content as there are both UCIs that cannot be repeated without redundancy and truth-conditional expressions that can. 7. The Individual Papers The UCIs that have received, by far, the most attention, are what I have called expressives in the narrow sense. However, as this survey has shown, use-conditional content comes in a great variety of forms. The papers in this collection address some of these. In their contribution German non-inflectional constructions as separate performatives, Sebastian Bücking and Jennifer Rau study the form and interpretation of German non-inflectional constructions like those presented in (60). Making use of Portner’s (2007) modification of Potts’s (2005) logic, they analyze the special syntax and semantics of these morphological UCIs and argue that they can be viewed as separate performatives. As discussed in §5.2, nondisplaceability has been questioned as a defining feature of use-conditional content. However, as Sophia Döring shows in her empirical investigation of the relation between Modal particles and context shift, not all UCIs can be context-shifted in the same way. Presenting new data on modal particles drawn from German corpora, she shows

50

daniel gutzmann

that modal particles show variation regarding the linguistic contexts that can induce a shifted interpretation. Modal particles are also dealt with in Markus Egg’s paper Discourse particles, common ground, and felicity conditions. Instead of discussing their shiftability, Egg builds on the vast descriptive literature on modal particles in German to delevop a formal approach to their meaning. In order to capture their use-conditional nature, he relates them to the common ground and analyzes them as expressions imposing felicity conditions on the relation an utterance bears to the previous one. More radically, he argues that even the felicity conditions of the two utterances can serve as the semantic argument of a particle. Besides particles, personal pronouns seem to be well suited to express use-conditional content on the word level, as we have seen in § 2.3. In his contribution I love me some datives: Expressive meaning, free datives, and Fimplicature, Laurence R. Horn presents a detailed analysis of the personal dative in English, relating it to other variants of free datives in different languages and to other UCIs like modal particles. Moreover, going back to the writings of Frege, he embeds his investigation into a historical and theoretical discussion of how use-conditional content fits into the division of labour between semantics and pragmatics, reaching back to the writings of Frege. In some respects, Eric McCready and Yohei Takahashi step into Grice’s shoes and take up his original example of a conventional implicature, when they divide the meaning of the Japanese connective mono into a truthconditional, asserted component on the one hand, and a non-truth-conditional but nevertheless conventional dimension of meaning on the other hand. Besides discussing the notion of Good reasons, they show how the formal logic developed in McCready 2010 can be used to deal with mixed UCIs. In her detailed investigation, Sophie Repp examines various means of Common ground management: Modal particles, illocutionary negation and verum. She argues that modal particles, illocutionary negation as expressed by the operator falsum, and the operator verum are common-ground managing operators, which indicate the status of a proposition relative to the common ground (newness, expectedness, speaker commitment etc.). Although they are non-truth-conditional, common-ground managing operators can nevertheless influence the truth-conditional meaning of a sentence, as Repp argues. She illustrates this in detail for the scopal interaction of negation and epistemic modal verbs in German. The observed effects are argued to be due to the negative marker denoting either propositional nega-

expressives and beyond

51

tion or the operator falsum, and to common-ground managing operators determining to a large degree the discourse appropriateness of the utterance they occur in. Yasutada Sudo’s contribution deals with the use conditions of Biased polar questions in English and Japanese. Starting with English and the classical distinction between negative and positive bias, Sudo uses data from Japanese to show that this distinction is too coarse-grained. Accordingly, he extends the distinction by adding the notion of evidential and epistemic bias. Sudo examines how these attitudes—which are not part of the truthconditionally relevant question meaning but impose use conditions on the proper use of the question—can be expressed in Japanese by the conventional means of different intonation patterns and certain question particles. Concluding this collection, Henk Zeevat examines the borderline between expressive meaning and presuppositions in his paper Expressing surprise by particles, in which he develops a new approach to the meaning of focus particles. In contrast to the vast majority of accounts of such expressions, his approach is non-truth-conditional in nature and is based on the idea that emotions like surprise should play a major role in linguistic semantics. He furthermore argues that focus particles are a case in which the useconditional and the truth-conditional dimension of meaning cannot be fully separated, shedding doubt on the property of independence discussed in §5.1 above. Acknowledgements This work is for the most part based on the second chapter of my doctoral dissertation (Gutzmann 2012). Besides Ede Zimmermann, Eric McCready, and Hans-Martin Gärtner, who deserve many thanks for being very helpful and supportive in that context, I would like to thank all the contributors to this volume and the speakers of the Osnabrück workshop for inspiring ideas. My co-editor Hans-Martin Gärtner provided plenty of helpful comments and references during various stages of this paper and contributed the brief historical introduction. Furthermore, Antonio Fortin also gave insightful remarks as well as helping with correcting the final manuscript. All remaining infelicities are my own.

52

daniel gutzmann References

Altmann, Hans (1976): Die Gradpartikeln im Deutschen. Untersuchungen zu ihrer Syntax, Semantik und Pragmatik. Tübingen: Niemeyer. ——— (1981): Formen der “Herausstellung” im Deutschen. Rechtsversetzung, Linksversetzung, Freies Thema und verwandte Konstruktionen. Tübingen: Niemeyer. Amaral, Patricia & Craige Roberts & E. Allyn Smith (2007): “Review of The logic of conventional implicatures by Chris Potts”. Linguistics and Philosophy 30, 707–749. DOI: dx.doi.org/10.1007/s10988-008-9025-2. Ameka, Felix (1992): “Interjections: The universal yet neglected part of speech”. Journal of Pragmatics 18, 101–118. DOI: dx.doi.org/10.1016/03782166(92)90048-G. Anand, Pranav (2007): “Re-expressing judgment”. Theoretical Linguistics 33, 199– 208. DOI: dx.doi.org/10.1515/TL.2007.012. Asudeh, Ash (2005): “Control and semantic resource sensitivity”. Journal of Linguistics 41, 465–511. DOI: dx.doi.org/10.1017/S0022226705003427. Austin, John L. (1962): How to do things with words. Oxford: Oxford University Press. Bar-Hillel, Yehoshua (1971): “Out of the pragmatic wastebasket”. Linguistic Inquiry 2, 401–407. URL: http://www.jstor.org/stable/4177643. Barker, Chris (2010): “Free choice permission as resource-sensitive reasoning”. Semantics and pragmatics 3, 1–38. DOI: dx.doi.org/10.3765/sp.3.10. Bauer, Laurie (1997): “Evaluative morphology. In search of universals”. Studies in Language 21, 533–575. DOI: dx.doi.org/10.1075/sl.21.3.04bau. Bayer, Josef & Hans-Georg Obenauer (2011): “Discourse particles, clause structure, and question types”. The Linguistic Review 28, 449–491. DOI: dx.doi.org/10.1515/ tlir.2011.013. Beaver, David I. & Brady Clark (2008): Sense and Sensitivity. How Focus Determines Meaning. Oxford: Blackwell. Beaver, David I. & Bart Geurts (2007): “Discourse representation theory”. Stanford Encyclopedia of Philosophy (Fall 2011 edition). URL: http://plato.stanford.edu/ archives/fall2011/entries/discourse-representation-theory. ——— (2011): “Presupposition”. Stanford Encyclopedia of Philosophy (Summer 2011 edition). URL: http://plato.stanford.edu/archives/sum2011/entries/ presupposition/. Bezuidenhout, Anne (2004): “Procedural meaning and the semantics/pragmatics interface”. In: Bianchi, Claudia, ed.: The semantics/pragmatics distinction. Stanford, CA: CSLI, 101–131. Birner, Betty J. & Gregory Ward (1998): Information status and noncanonical word order in English. Amsterdam: John Benjamins. Borer, Hagit & Yosef Grodzinsky (1986): “Syntactic cliticization and lexical cliticization. The case of Hebrew dative clitics”. In: Borer, Hagit, ed.: The syntax of pronominal clitics. New York: Academic Press, 175–217. Bücking, Sebastian & Jennifer Rau (2013): “German non-inflectional constructions as separate performatives”. In: Gutzmann, Daniel & Hans-Martin Gärtner, eds.: Beyond expressives. Explorations in use-conditional meaning. Leiden: Brill, 59–94. [This volume].

expressives and beyond

53

Bühler, Karl (1934): Sprachtheorie. Jena: Gustav Fischer Verlag. Büring, Daniel (1997): The meaning of topic and focus. The 59th street bridge accent. London: Routledge. ——— (2003): “On D-trees, beans, and B-accents”. Linguistics and Philosophy 26, 511–545. DOI: dx.doi.org/10.1023/A:1025887707652. Büring, Daniel & Katharina Hartmann (2001): “The syntax and semantics of focussensitive particles in German”. Natural Language & Linguistic Theory 19, 229– 281. DOI: dx.doi.org/10.1023/A:1010653115493. Castroviejo Miró, Elena (2008): “An expressive answer. Some considerations on the semantics and pragmatics of wh-exclamatives”. Proceedings of CLS 44 (Vol. 2), 3–17. Claudi, Ulrike & Bernd Heine (1986): “On the metaphorical base of grammar”. Studies in Language 10, 297–335. DOI: dx.doi.org/10.1075/sl.10.2.03cla. Coleman, John (2006): “Design features of language”. In: Brown, Keith, ed.: Encyclopedia of language & linguistics. 2nd ed. Oxford: Elsevier, 471–475. DOI: dx.doi .org/10.1016/B0-08-044854-2/04743-X. Cruse, David Alan (1986): Lexical Semantics. Cambridge: Cambridge University Press. ——— (2004): Meaning in language. An introduction to semantics and pragmatics. 2nd ed. Oxford: Oxford University Press. d’Avis, Franz Josef (2002): “On the interpretation of wh-clauses in exclamative environments”. Theoretical Linguistics 28.1, 5–31. DOI: dx.doi.org/10.1515/thli.2002.28 .1.5. Döring, Sophia (2013): “Modal particles and context shift”. In: Gutzmann, Daniel & Hans-Martin Gärtner, eds.: Beyond expressives. Explorations in use-conditional meaning. Leiden: Brill, 95–123. [This volume]. Dowty, David R. & Robert E. Wall & Stanley Peters (1981): Introduction to Montague semantics. Dordrecht: Reidel. Dressler, Wolfgang Ulrich & Lavinia Merlini-Barbaresi (1994): Morphopragmatics. Berlin: de Gruyter. Dummett, Michael (1978): “Frege’s philosophy”. In: Truth and other enigmas. Cambridge, MA: Harvard University Press, 87–115. Ebert, Cornelia & Stefan Evert & Katharina Wilmes (2011): “Focus marking via gestures”. Proceedings of Sinn und Bedeutung 15, 193–208. URL: http://universaar.uni -saarland.de/monographien/volltexte/2011/30/artikel/Ebert_et_al_sub15.pdf. Egg, Markus (2013): “Discourse particles, common ground, and felicity conditions”. In: Gutzmann, Daniel & Hans-Martin Gärtner, eds.: Beyond expressives. Explorations in use-conditional meaning. Leiden: Brill, 125–149. [This volume]. Fortin, Antonio (2011): The morphology and semantics of expressive affixes. PhD thesis. Oxford: University of Oxford. Frege, Gottlob (1897/1979): “Logic”. In: Hermes, Hans & Friedrich Kambartel & Friedrich Kaulbach, eds.: Posthumous writings: Gottlob Frege. Translated by Peter Long and Roger White. Oxford: Blackwell, 126–151. Frey, Werner (2010): “A’-movement and conventional implicatures: About the grammatical encoding of emphasis in German”. Lingua 120, 1416–1435. DOI: dx.doi .org/10.1016/j.lingua.2008.09.016. Fries, Norbert (2007): “Die Kodierung von Emotionen in Texten. Teil 1: Grundlagen”. Journal of Literary Theory 1, 293–337. DOI: dx.doi.org/10.1515/JLT.2007.020.

54

daniel gutzmann

——— (2009): “Die Kodierung von Emotionen in Texten. Teil 2: Die Spezifizierung emotionaler Bedeutung in Texten”. Journal of Literary Theory 3, 19–71. DOI: dx.doi.org/10.1515/JLT.2009.003. Gamut, L.T.F. (1991): Logic, language, and meaning. Chicago: University of Chicago Press. Gärtner, Hans-Martin (2013): “Infinite Hauptsatzstrukturen”. In: Meibauer, Jörg & Markus Steinbach & Hans Altmann, eds.: Satztypen des Deutschen. Berlin: de Gruyter, 202–231. Gazdar, Gerald (1979): Pragmatics. Implicature, presupposition and logical form. New York: Academic Press. Geach, Peter (1965): “Assertion”. The Philosophical Review 74, 449–465. DOI: dx.doi .org/10.2307/2183123. Geurts, Bart (2007): “Really fucking brilliant”. Theoretical Linguistics 33, 209–214. DOI: dx.doi.org/10.1515/TL.2007.013. Girard, Jean-Yves (1987): “Linear logic”. Theoretical computer science 50, 1–102. DOI: dx.doi.org/10.1016/0304-3975(87)90045-4. Green, Keith & Richard Kortum (2007): “Can Frege’s färbung help explain the meaning of ethical terms?” Essays in Philosophy 8. URL: http://commons.pacificu. edu/eip/vol8/iss1/10. Grice, H. Paul (1975): “Logic and conversation”. In: Cole, Peter & Jerry L. Morgan, eds.: Speech acts. New York: Academic Press, 41–58. Groenendijk, Jeroen & Martin Stokhof (1991): “Dynamic predicate logic”. Linguistics and Philosophy 14, 39–100. DOI: dx.doi.org/10.1007/BF00628304. Gunlogson, Christine (2003): True to form: Rising and falling declaratives as questions in English. New York: Routledge. Gutzmann, Daniel (2007): “Eine Implikatur konventioneller Art: der Dativus Ethicus”. Linguistische Berichte 211, 277–308. URL: http://www.ingentaconnect.com/ content/buske/lingber/2007/00002007/00000211/art00002. ——— (2008): On the interaction between modal particles and sentence mood in German. MA thesis. Mainz: Johannes Gutenberg-Universität. ——— (2009): “Hybrid semantics for modal particles”. Sprache und Datenverarbeitung 33, 45–59. ——— (2012): Use-conditional meaning. Studies in multidimensional semantics. PhD thesis. University of Frankfurt. Gutzmann, Daniel & Elena Castroviejo Miró (2011): “The dimensions of verum”. In: Bonami, Olivier & Patricia Cabredo-Hofherr, eds.: Empirical Issues in Syntax and Semantics 8, 143–165. URL: http://www.cssp.cnrs.fr/eiss8/gutzmann-castroviejo -miro-eiss8.pdf. Hare, Richard M. (1952): The language of morals. Oxford: Oxford University Press. Harris, Jesse A. & Christopher Potts (2009a): “Perspective-shifting with appositives and expressives”. Linguistics and Philosophy 32.6, 523–552. DOI: dx.doi.org/ 10.1007/s10988-010-9070-5. ——— (2009b): “Predicting perspectival orientation for appositives”. Proceedings of CLS 45 (Vol. 1), 207–221. Heim, Irene (1982): On the semantics of definite and indefinite noun phrases. PhD thesis. Amherst, MA: University of Massachusetts. URL: http://semanticsarchive .net/Archive/Tk0ZmYyY/.

expressives and beyond

55

Helbig, Gerhard (1977): “Partikeln als illokutive Indikatoren im Dialog”. Deutsch als Fremdsprache 14, 30–44. Hockett, Charles (1958): A course in modern linguistics. New York: Macmillan. Hockett, Charles & Stuart A. Altmann (1968): “A note on design features”. In: Sebeok, Thomas A., ed.: Animal communication. Techniques of study and results of research. Bloomington: Indiana University Press, 61–72. Höhle, Tilman N. (1992): “Über Verum-Fokus im Deutschen”. In: Jacobs, Joachim, ed.: Informationsstruktur und Grammatik. Opladen: Westdeutscher Verlag, 112–141. Hom, Christoper (2008): “The semantics of racial epithets”. Journal of Philosophy 105, 416–440. ——— (2010): “Pejoratives”. Philosophy Compass 5, 164–185. DOI: dx.doi.org/10.1111/ j.1747-9991.2009.00274.x. Horn, Laurence R. (1969): “A presuppositional analysis of only and even”. Proceedings of CLS 5, 98–107. ——— (1996): “Exclusive Company: Only and the dynamics of vertical inference”. Journal of Semantics 13.1, 1–40. DOI: dx.doi.org/10.1093/jos/13.1.1. ——— (2007): “Toward a Fregean pragmatics. Voraussetzung, Nebengedanke, Andeutung”. In: Horn, Laurence R. & Istvan Kecskes, eds.: Explorations in pragmatics. Linguistic, cognitive, and intercultural aspects. Berlin: de Gruyter, 39–69. ——— (2008): “I love me some him. The landscape of non-argument datives”. In: Bonami, Olivier & Patricia Cabredo Hofherr, eds.: Empirical Issues in Syntax and Semantics 7, 169–192. ——— (2013): “I love me some datives: Expressive meaning, free datives, and Fimplicature”. In: Gutzmann, Daniel & Hans-Martin Gärtner, eds.: Beyond expressives. Explorations in use-conditional meaning. Leiden: Brill, 151–199. [This volume]. Horn, Laurence R. & Gregory Ward, eds. (2004): The handbook of pragmatics. Oxford: Oxford University Press. Hornsby, Jennifer (2001): “Meaning and uselessness: How to think about derogatory words”. Midwest Studies in Philosophy, 25, 128–141. DOI: dx.doi.org/10.1111/ 1475-4975.00042. Jacobs, Joachim (1983): Fokus und Skalen. Zur Syntax und Semantik der Gradpartikeln im Deutschen. Tübingen: Niemeyer. Jakobson, Roman (1960): “Linguistik und Poetik. Ausgewählte Aufsätze 1921–1971”. In: Poetik. Frankfurt: Suhrkamp, 83–121. Jay, Tomothy & Kristin Janschewitz (2007): “Filling the emotion gap in linguistic theory. Commentary on Potts’ expressive dimension”. Theoretical Linguistics 33, 215–221. DOI: dx.doi.org/10.1515/TL.2007.014. Jayez, Jacques & Corinne Rossari (2004): “Parentheticals as conventional implicatures”. In: Corblin, Francis & Henriëtte de Swart, eds.: Handbook of French semantics. Stanford, CA: CSLI, 211–229. Kamp, Hans & Uwe Reyle (1993): From discourse to logic. Dordrecht: Kluwer. Kaplan, David (1999): “The meaning of ouch and oops. Explorations in the theory of meaning as use”. Ms. University of California at Los Angeles. König, Ekkehard (1991): The meaning of focus particles. A comparative perspective. London: Routledge. Kratzer, Angelika (1999): “Beyond ouch and oops. How descriptive and expressive

56

daniel gutzmann

meaning interact”. Cornell Conference on Theories of Context Dependency. URL: http://semanticsarchive.net/Archive/WEwNGUyO/. ——— (2004): “Interpreting focus: Presupposed or expressive meanings? A comment on Geurts and van der Sandt”. Theoretical Linguistics 30, 123–136. DOI: dx .doi.org/10.1515/thli.2004.002. Ladd, D. Robert (1990): “Intonation: Emotion vs. grammar”. Language 66, 806–816. DOI: dx.doi.org/10.2307/414730. Lakoff, Robin (1974): “Remarks on ‘this’ and ‘that’”. Proceedings of CLS 10, 345–356. Lambert, Silke (2007): “Pragmatic and stylistic relevance of free datives in German”. Ms. Buffalo, NY. Lappin, Shalom, ed. (1996): The handbook of contemporary semantic theory. Oxford: Blackwell. Lasersohn, Peter (2005): “Context dependence, disagreement, and predicates of personal taste”. Linguistics and Philosophy 28.6, 643–686. DOI: dx.doi.org/10.1007/ s10988-005-0596-x. Levinson, Stephen C. (1983): Pragmatics. Cambridge: Cambridge University Press. Lyons, John (1977): Semantics. Oxford: Oxford University Press. Maienborn, Claudia & Paul Portner & Klaus von Heusinger, eds. (2011): Handbook of semantics. 2nd ed. Berlin: de Gruyter. Marty, Anton (1908): Untersuchungen zur Grundlegung der allgemeinen Grammatik und Sprachphilosophie. Halle: Niemeyer. McCready, Eric (2009b): “What man does”. Linguistics and Philosophy 31.6, 671–724. DOI: dx.doi.org/10.1007/s10988-009-9052-7. ——— (2010): “Varieties of conventional implicature”. Semantics & Pragmatics 3.8, 1–57. DOI: dx.doi.org/10.3765/sp.3.8. McCready, Eric & Yohei Takahashi (2013): “Good Reasons”. In: Gutzmann, Daniel & Hans-Martin Gärtner, eds.: Beyond expressives. Explorations in use-conditional meaning. Leiden: Brill, 201–229. [This volume]. Morzycki, Marcin (2010): “Expressive Modification and the Licensing of Measure Phrases”. Journal of Semantics 28, 401–411. DOI: dx.doi.org/10.1093/jos/ffq021. Nouwen, Rick (2007): “On appositives and dynamic binding”. Research on Language and Computation 5, 87–102. DOI: dx.doi.org/10.1007/s11168-006-9019-6. URL: http://www.let.uu.nl/~Rick.Nouwen/personal/papers/appdyn.pdf. Ogden, Charles K. & Ivor A. Richards (1923): The meaning of meaning. London: Routledge & Kegan Paul. Oppenrieder, Wilhelm (1989): “Selbständige Verb-letzt-Sätze. Ihr Platz im Satzmodussystem und ihre intonatorische Kennzeichnung”. In: Altmann, Hans & Anton Batliner & Willhelm Oppenrieder, eds.: Zur Intonation von Modus und Fokus im Deutschen. Tübingen: Niemeyer, 163–244. Portner, Paul (2007): “Instructions for interpretation as separate performatives”. In: Schwabe, Kerstin & Susanne Winkler, eds.: On information structure, meaning and form. Amsterdam: John Benjamins, 407–426. Potts, Christopher (2002): “The syntax and semantics of as-parentheticals”. Natural Language and Linguistic Theory 20, 623–689. DOI: dx.doi.org/10.1023/A:1015892 718818. ——— (2005): The logic of conventional implicatures. Oxford: Oxford University Press.

expressives and beyond

57

——— (2007b): “The centrality of expressive indices”. Theoretical Linguistics 33, 255–268. DOI: dx.doi.org/10.1515/TL.2007.019. ——— (2007c): “The expressive dimension”. Theoretical Linguistics 33, 165–197. DOI: dx.doi.org/10.1515/TL.2007.011. ——— (2012): “Conventional implicature and expressive content”. In: Maienborn, Claudia & Klaus von Heusinger & Paul Portner, eds.: Semantics. An international handbook of natural language meaning. Berlin: de Gruyter, 2516–2535. Potts, Christopher & Florian Schwarz (2010): “Affective ‘this’”. Linguistic Issues in Language Technology 3, 1–30. URL: http://www.elanguage.net/journals/index. php/lilt/article/viewFile/664/525. Recanati, François (2004): “Pragmatics and semantics”. In: Horn, Laurence R. & Gregory Ward, eds.: The handbook of pragmatics. Oxford: Blackwell, 442–462. Reis, Marga (1995): “Über infinite Nominativkonstruktionen im Deutschen”. Sprache und Pragmatik Sonderheft. Festvorträge anläßlich des 60. Geburtstags von Inger Rosengren, 114–156. ——— (2003): “On the form and interpretation of German wh-infinitives”. Journal of Germanic Linguistics 15, 155–201. DOI: dx.doi.org/10.1017/S147054270300028X. Repp, Sophie (2013): “Common ground management: Modal particles, illocutionary negation and verum”. In: Gutzmann, Daniel & Hans-Martin Gärtner, eds.: Beyond expressives. Explorations in use-conditional meaning. Leiden: Brill, 231– 274. [This volume]. Rett, Jessica (2008): “A degree account of exclamatives”. Proceedings of SALT 18, 601–608. URL: http://ecommons.library.cornell.edu/bitstream/1813/13058/2/Rett .pdf. Romero, Maribel (2005): “Two approaches to biased yes/no questions”. Proceedings of the WCCFL 24, 352–360. Romero, Maribel & Chung-hye Han (2004): “On negative yes/no questions”. Linguistics and Philosophy 27, 609–658. DOI: dx.doi.org/10.1023/B:LING.00000338 50.15705.9. Rooth, Mats (1985): Association with focus. PhD thesis. Amherst, MA: University of Massachusetts. ——— (1992): “A theory of focus interpretation”. Natural Language Semantics 1, 75–116. DOI: dx.doi.org/10.1007/BF02342617. Russell, Bertrand (1905): “On denoting”. Mind 14, 479–493. URL: http://www.jstor.org .ubproxy.ub.uni-frankfurt.de/stable/2248381. Saka, Paul (2007): How to think about meaning. Dordrecht: Springer. Schlenker, Philippe (2007): “Expressive presuppositions”. Theoretical Linguistics 33.2, 237–245. DOI: dx.doi.org/10.1515/TL.2007.017. Schmitz, Hans-Christian (2008): Accentuation and interpretation. Basingstoke: Palgrave Macmillan. Schwarzschild, Roger (1999): “GIVENness, avoidF and other constraints on the placement of accent”. Natural Language Semantics 7, 141–177. DOI: dx.doi.org/10 .1023/A:1008370902407. Searle, John R. (1969): Speech acts. An essay in the philosophy of language. Cambridge: Cambridge University Press. von Stechow, Arnim & Dieter Wunderlich, eds. (1991): Semantik. Ein internationales Handbuch zeitgenössischer Forschung. Berlin: de Gruyter.

58

daniel gutzmann

Stenius, Eric (1967): “Mood and language game”. Synthese 17, 254–274. DOI: dx.doi .org/10.1007/BF00485030. Stevenson, Charles L. (1937): “The emotive meaning of ethical terms”. Mind 46, 14–31. DOI: dx.doi.org/10.1093/mind/XLVI.181.14. Strawson, Peter F. (1950): “On referring”. Mind 59, 320–344. DOI: dx.doi.org/ 10.1093/mind/LIX.235.320. Stump, Gregory T. (1993): “How peculiar is evaluative morphology?” Journal of Linguistics 29, 1–36. DOI: dx.doi.org/10.1017/S0022226700000037. Sudhoff, Stefan (2010): Focus particles in German. Syntax, prosody, and information structure. Amsterdam: John Benjamins. Sudo, Yasutada (2013): “Biased polar questions and Japanese question particles”. In: Gutzmann, Daniel & Hans-Martin Gärtner, eds.: Beyond expressives. Explorations in use-conditional meaning. Leiden: Brill, 275–295. [This volume]. Traugott, Elizabeth Closs & Richard B. Dasher (2001): Regularity in semantic change. Cambridge: Cambridge University Press. Truckenbrodt, Hubert (2006a): “On the semantic motivation of syntactic verb movement to C in German”. Theoretical Linguistics 32, 257–306. DOI: dx.doi.org/10 .1515/TL.2006.018. ——— (2006b): “Replies to the comments by Gärtner, Plunze and Zimmermann, Portner, Potts, Reis, and Zaefferer”. Theoretical Linguistics 32, 387–410. DOI: dx.doi.org/10.1515/TL.2006.025. Wegener, Heide (1989): “Eine Modalpartikel besonderer Art: Der Dativus Ethicus”. In: Weydt, Harald, ed.: Sprechen mit Partikeln. Berlin: de Gruyter, 56–73. Weydt, Harald (1969): Abtönungspartikel. Die deutschen Modalwörter und ihre französischen Entsprechungen. Bad Homburg: Gehlen. Williamson, Timothy (2009): “Reference, inference and the semantics of pejoratives”. In: Almog, Joseph & Paolo Leonardi, eds.: The philosophy of David Kaplan. Oxford: Oxford University Press, 137–158. Yule, George (2006): The study of language. 3rd ed. Cambridge: Cambridge University Press. Zanuttini, Raffaella & Paul Portner (2003): “Exclamative clauses: At the syntaxsemantics interface”. Language 79.1, 39–81. DOI: dx.doi.org/10.1353/lan.2003.0105. Zeevat, Henk (2013): “Expressing surprise by particles”. In: Gutzmann, Daniel & Hans-Martin Gärtner, eds.: Beyond expressives. Explorations in use-conditional meaning. Leiden: Brill, 297–320. [This volume]. Zimmermann, Malte (2004): “Discourse particles in the left periphery”. In: Shaer, Benjamin & Werner Frey & Claudia Maienborn, eds.: ZAS Papers in Linguistics 35, 543–566. Zwicky, Arnold (1985): “Clitics and particles”. Language 61, 283–305. DOI: dx.doi .org/10.2307/414146. Zwicky, Arnold & Geoffrey K. Pullum (1987): “Plain morphology and expressive morphology”. Proceedings of BLS, 330–340. URL: http://ling.ed.ac.uk/~gpullum/ bls_1987.pdf.

GERMAN NON-INFLECTIONAL CONSTRUCTIONS AS SEPARATE PERFORMATIVES

Sebastian Bücking and Jennifer Rau

1. Introduction In this paper, we will be concerned with the form and interpretation of German Non-Inflectional Constructions (= NICs), which are widely used in web-based communication like chat or (informal) email, as well as in comics. Following Teuber (1998) and Schlobinski (2001), we will argue that NICs do not build upon an impoverished grammatical competence, but in fact exhibit intriguing formal and functional characteristics worthy of linguistic scrutiny. Simple examples are given in (1) and (2); (3) provides attested NICs in a more elaborate context. English does not have corresponding forms containing modification and arguments, so any close idiomatic translation would suggest unintended similarities with full-fledged constructions. Therefore, we give only word-for-word glosses for NICs; for the sake of transparency, we use the gloss stem for verbal stems lacking inflectional morphology. Contextualized examples are amended by full translations in order to assure readability.1 (1) (2) (3)

•grins• smile.stem

•dich in den Arm nehm• you.acc in the arm take.stem

a. Synchro-Frank: Lichtausknips …… •Klick• […] Gute Nacht. […] light.off.turn.stem …… click.stem Good night

1 (3) is taken from http://www.busfreunde.de/read.php?1,293725, date: 12/29/2009. Authentic NICs are usually marked by surrounding “*…*”. Since “*” conventionally mark ungrammatical examples in linguistic texts, we will use • instead for highlighting NICs. This means that we slightly adapt authentic examples in order to avoid confusion. (3) shows that complex NICs are sometimes written in one word. We will adopt this for the rating study; but otherwise we will not consider orthography in any detail (see Schlobinski 2001 and footnote 7 for remarks on the orthographical realization of NICs).

60

sebastian bücking and jennifer rau b. PxCbulli: Wie? So früh schon? Nee, ein paar Stündchen What? That early already? Come on, some hours müssen noch! •lichtwiederanknips• must still! light.again.on.turn.stem

Hast Du was Wichtiges vor heute Nacht so um 4, oder Have you something important on tonight around 4, or warum steigst Du schon so früh aus? Gruß, Henrik why get you already so early off? Regards, Henrik

c. sMurF: •birne rausdreh• light.bulb out.take.stem

so nu ist wieder dunkel und ich kann in ruhe abdösen hier :zzz Well, now is again dark and I can in peace snooze here :zzz ‘Synchro-Frank: (NIC: light.off.turn.stem), (NIC: click.stem). Good night PxCbulli: What? That early already? Come on, some hours still have to be! (NIC: light.again.on.turn.stem). Do you have something important on tonight around 4 or why do you get off that early? Regards, Henrik sMurF: (NIC: light bulb out.take.stem) Now it is dark again and I can snooze unhindered.’

All three examples suggest that NICs are characterized by specific formal traits: they seem to possibly include VP-internal structure with the V-head in final position, but there are neither obvious inflectional morphemes nor overt subjects. The elaborate example in (3) is suggestive of the particular function of NICs: the first user Synchro-Frank turns the lights off, expressed by the formally prototypical NIC •lichtausknips• (‘light.off.turn.stem’), intending to close the discussion in the newsgroup for this day. Pragmatically, it does not make sense to assert that the user is turning lights off in “real life”, for instance at his own home. His lights at home are irrelevant for the virtual lights in the newsgroup.2 In a creative way, this performance can be responded to by turning the light on again and, finally, by taking the light bulb out in the virtual reality of the newsgroup, cf. the contributions by the users PxCbulli and sMurF. Crucially, turning off the light, turning it on and taking out the light bulb can be performed by uttering it using a NIC. We will

2 Of course, the speaker could also utter Ich mach jetzt das Licht aus (‘I turn off the light now’)—referring to the real lights at his home—and thereby convey that he is about to quit and go to bed. However, this message is only indirectly deduced from an assertion; other than in the example above, no lights are turned off by the utterance. We come back to this major point in Section 4.

german non-inflectional constructions

61

argue that such performative force is characteristic for the function of NICs in general. The paper is structured as follows: after giving an overview of the previous work on NICs in Section 2, we will show in Section 3 that NICs are robust constructions which have the following formal properties: they are syntactically independent and consist of verb stems potentially projecting internal structure; but they lack any inflectional morphology and overt subjects and are deficient in terms of functional structure. These assumptions will be supported by the results of a study on grammaticality judgements of NICs and by corpus data. In Section 4, we will show that NICs display typical properties of separate performatives according to criteria as discussed e.g. in Potts 2005, 2007c and Portner 2007. In particular, it will be argued that NICs are qua their form explicit performatives. In order to capture these findings, we will suggest a formal representation in Section 5. In Section 6, we will discuss how the pragmatic function of NICs can be linked to their specific formal properties as bare verbal structures. Section 7 summarizes our results. 2. Previous Work There are basically two linguistic papers on NICs in German, the first one by Teuber (1998) who describes basic properties of NICs and in particular addresses formal and morphological issues.3 The second one, Schlobinski 2001, investigates the syntactic and pragmatic characteristics of NICs including their sociolinguistic dimension. The German term for Non-Inflectional Constructions, “Inflektiv”, was established by Teuber (1998). He argues that NICs are neither interjections (as maintained in Hentschel & Weydt 1994: 299; see also Duden 2006: § 892) nor truncated verbal forms but that NICs are basic verbal forms in the verbal paradigm. Diachronically, Teuber assumes that NICs were permitted due to a morphological change: in previous stages only verbal forms contained the root forming suffix ô, which was attenuated to ə. Due to this change, 3 To be fair, Teuber also considers pragmatics by relating NICs to “one-word-sentences” and assigning them the illocutionary function of “comments”. However, although NICs may have a one-word-flavor, they are more complex structures (see below for a detailed discussion of corresponding evidence). Moreover, the notion of “comments”—being defined as expressions that convey a subjective feeling, an evaluation or an opinion towards a given situation—remains rather vague as Teuber himself concedes, cf. Teuber 1998: 23. For instance, any ordinary assertion can convey subjective feelings, evaluations or opinions as well. In our opinion, the crucial question of how specifically NICs take over such a function remains unanswered in his approach.

62

sebastian bücking and jennifer rau

verbal stems and roots became identical. This allowed the emergence of non-inflected forms like NICs—being identical to the verbal stem—as part of the verbal paradigm.4 Modern High German thus offered a construction which is verbal but not inflected. We will follow his insight here and assume that NICs are verbal stems, cf. (4). Further evidence for this position will be drawn from the empirical studies in Section 3. (4)

a. schluck-t gulp-prs.3sg b. •schluck• gulp.stem

Consequently, Teuber assumes that NICs are not a recent construction (which is due to comics or web language) but something which is generally provided in German grammar. On the one hand, it is thus highly special and peculiar morphologically, but on the other hand, being the verbal basic form, it corresponds to a fairly ordinary part of the verbal paradigm. He points out that NICs like knall (‘bang.stem’) were already reported by Adelung (1782: 207), hence claiming that the rise of NICs lies before the 20th century and has typological rather than sociolinguistic roots. In contrast, Schlobinski (2001) assumes that NICs may have appeared sporadically before but root as a specific construction in the language of comics and spread out from this genre into other registers. In particular, he considers the proto-NICs translations of English “sound words”, for instance klopf (‘knock.stem’) or bimmel (‘jingle.stem’) in Mickey Mouse translations from the early 1950s. Functionally, Schlobinski divides noninflectional constructions in two broad classes: first, NICs can correspond to immediate representations of actions and internal states as in the case of •grins• (‘smile.stem’) and •bibber• (‘jitter.stem’). Schlobinski considers this function the original one that is already exemplified by sound words in comics. These—in virtue of being iconic signs of the respective “sounding” events—also serve as immediate representations. Second, according to Schlobinski, there are NICs which are ordinary illocutionary acts, having assertive, expressive and regulative illocutionary forces. For instance, he argues that NICs as in (5) (= his example (39)) are assertive and form a narrative plot:

4 See Teuber 1998: 12–18 for details concerning the role of non-inflected forms in the whole verbal paradigm of Modern High German.

german non-inflectional constructions (5)

63

a. Nobse: •zum See sprint• to.the lake sprint.stem

b. KITE: •hinter Nobse herspring• behind Nobse jump.stem

c. Nobse: •in den See spring• […] in the lake jump.stem

d. Nobse: •Kite unter Wasser tunk• Kite under water dunk.stem

e. KITE: •japs und wasser schluck• gasp.stem and water swallow.stem

Although we agree with many of his intuitive observations, we do not share his conclusions and way of approaching and capturing the findings. First, it would be too simple to merely trace back NICs to sound word translations from Mickey Mouse. This extralinguistic explanation underestimates the impact of structural prerequisites; in particular, it is unable to explain the contrast between English and German because one would expect (complex) NICs in both languages if the grammar provided the expressions. To be sure, NICs are in fact distributed frequently in comics, in chats, informal emails, and recently also in newspaper texts. Nevertheless, we will show that such a distribution is an epiphenomenon: the typical pragmatic properties of NICs predestine them for virtual contexts. Second, it is misleading to identify the functional contribution of NICs with ordinary illocutions. In fact, the NICs in (5) are intuitively not identical in meaning to corresponding assertions as Ich sprinte zum See (‘I sprint to the lake’). Instead, by heavily relying on the notion of explicit performative, we will stick to Schlobinski’s original intuition that NICs are in a certain sense “immediate representations” and thus basically uniform in their function. This perspective will be generalized. The seemingly “assertive” effect of examples like (5) is derived from more general considerations.5 Third, Schlobinski’s (2001) approach does not provide

5 In our opinion, Schlobinski’s (2001: 197–204, 214 f.) classification is not fully transparent to begin with. For instance, a NIC as stamp (sic!) (‘stamp.stem’) is classified both as immediate representation of a certain action without being an illocution proper and, at the same time, as a communicative action with a—rather indirectly deduced—illocutionary potential. In a way, such a split correctly describes its peculiar function. However, what exactly does it mean to have and not have illocutionary force? And are other NICs really different from stamp in being clear-cut exemplifications of either class? Our approach will avoid these problems by relying on a more rigid understanding of (assertive) illocution and by explicitly deducing illocutionary effects on the basis of a uniform analysis of NICs as explicit performatives.

64

sebastian bücking and jennifer rau

a way of linking form and function. Actually, the idea of NICs as assertions runs into obvious troubles with the more or less standard assumption that assertions are closely connected to finiteness, cf. Klein 1998.6 Although we will not be able to give a full account, our analysis will bear upon the desideratum of an appropriate form-function fit. 3. Formal Properties The examples given in (1), (2) and (3) suggest that there is a specific makeup of NICs, i.e. no inflecional morphology, verb final VP-construction and lack of (overt) subjects.7 We will confirm this hypothesis on the basis of a

6 Similarily, Reis (2003: 197) points out that “finiteness is a necessary condition for SAYing X”, cf. (i) (= her example (64a)):

(i)

a. Du Linguist! you linguist b. Du bist ein Linguist. you are a linguist

Verbless constructions like (ia) cannot be assertional; there is no anchoring in terms of grammar proper. 7 This corresponds to the observations in Schlobinski 2001, who also assumes that the word order corresponds to verb-final (infinitival) word order. We will not be concerned with incorporation here. Under this label, Schlobinski discusses correlations between the feasibility of NICs, their orthographic realization, the construction’s length and complexity, and different kinds of “relatedness to the verb”. To our mind it remains unclear what exactly constitutes incorporation in his approach. He seems to heavily rely on the writing of NICs in one word. However, we see two related problems with his perspective: first, his discussion blurs the distinction between competence and performance. Orthography as extragrammatical phenomenon can merely suggest but never provide conclusive evidence in favor of a specific structural make-up. Second, if orthography is considered a reflex of grammatically relevant features, these have to be made transparent. This remains a desideratum in Schlobinski’s work. In the case at hand, we suspect that orthography might be used in order to point to phonological and corresponding informational differences in terms of “integration” (rather than “incorporation”), cf. Jacobs 1999, 2005 (we thank one reviewer for proposing this line of thought). We cannot go into details here; let us just remark that the VP-structure in fact allows for corresponding fine-grained structural differences. For instance, building inter alia upon Jacobs 1999, Maienborn (2001) distinguishes V-adjoined locatives and VP-adjoined cases. Under neutral stress conditions, the former is considered prosodically and semantically integrated, whereas the latter is not, cf. the interpretive and prosodic differences of (ia) versus (ib) (following her example (11); capital letters mark primary sentence accent, stress marks secondary accent): (i)

a. Angela hat sich mit Bardo im MUSEUM verabredet. Angela has refl with Bardo in the museum arranged.to.meet ‘The appointed event is located in the museum’

german non-inflectional constructions

65

grammaticality judgement survey and on corpus data. Subsequently, we will propose and discuss a basic syntactical structure for NICs. 3.1. Grammaticality Judgements In a web-based study, 41 German native speakers were asked to rate sample expressions on a scale from 1 to 7 (1: little acceptance, 7 highest acceptance). Appropriate chat-like samples were followed by a NIC which was marked with “*”8 (and written in one word) in order to trigger the environment typical for NICs. Participants were presented 12 items in 4 conditions. For illustration, see (6): (6)

Hi zusammen! So doof es klingt: Ich krieg meinen PC net an. Hab dem ein neues Gehäuse spendiert, alles umgebaut, angeschlossen, aber nichts funktioniert. *hilfebrauch* ‘Hi everyone! It may sound stupid, but I can’t get my PC going. Gave it a new case, reassembled everything, plugged it in, but nothing works.’ help.need.stem

In this study, NICs were presented under the following conditions: a combination of the opposing features ± subject and verb final vs. verb fronted. To be clear, hilfebrauch was thus presented in four versions:

b. Angela hat sich mit Bardo im Muséum VERABREDET. Angela has refl with Bardo in.the museum arranged.to.meet ‘The arrangement takes place in the museum’ NICs allow for the same distinction with the same interpretive effect, cf. the following contrast: (ii)

a. •im MUSEUM verabred• in.the museum arranged.to.meet.stem b. •im Muséum VERABRED• in.the museum arranged.to.meet.stem

We leave for further research the question of whether, in fact, cases such as (iia) are preferably written in one word whereas those in (iib) are more likely to be written separately (here, the modifier is a full phrase; we speculate that this is an important intervening factor). One may add that those modifiers for which Schlobinski (2001: 212) notes a clear preference for one word spelling are manner modifications. For these a very close relation to the verbal head is independently motivated. See the discussion on modification below. 8 Note that in this case, we present the example here as used in the study, i.e. using the typical environment “*…*”.

66 (a) (b) (c) (d)

sebastian bücking and jennifer rau hilfebrauch brauchhilfe ichbrauchhilfe ichhilfebrauch

(– subject, verb final) (– subject, verb fronted) (+ subject, verb fronted) (+ subject, verb final)

The mean judgements for all NICs under these conditions are summarized in Table 1. Table 1: Mean judgements for NICs depending on distribution of verb and subject

The differences between conditions across items were highly significant (p frame and domain adverbials (e.g. reference time related temporals) > event-external adverbials (e.g. causals) > highest ranked argument > event-internal adverbials (e.g. eventrelated locatives and temporals) > process-related adverbials (e.g. manner adverbials)

According to the corpus data, the majority of modified NICs include manner adverbials, as in (14): (14) •schnellersegel•, •traurigguck•, •fiesgrins•, •fix fast.er.sail.stem, in.a.sad.way.look.stem, in.a.nasty.way.grin.stem, quickly die folie abpiddeln tu• the film scrape.off do.stem

(Chat-Corpus)

These adverbials are considered structurally very low, projected in direct proximity to the verbal head.21 Albeit less frequent, locatives, temporals and causals are also attested, cf. (15): (15) a. •zündel hier oben• kindle.stem here above

b. •gleich mal erstmal pennen geh• shortly part for.the.time.being sleep.inf go.stem c. •rumhust wegen der staubwolke• cough.stem because.of the cloud.of.dust

(Chat-Corpus) (Chat-Corpus) (Chat-Corpus)

19 In terms of incorporation, Schlobinski (2001) briefly discusses adverbial modification in NICs. However, as noted in footnote 7, his discussion is mainly about orthography and does not spell out where grammar proper comes into play. Moreover, Schlobinski does not relate his remarks to current semantic and/or syntactic analyses of adverbial modification. So he speaks of verb-related versus proposition-related adverbials without further qualifications. Therefore, his argument remains quite opaque. In particular, he says that 31% of NICs with sentence adverbials are written in one word: this presupposes the existence of NICs with sentence adverbials; however, he does neither define his notion of sentence adverbial nor give an example. To our findings, sentence adverbials (in the sense of commenting on an asserted proposition) are not feasible to begin with, see below. 20 The notation follows the adaption of Frey’s proposal in Maienborn & Schäfer 2011. 21 A very low position is also maintained for directional adverbials (cf. Maienborn 1996); these can be found as well, as in (i):

(i)

•inneEckekriech• in.a.corner.crawl.stem

(Chat-Corpus)

74

sebastian bücking and jennifer rau

According to Frey’s hierarchy, all of these, pace differences with regard to the subject position (see the discussion above), are projected in the direct vicinity of VP. The given VP-analysis of NICs thus correctly predicts that there is room for modifiers that directly target different verbal layers. In Frey’s (2003) classification, sentence adverbials describe how the speaker comments on the clause’s proposition, cf. evaluatives as leider (‘unfortunately’) or epistemic modifiers as wahrscheinlich (‘probably’).22 The chat corpus does not contain any NICs with a sentence adverbial. In fact, constructed examples like (16) seem to be highly marked, if not outright ungrammatical: (16) a. *?•leider das Problem nicht versteh• unfortunately the problem not understand.stem b. *?•wahrscheinlich bald schlafen geh• probably soon sleep.inf go.stem

Frey assigns sentence adverbials a structurally high attachment site; in particular, he argues that sentence adverbials have to c-command the base position of the finite verbal form (cf. also Haider 2000). One may add that approaches with an articulated functional structure often associate sentence adverbials with structural levels higher than the VP, e.g. the C-domain (cf. the discussion in Grohmann & Etxepare 2003). The lack of sentence adverbials in NICs thus provides additional evidence for our claim that the structure of NICs does not contain finite morphology; furthermore, no need for a corresponding functional structure above the VP arises. What is going on between the C- and V-domain is rather difficult to assess, both for theoretical and empirical reasons: there is one good candidate for a frame-setting interpretation of an adverbial in the corpus, cf. (17a). Here, the PP seems to restrict the domain in relation to which the speaker considers something a justified action. Similarly, the domain adverbial politisch (‘politically’) in the constructed example (17b) may be possible: (17) a. •korrekte aktion find bei solchen nicknames• correct action consider.stem in.view.of such nicknames (Chat-Corpus) b. •politisch so gut wie tot sei• politically virtually dead be.stem

22 Other than e.g. Maienborn & Schäfer (2011), Frey does not subsume frame adjuncts under the label “sentence adverbial”, see below for discussion.

german non-inflectional constructions

75

Frey (2003) shows that frame adverbials do not enforce a c-commandrelation with the finite verb; this might explain why they are not generally ruled out in NICs. However, it nevertheless remains an open question whether some kind of functional structure above the VP is involved. As far as we can see, there is no clear consensus on what kind of entity and structure frame-setting adverbials relate to.23 In view of missing clear-cut empirical evidence and these theoretical shortcomings, we leave this issue for further research and keep the structure simple.24 To sum up: the distribution of adverbials in NICs confirms the basic syntactic analysis of NICs as simple VPs involving a complex internal structure but remaining deficient with regard to functional levels above.25

23 Maienborn (2001) adjoins locative frames to a TopP-phrase (hosting the sentence topic) in the C-domain above sentence adverbials. However, in our opinion, Frey convincingly shows that frame adverbials have a lower attachment site. Alternatively, one might think of frame-setters as modifiers to a topic situation, cf. Klein’s (1994) notion of topic time and remarks in Maienborn & Schäfer 2011. If this situation is introduced via an Asp-head, one could conjecture that NICs include an AspP-level above the VP. 24 In the chat corpus, we found two further adverbial types in NICs, not yet discussed, namely (ia) with a mental-attitude adverbial and (ib) with a discourse-anaphoric adverbial:

(i)

a. •skeptischzuhör• skeptically.listen.to.stem

b. •allerdings gerade nen soundserver am testen bin• however at.the.moment a sound.server at testing am

(Chat-Corpus) (Chat-Corpus)

Mental-attitude adverbials are usually assigned a VP-internal attachment site (cf. Maienborn & Schäfer 2011); therefore no problem with the VP-analysis arises. Discourse-anaphoric adverbials, in being responsible for linking discourse parts, may not contribute to the internal make-up of their host expressions anyway. We leave aside how and where they should be attached in the case of NICs. 25 One reviewer asks whether there are verb type restrictions for NICs. As far as we can tell, there are no clear restrictions with regard to usual distinctions such as “unaccusatives vs. ergatives” or Vendler’s (1967) aspectual classes. However, questions arise if one looks at auxiliaries, in particular modals. There are few examples, cf. (i): (i)

•malmeinauftretenhiergewaltigüberdenkenmuss• part.my.behavior.here.enormously.rethink.inf.must.stem

(Chat-Corpus)

This shows that modals are not generally ruled out. Note that (presumably) ungrammatical forms as *•könn• (‘can.stem’), *•müss• (‘must.stem’), etc. do not contradict this claim. NICs are not derived from the infinitive forms, i.e. können (‘can.inf’), müssen (‘must.inf’), etc.; they are stems that can be different from the ones in infinitive forms (cf. the well-known peculiarities of the preterite-present inflection of modals in German; see also Schlobinski 2001: 209 on the alternating stems will/woll (‘want.stem’)). The main question is to what extent modal NICs bear upon the given syntactic and semantic analysis. The issue deserves careful scrutiny, both theoretically and empirically; here, we can only make some preliminary remarks. First, modals seem to be rather marginal. Second, if used, they presumably have a

76

sebastian bücking and jennifer rau

So far, we have shown that there is a clear-cut non-inflectional construction which can be constructed and identified according to distinctive formal properties. In the following sections, we will argue that it has specific pragmatic properties which can be related to its formal make-up. 4. NICs as Separate Performatives Potts (2005, 2007c) and Portner (2007) argue that expressives such as bastard or damn as well as certain constructions like appositions and vocatives contribute to a meaning dimension that is both separate from the ordinary truth-conditional content and performative in character. In order to substantiate this claim, Potts enumerates a range of criteria that identify linguistic expressions as bearers of such additional performative meaning components. In particular, Potts argues that expressives are (i) nondisplaceable (i.e. they are bound to the utterance situation), (ii) perspective dependent (i.e. always viewed from a certain perpective, most often the speaker’s, possibly also from that of a “contextual judge”), (iii) descriptively ineffable (i.e. difficult to exactly reformulate in propositional terms), and (iv) immediate (i.e. performative due to being automatically true by virtue of being uttered). In the following we will show that NICs obey these criteria in a characteristic way; their behavior thus provides good evidence for the analysis of NICs in terms of separate performatives. We will first focus on descriptive observations; technical details follow in Section 5. First, that NICs belong to a separate meaning dimension is indicated by their being nondisplaceable: unlike ordinary descriptive meaning which can be temporally shifted and may relate to different worlds, NICs are invariable predicates of the utterance situation, as the following contrast illustrates: root but not an epistemic reading, cf. for an overview of this distinction e.g. Hacquard 2011 and example (i) for exemplification. In syntactic analyses, root modals are usually associated with scope at the VP-level and epistemics with scope at the sentence level (e.g. the C-domain); see again Hacquard 2011 for a summary of arguments and more precise suggestions. Structurally, our simple VP-analysis of NICs hence seems to be basically correct (pace possible ModPs above, depending on the general syntactic framework). However, it is not clear to us how to interpret an example as (i): in our analysis (see the following sections for details), it is supposed to be an explicit performative. Consequently, does “I am hereby obliged to think about my behavior” correctly render the meaning of (i)? Maybe, we have to accept that such a performance is in fact possible in the virtual extension of the real world (see Section 5 for the notion of virtual extension). The rather marginal status of NICs with modals might be explained by the assumption that such modal interpretations are not very plausible candidates for performances.

german non-inflectional constructions

77

(18) Gestern hatte ich ein Date mit meinem Ex-Freund. ‘Yesterday, I had a date with my ex-boyfriend.’ a. Ich werde / würde / wurde rot. I.nom become.prs.1sg become.irr.1sg become.pst.1sg red. ‘I am blushing/would blush/blushed’ b. •rot werd• red become.stem

In the case of continuation (18a), the dating event and the blushing event can be freely related in terms of worlds and time, while in (18b), blushing is necessarily situated in the utterance time and world, i.e. the speaker turns red while making the utterance.26 Strikingly, the semantic nondisplaceability of NICs is reflected in the fact that they cannot be embedded under complementizers or in bare form. This is illustrated in (19):27 (19) a. A: Was soll jeder sehen? A: What should everyone see? B: *•dass Dich in den Arm nehm• B: that you in the arm take.stem b. A. Was möchtest Du tun? A: What want you do?

B: *Ich möchte •dich in den Arm nehm• B: I want you in the arm take.stem

26 He does so in a virtual extension of the real world. That is, he does not have to blush in the real physical sense; cf. Section 5 for qualifications. One reviewer points out a picture of a comic strip cited in Schlobinski 2001: 200 where the NICs •kick• (‘kick.stem’), •kuppel• (‘gang.stem’) and •schalt• (‘change.gears.stem’) seem to refer to the past whereas the NIC •schnarr• (‘vibrate.stem’) refers to the present. However, the example does not disprove the nondisplaceability of NICs; on closer inspection, it rather offers evidence in favor of it. The picture illustrates the initial stages of a motorcyclist’s speedy start. Those three NICs with putative past reference are smaller than the one with present reference, thereby giving the impression of background information, and, more importantly, they are localized not with the motorcylist but with the cloud of dust left behind. That is, the NICs are drawn where the corresponding actions were carried out. In this sense, they do not refer to the past but to the present time of the situation they were in. Since this situation is over, a putative past reference arises. More generally, the example exhibits iconic strategies to bundle a temporal sequence within one picture; usually, the order of events is reflected by multiple pictures. 27 To be sure, examples like (9) with embedding do not provide any counterevidence. In (9), the NIC is not embedded but provides itself the embedding matrix constituent.

78

sebastian bücking and jennifer rau

We conclude that NICs are per se syntactic orphans without any potential of being integrated into a higher structural configuration.28 Second, (18b) also shows that NICs are generally bound to the speaker’s perspective: the speaker is usually identified with the implicit subject. In particular, it would be infelicitous here to identify the referent that is turning red with the hearer.29 Note that this is not a conceptual restriction since a continuation as in (20) that makes use of an ordinary finite form is perfectly possible: (20) Gestern hatte ich ein Date mit meinem Ex-Freund. Du wirst ja rot! Was zum Teufel hast Du damit zu tun? ‘Yesterday I had a date with my ex-boyfriend. You are blushing. What the hell do you have to do with that?’

Third, NICs are in a telling way descriptively ineffable. At first sight, an example as (21) could just be reformulated by saying Ich grübel gerade (‘I am meditating at the moment’). (21) •grübel• meditate.stem

In fact, however, such a paraphrase is not satisfying because it merely describes the assumed action but lacks the performative flavor of the original NIC. This observation is closely related to the following core trait NICs exhibit. Fourth, and most importantly, NICs are performative in a very peculiar sense: uttering a NIC is an act of performing something, thus true by virtue of being uttered. Moreover, the performed action is directly encoded in the NIC itself, i.e. NICs are explicit performatives. Therefore, one can react to the NIC in (22a) by rejection like in (23b) but not by negation like in (23a). In contrast, the assertion in (22b) can be negated. Hence, as generally assumed

28 This contrasts with expressives like bastard, which are syntactically embedded though semantically separate. While Potts (2007c) therefore must assume a mismatch between structure and meaning for a range of expressives, the isolated structure of NICs is perfectly in line with its separate meaning. Moreover, the inevitable structural independence of NICs distinguishes them also from other orphan-like constructions such as certain appositives, corrections or dislocations. These might be analyzed as structurally independent units as well, but in contradistinction to NICs, they in principle have the formal potential to get linked to ordinary full-fledged linguistic material. 29 As suggested for non-speaker oriented expressives by Potts (2007c), one can imagine an implementation of a judge as well. Thereby the speaker would be allowed to play different “roles”. In fact, an identification of the denotation of the silent subject with the hearer in example (18b) seems to be possible if the speaker puts himself into the hearer’s role.

german non-inflectional constructions

79

for performatives, NICs are evaluated with regard to felicity conditions and not to truth conditions.30 (22) a. A: •grübel• meditate.stem

b. A: Ich grüble. I meditate.prs.1sg (23) a. B: Du grübelst nicht, Du tust nur so. ‘B: You are not meditating, you are just pretending.’ b. B: Hör auf damit. ‘B: Stop it.’

The specific performative effect of NICs is also evident in attested examples. Recall example (3) in the introduction, whose interpretation has to make use of performances in a virtual reality. Such a performative effect is illustrated here once more by the following attested example:31 (24) a. horsedog: guten Morgen!!!! •Kaffeeausschenk• good morning!!!! coffee.serve.stem b. Meuli: •Kuchen in den Thread stell• cake into the thread put.stem

c. Flauscheding: […] •Becher hinhalt• :-) Guten Morgen ihr Lieben! cup hand.stem :-) Good morning, my dears! […] d. horsedog: es MUSS ein guter Tag werden – die Sonne scheint, die it MUST a good day become – the sun shines, the 30 One reviewer suggests that it might be possible to react to (22a) by uttering something like (i).

(i)

Komm, tu doch nicht so. Du erwartest doch nur, dass die anderen come.on, do part not like.this. You expect part only that the others für dich denken. for you think. ‘Stop pretending. You just expect the others to think for you.’

We would like to point out that the comment that speaker A was pretending presupposes nonetheless that speaker A was performing a meditating action instead of just asserting it. It is true that it does not necessarily have to be true because the speaker could simulate. Similarly, it is possible to lie using expressive interjections like aua (‘ouch’ in order to signal pain) as well. This, however, is a secondary effect and does not weaken the point. 31 http://www.medi-learn.de/medizinstudium/foren/archive/index.php/t-47304-p-12 .html; date: 12/29/2009.

80

sebastian bücking and jennifer rau Vögel singen :-)) Ich hab super gut geschlafen! Danke für den birds sing I have super well slept! Thanks for the Kuchen, Meuli! cake, Meuli! ‘horsedog: good morning!!!! (NIC: coffee.serve.stem) Meuli: (NIC: cake into the thread put.stem) Flauscheding: […] (NIC: cup hand.stem) :-) Good morning my dears! […] horsedog: it must become a good day—the sun is shining, the birds are singing:-)) I have slept very well! Thanks for the cake, Meuli!’

In this conversation, speakers explicitly refer to the virtual reality of the coffee and of the cake, since the latter is served “into the thread”.32 At the same time, the NICs are responded to by actions, one is handing the cup (again via NIC), the second one is to thank for the cake (not via NIC but via a typical explicit performative). Both reactions provide evidence for our claim that NICs are interpreted as actions. Note that the observations in Schlobinski 2001 can be captured in our approach as well, based upon general considerations. Recall first that Schlobinski considers original proto-NICs “immediate representations”. In our account, such immediacy follows directly from the analysis of NICs as separate performatives.33 Second, Schlobinski claims that NICs can be analyzed as plot-oriented assertive contributions. Based on the argument and evidence from above, we reject this assumption and deduce the putative assertive force from the performative analysis. Consider the following minimal pair: (25) a. •kaffee ausschenk• coffee part.serve.stem

b. Ich schenke Kaffee aus. I serve.prs.1sg coffee part

Under Schlobinski’s approach, there is no way to distinguish between the two constructions. However, they clearly differ in function. (25b) is only 32

A thread is a chain of postings on a single subject. One might object that examples with sound words/NICs provide counterevidence to the performative analysis. Prima facie, in these cases there is no speaker-dependence. If a sound word/NIC like quietsch (‘squeak.stem’) is used in a comic, the reader fills the subject slot by a car or its wheels but not by a speaker or some other protagonist. However, we argue that in these cases the “speaker” is in fact the car or the wheels. This assumption is corroborated by the fact that in a corresponding drawing the sound word/NIC would be localized next to the presumable causer, i.e. to the car or more precisely the wheels. 33

german non-inflectional constructions

81

performative in the sense of being an assertion. Therefore it is not an explicit performative. The utterance describes the serving of coffee which is performed non-virtually (or virtually) but it does not amount to be this action itself. That is, assuming that the speaker makes a true statement (cf. Grice’s maxim of quality), the addressee only infers that the speaker in fact serves coffee. In (25a), the situation is reversed. By uttering (25a), the coffee is (virtually) served and the assertional effect is an epiphenomenon of the force of such an explicit performative. That is, since a performative is true by virtue of being uttered, the addressee can conclude that a corresponding assertion would be justified. We conclude with remarks on a fifth criterion Potts adduces for the identification of expressives: typically, expressives should be repeatable without redundancy. Instead, as shown in example (26), there is an effect of “strengthening the emotive content” (Potts 2007c: 167). (26) Damn, I left my damn keys in the damn car.

[= Potts 2007c, (34c)]

It has to be stressed that the property of repeatability applies only for emotional content. It seems that redundancy is not a core property of expressives. Beyond examples like (26), repeating non-truth-conditional content has a variety of effects. For instance, repeated calls, which usually serve to catch the addressee’s attention (see Portner 2007, who relies on Zwicky 1974), can be thought of as a strengthening of the request for attention, but they can also be seen as iterative sequences of (single) calls or stress an ongoing request to catch attention. The effects of repetition of expressive content needs much more research. We think that the case of NICs can be enlightening for this case. For NICs, doubling seems to open a variety of interpretations. There are NICs from emotional predicates which show the strengthening effect, e.g. see (27). (27) •freufreu• be.happy.stem.be.happy.stem

(Chat-Corpus)

In contrast, there are iterative and progressive readings of NICs as well: (28) a. •winkwink• wave.stem.wave.stem b. •klatschklatsch• clap.stem.clap.stem

(29) a. •schleppschlepp• carry.stem.carry.stem

b. •wartwartwartwart• wait.stem.wait.stem.wait.stem.wait.stem

(Chat-Corpus) (Chat-Corpus) (Chat-Corpus) (Chat-Corpus)

82

sebastian bücking and jennifer rau

We assume that generally, the interpretation of multiple NICs depends on the lexical semantics of the verb. Their common property is that doubling does not lead to redundancy as is the case for asserted (at-issue) content.34 Having described the central pragmatic properties of NICs, we will now turn to their formal representation. .

5. The Formal Representation of NICs as Separate Performatives Building upon Potts 2005; 2007c, Portner (2007) assigns two separate dimensions of meaning to sentences: (i) “at issue-meaning”, i.e. a sentence’s ordinary truth-conditional content, and (ii) “separate performatives”, i.e. additional non-truth-conditional content that supplies further instructions for the interpretation of the sentence. See example (30) from Portner (2007) for illustration; here, the epithet and appositive yield additional performatives:35 (30) Paul met that bastard Lance, a cyclist. (31) a. at issue-meaning: [λw. Paul met Lance in w] b. separate performatives: {[λw. I hereby assert that Lance is a cyclist in w], [λw. I hereby assert that Lance is a bad guy in w]}

34

(i)

Obviously, assertions follow different rules; their repetition leads to redundancy, cf. (i): #I’m angry! I forgot my keys. I’m angry! They are in the car. I’m angry! [= Potts 2007c, (35)]

This may carry over for separately performed assertions. The case of appositions and appositive relatives (analyzed as expressives in Potts 2005, 2007a) shows that repetition of nontruth-conditional content may turn out to be redundant as well, see (ii). (ii)

#Lance, a cyclist, who is a cyclist, is from Texas.

Assertions, no matter whether their content is at-issue or non-at-issue, have different felicity conditions than other speech acts. For performances like clapping, there are no restrictions on iterating these acts. 35 In order to make the performative character explicit, we follow Portner’s 2007: 413 suggestion to include “I hereby assert” in the paraphrase of both cases; an alternative would be “the speaker asserts”. Note that Portner himself omits this part in his formal derivation of the appositive case whereas he includes the speaker’s action in the case of vocatives (cf. “the speaker requests John’s attention” for John, your dinner is ready!). See the contrast between the representations (15a) and (17a) in Portner 2007: 414. In order to uniformly make the speaker’s action transparent, we opt for the explicit integration of the explicit performative in the form of “I hereby …” / “the speaker …”.

german non-inflectional constructions

83

The meaning of a sentence is thus a pair ⟨AS , CS ⟩ where AS corresponds to the regular compositional content and CS to the simple sum of all separate performatives introduced anywhere in the sentence. For the calculation of the two dimensions, Portner assumes two different interpretation functions, ⟦⋅⟧c for the ordinary content and ⟦⋅⟧Cc for the additional separate performatives.36 We propose to capture NICs as separate performatives in Portner’s terms, as in the following meaning representation: (32) a. ⟦NIC⟧c = ∅ b. ⟦NIC⟧Cc = [λPλw. I hereby make P(v) true in w]

P corresponds to the descriptive content of the subjectless verbal structure projected within NICs. Following our basic syntax from Section 3.3, this structure can include VP-internal arguments, predicates and modifiers. In order to integrate the missing argument (for the subject), P is applied to a free variable v that must be instantiated via contextual means.37 Applying the proposed semantics of NICs to the verbal structure rot werd (‘red become’) yields the meaning representation in (33) for the corresponding NIC •rot werd•:38 (33) ⟦NIC⟧(⟦rot werd⟧) =

a. ⟦•rot werd•⟧c = ∅ b. ⟦•rot werd•⟧Cc = [λw. I hereby make blush(v) true in w]

36 This strategy follows Rooth’s (1985) approach to focus semantics and is different from Potts’s type-theoretical solution for the parallel calculation of both dimensions, cf. Potts 2005, 2007c for details. Note furthermore that Portner (2007) uses both “separate performative” and “expressive meaning” for the separate meaning dimension. We will mainly use the first term because it directly captures the idea of a separately performed speech act that does not affect truth conditions. 37 As noted, the semantic proposal is based on the subjectless syntactic structure, cf. the discussion in Section 3.3. Hence, there is no grammatical guide for integrating the corresponding argument; that is why we resort to a free variable. If one instead adopts an analysis with a silent PRO for the subject, the semantic representation would be the following:

(i)

a. ⟦NIC⟧c = ∅ b. ⟦NIC⟧Cc = [λpλw. I hereby make p true in w]

Here, NIC takes as its first argument a full proposition with PRO being part of it. However, since PRO cannot be bound within the structure (recall that there is no embedding), it actually behaves as a variable to be filled contextually. This leads to the very same analysis as with a free variable v. So, we do not see how in our particular case the alternative syntactic option should make a semantic difference. 38 See Section 6 for compositionality issues.

84

sebastian bücking and jennifer rau

The representation in (33) captures the aforementioned traits of NICs as separate performatives in a fairly straightforward way: their nondisplaceability and immediacy follow automatically from their being mapped onto the second meaning dimension ⟦⋅⟧Cc whereas ⟦⋅⟧c remains empty. Such mapping encodes directly that NICs are performatives, i.e. bound to the utterance situation and true by virtue of being uttered. Since the “performance” is directly related to the verbal content, NICs are, moreover, explicit performatives; see below for qualifications. Since the speaker is explicitly assigned the role of a performer making P(v) true, it is natural to identify the free variable v with the denotation of the speaker. This correctly predicts the dependence of NICs on the speaker’s perspective. What about the particular descriptive ineffability of NICs? This nicely shows up in the way we have set up the contruction’s meaning contribution. If we take the identification of v with the speaker for granted, the NIC •rot werd• in less formal terms has the denotation in (34): (34) ⟦•rot werd•⟧Cc = [λw. I hereby blush in w]

That is, by uttering the NIC •rot werd• the speaker blushes. Prima facie, this is obviously impossible: in other words, the meta-language used in order to describe the meaning contribution of NICs resorts to verbalizations which are considered anomalous if taken at face-value.39 This “descriptive ineffability” results from the specific performative character we argue for. In the following, we briefly elaborate on this aspect and link the discussion to the question why there is an anchoring to the actual world w. The anchoring in w might come as a surprise since the corresponding actions do not seem to be really performed in the actual world. By using the NIC •rot werd• (‘red become.stem’), one does not in fact blush; in the case of •Dich umarm• (‘you embrace.stem’), if used in a chat, physical distance between the interlocutors even renders any actual performance impossible. However, according to the present idea, the construction’s specific effect is to do things linguistically which under normal circumstances cannot be instantiated by linguistic means. Especially the attested examples (3) and

39 One might suggest resorting to more involved formulations in order to avoid such a “problem”, cf. for instance:

(i)

⟦•rot werd•⟧Cc = [λw. speaker(c) is an instantiator who makes blush(v) true in w]

But this only clouds what is going on. We would rather say that there is no “problem” with the anomalous formulation to begin with; rather, it exactly reflects what NICs are about, see below.

german non-inflectional constructions

85

(24) from the introduction and Section 4 made this intuition palpable. In order to handle this seemingly inconsistent state of affairs, we argue for the following: NICs simulate an extension of regularly operative felicity conditions in such a way that verbs which are canonically non-performative can come across as performative. In this sense, the actions embedded within NICs are performed in the actual world, that is, they are performed in a virtual extension of the real world.40 This perspective can be seen as an amendment to Searle’s (1989) answer to the question of how performatives work. Searle (1989: 555 f.) argues that performative utterances are “self-referential to a verb which contains the notion of an intention as part of its meaning, and [that] the acts in question can be performed by manifesting the intention to perform them”. Crucially, in his account, performatives are basically declarations, and limitations of performativity are merely due to facts of nature and the real world, not to some special “performative” semantics. That is, the world as we know it confines performatives to those few cases where there is an institutional convention that words alone can change the world. As Searle (1989: 554) puts it: “If God decides to fry an egg by saying, ‘I hereby fry an egg,’ […] He is not misusing English”. In the case of NICs, the speaker overrules such conventional restrictions by a specific linguistic form. In a sense, the speaker plays god by virtue of linguistic means. The predominant use of NICs in chat communication and informal e-mail is instructive here: while in these letter-based forms of communications the interlocutors are physically distant from each other, they often aim at integrating some facets of immediate oral communication, thereby going beyond a simple exchange of information. One such facet is direct perception of actions that accompany the communicational setting. Therefore it seems highly desirable to have a linguistic form that could give the impression of such direct interaction. We argue that NICs in fact constitute such a form: they do not merely describe concomitant actions but present them as being in fact performed and thus despite factual distance accessible to immediate perception. Such reasoning also explains why expressive-emotive NICs as •lach• (‘smile.stem’), •grins• (‘smile.stem’), •gähn• (‘yawn.stem’) or •seufz• (‘sigh. stem’) are very widely used, cf. Schlobinski 2001: 204. Their base verbs describe actions which typically accompany oral communication; in this 40 We speak of a “virtual” extension of the real world in order to make a distinction from the notion of possible worlds. Virtual worlds are considered parts of the actual world and so are the performances via NICs.

86

sebastian bücking and jennifer rau

sense, the NICs imitate or double such concomitant circumstances. Hence the use of NICs is facilitated if the performances can be easily thought of as accompanying real actions. Note, however, that NICs are independently performative, cf. examples as •dich drück• (‘you hug.stem’) or •Dir Kaffee ausschenk• (‘you coffee serve.stem’) for which, as noted already above, there are no plausible counterparts in reality.41 Finally, the fact that repetitions in NICs are never redundant is predicted by our analysis as well. It is plausible that actions as clapping and carrying can be carried out multiple times, which captures iterative and progressive interpretations of examples like (28) and (29). For the strengthening effect with emotive content, general mechanisms seem to be at work: iterative readings of multiple emotive NICs are possible, as in (27), but are usually reinterpreted as referring to the same performative expressive act. As shown in Section 4, this phenomenon is not tied to NICs but holds for emotive expressives in general. As far as NICs are concerned, the predictions of our performative approach seem to be correct. So far we have merely described the peculiar performative character of NICs, disregarding the question of where the construction’s specific effects come from. In other words: how does the formal make-up of NICs shape and thereby explain their function? This issue is addressed in the following section. 6. Remarks on the Form-Function Fit In order to explain the specific performative function attested for NICs, two different approaches seem feasible. One may opt for a “constructional” account with an arbitrary relation between the form and function of NICs, or for a compositional “derivational” account that relates formal properties to function in a non-arbitrary way by resorting to independently motivated grammatical principles.42 In the following, we will not be able to give a clearcut analysis, but merely sketch strategies to explain the specific performative force of NICs. 41 Accordingly, using e.g. the NIC •lach• (‘laugh.stem’) does not necessarily include real concomitant laughing. The speaker can either perform it only via NIC (i.e., the NIC imitates the action) or he can both perform it via NIC and in real (i.e., the NIC doubles the action). 42 These two fundamentally different options are well-known from the literature on sentence types, cf. e.g. Altmann 1993 for a “constructional” account of sentence types and Brandt et al. 1992 for an early “derivational” approach.

german non-inflectional constructions

87

Within his proposal for appositives, Portner (2007) does not discuss in any detail how form and function are mapped onto each other. He merely assumes that the appositive structure triggers the following meaning shift:43 (35) a. ⟦a cyclist⟧c = [λxλw. x is a cyclist in w] b. ⟦a cyclist⟧Cc = ∅

(36) a. ⟦a cyclistappos ⟧c = ∅ b. ⟦a cyclistappos ⟧Cc = [λxλw. I hereby assert that x is a cyclist in w]

Our analysis of NICs from above can be viewed as a similar meaning-shift that empties the ordinary meaning component and renders it as a specifically modified separate performative. As they stand, both projections onto the second meaning dimension seem to be just construction-dependent effects. In contrast to this approach, we will now sketch two alternative proposals that are in the spirit of derivational accounts. 6.1. The Interpretation of NICs via a Silent Force-Indicating Operator There is a well-known tradition of encoding illocutionary force via forceindicating operators. The basic idea here is to analyze NICs in terms of such an operator, analogously to corresponding approaches to the semantics and pragmatics of canonical declaratives, interrogatives and imperatives. To be more specific, we briefly introduce one such approach, namely Han 2000 for imperatives, and then present a similar solution for NICs. We have chosen imperatives because they resemble NICs in a particular way, see below. Han assumes that there is a morpho-syntactic feature [imp] in the Cdomain of an imperative clause, as in the following simplified structure.44 (37)

CP C | [imp]

… … VP …V…

43

Again, we add the explicit performative marker to Portner’s representation, cf. footnote

35. 44 Most notably, Han (2000) assumes a separate Mood-projection in between; we will not address this issue here.

88

sebastian bücking and jennifer rau

Her main point with regard to the imperative’s force is that [imp] is directly mapped onto a force-indicating operator that she labels “directive”.45 Following the basic idea that such operators allow speakers to perform a speech act by their utterrance, “directive” applies to a proposition and returns a directive action. More specifically, according to Han, a directive action amounts to instructing the addressee to update his so called Plan Set, i.e. the discourse component where the addressee’s intentions are represented. Transferring such an operator-analysis to NICs, we propose an invisible morphosyntactic feature [ninfl] (symbolizing the lack of any inflection) on top of the non-inflected VP. For the sake of analogy, we let it project in C as well, cf. the structural sketch in (38).46 (38)

CP C | [ninfl]

VP …V…

If one then assumes that [ninfl] is directly mapped onto a force-indicating NIC-operator as defined in the preceding section, cf. (32), the desired effect of NICs is derived in a systematic bottom-up compositional fashion. The merits of this approach are the following: NICs are independent utterances with their own force; therefore, resorting to the very same mechanisms as for standard illocutions is attractive. Moreover, following the standard assumption that embedded clauses do not contain force-indicators, the principled non-embeddability of NICs results from the presence of the NIC-operator automatically. Finally, a neat parallel between imperatives and NICs is captured: in both cases, there is a close connection between morphological form and illocution. Considering [ninfl] a specific verbal mood feature, one may say that in both imperatives and NICs verbal and sentence mood correspond. 45 This is not meant to constrain imperatives to directives proper; Han rather considers the operator involved a place holder for sentential force. The exact interpretation is guided via pragmatic reasoning. 46 Note that we do not have any empirical evidence for such a position (cf. the discussion in Section 3.3); the proposal just follows Han’s (2000) solution. In particular, this means that that there might be reasons to assign [ninfl] to a separate ForceP (cf. Rizzi’s (1997) split of the C-domain) or to adjoin it to the VP without in fact having a functional structure as such (cf. remarks in Sternefeld 2006, who favors a CP-adjunction analysis of illocutionary operators without localizing them in a head position; analogously, we would then just opt for VP-adjunction in the case of NICs). We have to leave a decision about these options to further research.

german non-inflectional constructions

89

In order to avoid potential confusion, we would like to point out similarities and differences in the performative character of imperatives (and declaratives and interrogatives) on the one hand and NICs on the other. These are encapsulated in the meaning contribution of the operators. The NIC- and the “directive”-operator can be considered similar in adding the “content” of their argument—a proposition in Han’s (2000) analysis of the imperative, a property in the analysis of NICs above—in a particular way to a particular discourse component. With an imperative, the speaker instructs the addressee to add a proposition to his Plan Set. In the case of NICs, there is no plausible instruction to the addressee; instead, via NIC the speaker just adds a property to his Action Set.47 Informally, we define the Action Set as the discourse component that pairs each participant of a communicational setting with those properties that describe his actions during conversation. In this sense, the NIC is an explicit performative because the property as such gets listed in the speaker’s Action Set. In the case of imperatives, it is the instruction that gets listed. This instruction, however, is not part of the imperative’s underlying “content”, i.e. its proposition. Therefore, the descriptive linguistic material does not correspond to the action as such and we are not faced with an explicit performative.48 6.2. The Interpretation of NICs via Pragmatic Reasoning As an alternative to the operator approach, one might attempt a structurally more parsimonious account by making use of the most prominent formal feature in NICs, i.e. the lack of any evidence for inflectional morphemes distinguishing NICs from both finite sentences and bare infinitive forms.

47 Of course, the addressee is involved in the sense of perceiving this amendment. However, he cannot refuse it and there is nothing else to refuse. Looking at imperatives, interrogatives and declaratives: again, the speaker cannot refuse the perception of the performative as such (i.e. he has to accept that there is an instruction). However, he can refuse to follow the speaker’s instruction. 48 With regard to the two dimensions of meaning as assumed in Section 5, there remains an open question. In her analysis of the imperative, Han builds upon the idea that the meaning of a sentence is split into a force-component and a propositional component. In a similar way, Portner (2007) considers an analysis of imperatives in terms of a force operator that introduces expressive meaning, i.e. a separate performative. According to Portner’s proposal, the force marker for the imperative does not render the at-issue-content empty; see his representation (28) with the imperative’s “content” (in his analysis, this is a property) in the ordinary meaning dimension. If this is the case, NICs and ordinary illocutions differ in this respect since on our analysis the NIC-operator assigns the at-issue-meaning no value at all. However, we do not see whether, and if so, how such a difference becomes (empirically) relevant.

90

sebastian bücking and jennifer rau

Without any morphological marking, it seems reasonable that NICs are not part of syntax proper: they do not give rise to any potential functional anchoring; other than bare infinitive forms, they even do not allow for a parasitic connection to functional layers via embedding. Being in this sense necessary structural orphans, we conclude that NICs cannot be linked to the ordinary meaning dimension but must be automatically mapped onto the instructional secondary meaning component. Admittedly, this reasoning still does not explain how NICs receive their particular performative character. One could think of it in terms of pragmatic inferences. A reasonable sketch of the procedure might look as follows: due to their absolutely bare form, NICs are semantically underspecified in a radical way. However, the linguistic system—making use of inflection, word-order or prosody—provides many formally specific forms in order to convey ordinary illocutionary acts as assertions or directives (requests, commands). Consequently, in order to express these functions, one should choose the linguistic forms that are specified in the appropriate way. NICs as highly non-specific forms are therefore ruled out as potential bearers of such illocutions. In contrast, performatives as declarations in Searle’s sense, cf. the discussion in Section 5, do not seem to be formally defined. Instead, an explicit performative is an explicit performative if one conceptualizes it as such. Therefore, the according specification can be seen as a last resort strategy in order to fill the radically underspecified representation with interpretable content, i.e. with performative content.49 The identification of the VP’s open argument slot with the speaker can be seen as a further by-product of this strengthening: the speaker can describe some other person’s actions but he cannot possibly do them in his place. That is, ascribing one’s own (linguistic) performance to another person is generally ruled out. Therefore it is the speaker himself who is identified as the “instantiator”.

49 The indicated process of pragmatic strengthening, i.e. the blocking of certain interpretations on the one hand and the appropriate enriching of the given material on the other, can be modeled along the lines of some version of Grice’s (1975) cooperative principle of conversation and his conversational maxims. For instance, exploiting the maxim of quantity—more specifically its maximizing part “Say as much as you can” (= Horn’s (1984) Q-principle)—it seems to be predicted that a non-specific form cannot convey information for which a more specific, i.e. formally unequivocal form exists. Exploiting the cooperative principle and the minimizing part of the quantity maxim “Say no more than you must” (= Horn’s (1984) Rprinciple), every utterance must make communicational sense and can be appropriately strengthened. A more careful application of such maxims to NICs, e.g. framed in Blutner’s (2000) bidirectional optimality theory and its balanced view on the countervailing Q- and R-principles, must be left to future research.

german non-inflectional constructions

91

The main challenge for the pragmatic approach is to keep NICs apart from other (non-canonical) sentence types. For instance, Reis (2003) envokes a pragmatic account of wh-root infinitives as Wohin gehen? (“Where (should one/we) go”) that also relies heavily on missing finiteness (in root position) and the lack of an overt subject. Crucially, however, it yields fairly different results; specifically, it generates a modal interpretation that is not observed with NICs. While it seems obvious to localize the difference in the lack of any possible inflection in the case of NICs as opposed to non-finite morphology in the case of root infinitives, it is not clear to us where and how exactly to integrate this information into the pragmatic inference process. 7. Conclusion This paper was concerned with the form and function of German NonInflectional Constructions (= NICs). We argued that NICs are robust verbal structures with clearly distinct formal and functional properties: formally, NICs are potentially complex VPs with a verb-final head lacking both an overt subject position and any kind of inflectional morphology; functionally, NICs are separate performatives enabling otherwise not explicitly performative verbal content to come across as explicit performative. In Section 3, data from both a grammaticality judgement survey and a chat corpus provided clear empirical evidence for the analysis of NICs as bare verbal structures: in accordingly identified slots, subjectless VPs without overt verbal morphology are both rated best, and by far most frequently used. Furthermore, the discussion of complex examples from the corpus revealed that NICs allow for an elaborate internal structure but are deficient in terms of the functional layers above VP. Turning to the function of NICs, we have shown in Section 4 that NICs display typical properties of separate performatives as discussed in Potts 2005, 2007c and Portner 2007, i.e. they are nondisplaceable, bound to the speaker’s perspective, non-redundant in case of repetition, descriptively ineffable, and true in virtue of being uttered. In particular, we provided evidence in favor of their analysis as explicit performatives: by uttering a NIC, the speaker accomplishes the action that is described by the verbal content. In Section 5, the specific performative meaning contribution of NICs is captured within Portner’s (2007) model. Accordingly, NICs do not contribute to the ordinary meaning dimension but map the verbal content— modified to the extent that the speaker is said to automatically guarantee

92

sebastian bücking and jennifer rau

the truth of the corresponding proposition—to the separate performative dimension. This encodes the performative character of NICs in a direct way and, most importantly, captures why NICs are explicit performative speech acts used in order to “do” things linguistically which normally cannot be instantiated by linguistic means. Finally, in Section 6, we considered whether to link the form and function of NICs in an arbitrary “constructional” way or rather via non-arbitrary “derivational” means. Following the derivational paradigm, we presented two potential solutions. It was suggested to map form and function either by resorting to a silent force-indicating NIC-operator or else by pragmatic reasoning on the basis of a radically underspecified form. The most important remaining question, to our mind, thus seems whether one can find independent evidence in favor of either of these approaches. Acknowledgements We would like to thank the audience of AG 10 Expressives and other kinds of non-truth-conditional meaning at the DGfS 2009 in Osnabrück for helpful comments. In particular, we thank both reviewers and the editors, Daniel Gutzmann and Hans-Martin Gärtner, for very supportive and instructive suggestions on an earlier draft of this paper. All remaining errors are of course ours. We are indebted to Noah Constant for checking the English manuscript. References Adelung, Johann C. (1782): Umständliches Lehrgebäude der deutschen Sprache zur Erläuterung der Deutschen Sprachlehre für Schulen. Vol. II. Leipzig. Reprinted 1971 Hildesheim: Olms. Altmann, Hans (1993): “Satzmodus”. In: Jacobs, Joachim & Arnim von Stechow & Wolfgang Sternefeld & Theo Vennemann, eds.: Syntax. Ein internationales Handbuch der zeitgenössischen Forschung. Syntax. An international handbook of contemporary research. Berlin: de Gruyter, 1006–1029. Blutner, Reinhard (2000): “Some aspects of optimality in natural language interpretation”. Journal of Semantics 17, 189–216. DOI: dx.doi.org/10.1093/ jos/17.3.189. Brandt, Margareta & Marga Reis & Inger Rosengren & Ilse Zimmermann (1992): “Satztyp, Satzmodus und Illokution”. In: Rosengren, Inger, ed.: Satz und Illokution. Vol. I. Tübingen: Niemeyer, 1–90. Chomsky, Noam (1995): The minimalist program. Cambridge, MA: MIT Press. Cinque, Guglielmo (1999): Adverbs and functional heads. A cross-linguistic perspective. Oxford: Oxford University Press.

german non-inflectional constructions

93

Duden (2006): Duden. Band 4. Die Grammatik. 7th revised edition. Mannheim: Dudenverlag. Ernst, Thomas (2002): The syntax of adjuncts. Cambridge: Cambridge University Press. Frey, Werner (2003): “Syntactic conditions on adjunct classes”. In: Lang, Ewald & Claudia Maienborn & Cathrine Fabricius-Hansen, eds.: Modifying adjuncts. Berlin: de Gruyter, 163–209. Grice, H. Paul (1975): “Logic and conversation”. In: Cole, Peter & Jerry L. Morgan, eds.: Speech acts. New York: Academic Press, 41–58. Grohmann, Kleanthes K. & Ricardo Etxepare (2003): “Root infinitives. A comparative view”. Probus 15, 201–236. DOI: dx.doi.org/10.1515/prbs.2003.008. Hacquard, Valentine (2011): “Modality”. In: Maienborn, Claudia & Klaus von Heusinger & Paul Portner, eds.: Semantics. An international handbook of natural language meaning. Volume 2. Berlin: de Gruyter, 1484–1515. Haider, Hubert (2000): “Adverb placement. Convergence of structure and licensing”. Theoretical Linguistics 26, 95–134. DOI: dx.doi.org/10.1515/thli.2000.26.1-2.95. Han, Chung-hye (2000): The structure and interpretation of imperatives. Mood and force in universal grammar. New York: Garland. Hentschel, Elke & Harald Weydt (1994): Handbuch der deutschen Grammatik. Berlin: de Gruyter. Horn, Laurence R. (1984): “Toward a new taxonomy for pragmatic inference. Qbased and R-based implicature”. In: Schiffrin, Deborah, ed.: Meaning, form, and use in context. Linguistic applications. Washington, DC: Georgetown University Press, 11–42. van Hout, Angeliek & Tom Roeper (1998): “Events and aspectual structure in derivational morphology”. In: Harley, Heidi, ed.: Papers from the UPenn/MIT roundtable on argument structure and aspect 32. MIT Working Papers, 175–220. Jacobs, Joachim (1999): “Informational autonomy”. In: Bosch, Peter & Rob van der Sandt, eds.: Focus. Linguistic, cognitive, and computational perspectives. Cambridge: Cambridge University Press, 56–81. ——— (2005): Spatien – Zum System der Getrennt- und Zusammenschreibung im heutigen Deutsch. Berlin: de Gruyter. Klein, Wolfgang (1994): Time in language. London: Routledge. ——— (1998): “Assertion and finiteness”. In: Dittmar, Norbert & Zvi Penner, eds.: Issues in the theory of language acquisition. Essays in honor of Jürgen Weissenborn. Frankfurt: Peter Lang, 225–245. Kratzer, Angelika (1996): “Severing the external argument from its verb”. In: Rooryck, Johan & Laurie Zaring, eds.: Phrase structure and the lexicon. Dordrecht: Kluwer, 109–137. Maienborn, Claudia (1996): Situation und Lokation. Die Bedeutung lokaler Adjunkte von Verbalprojektionen. Tübingen: Stauffenburg. ——— (2001): “On the position and interpretation of locative modifiers”. Natural Language Semantics 9, 191–240. DOI: dx.doi.org/10.1023/ A:1012405607146. Maienborn, Claudia & Martin Schäfer (2011): “Adverbs and adverbials”. In: Maienborn, Claudia & Klaus von Heusinger & Paul Portner, eds.: Semantics. An international handbook of natural language meaning. Volume 2. Berlin: de Gruyter, 1390–1420.

94

sebastian bücking and jennifer rau

Portner, Paul (2007): “Instructions for interpretation as separate performatives”. In: Schwabe, Kerstin & Susanne Winkler, eds.: On information structure, meaning and form. Amsterdam: John Benjamins, 407–426. Potts, Christopher (2005): The logic of conventional implicatures. Oxford: Oxford University Press. ——— (2007a): “Conventional implicatures, a distinguished class of meanings”. In: Ramchand, Gillian & Charles Reiss, eds.: The Oxford handbook of linguistic interfaces. Oxford: Oxford University Press, 475–501. ——— (2007c): “The expressive dimension”. Theoretical Linguistics 33, 165–197. DOI: dx.doi.org/10.1515/TL.2007.011. Reis, Marga (2003): “On the form and interpretation of German wh-infinitives”. Journal of Germanic Linguistics 15, 155–201. DOI: dx.doi.org/10.1017/ S1470542703 00028X. Rizzi, Luigi (1997): “The fine structure of the left periphery”. In: Haegeman, Liliane, ed.: Elements of grammar. Dordrecht: Kluwer, 281–337. Rooth, Mats (1985): Association with focus. PhD thesis. Amherst, MA: University of Massachusetts. Schlobinski, Peter (2001): “*knuddel – zurückknuddel – dich ganzdollknuddel*. Inflektive und Inflektivkonstruktionen im Deutschen”. Zeitschrift für germanistische Linguistik 29, 192–218. DOI: dx.doi.org/10.1515/zfgl.2001.013. Searle, John R. (1989): “How performatives work”. Linguistics and Philosophy 12, 535–558. DOI: dx.doi.org/10.1007/BF00627773. Sternefeld, Wolfgang (2006): Syntax. Eine morphologisch motivierte generative Beschreibung des Deutschen. Tübingen: Stauffenburg. Teuber, Oliver (1998): “fasel schreib erwähn – Der Inflektiv als Wortform des Deutschen”. Germanistische Linguistik 141, 7–26. Vendler, Zeno (1967): Linguistics in philosophy. Ithaca, NY: Cornell University Press. Zwicky, Arnold (1974): “Hey, whatsyourname!” Proceedings of CLS 10, 787–801.

MODAL PARTICLES AND CONTEXT SHIFT

Sophia Döring

1. Introduction The phenomenon of context shift has received much attention due to Kaplan’s fixity thesis (Kaplan 1989), which claims that there are no ‘monster operators’ that shift the interpretation context of indexicals. Meanwhile, a number of counterexamples, i.e. several cases of shifted indexicals in different languages, have been reported. The present work deals with expressive words, viz. German modal particles (MPs). I claim that, referring to the speaker’s attitude, the expressive meaning of MPs is context-dependent, just as indexicals. The phenomenon of context shift, then, can shed light on this context dependency. For this article, three modal particles have been analyzed in different environments to investigate whether they have to refer to the actual speaker in every context. The results not only prove that the context for interpreting modal particles can be shifted, but they also show interesting differences between the selected particles. This might suggest that there are different classes of modal particles, which of course gives rise to the question how to categorize them. The first part of this article discusses the nature of modal particles, followed by an analysis of the meaning of ja, doch, and wohl. Section 3 introduces the notion of context shift. Afterwards the particles’ behavior will be analyzed in a number of constructions, using, when possible, data drawn from German corpora. In this section, I also want to look at those criteria that influence whether a particle is shifted in a particular environment or not. Finally, I will discuss how differences between the three MPs with respect to shiftability can be accounted for. 2. The Nature of Modal Particles German modal particles are an expressive device that the speaker employs to organize discourse. MPs can be used for common ground management,

96

sophia döring

i.e. they indicate whether a proposition is mutually known to the participants, uncertain or inactive. In other words, they convey the epistemic state of the discourse participants with respect to the descriptive content of the utterance. 2.1. General Remarks Modal particles are used a lot in German spoken communication and have been subject to considerable linguistic discussion (cf. Weydt 1977; Doherty 1985; Thurmair 1989; König 1991; Karagjosova 2004; Gutzmann 2008, 2009; Coniglio 2006, 2007, 2008, 2011; Kratzer 1999; Zimmermann 2004, 2011, etc.). Most of the expressions that serve as modal particles can also constitute different parts of speech, which is one of the factors that makes their analysis challenging. For instance, ja and doch can be used as response particles, too, eben and schon as temporal adverbs and auch as focus particle. When used as modal particles, the truth conditions of the sentence are not affected, but rather its felicity conditions, which is illustrated by the contrast between (1a) and (1b). The two sentences would be acceptable in different contexts, although they only differ in the choice of the modal particle: (1)

a. Peter ist doch diese Woche in Amsterdam. Peter is PART this week in Amsterdam ‘But Peter is in Amsterdam this week.’ b. Peter ist wohl diese Woche in Amsterdam. Peter is PART this week in Amsterdam ‘Peter is presumably in Amsterdam this week.’

By using doch in (1a), the speaker utters his belief that the proposition Peter is in Amsterdam is not activated in the addressee’s mind at the time of utterance, so he reminds him of it. Besides, doch always indicates that the speaker does not consider the proposition to be controversial. Wohl in (1b) on the other hand is used by the speaker if he is not sure whether it is true that Peter is in Amsterdam, independent of what the addressee knows. In a context in which the speaker is not convinced that Peter is in Amsterdam, an utterance of (1b) would be felicitous, (1a) would not be. Their truth conditions, however, are the same, and are identical to those of the respective sentence without a modal particle. German modal particles, then, constitute a closed class of items that speakers use to organize discourse and that do not affect truth conditions.

modal particles and context shift

97

This characterization may be uncontroversial, but beyond that there still is a number of questions that have not yet been answered. It is often assumed that the class of modal particles is homogenous in that the mechanism underlying different particles is the same. There is, however, some evidence that this is not true. Zimmermann (2011) indicates that German ja, doch and wohl combine with different semantic components and thus have different effects. I claim in Section 5 that a uniform analysis for items of this class is probably not available. Another aspect that is still under discussion is whether their semantic contribution has the status of a conventional implicature (as e.g. Potts (2005) proposes) or rather that of a presupposition (Schlenker 2007). There are arguments in favor of both positions, but this question has to be left open here. Finally, their context dependency has not been exhaustively studied and this is where a context shift analysis can be promising. German modal particles can be compared to indexicals in that they are dependent on the context for interpretation. The reference of indexicals— like the first person pronoun or temporal adverbs such as today—can only be fixed when considering the context of utterance. Modal particles, on the other hand, refer to the speaker’s belief or attitude with respect to the proposition expressed. From this reference to the speaker of the utterance, a context dependency follows automatically. If MPs depend on the context of utterance in the same way as indexicals do, the question arises whether a context shift can affect their interpretation as well, e.g. when used in reported speech, where one utterance situation is embedded in another one. For indexicals, it has been observed that—amongst others—attitude operators can shift their context of interpretation to an embedded context (cf. Kammerzell & Peust 2002, Schlenker 2003, Anand & Nevins 2004, Quer 2005, McCready 2007, Sharvit 2004, 2008, etc.). When considering certain instances of expressive devices, however, as for example epithets (like that idiot), different authors take them to be invariably speaker-oriented (cf. Corazza 2005 or Potts 2005), which suggests that the same holds for modal particles. Up to now this aspect of modal particles has not been investigated systematically. Accordingly, three frequently used German modal particles have been chosen and will be analyzed in five different constructions. As mentioned before I assume that modal particles probably work in different ways, i.e. they cannot be regarded as one homogeneous class. Therefore, looking at constructions like reported speech might be interesting not

98

sophia döring

only to find out whether the particles’ reference can be shifted at all, but also to compare different particles’ behavior with regard to shiftability. 2.2. Basic Meaning of Selected MPs The three particles chosen for the present study are ja, doch, and wohl. Pinning down the meaning of MPs is a well-known problem as it may vary according to the context they are used in. Due to this context dependency, it has often been discussed whether there is one underlying meaning for each particle at all or whether an analysis with several meaning variants is more appropriate. I believe that a basic meaning for each particle can indeed be found and other uses can be ascribed to it. In the following, however, there will only be a short sketch of the particles’ meaning that enables the reader to follow the context shift analysis. For grasping the MP’s meaning, the notion of common ground proves to be highly valuable. Common ground, as introduced by Stalnaker (1973, 1974, 2002) for the description of presuppositions, denotes the shared knowledge of discourse participants. It can be understood as a set of assumptions and beliefs that speaker and addressee share and that they recognize that they share. During discourse, every utterance updates this common knowledge of speaker and addressee. German MPs make reference to the common ground in different ways, and thus can be defined based on this notion. In order to lay the groundwork for an analysis of the sentences to be discussed in Section 4, I will now briefly sketch the meaning of ja, doch, and wohl, using the concept of common ground. 2.2.1. Ja The meaning of the particle ja is well understood (cf. Doherty 1985, König & Requardt 1991, Jacobs 1991, Lindner 1991, Kratzer 1999). As described above, one component of its meaning is its reference to the common ground: Zimmermann (2011: 2016) describes its semantic function as “establishing or reconfirming a proposition p as part of the common ground, often based on perceivable contextual evidence”. By this, ja expresses that the proposition p is uncontroversial at the time of utterance, i.e. that there is no proposition q in the common ground that contradicts p. At the same time, the speaker assumes that the addressee shares his knowledge that p holds (i.e. p is already part of the common ground) or is able to gain evidence for the fact that p is true, as in (3):

modal particles and context shift (2)

Max hat ja eine neue Freundin. Max has part a new girlfriend ‘Max has a new girlfriend.’

(3)

Deine Schuhe sind ja ganz dreckig! your shoes are part totally dirty ‘Your shoes are totally dirty!’

99

This postulation of the meaning and function of ja is supported by the fact that it is inadequate in answers to a question (cf. (4)), as well as in corrections or breaking news contexts in which p obviously is not part of the addressee’s knowledge. (4) A: Wann kommt denn der Bus? When arrives part the bus “When does the bus arrive?” B: #Er kommt ja um 12:40 Uhr. It arrives part at 12:40 o’clock ‘It arrives at 12:40.’

Kratzer (1999) proposes that the descriptive meaning of an utterance with ja (which in (2) would be Max hat eine neue Freundin, p minus the MP’s contribution) serves as input for the computation of the expressive meaning. Thus, ja is not part of the descriptive meaning of an utterance, it operates on a different level. 2.2.2. Doch Similar to ja, the particle doch also makes an assertion about the addressee’s knowledge (cf. Thurmair 1989, Karagjosova 2004, Egg 2013 [this volume], Repp 2013 [this volume], for remarks on doch). As stated above, it conveys the speaker’s belief that p is not activated in the addressee’s mind at the time of utterance—or at the current stage of discourse: […] the speakers using doch have some evidence from the behaviour of their addressee that the latter is not taking into account some particular state of affairs. (Lindner 1991: 183)

Thus, by uttering doch p, the speaker reminds the addressee, who may not be aware of p or may assume at the time of utterance that p is not true, which is illustrated by the examples below. As a consequence, doch indicates that p is in contrast, either with another proposition explicitly given in discourse (as in (6)) or with the speaker’s assumption about the addressee’s background knowledge (as in (5)).

100

sophia döring

(5)

Du gehst schon? Ich habe doch Tee gemacht! You leave already I have part tea made ‘You’re leaving? But I made tea!’

(6)

A: Maria hat einen VW. Maria has a VW ‘Maria owns a VW.’ B: Nein, Maria hat doch einen Porsche! No Maria has part a Porsche ‘No, Maria owns a Porsche.’

The two important components of the meaning of doch therefore are reference to common ground and contrast (cf. Karagjosova 2004). The first is shared with ja, while the second constitutes a distinction. 2.2.3. Wohl The particle wohl is discussed in detail by Zimmermann (2004), who analyzes it as a modifier of sentence type indicators. This point can be illustrated by the difference between wohl in declaratives and wohl in interrogatives (which will be addressed in Section 4.4). According to Zimmermann (2004), wohl weakens the speaker’s commitment to the proposition in declaratives, while in interrogatives, it leads to “a request for a weaker commitment concerning the questioned proposition” (Zimmermann 2004: 253).1 By using wohl, the speaker indicates that he does not want to completely commit himself to the truth of the proposition. Consequently, in contrast to ja and doch described before, a speaker uttering wohl p is not sure whether p is true: “the descriptive content of the clause is not presented as secure knowledge, but rather as an assumption or a conjecture” (Zimmermann 2011: 2018). An example for the use of wohl is given in (7): (7)

Du hast die Schlüssel wohl verloren. You have the keys part lost ‘You must have lost the keys.’

wohl is infelicitous in contexts where the speaker signals that he is strongly convinced of the truth of the proposition, as in (8): (8) #Ich bin mir ganz sicher: Du hast die Schlüssel wohl verloren. I am me absolutely sure: You have the keys part lost ‘I’m absolutely sure: You presumably lost the keys.’

1 Zimmermann (2004) accounts for these data by analyzing wohl as moving to Spec, ForceP on LF with the result that wohl can code the force of commitment to the proposition.

modal particles and context shift

101

Apart from the missing truth claim, wohl differs from ja and doch in that it does not make a statement about the assumed knowledge of the addressee. As a consequence, it can occur in questions2 (cf. (9)) and in cases where it is obvious that the addressee does not share the knowledge that p is true (cf. (10)): (9)

Ob Peter wohl Wein mitbringt? Whether Peter part wine brings ‘I wonder whether Peter will bring wine?’ lit: ‘Will Peter presumably bring wine?’

(10) A: Was bringt Peter mit? What brings Peter with ‘What does Peter bring?’ B: Peter bringt wohl Wein mit. Peter brings part wine with ‘Peter presumably brings wine.’

On the first glance, we see that ja and doch share their reference to the common ground. They indicate that the speaker assumes the proposition to be part of the common ground. With wohl, in contrast, the speaker expresses that the respective proposition shall not automatically be added to the common ground. (Note that there is also a use of wohl which is rather affirming, but this cannot be discussed here.) A statement about the speaker’s attitude or beliefs, however, is made by all of the three particles, which makes them dependent on the context of utterance. I will take a closer look at this context dependency in Section 4, which presents the results of the context shift analysis. The next section will introduce the phenomenon of context shift in more general terms. 3. Context Shift Kaplan’s fixity thesis has drawn attention to context shift, with its claim that the interpretation of context-dependent expressions is necessarily fixed to the context of utterance: [N]o operator can control the character of the indexicals within its scope, because they will simply leap out of its scope to the front of the operator. (Kaplan 1989: 510)

2

For a further discussion of wohl in questions, cf. Section 4.4.

102

sophia döring

Kaplan (1989) distinguishes between the content and the character of an expression, where the content of, for instance, a predicate with respect to a context is a property or relation. The content of sentences consists of structured propositions. The character of an expression or sentence is its literal meaning. With the character of an expression, as a function from contexts to contents, Kaplan accounts for the influence of context: It provides an expression’s or sentence’s content in different contexts, i.e. the content of a sentence may be true or false with respect to a specific context. Kaplan calls operators that can shift the character of a context-dependent expression ‘monster operators’ and claims that they do not exist in natural language. An indexical expression, such as the first-person pronoun in (11) below, therefore always has to refer to the actual speaker, a co-reference with the speaker of the reported situation is not possible: (11)

*Maryi said that Ii was an idiot.

This impossibility of context shifts proposed by Kaplan (1989) has in the meantime been proven wrong with regard to several types of expressions in different languages (cf. Anand & Nevins 2004; Kammerzell & Peust 2002; McCready 2007; Schlenker 2003; Sharvit 2004; Quer 2005). Most of these studies, however, concentrate on the shifting behavior of indexicals, showing that indexical pronouns as in (11) can in fact be interpreted with respect to the reported context in some languages, as for instance in the often quoted case of Amharic logophoric pronouns given in Schlenker 2003: 68. (12) Situation to be reported: John says: ‘I am a hero’. ǰon ǰəgna nə- n n yɨl-all john hero be.pf-1so 3m.say-aux.3m ‘Johni says that Ii am a hero.’

In Amharic, the first person pronoun can be coreferential with the speaker of the embedded speech act. (Note that this is not obligatory.) As modal particles display a clear context dependency as well, we want to examine whether they are able to shift in a similar way. The phenomenon of context shift, therefore, helps us gain a better understanding of the nature of context-dependent expressions. Formally, context shift can be captured by operators that manipulate the context. Different kinds of such operators for the various environments facilitating context shift have been proposed in previous studies (cf. Schlenker 2003; Anand & Nevins 2004; Sharvit 2004; 2008; McCready 2007).

modal particles and context shift

103

Context will be defined as a triple of parameters here, determining its relevant properties (most importantly speaker, interlocutor and judge): c = ⟨ac , ic , jc ⟩. This kind of triple is an extended version of what is proposed by Kaplan (1989), with a judge added that was not present in the original version (cf. McCready 2007). The judge is the one whose attitude is conveyed. Parameters for time and place of utterance can be added as well but as they are not relevant for context shift, they are left out now for simplicity. By default, the judge parameter is set to the actual speaker. For an account of attitude verbs, a monster operator was proposed (cf. Schlenker 2003). It manipulates the context parameter with the consequence that the reference of indexicals can no longer be interpreted within the actual context of utterance. We already know, however, that the perspective cannot only be shifted to an embedded speaker, but also to the addressee (cf. McCready 2007 for context shift in questions). I therefore assume two operators, both working on the same principle: They manipulate context triples by shifting the judge parameter. One operator accounts for a shift to the addressee’s point of view by setting the judge parameter to the interlocutor, which occurs in questions. This is illustrated below: (13) ⟦OPADR Φ⟧⟨ac,ic,jc⟩ = ⟦Φ⟧⟨ac,ic,ic⟩

In most of the other constructions, the reference might shift to what I call the protagonist3 of the utterance. Thus, we need an operator which does that. I assume this second operator shifts the judge parameter to a logophoric center (l) in the context: (14) ⟦OPLOG Φ⟧⟨ac,ic,jc⟩ = ⟦Φ⟧⟨ac,ic,⟦l⟧⟩

Both operators should be understood as free operators that arise whenever they are needed. Let’s say that the non-shifted interpretation is the default, there is no shift without a reason. The operators only appear when the straightforward interpretation for a sentence is not plausible. Establishing the exact conditions for their appearance, however, rather falls within the field of language usage.

3 I will use the term protagonist for the instances of context shift, where the perspective is not shifted to the addressee but to a person mentioned in the sentence, not necessarily the speaker of an embedded utterance.

104

sophia döring 4. Analysis

From recent work on context shift we know that there is a number of constructions that allow the interpretation of a context-dependent expression in a context other than that of the current utterance. Reported speech is the prototypical context to be considered, as well as free indirect discourse (cf. Sharvit 2004). As another perspective is reported, these constructions should constitute the perfect environment for shifts. Beside embeddings under verbs of saying, sentences with attitude verbs are of interest, too, as—in contrast to reported speech—no proper speech act is embedded here. In addition, these embeddings have not been analyzed in depth so far, so I will look at them in Section 4.2. Other constructions that involve reference to someone else’s utterance— although more liberally—are evidential constructions, in which the speaker does not report a speech act directly, but gives the source of the information. Finally, questions may license context shifts due to their focus on the addressee (cf. McCready 2007). So these are the five types of constructions chosen for the present study. The aim of the following analysis is to find out whose attitude the modal particles refer to in constructions where not only the current speaker, but another possible referent is available. Finally a note on terminology: Following the theory of Kaplan (1989) sketched in 3, the term judge will be used to refer to the person whose attitude is expressed. 4.1. Reported Speech Peust (2005: 97) describes reported speech as an ‘island’4 for a different deictic perspective, which immediately indicates its relevance with respect to context shift. While in some languages, as in Amharic above, the interpretation of indexical pronouns and temporal adverbs can be shifted to the reported context, this is not possible in German: These context-dependent expressions are necessarily fixed in the context of utterance (cf. (11) above). Modal particles, however, can be shifted in German, as the following corpus sentence illustrates:5

4 5

Note that the term island is not used in the common syntactic sense here. I used the DWDS corpus for my analysis.

modal particles and context shift

105

(15) […] Ich hörte Marcel Reif, der das Spiel kommentierte, noch sagen, I heard Marcel Reif who the game commented just say dass die Bayern es wohl geschafft hätten, in dem Moment schieben that the Bavarians it part made had in this moment move die Engländer den Ball rein. the English the ball in ‘I heard how Marcel Reif, who commented on the match, said that Bayern presumably made it—right at that moment England scored a goal.’6

The current speaker of this utterance, Helmut Kohl, uses the particle wohl, which expresses an uncertainty of the speaker. It is obvious in (15) above that this uncertainty is not assigned to Kohl, but to the speaker of the reported situation, Marcel Reif. We say that the context of interpretation of wohl is shifted to the reported situation. Wohl, therefore, allows for a shift of the perspective here. As a side note, it should be mentioned here that sentence mood plays a role. Analyzing the three MPs in reported speech, as well in embeddings under attitude verbs, we find that the use of subjunctive frequently facilitates a context shift reading. Still, subjunctive mood is not decisive for such an interpretation in most of the cases. (15), for instance, would not be as natural without the subjunctive, but the sentence would not be ruled out, and the shifted reading for wohl would not be either. For a clearer picture, (16) and (17) now illustrate reported speech contexts containing doch and ja: (16) Der kleine Saal platzt aus allen Nähten; […] und noch mitten in The small hall bursts out all seams and yet in.the.middle in Baudrillards Vortrag ruft jemand verzweifelt, dass doch in Raum Baudrillard’s lecture shouts somebody desperately that part in room 104 mehr Platz sei. 104 more space was ‘The small hall bursts at the seams; […] and in the middle of Baudrillard’s lecture somebody shouted that there was more free space in room 104.’7 (17) AM Fischer warf hier ein, dass Botschafter Huber ja die AM Fischer added here in that ambassador Huber part the Botschaftstore nicht allzuweit aufmachen musste. embassy.doors not too.wide open must 6 7

DWDS Corpus: “NEAR(sagen,wohl,5)”, #1302. Berliner Zeitung, 28.02.2002, Christian Esch: ‘Ich selbstmorde die Begriffe’.

106

sophia döring ‘Foreign minister Fischer added here that ambassador Huber would not have to open the doors of the embassy too wide.’8

I have analyzed doch as reminding the addressee of p; in (16) we intuitively understand that not the actual speaker, but the person in the audience reminds the other people. The actual speaker just picks the particle up and thereby reports his attitude. The same seems to hold for (17): We can readily think of a situation in which the actual speaker has no opinion and attitude towards opening the gates of the embassy at all—still an utterance of (17) is fine as the influence of ja is attributed to the foreign minister. Again it is more plausible to ascribe the evoked attitude to the speaker of the reported situation, rather than to the actual speaker. These conclusions, however, rely on intuitions and estimations of plausibility. To make judgements more reproducible, constructed contexts as in (18) are necessary. While in (18a) the actual speaker expresses his strong belief in the truth of the proposition (ich weiß genau), in (18b) the embedded speaker Paul does so (versichern). As stated before, wohl expresses a lowered estimation of trustworthiness in the proposition, which is incompatible with expressing a strong conviction of the proposition’s truth. The fact that (18a) works fine, but (18b) does not, proves that the judge of wohl is the speaker of the embedded utterance, i.e. Paul: (18) a. Ich weiß genau, dass Maria als Prinzessin zum Fasching geht I know exactly that Maria as princess for costume.party goes aber Paul erzählte seiner Tante vorhin, dass sie wohl als Flamingo but Paul told his aunt earlier that she part as flamingo gehe. went ‘I’m sure that Maria will be disguised as a princess for the costume party, but Paul told his aunt that she presumably would be disguised as a flamingo.’ b. #Paul hatte seinen Freunden seit Wochen versichert, dass Maria wohl Paul had his friends for weeks reassured that Maria part als Prinzessin zum Fasching gehen würde. as princess for costume.party go would ‘Paul reassured his friends for weeks that Maria presumably would be disguised as a princess for the costume party.’ 8

DWDS Corpus: ‘NEAR(daß,ja,5)’, #3458.

modal particles and context shift

107

What these two versions demonstrate is that in some cases, the reference of wohl even has to be interpreted as shifted.9 At first glance, shifted readings for modal particles in reported speech seem to be unproblematic: All of the analyzed sentences allow for such an interpretation readily, although this is not obligatory. On closer inspection, however, the embedding verb appears to play a role, cf. the two versions of (19): (19) a. Er flüsterte ihr zu, dass er ja Pirat werden wolle. He whispered her to that he part pirate become wanted ‘He whispered to her that he wanted to become a pirate.’ b. ?Er erzählte ihr, dass er ja Pirat werden wolle. He told her that he part pirate become wanted ‘He told her that he wanted to become a pirate.’

There is no difference in grammaticality, but (19a) sounds more natural than (19b). Such a variation is probably due to the difference between flüstern, as a verb describing the manner of speaking, and erzählen, which makes no statement about manner. As a first guess, it seems to be plausible that manner of speaking verbs far more strongly commit a speaker to report an utterance word by word than do verbs of saying such as erzählen, which 9 It is difficult to construe a sentence pair like this for ja and doch: As an addressee has to be involved (an actual as well as an embedded one), made-up discourse often sounds unnatural and nested. The following example, however, shows that the use of ja in the embedding is strange if the embedded speaker does not know that p is true. In (ii), it is the actual addressee who indicates by his question that he does not know that p is true. The sentence is acceptable nevertheless, as the judge here is Paul, the protagonist.

(i)

A: Wissen eigentlich die anderen, dass Maria im Urlaub ist? Know actually the others that Mary in.the vacation is ‘Do the others know that Mary is on vacation?’ B: Paul auf jeden Fall. #Als Peter vorhin fragte, wo eigentlich Maria bleibt, Paul in any case when Peter earlier asked where actually Maria is antwortete Paul, dass sie ja im Urlaub sei und nicht komme. answered Paul that she part in.the vacation was and not came ‘Paul definitely knows. When Peter asked where Mary is, Paul answered that she was on vacation and did not come.’

(ii)

A: Wo bleibt eigentlich Maria? ‘Where is Mary?’ B: Ich habe gehört, wie Paul vorhin zu Peter sagte, dass sie ja im Urlaub sei und nicht komme. ‘I heard Paul telling Peter that she was on holiday and did not come.’

108

sophia döring

rather have to do with reporting the facts. But we will have a closer look at these predicates later on in Section 5. What is striking, apart from that, is that in the set of sentences in reported speech, those containing ja are not as easily embeddable as those containing wohl: the respective sentences often appear to be unnatural. Whether this has to do with some kind of difficulty in shifting the reference of ja will be discussed later on. 4.2. Embedding under Attitude Verbs It is interesting to look at attitude verbs because what we find here is not really an instance of reported speech: no implicit speech act is involved. Does an embedding predicate of this type nevertheless offer something like an island for a different perspective? In the literature on modal particles, it is often claimed that they cannot be embedded under attitude verbs, as for instance in Thurmair 1989 or Zimmermann 2011: […] ja is generally impossible in complement clauses […], except under verba dicendi (often with subjunctive mood), even though there is no relation involved and even though a plausible interpretation is available in principle. […] ja is always evaluated with respect to the utterance context. Hence, it cannot be embedded, unless it forms part of a reported speech act under a verb of saying […]. (Zimmermann 2011: 2023)

On first glance, this seems to be true: In (20), the insertion of ja in a sentence embedded under different types of German attitude verbs leads to unacceptability in most cases: (20) Max weiß / #glaubt / #vermutet, dass Paul ja niemandem Max knows believes assumes that Paul part nobody etwas sagen würde. something say would ‘Max knows/believes/assumes that Paul would not tell anybody anything.’

Ja is incompatible with glauben and vermuten. An embedding under wissen is completely acceptable: wissen is a factive verb and thus compatible with ja, which implies the truth of the proposition in question. On closer inspection, we indeed find a number of completely acceptable instances of a modal particle embedded by an attitude verb. Examples are given in (21) to (23): (21) Lisa sah ihn verwundert an. Dann wurde ihr bewusst, dass sie Lisa looked him puzzled at then became her aware that she

modal particles and context shift

109

ja in einem Firmenjet saß. part in a company-plane sat ‘Lisa looked at him puzzled. Then she realized that she was sitting in a company plane.’10 (22) Max schüttelte den Kopf und dachte bei sich, dass es doch nun Max shook the head and thought at himself that it part now wirklich nicht das erste Treffen war, das Paul absagte. really not the first meeting was that Paul cancelled ‘Max shook his head and thought to himself that it indeed was not the first meeting Paul cancelled.’11 (23) A hat sich in der Dienststelle einen neuen Aufkleber auf das Auto A had himself in the office a new sticker on the car geklebt, weil er dachte, dass er wohl irgendwie abgegangen ist oder stuck because he thought that it part somehow come.off is or sich jemand einen Scherz erlaubt hat. himself somebody a joke allowed had ‘A stuck a new sticker, which he got at work, on his car as he thought that it presumably somehow came off or he fell for a trick.’12

These sentences not only provide evidence for the claim that modal particles can be embedded by attitude verbs, but also show that the context of interpretation in these sentences may be shifted: It is natural to ascribe the expressed attitude in (21) to the protagonist of the situation, i.e. Lisa would be the one who thinks that she should have already known that she is sitting in a company plane (in this case no addressee is involved, but Lisa herself is the one who already knows p). This reading, however, is not available in the case of (20) as no common ground or addressee can be inferred from the context, and we do not interpret it as Max addressing himself. We face two problems here: Firstly, ja and doch both make reference to the common ground of speaker and addressee. In embeddings under attitude verbs, we find (by definition) no addressee. Thus, we have to interpret the speaker’s attitude as directed towards himself and his own knowledge in order to get a shifted reading. This interpretation is, however, not available in all contexts. 10

DWDS Corpus: “NEAR(daß,ja,5)”, #3528. DWDS Corpus: “NEAR(daß,ja,5)”, #3528. 12 http://www.talkteria.de/forum/topic-146892.html: “Streit mit Zeugen Jehova wegen Autoaufkleber”. 11

110

sophia döring

Secondly, ja and doch both claim that the expressed proposition is true. As a result, they cannot occur in the scope of non-factive predicates such as to believe as these do not presuppose the truth of their sentential complements. Thus, instances of MPs embedded under attitude verbs do allow the judge to be someone else other than the actual speaker. Not all attitude verbs, however, can embed all MPs in a meaningful way. 4.3. Free Indirect Discourse After this side note on attitude verbs, we turn to another type of reported speech. Free indirect discourse is used almost exclusively in narrating literature and can be described as an implicit way of reporting speech that does not make use of verbs of saying (cf. Banfield 1982; Sharvit 2004, 2008, etc.). As there is no formal device that shows that someone else’s utterance or thought is reported, the addressee has to infer from the context whose attitude is expressed. Free indirect discourse, then, cannot be distinguished from its context by grammatical means, but only by semantic features (cf. Vuillaume 2002). It has been observed that temporal adverbs often can be shifted in free indirect discourse, as (24) illustrates, while personal pronouns again do not allow for shifts in most languages, cf. (25): (24) George sighed. Tomorrow he would buy a wedding ring for Pattie. (25) *Georgei sighed. Ii would buy a wedding ring for Pattie.

Tomorrow here does not refer to the day following the day of the utterance, it has to be interpreted within the reported context. When analyzing modal particles in free indirect discourse, we expect them to be able to shift, as even reported speech allows them to do, which in a way is a ‘more restricted’ environment (temporal adverbs, for instance, cannot shift in reported speech), cf. also Eckardt (2012), who analyzes particles in free indirect discourse as shifting indexicals. The examples in (26) and (27) below prove that the MPs here are interpreted from the protagonist’s point of view. More than that: We can hardly even think of a non-shifted interpretation, such that the speaker wants to remind the addressee that Paul spent the whole evening in the kitchen cooking. Free indirect discourse is clearly associated with prose texts which not only facilitates, but even forces the shifted reading.

modal particles and context shift

111

(26) Die Enttäuschung war Paul ins Gesicht geschrieben. Er hatte doch The disappointment was Paul in.the face written He had part den ganzen Abend für diesen Eintopf in der Küche gestanden. the whole evening for this stew in the kitchen stood ‘The disappointment was written all over Paul’s face. He had stood in the kitchen all evening for this stew.’ (27) Die Aufgabe machte Maria keine Angst. Sie hatte ja ein Ass im The task made Maria no fear she had part an ace in.the Ärmel. sleeve ‘The task did not frighten Maria. She had an ace up her sleeve.’

When analyzing embedded sentences, we can also discuss subordinate clauses (for instance relative clauses or concessive clauses) including particles, as they also seem to belong in this section. To start with, (28) shows that modal particles can occur in relative clauses as well, but only in appositive ones (cf. Thurmair 1989: 79).13 The uncertainty of wohl, again, is rather ascribed to the protagonist (er ‘he’) of the situation than to the actual speaker, although this reading is not impossible, which becomes evident in the paraphrases in (28ʹ): (28) Auf der anderen Seite sah er […] einen Mann, der wohl der Vater On the other side saw he a man who part the father dieser Kinder war. of.these children was ‘On the other side he saw a man who presumably was the father of these children.’14 (28ʹ) Auf der anderen Seite sah er einen Mann und er vermutete / ich On the other side saw he a man and he assumed I vermute, dass es der Vater der Kinder war. assume that it the father of.the children was ‘On the other side he saw a man and he assumed/I assume that he was the father of these children.’

13 This is due to the fact that modal particles ask for a completely specified proposition. Restrictive relative clauses form open propositions and therefore cannot be combined with them. 14 http://www.fantasyguide.de: Andreas Fischer: “Der Brückentroll”.

112

sophia döring

Therefore, (28) is ambiguous with respect to whose attitude is conveyed by wohl. In (29) and (30), a causal and a concessive subordinate clause are given, as another case of a syntactic embedding of particles and of free indirect discourse at the same time: (29) Kurt war empört, weil er ja auch etwas getan hatte. Kurt was indignant because he part too something done had ‘Kurt was indignant since he had done something as well.’ (30) Kurt war verärgert, obwohl er ja selbst daran schuld war. Kurt was annoyed although he part himself at.it guilty was ‘Kurt was annoyed although he himself was to blame for it.’

Again, both readings are accessible in these two sentences. The context may be shifted to the view of the protagonist. Imagine a context where Kurt’s mother complimented his two sisters for cleaning the kitchen, although he helped as well and thinks she should know that. The speaker himself, however, may be the judge, too. Constructing a suitable context permits both possibilities. Nevertheless, unambiguous contexts can be construed as in (31) where the causal relation is one the protagonist himself cannot express, consequently there is no context shift. (31) Kurt war niedergeschlagen, weil er ja, (ohne es zu wissen), Kurt was sad because he part (without it to know) unter einer klinischen Depression litt. under a clinical depression suffered ‘Kurt was sad because he was depressed (without knowing it).’

4.4. Questions As questions refer to the epistemic knowledge of the addressee (or sometimes the knowledge of the speaker and the addressee), it is probable that they license perspective shifts. McCready (2007) showed that Japanese experiencer predicates, as well as predicates of personal taste in English, can be subject to context shift. While these expressions refer to the taste or feelings of the speaker in assertions, they describe them from the perspective of the addressee in questions, as is illustrated by (32): (32) a. Smurf ice cream is tasty. b. Is smurf ice cream tasty?

modal particles and context shift

113

While a speaker utters his own taste in (32a), he wants to know something about the addressee’s taste in in (32b). Analyzing the behavior of German modal particles in this sentence type is problematic as only one of the three particles is licensed in non-rhetorical questions (cf. Egg 2013 [this volume] on doch in rhetorical questions). Ja, as well as doch, state the truth of the proposition which makes them inappropriate for questions. Thus, these particles’ meaning is the reason why they are infelicitous for the sentence type meaning of questions (cf. Lindner 1991; Doherty 1985; Truckenbrodt 2004), cf. (33): (33) #Hat Moritz ja/doch den Kuchen allein verziert. Has Moritz part the cake on.his.own decorated ‘Did Moritz decorate the cake on his own?’

If one nevertheless finds one of the above particles in a question, it is its stressed variant, as in the case of DOCH 15 below, or it is the speech act of a question expressed by a rising declarative as in (35): (34) Gibt es doch eine “Zweite Erde”? Gives it part a second earth ‘Is there a “Second Earth”?’16 (35) Du hast doch den Herd ausgemacht? You have part the stove turned.off ‘You turned off the stove, right?’

I will come back to cases like these after the analysis of wohl. As the following examples illustrate, wohl can appear in questions easily, maintaining its usual effect, namely an uncertainty towards the truth of the proposition: (36) Wird Peter wohl Anna mitbringen? Will Peter part Anna bring ‘(I wonder) Will Peter bring Anna?’ (37) Ob er wohl schon lange auf der Straße lebt? Whether he part already long on the street lives ‘I wonder whether he lives on the streets for a long time already.’ (lit.: ‘Does he live on the streets for a long time already?’) 15 The contrasted DOCH—in contrast to doch—does not express that the speaker assumes p to be not activated in the addressee’s mind (although it is supposed to be common ground); instead the speaker appears to believe that not-p is activated in the addressee’s set of beliefs. This is reversed in questions: The speaker signals that he himself considered not-p to be the case. This explains why the meaning of doch cannot be reversed: If p were not active in the mind of the speaker, he would not be able to ask the corresponding question. 16 http://www.heise.de: Petra Vitolini Naldini: Gibt es doch eine “Zweite Erde”? December 29th 2007. The capitals to indicate stress are mine.

114

sophia döring

Interestingly, the speaker in these questions does not seem to expect an answer from the addressee, but rather something like a guess. Note, however, that questions in the form of ob-sentences display this behavior anyway (and wohl has a preference to appear in them), but the same holds for (34). This is also noted by Zimmermann (2004) who describes the interpretation of wohl as sensitive to different sentence types (cf. Zimmermann 2004: 258ff.). We can say that the attitude expressed by wohl is ascribed to the addressee of the question, which then would be a case of context shift.17 Of course the speaker is not sure about the truth of the proposition either, but this apparently is due to the sentence type, not to the use of the modal particle. This picture becomes even clearer when looking at specific instances and uses of questions. In a question that a teacher asks a student, wohl is fine, although we know that the teacher most probably knows the answer. The judge is the student. If a question is one that should be easy to answer for the addressee, as in (39), on the other hand, wohl is inappropriate as the speaker indicates that he does not expect the addressee to know the answer. (38) Lehrer: Was ist wohl die Hauptstadt von Argentinien? Teacher: What is part the capital of Argentina ‘Teacher: (I wonder) What is the capital of Argentina?’ (39) #Bist du wohl verheiratet? Are you part married ‘Are you presumably married?’

Finally, wohl is fine in rhetorical questions (cf. Egg 2013 [this volume]) which further supports the claim that the particle’s effect is interpreted from the addressee’s perspective here because the speaker in rhetorical questions by definition does not expect an answer. The appearance of wohl in questions, thus, might be analyzed as another instance of context shift. Let’s come back to the appearance of doch in rising declaratives, as syntactically they are not interrogatives but they constitute the speech act of a question. The following sentence from the corpus gives a further example:

17 This is not the only possible analysis. Following Zimmermann (2004), we could argue that wohl does not involve reference to the speaker but rather to the epistemic center encoded in the sentence type. We do not have to assume a shift. In this paper, I follow McCready (2007), who describes the same phenomenon for English predicates of personal taste and Japanese experiencer predicates and categorizes it as an instance of perspective shift.

modal particles and context shift

115

(40) “Aber ihr kommt doch beide?”, fragte Heinrich, sich zu den But you come part both asked Heinrich himself to the Schwestern wendend. sisters turning ‘ “But you will come, won’t you?”, Heinrich asked, turning to his sisters.’18

The use of rising declaratives is acceptable only if the speaker can be sure that the addressee knows the answer (cf. Gunlogson 2001: 10). At the same time, there is always a preference for one answer: rising declaratives are used by the speaker to make sure the addressee shares his knowledge. In assertions, a speaker using doch assumes that the proposition is not active in the addressee’s mind. In rising declaratives as in (40) on the other hand, the speaker wants to find out whether p is not activated in the addressee’s mind: The speaker wants the addressee to tell him if the addressee is not considering coming tonight. (Lindner 1991: 189)

Thus, the meaning of doch in questions slightly differs from that in assertions, but still the attitude expressed by the particle is ascribed to the speaker and the context of interpretation therefore is not shifted. 4.5. Evidential Constructions Finally, I want to analyze the appearance of modal particles in evidential constructions. What is relevant here, is that evidential constructions mention the source of the information without using verbs of saying—as opposed to the instances of reported speech presented in § 4.1. An example for such a construction is given in (41), the passage in italics introduces the source of the given information: (41) According to the Rolling Stone Magazine, Steven Tyler has wanted to leave Aerosmith for two years.

Do evidential constructions allow for a context shift of expressive devices? On the one hand, the source of information is given and so a shift to the source seems possible but on the other hand, evidential constructions do not equal reported speech as the actual speaker most probably uses his own words: it is not his aim to report someone else’s speech, but rather to verify his utterance.

18

http://www.zeno.org/Literatur: Malwida von Meysenbug: Unerfüllt.

116

sophia döring

Considering this, we would expect that the whole utterance, including the attitude expressed by modal particles, rather is attributed to the actual speaker. First consider examples containing doch and ja: (42) Laut Medien ist diese Frage doch längst geklärt. According.to media is this question part long.ago answered ‘According to the media this question has been answered long ago.’ (43) Berichten zufolge soll ja Peking während der Reports According.to shall part Peking during the Olympiabewerbung braune Grasflächen grün angestrichen haben. Olympia.application brown grassland green painted have ‘According to reports Peking has painted brown grassland green during the application for the Olympic games.’19

In these sentences, a context shift reading is not easily accessible. Actually, we consider it possible to find sentences containing MPs in reports or news articles, but (43) could hardly be uttered as an answer to a question like “I have not heard anything about Peking’s application for the Olympic games! Was it according to the rules?”, in which the speaker signals that he does not know anything about the application. If the attitude conveyed by ja was ascribed to the reports (they might assume that their readers already know this fact), (43) should be an appropriate answer, but in fact it is not. A reading where the reports include the particle ja and its claim that p is uncontroversial is not ruled out but less probable and hard to get. At first glance, evidential constructions seem not to allow for context shifts which would not be surprising as they do not report someone else’s utterance and thus his attitude and feelings. Sentences with wohl, however, do not behave like the ones with ja and doch, they allow for a shifted reading: (44) Auch wenn aus dem schönen Wetter laut Wetterbericht wohl Also if off the nice weather According.to weather.report part vorläufig nichts wird. temporarily nothing becomes ‘Even though according to the weather report there won’t be pleasant weather for the time being.’20

For (44), it is perfectly natural to assume that wohl is originally used by the weather report in its forecast to weaken the claim ‘There won’t be pleasant

19 20

http://blogger.chinaseite.de: Entry by Kim Schiefer, March 07th 2007. http://www.3land.info: 3land-News, 10.02.2007.

modal particles and context shift

117

weather for the time being’. Such a reading is available easily because of world knowledge: Weather reports do not offer guarantees. The judge, then, is not the actual speaker. Accordingly, wohl in evidential constructions at least appears to allow for a context shift much more easily than ja and doch do. Still, the shiftability of wohl is not independent of contextual criteria, such as the reliability and authority of the source. That explains why in (45) the judge parameter is set to the speaker: We would not expect that a text of law includes this uncertainty. (45) Laut Gesetz ist das Parken von Wohnwagen in dieser Straße According.to law is the parking of trailers in this street wohl verboten. part forbidden ‘According to the law, parking of trailers is not permitted in this street.’

4.6. Intermediate Summary So far we have seen that the context of modal particles can be shifted quite systematically: in reported speech, embeddings under attitude verbs, free indirect discourse, questions and in evidential constructions. Although a number of shifted indexicals have been found, Kaplan’s fixity thesis seems to hold for the majority of indexical expressions. Under this assumption, our observations rather suggest that German MPs differ from indexicals as they can be shifted in a number of constructions. These five constructions, reported speech, free indirect discourse, embeddings under attitude verbs, questions and evidential sentences, all have in common that—in German or English—they do not license shifts for indexicals as easily. Modal particles and indexicals should thus not be simply lumped together. What is even more interesting is the fact that there are differences in the shiftability of the three chosen particles. Reported speech and free indirect discourse license context shifts for all of the three particles but sentences with ja and doch often are marked compared to those with wohl. The same is true for embeddings under attitude verbs. This difference is most obvious in the case of evidential constructions. It is, however, not a matter of acceptability or even grammaticality, but rather a subtle difference in the naturalness of sentences, which makes the analysis difficult. In the following, possible reasons for these differences will be discussed.

118

sophia döring 5. Different Classes of Modal Particles?

As mentioned before, wohl allows for a shifted reading in all analyzed contexts and under all embedding predicates without any problems. Ja and doch on the other hand do not: a shifted reading is hardly accessible in evidential constructions and they are sensitive to differences between verbs of saying. How can we account for this observation? There are various starting points. I want to return to the different types of verbs of saying first, which were addressed in Section 4.1. To understand the difference between these verbs may also offer an explanation to why an MP is sensitive to them or not. A classification of verbs is proposed by Bolinger (1968), as well as by Dor (2005), who describe differences between verbs of saying with regard to the phenomenon of that-deletion. It was often claimed that verbs of saying can embed their propositional complements with the insertion of a complementizer that— or just as well without one. As Bolinger (1968) and others point out, this is not true for all predicates, as some of them do require the complementizer. Dor (2005) suggests that what plays a role here is a semantic property of predicates he calls ‘truth claim’: that-deletion is possible under predicates which semantically entail that a cognitive agent (most often their subject) has made an epistemic claim concerning the truth of the proposition denoted by the embedded clause (Dor 2005: 345)

Using this criterion, Dor (2005) divides predicates into three classes and states that his type 3 verbs, which do not claim the truth of the proposition, do not license that-deletion. Type 1 predicates, on the other hand, like to tell, to state or to admit, denote illocutionary acts of the asserting type (cf. Dor 2005: 353): They state that the proposition is true and allow that-deletion. Type 2 predicates are virtually in between. Returning to modal particles, we noticed before that ja and doch involve the claim that the expressed proposition is true. At the same time, ja and doch—when compared to wohl—are problematic with respect to context shift. That is why it seems reasonable to analyze whether there is a correspondence between these three types of predicates established by Dor (2005) and their licensing of modal particles. If there was a correspondence, embeddings under type 3 predicates should also fail to allow context shifts for ja and doch: The reason is that type 3 predicates do not involve a truth claim, but ja and doch imply the truth of the proposition they appear with.

modal particles and context shift

119

This claim, however, is proven wrong, as (46) shows, where a type 1 predicate is given in (46a), type 2 in (46b) and type 3 in (46c): (46) a. Peter sagte, dass Maria ja/wohl kündigen wolle. b. Peter flüsterte, dass Maria ja/wohl kündigen wolle. c. Peter las aus dem Brief vor, dass Maria ja/wohl kündigen Peter read from the letter V-part that Maria part resign wolle (und deswegen nicht mehr an den Treffen teilnehmen wanted and thus not any.longer to the meetings join werde). would ‘Peter said/whispered/read aloud from the letter that Maria wanted to quit the job (and therefore no longer wanted to participate in the meetings).’

The verb vorlesen in (46c) as a type 3 predicate does not involve the speaker’s claim that the proposition is true—still the particle ja is just as acceptable as wohl is. So obviously the truth claim of the predicate does not play a decisive role. Instead, we notice something else when testing the different types of verbs: It is not felicitous for all particles to be embedded under emotive predicates. These verbs belong to Dor’s Type 2 that do not entail a truth claim, ‘but allow their meanings to be pragmatically extended to approximate that of a truth claim predicate’ (Dor 2005: 348): (47) Peter war zuversichtlich, dass er es #ja/#doch/wohl schaffen würde. Peter was confident that he it part make would ‘Peter was confident that he would make it.’

Once again, wohl is fine, while ja and doch are not as acceptable here.21 Obviously there is no correspondence between the acceptability of particles in embedded sentences and the predicates that allow for that-deletion. Instead, we found another case where ja and doch differ from wohl: embeddings under emotive predicates. This leads us to another possible difference between the particles. Apparently, it is not a truth claim that is crucial, but instead the differences can be explained when considering whether a speech act is involved. The meaning of ja and doch makes a claim about the addressee’s knowledge and the common ground, in ja for instance that p should already be

21 There seem to be different judgements on that. The construction with wohl, however, is more transparent. (Note, furthermore, that the stressed variant of doch would work fine.)

120

sophia döring

part of the shared background assumptions of speaker and addressee. We could argue therefore that they depend on a construction involving (at least implicitly) an underlying speech act (and thus including an addressee). Let’s look at a version of the weather report example, contrasting wohl and ja. As seen before, wohl can be shifted easily, while ja in the same sentence is problematic: (48) Laut Wetterbericht wird es wohl/ja morgen regnen. According.to weather.report will it part tomorrow rain ‘According to the weather report it will rain tomorrow.’

The source of the uncertainty conveyed by wohl can be the speaker or the weather report. In the case of ja, however, the perspective of the speaker cannot be shifted as easily, which may be due to the fact that this particle has this addressee-related component, as described above. In contrast to indirect speech, evidential constructions are not supposed to report the underlying speech act. Accordingly, that is why in (48) ja with its addressee-related meaning can only be interpreted from the speaker’s point of view, i.e. in the frame context that involves an addressee. Since evidential constructions do not imply a speech act and emotive predicates do not either, they do not license context shift for ja and doch. Wohl, on the other hand, only makes an assertion about the speaker’s knowledge and therefore can be shifted in both kinds of construction.22 6. Conclusion In this paper I have analyzed three German modal particles with respect to the phenomenon of context shift. While this issue has been much discussed for indexicals recently, it was not clear whether expressive devices such as modal particles act in the same way. I claim that they depend on the context just as indexicals do and used context shift to gain a better understanding of this dependency. The analysis of the three particles ja, doch, and wohl revealed that they can be shifted in a number of environments, but not all of them allow for shifts in the same way. I believe that these differences are due to differences in the semantics of the particles. Ja and doch refer to the common ground, thus making an assertion about the addressee’s knowledge. Because of that 22 Note that Zimmermann (2011) also assumes that the embeddability of wohl is due to its meaning. A more detailed account for this, as well as for the other MPs is still to be worked out, but these first observations point the direction.

modal particles and context shift

121

they ask for constructions involving a speech act, at least one that is implicit. Wohl, on the other hand, only makes an assertion about the speaker’s knowledge and is more flexible with respect to context shifts in different constructions. This may be only part of the explanation for differences in the behavior of modal particles. For the analysis of the phenomenon of context shift, however, the reference to common ground seems to be crucial. A discussion of possible implications of this context shift data for the right analysis of MPs may be subject to future research. It appears to be worthwhile, furthermore, to look for differences between the particles in other domains as well. They obviously do not form a homogenous class and it remains challenging to systematically find and describe the criteria responsible for these differences. Acknowledgements This paper is based on work done for my B.A. thesis from 2008. Thanks to Manfred Krifka, Markus Egg, George Smith as well as two anonymous reviewers for their helpful remarks. All remaining mistakes are mine. References Anand, Pranav & Andrew Nevins (2004): “Shifty operators in changing contexts”. Proceedings of SALT 14, 20–37. URL: http://elanguage.net/journals/index.php/ salt/article/view/14.20/1732. Banfield, Ann (1982): Unspeakable sentences. Narration and representation in the language of fiction. Boston: Routledge & Kegan Paul. Bolinger, Dwight (1968): “Postposed main phrases: An English rule for the Romance subjunctive”. Canadian Journal of Linguistics 14.1, 3–30. Coniglio, Marco (2006): “German modal particles in the functional structure of IP”. University of Venice Working Papers in Linguistics 16, 57–95. URL: http://lear.unive .it/bitstream/10278/204/1/2006-2s-Coniglio.pdf. ——— (2007): “German modal particles in root and embedded clauses”. University of Venice Working Papers in Linguistics 17, 109–141. URL: http://lear.unive.it/ bitstream/10278/142/1/Coniglio-WP-2007-pp109-141.pdf. ——— (2008): Die Syntax der deutschen Modalpartikeln. Ihre Distribution und Lizensierung in Haupt-und Nebensätzen. PhD thesis. Venice: Università Ca’ Foscari. ——— (2011): Die Syntax der deutschen Modalpartikeln. Ihre Distribution und Lizenzierung in Haupt-und Nebensätzen. Berlin: Akademie Verlag. Corazza, Eros (2005): “On epithets qua attributive anaphors”. Journal of Linguistics 41, 1–32. DOI: dx.doi.org/10.1017/S0022226704003044. Doherty, Monika (1985): Epistemische Bedeutung. Berlin: Akademie-Verlag.

122

sophia döring

Dor, Daniel (2005): “Toward a semantic account of that-deletion in English”. Linguistics 43.2, 345–382. DOI: dx.doi.org/10.1515/ling.2005.43.2.345. Eckardt, Regine (2012): “Particles as speaker indexicals in free indirect discourse”. Sprache und Datenverarbeitung 35–36. Special issue on formal approaches to discourse particles and modal adverbs. Ed. by Hogeweg, Lotte & Eric McCready & Grégoire Winterstein, 99–109. Egg, Markus (2013): “Discourse particles, common ground, and felicity conditions”. In: Gutzmann, Daniel & Hans-Martin Gärtner, eds.: Beyond expressives. Explorations in use-conditional meaning. Leiden: Brill, 125–149. [This volume]. Gunlogson, Christine (2001): True to form: Rising and falling declaratives as questions in English. PhD thesis. Santa Cruz: University of California. Gutzmann, Daniel (2008): On the interaction between modal particles and sentence mood in German. MA thesis. Mainz: Johannes Gutenberg-Universität. ——— (2009): “Hybrid semantics for modal particles”. Sprache und Datenverarbeitung 33.2, 45–59. Jacobs, Joachim (1991): “On the semantics of modal particles”. In: Abraham, Werner, ed.: Discourse particles. Amsterdam: John Benjamins, 141–162. Kammerzell, Frank & Carsten Peust (2002): “Reported speech in Egyptian: Forms, types, and history”. In: Güldemann, Tom & Manfred von Roncador, eds.: Reported speech as a meeting ground for different linguistic domains. Amsterdam: John Benjamins, 289–322. Kaplan, David (1989): “Demonstratives. An essay on the semantics, logic, metaphysics, and epistemology of demonstratives and other indexicals”. In: Almog, Joseph & John Perry & Howard Wettstein, eds.: Themes from Kaplan. Oxford: Oxford University Press, 481–563. Karagjosova, Elena (2004): The meaning and function of German modal particles. PhD thesis. Saarbrücken: Universität des Saarlandes. König, Ekkehard (1991): The meaning of focus particles. A comparative perspective. London: Routledge. König, Ekkehard & Susanne Requardt (1991): “A relevance-theoretic approach to the analysis of modal particles”. Multilingua 10, 63–77. DOI: dx.doi.org/10.1515/mult .1991.10.1-2.61. Kratzer, Angelika (1999): “Beyond ouch and oops. How descriptive and expressive meaning interact”. Cornell Conference on Theories of Context Dependency. URL: http://semanticsarchive.net/Archive/WEwNGUyO/. Lindner, Karin (1991): “‘Wir sind ja doch alte Bekannte’—the use of German ja and doch as modal particles”. In: Abraham, Werner, ed.: Discourse particles. Amsterdam: Benjamins, 303–328. McCready, Eric (2007): “Context shifting in questions and elsewhere”. Proceedings of Sinn und Bedeutung 11, 433–447. URL: http://parles.upf.edu/llocs/glif/pub/sub11/ individual/mccready.pdf. Peust, Carsten (2005): “Weiteres zur Personenverschiebung in der ägyptischen indirekten Rede”. Lingua Aegyptica 13, 77–102. Potts, Christopher (2005): The logic of conventional implicatures. Oxford: Oxford University Press. Quer, Josep (2005): “Context shift and indexical variables in sign languages”. Proceedings of SALT 15, 152–168. URL: http://elanguage.net/journals/salt/article/ view/15.152/1760.

modal particles and context shift

123

Repp, Sophie (2013): “Common ground management: Modal particles, illocutionary negation and verum”. In: Gutzmann, Daniel & Hans-Martin Gärtner, eds.: Beyond expressives. Explorations in use-conditional meaning. Leiden: Brill, 231– 274. [This volume]. Schlenker, Philippe (2003): “A plea for monsters”. Linguistics and Philosophy 26, 29–120. DOI: dx.doi.org/10.1023/A:1022225203544. ——— (2007): “Expressive presuppositions”. Theoretical Linguistics 33, 237–245. DOI: dx.doi.org/10.1515/TL.2007.017. Sharvit, Yael (2004): “Free indirect discourse and de re-pronouns”. Proceedings of SALT 14, 305–322. URL: http://elanguage.net/journals/salt/article/view/14.305/ 1747. ——— (2008): “The puzzle of free indirect discourse”. Linguistics and Philosophy 31, 353–395. DOI: dx.doi.org/10.1007/s10988-008-9039-9. Stalnaker, Robert (1973): “Presuppositions”. Journal of Philosophical Logic 2, 447– 456. DOI: dx.doi.org/10.1007/BF00262951. ——— (1974): “Pragmatic presuppositions”. In: Munitz, Milton K. & Peter Unger, eds.: Semantics and philosophy. New York: New York University Press, 197–213. ——— (2002): “Common ground”. Linguistics and Philosophy 25, 701–721. DOI: dx.doi.org/10.1023/A:1020867916902. Thurmair, Maria (1989): Modalpartikeln und ihre Kombinationen. Tübingen: Niemeyer. Truckenbrodt, Hubert (2004): “Zur Strukturbedeutung von Interrogativsätzen”. Linguistische Berichte 199, 313–350. Vuillaume, Marcel (2002): “Die impliziten Formen der Redewiedergabe”. In: Baudot, Daniel, ed.: Redewiedergabe, Redeerwähnung. Formen und Funktionen des Zitierens und Reformulierens. Tübingen: Stauffenburg, 83–96. Weydt, Harald (1977): Aspekte der Modalpartikeln: Studien zur deutschen Abtönung. Tübingen: Niemeyer. Zimmermann, Malte (2004): “Discourse particles in the left periphery”. In: Shaer, Benjamin & Werner Frey & Claudia Maienborn, eds.: ZAS Papers in Linguistics 35, 543–566. ——— (2011): “Discourse particles”. In: Maienborn, Claudia & Klaus von Heusinger & Paul Portner, eds.: Semantics. An international handbook of natural language meaning. Berlin: de Gruyter, 2011–2038.

DISCOURSE PARTICLES, COMMON GROUND, AND FELICITY CONDITIONS

Markus Egg

1. Introduction The goal of this paper is to argue for a novel approach to the semantics of discourse particles.1 This topic is a challenge in that the range of concrete usages of discourse particles in context can be extremely varied, which raises the question of whether it is sensible at all to presume that they contribute uniformly to the semantics of the discourses they are a part of. In order to be a serious competitor to analyses that regard discourse particles as polysemous (e.g., Helbig 1988), such a uniform semantic interpretation of a discourse particle would have to comply with two seemingly conflicting requirements: – it must be sufficiently specific – to allow the derivation of the interpretation of concrete examples – to prevent overgeneration – it must be sufficiently general to cover a wide range of concrete usages Previous approaches in the literature typically concentrate on the second of these requirements. E.g., Thurmair (1989), König (1997), Karagjosova (2004), or König & Requardt (1991) capture a wide range of usages of discourse particles in their analyses by assuming rather general meanings. As an example, consider the case of doch. It contributes a two-place relation to the meaning of a larger discourse, which relates the utterance of which doch is a part to information from the context (often this information is specified by a previous utterance to which the doch-utterance is a reaction).2 1 To avoid terminological confusion, let me point out that these particles are sometimes referred to as modal particles, whereas discourse particle has been used to denote another class of particles that is excluded from consideration here, viz., the group of particles like äh ‘eh’ or also ‘well’, which are used to handle the flow of conversation, turn taking, and problems of language production. 2 While many discourse particles introduce such a two-place relation in their semantics,

126

markus egg

This relation is described in terms of the features ‘known’ and ‘correction’ in Thurmair (1989), which means that doch-utterances introduce old information which corrects previously uttered incompatible information. Similarly, Karagjosova (2004: 137) states that doch introduces a contrast between the status of a proposition as part of the common ground and the fact that it is not available to one of the interlocutors. Usually, this interlocutor is the hearer, and then doch-utterances can act as reminders in that they present old information in situations in which hearers seem to be unaware of it. König & Requardt (1991) assign to doch-utterances the function of pointing out inconsistencies between old information and a new piece of information or action. Such general descriptions are applicable to cases like Karagjosova’s (2004) example (1), in which B’s doch-utterance reminds A of the (known) fact that Peter is ill, which is (at least prima facie) inconsistent with A’s announcement that Peter will also come along and therefore can act as a correction of this announcement: (1)

A: Peter wird auch mitkommen. B: Er ist doch krank. ‘A: Peter will come along, too. B: But he is ill.’

These descriptions of doch in the literature capture the intuitions about its semantic contribution, but do not attempt to spell out in detail the way in which the meaning of doch interacts with the meaning of utterances in the semantic construction of larger discourses. This very abstract nature of these general descriptions allows the subsumption of very many different uses of these particles. However, there are limits, e.g., in discourse-initial uses of doch-utterances like König & Requardt’s (1991) example (2), no such explanations apply. Such uses function as opening lines in conversations, they neither correct nor remind the hearer, nor is there an inconsistency between the utterance and the context:3 (2)

Sie sind doch Paul Meier. ‘You must be Paul Meier.’

In this paper, I will propose an analysis of discourse particles that offers the necessary generality to account for a wide range of uses while at the same e.g., schon, auch, or halt, others do not, e.g., ja, which merely characterises the utterance in which it occurs as old (or uncontroversial) information (Zimmermann 2011). In the remainder of the paper, I will concentrate on the former group of particles. 3 The cited analyses differ in coverage, e.g., Karagjosova’s (2004) account covers cases like (1), but does not extend to (2) (in spite of its wide coverage). Such examples would also be modelled in terms of a contrast between between CG status and availability for the hearer in her analysis (p. 134), which would characterise (2) as a reminder, which it is not.

discourse particles, common ground, felicity conditions 127 time being specific enough to make possible a semantic construction for discourses that comprise these particles. My attempt to pursue these two competing goals simultaneously rests on the following observation: Typically, e.g., in (1), an utterance with a discourse particle is used as a reaction to a previous utterance. It is sometimes possible to identify the (propositional) semantic arguments of the particle with the meanings of these two utterances, called ‘involved utterances’ for short. I will show that this is indeed the case for (1) in section 3. However, this is only the simplest case. Example (3), taken from Thurmair (1989), shows that semantic arguments of a discourse particle may differ from the meanings of the involved utterances: (3)

A: Seit wann hast du denn den Zauberberg? B: Den hast du mir doch vor zwei Jahren geschenkt. ‘A: Since when have you owned the Zauberberg? B: But you gave it to me two years ago.’

B’s utterance is a reaction to the implicit statement that A does not know the answer to his question (otherwise he would not have asked). This statement emerges as one of the arguments of doch in (3), although it differs from the meaning of A’s utterance. Therefore it is necessary to distinguish the semantic arguments of a discourse particle from the meanings of the involved utterances. The following nomenclature is used to highlight this observation: An utterance with a discourse particle is a “p(article)-utterance” (or a “partutterance”) for a given particle part; if it is used as a reaction to a preceding utterance, this second utterance is called “a(ntecedent)-utterance”. The semantic arguments of the particle are referred to as “a-proposition” and “pproposition”, respectively. This perspective on discourse particles assigns them a different role in discourse processing. Rather than merely indicating the relation between two already identified propositions, the meaning of the particle applied to its first argument (which is very often but not always the interpretation of the p-proposition) determines the range of potential a-propositions in the context of utterance. From this range, the hearer of the p-utterance must select the appropriate proposition. This intuition is similar to König & Requardt’s (1991) claim that such particles function as “metapragmatic instructions” that relate the p-proposition to a suitable proposition that is already assumed. For the working linguist, this observation entails that it is not possible to determine an a-proposition in terms of an utterance preceding the respec-

128

markus egg

tive p-utterance in all cases. This might explain (part of) the problems of pinning down the semantics of these particles in the literature, which characteristically compares such pairs of utterances to derive the semantics of discourse particles. However, merely decoupling the semantic arguments of a discourse particle from the meanings of the two involved utterances would run the risk of obtaining a too liberal analysis. In this paper, I will put forward the idea that the choice of semantic arguments is severly limited, apart from the meanings of the involved utterances, only utterance-related statements are feasible semantic arguments, in particular, those that emerge through the felicity conditions (Searle 1969) of these utterances. This approach to the semantics of discourse particles is in line with much earlier work that assigned speech acts a crucial role in the interpretation of discourse particles. They were analysed as speech act modifiers (Helbig 1977; Jacobs 1991; Lindner 1991; Sandig 1979; Zeevat 2000, 2004), some authors even put down this modification potential to the ability of discourse particles to add and/or modify felicity conditions of these speech acts (Karagjosova 2004; Waltereit 2001). However, in contrast to these analyses, I claim that the role of discourse particles lies not so much in a modification of speech acts or their felicity condition of the utterances they are a part of. Rather, discourse particles can relate these (p-)utterances to felicity conditions of other utterances (their respective a-utterances). In this way, one can account for the influence these particles exert on the way in which the utterances they are a part of are connected to the previous discourse. I will show in Section 3.2 that such a reference to felicity conditions of a previous utterance shows up e.g. in (3); here the doch-utterance reminds A of the fact that the first preparatory condition for a question (that the speaker does not know the answer) does not hold, since A (being the one who gave the book to B) should know since when the book has been in B’s possession.4 The distinction of involved utterances and semantic arguments of discourse particles makes it possible to follow the standard practice of investigating the semantics of discourse particles by analysing pairs of a- and p-utterances.

4 Such an analysis is compatible with the observation that discourse particles can show up in dependent clauses, which do not constitute a speech act by themselves (OrmeliusSandblom 1997: 123ff.): The doch-utterance itself is not presupposed to introduce a speech act, only the utterance to which it reacts is.

discourse particles, common ground, felicity conditions 129 The paper is organised as follows: I will introduce some background assumptions on discourse particles in general in Section 2, then Section 3 illustrates the proposed approach to these particles in detail for the particle doch. In Section 4 I will then show that the findings for doch carry over to the notoriously difficult discourse particle schon (without attempting an exhaustive analysis of schon). This section will also discuss ‘minimal pairs’ of doch- and schon-utterances (which only differ in that they comprise either doch or schon, respectively). Finally, section 6 will offer an outlook to further research. 2. Formal Background This paper follows much previous work in assuming that the meaning of discourse particles involves reference to the common ground (Franck 1980; Karagjosova 2004; König 1997; Lindner 1991; Zeevat 2000, 2004; Zimmermann 2011). Common ground (CG) and the interlocutors’ individual backgrounds are modelled in terms of common or individual belief (Stalnaker 2002). Individual belief is equated with the set of propositions that are true in all possible worlds compatible with the individual’s beliefs; common belief, with the set of propositions believed by all members of the respective group of believers. Stalnaker notes that this is an idealisation in that the CG might comprise propositions not shared by the background of every member of the group. This possibility will play a crucial role in the interpretation of discourse particles, which can be used to point out potential conflicts between individual backgrounds and common ground (see also Zeevat 2000, 2004). Reasoning on the CG and these backgrounds (which I assume to be not closed under deduction) often employs inference patterns, which are part of the CG and will be modelled in terms of defeasible deduction (Asher & Lascarides 2003). I.e., CG and backgrounds include statements of the form ‘p > q’ (p defeasibly entails q), which together with p allows deducing q defeasibly.5 This reference to the common ground makes the semantics of discourse particles dependent on the context, because the CG is relative to the interlocutor(s) of a- and p-utterances. This dependency triggers the shifting 5 This defeasible Modus Ponens applies only if ¬q does not already hold and if it is not possible to deduce ¬q at the same time (the latter constellation is known as “Nixon diamond”; Asher & Lascarides 2003).

130

markus egg

effects observed in Döring 2013 [this volume]. As an example, consider embedding the discourse (1) in a report like in (4) (for further examples, see Thurmair 1989: 75f.): (4)

A sagte, Peter komme auch mit. B entgegnete, er sei doch krank. ‘A said that Peter would come along, too. B retorted that he was ill.’

Assuming that one of the effects of doch is to present a proposition (here, that Peter is ill) as part of the common ground, the shifting effect emerges in (4) in that the CG relevant for the interpretation of doch is calculated with respect to A and B, not with respect to the speaker and hearer of (4). This means in particular that the speaker of (4) does not commit himself to the statement that Peter’s illness is in the common ground of himself and his hearer. 3. The Case of Doch The discourse particle doch introduces a notion of tension between the aand the p-proposition. I will first make this notion precise and explain it with the simplest group of examples before moving on to more involved examples. 3.1. Declarative a- and p-Utterances In the simplest examples, the a-proposition is expressed directly by the a-utterance, and the meaning of the declarative p-utterance provides the p-proposition.6 Consider e.g. (5) [= (1)] and (6), adapted from Karagjosova (2004); what is in tension in these examples is being ill on the one hand and going out and living healthily on the other hand, respectively. (5)

A: Peter wird auch mitkommen. B: Er ist doch krank. ‘A: Peter will come along, too. B: But he is ill.’

(6)

Ich bin oft krank. Dabei lebe ich doch gesund. ‘I am often ill. But I have a healthy lifestyle.’

The intuitive notion of tension is formalised as follows. The effect of doch p as a reaction to an a-proposition q against the common ground C is to

6 I cover only the unstressed particle doch, not the stressed version that could be paraphrased by trotzdem ‘nevertheless’. See Section 5 and Meibauer 1994 for the relation between differently stressed versions of doch and other discourse particles.

discourse particles, common ground, felicity conditions

131

remind the hearer that C comprises a potential impediment p for q, which can express e.g. surprise at the fact that nevertheless q should hold or put doubt on q. Still, q is not explicitly denied, since it is compatible with p. Formally, the common ground C comprises both p and the fact that p defeasibly entails ¬q (which by defeasible Modus Ponens would allow one to infer ¬q, if q did not hold): (7)

⟦doch⟧(p)(q) iff both p and p > ¬q are part of the common ground

In other words, the tension or contrast introduced by doch emerges because the inference pattern p > ¬q is not applicable to the a- and p-proposition. Note that there is a certain asymmetry between these propositions: While the p-proposition is closely related to the p-utterance, and must be given at the time of its utterance, there is more leeway for the a-proposition: It can but need not be related to an a-utterance and need not be available in the CG at the time of the p-utterance (see the discussion on p. 136f. below). This analysis is based on similar intuitions as the one of Weydt (1969) or König (1997), who assume that doch p points out a contradiction in the CG, in that q is incompatible with a consequence of p. The difference, however, is that I regard this incompatibility as a default only, following proposals of Karagjosova (2004) and Zeevat (2004). Such defaults can be overridden, which is illustrated by (6).7 The CG status of p is also expressed in Karagjosova’s (2004) analysis according to which doch can introduce a proposition as a reminder and Thurmair’s (1989) feature ‘known’. There is also disagreement in the literature about the direction of the implicature, in particular, Abraham (1991) and Ormelius-Sandblom (1997: 83ff.) assume the implicature q > ¬p for doch. As I will argue in Section 4, this is the implicature that characterises the discourse particle schon. In examples such as (5) and (6), p is the direct interpretation of the p-utterance. This analysis straightforwardly accounts for (5): Being ill is a potential impediment for going out, so, by pointing out Peter’s illness, B expresses surprise or disbelief at A’s announcement that Peter will also come along (without necessarily correcting it or refuting it completely, because even ill people can go out in principle). Similarly, in (6) the speaker expresses surprise at his frequent illness in spite of his healthy lifestyle. This example highlights the default status of the 7 This example is also at variance with Gast’s (2008) analysis according to which doch is used to remove an inconsistency from the common ground: In (6), the potentially incompatible propositions both remain in the common ground.

132

markus egg

impediment p: If p and q were contradictory, the speaker (who has uttered them both and is thus committed to the truth of either) would have uttered something contradictory, which intuitively is not the case. This analysis predicts that variants of (6) with contradictory a- and ppropositions q and p should be ruled out if spoken by one person, and function as corrections if p and q are uttered by different interlocutors, because then p and q could not be true simultaneously. Assuming that being often ill and being always well are contradictory, this prediction is borne out e.g. by (8) and (9): (8) *Ich bin oft krank, dabei bin ich doch immer gesund. ‘I am often ill. But I am always well.’ (9)

A: Ich bin oft krank. B: Du bist doch immer gesund. ‘A: I am often ill. B: But you are always well.’

Finally, if one assumes an inference whose conclusion is the negation of a proposition q and whose premiss is the negation of a presupposition of q (e.g., if it is not true that Max ever smoked then he cannot quit smoking), the analysis can also account for the fact that doch-utterances are often used to point out presupposition failures. However, I will offer a more generalisable analysis of such examples below, see p. 134 for details. (10) A: Max hat aufgehört zu rauchen. B: Er hat doch niemals geraucht. ‘A: Max has stopped smoking. B: But he never smoked at all.’

3.2. Declarative p-Utterances and Non-Declarative a-Utterances The analysis so far covers cases in which an a-proposition is provided by the meaning of an a-utterance. But in the case of doch, the a-proposition can also emerge as a felicity condition of the a-utterance. As an example, consider doch-utterances as reactions to questions, as in example (11) [= (3)]: (11)

A: Seit wann hast du denn den Zauberberg? B: Den hast du mir doch vor zwei Jahren geschenkt. ‘A: Since when have you owned the Zauberberg? B: But you gave it to me two years ago.’

Intuitively, doch in (11) suggests surprise at the question being asked at all, considering the fact that A himself gave the book to B and therefore should know that B possesses it. This intuition can be reconstructed as follows. B’s utterance expresses a proposition p (that A gave the book to B) and points out that p is part of the common ground. Furthermore, it is part of the CG that p is a potential obstacle for a specific a-proposition q (p > ¬q).

discourse particles, common ground, felicity conditions 133 As outlined in Section 1, p-utterances restrict the range of potential apropositions, and the hearer of a p-utterance tries to identify the a-proposition q in the given context. Since the meaning of the a-utterance in (11) is not a proposition, it cannot directly contribute q. But since B assumes that A is cooperative, A’s question has introduced into the CG all the felicity conditions that there are for questions, among them the first preparatory condition, viz., that A does not know the answer to his question. This is a suitable q, because it is reasonable to assume that if A gave the book to B (= p), he should know the answer to the question (= ¬q). Examples in which doch-utterances are used in reaction to imperatives work in analogy, e.g., (12): (12) A: Übersetze mir bitte diesen Brief. B: Ich kann doch kein Englisch. ‘A: Please translate this letter for me. B: But I do not know English.’

What is in tension here are B’s lack of proficiency in English (= p) and A’s belief that B can translate a English letter (i.e., the first preparatory condition of the request, which serves as q). Since it is CG information that B does not speak English, A should not have required B to translate letters in English. Doch-reactions can also target imperatives that express a piece of advice, e.g., in (13). Here B’s reaction introduces a potential impediment for the first preparatory condition for advice (that A has some reason to believe that his advice will benefit B). (13) A: Bestell dir die Schweinshaxe! B: Ich bin doch Vegetarier. ‘A: Do order the pork knuckle. B: But I’m a vegetarian.’

This use of doch shows up in reactions to declarative statements, too, as Karagjosova’s (2004) example (14) shows. Here the p-utterance clearly does not present a potential impediment for the proposition expressed by A. Rather, B’s doch-utterance focusses on the fact that A’s utterance expresses surprise, and suggests that A should not be surprised at all. (14) A: Peter sieht schlecht aus B: Er war doch lange im Krankenhaus. ‘A: Peter does not look healthy. B: But he has been in hospital for a long time.’

In (14), Peter’s long stay in the hospital (= p) is not a potential obstacle to not looking healthy, on the contrary. The answer to this puzzle lies in the observation that B’s utterance targets a felicity condition of A’s expression of surprise: One of the relevant felicity conditions is that the speaker considers the fact about which he is surprised as something extraordinary, which would not have happened in a normal course of events. I.e., Peter’s unhealthy appearance is presented as an extraordinary fact (our q).

134

markus egg

B presents a potential impediment for this estimation by pointing out that Peter has been in hospital for a long time (p), which defeasibly implies that his looking unhealthy is normal (= ¬q). Examples in which a doch-utterance is used to point out violated presuppositions like (10) above could be given a uniform analysis along these lines, too. For presupposition violations in assertive a-utterances, the dochutterance can be regarded as targeting the sincerity condition of statements, which requires the speaker to believe in what he states: He cannot do so, if it is CG information that a presupposition of this statement does not hold. For questions, one would have to augment Searle’s (1969) list of felicity conditions and assume the additional condition that the speaker expects the hearer to know the answer, which characterises questions as requests for information.8 This condition is violated if it is CG information that a presupposition of the question is not fulfilled. E.g., A’s question in (15) presupposes that B has a superior, and B’s reaction points out that the contrary is CG information: (15) A: Wie heißt Ihr Dienstvorgesetzter? B: Ich habe doch keinen. ‘A: What is the name of your superior? B: But I have none.’

A final example of a doch-utterance that targets a felicity condition is (16), a variation of an example of Thurmair’s (1989) (discussed in the original version as (36) below). (16) is uttered in a context in which the speaker, Katharina Blum, who has been libelled by the gutter press, is shown newspapers with counterstatements whose intention is to restore her good name. In such a context, it is possible to react with the doch-utterance (16); the a-utterance is not verbalised but can be paraphrased as “there are counterstatements that will restore your good name”, which is an attempt to console the speaker. (16) Das liest doch keiner. ‘But no one will read this.’

(16) is possible as a reaction in the described context, even though it does not express surprise or doubt at the fact that there are counterstatements by offering a potential impediment for it. Rather, it points out a potential obstacle from the CG for the sincerity condition of the consolation offered in terms of the implicit a-utterance: The speaker of a consolation believes that the fact expressed in the consolation will alleviate the problem which is bothering the hearer (= q). But 8 Knowing the answer is a precondition to the first preparatory condition of a request for this answer, viz., that the hearer can comply with the request.

discourse particles, common ground, felicity conditions 135 (16) points out that, according to the CG, no one will read the counterstatements that try to restore Katharina’s name (= p). This should prevent the persons showing her the counterstatements from believing that their sheer existence already alleviates her problem (that she was libelled by a tabloid). 3.3. Non-Declarative p-Utterances The next class of doch-utterances are not declarative (as noted by Helbig 1977 and many others). Doch shows up in imperative as well as in interrogative utterances (the latter adapted from Thurmair 1989): (17) Verklag mich doch! ‘Go ahead and sue me.’ (18) Komm doch nach Hause! ‘Do come home.’ (19) Wie heißt doch diese Kneipe in der Sredzkistraße? ‘What is the name of this pub in the Sredzkistraße?’ (20) Wie sagt Goethe doch so treffend? ‘What is this piercing remark of Goethe?’

In imperative sentences, doch can be used provocatively, as in (17), because it suggests that the hearer is not able to do what is requested of him. In imperatives like (18), doch expresses that the action requested/suggested to the hearer is a very obvious and natural thing to do. As for doch-questions, they refer to a piece of knowledge that the speaker knows or is supposed to know (Thurmair 1989), e.g., in (19), which implies that the speaker knows the answer at least in principle, or in (20), which announces quotations. What all these utterances have in common is that speakers use doch to point out that they are aware of evidence from the CG which suggests that one of the felicity conditions of the utterance itself does not hold. Interpreting these sentences poses additional challenges in that the putterance itself no longer provides a proposition which could serve as the semantic argument of doch,9 furthermore, it need not be accompanied by an a-utterance from which to derive the a-proposition at all. For these cases, I assume that the p-proposition p (the argument of doch) is the fact that the sentences were uttered, which (trivially) is part of the common ground C (the first condition in (7)). Formally, this notion could be

9 This holds expecially for interrogative sentences; for imperative sentences, this problem would disappear in Schwager’s (2006) modal, hence, declarative, interpretation of imperatives.

136

markus egg

made precise e.g. in terms of a conversation table as in Farkas & Bruce (2010). Given the general assumption of cooperativity of interlocutors, the felicity conditions associated with different kinds of illocutionary acts emerge from the common ground C as default entailments from the utterance of a specific illocutionary type to each of its felicity conditions (the CG implicature p > ¬q in (7); here ¬q refers to the respective felicity condition). This means that using doch in these cases triggers the search for a suitable a-proposition q, which is derivable from the common ground and negates one of the felicity conditions of the utterance. This means that the a-proposition is not yet in the CG at the time of the p-utterance, it is only introduced by the p-utterance itself. E.g., using doch in (17) suggests that the first preparatory condition of a request (that the speaker believes that the hearer can do what is requested of him) does not hold, even though this could be deduced defeasibly from the fact that the request was made. This example also violates the sincerity condition for requests because the speaker does not want the hearer to actually go ahead and sue him,10 but its challenging and provoking flavour can only be explained in terms of a violation of the first preparatory condition of requests. It is also possible to target the second preparatory condition of a request/ piece of advice (that it is not obvious to speaker and hearer that the hearer does what is requested of him in a normal course of events) by using doch in an imperative. This is illustrated by (18), which can serve as a request or as a piece of advice. (18) thus suggests that it is obvious that the hearer will do anyway what is required/advised, even though uttering the request/piece of advice defeasibly entails the contrary. The effect is that the request/advice is presented as a very natural thing to do, which should need no further prompting by the speaker. My interpretation of such examples deviates from Thurmair’s (1989), who interprets them also in terms of a feature ‘correction’, i.e., she claims that doch-imperatives are used to correct an unwanted behaviour of the hearer. However, this explanation would be incompatible with examples like (21): (21) A: Ich würde dich gerne mal besuchen. B: Dann komm doch mal vorbei! ‘A: I’d love to call on you some time. B: Well, just drop by.’

B’s reaction is not a request to the hearer to change his behaviour in any way. Instead, it confirms A’s behaviour by presenting the call he wants to pay on

10

This was noted by an anonymous reviewer.

discourse particles, common ground, felicity conditions 137 B as a very natural thing to do for A, in other words, the second preparatory condition of a request is violated in this example once again. For questions, the effect of doch can be explained in an analogous fashion. It indicates that the utterance violates a felicity condition whose validity could be deduced defeasibly from the fact that the question has been asked, viz., the first felicity condition for questions (that the speaker does not know the answer already). By using doch, the speaker points out that this expectation is not fulfilled, either because the answer escapes the speaker only momentarily, as in (19), or because the speaker obviously knows, as in the formulaic (20), which is conventionalised as an announcement. If we stick to Searle’s (1969) conception of felicity conditions for questions, this analysis also explains why ordinary (non-rhetorical) wh-questions do not allow doch, e.g., in (22); here there is no tension between making the p-utterance and potential obstacles for its felicity conditions: (22) *Wer schreibt dir doch? ‘But who is corresponding with you?’

What the analysis does not yet explain exhaustively, however, is the question of why doch cannot be used in polar questions, even in a context in which the speaker knows the answer in principle but has a temporary lapse of memory with respect to this answer (in analogy to (19)): (23) *Habe ich am Freitag doch einen Arzttermin? ‘Do I have a doctor’s appointment on Friday?’

A first explanation would be to point out that a temporary lapse of memory is far more likely to involve a choice from a large range of alternatives (as introduced by wh-questions, e.g., other potential pub names or Goethe quotes) than a decision between a proposition and its negation. Rhetorical questions are also incompatible with doch. Consider e.g. the contrasting pairs (24a)/(24b) and (24a)/(24c): (24) a. A: Ich werde meinen 30. Geburtstag mit einem großen Fest feiern. ‘A: I’ll throw a big party on the occasion of my 30th birthday.’ b. B: Es würde doch keiner zu deinem Fest kommen. ‘B: But no one would come to your party.’ c. B: *Wer würde doch zu deinem Fest kommen? ‘B: But who would come to your party?’

The contrast in acceptability shows that the inacceptability of the latter pair (and of the rhetorical doch-question in particular) cannot be due to the interpretation of the rhetorical question as a negated statement, because in this case, (24b) should be as unacceptable as a response to (24a) as (24c).

138

markus egg

The problem with (24c) is due to the fact that rhetorical questions characterise the statement expressed by them as information that is derivable from the common ground, just like doch (Egg 2007). Therefore, (24c) emerges as more complex yet not more informative than (24b), thus falls prey to Gricean reasoning based on the maxim of manner (Grice 1975). In sum, the difference between the non-declarative doch-utterances and declarative doch-utterances like in (11) or (12), which react to a non-declarative a-utterance, is that the latter target felicity conditions of this autterance, whereas the former refer to their own felicity conditions. Since non-declarative doch-utterances are not propositions semantically, the semantic argument of doch cannot be the meaning of the p-utterance directly. Instead, doch applies to the fact that the speaker uttered the sentence. Once this analysis of non-declarative doch-utterances is adopted, one can also identify a group of declarative doch-utterances that lend themselves to such an analysis straightforwardly, viz., discourse-initial doch-utterances: (25) Morgen fahre ich doch nach Wien. ‘Well, I’ll go to Vienna tomorrow’ (26) Du hast doch ein Auto ‘Well, you have a car’

Here the intuitive effect of doch is one of a reminder. This is compatible with assuming that it characterises the p-utterance as old, here, the speaker’s travel plans or the fact that the hearer has a car. These examples are nevertheless challenging because there is no obvious tension to information from the CG. But if one assumes that the tension arises between the CG and felicity conditions of the utterance itself, then the explanation is that doch targets the second preparatory condition for statements here, viz., that it is not obvious to the speaker that his information is already known to the hearer (the fact that this condition is not met in these cases is also noted in Karagjosova 2004: 167). This condition (= ¬q) is defeasibly implied by making the statement (= p), but it is part of the CG that the speaker knows that the hearer knows (= q). The prediction that such discourse-initial doch-utterances are only acceptable if they introduce old information is borne out by examples like (27). They are fine if the hearer has been informed about the identity of the speaker before, but unacceptable if uttered out of the blue between two interlocutors that have not met before: (27) Ich bin doch Paul Meier. ‘But I am Paul Meier.’

discourse particles, common ground, felicity conditions 139 The use of doch-utterances in w-exclamatives can also refer to their felicity conditions, as in (28). If we assume following Zanuttini & Portner (2003) and Rett (2008) that (28) characterises the degree of Amélie’s beauty as unexpectedly or surprisingly high, the utterance introduces a potential obstacle to the sincerity condition of the statement (that the speaker believes in what he says). (28) Wie schön Amélie doch ist! ‘How beautiful Amélie is!’

This explanation also covers uses of doch in expressions of outrage. Here doch indicates that it is common knowledge that the hearer does know that the situation or action to which the speaker refers is indeed outrageous (i.e., the second preparatory condition for statements is not met): (29) Das ist doch die Höhe! ‘That is the limit!’

This analysis raises the question of how the interpretation of the p-proposition p (the first semantic argument of doch) as the fact that the utterance has been made can be restricted. If this interpretation were freely available for all declarative doch-utterances, the analysis would overgenerate considerably, e.g., doch is predicted to be acceptable in any utterance as long as the utterance presents material from the common ground, because then invariably the second preparatory condition for statements is not met. But then (30) should be as acceptable as its counterpart (5), which it is not: (30) A: Peter wird auch mitkommen. B: *Er ist doch gesund. ‘A: Peter will come along, too. B: But he is well.’

My intuition is that this interpretation of the p-proposition is only available as a last resort, if the p-utterance does not yield a proposition in its semantics, or if there is no potential a-proposition available in the context (which happens particularly at the beginning of a discourse). In sum, it is possible to account for a wide range of doch-sentences with a uniform semantic analysis: Doch relates two propositions p and q iff p is part of the common ground and so is the fact that p defeasibly implies ¬q, i.e., p presents a potential impediment for q. The necessary flexibility of the analysis comes in during the correlation of p and q with utterances: While p often emerges as the meaning of the doch-utterance, it is identified with the fact that this utterance has been made in the case of non-declarative and discourse-initial declarative p-utterances. The other proposition q can be given as the meaning of a preceding a-utterance to which the doch-utterance reacts, but it can also be one of the felicity conditions of the a-utterance, or, if p is the fact that the doch-

140

markus egg

utterance has been made, a felicity condition of the doch-utterance itself (i.e., in these cases there is no need to fall back upon an a-utterance). 4. The Case of Schon In this section, I will show that it is possible to employ the analysis sketched in the previous section for doch also for an account of the semantics of schon (the unstressed discourse particle). This section is not meant to offer a comprehensive account of the particle, its more moderate goal is to advocate for this particle a specific interpretation along the lines sketched for doch. I.e., for schon, too, the interpretation can be specific because it combines in a flexible way with the involved utterances or their felicity conditions. Like in the case of doch, I will start my account of schon with a simple example, in which the semantic arguments of the particle are identical to the meanings of the involved utterances: (31) A: Peter ist nicht sehr klug. B: Er wird seine Abschlußprüfungen schon bestehen. ‘A: Peter is not very clever. B: He will pass his finals, don’t worry.’

Here the intuition is that the schon-utterance is used to confirm the autterance, but at the same time, rules out one of its potential consequences, i.e., B accepts A’s claim that Peter is not very intelligent, but rules out the tacit implication that this will make him fail his finals. This characterises the p-proposition (here, that Peter will pass) as unexpected and somewhat surprising. This description ties in with, but at the same time makes more precise, Thurmair’s (1989) description that “parts of the preceding utterance are confirmed, but the overall validity of the utterance is restricted”. An additional feature is that declarative schon-utterances must refer to something positive. A variant of (31) in which the polarities are reversed sounds strange because it would only be acceptable if Peter’s failure is regarded as positive : (32) A: Peter ist sehr klug. B: Er wird seine Abschlußprüfungen schon nicht bestehen. ‘A: Peter is very clever. B: He will fail his finals, don’t worry.’

This explains why schon-utterances often have a soothing flavour, because they point out an unexpected positive fact in the light of information that suggests the opposite (i.e., the a-proposition). This effect can be seen very clearly in the jocular (33), whose literal interpretation is contradictory (because failing is not positive), and in whose eventual interpretation schon

discourse particles, common ground, felicity conditions

141

enforces a non-literal interpretation of schiefgehen “fail” as its antonym, i.e., the eventual interpretation is ‘it will go fine, don’t worry’: (33) Wird schon schiefgehen ‘It will go wrong, don’t worry.’

Formally, the meaning of schon is reconstructed as follows. (Note, however, that no positive evaluation of p is required for non-declarative schonutterances.)

(34) ⟦schon⟧(p)(q) iff both p and q hold, p is evaluated positively, and, according to the common ground C, q defeasibly entails ¬p (q > ¬p).

Another example of Thurmair’s that illustrates this pattern (adapted) is (35); here p is the claim that no one will shoot at B (which is evaluated as positive), even though the fact that the situation is dangerous (= q) defeasibly implies the opposite: (35) A: Es ist gefährlich. B: Es wird schon keiner auf dich schießen. ‘A: It is dangerous. B: No one will shoot you, don’t worry.’

This account also covers rhetorical schon-questions whose interpretation is negative (no entity fits the question; see also Meibauer 1986), as in the original Katharina Blum example. Recall that it is uttered by Katharina Blum as a reaction to being shown counterstatements that intend to restore her reputation after having been libelled by a tabloid: (36) Wer liest das schon? ‘But who will read this?’

The interpretation of (36), our p, is that no one will read these counterstatements, and that this is old information (which can be derived from the CG).11 This is unexpected, given the fact that there are counterstatements (our q, which is not verbalised in the example but expressed by showing the counterstatements to Katharina), and that, according to the common ground, one would expect that, once these counterstatements are there, they will be read by some. In this example, the statement that no one will read this is not evaluated positively, on the contrary, which exemplifies the observation that schon introduces a positive evaluation for declarative p-utterances only. Note that one could not explain the negative evaluation of the p-utterance (that no one will read the counterstatements) in terms of a scope of the negation (as 11 Here the fact that the rhetorical question characterises its information as old is not redundant, because the semantics of the particle itself does not do this (as opposed to doch).

142

markus egg

introduced by the rhetorical question) over schon and a subsequent positive evaluation of the fact that someone reads the counterstatements.12 This analysis is corroborated by comparing (36) to (37). Even though (37) is a declarative paraphrase of (36),13 it could not substitute for (36) as a reaction in the context sketched for (36): (37) Das liest schon keiner. ‘No one will read this, don’t worry.’

In the Katharina Blum context, (37) would have to be interpreted (nonsensically) as an attempt of Katharina to comfort those showing her the counterstatements by pointing out to them that—contrary to what they expected—the (unpleasant) event of the counterstatements being read is not going to happen. This contrast between (36) and (37) follows from the fact that (37) is declarative and therefore is evaluated positively. Schon-utterances can also react to non-declarative a-utterances, e.g., in the following two examples. (38) A: Paß auf Dich auf! B: Mir wird schon nichts passieren. ‘A: Take care. B: Nothing will happen to me, don’t worry’ (39) A: Was soll ich meinen Eltern zu Weihnachten schenken? B: Dir wird schon etwas einfallen. ‘A: What should I give to my parents for Christmas? B: You will be able to think of something in time, don’t worry.’

In (38), B’s reaction targets the first preparatory condition of A’s warning, viz., that he has reason to believe that something will happen that is not in the interest of B (= q). This defeasibly entails that B will be harmed in some way, which is ruled out by B in his reply to A’s warning (= p). The discourse (39) starts with a question whose first preparatory condition states that A does not know the answer. B refers to this condition in his reply, he does not refute it, but points out that a proposition defeasibly deducible from this condition (that A’s state of ignorance will persist) does not hold.

12 Such an explanation would run counter to the fact that B’s schon-utterance in (35) (that no one will shoot at A, which is evaluated positively) could be replaced by the rhetorical question Wer wird schon auf dich schießen? ‘Who (i.e., no one) will shoot at you’. Assuming narrow scope of the evaluation w.r.t. the negation would falsely assume a positive evaluation of B being shot at. Thanks to an anonymous reviewer for this example. 13 As opposed to (36), the fact that no one reads the counterstatements is not presented as CG information in (37), but this is of no avail for the present argument.

discourse particles, common ground, felicity conditions 143 These examples once again show that the semantic arguments of a discourse particle and the involved utterances need not coincide, in particular, this holds for non-declarative a-utterances whose interpretation is not a proposition. Other instances of this kind of use of schon are non-negative rhetorical questions like in (40): (40) A: Wen wird Günther zu seiner Feier einladen? B: Wen wird er schon einladen? (Seine langweiligen Kollegen natürlich.) “A: Who will Günther invite to his party? B: Well, who will he invite? (His boring colleagues, of course.)”

Here the interpretation of B’s rhetorical question, our p, is non-negative, it can be paraphrased as ‘it is known information that Günther will invite specific persons to his party (viz., his boring colleagues)’. This statement targets the first preparatory condition of questions (that A does not know the answer to his question, our q): If A does not know the answer then it makes sense to assume (to derive defeasibly) that the answer is not already in the common ground. B points out, however, that this assumption does nevertheless not hold (i.e., that the answer is CG information). Just like for doch, there are also cases of schon-utterances that target a felicity condition of an a-utterance even though the a-utterance is declarative: (41) A: Du hättest Hans helfen müssen. B: Was hätte ich schon tun können? ‘A: You should have helped Hans. B: What could I have done?’

B’s reaction in (41), a rhetorical question, can be paraphrased as “B could not have done anything (and this is old information)”. A’s utterance is a reproach, he is telling B off for not offering any help. The a-proposition q in this example then is a felicity condition for reproaches which states that the speaker believes that the hearer would have been able to act in a different way. This defeasibly entails the opposite of B’s statement, and is thus ruled out by B’s reaction. Another example that works exactly in the same way is (42). Here B’s reaction (that it is part of the CG that no one likes losing) targets the second preparatory condition for statements (that it is not obvious to the speaker that the hearer knows already). This condition suggests deducing defeasibly that there are some people (at least Karen) for whom it is not CG information that they are angry at being defeated, which is blocked by B’s utterance. (42) A: Karen hat sich sehr über ihre Niederlage geärgert. B: Wer verliert schon gern? “A: Karen was very angry at her defeat. B: Well, who likes losing in the first place?”

144

markus egg

Finally, there are also cases of non-declarative schon-utterances. They are restricted to imperatives since all schon-interrogatives are rhetorical (Meibauer 1986). (43) Nun mach schon! “Come on!”

Following the approach sketched for doch, p in such cases is the fact that the utterance was made. I.e., the utterance itself is unexpected due to information from the CG, our q. This information might have been introduced in terms of a previous utterance of the hearer, which introduced a potential obstacle for complying with the speaker’s request, or merely been conveyed by the hearer’s hesitation to comply. Consequently, the CG suggests that no such utterance should be made at all. This reconstruction fits in with Thurmair’s (1989) intuition that these imperatives relate to obstacles/hesitations in the context. 5. The Next Steps In this section, some further steps of the presented approach to discourse particles are outlined in order to show that this approach has further scientific potential and links up with ongoing lines of research in the literature on discourse particles. Once semantic descriptions of individual discourse particles like schon and doch have been formulated, they can be tested (and, possibly, be refined further) by contrasting ‘minimal pairs’ of utterances that only differ in including different discourse particles. For such a pair that differs only in doch and schon, respectively, consider the dialogues (44a)/(44b) in contrast to (44a)/(44c): (44) a. A: Ich habe ein Loch in meiner Hose. “A: There’s a hole in my trousers.” b. B: Das merkt doch keiner. “B: But no one will notice.” c. B: Das merkt schon keiner. “B: No one will notice, don’t worry.”

Intuitively, both (44b) and (44c) as a reaction to (44a) express that the hole is not a problem, but (44b) sounds less friendly than (44c). This intuition can be explained in the following way. B’s reaction in (44a)/ (44b) casts doubt on the speech act (complaint) of A by pointing out information accessible to A as part of the CG (that no one will notice the hole), which suggests that one of the felicity conditions of the complaint

discourse particles, common ground, felicity conditions 145 (that the situation referred to is a problem for A) does not obtain. I.e., if A had taken into account this information, he would not have uttered the complaint at all. But in (44a)/(44c), B accepts A’s speech act, his reaction merely tries to soothe A by pointing out that a plausibly expectable negative consequence of the fact about which A is complaining is not going to happen (i.e., people will not notice). A similar explanation can be offered for the two Katharina Blum reactions (46) and (45) [= (16) and (37), respectively]: (45) Das liest doch keiner. “But no one will read this” (46) Das liest schon keiner. “No one will read this, don’t worry.”

While the doch-variant expresses doubt at the consoling a-utterance by pointing out evidence for the claim that one of its felicity conditions might not hold, the schon-variant would accept it, but point out that an unpleasant yet expectable consequence (that people will read it) will not occur after all, which makes (45) odd in the given context. Such contrastive studies of discourse particles need to be informed about the properties of several discourse particles simultaneously; this is also the case for the second line of research to be listed in this section, viz., investigating the cooccurrence of discourse particles in one single utterance. (47) Das ist ja doch schon ein Unterschied. “Well, but this does make a difference, you know.”

Semantic descriptions of individual discourse particles must furthermore be able to explain the range of possible combinations (see e.g Thurmair 1989 for a list of such combinations) and for their interpretation. Integrating isolated semantic accounts of discourse particles is also called for in analyses of the relation between lexically ambiguous entities that include uses as discourse particles. E.g., schon in (48), indicates that a proposition (in our example, A’s claim) is a possibility, but only just. I.e., it is above, but very close to, the threshold that divides possible from impossible propositions. Such an account could be formalised in terms of Kratzer’s (1991) graded modality. (48) A: Amelie ist begabt. B: Schon, aber … “A: Amelie is talented. B: Yes, she is, but …”

146

markus egg

This closely resembles the spatial use of schon in (49), which expresses that Basel does lie in Switzerland, but just behind the border (from a German point of view). (49) Basel ist schon in der Schweiz. “Basel is already in Switzerland.”

The connection between the spatiotemporal uses of schon and the use in (48) is metaphorical in that the structuring schon imposes on the temporal or spatial domain is mapped onto the abstract domain of modality (Klein 2010; Löbner 1989; Meibauer 1994). Finally, there is the question of how to relate unstressed and stressed uses of discourse particles (Brauße 1988; Féry 2010). I will illustrate this problem with an initial analysis of stressed doch, which can function as an utterance by itself and is interpreted as a correction of a preceding negated utterance. Inference patterns from the context are not relevant anymore: (50) A: Peter ist nicht krank. B: DOCH./Er ist DOCH krank. “A: Peter is not ill. B: Yes, he is.”

This example can be explained without having to assume ambiguity (Egg & Zimmermann 2012): In (50), all constituents of B’s utterance have already been mentioned, therefore only the discourse particle can carry the (obligatory) sentence accent. Since previously mentioned material can be elided, the discourse particle may remain alone and constitute the utterance by itself. The interpretation of stressed doch (without inference patterns) follows from the fact that the a- and p-utterances are each other’s negation (‘Peter is ill’—‘he is not ill’). This means that the inference pattern becomes trivial (p > p), hence, can be ignored in the interpretation of the dialogue in (50). In sum, while I did not attempt to cover all these lines of research exhaustively, it could be shown that the proposed analysis of discourse particles can be integrated into them and has the potential to give rise to further research questions. 6. Conclusion In this paper, a research programme for discourse particles was presented, which aims at deriving a very specific unified semantic analysis of these particles that nevertheless manages to capture a wide range of uses of the particles. These two competing goals could be pursued in one analysis because it was assumed that there is more flexibility in combining the meaning of a

discourse particles, common ground, felicity conditions 147 discourse particle with the meanings of surrounding utterances. Discourse particles like doch and schon relate two propositions semantically, but the meaning of the utterance of which the particle is a part, and the meaning of the utterance to which this first utterance reacts are not the only feasible semantic arguments of the particles. Instead, particles can have felicity conditions of these two utterances as semantic arguments, too. This research programme was illustrated in detail by investigating the particle doch and was shown to be applicable to schon as well. Finally, links of the programme to other lines of research on discourse particles and further research questions were sketched. Acknowledgements The author wishes to thank two anonymous reviewers for extensive comments on an earlier version of this paper. References Abraham, Werner (1991): “Discourse particles in German. How does their illocutive force come about?” In: Abraham, Werner, ed.: Discourse particles. Amsterdam: John Benjamins, 203–252. Asher, Nicholas & Alex Lascarides (2003): Logics of conversation. Cambridge: Cambridge University Press. Brauße, Ursula (1988): “Modalpartikeln in Fragesätzen”. Linguistische Studien 177, 77–113. Döring, Sophia (2013): “Modal particles and context shift”. In: Gutzmann, Daniel & Hans-Martin Gärtner, eds.: Beyond expressives. Explorations in use-conditional meaning. Leiden: Brill, 95–123. [This volume]. Egg, Markus (2007): “Meaning and use of rhetorical questions”. In: Proceedings of the 16th Amsterdam Colloquium. Ed. by Aloni, Maria & Paul Dekker & Floris Roelofsen. Universiteit van Amsterdam, ILLC, 73–78. Egg, Markus & Malte Zimmermann (2012): “Stressed out! Accented discourse particles—the case of ‘DOCH’.” Proceedings of Sinn und Bedeutung 16. 225–238. Farkas, Donka F. & Kim B. Bruce (2010): “On reacting to assertions and polar questions”. Journal of Semantics 27.1, 81–118. DOI: dx.doi.org/doi:10.1093/jos/ffp010. Féry, Caroline (2010): “Information structure of schon”. In: Hanneforth, Thomas & Gisbert Fanselow, eds.: Language and logos. A festschrift for Peter Staudacher. Berlin: Akademie-Verlag, 160–175. Franck, Dorothea (1980): Grammatik und Konversation. Stilistische Pragmatik des Dialogs und die Bedeutung deutscher Modalpartikeln. Königstein: Scriptor. Gast, Volker (2008): “Modal particles and context updating—the function of German ja, doch, wohl and etwa”. In: Vater, Heinz & Ole Letnes, eds.: Modalverben und Grammatikalisierung. Trier: Wissenschaftlicher Verlag, 153–177.

148

markus egg

Grice, H. Paul (1975): “Logic and conversation”. In: Cole, Peter & Jerry L. Morgan, eds.: Speech acts. New York: Academic Press, 41–58. Helbig, Gerhard (1977): “Partikeln als illokutive Indikatoren im Dialog”. Deutsch als Fremdsprache 14, 30–44. ——— (1988): Lexikon deutscher Partikeln. Leipzig: Verlag Enzyklopädie. Jacobs, Joachim (1991): “On the semantics of modal particles”. In: Abraham, Werner, ed.: Discourse particles. Amsterdam: John Benjamins, 141–162. Karagjosova, Elena (2004): The meaning and function of German modal particles. PhD thesis. Saarbrücken: Universität des Saarlandes. Klein, Wolfgang (2010): “About the German particles schon and noch”. Ms. MPI Nijmegen. König, Ekkehard (1997): “Zur Bedeutung von Modalpartikeln im Deutschen. Ein Neuansatz im Rahmen der Relevanztheorie”. Germanistische Linguistik 136, 57– 75. König, Ekkehard & Susanne Requardt (1991): “A relevance-theoretic approach to the analysis of modal particles”. Multilingua 10, 63–77. DOI: dx.doi.org/10.1515/ mult.1991.10.1-2.61. Kratzer, Angelika (1991): “Modality”. In: von Stechow, Arnim & Dieter Wunderlich, eds.: Semantik. Ein internationales Handbuch der zeitgenössischen Forschung. Berlin and New York: de Gruyter, 639–650. Lindner, Karin (1991): “ ‘Wir sind ja doch alte Bekannte’—the use of German ja and doch as modal particles”. In: Abraham, Werner, ed.: Discourse particles. Amsterdam: Benjamins, 303–328. Löbner, Sebastian (1989): “German schon—erst—noch: An integrated analysis”. Linguistics and Philosophy 12, 167–212. DOI: dx.doi.org/10.1007/BF00 627659. Meibauer, Jörg (1986): Rhetorische Fragen. Tübingen: Niemeyer. ——— (1994): Modaler Kontrast und konzeptuelle Verschiebung. Studien zur Syntax und Semantik deutscher Modalpartikeln. Tübingen: Niemeyer. Ormelius-Sandblom, Elisabet (1997): Die Modalpartikeln ja, doch und schon. Zu ihrer Syntax, Semantik und Pragmatik. Stockholm: Almqvist & Wiksell. Rett, Jessica (2008): “A degree account of exclamatives”. Proceedings of SALT 18, 601–608. URL: http://ecommons.library.cornell.edu/bitstream/1813/13058/2/Rett .pdf. Sandig, Barbara (1979): “Beschreibung des Gebrauchs von Abtönungspartikeln im Dialog”. In: Weydt, Harald, ed.: Die Partikeln der deutschen Sprache. Berlin: de Gruyter, 84–94. Schwager, Magdalena (2006): “Conditionalized imperatives”. Proceedings of SALT 16, 241–258. URL: http://ecommons.cornell.edu/bitstream/ 1813/7591/1/salt16_ schwager_241_258.pdf. Searle, John R. (1969): Speech acts. An essay in the philosophy of language. Cambridge: Cambridge University Press. Stalnaker, Robert (2002): “Common ground”. Linguistics and Philosophy 25, 701– 721. DOI: dx.doi.org/10.1023/A:1020867916902. Thurmair, Maria (1989): Modalpartikeln und ihre Kombinationen. Tübingen: Niemeyer. Waltereit, Richard (2001): “Modal Particles and their Functional Equivalents. A Speech-Act-Theoretic Approach”. Journal of Pragmatics 33, 1391–1417. DOI: dx.doi.org/10.1016/S0378-2166(00)00057-6.

discourse particles, common ground, felicity conditions 149 Weydt, Harald (1969): Abtönungspartikel. Die deutschen Modalwörter und ihre französischen Entsprechungen. Bad Homburg: Gehlen. Zanuttini, Raffaella & Paul Portner (2003): “Exclamative clauses: At the syntaxsemantics interface”. Language 79, 39–81. DOI: dx.doi.org/10.1353/lan.2003.0105. Zeevat, Henk (2000): “Discourse particles as speech act markers”. LDV-Forum 17, 74–91. ——— (2004): “Particles. Presupposition triggers, context markers, or speech act markers?” In: Blutner, Reinhard & Henk Zeevat, eds.: Optimality theory and pragmatics. Houndmills: Palgrave, 91–111. Zimmermann, Malte (2011): “Discourse particles”. In: Maienborn, Claudia & Klaus von Heusinger & Paul Portner, eds.: Semantics. An international handbook of natural language meaning. Berlin: de Gruyter, 2011–2038.

I LOVE ME SOME DATIVES: EXPRESSIVE MEANING, FREE DATIVES, AND F-IMPLICATURE

Laurence R. Horn

1. Introduction In addition to introducing the very successful product line of conversational implicatures, Paul Grice and his successors have assembled an arguably inchoate class of phenomena under the trade name of conventional implicature, whose reception in the scholarly marketplace has been somewhat bridled. A conventional implicature associated with expression E is a non-cancelable contribution to the content of an expression whose falsity does not affect the truth conditions of E. This construct has evoked much recent skepticism—Bach (1999) has consigned it to the dustbin of mythology, while Potts (2005 et seq.) has undertaken a pyrrhic rehabilitation via redefinition—but Grice’s admittedly sketchy device for treating aspects of content that are irrelevant to the truth conditions of an asserted proposition has a rich lineage. In delineating meanings that do not “affect the thought” or “touch what is true or false”, Frege (1892, 1897/1979, 1918) directly prefigured Grice’s conventional implicature. While much recent scholarship has followed Dummett (1973) in dismissing Frege’s positive proposals in this area as representing a confused, subjective notion of “tone”, this fails to do justice to Frege’s intention and practice. For a range of connectives, expressive particles, pronouns, and syntactic constructions, some proposed for the role by Frege and/or Grice and others not considered by them, such an approach remains eminently plausible. 2. The Fregean Landscape of Sub-Sense Relations In addition to the presupposition (Voraussetzung) of reference for proper names in sentences like Kepler {died/didn’t die} in misery (Frege 1892: 40), Frege also allows for the weaker relation of “side-thought” (Nebengedanke) of existence for universally quantified statements, an essentially pragmatic

152

laurence r. horn

relation which involves material neither meant nor presupposed as admitted (Frege 1906: 306–307). The former constitutes a necessary condition for an assertion to be made; the latter does not. (See Horn 2007 for elaboration.) But this does not exhaust the inventory of Fregean relations for implication beyond—or below—the domain of sense and reference. In fact, the issue arises for Frege in the Begriffsschrift, before the distinction between Sinn and Bedeutung is even drawn: The difference between ‘and’ and ‘but’ is of a kind that has no expression in this Begriffsschrift. A speaker uses ‘but’ when he wants to hint [einen Wink geben] that what follows is different from what might at first be supposed. (Frege 1879: 63)

Similar hints or suggestions recur elsewhere. Considering a spectrum of linguistic phenomena ranging from particles like although, but, yet, still, and already to active/passive alternations and word order, Frege devotes several passages throughout his works to describing such expressions and constructions that “aid the hearer’s understanding” without, however, affecting the propositional content—or, in Fregean parlance, the thought. Here is Frege in “Über Sinn und Bedeutung” on the meaning contributions of adversative particles (here and below, boldface is added): Subsidiary clauses beginning with ‘although’ [obgleich] also express complete thoughts. This conjunction […] does not change the sense of the clause but only illuminates it in a peculiar fashion (footnote: Similarly in the case of ‘but’ [aber], ‘yet’ [doch].) We could indeed replace the concessive clause without harm to the truth of the whole by another of the same truth value; but the light in which the clause is placed by the conjunction might then easily appear unsuitable, as if a song with a sad subject were to be sung in a lively fashion. (Frege 1879: 167)

Crucially, then, the difference between p and q and p although q does not affect the truth conditions of the proposition involved, a claim that appears correct. Imagine I bet you $100 that Robin will marry Chris although Chris is extremely poor and it develops that Robin does marry Chris, but precisely because of, not in spite of, Chris’s impecuniary state. Then the although-condition is not satisfied, but it is clear (to me!) that I have won the bet, for which all that matters is the two parties wed. Frege provides an inventory of phenomena lending themselves to similar analyses in “Der Gedanke”, beginning with the choice between the neutral horse [Pferd] and its evaluatively laden counterparts like steed or nag [Roß, Gaul, Mähre]:

expressive meaning, free datives, and f-implicature

153

It makes no difference to the thought whether I use the word ‘horse’ or ‘steed’ […] The assertive force does not extend over that in which these words differ […] Much in language serves to aid the hearer’s understanding, for instance emphasizing part of the sentence by stress or word-order. Here let us bear in mind words like ‘still’ and ‘already’. Somebody using the sentence ‘Alfred has still not come’ [Alfred ist noch nicht gekommen] actually says ‘Alfred has not come’ and, at the same time hints [andeutet]—but only hints—that Alfred’s arrival is expected. Nobody can say: Since Alfred’s arrival is not expected, the sense of the sentence is therefore false. The way that ‘but’ differs from ‘and’ in that we use it to intimate [andeuten] that what follows it contrasts with what was to be expected from what preceded it. Such conversational suggestions make no difference to the thought. A sentence can be transformed by changing the verb from active to passive and at the same time making the accusative into the subject. In the same way we may change the dative into the nominative and at the same time replace ‘give’ by ‘receive’. Naturally such transformations are not indifferent in every respect but they do not touch the thought, they do not touch what is true or false […] It is just as important to ignore distinctions that do not touch the heart of the matter, as to make distinctions which concern essentials. But what is essential depends on one’s purpose. To a mind concerned with the beauties of language, what is trivial to the logician may seem to be just what is important. (Frege 1918: 331)

The thought in Frege’s theory of meaning corresponds to the notion of what is said in (neo-)Gricean theory;1 in current terminology the key claim is that the phenomena under discussion here do not affect the determination of what is said. In stating Alfred still has not come, I say that he has not come, while “hinting” that his arrival is expected; p but q differs from p and q in “intimating” a sense of contrast. These two verbs in Geach’s rendering—hint and intimate—both translate Frege’s andeuten, a verb we can gloss (with Gricean hindsight) as ‘(conventionally) implicate’ and nominalize as Andeutung. While the historical significance of Frege’s remarks in this area is often overlooked (for example, Frege is nowhere mentioned in Chris Potts’s (2005) major monograph on conventional implicature), the Andeutung relation, for a component of linguistic meaning that does not affect propositional content or “touch what is true or false”, is a direct precursor of Grice’s conventional implicature.

1 To be sure, this is not an entirely straightforward matter. The delineation of what is said remains at the heart of scholarly disputes in neo-Gricean and post-Gricean pragmatics; see discussion below and Bach 2001, Saul 2002, Carston 2002, and Horn 2009 for elaboration and references.

154

laurence r. horn

Among Frege’s intriguing, if largely overlooked, commentaries in this area is a passage in the unpublished 1897 Logic (Frege 1897/1979: 242) touching on the division of labor between semantics and discourse pragmatics. After noting that the addition of particles like ach ‘ah’ and leider ‘unfortunately’ or the replacement of Hund ‘dog’ with Köter ‘cur’ “makes no difference to the thought”, Frege continues: The distinction between the active and passive voice belongs here too. The sentences ‘M gave document A to N’, ‘Document A was given to N by M’, ‘N received document A from M’ express exactly the same thought; we learn not a whit more or less from any of these sentences that we do from the others. Hence it is impossible that one of them should be true whilst another is false. It is the very same thing that is here capable of being true or false. For all this we are not in a position to say that it is a matter of complete indifference which of these sentences we use […] If someone asks ‘Why has A been arrested?’ it would be unnatural to reply ‘B has been murdered by him’, because it would require a needless switch of the attention from A to B. Although in actual speech it can certainly be very important where the attention is directed and where the stress falls, it is of no concern to logic. (Frege 1897/1979: 242)

At issue here are those alternations that involve the packaging of content motivated by considerations of information structure, or what the Prague school linguists would later dub functional sentence perspective: the choice between members of (quantifier-free) active/passive pairs or alternants involving indirect converses (Cruse 1986: § 10.7) like give and receive, motivated in each case by the goal of mapping topics into subject position. Frege’s pairs are allosentences (Lambrecht 1994), in which differences in syntactic or prosodic form correlate with differences in information packaging while the information itself (the propositional content) remains constant. Frege also presciently foreshadows the modular approach to meaning in natural language: render unto logic what is relevant to sense, reference, and truth conditions, render unto pragmatics what is relevant to usage so long as it concerns distinctions without a difference to the thought or propositional content. The examples motivating Frege’s appeal to subsense relations, which we provisionally denote as Andeutungen (‘suggestions, hints’), thus extend to the range of cases as given in Table 1 on the facing page.

expressive meaning, free datives, and f-implicature

155

Table 1: Frege’s “Andeutungen” (proto-conventional implicatures) Expressions [vs. unmarked alternative] Alfred has not yet come [vs. Alfred has not come] B received C from A, B was given C by A [vs. A gave C to B] B was murdered by A [vs. A murdered B] A murdered B [vs. A murdered B] p but q [vs. p and q] p although q [vs. p and q] ah, p; (un)fortunately p [vs. p simpliciter] The cur [vs. The dog] howled the whole night The steed [vs. The horse] raced around the track

Andeutung Alfred’s coming is expected B is the topic of discussion B is the topic of discussion B is the topic of discussion there is a contrast between p, q p is surprising, given q S has relevant attitude toward p negative evaluation of referent positive evaluation of referent

The legacy of Fregean subsense relations has been anything but straightforward. One reason is that Frege was more concerned in collecting such phenomena for the purpose of setting them aside rather than in any attempt to characterize them positively. This has not prevented analysts from tossing Frege’s examples of the kind in Table 1 into a mixed bag labeled, from Dummett on, as tone. For Dummett (1973: 2–3, 83–89), tone—conflating the Fregean notions of Färbung ‘coloring’ and Beleuchtung ‘illumination’—is problematic because of its (allegedly) inherent subjectivity, characterized in terms of ideas (Vorstellungen) or mental images: Frege makes a poor explanation worse by suggesting that mental images are incommunicable in principle: no two people can ever know that they have the same mental image. It would follow that tone was a feature of meaning which was, in principle, subjective. This conclusion is a simple contradiction. Meaning, under any theory, cannot be in principle subjective […] Tone is not, however, in itself any more subjective than sense. (Dummett 1973: 85)

Dummett’s conclusion that Frege’s characterization of subsense relations is incoherent due to this subjectivity has become received wisdom; cf. e.g. Neale (1999, 2001). As Neale (1999) puts it, “Dummett shows decisively that Frege’s positive position on colouring is untenable, so I will spend no more time on it”. But “Frege’s positive position” is nothing Frege would have recognized, much less endorsed. When discussing the examples that have been most closely scrutinized, including the but vs. and distinction, the contribution of noch in noch nicht, or the non-referential meaning associated with cur, Frege writes not of any Vorstellungen communicated here but rather of what the speaker andeutet (‘suggests, hints’). In the passages from the Begriffsschrift or those from the Logik (Frege 1897/1979: 152) two decades later, the speaker gibt einen Wink (‘hints’) that some contrast exists between

156

laurence r. horn

the clauses that aber conjoins or that the Köter is not a highly regarded canine. Crucially for Frege (as for Grice), this is a distinction without a truthconditional difference; the thought would be the same whichever lexical item is chosen. But while the two terms in each case are not interchangeable, a hint or suggestion is not ipso facto a subjective “idea”. Since Frege alternates between the two terms andeutet and gibt einen Wink to identify a non-truth-conditional meaning that a speaker suggests, hints, or intimates (prefiguring conventional implicature), while never considering the kind of data for which Grice develops the notion of conversational implicature, I shall dub this subsense ingredient of conventional content f-implicature. While conceding that “a hint is evidently not the production of a mental image”, Dummett still appears to mischaracterize Fregean doctrine (to the extent that it is a doctrine) of subsense relations. Dummett (1973: 88) sees Fregean “tone” as more plausibly applying to cases of “expressive meaning”, whereas the choice of but as against and “does not serve to convey any attitude on the part of the speaker, in the sense in which a speaker may evince, e.g., a respectful, apologetic or regretful attitude”. This would be an easier claim to evaluate if we knew just what sorts of “expressive” examples Dummett has in mind, but it is certainly true that current work on conventional implicature has focused largely, although not exclusively, on expressive constructions (see e.g. Davis 2009; Kim & Sells 2007; McCready 2004; Potts 2005, 2007c; Potts & Kawahara 2004). 3. Frege, Grice, and the Contribution of But Grice’s first differentiation between conventional implicature and the relations on either side of it (conversational implicature, semantic presupposition) is presented (without labels) in the course of his defense of the causal theory of perception (Grice 1961: 126–132). Grice’s analysis of the relation of (1c) to (1a) includes the claims in (2). (1)

a. She was poor but she was honest. b. She was poor and she was honest. c. There is some contrast between poverty and honesty, or between her poverty and her honesty.

(2)

– the truth status of (1c) is irrelevant to the truth conditions of (1a), as distinct from the case of (semantic) presupposition: “Even if the implied proposition were false, i.e. if there were no reason in the world to contrast poverty with honesty either in general or in her case, the original statement could still be false [rather than devoid of truth value];

expressive meaning, free datives, and f-implicature

– – – –

157

[…] if for example she were rich and dishonest”. Of course (1a) can also be true on Grice’s account if (1b) is true and (1c) false. what is said in (1a) does not semantically imply (1c), whence the anomaly of #If she was poor but honest, then there is some contrast between (her) poverty and honesty. the implication of (1c) is detachable, i.e. removable by replacement of but with and as in (1b). the implication of (1c) is non-cancellable (#She is poor but she is honest, but I do not mean to suggest there is any contrast between poverty and honesty.)2 the inference from (1a) to (1c) is “a matter of the meaning of the word ‘but’ ” (unlike the inferences drawn in the relevant contexts from Jones has beautiful handwriting and his English is grammatical or My wife is either in the kitchen or in the bedroom)

Six years later, when Grice affixes the conventional implicature label, but again has pride of place, along with therefore (Grice 1989: 25); apparently without recognizing it, Grice thus traces the footsteps of Frege, for whom the contrast suggested or hinted at by adversatives is the primary instance of Andeutung or F-implicature. Expressions falling under this analysis represent a recalcitrant residue for Grice (who was concerned with delineating what is said and what is conversationally, and hence calculably, implicated) as they did for Frege (who was concerned with the thought, i.e. with sense and potential reference); for both, detecting an F-implicature facilitates the real work of semantics by clearing away the brush. But Grice also situates this relation within a map of what we refer to (though he does not) as the semantics/pragmatics divide. If not always accepted, his effect on the landscape is widely recognized, as here by Davidson: It does not seem plausible that there is a strict rule fixing the occasions on which we should attach significance to the order in which conjoined sentences appear in a conjunction: the difference between ‘They got married and had a child’ and ‘They had a child and got married.’ Interpreters certainly can make these distinctions. But […] much that they can do should not count as part of their linguistic competence. The contrast in which is meant or implied by the use of ‘but’ instead of ‘and’ seems to me another matter, since no amount of common sense unaccompanied by linguistic lore would enable an interpreter to figure it out. Paul Grice has done more than anyone else to bring these problems to our attention and help to sort them out. (Davidson 1986: 161–162)

2 Note that even metalinguistic negation is ruled out here: #She isn’t poor but honest, she’s poor and honest. Rejection of the implicature requires a more explicitly metalinguistic device, as noted in the text below.

158

laurence r. horn

But how, exactly, does the sorting work? If descriptive content, reflecting what is said, falls within semantics and if what is conversationally implicated (e.g. the ‘for all I know, not both p and q’ upper-bounding implicatum associated with the utterance of the disjunction “p or q” or the negative effect of the Gricean letter of recommendation) falls within pragmatics, where does conventional implicature fall? One standard view—impossible to confirm directly, since Grice never refers to pragmatics as such3—is that by falling outside what is said, the conventionally implicated is pragmatic (see e.g. Gutzmann 2008: 59). One argument on this side is terminological; in Kaplan’s words, According to Grice’s quite plausible analysis of such logical particles as “but”, “nevertheless”, “although”, and “in spite of the fact”, they all have the same descriptive content as “and” and differ only in expressive content […]. The arguments I will present are meant to show that even accepting Grice’s analysis, the logic is affected by the choice of particle, as it should be on my view of logical validity as the preservation of truth-plus rather than (merely) descriptive truth. If this is correct, then generations of logic teachers, including myself, have been misleading the youth. Grice sides with the logic teachers, and though he regards the expressive content as conventional and hence (I would say) semantic (as opposed to being a consequence of his conversational maxims), he categorizes it with the maxim-generated implicatures. (Kaplan 1999: 20–21)

To be sure, conventional implicatures are implicatures. But then again, they are conventional; we are indeed dealing here, unlike in the maxim-based cases, with aspects of content. In essence, the question comes down to which of the two diagrams in Figure 1 we take to more faithfully represent Grice’s thinking. What U[tterer] meant what U said

what U implicated

conventionally

non-conventionally (e.g. conversationally)

3 Of course, Grice could have referred to pragmatics. We need to be even warier of similar representations about Frege, as when Kaplan (1999: fn. 12) claims Frege would have said that epithets “do not contribute to cognitive content and thus the study of their use belongs not to semantics but to pragmatics”. Given that the semantics/pragmatics distinction postdated Frege by several decades, we cannot be certain just how Frege would have classified his curs and nags, much less his Boches.

expressive meaning, free datives, and f-implicature

159

What U[tterer] meant conventionally what U said

non-conventionally (e.g. conversationally)

what U conventionally implicated

Figure 1: The meaning tree: two versions

While most exegesis (including Horn 1989: 146 as well as Kaplan above) focuses on the unity of the implicature category and thus opts for the first tree, it is arguably the second tree, adapted from Neale 1992: 523, that more accurately reflects what is said by Grice. Two decades after the William James lectures, Grice revisits the status of these categories in his Retrospective Epilogue (“Strand Five”, Grice 1989: 359–365), where he differentiates central and non-central modes of meaning by invoking the two criteria of formality (“whether or not the relevant signification is part of the conventional meaning of the signifying expression”) and dictiveness (“whether or not the relevant signification is part of what the signifying expression says”). If, for example, a speaker says “p; on the other hand, q” in the absence of any intended contrast of any kind between p and q, one would be inclined to say that a condition conventionally signified by the presence of the phrase “on the other hand” was in fact not realized and so that the speaker had done violence to the conventional meaning of, indeed had misused, the phrase “on the other hand.” But the nonrealization of this condition would also be regarded as insufficient to falsify the speaker’s statement. (Grice 1989: 361)

It is formality without dictiveness that yields conventional implicature, although the latter term itself appears nowhere in the Epilogue. If conventional (or F-)implicature for adversatives like but and on the other hand is a matter of non-truth-conditionally projecting but encoded content (cf. Barker 2003), just what is this content? Dummett (1973), like Bach decades later, points out that the implicature contributed by but cannot involve simple “contrast” between the clauses it connects: Frege’s account of ‘but’ is incorrect: the word is indeed used to hint at the presence of some contrast; but not necessarily one between what the second half of the sentence asserts, and what you would expect, knowing the first half to be true […]. If a club committee is discussing what speakers to invite, and someone says, ‘Robinson always draws large audiences’, a reply might be ‘He always draws large audiences, but he is in America for a year’; the objector is not suggesting that a popular speaker is unlikely to go to America, but that, while Robinson’s popularity as a speaker is a reason for inviting him, his being

160

laurence r. horn in America is a strong reason against doing so. The word ‘but’ is used to hint that there is some contrast, relevant to the context, between the two halves of the sentence: no more can be said, in general, about what kind of contrast is hinted at. (Dummett 1973: 86)

While Dummett (and Bach) find Frege (and Grice) too restrictive in spelling out what p but q adds to p and q, others have proposed alternative treatments, invoking not contrast—even context-dependent contrast—but surprise or denial of expectation, i.e. the unexpectedness of q given p. Relevance theorists have argued that but and related discourse connectives involve procedural rather than conceptual meaning (Blakemore 2002; Hall 2007), but others have questioned this distinction or its application (Mosegaard Hansen 2003; Pons Bordería 2008; Rieber 1997). But neither contrast nor unexpectedness in themselves account for the role played by but clauses in argumentation (Anscombre & Ducrot 1983; cf. also Merin 1999: 203–209 for a formalization within decision-theoretic semantics): for a speaker uttering p but q, p constitutes an argument for r, while q constitutes a stronger argument for ¬r. Dummett’s first committee member presents a reason for inviting Robinson; the objector, while conceding this point, offers a stronger reason for not inviting him. The lack of any intrinsic contrast between the two clauses connected by but is underlined by the possibility of p, but q too or the frame illustrated in the sampler in (3); here and below, the diacritic 𝛾 marks examples obtained by Googling. (3)

a. b. c. d.

u� War is hell, but so is withdrawal. u� Breast cancer is unfair, but so is life. u� Divorce is hell, but so is a bad marriage. u� Boyfriend is visiting soon, but so is my period!!?

An argumentation-sensitive account of the non-truth-conditional contribution of but also makes sense of an otherwise puzzling asymmetry in the interpretation of p but q clauses; the perspectives urged toward the subject in (4a) and (4b) are very different indeed. (4)

a. He is rich but he is dissolute. b. He is dissolute but he is rich.

(Cf. Blakemore 2002: 103 and Vallée 2008: 422 on the asymmetry of but.) So just what is the contribution of but and why is it so hard to pin down?4 This difficulty is widely recognized. Blakemore (2002: 53–54), for example

4 There are occurrences of but that conform more naturally to a contrast analysis than a denial-of-expectation or argumentative approach, as in (i), from a New York Times article

expressive meaning, free datives, and f-implicature

161

cites the “elusive quality of but”, noting elsewhere that other discourse particles like well are also “frustratingly elusive” (Blakemore 2002: 128); cf. Hall (2007: 155–156) for similar observations. More generally, as we shall observe, this “descriptive ineffability” (to adopt the term from Potts 2007) is a trait symptomatic of conventional implicatures in general. (But see Gutzmann 2013 [this volume] for some reservations.) 4. Conventional Implicature and Its Discontents For Grice (1989), a conventional implicature C associated with an expression E manifests the following two definitional properties: – by virtue of being conventional, C constitutes a non-cancelable aspect of the meaning of E – by virtue of being an implicature, C’s truth or falsity has no effect on the truth conditions of E. As we have seen, Frege’s comments on (what Dummett calls) “tone” have received mixed reviews at best, in part because Frege collected such phenomena largely for the purpose of setting them aside rather than in any attempt to characterize them positively. In this respect, too, he foreshadows Grice, whose goal in (briefly) delineating the category of conventional implicature was more to say what token exemplars are not—viz., part of what is said or of what can be rationally calculated—than to say what they are. The first attempt to formalize Grice’s notion is Karttunen & Peters’s (1979) proposal to fold the better established—but, they argue, inconsistently defined—relation of presupposition into conventional implicature, for which they offer a multidimensional non-truth-conditional treatment of even and related particles within an extended version of Montague Grammar. On the their analysis, (5a) entails (and indeed is logically equivalent to) (5b) while conventionally implicating both (5c) and (5d) (or the conjunction of the two).

“In North Carolina, Lawsuit Is Threatened Over Councilman’s Lack of Belief in God” (Dec. 13, 2009): (i)

City Councilman Cecil Bothwell of Ashville believes in ending the death penalty, conserving water and reforming government, but he does not believe in God. http://www.nytimes.com/2009/12/13/us/13northcarolina.html

162 (5)

laurence r. horn a. Even [Bill]F likes Mary. b. Bill likes Mary. c. Existential implicature: There are other x under consideration besides Bill such that x likes Mary. d. Scalar implicature: For all x under consideration besides Bill, the likelihood that x likes Mary is greater than the likelihood that Bill likes Mary.5

For Bach (1999), on the other hand, the putative instances of conventional implicature nominated by Grice or Karttunen & Peters—or by Frege under other labels—involve either secondary aspects of what is said (as with even, but, although, still, and similar particles) or higher-level speech acts (as with adverbial modifiers like frankly or to tell the truth). Bach’s principal argument against the Fregeo-Gricean line on but centers on the behavior of “ACIDs” (alleged conventional implicature devices) in embedded contexts. Without rehearsing the issues here, I note that Bach’s arguments based on his “IQ [indirect quotation] test”— An element of a sentence contributes to what is said in an utterance of that sentence if and only if there can be an accurate and complete indirect quotation of the utterance (in the same language) which includes that element, or a corresponding element, in the ‘that’-clause that specifies what is said. (Bach 1999: 340)

—have been challenged on a number of fronts by, among others, Carston (2002: §2.5), Blakemore (2002: 56–58), Barker (2003), Hall (2007), and Greenhall (2007).6 In addition to his arguably myth-conceived dismissal of conventional implicatures, Bach (1999) also objects to the particular analysis of but sketched by Frege and Grice, and in this he (like his precursor Dummett) is quite correct, as we have noted. Similarly, Grice’s invocation of conventional implicature for the analysis of therefore— In some cases the conventional meaning of the words used will determine what is implicated, besides what is said. If I say (smugly), He is an Englishman; he is, therefore, brave, I have certainly committed myself, by virtue of the

5 As with but, the content of the implicature(s) associated with even have been matters of contention for some time; cf. Kay 1990, Horn 1992, Francescotti 1995, Rieber 1997, and Schwenter 2002 for some relevant considerations. 6 One germane objection raised by Barker (2003: fn. 21) is that Bach, in applying the IQ test to ACIDs embedded under said that, “assumes that the ‘said-content’ locution as used in ordinary language tracks the theoretical notion of said-content, but there are no cogent reasons for holding this.”

expressive meaning, free datives, and f-implicature

163

meaning of my words, to its being the case that his being brave is a consequence of (follows from) his being an Englishman. But while I have said that he is an Englishman, and said that he is brave, I do not want to say that I have said (in the favored sense) that it follows from his being an Englishman that he is brave, though I have certainly indicated, and so implicated, that this is so. I do not want to say that my utterance of this sentence would be, strictly speaking, false should the consequence in question fail to hold. (Grice 1989: 25–26)

—has been challenged on empirical grounds by Bach (1999), Neale (1999, 2001), and Predelli (2003). Defending this analysis is certainly a more difficult task in the case of therefore than it is for its fellow candidates but and after all (along with several of Frege’s nominees). For Grice, while q because p says that the truth of q follows from that of p, p therefore q only implicates this connection, but this judgment is not widely shared. This is likely a point on which speakers differ; it is certainly possible that the status of aspects of word meaning could shift from beyond to within the umbrella of what is said (or what Potts calls “at-issue” meaning), as indeed recognized by Frege: Of course borderline cases can arise because language changes. Something that was not originally employed as a means of expressing a thought may eventually come to do this because it has constantly been used in cases of the same kind. A thought which to begin with was only suggested by an expression may come to be explicitly asserted by it. (Frege 1897/1979: 241)

It is plausible that this kind of shift, exemplified in the well-known euphemism treadmill, could be responsible for the instability of implicated vs. “at-issue” meanings of particles and other expressions. Bach’s revised analysis of but and therefore proceeds from his rejection of the “OSOP” (One-Sentence-One-Proposition) assumption and his consequent advocacy of multi-dimensional semantic analyses. For a (paleoor neo-) Fregean or Gricean addressing the status of connectives like but, the relevant question here is not whether two or more propositions may be expressed (directly or indirectly) by the utterance of a given sentence, but what the consequences for the truth conditions of that sentence are when one of these propositions is false. For Bach, both the conjunction and the (context-dependent) contrast between the conjuncts qualify as components of what is said in p but q; our reluctance to brand (1a) false if just the (1b) component fails to hold derives not from (1b)’s status as (merely) implicated but from the fact that it is not the primary assertion.7 7 Other critics of the Gricean notion include Stanley (2002), whose constraint on the relation between what is said to what is communicated, his “Expression-Communication

164

laurence r. horn

In his book-length treatment of the relation, Potts (2005) follows Bach in arguing for the asserted or, in his terminology, “at-issue” status of the but/even class but at the same time he rebrands Grice’s conventional implicature category for his own class of “CIs”, encompassing (what he takes to be) the non-at-issue contributions of expressives (that jerk), supplements (e.g. non-restrictive relatives and appositives), epithets, and honorifics. But while the notion of secondary assertion, invoked in one form or another by both Bach (1999) and Potts (2005), is a useful construct, it appears to be a better fit for some of the “CI” cases in Potts’s catalogue than it is for the original range of phenomena falling under F-implicature. Thus, consider the case of appositive or non-restrictive relative (NRR) clauses as in Frege’s (1892: 38) celebrated example in (6). (6)

a. Napoleon, who recognized the danger to his right flank, himself led his guards against the enemy position. b. Napoleon recognized the danger to his right flank. c. Napoleon himself led his guards against the enemy position.

Such cases, pace Potts 2005, do involve two asserted propositions, although the assertion in the relative clause (6b) is backgrounded or secondary. Frege maintains that the falsity of either (6b) or (6c) results in the falsity of (6a), although his intuition is not universally shared. In fact, whether we are inclined to judge a sentence false on the basis of the falsity of its contained NRR clause depends on our assessment of the significance of that information in that clause for the overall claim. Thus, New Haven, which is the capital of Connecticut, is the home of Yale University might count as (misleading but) true as an answer to a question about the location of various Ivy League colleges but false as an answer to a question about which state capitals host Ivy League institutions, given that Connecticut’s capital is Hartford. A strong argument for Frege’s dual-assertion analysis for such cases is the insertability of an overt performative within the NRR, which indicates that such clauses are part of at-issue (even if backgrounded) rather than merely implicated (or presupposed) content. (7)

a. The bank bailout bill, which I hereby pledge to support, is immoral and unconscionable. b. The qualifications of this woman, whom I hereby nominate, are unquestionable.

Principle”, is inconsistent with conventional implicature, and Iten (2005), who challenges the Fregean-Gricean approach from the Relevance-theoretic perspective. (See also the Carston 2002 and Blakemore 2002 references on procedural meaning cited above.)

expressive meaning, free datives, and f-implicature

165

Similar results hold for the not only p but (also) q construction (Horn 2000), where p and q are both at issue in Potts’s sense but q is foregrounded with respect to p, whether or not q in fact entails p. Once again, evidence from the distribution of overt performatives indicates that both p and q are put forward as assertions, promises, etc., as seen in the attested examples in (8), due respectively to Oprah Winfrey and http://www.a-weddingday.com. (8)

a. b.

u� Not only do I hereby retract my claim, but I also hereby apologize to the

cattlemen in the great state of Texas. u� Not only do I promise to be a good and faithful husband to you, but also to be a patient, loving father to [children’s names].

But does the pattern seen in such clausal constructions extend to still, even, and similar particles? It’s hard to find any examples that would support the argument. A more direct objection to the dual-said-content analysis of such cases is that the falsity of the “backgrounded” content (relating to contrast, unexpectedness, etc.) may render the host sentence inappropriate but not false (cf. Barker 2003). In these cases—contra Bach (1999) and Potts (2005)—the traditional Fregeo-Gricean line, distinguishing what is F-implicated from what is said remains eminently solid.8 Nor are these the only expressions that lend themselves to this approach, as exemplified by the case of epithets and other expressives, where the Potts (2007) approach is consistent with the Fregean tradition. Williamson (2003, 2009), in his spirited defense of Frege’s account of pejoratives (e.g. cur, Boche) in terms of (proto-)conventional implicature, insightfully addresses a potential objection to an F-implicature-based analysis of pejoratives: How can we reconcile the intuition that someone might

8 The argument here does not automatically extend to nominal appositives, which do appear to be robustly non-at-issue. I cannot discuss Potts’s treatment of parenthetical supplements as CIs here except to note that supplements do not map onto simple embedded structures. While Potts (2005: 92) takes (i) and (ii)

(i) (ii)

Max, it seems, is a Martian. It seems that Max is a Martian.

to be equivalent, this cannot be the case, since “but he isn’t really” can only be felicitously appended in the latter case. Similarly, compare (iii) and (iv) (assuming coreference): (i) (ii)

It seems that Max is older than he (really) is. #Max, it seems, is older than he (really) is.

See Amaral et al. (2007) for other problems with Potts 2005, and Harris & Potts (2009) for a response, focusing on the controversial speaker-oriented criterion for CIs.

166

laurence r. horn

assent to (9a) and reject (9b) with the claim that the difference between dog and cur is merely a matter of Fregean “tone” and not sense? (9)

a. Every dog is a dog. b. Every dog is a cur.

Williamson (2003) responds:9 Such problems undermine Frege’s simple account of propositional attitude ascriptions, not his claim that pairs like ‘cur’ and ‘dog’ have the same truthconditional meaning. For similar problems arise even for pairs of synonyms with the same tone. Kripke (1979) gives the example of the synonymous natural kind terms ‘furze’ and ‘gorse’. A speaker might […] acquire normal competence with both without being sure that they refer to exactly the same kind of plant. Thus ‘He believes that all furze is furze’ and ‘He believes that all furze is gorse’ appear to differ in truth-value. It does not follow that ‘furze’ and ‘gorse’ are not synonyms after all. Such problems show nothing special about pejoratives. Whatever the right account of propositional attitude ascriptions, it is compatible with the Fregean view that ‘cur’ and ‘dog’ differ in tone but not sense (truth-conditional meaning). (Williamson 2003: 262)

Williamson’s line on epithets is challenged by Hom (2008), but this critique is based on a misunderstanding, as shown by Hom’s claim that on the conventional implicature account of ethnic slurs “the derogatory content is merely implicated and not part of what the sentence literally says (i.e. derogatory content is not part of the semantic content of the sentence)” (Hom 2008: 423).10 But semantic content is not identical to truth-conditional

9 As the editors, Daniel Gutzmann and Hans-Martin Gärtner, point out, cur may shift to incorporate the negative ascription (= ‘mongrel’ or ‘obnoxious dog’) into its truth-conditional content. This shift, an instance of the reanalysis envisaged by Frege in the passage cited above, may be coerced in contexts like “I like (most) dogs, but I despise curs.” Similar remarks apply to human-referring epithets. 10 While Hom (2008: 423) also finds “pragmatic minimalists” like Williamson guilty of holding epithets to be entirely synonymous with their nonpejorative correlates, no theorist who distinguishes what is meant from what is said (allowing for aspects of encoded meaning in the former category) actually regards dog and cur, or (to take Hom’s example) Chinese and Chink, as synonymous pairs. The same objection applies to Ludlow’s (2012) defense of the claim that the “differences in application” between a and the do not preclude their sharing the same “semantical content”:

[M]any synonyms [sic] customarily are put to different uses. Take, for example, Grice (1961, 1975) on the distinction between ‘but’ and ‘and’. On Grice’s account ‘but’ and ‘and’ literally mean the same thing, but different “conventional implicatures” are associated with them; ‘but’ implicates a sense of contrast between the conjuncts. (Ludlow 2012: 386) As argued in Horn & Abbott 2012, if there is no semantic (encoded) difference between

expressive meaning, free datives, and f-implicature

167

content. Conventional implicatures constitute part of encoded content but not part of truth-conditional content (differing in this respect, as argued above, from appositives and non-restrictive relatives): The conventional implicature possessed by a sentence S is not part of its force, but is a part of S’s semantic content […] Nevertheless, S’s implicature makes no contribution to S’s truth-conditions. (S. Barker 2003: 3)

Beyond discourse particles (but, even, Ger. ja, doch, Jap. yoku(mo)), there are more candidates for F-implicature status than were dreamt of in Frege’s or Grice’s philosophy, ranging from second person pronouns in most Romance, Germanic, and Slavic languages, the full domain of expressive language (from slurs and epithets to honorifics and intensifiers), evidentiality markers bearing on the source of information or the strength of speaker’s commitment (cf. Aikhenvald 2004; Davis et al. 2007; Faller 2012; Matthewson et al. 2007), and the uniqueness or maximality condition on definites (Horn & Abbott 2012). One of the strongest cases for F-implicatures involves second person “T/V” pronoun choice, illustrated by the distinction between tu and vous in French or du and Sie in German. More clearly than but (and certainly therefore), the T/V distinction does not involve what is said, “makes no difference to the thought”, and leaves truth conditions unaffected, but it does involve conventional, non-calculable aspects of meaning.11 Notice that even if we accepted the verdict of Bach’s IQ tests for adversative connectives like but, still, and even, the contribution of the familiar vs. formal pronouns fails these tests and thus cannot constitute part of what is said: if I am on formal terms with you, I cannot remind you that #Votre mère a dit que tu dois lui donner un coup de téléphone (‘Your (formal) mother said that you (fam.) should give her a phone call’) even if what your mother said to you

indefinites and definites, there is nothing from which to derive their putative pragmatic differences. 11 Levinson (1983: 128–129) suggests, but does not pursue in detail, a conventional implicature-based analysis of T/V pronoun choice. Note, however, that a speaker’s move from the T to the V form, or vice versa, will often generate conversational implicatures signaling an increase in intimacy, alienation, contempt, etc.; it is the static if complex values of the pronouns of power and solidarity (Brown & Gilman 1960; Mühlhäusler & Harré 1990; Zimmermann 1997; Taavitsainen & Jucker 2003) that lend themselves to an F-implicature treatment. See Keenan 1971 for an earlier treatment of T/V pronouns in terms of pragmatic presuppositions, Kaplan 1999 and Greenhall 2007 for use-conditional approaches, Potts 2007: § 4 for a CI analysis, and Sauerland 2007 for a critique of unified accounts of pronouns and epithets based on their differential behavior with respect to perspective shifting in indirect discourse.

168

laurence r. horn

earlier was indeed Tu dois me donner un coup de téléphone. In Potts’s terms, the T/V choice is very much “speaker-oriented” in the fashion of canonical CIs.12 The choice between Tu as tort and Vous avez tort (‘You’re wrong’) is based on the speaker’s assessment of the social context, not truth conditions. Similar remarks could be entertained for natural gender, such as the choice between he and she in English or analogous gender distinctions in other languages, which have been treated in presuppositional terms (Büring 2005; Cooper 1983). Thus in affirming “Tu es soûl”, my belief that a certain social relationship obtains between us and that you are male is not part of the thought or of what is said; both propositions are indeed communicated, but what is said is simply that you’re drunk. If this is right, then what is said in all of the variants in (10) is identical. (10) a. b. c. d.

Tu es soûl. Tu es soûle. Vous êtes soûl. Vous êtes soûle.

‘You (sg., masc., familiar) are drunk’ ‘You (sg., fem., familiar) are drunk’ ‘You (sg., masc., formal) are drunk’ ‘You (sg., fem., formal) are drunk’

Kaplan (1999) has advocated extended notions of “validity-plus” and “truthplus” within the “semantics of use” to deal with the expressive aspects of meaning contributed by hypocoristics, ethnic slurs, epithets, and interjections, as well as second person pronouns and other honorifics,13 but his observations are in fact consistent with the spirit of the Fregean/Gricean approach; cf. Gutzmann 2008 for a use-conditional analysis of German modal particles (ja, doch, wohl, halt, et al.) that bridges the gap between Fimplicatures and Kaplan’s semantics of use. In this connection it may be worth responding to one attempt to refute the view of the T/V distinction through the lens of conventional implicature. 12 Such pronouns, in indexing both a second-person referent and an appropriateness condition, seem to cast doubt on Potts’s (2005: 7) generalization (2.5) that “No lexical item contributes both an at-issue and a CI meaning”, but—as Chris Potts reminds me—pronouns are feature complexes that do not constitute “lexical items” in the relevant sense. Another problem for generalization (2.5), as noted in Horn 2007: fn. 9 and Williamson 2009: fn. 16, is posed by expressives that take the form not of adjectival modifiers or appositives (“that damn guy”, “that asshole Joe”) as in Potts 2005 but of nominal epithets per se (“I wouldn’t hire that scoundrel”, “Why did you vote for that asshole?”). In these cases, a lexical item does “contribute both an at-issue and a CI meaning”—no problem, of course, for the echt Fregean/Gricean view. More recently, McCready (2009) proposes a “mixed content” line on numeral classifiers in Japanese, on which the truth-conditional contribution of the numeral combines with the CI meaning of the nominal classifier. 13 Analyses of honorifics and other expressives in terms of Pottsian CIs are given in Potts & Kawahara 2004, McCready 2004, 2010, Potts 2007, and Kim & Sells 2007.

expressive meaning, free datives, and f-implicature

169

Tsohatzidis (1992: 572) observes that while (11a) could be true (if I do always use the V form in speaking to Paul), its counterpart in (11b) is necessarily false, indeed self-falsifying. The same facts hold in reverse for (12a), (12b); it is an illocutionary contradiction for me to use the second-person singular to Paul while denying I would ever do so. (11)

a. Je vous parle toujours au pluriel, mon cher Paul. b. Je te parle toujours au pluriel, mon cher Paul.

(12) a. Je ne vous parle jamais au singulier, mon cher Paul b. Je ne te parle jamais au singulier, mon cher Paul.

[◇T] [□F] [◇T] [□F]

Similar considerations apply to the contrast in (13) (= ‘I now begin to use the T form to you’). (13) a. En ce moment, je commence à te tutoyer. b. En ce moment, je commence à vous tutoyer.

[◇T] [□F]

Thus, Tsohatzidis concludes, the T/V distinction must be truth-conditionally relevant, ruling out a conventional-implicature-based analysis. But this argument really hinges not on the truth-conditional contribution of second person pronouns but on the formal role of self-reference here and in analogous cases, e.g. those in (14) and (15), where it would be unwarranted to infer that I and yours truly are referentially distinct or that we must distinguish the contribution to compositional semantics of Bart and Lisa vs. Lisa and Bart.14 (14) a. I never refer to myself in the third person. b. Yours truly never refers to himself in the third person.

[◇T] [□F]

(15) a. I never refer to Bart and Lisa in alphabetical order. b. I never refer to Lisa and Bart in alphabetical order.

[□F] [◇T]

Thus, pace Tsohatzidis, self-referential or metalinguistic minimal pairs of the kind seen in (11)–(13) do not disprove the thesis that the T/V distinction is truth-conditionally irrelevant. (Arguably, other considerations are involved here; see Zimmermann 1997 for elaboration.) Summarizing where we stand at this point, I have supported the Fregean approach positing non-truth-conditional “F-implicatures” for a wide range of cases, including particles like but and even where Potts (following Bach) diagnoses the relevant backgrounded contributions of meaning as second14 As the editors point out, the metalinguistic nature of examples like (11)–(13) is further supported by the non-content-driven “contradiction” of Je te parle toujours en anglais, mon cher Paul.

170

laurence r. horn

ary aspects of what is said. I have maintained that even if we grant the two dimensions of meaning posited by Bach and Potts (and indeed by Karttunen & Peters), we have no compelling evidence against the FregeoGricean position and for the claim that the second dimension (contrast, unexpectedness, etc.) affects the truth conditions or “thought” of the containing sentence. On the other hand, a Pottsian CI analysis of “supplements” like the non-restrictive relative clauses of Frege’s Napoleon example (6) or the related not only clauses in (8) is less compelling; in this case, we do indeed have evidence for two at-issue clauses, where the failure of either may affect the truth conditions of the sentence as a whole. In a third set of cases involving expressive meaning, including epithets and T/V pronouns, the F-implicature approach and the CI approach elaborated in Potts 2007 and similar work coincide. Before turning to a construction (and its relatives) that provides additional grist for the Fregeo/Gricean mill, let us attempt to depict the landscape of (non-)at-issue meaning. The F-implicatures we have discussed— for but, even, T/V pronouns, epithets, etc.—are distinct from the assertorically inert entailments discussed in particular by Horn (2002, 2009) as exemplified in exclusives (e.g. only), exceptives (e.g. nobody … but) and approximatives (almost, barely). An inert entailment is not asserted but it is entailed, thus constituting a truth condition for the entailing sentence: Only Pat passed is false if nobody passed. An F-implicature, by definition, cannot be a truth condition for the implicating statement.15 Thus we have the (non-)at-issue relations as given in Table 2.

15 Thus, for example, the exclusive proposition (ia) has the inert entailment in (ib) and the ordinary (asserted) entailment in (ic).

(i) a. Only Lee can pass the test. b. Lee can pass that test. c. No one other than Lee can pass that test. Similarly, the approximative in (iia) entails both (iib) and (iic), but only the latter entailment is asserted. (ii) a. Robin almost passed the test. b. Robin didn’t pass the test. c. Robin “came close to” passing. See Horn 2002, 2009 for elaboration, argumentation, and implications.

expressive meaning, free datives, and f-implicature

171

Table 2: (non-)at-issue meaning relations

5. Frege Got Him a New Locus Classicus: The Personal Dative and F-Implicature A shibboleth of Southern and Appalachian non-standard U.S. English is the optional appearence of a nonsubcategorized “personal dative” pronominal in transitive clauses which obligatorily coindexes the subject and whose semantic contribution is ill-understood. This personal dative (PD) bears suggestive if not always straightforward relations to constructions involving the distribution of morphosyntactically distinct dative marking in such languages as French, German, Warlpiri, Hebrew, and Old English involving what have been variously termed ethical, free, non-lexical, or affected datives. Some of these non-indirect object datives are coreferential with the subject (e.g. Je me prends un petit café) while others (to be surveyed in § 6 below) are non-coreferential (e.g. Ils lui ont tué son oiseau); they typically invite benefactive and malefactive (adversative) understandings respectively. Our focus in this section is the English personal dative, on display in a range of traditional country and mountain ballads and their modern descendants [boldface used here to indicate coreference, not contrast or focus]: (16) a. And now I’ve married me a pretty little wife And I love her dearer than I love my life. (“Rake and Rambling Boy”, trad.) b. I’m gonna buy me a shotgun, just as long as I am tall. (Jimmie Rodgers, “T for Texas”) c. When I was a young girl, I had me a cowboy. (John Prine, “Angel From Montgomery”) d. Now the Union Central’s pulling out and the orchids are in bloom, I’ve only got me one shirt left and it smells of stale perfume. (Bob Dylan, “Up to Me”)

172

laurence r. horn

The ordinary pronominals here contrast minimally with the reflexive in e.g. I’m gonna sit right down and write myself a letter. (The PD counterpart I’m gonna write me a letter would also be possible in the relevant dialect when me is not an indirect object/goal argument.) While first person singular “bound” pronominals predominate, second and third person cases are also possible in the backwoods: (17) a. ∅i Get youi a copper kettle, ∅i get youi a copper coil, Cover with new-made corn mash and never more you’ll toil. (“Copper Kettle”, traditional ballad) b. My daddy he once told me Don’t you love you any man (Dusty Springfield, “We’ll Sing in the Sunshine”) (18) Raised in the woods so’s he knew every tree [proi ] Kilt himi a b’ar when he was only three.16 (“Ballad of Davy Crockett”)

Note that the PDs above co-occur with other well-known instances of Appalachian English features (cf. Wolfram & Schilling-Estes 1998)—the determiner in them jugs, the verb forms lay and kilt, the noun b’ar [= bear], contracted so-[a]s, a-prefixation and “g-dropping” in a-fillin’.17 Moving from lyrics to prose, we find PDs—while still restricted to (informal) register—ranging freely over person, number, and geography; these examples are set in Chicago, New York, Minnesota, and Philadelphia respectively. (19) a. “I’m going to have to hire me a detective just to follow you around.” (1988 Sara Paretsky novel, Blood Shot, p. 191) b. “It’s too bad we don’t have any of those hellebores”, I say. “We could drop them in the Meer and poison us some fish.” (2006 Ayelet Waldman novel, Love and Other Impossible Pursuits, p. 224)

16 Alternatively, the PD may be cataphoric to the subject of the next line, “Davy, Davy Crockett/King of the wild frontier”, rather than linked to a pro subject. Unfamiliarity with the regional markers in the relevant line in the insular 1950s led some hearers to misparse the relevant line as “Killed in a bar when he was only three” despite the pragmatic implausibility of such a reading within the context. 17 In the absence of such features, the PD may strike listeners as anomalous or even patronizing, as seen in the negative reaction by a blogger to the line “She said she married her an architect” in Dan Fogelberg’s 1980 pop classic “Same Old Lang Syne”: “Did she really say ‘I married me an architect?’ Or is Fogelberg, who seems capable of standard usage, the kind of guy who would say, ‘Dag nabbit, she up ‘n’ married her an architect’” (http://whatstherumpus .blogspot.com/2005/12/more-stupid-holiday-songs.html). This same Dagnabbit Effect may have cost John Kerry the 2004 election when he entered a small town grocery in rural

expressive meaning, free datives, and f-implicature

173

c. “If you attend church just to go through the motions, God’d rather you get you a bottle of bourbon and a whore and go to a hotel and have you a good time.” (Uncle Al in Garrison Keillor’s Lake Wobegon Summer 1956 (2001), p. 274) d. I keep logs of illegal huntin’ here on the wildlife preserve. Poachers, hunters—they come by at night, tryin’ to pinch ’em some deer meat. (Ranger to detectives on “Cold Case”, CBS, 28/11/04)

As seen in (20)–(23), PDs are not restricted to argument positions; they occur where true datives (including true reflexive goals) cannot, they fail to saturate the subcategorization frame of ditransitives, and they co-occur with an external subcategorized indirect object: (20) a. I need me a Coke. b. I seen me a mermaid once. (from movie “Hunt for Red October”) (21) a. He bought {himself/him} a new pick-up. b. He needs {*himself/him} just a little more sense. c. What I like is goats. I jus’ like to look at me some goats. (Sroda & Mishoe 1995) (22) a. She fed {*her/herself} some chitlins. b. She gave {*her/herself} a big raise. (vs. She got her a big raise.) (23) a. He’s gonna buy {him/*himself} a pick-up for his son. b. He’s gonna buy (*him) his son a pick-up. c. I need me a little more time for myself.

Semantically, the salient property of PDs is their lack of effect on the truth conditions of the sentences in which they occur. Syntactically, the salient features of PDs are their restriction to appearance as clitics or weak pronouns attaching to the verb (*I want yours truly some grits), their optionality, and their apparent violation of Principle B, their appearance outside subcategorized positions—and their failure to satisfy subcategorization requirements, as in (22). I argue in Horn 2008 that because PDs are not arguments, they are ipso facto not co-arguments with the subject and therefore do not trigger the co-argument version of Principle B advocated by Pollard & Sag (1992) and Reinhart & Reuland (1993).18 Ohio and infamously asked “Can I get me a hunting license here?” (See Horn 2008: §4 for elaboration.) 18 While pronominal PDs are thus allowed, this does not in itself rule out the appearance of reflexives as an alternate option, as the editors point out. As is well known, the usual complementary distribution of pronominals and anaphors does not preclude their overlap in some environments, such as picture noun contexts (She read a story about her(self )) and

174

laurence r. horn

The grammatical and sociolinguistic characteristics of the Personal Dative construction (PDC) are explored in more detail in Christian 1991, Webelhuth & Dannenberg 2006, Conroy 2007, Horn 2008, and Hutchinson & Armstrong to appear; we shall concentrate here on the semantic and pragmatic implications of its distribution. Just what is the semantic contribution of the non-argument pronominal to the clause in which it occurs, given that it does not alter truth conditions? What are the characteristics of predicates that license PDs? First, while many naturalistic occurrences of PDs involve such verbs as get, buy, make, have, want, and need, one relatively recent addition to the set of licensers is worth noting. Since Toni Braxton’s pop song “I Love Me Some Him” (lyrics by SoulShock & Karlin, Andrea Martin, and Gloria Stewart) reached the top of the charts in 1997— I love me some him I’ll never love this way again I love me some you Another man will never do

—the title has generated a snowclone19 of the form I ( Just) Love Me Some X. Note that X here is not semantically quantified (the singer does not adore just an unspecified subpart of her beloved) but is a name, pronoun, generic, etc. that only occurs with an indefinite to satisfy the constraints on the direct object (theme) argument in the PDC. Thus Owens’s declaration below essentially reduces to the observation “I love myself.” (24) The “I love me some” snow clone I LOVE ME SOME ME – slogan popularized by American football player Terrell Owens (a.k.a. “T.O.”) u� I just love me some Jude Law.

– posting on salon.com

My husband used to love him some Jack Daniels. – Halle Berry’s character to Billy Bob Thornton’s in “Monster’s Ball”

non-argument PPs (He pulled a robe around him(self )), and this may apply here as well; see fn. 23 below. 19 See the wikipedia entry at http://en.wikipedia.org/wiki/Snowclone for much more on snowclones. PDs themselves (along with their international cousins; cf. §6 below) have become trendy enough to make it into Language Log; see Mark Liberman’s “On beyond personal datives?” November 2009 post on “I nearly stepped on me a dog” (http://languagelog .ldc.upenn.edu/nll/?p=1863), my guest post (http://languagelog.ldc.upenn.edu/nll/?p=1868), and the (over 80) comments they sparked.

expressive meaning, free datives, and f-implicature

175

I just love me some Zzzzzzzz’s. It’s been nothing but all nighters these past few months. [generic, not existential some!] – blog posting

As in the examples cited earlier, these PDs typically involve positive affect, often conferring a benefit to the subject. Thus compare the minimal pairs in (25) and (26), in which the personal dative is marked or unacceptable for most speakers in the absence of intention. (25) a. He shot him two squirrels. b. #He (got drunk and) shot him two coonhounds (by mistake). (26) a. She caught her a catfish. b. #She caught her a {cold/case of the clap}.

Predictably, the versions in (26b) are fine in the unlikely event that the cold or case of the clap was contracted intentionally. Affect-linked asymmetries in the licensing of PDs are reflected in the data in (27), reflecting google entries surveyed on March 1, 2009. (27) a. I love me some X: 1,020,000 vs. I hate me some X: 23,400 (Yankees, exams, emo) b. She loves her some X: 833 (grapefruit, sparkly dance boys, Ozzy, chocolate, jesus, Halloween, Z cars, kraft dinner) vs. She hates her some: 7* (J. Lo, Mao) c. I want me some X: 650,000 (fonts, Krispy Kremes, candy, monitors, …) d. I saw me some X: 28,400 (relating to entertainment, fun, goal attained, etc.) e. I found me some X: 142,000 (happiness, friends) vs. I lost me some: 7* (clothes) [*of which one is a link to Horn 2008]

When see licenses PDs, it typically alludes to the result of a conscious effort of looking; along the same lines, consider the 2007 Toby Keith song lyric “I’m gonna get my drink on/I’m gonna hear me a sad song” (gratia Will Salmon), in which the sad song is not encountered accidentally but deliberately sought out.20 Even the examples with apparent negative affect are often more positive than it may initially appear; many of the I lost me some X examples occur in the frame I lost me some weight, where the loss is the result of intentional action. 20 Fellbaum (2005: 236) notes that the “agent as beneficiary” construction tends to exclude perception verbs, as in her ??I saw me an accident on the road and ??I hear me some noises in the street, but this exclusion is based on the assumption that the subject lacks the relevant Toby Keith-style intention.

176

laurence r. horn

In other cases, a PD with negative affect is facilitated by local syntagmatic priming, often in a contrastive context. Thus a blog evaluating the movie Serendipity, which featured John Cusack as protagonist and fate and destiny as plot elements, includes the verdict “I love me some John Cusack. I hate me some Fate and Destiny.” Another factor favoring the appearance of PDs is the spontaneous, occasion-specific nature of the utterance, typically signifying the satisfaction of a current intention, need, or desire; this is the nondisplaceability property of CIs noted by Potts 2007c, although others (e.g. Amaral et al. 2007, Gutzmann 2013 [this volume]) have challenged the reliability of this diagnostic. In (28)—the response of Miss South Carolina (the geographically challenged contestant in the 2007 Miss Universe pageant21) to the query “What’s the first thing you’ll do when you get home?”—the PD expressing the speaker’s current dining plans disappears in the reportive follow-up. (28) [I’m gonna] eat me some hamburgers. I haven’t eaten hamburgers in three years.

Moving the PD from the first sentence to the second (“I haven’t eaten me hamburgers in three years”) renders the utterance less likely, even for Miss South Carolina. While many PDs (with get, buy, etc.) directly involve possession, others— in particular with need or want—look forward to a future possession marking the completion or satisfaction of a current modal or propositional attitude, as in (29), from Michael Montgomery’s extensive database. (29) He needs him just a little more sense.

Both need and want are typically analyzed as embedding possession—to need (or want) is to need (or want) to have22—and have is a canonical PD predicate. Other attested examples are more recalcitrant, extending the construction to contexts in which the “personal” dative is impersonal (although still benefactive in a sense) or affective but not obviously benefactive, even in an extended sense: (30) a. That house needs it a new roof. (Sroda/Mishoe1995) b. He rode him around with a head in his trunk for a week. (M. Montgomery, p.c.)

21

Cf. http://www.youtube.com/watch?v=lj3iNxZ8Dww. Evidence includes the distribution of time adverbials modifying the interval of possession: I {need/want} your printer until tomorrow afternoon ( for a week, …); cf. e.g. Dowty 1979: 269–271. 22

expressive meaning, free datives, and f-implicature

177

Narrowing down the contexts in which PDs appear (or appear naturally; it’s hard to determine any absolute exclusions, especially as the construction spreads beyond its original home turf) helps determine the meaning they contribute. But what is the status of that meaning? If PDs are not subcategorized by the verb, and a fortiori not (indirect) objects (the “Southern Double Object” label of Webelhuth & Dannenberg 2006 notwithstanding), what are they? If personal datives do not constitute arguments of the predicate, what is their semantic contribution, if any, to the sentences in which they appear? Unsurprisingly, I take PDs to contribute an F-implicature of subject affect, imposing a use-conditional constraint on its felicitous assertion, viz. that the speaker assumes that the action expressed has or would have a positive effect on the subject, typically satisfying the subject’s perceived intention or goals. Note in particular that PDs share with (other) F-implicatures the property of descriptive ineffability cited above: the content of such implicatures is notoriously elusive.23 In this respect, it patterns with (31) – the implicature of effort or difficulty associated with manage – the source of the positive or negative assessment in the implicatures associated respectively with deprive and spare (Wilson 1975) – the nature of the contrast/unexpectedness implicated by but (see discussion in § 2) – the characterization of the scalar F-implicature associated with even (relative or absolute? unlikelihood or noteworthiness?) – the nature of the expressive attitude embodied in racial and ethnic slurs and other epithets (Potts 2007c; Williamson 2003, 2009) – the precise notion of uniqueness, maximality, or individuability constituting F-implicature of definite descriptions (cf. Horn & Abbott 2012 and references cited therein) – the appropriateness implicatures for tu vs. vous or other T vs. V 2nd person pronouns within a given context in a particular sociolinguistic community of practice (T can be affectionate, presumptuous, comradely, or condescending; V can be polite, aloof, diplomatic, or hostile; cf. Brown & Gilman 1960; Greenhall 2007; Mühlhäusler & Harré 1990; Taavitsainen & Jucker 2003; Zimmermann 1997)

23 As Geurts (2007) points out in his critique of Potts 2007, the presence of ineffability can at most be a necessary, not sufficient, condition on expressives and other candidates for non-at issue meaning, since at-issue meaning in many cases—including, although not limited to, those involving vagueness—may well be (relatively) ineffable in their own right. One response, which I cannot pursue in depth here, is that while there may be fuzzy boundaries in the extension of terms like green or languid, to take two of Geurts’s examples, the core of their meaning contribution is not infected with ineffability as it is for the examples of non-truth-conditional effects under discussion here (extending to but and even, where,

178

laurence r. horn

Thus the fact that it is difficult to pin down precisely what it is that PDs contribute to the semantics of the sentences in which they occur, as eloquently demonstrated by the literature on the construction, is an indirect argument for situating that meaning—however it is to be represented—as an F-implicature. But what motivates (or permits) this property of ineffability for F-implicatures? It is plausible that the edges of truth-conditional meaning should be discrete, while inconsistency in the mental representation of non-truthconditionally relevant content is less pernicious. If you know generally that my use of vous rather than tu signals something in the range of formal respect, distancing, and/or lack of intimacy, my precise motives can remain underdetermined, but if you do not know whether I’m using a 2nd person or 3rd person pronoun, the indeterminacy would be more serious. Similarly, you will want to know whether I bought the car for myself or for my son, and hence to whom an indirect object pronoun refers, but whether or not you can figure out precisely why I affirm “I bought me a car for my son” rather than simply “I bought a car for my son”, no difference in argument structure or truth conditions will arise. Another feature of the PD that speaks to its F-implicature status is its resistance to negation. We saw in (27) above that PDs generally prefer emotively positive contexts that reflect the fulfillment of the subject’s intentions or goals. More generally, we noted the contrast between love (± the relevant snowclone) and hate. In fact, there are over 300,000 raw google hits for “I don’t love me some X”, but these tend overwhelmingly to involve either syntagmatic priming or the canceling effect of double negation: (32) “love me some” under negation

u� Okay, I don’t love me some Adam Sandler, the way I love me some Cadbury

Eggs and the way I love me some latex kitchen gloves. But his new movie, Punch-Drunk Love. … u� Just because I’m not watching Elf repeatedly does not mean I don’t love me

some Christmas.

u� Which is not to say I don’t love me some Wham!

u� At what point do fanatics say to themselves, ‘Okay, I know killing is supposed

to be all wrong and shit, but dammit if I don’t love me some God!’?

unlike Potts, I invoke conventional implicature). Thanks to the editors for drawing my attention to the relevance of Geurts’s discussion.

expressive meaning, free datives, and f-implicature

179

Indeed, most negated love me some cites are of the form “(It’s) not/It isn’t that I don’t love me some X ”, “Don’t think that I don’t love me some Y ”, “I can’t say I don’t love me some Z”, etc. When we move to other predicates, the results are similar. Some empirical contrasts, courtesy once again of Google (Nov. 1, 2009), with some outliers indicated: (33) Polarity constrasts with PDs “I have me some” “I have me a” “I don’t have me a/any” “I lack me a/any” “I want me some” “I don’t want me any”

22,600 37,300 89 Pity I don’t have me any beers 6 i lack me a microwave oven and didn’t feel like frying 677,000 3

“I like me some/a” 130,000 “I don’t like me any/a” 7 Not that I don’t like me a bit of Matt “I dislike me some/any/a” 6 I dislike me some Parasite Hilton

As is well known, many of the standard examples of F-implicature are essentially impermeable to ordinary negation.24 If you tell me She’s poor but happy and I am willing to grant that she possesses both properties but reject any expectation that poverty and happiness generally contrast, it’s not clear how I can convey this, especially with a simple negative (?She’s not poor but happy). Not even Hercules can lift that stone does not negate Even Hercules can lift that stone, but instead paraphrases Even Hercules can’t lift that stone. Some F-implicata can be attacked only with metalinguistic or echoic negation, while others (e.g. the assumptions conveyed by the use of epithets or T/V pronouns) consistently scope out of negation. Once again, the behavior of the PD construction as positive polarity items effectively resisting the scope of negation is consistent with their treatment in terms of F-implicature, a conventional but non-truth-conditional contribution to content. With PDs, a complicating factor may be the typically benefactive nature of the construction; the subject is understood to intend and/or be positively affected by, the action in question. There is thus a conflict between the negativity, in the emotive or evaluative sense, of negation (cf. Horn 1989; 24 Gutzmann (2008), for example, shows that ja (implicating that the proposition is already known to the hearer), along with other modal particles like halt (implicating the obviousness of p), doch, and wohl, as opposed to truth-conditionally relevant particles like nur, allein, vielleicht, scopes out of negation. We return to the link between German Modalpartikeln and non-argument datives in the following section.

180

laurence r. horn

Potts 2010) and the positivity of PDs. Thus, it is not surprising when a television commentator declares of a college football team that “Auburn hired them one heck of an offensive coordinator” (heard 6 Jan. 2009); a parallel declaration following a characteristically disappointing season that “The Mets fired them another manager” would be far less likely, even controlling for regional differences.25 6. Around the World with Non-Argument Datives In the Construction Grammar based treatment of the “Southern double object” advocated by Webelhuth & Dannenberg (2006), the personal dative construction is taken to be a sui generis idiom, precluding any effort to seek a systematic relationship between the non-standard English construction and non-subcategorized datives in other languages that index either the subject or non-subject referents.26 The term free dative (cf. also non-valence, extended, or sentence dative) is sometimes employed as a cover term for what we will see is a family of related constructions. This landscape features specimens that have been variously termed the ethic(al) dative, the dative of interest or pertinence, and the dativus (in)commodi, among others; the languages on display range over French and its Romance cousins (Leclère 1976, Barnes 1985, Authier & Reed 1992 Herschensohn 1992, Cuervo 2003), German (Abraham 1973, Wegener 1989, Draye 1996, Maling 2001, Burridge 2003, Cook 2006, McIntyre 2006, Gutzmann 2007), Old and Middle English (Sweet 1900, Keenan 2003), Hebrew—both Ancient (Gesenius 1910, Muraoka 1978) and Modern (Berman 1982, Borer & Grodzinsky 1986, Landau 1998, Halevy 2007, Linzen 2008)—and Warlpiri (Simpson 1991, Legate 2001). Overviews ranging across many languages and language families are offered by Lamiroy & Delbecque (1998), Hole (2006), Bosse et al.

25 The editors point out another factor for the asymmetry here, viz. the “negativity of negation” (Horn 1989, Potts 2010). In Horn 2008: 179, I attribute the virtual lack of anaphoric X’s ass (cf. Beavers & Koontz-Garboden 2006) in PDs—despite the very occasional u� God loves his ass some kidneys (‘God loves kidneys’) or u� I need my ass some gingko biloba—to the clash between the adversative nature of metonymic X’s ass and the benefactive nature of PDs. In naturally occurring instances of this kind, e.g. u� (You should) get your ass some help/therapy, the metonymic (non-proctological) occurrence of your ass “marks not simply a pejorative attitude, but rather the speaker’s impatience toward the addressee/subject”, who will indeed achieve a positive affect by following the prescription. 26 As Adele Goldberg reminds me, there is nothing intrinsic to Construction Grammar itself that rules out a systematic attempt to relate the PDC to non-argument datives in other languages.

expressive meaning, free datives, and f-implicature

181

(2012), and the papers in Hole et al. 2006. While space prevents a full travelogue, some relevant high points of the journey will be explored briefly here. We begin our tour with French, in which a coreferential construction bearing apparent connections with the PD occurs in informal or colloquial usage, typically (but not exclusively) with 1st person subjects and common monosyllabic verb forms: (34) a.

u� Je me bois un bon café chaud.

lit., ‘I drink me a good hot coffee’ u� J’ ouvre le frigo, je me bois un verre de vin rouge, je me détend dans mon canapé. ‘I open the fridge, I drink (me) a glass of red wine, I relax on my couch’ c. u� Je me lis tantôt la Bible et le Coran, du Porno et du mystère. ‘I read (me) sometimes the Bible and the Koran, porno and mysteries’ d. u� Je me fais un voyage. [6210 hits, mostly 1st person] ‘I make (me) a trip’ e. u� Manger pour elle devient secondaire ou alors elle se prend un repas devant le pc. ‘Eating becomes secondary to her, or she has (her[self]) a meal in front of her p.c.’ b.

Note that when a 3rd person example is attested, as in (34e), the reflexive clitic is mandatory; *Ellei luii prend un repas … is impossible.27 While standard German lacks directly parallel subject co-referential datives (as opposed to the non-coreferential non-argument datives discussed below), Silke Lambert has called my attention to a similar construction in the Westphalian/Niederrhein dialect area: (35) a. Ich trinke mir jetzt einen Kaffee. lit. ‘I drink me a coffee now’ b. {Er/Sie} trinkt sich jetzt einen Kaffee. lit. ‘He/She drinks himself/herself a coffee now’

As in (34e), only a reflexive is possible in 3rd person cases, not the simple pronominal.28

27 The appearance of non-argument reflexive clitics in such cases can be taken as evidence for different OT-style rankings of the relevant soft constraints, where the preference for reflexive marking of locally coreferential nominals clashes with the preference for restricting non-logophoric reflexives to co-arguments. It is worth noting that for some speakers English PDs may themselves take the form of reflexives rather than pronominals, as in the 2008 Mariah Carey song lyric “The whole entire world can tell/That you love yourself some me” [gratia Ben Zimmer]. 28 For additional related discussion, see the WordReference Forum thread on sich etwas essen/trinken at http://forum.wordreference.com/showthread.php?t=214623.

182

laurence r. horn

The constructions in (34) and (35) are similar to, but more constrained than, the English PD; cf. also the range of Polish datives described by Da̧browska (1997). The “Lamedh” or le-marked dative of Biblical Hebrew is a more distant cousin, occurring in collocations glossed as ‘Get thee away’, ‘Turn thee aside’, or ‘Our bones are dried up, our hope is lost, we are cut off’ (Ez. 37:11, lit. ‘We are cut us off the parts’): cf. Gesenius 1910: § 119, Muraoka 1978. In fact, as Elitzur Bar-Asher informs me (p.c.), while translators are typically advised to leave the Lamedh untranslated, Gesenius (1910: 381) describes it as a dativus commodi or incommodi, i.e. benefactive/malefactive dative, “used—especially in the colloquial language and in later style—in the form of a pronoun with [le-] as an apparently pleonastic dativus ethicus, with many verbs, in order to give emphasis to the significance of the occurrence in question for a particular subject. In this construction the person of the pronoun must always agree with that of the verbal form”. This construction is an intransitive cousin of the PD; note the reference to its “apparently pleonastic” character. A similar extended use of subjectcoreferential datives is found in late Latin “dans l’ usage populaire, pour indiquer d’une façon plus intensive la part que le sujet prend à l’ action; beaucoup de verbes se sont ainsi construits, notamment des verbes de mouvement” (Bourciez 1930: §118c). A millennium later, we find something similar closer to home—not only in archaic intransitives (“Now I lay me down to sleep”, “Get thee hence”) but in a more direct forerunner to the PDC.29 Sweet (1900) describes the “pleonastic dative” of Old English in analogous terms (boldface added as usual for coreference and underlining to highlight passages of particular relevance): In OE a personal pronoun in the dative is often added reflexively to a pronoun in the nominative but without materially affecting the meaning, as in hē ondrēd him þone mann ‘he was afraid of the man’, literally ‘feared for himself’, hi ̄e ġewiton him ‘they departed’. (Sweet 1900: § 1106)

Over a century later, Keenan (2003) independently provides an updated take on this construction: [F]rom Late OE through ME we also find many non-theta (pleonastic) occurrences of pronouns. They do not satisfy either a semantic role requirement or 29 Curme (1931: 107) describes both the “reflexive dative” PD-style transitive construction (“Let every soldier hew him down a bough” [Macbeth]), now “more common in popular speech than in the literary language” (“I want me a woman who can milk” [Lucy Furman’s “The Quare Women”, 1923]), and its intransitive counterpart in the form of a “weak, almost pleonastic, dative of interest […] after sit, lie, and verbs of motion and fearing” (I sit me down, I dread me, He walked him forth) in texts from Shakespeare and Marlowe to Byron and Tennyson.

expressive meaning, free datives, and f-implicature

183

a syntactic requirement of the verb. In OE they are usually dative […]. They are always bound to the local subject, agreeing with it in person, number and gender and serve semantically to heighten the involvement of its referent: e.g. the subject acted intentionally or was involved in the action in some way other than the role it has in virtue of being the subject argument. (Keenan 2003: § 1.2)

Keenan’s examples include the sentences in (36) (36) a. forðæm hi him ondrædað ða frecenesse ðe hi ne gesioð because they them fear the danger that they not see CP.433 c880 b. ac he ne wandode na him metes to tylienne. … & nam but he not hesitated at all him provisions to provide … and took him on orfe & on mannum & … gewende him þa east werd him in cattle and in men and … went him then eastward to his feder & gewendon heom þa begen east weard … to his father and went them then both eastward … Chron(E)1052

Like those in ancient Hebrew and Latin, the Old/Middle English “pleonastic” datives mark heightened subject involvement but occur in transitive as well as intransitive clauses. Traveling to yet another continent, we find pronominal non-argument coreferential datives Down Under in Warlpiri, as described by Simpson (1991: 382): (37) a. ka-nyanu kuyu nyanungu-ku pi-nyi liwirringki-rli-ji. pres-refl meat it-dat hit-nonpast Lizard sp.-erg-euph ‘… it kills itself animals, that Lizard’ b. Palkarni-rlipa-nyanu yalumpuju ngalipa-ku-jala marda-rni. scarce-1pl.subj-refl that.near we.pl.incl.-dat-clear hold-nonpast ‘We’ll keep these scarce things just for ourselves’

Commenting on this construction, Legate (2001) notes its similarity to English ‘I’m gonna bake me a cake’, i.e. the PDC. In addition to the non-argument subject-coreferential datives surveyed above, in many languages the involvement of non-subject participants not subcategorized by the predicate may also be indexed by dative (pro)nominals. Typically (although not invariably), these non-argument datives are restricted to pronouns (sometimes to “speech-act”, i.e. 1st and 2nd person, pronouns) and often serve to mark adversative or maleficiary rather than

184

laurence r. horn

beneficiary affect, as in the French imprecation Foutez-moi le camp. One language with a particularly robust array of free datives is German. Discussing the examples in (38), (38) a. Helf mir mal deinem Vater in der Küche. help me-dat a.minute your-sg father in the kitchen ‘Go help your father in the kitchen for a minute for me’ b. Der David hat mir der Claudia schon zuviel Geschenke the David has me.dat the Claudia.dat already too.many gifts.acc gegeben. given ‘I think [lit., ‘To me’] David has already given Claudia too many presents’

Maling (2001: 432) comments: This extra dative […] is interpreted as a beneficiary or person adversely affected by the event […]. I assume that this dative is not subcategorized for by the verb. As an adjunct rather than an argument, it is not a grammatical object, and thence not a counterexample to the descriptive generalization that German allows at most one dative object per clause. (Maling 2001: 432)

Although not co-indexing the subject in the manner of PDs, these datives are thus similarly non-subcategorized, non-arguments, and non-objects, and Maling’s reference to adverse affect is the negative counterpart of our characterization of the usual semantic value of PDs. Indeed, Gutzmann (2007) analyzes the contribution of the “dativus ethicus” in contexts like (38a) as a conventional implicature expressing “that the speaker has some personal interest in the hearer’s execution of the action requested.” However, Gutzmann (p.c.) notes that (38b) is not in fact a dativus ethicus but a dativus iudicantis, allowing 3rd person (and non-pronominal) objects that can be fronted and questioned, unlike ethical datives. A case can be made that such examples, as distinct from true ethical datives, do contribute to truth-conditional rather than just use-conditional meaning and are thus less suited to an F-implicature-based analysis.30 As stressed by Draye (1996), Gutzmann (2007), Lambert (2007, 2010), and others, “free datives” represent a continuum, members of a dative family. The relevant properties of the family members include their syntactic optionality, non-argument status, non-extractability, preference for pronominal and in particular 1st person 30 Parallel situations are attested elsewhere. Thus Sawada (2010) argues that the Japanese particle chotto contributes at-issue content when it functions as an amount minimizer but contributes only CI content when it functions as an expressive minimizer. Cf. Davis et al. 2007 for similar observations about the epistemic and expressive value of various evidentials.

expressive meaning, free datives, and f-implicature

185

form, and affectedness (Betroffenheit), unpacked by Lambert as the cognitive/emotional involvement of the referent (based on the speaker’s assessment). By these criteria, it is only the true ethical datives that are clear candidates for the analytical option of F-implicature. Similar “affected dative” constructions have been posited for older English (Curme 1931: 106–109), for Romance (see e.g. Leclère 1976, Barnes 1985, Authier & Reed 1992, and Herschensohn 1992 on the “non-lexical” datives of French and Cuervo 2003 on Spanish), and for Modern Hebrew (Berman 1982, Halevy 2007, Linzen 2008). In French, German, Serbian/Croatian, Modern Hebrew, and other languages, the non-coreferential cases are often seen as extending to possessive datives. Thus we obtain paradigms like (39) (culled from above sources; glosses mine); note that lexical dative NPs are ruled out in these environments, as seen in (39d). (39) a. Paul se tape un pastis. ‘Paul knocks (him) back an anisette’ [“Reflexive dative”, marking the subject’s interest; cf. (34)] b. Au mont St. Michel, la mer te monte à une de ces vitesses! ‘At Mont St. Michel, the sea rises (on you) at an incredible speed’ [“Ethical dative”] c. Il te lui a donné une de ces gifles. ‘He gave her a slap (on you)’ d. Jean lui a attrapé deux rhumes. (*Jean a attrapé deux rhumes à sa mere.) ‘Jean caught {her/*his mother} two colds.’ e. Les mains lui tremblent. ‘His hands are shaking’ [“Possessive dative”]

Lamiroy & Delbecque (1998) gather this family together under a single umbrella: [T]he possessive and the ethical dative are different manifestations of one and the same basic phenomenon, viz. that of introducing entities into the sentence structure which, from a syntactic point of view, are not lexically predicted by the verb and which semantically correspond to entities that are not actively involved in the process but nonetheless affected by it, in one way or another. (Lamiroy & Delbecque 1998: 63)

Possessive datives have unsurprisingly been candidates for F-implicatureoriented analyses, complete with ineffability effects. In her study of external possessive constructions in Northern Pomo, Spanish, and Czech, O’Connor (2007) argues that EPCs, which index “[the] speaker’s stance or judgment with respect to the outcome of the event for the possessor”, allow the speaker to “exploit a relatively underspecified conventional implicature

186

laurence r. horn

to convey scorn, respect, distance, or empathy, or to otherwise enrich the hearer’s understanding of their stance toward the fate of the extra-thematic possessor”. But there are differences among the members of the dative family as well as similarities, especially in their syntax. Berman (1982: 49–50), while acknowledging that “extended” datives in Hebrew (including possessors, benefactives, and others) “have in common the sense of association or involvement of someone with an event for which he is not responsible, and of which he is not the direct patient, an event by which he is nonetheless affected, either favorably or adversely”, distinguishes the more restricted ethical datives from other members of its “extended” family, including the possessives. Extended datives, such as those in (38), are indeed all animate31 and “affected” by the action of the clause without being related to the valency of the verb. Instead, they involve “sentence datives” à la Curme (1931), Draye (1996), or Burridge (2003). But following Barker & Dowty (1993), possessives are arguments—viz., nominal arguments of the possessee—while ethical datives are echt non-arguments, which may be one factor for the differences cited in the literature, e.g. the fact that passives are fine with lexical datives in French, somewhat unnatural with possessive datives, and totally out with ethical datives (Lamiroy & Delbecque 1998: 64; glosses theirs). (40) a.

Ce livre lui a éte donné par son grand-père. ‘This book was given to him by his grandfather’ b. ??La figure lui a été cassée par la police. ‘His face was broken [i.e. he was stabbed] by the police’ c. *Une de ces gifles te lui a été donnée. ‘He got one of those smacks (to your detriment)’

Similarly, in Hebrew it is only the ethical datives and the PD-type subject coreferential datives that completely block extraction and allow only clitic form (cf. Borer & Grodzinsky 1986; Halevy 2007).

31 Whence the “Dead Possessor Effect” described in Burridge 2003, Hole 2006, and elsewhere. The minimal pair in (i) is from Burridge 2003 (her (43)/(44)), although the editors point out that the anomaly is reduced with an impersonal subject (Man zog dem Toten …).

(i)

Wir zogen dem Verletzten/#Toten die Jacke aus. ‘We took off the injured/#dead person’s coat’

(See Hole 2006, Linzen 2008, and Bosse et al. 2012 for reflexes of the dead possessor effect in Mandarin, Hebrew, and Albanian respectively.) A genitive in an example like (i) would freely allow either a living or dead possessor, since it is the link between datives and affectedness that is crucial.

expressive meaning, free datives, and f-implicature

187

A similar conclusion is reached independently by Gutzmann (2007) for ethical datives, the title of whose paper, “Eine Implikatur konventioneller Art: Der Dativus Ethicus”, plays off that of Wegener’s (1989) paper on modal particles, “Eine Modalpartikel besonderer Art: Der Dativus Ethicus.” Indeed, we find sentences that allow a variety of readings extending from the members of the dative family (including those manifesting the EPC) to modal particles, involving in each case the contribution of an F-implicature. Thus (41), from Lambert (2007), following Schmid (1988), allows dativus commodi, dativus incommodi, and/or external possessive readings: did Max cut the roses for his girlfriend, did he spite her by cutting them (cf. “Max cut the roses off on her”), or did he just cut off her roses? (41) Max schneidet seiner Freundin die Rosen ab. Max cuts his-fem.dat girlfriend the roses off (Lambert 2007: (17), from Schmid 1988: 153)

Similarly, Lambert (2007: (21)) notes, the implicature contributed by the ethical dative in (42a) (a line from Empfang, a poem by Richard Dehmel) could equally well be contributed by ja, the Modalpartikel in (42b), either alongside or instead of the dative particle.32 (42) a. Aber komm mir nicht im langen Kleid! but come me-dat not in.the long dress b. Aber komm (mir) ja nicht im langen Kleid!

Studies of German modal particles (ja, doch, denn, halt, bloß, …) by Meibauer (1994), Wegener (2002), and especially Gutzmann (2008) have shown that while MPs often have homophonous non-MP doppelgängers, true MPs can be distinguished by their inability to bear stress and their tendency to be phonetically elided, by their semantically bleached meanings (as against their adverbial homophones), by their syntactically constrained distribution (appearing only in the middle field, and sometimes restricted by mood or illocutionary force), and by their inability to be questioned, focused, or negated, traits we have already seen to cluster with F-implicated material.33 Most strikingly, MPs can be suppressed salva veritate and with no

32 Other speakers find (42b) acceptable only with stressed ja, which rules out an MP reading given the diagnostics noted in the text. (Cf. Gutzmann 2008: 74 for related discussion.) 33 In these and other respects, the “evidential” markers in St’át’imcets described by Matthewson et al. (2007)—reportative ku7, inferential k’a, and perceived-evidence -an’—pattern strikingly like the German Modalpartikeln. For more on the semantic status of the meaning contribution of evidentials cross-linguistically, see also Faller 2002, 2012; Aikhenvald 2004; and Davis et al. 2007.

188

laurence r. horn

palpable loss of informative content; they function epistemically to signal the speaker’s attitude about the proposition, rather than contributing to truth conditions. Thus, Gutzmann (2008: 52) notes that the four truthconditionally identical statements in (43) differ in what is implicated, yielding different appropriateness or use conditions for all four: (43) a. b. c. d.

Peter ist Peter ist Peter ist Peter ist

∅ ja doch halt

ein Linguist. ein Linguist. ein Linguist. ein Linguist.

‘Peter is a linguist’ ‘… as you (should) know’ ‘… in contradiction to an earlier claim’ ‘… which is an incontrovertible fact’

As Gutzmann (2008: 62) observes, use-conditional meaning (à la Kaplan 1999) may be a more accurate label than expressive meaning for these phenomena. Our whirlwind tour of some of the world’s celebrated non-argument dative sites has barely scratched the surface of the complex range of phenomena involved, but it does suggest that the personal dative of non-standard varieties of American English, rather than constituting an isolated idiom, is in fact one representative of a wide class of non-argument affectees.34 Such affectees are typically marked as datives in languages with a more sophisticated panoply of case options than modern English retains, whence the semi-misnomer of “personal dative” for what is not formally a dative at all. In English, which lacks a weak clitic reflexive like Dutch zich or French se (cf. Reuland 2001), the non-argument status of the locally co-indexed element suffices to allow, or for most speakers in the relevant dialect to require, its representation as an ordinary pronominal, in apparent (but, I have suggested, not real) violation of Principle B.35

34 Mention should be made of related constructions in which non-subject affectees appear as obliques (as in the English adversative My dog died on me) or as subjects (as in the adversative passives found in many languages); cf. Hole 2006 for a comprehensive study of “extra arguments”. To the extent that this terminology is well-taken, we are not dealing with (mere) F-implicature as with PDs and non-coreferential ethical datives; extra arguments, after all, are arguments. 35 While I have argued that PDs, as non-co-arguments, are exempt from Condition B, they can represent bound variables, as seen in the quantifier in (i) and the requirement for sloppy readings in (iia) as opposed to the strict reading preferred in (iib), (iic) for contrastive focus (non-PD) pronominals (cf. Horn 2008: § 2).

(i) {Everyone here/Only Billy Joe} needs him a new pickup truck. (ii) a. I need me a shotgun and so does Billy Bob. [sloppy identity only: we both do] b. I want ME to win and so does Hubert. [pref.: ‘Hubert wants me to win too’]

expressive meaning, free datives, and f-implicature

189

7. F-Implicature: Some Residual Issues I have tried to demonstrate the utility of what I call F-implicature, subsuming a variety of relations Frege introduces under different labels (or non-labels) along with Grice’s relation of conventional implicature (a.k.a. “formality without dictiveness”). Symptoms of F-implicature status include detachability, non-cancellability, and irrelevance to the truth conditions of (or the thought expressed by) the sentence that contains them.36 Additional characteristics include projection out of embedded contexts, the immunity to certain kinds of objection, and ineffability or context variation in the content of the implicature.37 For both Frege and Grice, but not necessarily for linguists “concerned”, in Frege’s words, “with the beauties of language”, identifying the constructions in question—a motley collection of particles, speaker-oriented sentence adverbs, prosodic features, word order effects—serves largely to isolate them in terms of what they are not: they do not affect the thought, the truth-conditionally relevant meaning of a given expression, and at the same time they cannot be computed by general principles of rational interchange. We have expanded the collection (or cited other work that expands it) to encompass a number of other candidates for F-implicature status, including additive and scalar particles, evidential markers, 2nd person pronouns, c. I voted for ME and so did Hubert. [pref.: ‘Hubert voted for me too’] For Conroy (2007), PDs—notwithstanding their pronominal appearance—are in fact anaphors, bound variables assigned case but no theta role, akin to SE anaphors on the theory of Reuland (2001). Note too that non-subject-coreferential (ethical) dative pronominals in Hebrew can be bound variables, as with the “affected experiencer” lo (iii), from Bosse et al. (2012: (98)); translation modified: (iii)

Kol hore1 hit-acben she-ha-yeled pit’om baxa lo1 be-emca ha-seret. every parent got-irritated that-the-child suddenly cried to.him/her in.middle the movie ‘Every parent1 got irritated when their1 child suddenly cried in the middle of the movie on them1 ’

Bosse et al. (2012: 1224) argue that in general “binding from the assertive content into the non-assertive content is allowed in natural language”. 36 It should be clear that in referring to F-implicatures (or conventional implicatures) as truth-conditionally irrelevant (to what is said), I am not claiming—nor is Frege or Grice— that their own content is itself non-truth-conditional; cf. Bach 1999: 331–332 on this fallacy. 37 Thanks to Eliza Block for helping me organize these properties more coherently, while also raising doubts about whether the putative F-implicating devices really form a natural class. My current speculation is that the very salient distributional differences Block isolates within this category are attributable to the syntactic differences among the implicature triggers but I will have to defer these questions to another occasion.

190

laurence r. horn

epithets and other expressives, the uniqueness condition on definites, and non-argument personal datives. Along the way, we have touched on Pottsian CIs, a class disjoint from Gricean conventional implicatures (and overlapping with their Fregean precursors).38 While I find Potts’s (and Bach’s) reanalysis of the classical implicature examples as part of what is said unpersuasive, his multiple-propositions approach does allow for an insightful description of clausal appositives and not only … but also constructions, where both expressed propositions are in principle truth-conditionally (rather than merely use-conditionally) relevant. But these are distinct from true F-implicatures, which do not involve “said-content”. Finally, and again very much in the spirit of Grice’s (and Frege’s) views, Jayez (2004) and Jayez & Rossari (2004) argue that discourse markers are not part of what is said or asserted but constitute conventional implicatures, as they may be explicitly communicated and grammatically embedded, but are not asserted and cannot be directly refuted. They argue further that presuppositions, as traditionally conceived, are a subset of conventional implicatures—those constituting part of given information. On the other hand, one argument for distinguishing between presuppositions and F-implicatures is based on the property of and filtering that has been recognized since Karttunen (1973) as characterizing presuppositions but that seems not to apply to F-implicatures: (44) a. If he’s married, his wife is unhappy. b. #If there’s a contrast between poverty and happiness, they’re poor but honest. c. #If Max is obnoxious, that bastard just got promoted. d. #Si nous nous connaissons, tu es soûl. ‘If we know each other, you [fam.] are drunk’ e. #If he wanted to listen to something sad, he just heard him a sad song.

On the topic of non-refutability, it is worth noting that besides the responses of “You’re wrong” or “That’s false”, which Jayez & Rossari (2004) show cannot attach to conventionally implicated material, another all-purpose refuter is “Bullshit!”, as explored by Ward (2003), who credits Craige Roberts for seminal work in this area. Substitution of bullshit for whaddaya mean produces anomaly:39 38 See Salmon 2011 for an empirical argument that an analysis based on F-implicature (or classical conventional implicature) provides a better account of epistemic must than one invoking Pottsian CI or presupposition. 39 Geoff Shaw (p.c.) notes that there is considerable variation on the relevant judgments,

expressive meaning, free datives, and f-implicature

191

(45) A: B1:

“Her name is Caroline. She’s an Italian girl but she’s pretty.” “What do you mean, but she’s pretty, Ma?” Frank said. “Why not ‘and she’s pretty’?” B2: #“Bullshit, Ma.” Frank said. “Why not ‘and she’s pretty’?”

Bullshit! is equally incapable of overriding the F-implicature of even, as Merde! is for that of an inappropriate tu or vous. On the other hand, as Gregory Ward points out (p.c.), the causal component of therefore is a viable candidate for bullshit-rejection, which supports its banishment from the ranks of F-implicature: (46) A: Lance loves musical comedies and has seen “Mamma Mia” twice. He is, therefore, gay. B: Bullshit. That doesn’t follow.

The PDC patterns with other cases of F-implicature on the bullshit diagnostic: (47) A: Toby heard him a sad song. B: #Bullshit. He heard one, but no way it was on purpose.

As Grice (1989: 46) acknowledges, “The nature of conventional implicature needs to be examined before any free use of it, for explanatory purposes, can be indulged in”. I hope to have shown here that for a range of expressive meanings in natural language, including a subset of non-argument datives, the careful indulgence in a coherent program for conventional but nontruth-conditionally relevant content, in the tradition of Frege (1879, 1892, 1897/1979, 1918) and Grice (1961, [1967] 1989), is free from bullshit. Acknowledgements I am grateful to Barbara Abbott, Grant Armstrong, Kent Bach, Suzanne Bobzien, Solveig Bosse, Stacey Conroy, Philippa Cook, Clare Dannenberg, Christiane Fellbaum, Itamar Francez, Adele Goldberg, Daniel Gutzmann, Owen Greenhall, Corinne Hutchinson, Silke Lambert, Julie Legate, Mark Liberman, Liela McLachlan, Michael Montgomery, Cathy O’Connor, Chris and that for many speakers an implicature-rejecting “Bullshit” response may be felicitous in some contexts. Thus in Dummett’s example from § 2 above, someone might reject the observation “Robinson gives good lectures but he is in Australia” by exclaiming “Bullshit! Our unlimited budget would cover the cost of the flight”. But even if this is acceptable (it’s marginal for me), it becomes totally impossible if the rebuttal consists of a bare “Bullshit!” This could only be used to refute one or both of the actual conjuncts, not the contrast between the reasons for and against inviting Robinson.

192

laurence r. horn

Potts, Kurt Queller, Craige Roberts, Molly Rubenstein, Will Salmon, Scott Schwenter, Geoff Shaw, Gregory Ward, Gert Webelhuth, Ben Zimmer, and those attending the Yale Philosophy Faculty Lunch (Feb. 2007), the LAUD Symposium (Landau, March 2006), the Colloque de Syntax et Sémantique (Paris, Oct. 2007), and especially the DGfS Arbeitsgruppe on Expressives (Osnabrück, March 2009) for comments, objections, and suggestions. Thanks to Eliza Block, my discussant in Oslo, and Mandy Simons and Dustin Tucker, my discussants in Ann Arbor. Feedback from participants in my 2009 Conventional Implicature seminar was also helpful. My deepest debt is to the editors of this volume, Daniel Gutzmann and Hans-Martin Gärtner, for their many helpful comments on the first draft of the current paper. Everyone but me is hereby absolved from anything I got wrong. References Abraham, Werner (1973): “The ethic dative in German”. In: Kiefer, Ferenc & Nicolas Ruwet, eds.: Generative grammar in Europe. Dordrecht: Reidel, 1–19. Aikhenvald, Alexandra (2004): Evidentiality. Oxford: Oxford University Press. Amaral, Patricia & Craige Roberts & E. Allyn Smith (2007): “Review of The logic of conventional implicatures by Chris Potts”. Linguistics and Philosophy 30, 707–749. DOI: dx.doi.org/10.1007/s10988-008-9025-2. Anscombre, Jean-Claude & Oswald Ducrot (1983): L’argumentation dans la langue. Bruxelles: Pierre Mardaga. Authier, J.-Marc & Lisa Reed (1992): “On the syntactic status of French affected datives”. The Linguistic Review 9, 295–311. DOI: dx.doi.org/10.1515/tlir.1992.9.4.295. Bach, Kent (1999): “The myth of conventional implicature”. Linguistics and Philosophy 22, 327–366. DOI: dx.doi.org/10.1023/A:1005466020243. ——— (2001): “You don’t say?” Synthese 128, 15–44. DOI: dx.doi.org/10.1023/A: 1010353722852. Barker, Chris & David R. Dowty (1993): “Non-verbal thematic roles”. Proceedings of NELS 23, 49–62. Barker, Stephen (2003): “Truth and conventional implicature”. Mind 112, 1–33. DOI: dx.doi.org/10.1093/mind/112.445.1. Barnes, Betsy (1985): “A functional explanation of French nonlexical datives”. Studies in Language 9, 159–195. DOI: dx.doi.org/10.1075/sl.9.2.02 bar. Beavers, John & Andrew Koontz-Garboden (2006): “A universal pronoun in English?” Linguistic Inquiry 37, 503–513. DOI: dx.doi.org/10.1162/ling.2006.37.3.503. Berman, Ruth (1982): “Dative marking of the affectee role: Data from modern Hebrew”. Hebrew Annual Review 6, 35–59. Blakemore, Diane (2002): Relevance and linguistic meaning. The semantics and pragmatics of discourse markers. Cambridge: Cambridge University Press. Borer, Hagit & Yosef Grodzinsky (1986): “Syntactic cliticization and lexical cliticization. The case of Hebrew dative clitics”. In: Borer, Hagit, ed.: The syntax of pronominal clitics. New York: Academic Press, 175–217.

expressive meaning, free datives, and f-implicature

193

Bosse, Solveig & Benjamin Bruening & Masahiro Yamada (2012): “Affected experiencers”. Natural Language and Linguistic Theory 30, 1185–1230. DOI: dx.doi.org/ 10.1007/s11049-012-9177-1. Bourciez, Édouard (1930): Éléments de linguistique romane. 3rd ed. Paris: C. Klincksieck. Brown, Roger & Albert Gilman (1960): “The pronouns of power and solidarity”. In: Sebeok, Thomas A., ed.: Style in language. Cambridge, MA: MIT Press, 253– 277. Büring, Daniel (2005): Binding theory. Cambridge: Cambridge University Press. Burridge, Kate (2003): “Sentence datives and the grammaticalization of the dative possessive”. LaTrobe University Working Papers in Linguistics. URL: http://www .latrobe.edu.au/linguistics/LaTrobePapersinLinguistics/Vol%2003/03Burridge .pdf. Carston, Robyn (2002): Thoughts and utterances. The pragmatics of explicit communication. Oxford: Blackwell. Christian, Donna (1991): “The personal dative in Appalachian English”. In: Trudgill, Peter & Jack K. Chambers, eds.: Dialects of English. London: Longman, 11–19. Conroy, Anastasia (2007): “The personal dative in Appalachian English as a reflexive pronoun”. University of Maryland Working Papers in Linguistics 16, 63–88. Cook, Philippa (2006): “The datives that aren’t born equal: Beneficiaries and the dative passive”. In: Hole, Daniel & André Meinunger & Werner Abraham, eds.: Datives and other cases. Between argument structure and event structure. Amsterdam: John Benjamins, 141–184. Cooper, Robin (1983): Quantification and syntactic theory. Dordrecht: Reidel. Cruse, David Alan (1986): Lexical semantics. Cambridge: Cambridge University Press. Cuervo, Maria Cristina (2003): Datives at large. PhD thesis. Cambridge, MA: MIT. Curme, George (1931): Grammar of the English language. II: Syntax. Boston: D.C. Heath & Co. Da̧browska, Ewa (1997): Cognitive semantics and the Polish dative. Berlin: de Gruyter. Davidson, Donald (1986): “A nice derangement of epitaphs”. In: Grandy, Richard E. & Richard Warner, eds.: Philosophical grounds of rationality. Intentions, categories, ends. Oxford: Oxford University Press, 157–174. Davis, Christopher (2009): “Decisions, dynamics and the Japanese particle yo”. Journal of Semantics 26.4, 329–366. DOI: dx.doi.org/10.1093/jos/ffp007. Davis, Christopher & Christopher Potts & Margaret Speas (2007): “The pragmatic value of evidential sentences”. Proceedings of SALT 17, 71–88. URL: http:// elanguage.net/journals/salt/article/view/17.71/1845. Dowty, David R. (1979): Word meaning and Montague grammar. The semantics of verbs and times in generative semantics and Montague’s PTQ. Dordrecht: Kluwer. Draye, Luk (1996): “The German dative”. In: van Belle, William & Willy van Langendonck, eds.: The dative. Vol 1: Descriptive studies. Amsterdam: John Benjamins, 155–215. Dummett, Michael (1973): Frege: Philosophy of language. Oxford: Duckworth. Faller, Martina (2002): Semantics and pragmatics of evidentials in Cuzco Quechua. PhD thesis. Stanford University.

194

laurence r. horn

——— (2012): “Evidential scalar implicatures”. Linguistics and Philosophy 35, 285–312. DOI: dx.doi.org/10.1007/s10988-012-9119-8. Fellbaum, Christiane (2005): “Examining the constraints on the benefactive alternation by using the world wide web as a corpus”. In: Reis, Marga & Stephan Kepser, eds.: Evidence in linguistics: Empirical, theoretical, and computational perspectives. Berlin: de Gruyter, 207–236. Francescotti, Robert (1995): “Even: the conventional implicature approach reconsidered”. Linguistics and Philosophy 18, 153–173. DOI: dx.doi.org/10.1007/BF 00985215. Frege, Gottlob (1879): “Begriffsschrift, selections”. In: Beaney, Michael, ed.: The Frege reader. Oxford: Blackwell, 47–78. ——— (1892): “Über Sinn und Bedeutung”. Zeitschrift für Philosophie und philosophische Kritik 100, 25–50. ——— (1897/1979): “Logic”. In: Hermes, Hans & Friedrich Kambartel & Friedrich Kaulbach, eds.: Posthumous writings: Gottlob Frege. Translated by Peter Long and Roger White. Oxford: Blackwell, 126–151. ——— (1906): “Letter to Husserl, 9 Dec. 1906”. In: Beaney, Michael, ed.: The Frege reader. Oxford: Blackwell, 305–307. ——— (1918): “Thought”. In: Beaney, Michael, ed.: The Frege reader. Translated by Peter Geach and Robert H. Stoothoff. Oxford: Blackwell, 325–345. Gesenius, Wilhelm (1910): [orig. early 19th c.] Gesenius’ Hebrew grammar. Ed. by Emil Kautsch, revised by Arthur Ernest Cowley. Second English edition. Oxford: Oxford University Press. Geurts, Bart (2007): “Really fucking brilliant”. Theoretical Linguistics 33.2, 209–214. DOI: dx.doi.org/10.1515/TL.2007.013. Greenhall, Owen (2007): “Pronouns and conventional implicature”. Paper presented at Fifth Barcelona Workshop on Issues in the Theory of Reference, June 2007. Grice, H. Paul (1961): “The causal theory of perception”. Aristotelian Society Supplement 35, 121–152. URL: http://www.jstor.org/stable/4106682. ——— (1975): “Logic and conversation”. In: Cole, Peter & Jerry L. Morgan, eds.: Speech acts. New York: Academic Press, 41–58. ——— (1989): Studies in the way of words. Cambridge, MA: Harvard University Press. Gutzmann, Daniel (2007): “Eine Implikatur konventioneller Art: der Dativus Ethicus”. Linguistische Berichte 211, 277–308. URL: http://www.ingentaconnect.com/ content/buske/lingber/2007/00002007/00000211/art00002. ——— (2008): On the Interaction between Modal Particles and Sentence Mood in German. MA thesis. Mainz: Johannes Gutenberg-Universität. ——— (2013): “Expressives and beyond. An introduction into varieties of use-conditional meaning”. In: Gutzmann, Daniel & Hans-Martin Gärtner, eds.: Beyond expressives. Explorations in use-conditional meaning. Leiden: Brill, 1–58 [This volume]. Halevy, Rivka (2007): “The subject co-referential l-pronouns in Hebrew”. In: Bar, Tali & Eran Cohen, eds.: Studies in Semitic and general linguistics in honor of Gideon Goldenber. Münster: Ugarit-Verlag, 299–321. Hall, Alison (2007): “Do discourse connectives encode concepts or procedures”. Lingua 117, 149–174. DOI: dx.doi.org/10.1016/j.lingua.2005.10.003.

expressive meaning, free datives, and f-implicature

195

Harris, Jesse A. & Christopher Potts (2009a): “Perspective-shifting with appositives and expressives”. Linguistics and Philosophy 32, 523–552. DOI: dx.doi.org/10 .1007/s10988-010-9070-5. Herschensohn, Julia (1992): “On the economy of Romance nonlexical datives”. In: Hirschbühler, Paul & E.F. Konrad Koerner, eds.: Romance languages and modern linguistic theory. Amsterdam: John Benjamins, 123–134. Hole, Daniel (2006): “Extra argumentality—affectees, landmarks, and voice”. Linguistics 44, 383–424. DOI: dx.doi.org/10.1515/LING.2006.014. Hole, Daniel & André Meinunger & Werner Abraham, eds. (2006): Datives and other cases: Between argument structure and event structure. Amsterdam: John Benjamins. Hom, Christoper (2008): “The semantics of racial epithets”. Journal of Philosophy 105, 416–440. Horn, Laurence R. (1989): A natural history of negation. Chicago: University of Chicago Press. ——— (1992): “The said and the unsaid”. Proceedings of SALT 2, 163–192. ——— (2000): “Pick a theory (not just any theory): Indiscriminatives and the freechoice indefinite”. In: Horn, Laurence R. & Yasuhiko Kato, eds.: Negation and polarity. Oxford: Oxford University Press, 147–192. ——— (2002): “Assertoric inertia and NPI licensing”. Proceedings of CLS 38 (Vol. 2), 55–82. ——— (2007): “Toward a Fregean pragmatics. Voraussetzung, Nebengedanke, Andeutung”. In: Horn, Laurence R. & Istvan Kecskes, eds.: Explorations in pragmatics. Linguistic, cognitive, and intercultural aspects. Berlin: de Gruyter, 39–69. ——— (2008): “I love me some him. The landscape of non-argument datives”. In: Bonami, Olivier & Patricia Cabredo Hofherr, eds.: Empirical Issues in Syntax and Semantics 7, 169–192. ——— (2009): “Only XL. The assertoric asymmetry of exponibles”. Proceedings of SALT 19, 198–212. URL: http://elanguage.net/journals/salt/article/view/19.12/ 1871. Horn, Laurence R. & Barbara Abbott (2012): “⟨the, a⟩: (In)definiteness and implicature”. In: Kabasenche, William & Michael O’Rourke & Matthew Slater, eds.: Topics in contemporary philosophy, vol. 10: Reference and referring. Cambridge, MA: MIT Press, 325–355. Hutchinson, Corinne & Grant Armstrong (to appear): “The syntax and semantics of personal datives in Appalachian English”. In: Zanuttini, Raffaella & Laurence Horn, eds.: Microsyntactic variation in North American English. Oxford: Oxford University Press. Iten, Corinne (2005): Linguistic meaning, truth conditions, and relevance. Houndsmill: Palgrave Macmillan. Jayez, Jacques (2004): “Presuppositions and pedigrees for discourse markers”. Empirical Issues in Syntax and Semantics 5, 89–110. URL: http://www.cssp.cnrs.fr/ eiss5/jayez/jayez-eiss5.pdf. Jayez, Jacques & Corinne Rossari (2004): “Parentheticals as conventional implicatures”. In: Corblin, Francis & Henriëtte de Swart, eds.: Handbook of French semantics. Stanford, CA: CSLI, 211–229. Kaplan, David (1999): “The meaning of ouch and oops. Explorations in the theory of meaning as use”. Ms. University of California at Los Angeles.

196

laurence r. horn

Karttunen, Lauri (1973): “Presuppositions of compound sentences”. Linguistic Inquiry 4, 169–193. URL: http://www.jstor.org/stable/4177763. Karttunen, Lauri & Stanley Peters (1979): “Conventional implicature”. In: Oh, ChoonKyu & David A. Dinneen, eds.: Presupposition. New York: Academic Press, 1–56. Kay, Paul (1990): “EVEN”. Linguistics and Philosophy 13, 59–111. DOI: dx.doi.org/10 .1007/BF00630517. Keenan, Edward (1971): “Two kinds of presupposition in natural language”. In: Fillmore, Charles J. & D. Terence Langendoen, eds.: Studies in linguistic semantics. New York: Holt, 45–54. ——— (2003): “An historical explanation of some binding theoretical facts in English”. In: Moore, John & Maria Polinsky, eds.: The nature of explanation in linguistic theory. Stanford: CSLI, 152–189. Kim, Jung-Bok & Peter Sells (2007): “Korean honorification. A kind of expressive meaning”. Journal of East Asian Linguistics 16, 303–336. DOI: dx.doi.org/10.1007/ s10831-007-9014-4. Lambert, Silke (2007): “Pragmatic and stylistic relevance of free datives in German”. Ms. Buffalo, NY. ——— (2010): Beyond recipients. Towards a typology of dative uses. PhD thesis. University of Buffalo. URL: http://linguistics.buffalo.edu/ graduate/phd/recent_ dissertations/. Lambrecht, Knud (1994): Information structure and sentence form. Cambridge: Cambridge University Press. Lamiroy, Béatrice & Nicole Delbecque (1998): “The possessive dative in Romance and Germanic”. In: van Langendonck, Willy & William van Belle, eds.: The dative. Vol. 2: Theoretical and contrastive studies. Amsterdam: John Benjamins, 29–74. Landau, Idan (1998): “Possessor raising and the structure of VP”. Lingua 107, 1–37. DOI: dx.doi.org/10.1016/S0024-3841(98)00025-4. Leclère, Christian (1976): “Datifs syntaxiques et datif éthique”. In: Chevalier, JeanClaude & Maurice Gross, eds.: Méthodes en grammaire française. Legate, Julie (2001): “A high applicative, a low applicative, and an ‘applicative’”. Paper presented at MIT Appl Fest. Cambridge, MA. Levinson, Stephen C. (1983): Pragmatics. Cambridge: Cambridge University Press. Linzen, Tal (2008): “Hebrew possessive datives: an affectedness-based approach”. Ms. Tel Aviv University. Ludlow, Peter (2012): “Descriptions”. In: Russell, Gillian & Delia Graff Fara, eds.: Routledge companion to the philosophy of language. New York: Routledge, 380– 391. Maling, Joan (2001): “Dative: The heterogeneity of the mapping among morphological case, grammatical functions, and thematic roles”. Lingua 111, 419–464. DOI: dx.doi.org/10.1016/S0024-3841(00)00039-5. Matthewson, Lisa & Henry Davis & Hotze Rullmann (2007): “Evidentials as epistemic modals: Evidence from St’át’imcets”. Linguistic Variation Yearbook 7, 201– 254. URL: http://www.ingentaconnect.com/ content/ jbp / livy / 2007 / 00000007 / 00000001/art00007. McCready, Eric (2004): “Two Japanese adverbials and expressive content”. Proceedings of SALT 14, 163–178. URL: http://elanguage.net/journals/salt/article/view/14 .163/1740.

expressive meaning, free datives, and f-implicature

197

——— (2009): “What man does”. Linguistics and Philosophy 31, 671–724. DOI: dx.doi .org/10.1007/s10988-009-9052-7. ——— (2010): “Varieties of conventional implicature”. Semantics & Pragmatics 3, 1–57. DOI: dx.doi.org/10.3765/sp.3.8. McIntyre, Andrew (2006): “The interpretation of German datives and English have”. In: Hole, Daniel & André Meinunger & Werner Abraham, eds.: Datives and other cases. Between argument structure and event structure. Amsterdam: John Benjamins, 185–212. Meibauer, Jörg (1994): Modaler Kontrast und konzeptuelle Verschiebung. Studien zur Syntax und Semantik deutscher Modalpartikeln. Tübingen: Niemeyer. Merin, Arthur (1999): “Information, relevance, and social decisionmaking”. In: Moss, Lawrence S. & Jonathan Ginzburg & Maarten de Rijke, eds.: Logic, language and computation. Vol. 2. Stanford: CSLI, 179–221. Mosegard Hansen, Maj-Britt (2003): “Review of [Blakemore 2002]”. Acta Linguistica Hafniensia 35, 190–206. Mühlhäusler, Peter & Rom Harré (1990): Pronouns and people. The linguistic construction of social and personal identity. Oxford: Blackwell. Muraoka, Takamitsu (1978): “On the so-called dativus ethicus in Hebrew”. Journal of Theological Studies 29, 495–498. DOI: dx.doi.org/10.1093/jts/XXIX.2.495. Neale, Stephen (1992): “Paul Grice and the philosophy of language”. Linguistics and Philosophy 15, 509–559. DOI: dx.doi.org/10.1007/BF00630629. ——— (1999): “Coloring and composition”. In: Stainton, Robert & Kumiko Murasugi, eds.: Philosophy and linguistics. Boulder, CO: Westview, 35–82. ——— (2001): “Implicature and colouring”. In: Consenza, Giovanna, ed.: Paul Grice’s heritage. Rome: Brepols, 135–180. O’Connor, Mary Catherine (2007): “External possession and utterance interpretation: a crosslinguistic exploration”. Linguistics 45, 577–613. DOI: dx.doi.org/10 .1515/LING.2007.018. Pollard, Carl & Ivan A. Sag (1992): “Anaphors in English and the scope of binding theory”. Linguistic Inquiry 23, 261–303. URL: http://www.jstor.org/stable/4178768. Pons Bordería, Salvador (2008): “Do discourse markers exist? On the treatment of discourse markers in relevance theory”. Journal of Pragmatics 40, 1411–1434. DOI: dx.doi.org/10.1016/j.pragma.2008.03.013. Potts, Christopher (2005): The logic of conventional implicatures. Oxford: Oxford University Press. ——— (2007): “The expressive dimension”. Theoretical Linguistics 33, 165–197. DOI: dx.doi.org/10.1515/TL.2007.011. ——— (2010): “On the negativity of negation”. Proceedings of SALT 20, 636–659. URL: http://elanguage.net/journals/salt/article/view/20.636/1414. Potts, Christopher & Shigeto Kawahara (2004): “Japanese honorifics as emotive definite descriptions”. Proceedings of SALT 14, 235–254. URL: http://elanguage .net/journals/salt/article/view/14.253. Predelli, Stefano (2003): “Scare quotes and their relation to other semantic issues”. Linguistics and Philosophy 26, 1–28. DOI: dx.doi.org/10.1023/A:102227 8209949. Reinhart, Tanya & Eric Reuland (1993): “Reflexivity”. Linguistic Inquiry 24, 657–720. URL: http://www.jstor.org/stable/4178836. Reuland, Eric (2001): “Primitives of binding”. Linguistic Inquiry 32, 439–492. DOI: dx .doi.org/10.1162/002438901750372522.

198

laurence r. horn

Rieber, Stephen (1997): “Conventional implicatures as tacit performatives”. Linguistics and Philosophy 20, 51–72. DOI: dx.doi.org/10.1023/A:1005383403910. Salmon, William (2011): “Conventional implicature, presupposition, and the meaning of must”. Journal of Pragmatics 43, 3416–3430. DOI: dx.doi.org/10.1016/j .pragma.2011.07.011. Sauerland, Uli (2007): “Beyond unpluggability”. Theoretical Linguistics 33, 231–236. DOI: dx.doi.org/10.1515/TL.2007.016. Saul, Jennifer (2002): “Speaker meaning, what is said and what is implicated”. Noûs 36, 228–248. DOI: dx.doi.org/10.1111/1468-0068.00369. Sawada, Osamu (2010): “The multidimensionality of the Japanese minimizers sukoshi/chotto ‘a little’”. Paper presented at annual meeting of the Linguistic Society of America, Baltimore. Schmid, Josef (1988): Untersuchungen zum sogenannten freien Dativ in der Gegenwartssprache und auf Vorstufen des heutigen Deutsch. Frankfurt: Peter Lang. Schwenter, Scott (2002): “Additive particles and scalar endpoint marking”. Belgian Journal of Linguistics 16, 119–133. DOI: dx.doi.org/10.1075/bjl.16.09sch. Simpson, Jane (1991): Warlpiri morpho-syntax: A lexicalist approach. Dordrecht: Kluwer. Sroda, Mary Sue & Margaret Mishoe (1995): “‘I jus like to look at me some goats’: Dialectal pronominals in Southern English”. Handout for NWAV 24 presentation. Stanley, Jason (2002): “Modality and what is said”. In: Tomberlin, James E., ed.: Philosophical perspectives 16: Language and mind. Oxford: Blackwell, 321–344. Sweet, Henry (1900): A new English grammar, logical and historical. Part I. Oxford: Oxford University Press. Taavitsainen, Irma & Andreas Jucker, eds. (2003): Diachronic perspectives on address term systems. Amsterdam: John Benjamins. Tsohatzidis, Savas (1992): “Pronouns of address and truth conditions”. Linguistics 30, 569–575. DOI: dx.doi.org/10.1515/ling.1992.30.3.561. Vallée, Richard (2008): “Conventional implicature revisited”. Journal of Pragmatics 40, 407–430. DOI: dx.doi.org/10.1016/j.pragma.2007.10.006. Ward, Gregory (2003): “Comments on ‘Felicity and presupposition triggers’ by Michael Glanzberg”. Linguistics & Philosophy Workshop, Ann Arbor. Webelhuth, Gert & Clare Dannenberg (2006): “Southern American personal datives: The theoretical significance of syntactic variation”. American Speech 81, 31–55. DOI: dx.doi.org/10.1215/00031283-2006-002. Wegener, Heide (1989): “Eine Modalpartikel besonderer Art: Der Dativus Ethicus”. In: Weydt, Harald, ed.: Sprechen mit Partikeln. Berlin: de Gruyter, 56–73. ——— (2002): “The evolution of the German modal particle denn”. In: Wischer, Ilse & Gabriele Diewald, eds.: New reflections on grammaticalization. Amsterdam: John Benjamins, 379–394. Williamson, Timothy (2003): “Understanding and inference”. Aristotelian Society Supplement 77, 249–293. URL: http://www.jstor.org/stable/4106998. ——— (2009): “Reference, inference and the semantics of pejoratives”. In: Almog, Joseph & Paolo Leonardi, eds.: The philosophy of David Kaplan. Oxford: Oxford University Press, 137–158. Wilson, Deirdre (1975): Presuppositions and non-truth-conditional semantics. New York: Academic Press.

expressive meaning, free datives, and f-implicature

199

Wolfram, Walt & Natalie Schilling-Estes (1998): American English. Dialects and variation. Oxford: Blackwell. Zimmermann, Thomas Ede (1997): “The addressing puzzle”. In: Künne, Wolfgang & Albert Newen & Martin Anduschus, eds.: Direct reference, indexicality, and propositional attitudes. Stanford: CSLI, 133–153.

GOOD REASONS

Eric McCready and Yohei Takahashi

1. Introduction The notion of conventional implicature as a distinct class of pragmatic meanings is due to Grice (1975), in a brief discussion. He provides only a single example, that of therefore, which he takes to be a connective semantically like and but carrying an additional (conventional) implicature of consequence: (1)

That book is a bestseller. Therefore, it must be interesting. ≈ It follows from the book being a bestseller that it must be interesting.

However, Grice’s characterization has been disputed by various authors in the recent upsurge of interest in conventional implicatures (e.g. Bach 1999; Potts 2005). Potts in particular argues for the existence of a class of conventionally implicated meanings, exemplified by (definite) appositive clauses and parentheticals, which share certain characteristics that therefore in part lacks.1 Bach, even more radically, disputes that a class of conventional implicatures exists at all, though the dispute may be in part a terminological one. However, the observation that therefore is not conventionally implicating (assuming that Potts is correct)2 does not mean that no connectives are. Scheffler (2005), for instance, argues that German denn has a conventionally implicating component. Our goal in this paper is to provide data relating to a connective in Japanese that similarly contributes expressive content.3

1 These characteristics will be reviewed below, in connection with our argument that the Japanese connective mono, which is the subject of this paper, carries conventionally implicated or expressive content. 2 We do not endorse the strong claim of Bach. 3 We will talk about “expressive content” in this paper with the intent of using the term to cover what Potts and others have termed conventional implicature. It is not clear to us where the dividing line between these content types lies. Our use of “expressive” here is roughly parallel to McCready’s (2010) use of the term “CIE (conventionally implicating or expressive) content”.

202

eric mccready and yohei takahashi

This connective, mono, indicates a causal relation between the two propositions it connects, but also expresses an emotive attitude of the speaker toward the causing proposition, and further expresses a judgement about the quality of the causal relation in question. We provide data about this connective in Section 2 and discuss some previous (informal) analyses from the Japanese-language literature in Section 5. In Section 3, we turn to the analysis of the three components of the meaning of mono: in brief, we take it to introduce a causal relation (which we state in terms of probabilities), a context-dependent emotive attitude, and to indicate that the causal relation holding between the propositions it connects is a good one. This last component is cashed out in terms of quality of reasons. Section 4 shows how to implement this analysis in the resource logic proposed by McCready (2010), building on the system of Potts 2005. Section 6 concludes. 2. Mono: The Facts We begin by exhibiting some known facts about mono from the Japaneselanguage linguistic literature, which will give the reader a sense of the content of mono and of the basic data we wish to characterize in this paper. Additional data will be introduced as the paper proceeds. It is generally recognized that mono carries a meaning of causality related to an event given in the preceding context.4 This usage can be exemplified as follows.5 (2)

A: aa, mata shiken ochite-shimatta ah again test failed-antihon.perf ‘Damn, I failed the test again …’ B: omae-ga benkyoo shinai nda mon you-Nom study do.not noda mono ‘That’s because you don’t study.’

4

For example, the Shin-Meikai-Daijiten dictionary defines mono as follows: Speaker justifies himself or blames hearers by making a claim, namely one in the form of a ‘reason’, about the topic raised in a discourse (Kindaichi et al. 1991: 1289).

5 A reviewer asks whether this causal meaning comes from the connective or can simply be inferred from the content of the sentences, in the manner of SDRT (Asher & Lascarides 2003). Such inferences are of course available. The presence of mono, like other causal connectives, makes the “inference” a monotonic one. However, it should be noted that in Japanese the use of connectives and other particles to indicate discourse relations is highly preferred, and perhaps even obligatory to induce the correct interpretation in certain contexts. This is a difficult issue and one that we cannot address in detail here.

good reasons

203

Here, the mono utterance of B serves to explain the content of A’s utterance, which directly precedes it in the dialogue. Hashimoto (1997) calls this usage of mono “Reason-Explanation”. The following are some additional examples from the literature (Tsubone 1996). (3)

ashita, umi-e iku-no-o yamechatta. ame-ga furisooda mono. tomorrow sea-to go-nm-acc cancelled rain-nom fall.may mono ‘I decided not to go to the sea. Because it will rain tomorrow.’ (Tsubone 1996: T’s (3a))

(4)

A: dooshite ookikunat-tara pairotto-ni naritai no? why get.big-when pilot-acc want.become q ‘Why do you want to be a pilot when you grow up?’ B: datte, kakkoii mono. come.on, cool mono ‘Well, because it’s cool.’

(Tsubone 1996: T’s (2b))

As is obvious from the instances above, the mono utterances above provide the reason or cause concerning the event depicted in the preceding context, mediating two events in terms of a causal relation. We have translated mono as “because” in the above. But does it differ in any way from that more standard causal connective? Above we suggested that it does, and that the difference is that it carries expressive content. Japanese indeed has a ‘plain vanilla’ causal connective, kara, which is intuitively quite different in quality from mono, to the extent that the latter has often been viewed as not being a connective at all, but rather a sentencefinal modal particle like the Japanese yo (Davis 2009; McCready 2009) or English man (McCready 2009).6 We will argue below that this characterization is not correct, but for now we only note the initial impression that is given by mono: first, that the sentence it appears in denotes a proposition which is a (or the) cause of a proposition that is salient in the context, usually because of being the denotation of the sentence that appears immediately prior to the instance of mono, and, second, that the speaker takes this causal relation to be highly evident or conclusive, i.e., to form a high quality reason for whatever is at issue. This initial characterization must be supplemented slightly, however, as is made clearer by the second use of mono discussed in the literature. The second type of mono usage is called “Exclamation” by Hashimoto, who defines this type as follows: “this group of mono utterances are used to let an addressee strongly recognize the addresser’s position or emotion 6

Note that kara, like mono, can appear with only a single clausal argument.

204

eric mccready and yohei takahashi

when the addresser thinks that the addressee does not understand thoroughly his opinion or feeling” (Hashimoto 1997: 210). As Hashimoto mentions, this use of mono is usually associated with children and young women, though the association with the latter group is far less strong, and seems to be weakening with the passage of time.7 An example is the following: (5)

A: ninjin-o tabenasai! carrot-acc eat.imp ‘Eat your carrot!’ B: yada mon! zettaini tabetaku-nai mon! no.way mono absolutely want.eat-neg mono ‘No! I don’t want to eat it!’

Hashimoto argues that this type of mono utterance should be differentiated from a more standard case of providing a reason in that in this case the speaker emphasizes his feelings rather than providing an addressee with a reason. But it seems pretty clear that a reason is indeed being provided here; it is just that this reason happens to consist of the speaker’s personal tastes and feelings. However, we think Hashimoto’s intuition has some bite. In fact, a major difference between mono and kara (for example) is that mono provides an emotive quality to the reason-giving: the speaker gives the impression of having some sort of emotive reaction to the proposition that serves as cause, so that the proposition is significant to her in some way. One other observation about mono that is notable is that it does not embed. It cannot appear in the scope of negation, conditionals, modals, or other semantic operators. We show examples involving negation, an epistemic modal, and a belief operator. This has consequences for how it should be analyzed semantically; as we will argue, these facts are evidence that mono carries a meaning that is, in part, expressive in nature.8 (6)

*omae-ga benkyoo shinai mon janai you-nom study do.not mono cop.neg ‘It’s not because you don’t study.’

7 It is interesting to speculate on why this might be. We would like to suggest that it follows from the way we characterize the particle’s meaning: mono marks a reason which is, subjectively, of high quality. It seems plausible that taking one’s feelings and attitudes as purely high quality reasons might be a childish act; and, if a particular group is allowed by a society to act childishly in some contexts (at least as a matter of social norms), it might be associated with that group too. 8 (8a) is grammatical on an irrelevant and somewhat incoherent reading on which mono is understood as a homophonic term meaning ‘thing’. We ignore this reading here and in what follows.

good reasons (7)

*Taro-ga asobi-sugita mon kamoshirenai Taro-nom play-too.much mono might ‘It might be because Taro played too much.’

(8)

*Watashi-wa aitsu-ga warui mono da to omou I-top [he-nom bad mono cop] c think ‘I think it’s because he’s bad.’

205

Mono also cannot appear in questions or imperatives, as shown in the following examples.9 Here, it differs from the vanilla connective kara and also from English therefore, which are grammatical in yes-no questions, though all three types are bad in imperatives. (9)

*ano sake umakatta mono (desu) ka that sake tasty mono (cop) q ‘Was it because that sake was tasty?’

(10) *hayaku shiro mon! quickly do mono ‘Because hurry the hell up!’

The analysis we propose in the next section has as its parts the three components of the meaning of mono motivated above. We turn to specifying these now, returning to issues concerning the Japanese literature and additional data in Section 5. 3. Analysis 1: Evaluating (and) Reasons In this section we will argue that the meaning of mono has three components. The first is that of a causal connective like because. The second is an emotive component indicating that the speaker holds an emotive attitude toward the proposition which forms the first (causing) argument of the causal relation: that is, the immediate sister of mono, its first argument in composition.10 The third is that the speaker regards this first argument not only as forming a cause, but as comprising a good reason for the caused proposition; we spell this out in terms of probabilities, which make it easy to compare the degree to which the truth of one proposition influences the likelihood of truth of another.11 The choice of a probabilistic 9 Analogues of (9) which use different question formation strategies are also ungrammatical. 10 We will not take a position on the syntax of these elements here. 11 As with knowledge and assertion, we must understand the notion of likelihood of truth here as one which is, while viewed as objective, ultimately a subjective judgement on the part

206

eric mccready and yohei takahashi

analysis of causation is made more or less for expository reasons. We indicate a possible alternative in 3.4. 3.1. Causation Since mono is a causal connective, we must make use of a notion of causation in our theory. Here, we review briefly two theories of causation from the philosophical literature, the counterfactual theory and the probabilistic theory. In this paper, we will make use of a version of the probabilistic theory, only because it comes with a property necessary for our analysis already built in: a (total) ordering of degrees of causation. Probabilistic theories of causation take it that causal dependencies should be stated in terms of probabilistic dependencies, so that A is taken to be a cause of B if the truth of A raises the probability of the truth of B. Some prominent expositions of this general view are Skryms (1980) and Eells (1991).12 On this view, causation by B of A is analyzed as increase in probability of A given that B as opposed to ¬B. The usual way of expressing this is in terms of conditional probabilities, as follows, where 𝒫 is a probability function obeying the usual conditions (additivity, etc.; see McCready & Ogata 2007 for one formulation) and 𝒫(p|q) indicates the probability of p conditional on q:13 (11)

𝒫(A|B) > 𝒫(A|¬B) ⟷ Cause(B, A)

This is extremely simple, in fact too simple: it ignores a number of complications familiar from the literature on probabilistic causation. For example, what happens in cases where there is a correlation in probabilities between A and B, but no causal connection, or in cases of “causal transitivity”?14 We put these problems aside for the purposes of our semantic analysis. As we

of the speaker (Williamson 2000). Thus, the notion of “good reason” for us is a normative one, though what counts as a good reason may vary across speakers due to the use of subjective probabilities. We think that this is a reasonable outcome. 12 In this paper, we ignore the standard problems for such theories (e.g. spurious correlations, cases of non-causes that increase probabilities of ‘effects’, and so on). 13 Here is one among many equivalent definitions of conditional probability: 𝒫(p|q) =

𝒫(p ∧ q) . 𝒫(q)

See e.g. Jeffrey 1983 for interesting discussion of related issues. 14 For example, my wakefulness may cause me to have a higher blood pressure than when I am sleeping; reading a book entails that I am awake; but my reading may not, intuitively, cause me to have a higher blood pressure than when I am sleeping.

good reasons

207

will detail in the following sections, our choice to use probabilistic causation in our definition of causal connectives allows a simple definition of quality of reasons in terms of the degree to which learning the reason induces increase in the likelihood of truth of what is being explained, which can easily be put in terms of the standard Bayesian notion of conditional probability. Another possibility would be to use a counterfactual analysis of causation. According to such theories, B is a cause of A to (roughly) the extent that sentences of the form “If B had been the case, A would have been the case” are true.15 Theories of this kind are certainly more well-known to the linguistic community than probabilistic analyses. However, in this paper, we will analyze mono as indicating quality of reasons, which makes it necessary to compare the quality of such reasons; this need, in turn, requires an ordering on degrees of causation, on the assumption that such an ordering allows comparison of reasons. These considerations lead us to make use of the probabilistic analysis in what follows. It would also have been possible to use a sophisticated version of a counterfactual analysis of causation, which in fact makes available a notion of degree: such an analysis is given by Lewis (2004) (building on the analysis of Lewis 1973a,b), who takes an event A to be a cause of another event B to the degree that A influences B, which is the case roughly if alterations in A result in alterations in B. This relational notion translates easily to comparisons, in that one can abstract scales of similarity from it, though the result is highly multidimensional, as noted by Lewis (2004:92); this multidimensionality might not be so problematic, however, if the techniques discussed by van Rooy (2010) are applied. However, spelling all this out would take us too far afield as well as introducing technical complexity that we feel is not strictly necessary for the current application; this is one major reason we use a simple analysis in terms of conditional probability. 3.2. Causality in Discourse Connectives In the previous section we discussed probabilistic analyses of causation, on which B is a cause of A just in case the probability of A is higher given B than given ¬B. This can be thought of as follows: 15 Such theories also have well-known problems, often reminiscent of problems with analyses of progressive aspect (Dowty 1979; Landman 1992; Smith 1997), where notions of “inertia worlds” and normality come into play. We will not consider these issues deeply here. Extensive discussion can be found in Collins et al. 2004.

208

eric mccready and yohei takahashi

(12) 𝒫(A|B) − 𝒫(A|¬B) > 0

This is the basic notion of causation we will assume going forward. However, (as a reviewer notes) this is an extraordinarily weak notion; even the tiniest increase in probability of A given B will be sufficient to make it true. We will therefore strengthen it in such a way that not just any increase in probability counts as a causing, but only increases that are sufficiently large. Exactly how much increase is required to count as sufficiently large is, we think, a contextually dependent notion, which we will implement by setting a degree which serves as a contextual standard for causation. The result of this change is the following: (13) 𝒫(A|B) − 𝒫(A|¬B) > stndc

where stndc represents a contextually determined standard, also a real number ∈ [0, 1], to which the gain in likelihood of q after conditionalization on p is compared. If the gain is greater than the standard, p can be considered a cause of q, and if the gain is not large enough, it cannot be so considered. This analysis follows the lines of Kennedy’s (1999; 2007) analysis of gradable adjectives. We then amend the basic definition of causation accordingly. (14) 𝒫(A|B) − 𝒫(A|¬B) > stndc ⟷ Cause(B, A)

On this basis we take the basic semantics of causal connectives to be the following. (15) λpλq[Cause(p, q)]

This formula will be the core of the semantics of mono we propose; indeed, the asserted or at-issue content of mono will be precisely the content in (15). The reader may be wondering: in the examples we have seen of mono, it appeared to modify only a single sentence. Why is it being modeled as a two-place predicate? The reason is that it has a causal meaning, which necessarily involves two arguments; in the discourses we have seen, the proposition being explained is already present in the discourse at the point mono is used. This is a quite general characteristic of the use of this lexical item. Thus, it is a connective in the sense that it indicates a causal connection between two propositions; it is more like a sentence-final particle in that it always appears modifying only a single sentential unit. (We believe that this is the reason it has been thought of as a true particle in the literature, though we dispute this characterization, as its meaning shows other aspects not at all characteristic of true modal particles—namely causativity.) Its semantics is therefore more like the “anaphoric connectives” discussed by Webber et al. (2003). In this paper, we will treat it as a true causal connective, one

good reasons

209

of whose arguments is made available by context. A fuller treatment would take facts about the discourse into account, ideally in a way that made principled predictions about possible anaphoric arguments in the way done in, for example, SDRT (Asher & Lascarides 2003). Space limitations preclude carrying out this analysis in the present paper, though a sketch might go something like the following: One would take mono to introduce a causal relation, namely a modified version of Explanation, taking the speech act associated with the proposition it modifies as argument. This speech act discourse referent would then be attached to the caused proposition via the usual sort of inference in the SDRT default logic, though in many cases a new proposition would need to be constructed, for example in cases where the first argument is a question such as Why can’t I get a job? (an issue we have glossed over in the above). The content we ascribe to mono here would then be the truth-conditional effect of inferring the modified Explanation. The whole package is complicated, and we do not have space to work it out here; however, in the conclusion, we indicate another reason to think that this kind of analysis should be worked out in detail. 3.3. Emotivity As we have seen, a use of mono(φ) carries the implication that the speaker is not neutral about the fact expressed by φ, instead finding it to be of subjective interest. Note, for example, that it is infelicitious to follow an utterance of a mono-sentence with a statement like (16), while it is fine to do so after an utterance including the vanilla connective kara; this indicates that the speaker must take the content marked by mono to be directly significant. (16) doo-demo ii kedo ne whatever good though pt ‘Whatever, it doesn’t really matter though.’

In other words, mono has an evaluative component to its semantics. We characterize this component as an emotive one, taking the speaker to be judging the content of φ as good, bad, or simply emotive in a polarityunspecific way (Wow!). For the analysis of this part of the meaning of mono, we will make use of an ‘emotive function’ proposed for the analysis of expressive content by McCready (2004, 2009, 2010). This function derives emotive attitudes from Kaplanian contexts (Kaplan 1989) construed as tuples of the speaker, hearer, time, world, and location of an utterance (as well as possibly many other coordinates). The idea is that these factors will determine what

210

eric mccready and yohei takahashi

attitude a speaker can be taken to hold toward the proposition that the function takes as argument: propositions that describe things normally taken as negative, or taken as negative by the speaker, yield the value bad, and similarly for the other possible values.16 The needed function is one which maps contexts to functions from propositions to emotional predicates (where “!” indicates emotivity unspecified for polarity): (17) E : c ↦ (℘(W) ↦ A), where A ∈ {bad, good, !}.

Here bad, good and ! are of type ⟨⟨s, t⟩, t⟩: functions from propositions into truth-values. Thus E maps contexts to functions from propositions into evaluative predicates of propositions. The question of which predicate exactly is returned by E is a difficult one. The function depends on context: the agent, addressee, time, and world (etc.) of the speech act determine what content is to be returned. The earlier work using this function (McCready 2004, 2009b, 2010) leaves the manner in which this is to be done highly underspecified. The intuition there was that these factors, together with the propositional content, determine an attitude; this determination, in fact, has a normative character, in that the attitude which can be expected to hold is dependent on how one would ordinarily react to the propositional content in question, though this can be overriden by other contextual factors (which, of course, is one reason for using an analysis of this kind). One possibility for full elucidation of the mechanism would therefore be to use a nonmonotonic logic which makes use of a notion of normality, for instance that of Asher & Morreau (1991). A further elaboration of this analysis would be to allow the interpreter to reason about the speaker’s probable attitudes, and about the speaker’s probable guesses about the reasoning the hearer would likely engage in, and so put the whole into a game-theoretic context. This view is worked out in detail in McCready 2012, but, for space reasons, we will use the earlier functional picture here. This part of the analysis of mono, therefore, cannot be called fully complete; but it suffices to show the character of the evaluative component of the connective. Should we conjoin the evaluative component of mono to the causal portion from the previous section, yielding the following formula? (18) λpλq[Cause(p, q) ∧ E p (p)]c

16 Compare recent work on evaluativity in exclamatives, such as Rett 2008. A reviewer comments that we may be mixing up emotivity and statements about expectation, which is one way to characterize the “!” operator. Here, we intend this operator not as a statement about probabilities or unexpectedness, but as something usable when the context does not allow recovery of a particular emotion.

good reasons

211

We should not. The reason is that the evaluative component of mono is not part of its at-issue content. The distinction between at-issue and expressive content is one that has sparked a good deal of recent research (Amaral et al. 2007; Potts 2005, 2007c, i.a.). Here, we will not take sides on these issues or the significance they might or might not have for defining ‘what is said’. Instead, we will focus on several tests for expressiveness that we think are reliable, at least when limited to certain domains.17 The particular tests are (i) unembeddability/independence and (ii) denials. We consider each in turn. The interpretation of test (i) is somewhat author-dependent. In the original formulation of Potts 2005 and in most subsequent work, independence of expressive content is understood as its invariable projection from operators in the (at-issue) semantics. For example, the following appositive is not interpreted in the scope of the external negation. (19) It is not true that John, (although) a golfer, is particularly rich.

Some authors (such as Scheffler 2005) instead understand this test in a stronger way: as ungrammaticality when an expressive content-bearing expression is embedded, due to the requirement for semantic projection. This is what we find in the case of mono, which cannot appear in the scope of negation, in conditionals, under modals, etc., as mentioned in Section 2. For space reasons, we repeat only the negation case from (6). Note the contrast with the non-expressive causal connective kara, which can easily be placed in the scope of negation, much like the English because. (20) a. *omae-ga benkyoo shinainda mon janai you-Nom study do.not mono cop.neg ‘It’s not because you don’t study.’ (+ CIE content) b. omae-ga benkyoo shinai kara janai you-nom study do.not because cop.neg ‘It’s not because you don’t study.’

Following the reasoning of Scheffler (2005), the explanation for this might be the following: since the expressive part of the content projects, that content is entailed by the sentence to be true, in a sense of entailment sufficiently broad to include expressive content. Suppose (as we will argue in a moment) that the expressive content of (20a) is a) that the speaker has an emotive attitude toward the addressee’s not studying, and b) that the lack 17 In particular, we will avoid embeddings under propositional attitudes, as it turns out to be easy to get (mixed) quotative interpretations there (Harris & Potts 2009).

212

eric mccready and yohei takahashi

of studying is a good reason for whatever is to be explained by the sentence, according to the speaker. Then it is entailed that the addressee’s not studying is a good reason for whatever is to be explained. But then it must be a reason, and so a cause;18 and so the negation is incoherent, for it is asserted that the addressee’s not studying is not a cause. Similar reasoning, modulo explanations based on Gricean Quantity instead of contradiction resulting in incoherence (explainable via an unavoidable Quality violation), applies to embeddings in conditionals, modals, attitudes, etc. The independence facts can thus be taken as an initial piece of evidence for a (partly) expressive status for mono. We now turn to test (ii). According to this test, expressive content cannot be targeted by denials, or, at least, not denials that dispute the truth of utterances. Consider the following example. (21) A: John, the banker, is pretty rich. B: That’s not true.

We follow Potts (2005) in assuming that the appositive carries a conventional implicature with content banker( j). It is an open question whether these meanings can be characterized as expressive, but we will put these issues aside here; their behavior in composition is very similar. The at-issue content of A’s utterance further has at-issue content rich( j). Now: which of these propositions can be targeted by B’s denial? It seems pretty clear that B’s utterance can only be interpreted as denying that John is rich, not the content of the appositive. Thus, only at-issue (non-expressive) content can be targeted by (simple) truth-oriented denials which lack metalinguistic interpretations.19 What happens in the case of mono? Consider the following dialogue. (22) A: omae-ga benkyoo shinai nda mon you-nom study do.not noda mono

18 In the sense we use “reason” here, at least, which is restricted to causal phenomena. A caveat and an interpretative comment are necessary here. The caveat: reason is a term with broad application, and relates to topics such as practical reason/action choice (‘reason to’) in addition to causal notions. This sense of ‘reason’ is not at issue in this paper. The comment: when we talk about changes in probability, what do we refer to? For probabilistic causation to make sense, we cannot use subjective probabilities here, for, if we do, we are only talking about changes in justification for belief in this or that proposition. Since we are concerned with ‘head-external’ causal relations, this interpretation cannot be the right one. Consequently, we must use a frequentist (or propensity-based) interpretation of probability (though see Skryms 1980). Thanks to a reviewer for raising comments related to these issues. 19 See also Jayez & Rossari 2004 for more on the denial test.

good reasons

213

B: sonna koto nai (yo) that thing cop.neg (part) ‘That’s not true.’

Here, B’s utterance can mean only that her lack of studying is not the cause of whatever is at issue, or that it’s not true that she did not study. It cannot mean that A lacks the relevant attitude toward this proposition, or that the quality of the causal relation is insufficiently high. We take this to indicate that these latter aspects of the meaning must be expressive content, a conclusion supported by the result of test (i) above. It would therefore be an error to build the emotive content directly into the semantics as an at-issue conjunction. The same holds for the reason-quality part of the content discussed in the following section. Our ultimate proposal will be that both of these things should be separated out as expressive content, making mono a bearer of mixed content along the lines argued for pejoratives and certain honorifics by McCready (2010). It should be noted that the reliability of the denial test has recently been called into question by Geurts (2010), who notes that the natural language truth predicate is often applied so flexibly as to be rather unuseful as a test for what counts as at-issue content. He cites examples where statements involving what appear to be subjective judgements (e.g. Lasersohn 2005) are qualified with truth predicates, as in the following example. (23) The coffee at that place is tasty, it’s true, but the service sucks.

Geurts notes that predications like that with tasty most likely do not involve truth in a strict sense of word-to-world mapping, but instead make ineliminable reference to the judgements of the speaker. While we think that the jury is still out on the proper analysis of predicates like tasty, it can be argued that use of the truth predicate is being stretched here. However, for us, it is important to note that the speaker of (23) proffers the judgement as a truth. We believe that this is a crucial difference between expressive predications and judgement predicates. Another important difference is the primacy of the predication by tasty; expressive predicates, of course, cannot be used as main predicates. (24) *John is damn.

This might suggest that truth claims require their content to be discourseprimary (cf. Simons et al. 2010), or that the content should be offered up as something for the hearer’s evaluation, rather than as a side effect in some sense of sentence interpretation. We will not pursue these points further here.

214

eric mccready and yohei takahashi 3.4. Quality of Reasons

One element of the meaning of mono remains to be formalized. As argued above, this connective indicates not only a causal relation but that the causer represents a good reason for believing the caused proposition. The notion of a good reason is a complex one, and can be viewed in a number of ways. For instance, we might follow Searle (2001) and take reasons to be answers to (discourse-relevant) questions; on this picture, a “good reason” would be one that provides a good answer to the relevant question, where this is defined in terms of (for example) minimization of entropy across the partition induced by the question (van Rooy 2003; McCready 2009). However, we will here take a simpler view. With this, it is easy to define a notion of ‘good reason’ in terms of the extent of the conditional probability change, as follows: (25) Good_reason(p, q) ⟷ 𝒫(q|p) − 𝒫(q|¬p) > stndgr

We take it to be obvious that there is no hard-and-fast standard for what counts as a good reason, which is the reason for adopting an analysis that makes use of a contextual standard. To make clear that for something to count as a good reason requires more than for it to count as a cause, we introduce the following constraint: (26) stndgr > stndc

Thus, the interval between the probability of A given B and the probability of A given ¬B is required to be larger for good reasons than for ordinary causation, which means that the truth of B must yield a larger increase in the probability of A in the former case, as desired. This strategy is similar to that followed by McCready 2009 for adverbial intensifiers like very: very red is taken to require for its truth a larger distance between the contextual standard for redness and the actual degree of redness of the relevant object than red alone does. We thus might, awkwardly, paraphrase good reason as very CAUSE. The question now is whether to take this indication of good reason-hood to be asserted or conventionally implicated. As we saw with the data in (22), the only content that can be targeted by denial is the causal relation itself; the emotive content was not an available target, and the quality of the reason was not either. Similarly, we saw in (20) that the statement about quality cannot be negated, and that this was indeed the reason for the ungrammaticality of sentences like (20a). The result is that the quality statement must be part of the expressive content of mono, together with the

good reasons

215

emotive component of the meaning. In the next section, we will see how to give a semantics to the connective that is faithful to this intuition. 4. Analysis 2: Mixed Content We have argued that mono is a connective that combines the semantics of a “standard” causal connective like the English because with the expressive content that the causing proposition is, subjectively, a reason which explains the effect well, in addition to being something the speaker has an emotive attitude toward. This complex combination means that to analyze the function of mono in a compositional manner, we need a system in which a single lexical item can simultaneously carry both “at-issue” and expressive content. This is something that is explicitly ruled out in the system proposed by Potts (2005) for the analysis of expressive content, because there are no such “mixed types” made available by the type specification. However, an extension of this system, ℒ+CI , has recently been proposed by McCready (2010) precisely to handle cases like the one we are presently concerned with. We will therefore adopt this system for our analysis. In this section, we first introduce the system,20 in Section 4.1, and then use it to give a lexical entry to mono, in 4.2. We finally show concretely how this lexical entry is used to derive the meanings of sentences containing mono in 4.3. 4.1. The System

The original system ℒCI is given by adding a set of special types to the standard Montagovian type system, along with rules to interpret them, and additional rules for interpretation of the resulting logical forms. The new types are called CI types and distinguished with a superscripted “c”; the “normal” types are called at-issue types and distinguished with a superscripted “a”. Application (etc.) of the at-issue types is standard; CI types are defined to interact with at-issue types in the way stated by the rule (R4).21 This rule guarantees the lack of resource sensitivity of the logic for CI content: expressive operators do not ‘use up’ the content they take as arguments. Instead, this content is duplicated and reused in the at-issue derivation. Objects of CI

20 For reasons of space, this introduction is of necessity somewhat telegraphic. The interested reader is directed to McCready 2010 for full details. 21 This formulation is taken from McCready 2010; the rule numberings follow those in that paper, where the full system is defined.

216

eric mccready and yohei takahashi

type t are removed from derivations via (R5); this ensures that the content on the root node of any tree is of at-issue type t, formalizing the intuition that CIE content does not form the primary content of utterances. (R4) (R5)

𝛼 : ⟨σ a , τ c ⟩ β : σ a 𝛼(β) : τ c • β : σ a 𝛼 : tc • β : τa β : τa

This system is highly constrained and only admits the creation of lexical objects of certain types. In particular, all lexical items are required to introduce only at-issue or expressive content, never both. For mixed-type objects like mono, ℒCI is thus too restricted. However, the simple expedient of ‘mixing’ content of CI type and at-issue type will not work for reasons related to the non-resource-sensitivity of the logic; full details are in McCready 2010, where a solution is proposed which involves adding types for resourcesensitive expressive content to ℒCI . The resulting system is called ℒ+CI . The new types are called shunting types and distinguished with a superscripted “s”. They are interpreted via the following rule, (R7). Note that this rule is distinct from ordinary functional application in that it involves objects of different type classes: the higher-typed object is a shunting type, and its argument is an ordinary at-issue type. The result is a transmutation of the at-issue typed argument into the result of functional application, which is of shunting type. (R7)

𝛼 : ⟨σ a , τ s ⟩ β : σ a 𝛼(β) : τ s

In this system, interpretation is performed in terms of entire derivations via a rule which collects all objects of type t which are of shunting or CI types and pairs them with the denotation of the root node. The result is taken to be always a pair of the form ⟨⟦𝛼 : σ a ⟧, {⟦β1 : t {c,s} ⟧, …, ⟦βn : t {c,s} ⟧}⟩, where σ {a,b} indicates that σ is of either the form σ a or σ b .22 Here, the first element of the pair represents the at-issue meaning of the sentence (utterance), its standard meaning, and the second element is a set containing any expressive content of propositional type associated with the sentence.

22 In fact, this represents an error in the McCready 2010 system: since shunting types are resource-sensitive, the possibility arises that the root node can be of type ts , which is not allowed for in the official ℒ+CI system. This problem is easily solved by allowing the root node to be of type t{a,s} .

good reasons

217

As might be clear, with the addition of shunting types it becomes possible to avoid problems associated with non-resource sensitivity and model mixed types. McCready 2010 adds a new rule for type creation which produces types of the following form:23 (27) ⟨σ, τ⟩a × ⟨ζ, 𝜐⟩s

This object is a product type where the conjoined types are an at-issue type and a shunting type. In practice, the inputs to both conjoined elements must be of the same type, though this is not required by the type creation rule itself. Mixed types like these are paired with λ-terms of the form 𝛼◆ β; “◆ ” (hereafter “diamond”) signifies a semantic object of mixed type. Such objects act as follows in composition. (R8)

𝛼◆ β : ⟨σ a , τ a ⟩ × ⟨σ a , 𝜐 s ⟩ 𝛾 : σ a 𝛼(𝛾)◆ β(𝛾) : τ a × 𝜐 s

The rule (R8) performs pointwise application of mixed-type objects on their inputs. As a reviewer notes, (R8) only allows for situations where both ‘sides’ of the mixed type take an argument of the same type. This is intentional. As far as is known at present, there are no mixed types which take arguments of different types. The reviewer suggests that the flexibility to allow for this case should be built into the system; but it seems to us preferable to restrict the available types as much as possible, insofar as the resulting system is able to characterize the operations available in natural language. In this, we follow the general strategy of Potts 2005. In any case, it is easy to loosen the rule to allow for such cases if it proves to be necessary. When no more arguments can be taken by the expressive part of the mixed type element, (R9) applies. (R9)

𝛼◆ β : σ a × t s 𝛼 : σa • β : ts

With this rule, the status of the expressive element of type t s is changed: it is no longer actively involved in the derivation. (Note, again, that this rule predicts that derivations involving shunting types always result in an object of type t s. As far as is presently known, this is indeed the case.) The result is that it can be stripped off by rule (R5), just as with other CIE 23 The clause of the type specification is as follows: “If σ, τ and ζ are at-issue types for ℒ+ CI and 𝜐 is a shunting type for ℒ+CI , then ⟨σ, τ⟩ × ⟨ζ, 𝜐⟩ is a mixed type for ℒ+CI .”

218

eric mccready and yohei takahashi

content.24 With this background, we are ready to consider the semantics of mono in full compositional detail. 4.2. Applying the System to Mono In this system, it is entirely straightforward to provide a lexical entry for mono that will have the required characteristics of providing an appropriate separation of at-issue and CIE content, while still allowing composition with multiple arguments. The desired semantics is the following one.25 (28) ⟦mono⟧ = λpλq[Cause(p, q)]◆ λpλq[E p (p) ∧ Good_reason(p, q)] : ⟨t, ⟨t, t⟩⟩a × ⟨t, ⟨t, t⟩⟩s

Here, mono is analyzed as having a denotation which is of the type of propositional connectives in both at-issue and CIE “dimensions”. (We use an extensional system for simplicity here.) It has as at-issue content that the two propositions it takes stand in a causal relation, and as CIE content that the causal relation is a “high-quality” one, and that the speaker has some emotive attitude toward the causing proposition. On the CIE side, it denotes an object of shunting type: the reason is the need to compose with two distinct propositions, which is difficult for the original Potts system ℒCI to deal with, as discussed in the previous section. Let us now see how this analysis works in an actual derivation. 4.3. A Derivation We take the following, by now familiar discourse as our basic test case. (29) A: aa, mata shiken ochite-shimatta ah again test failed-antihon.perf ‘Damn, I failed the test again …’ B: omae-ga benkyoo shinainda mon you-nom study do.not mono ‘That’s because you don’t study.’ 24 As a reviewer notes, this is not technically correct. (R5) applies only to CI-typed objects, terms of type tc , while here we are dealing with terms of type ts . There are two possible solutions to this problem. The first is to allow (R5) to apply to terms of type ts as well (which means changing tc to t{c,s} ). The second is to allow (R9) to alter the type of the expressive part of its mixed-type input to tc so that (R5) is able to apply. The first solution is less radical, but the second is closer to the intuitive idea behind the rule application sequence. 25 Note that this typing is not, strictly speaking, compatible with the requirements of E, which looks for an argument of type ⟨s, t⟩, and neither with the emotive predicates that it delivers, which require the same. We are using an extensional semantics for simplicity of exposition, but nothing serious hangs on this decision.

good reasons

219

Plainly, B is responding to A’s statement about failing the test, and indicating A’s lack of study as the cause. We take B’s utterance to have the following (highly simplified) logical form, where [p] is a proposition made available by context.26 (30)

S [failed_test(h)]

S | ¬study(h)

MONO

The meaning of this tree can be computed in the following manner, given the lexical entry for mono from the previous section and the rules of ℒ+CI :

¬study(h) : t a λpλq[Cause(p, q)]◆ λpλq[Good_reason(p, q)] ∧ E p (p) : ⟨t, ⟨t, t⟩⟩ a × ⟨t, ⟨t, t⟩⟩ s R8 λq[Cause(¬study(h), q)◆ λq[Good_reason(¬study(h), q) ∧ E ¬study(h) (¬study(h))] : ⟨t, t⟩ a × ⟨t, t⟩ s failed_test(h) : t a R8 Cause(¬study(h), failed_test(h))◆ Good_reason(¬study(h), failed_test(h)) ∧ E ¬study(h) (¬study(h)) : t a × t s R9 Cause(¬study(h), failed_test(h)) : t a • Good_reason(¬study(h), failed_test(h)) ∧ E ¬study(h) (¬study(h)) : t s R5 Cause(¬study(h), failed_test(h)) : t a

As desired, then, on the proposed analysis, we get a split in content between the at-issue statement that a causal relation holds between the two arguments to mono, and the two judgements made by the speaker of the monosentence: that the cause is emotively significant, and that the cause represents a good quality reason for the effect. The behavior discussed earlier with respect to the independence test and the denial test is thus predicted, and the lexical content of mono is successfully accounted for. 5. Comparison with Previous Accounts

To our knowledge, there is no other formal account of mono in the literature. However, substantial attention has been given to this connective in the informal Japanese literature (nihongogaku). In this section, we will compare our formal account to the informal analyses that have been offered there. In this literature, it is often the case that researchers offer a great deal of puzzling and heterogeneous data, which is then analyzed in a rather impressionistic way. The literature on mono is no exception. We have already considered some of the data from the literature; our focus in this section will therefore be on the adequacy of the analyses that have been proposed. The reader might not be surprised to find that our formal account comes out of the comparison looking pretty good. 26 Again, this really should be made to follow from a computation of discourse structure in the SDRT style.

220

eric mccready and yohei takahashi

In the recent literature, it seems that there have not been many discussions centering individually on the discourse function of mono in spite of the fact that many researchers regard mono as a sentence-final particle (e.g. Noda 2002; Teramura 1992), and such particles are a major focus of the informal Japanese linguistic tradition. As far as we know, the works that primarily focus on mono are confined to a series of papers by Tsubone (1994, 1996) and Hashimoto (1997). We consider the accounts proposed by these two authors in turn. 5.1. Tsubone (1994, 1996) Tsubone argues that the propositional content carried by mono includes the meaning of a ‘general concept’ associated with the topic given in the preceding context. Essentially, the idea is that we are to identify a causal relation based on generic properties of objects: if in general Ps are Qs, then an object being a P can be taken to be evidence that it is also a Q.27 Tsubone (1994) argues that the general property originally is observed from an abstract noun mono (in Japanese linguistics, it is called keishikimeishi): (31) dentoo to iu mono-wa, nihon-no dokodedemo kantan-ni mi-rareru tradition C call mono-top Japan-gen everywhere easily see-can mono janai. mono cop.neg ‘It is generally recognized that we cannot easily observe traditional practices everywhere in Japan.’ (Tsubone 1994: T’s (1a))

Tsubone then argues that this general interpretation to abstract noun mono can be observed in other lexical items with the same morphological shape, such as the causal connective we are concerned with. Indeed, when mono is used, it does seem that the causal relation it expresses holds quite generally in many cases: in most cases, if the causing proposition holds, the caused proposition’s probability should rise, given a causal relation between the

27

Tsubone (1996: 37–38) defines the concept of generality as follows:

(i) When we look at a certain point of a time-scale, we can find a uniformity or similarity between individuals belonging to a given group. (ii) When we look at each of the individuals, we can find that they share a trait across the times in the time-scale. Thus, Tsubone wants to make use of properties shared by the individuals of a group that persist across times in order to define causal relations.

good reasons

221

two. But is this always the case? That is, must the causing proposition always be a cause of the (potential) effect? It seems pretty clear that it need not be, any more than it is the case that causation itself is not influenced by contextual factors (cf. Swanson 2010) or dependent on the facts of a given case. Consider, for an arbitrarily chosen example, the following dialogue, produced in a situation where A has inherited a chicken farm and means to ruin it to collect the insurance, but has some qualms of conscience about the method of ruination she has agreed on with B, namely the release of a virus fatal to chickens. (32) A: kuso! bokujo-o tsubusu keikaku-wa daishippai-da! Damn farm-acc destroy plan-top totally-spoil. ‘Damn! Our plan to ruin the farm was totally spoiled!’ B: omae-ga kin-no sanpu-o ore-ni makasenaindamono! you-nom virus-gen spread-acc I-Dat did.not.let.mono ‘It’s because you didn’t let me spread the virus!’

Clearly, a causal relation does indeed hold here; but it seems difficult to say that there is a general correlation between releasing deadly viruses and ruining chicken farms. Examples like this one make us believe that Tsubone has gone astray by focusing on the wrong elements of the causal relation, i.e. its generality or lack thereof, which we take not to be truly necessary for causation, much less for the relation denoted by mono in particular. We take what is special about mono to be, instead, the emotive quality and strength of reason indicated by its CIE content. 5.2. Explanation via “Speaker’s Logic” The other account we will consider here is that of Hashimoto (1997), who takes the function of mono to be what she calls “explanation under the speaker’s logic”, defined roughly as the subjective views of the speaker. Hashimoto claims that, for explanations based on speaker logic to be accepted by a hearer (since they are, by definition, explanations that are not truly objective), the hearer must have a close and positive relationship with the speaker. The function of mono is then claimed to be “… to explain the speaker’s recognition to a hearer that shares a sense of closeness28 with the speaker based on the speaker’s logic”.

28 Hashimoto’s term is miuchi-ishiki, roughly translatable as feeling as if one is part of the same societal (in)group.

222

eric mccready and yohei takahashi

Hashimoto claims that the content of mono has to do with the speaker’s subjective opinion based on his own logic or knowledge, which mono indicates to be a subjective explanation for some contextually relevant proposition (in our terms; we doubt Hashimoto would have put it quite this way). Hashimoto argues that this analysis explains a well-known aspect of sentences with mono, that they frequently convey a sense of blame (when otherdirected) or self-justification (when speaking about one’s own actions); this effect, according to Hashimoto, arises because mono provides an explanation from the viewpoint of speaker’s logic. If the preceding objective context is consistent with the speaker’s opinions, then, Hashimoto has it, the utterance is expected to sound affirmative or friendly, but on the other hand, if the context is inconsistent with these opinions, then the utterance is expected to sound frustrated or reproachful. We think there is something to Hashimoto’s view; indeed, parts of our analysis can be understood as a formalization of it. In particular, our use of subjective probabilities is something we take to realize Hashimoto’s notion of speaker logic. Again, there are two possibilities: that the objective facts match the subjective assessment of them by the speaker, or not. If they do match, then a relatively positive impression may well arise (though not necessarily: as far as we can see, it could just as well be perfectly neutral). If they do not match, it is easy to see that a sense of self-justification could appear: if the speaker is explaining her actions with her subjective assessment, clearly she is justifying them, particularly if her assessment differs from the objective facts. The sense of blame that sometimes arises we think comes from a different source, which for us is the emotive component of mono’s meaning: if the speaker thinks that the cause is a negative thing, clearly an impression of blame will be given if that cause was brought about by some agent (e.g., as is often the case in uses of mono, the hearer). While we are sympathetic with her general intuition, we do not think Hashimoto’s (1997) analysis is faultless. First, the notion of speaker’s logic is not completely clear, and (because of lack of formalization) it is hard to tell exactly what she has in mind. For this reason we are not absolutely sure that our reconstruction of her reasoning above is faithful to her original idea. This vagueness is a serious weak point of her analysis. Second, there are empirical difficulties with the assumption that use of mono requires and indeed induces familiarity and even friendliness. We agree that an impression of familiarity can arise, but we think that this is simply due to the expression of emotive attitude conveyed by the connective: if one expresses strong emotions, some degree of familiarity is to be expected, for otherwise it would not be appropriate to do so, especially in Japanese society as traditionally

good reasons

223

construed. We doubt, however, that actual familiarity is required. Compare formal and informal pronouns in European (and other) languages: use of informal pronouns generally indicates familiarity or closeness, but a speaker can use them unilaterally even when it is not appropriate. Familiarity, then, cannot be taken to be a hard condition. Let us back this argument up with an example. If the argument that presumes the mono speaker to have a friendly feeling with a hearer is adopted, the following utterance should induce a feeling of closeness or even of belonging to the same in-group between speaker and hearer. (33) A: saiyoo-shite-itadake-tara takusan kinmu-ni tsuite hayaku shigoto-o adopt-do-hon-if many service-dat attach quickly work-acc oboe-tai desu. learn-want cop ‘If you hire me, I want to go on duty a lot and learn all aspects of the operation immediately.’ B: demo, anata-wa unten-menkyo-o motteinai nda mono. kore-wa but you-top driving-license-acc have-not noda mono this-top ookina furi da yo. big disadvantage cop part ‘I know what you’re saying, but you don’t have a driver’s license. This is a big disadvantage for you.’

Suppose that this discourse takes place in the following scenario. A applied for a job, for which this is the job interview. B is an interviewer. B read A’s resume and noticed that A did not have a driver’s license, which is almost necessary to perform the job. At the moment when B noticed A’s lack of a license, B essentially decided not to hire A, but still had to interview A. B’s purpose with the utterance above is thus to show A that he is not getting the job. B’s use of mono here actually sounds rather cold and chilling. But this is precisely the opposite of what is expected if Hashimoto is correct. For us, the incorrect implication does not arise: B simply holds a negative attitude (bad) toward the chance of A’s being hired; he feels comfortable using mono here because of power imbalances in this context. We take this to be a welcome non-consequence of our analysis.29

29 It should be noted that this example is somewhat special in that it is not clear precisely what is being explained by mono. This is a case in which the content of some immediately preceding assertion is not available as argument for the connective. Here, a special mechanism is needed. We suggest that what is required is a topic formation operator which is able

224

eric mccready and yohei takahashi 5.3. Summary

Summing up, we take it that our analysis captures the facts better than previous proposals. Neither of Tsubone’s and Hashimoto’s analyses are formalized, so it can be difficult to see exactly what their proposals amount to; this problem is resolved in our treatment due to our use of formal semantic and pragmatic techniques. Both of these authors also make wrong empirical predictions. Tsubone claims that mono can only be used with causal relations that are fully general across cases (‘generic causes’); as we showed, this view is too strong, a problem we diagnosed as following from a wrong view of causation. Hashimoto’s analysis, on the other hand, we found fairly close to the mark in its focus on subjectivity. However, her view suffers from incorrect assumptions about requirements on the relative social position and closeness of speaker and hearer. We showed that our analysis is not subject to these problems.30 6. Conclusion In this paper, we have given a formal semantics for the Japanese causal connective mono. We argued that mono has three meaning components: first, the indication of a causal connection between two propositions, which is its at-issue meaning, second, the expression of a speaker attitude toward the causing proposition, and third, the indication that the causal relation is a ‘high-quality’ one, which latter two meaning components are conventionally implicated or expressive. This analysis was implemented using subjective probabilities together with a contextual standard for quality of reasons, and the whole was expressed in ℒ+CI , a logic for lexical terms which introduce mixed expressive and at-issue content. Finally, we compared this to take into account the intentions of speakers in their linguistic acts; a possibility would be the ⇑ operator of SDRT, or a variant thereof. This issue connects directly to the final use of mono which we have left unanalyzed, discussed in Section 6. 30 A reviewer suggests that we are being a bit unfair to these authors in expecting them to give meanings for mono that satisfy a high standard of precision, given that our own analysis is underspecified in some respects: namely, we do not give a full treatment of how the emotive content of mono is determined (though, as mentioned, this is done in McCready 2012), and we do not commit to a particular analysis of causation, although we do give full details of how composition with mono works at the sentence level and of the meaning types its semantics involves. This is a fair point, but we nonetheless feel that our analysis is an improvement on previous attempts. Our formalization, while not fully complete, clarifies a number of issues related to the meaning of mono, also highlights the areas where more work needs to be done, and further is not subject to the wrong predictions of the earlier analyses.

good reasons

225

analysis with existing analyses in the informal Japanese-language literature on the topic, showing that our analysis retains many of the insights from that literature, while improving on them. One interesting point which we have not touched on much here is the classification of mono as sentential connective or sentence-final particle. As we indicated, both Tsubone and Hashimoto take mono to be—in general—a particle; but we have analyzed it uniformly as a connective, albeit a peculiar one which is intrinsically discourse-functional.31 Which of these characterizations is correct? More generally, what criteria can we give to distinguish between particles and connectives? The answer which has guided us here is this one: particles take only a single propositional argument, while connectives are necessarily two-place at a semantic level. To our knowledge, all lexical items that are universally agreed to qualify as particles are modifiers of single sentences (Davis 2009; McCready 2009; Noda 2002; Zimmermann 2011). Since mono is clearly semantically two-place, we believe it is inappropriate to classify it as a particle. A second point is the possibility of paraphrase: more or less universally, it has proven difficult or impossible to characterize the meaning of discourse particles via paraphrase, or even to say exactly what they do, as can be seen from the wide range of positions adopted in the literature.32 But the meaning of mono is plainly a causal one, a point which is not in any doubt. We take this solidity to be a second reason to classify mono as a connective. However, the expressive part of mono is indeed quite similar to the content of some particles, which makes it natural that analyses of this kind can be found in the literature. It seems fair to say that mono is, truly, an object of mixed type. We must leave one issue related to mono unsettled in this paper. There are instances of mono that provide explanations at a sort of metalevel, not at the level of content; Hashimoto characterizes these as explaining “the event that an addressee does not recognize or the ground of why an addresser thinks so” (Hashimoto 1997: 209). Hashimoto names this type of mono utterance “Detailed Description”, some instances of which are provided below. (34) A: Yamada-ga rikon-shitarashiina. Yamada-nom divorce.did.likely ‘I heard Yamada got divorced.’

31 Though not at all unique: a comparable connective without expressive content is the English therefore. 32 See especially Gutzmann 2008 for a summary of some analyses that have been proposed for German particles together with an insightful approach in terms of expressive meanings.

226

eric mccready and yohei takahashi B: aa. shikamo kodomo-to bekkyo-da mono. sainan-dayona. yes furthermore child-with separation mono misfortune-is ‘I know. To make the matter worse, he will live separately from his child. How awful it is!’

(35) A: ashita-wa yasumi-de ureshiina. tomorrow-top holiday-so glad ‘I’m glad since tomorrow is holiday.’ B: soreni tenki-mo hare-mitaidesu mon(o) ne. furthermore weather-also sunny-likely.is mono ne. ‘Furthermore, it will be sunny tomorrow.’

In this type of mono utterance, the speaker of the mono sentence picks up the event introduced in the preceding context and then provides new information about the event. Hashimoto (1997) says that this type of mono utterance is used in order to supplement the hearer’s (or even the speaker’s) lack of understanding of some area related to the contextually salient proposition with respect to which the mono sentence is stated, which we think is broadly the right characterization. But this is, canonically, an elaboration relation rather than a causal one, in the usual sense of theories of discourse relations like SDRT. Does this mean that mono is ambiguous? We think that it does not. In fact, this usage can be thought of as a way to provide additional evidence for what drives the addressee to have his judgment or opinion about the given topic, rather than a way to indicate the cause of the topical proposition. Thus, what we are dealing with is causation at a discourse level, or at the level of intention: “my use of mono indicates additional reasons for making the utterance that you made, as opposed to its propositional content”.33 If this is right, then we continue to have, in a general sense, a causal meaning for mono; but capturing this sense formally requires reasoning about speaker intentions in communication, and then modifying them; this is a complex operation which needs a great deal of machinery that we cannot introduce here for space reasons. Together with clarifying the way in which mono interacts with the previous context at the level of discourse structure, we think that characterizing this use of mono is the most pressing task facing our theory. One place we think our proposal here might be useful is in application to the German denn, which was mentioned in the introduction, as well as other German connectives with expressive meanings such as weil. Scheffler (2005)

33 A reviewer notes that the expressive use of German weil can have a similar meaning. This connection seems an interesting avenue for future work.

good reasons

227

provides an interesting analysis of this connective in terms of conventional implicature, but does not give a compositional implementation of the idea: in fact, in the original system of Potts (2005), such an analysis cannot be provided because of the way the resource logic behaves (as shown by McCready 2010). However, as we have shown here, ℒ+CI is well suited for analyzing connectives with mixed expressive and asserted content, and indeed for mixed content expressions in general. We think an analysis of denn might well be available along the lines of what we have proposed here for mono, though space limitations prevent a full exploration of the idea. We therefore leave it too for future work. Acknowledgements Thanks to Daniel Gutzmann and Hans-Martin Gärtner for giving us the opportunity to write this paper, and to two anonymous reviewers for extremely helpful and insightful comments. References Amaral, Patricia & Craige Roberts & E. Allyn Smith (2007): “Review of The logic of conventional implicatures by Chris Potts”. Linguistics and Philosophy 30, 707–749. DOI: dx.doi.org/10.1007/s10988-008-9025-2. Asher, Nicholas & Alex Lascarides (2003): Logics of conversation. Cambridge: Cambridge University Press. Asher, Nicholas & Michael Morreau (1991): “Commonsense entailment. A modal theory of non-monotonic reasoning”. In: Mylopoulos, J. & R. Reiter, eds.: Proceedings of the Twelfth International Joint Conference on Artificial Intelligence. Los Altos, CA: Morgan Kaufman, 387–392. Bach, Kent (1999): “The myth of conventional implicature”. Linguistics and Philosophy 22, 327–366. DOI: dx.doi.org/10.1023/A:1005466020243. Collins, John & Ned Hall & L.A. Paul, eds. (2004): Causation and counterfactuals. Cambridge, MA: MIT Press. Davis, Christopher (2009): “Decisions, dynamics and the Japanese particle yo”. Journal of Semantics 26, 329–366. DOI: dx.doi.org/10.1093/jos/ffp007. Dowty, David R. (1979): Word meaning and Montague grammar. The semantics of verbs and times in generative semantics and Montague’s PTQ. Dordrecht: Reidel. Eells, Ellery (1991): Probabilistic causality. Cambridge: Cambridge University Press. Geurts, Bart (2010): Quantity implicatures. Cambridge: Cambridge University Press. Grice, H. Paul (1975): “Logic and conversation”. In: Cole, Peter & Jerry L. Morgan, eds.: Speech acts. New York: Academic Press, 41–58. Gutzmann, Daniel (2008): On the interaction between modal particles and sentence mood in German. MA thesis. Mainz: Johannes Gutenberg-Universität.

228

eric mccready and yohei takahashi

Harris, Jesse A. & Christopher Potts (2009): “Perspective-shifting with appositives and expressives”. Linguistics and Philosophy 32, 523–552. DOI: dx.doi.org/ 10.1007/s10988-010-9070-5. Hashimoto, Yoshimi (1997): “Shuujoshi mono: ‘hanashite-no ronri’-kara-no setsumei [Sentence-final particle mono: Explanation under ‘speaker’s logic’]”. Studies on Japanese and Japanese Culture 7, 201–212. Jayez, Jacques & Corinne Rossari (2004): “Parentheticals as conventional implicatures”. In: Corblin, Francis & Henriëtte de Swart, eds.: Handbook of French semantics. Stanford, CA: CSLI, 211–229. Jeffrey, Richard (1983): The logic of decision. Chicago: University of Chicago Press. Kaplan, David (1989): “Demonstratives. An essay on the semantics, logic, metaphysics, and epistemology of demonstratives and other indexicals”. In: Almog, Joseph & John Perry & Howard Wettstein, eds.: Themes from Kaplan. Oxford: Oxford University Press, 481–563. Kennedy, Christopher (1999): Projecting the adjective. The syntax and semantics of gradability and comparison. New York: Garland. ——— (2007): “Vagueness and grammar: the semantics of relative and absolute gradable adjectives”. Linguistics and Philosophy 30, 1–45. DOI: dx.doi.org/10.1007/ s10988-006-9008-0. Kindaichi, Kyoosuke & Takeshi Shibata & Akio Yamada & Tadao Yamada, eds. (1991): Shin-Meikai Kokugo-Daijiten [New Meikai Dictionary of Japanese]. Vol. 4. Sanseidoo. Landman, Fred (1992): “The progressive”. Natural Language Semantics 1, 1–32. DOI: dx.doi.org/10.1007/BF02342615. Lasersohn, Peter (2005): “Context dependence, disagreement, and predicates of personal taste”. Linguistics and Philosophy 28, 643–686. DOI: dx.doi.org/10.1007/ s10988-005-0596-x. Lewis, David (1973a): “Causation”. Journal of Philosophy 70, 556–567. DOI: dx.doi.org/ 10.2307/2025310. ——— (1973b): Counterfactuals. Oxford: Blackwell. ——— (2004): “Causation as influence”. In: Collins, John & Ned Hall & L.A. Paul, eds.: Causation and counterfactuals. Cambridge, MA: MIT Press, 75–106. McCready, Eric (2004): “Two Japanese adverbials and expressive content”. Proceedings of SALT 14, 163–178. URL: http://elanguage.net/journals/salt/article/view/14 .163/1740. ——— (2009a): “Particles: Dynamics vs. utility”. In: Takubo, Y. & T. Kinuhata & S. Grzelak & K. Nagai, eds.: Japanese/Korean Linguistics 16. CSLI, 466–480. ——— (2009b): “What man does”. Linguistics and Philosophy 31, 671–724. DOI: dx.doi.org/10.1007/s10988-009-9052-7. ——— (2010): “Varieties of conventional implicature”. Semantics & Pragmatics 3, 1–57. DOI: dx.doi.org/10.3765/sp.3.8. ——— (2012): “Emotive equilibria”. Linguistics and Philosophy 35, 243–283. DOI: dx .doi.org/10.1007/s10988-012-9118-9. McCready, Eric & Norry Ogata (2007): “Evidentiality, modality, and probability”. Linguistics and Philosophy 30, 147–206. DOI: dx.doi.org/10.1007/s10988-0079017-7. Noda, Harumi (2002): “Shuujoshi no kinoo [The functions of sentence-final parti-

good reasons

229

cles]”. In: Miyazaki, K. & H. Noda & S. Adachi & T. Takanashi, eds.: Modariti [Modality]. Kuroshio Press, 261–288. Potts, Christopher (2005): The logic of conventional implicatures. Oxford: Oxford University Press. ——— (2007c): “The expressive dimension”. Theoretical Linguistics 33, 165–197. DOI: dx.doi.org/10.1515/TL.2007.011. Rett, Jessica (2008): “A degree account of exclamatives”. Proceedings of SALT 18, 601–608. URL: http://ecommons.library.cornell.edu/bitstream/1813/13058/2/Rett .pdf. van Rooy, Robert (2003b): “Questioning to resolve decision problems”. Linguistics and Philosophy 26, 727–763. DOI: dx.doi.org/10.1023/B:LING.0000004548.98658 .8f. ——— (2010): “Measurement and interadjective comparisons”. Journal of Semantics 28, 335–358. DOI: dx.doi.org/10.1093/jos/ffq018. Scheffler, Tatjana. (2005): “Syntax and semantics of causal denn in German”. In: Dekker, Paul & Michael Franke, eds.: Proceedings of the Fifteenth Amsterdam Colloquium. LLC/University of Amsterdam, 215–220. Searle, John R. (2001): Rationality in action. Cambridge, MA: MIT Press. Simons, Mandy & Judith Tonhauser & David I. Beaver & Craige Roberts (2010): “What projects and why”. Proceedings of SALT 20, 309–327. URL: http://elanguage .net/journals/index.php/salt/article/viewFile/20.309/1326. Skryms, Brian (1980): Causal necessity. New Haven, CT: Yale University Press. Smith, Carlota S. (1997): The parameter of aspect. 2nd ed. Dordrecht: Kluwer. Swanson, Eric (2010): “Lessons from the context sensitivity of causal talk”. Journal of Philosophy 107.5, 221–242. Teramura, Hideo (1992): Teramura Hideo ronbunshuu 1 [Collected Papers of Hideo Teramura 1]. Tokyo: Kuroshio Press. Tsubone, Yukari (1994): “Monoda-ni kansuru ichi-koosatsu”. Nihongo-Kyooiku [ Japanese Education] 84, 65–77. ——— (1996): “Shuujoshi, setuzokujoshi,-to shiteno mono-no imi: mono, mononara, monono, monoo [Semantics of mono as sentence-final particle: mono, mononara, monono, and monoo]”. Nihongo-Kyooiku [ Japanese Education] 91, 37–48. Webber, Bonnie & Matthew Stone & Aravind Joshi & Alastair Knott (2003): “Anaphora and discourse structure”. Computational Linguistics 29, 545–588. URL: http://www.aclweb.org/anthology-new/J/J03/J03-4002.pdf. Williamson, Timothy (2000): Knowledge and its Limits. Oxford: Oxford University Press. Zimmermann, Malte (2011): “Discourse particles”. In: Maienborn, Claudia & Klaus von Heusinger & Paul Portner, eds.: Semantics. An international handbook of natural language meaning. Berlin: de Gruyter, 2011–2038.

COMMON GROUND MANAGEMENT: MODAL PARTICLES, ILLOCUTIONARY NEGATION AND VERUM

Sophie Repp

1. Introduction The goals of this paper are twofold. The first goal is to argue that there are operators with a common ground (= CG) managing function that can influence the truth-conditional meaning of a sentence even though their meaning usually is taken not to be truth-conditional. These CG-managing operators indicate the status of a proposition relative to the CG: (i) they indicate whether the proposition is part of the CG, or whether it is new; (ii) they indicate the interlocutors’ current stance towards the proposition: whether the proposition is e.g. unexpected or obvious; and (iii) they indicate how the CG should develop: whether, and with what degree of commitment on the side of the speaker, the proposition should become part of the CG, or be removed from it. Illocutionary operators like assert take propositions as arguments that have been assigned a CG status. This means that utterances as semanticpragmatic units are structured as follows: (1)

[illocutionary operator [CG-managing operator [proposition]]]

I propose that modal particles, as well as conversational epistemic operators—like the verum operator (Romero & Han 2004), and illocutionary negation expressed as the operator falsum (Repp 2006, 2009)—are CGmanaging operators. If there is no overt CG-managing operator a default is assigned. The default is that the proposition is suggested to be added to the CG, i.e. that the proposition is new, that the speaker takes it to be true, and that the addressee’s stance towards it can be the subject of the next utterance. The second goal of this paper is an empirical one: to make sense of a—at first sight—very puzzling set of data involving modal verbs and negation in German. The scopal interaction of modal verbs and negation has received some attention in recent years (e.g. Ehrich 2001; Zeijlstra 2007; Penka 2010) but there are still many questions open. For instance, as we go along, it will

232

sophie repp

become clear that prosody and the presence or absence of the particle auch (‘also’) are very important in this interaction. The precise reasons for this are unexplored. I will not offer an account for the role of these two factors in this paper. I will only work with descriptive generalizations of the empirical patterns. The focus of this paper is a set of data that—as far as I am aware— has not been discussed at all yet. I am interested here in the scope relations between the negation and a modal verb in sentences that also contain a modal particle. I will show that these data are very well suited to sharpen our understanding of the working of CG-managing operators: of illocutionary negation (= falsum), modal particles and verum. I will argue that the presence of CG-managing operators, which bridge truth-conditional meaning and illocutionary meaning, can produce compatibility or incompatibility of a proposition with the CG, depending on what the CG is, and depending on which CG-managing operators are combined in an utterance. This then results in the availability or unavailability of certain scope relations between the negation and the modal verb. The paper is structured as follows. In the remainder of this section, I present the basic set of data, where I focus on sentences with an epistemic modal verb, negation and the modal particle doch. I do not consider root modal verbs for reasons of space, and to keep possible confounding factors to an absolute minimum.1 In Section 2, I discuss the notion of CG and look at the meaning and function of the falsum operator and the particles doch and ja. On the basis of these considerations, Section 3 then explains why the data I present in the introductory section pattern the way they do. Section 4 extends the empirical base to modal particles not presented in the original data set and explains the observed effects. Section 5 looks at verum and compares it to falsum. Section 6 concludes. The German modal verb können (‘can’) typically scopes under negation (e.g. Ehrich 2001; Öhlschläger 1989), see (2) with können in an epistemic reading. (2)

Paul kann nicht ins Schwimmbad gegangen sein. Paul can not to.the pool gone be ‘Paul can’t have gone to the pool.’

neg>mod

The scope relations can but need not change if the negative marker nicht (‘not’) receives prosodic prominence.2 With an accent on nicht (2) can be

1 For the interaction of illocutionary negation with root modals in gapping sentences see (Repp 2009: Chapter 4). 2 To elicit a ‘neutral’ prosody of (2), i.e. one that expresses wide focus, a question asking

common ground management

233

understood as a rejection of a previous utterance, i.e. a denial, still with wide scope of the negation: (3)

A: Paul kann ins Schwimmbad gegangen sein. B: Paul kann nicht ins Schwimmbad gegangen sein. Paul can not to.the pool gone be A: ‘Paul can have gone to the pool.’ B: ‘Paul cannot have gone to the pool.’

neg>mod

An accent on the negative marker can also elicit a narrow scope reading of the negation, which typically occurs with additional markers such as auch (‘also’) (e.g. Ehrich 2001; Penka & Stechow 2001), see (4). As a matter of fact, for a narrow scope reading to arise in sentences with epistemic können (‘can’) in German, the accent on the negative marker is obligatory.3 (4)

A: Paul kann ins Schwimmbad gegangen sein. B: Paul kann auch nicht ins Schwimmbad gegangen sein. Paul can also not to.the pool gone be A: ‘Paul can have gone to the pool.’ B: ‘It is (also) possible that Paul has not gone to the pool.’

mod>neg

The last option to interpret (2) with an accent on the negative marker is as an answer to a yes-no question such as Can Paul have gone to the pool?, which also results in wide scope of the negation. The addition of the modal particle doch, whose meaning roughly can be paraphrased as ‘why are you not considering p, you should know that p’ makes the discourses in (3) and (4) (i.e. with or without auch) infelicitous, see (5). The denial reading of (3B) is not available, and the narrow scope reading of (4B) (which is more easily attainable with auch) is incoherent: (5)

A: Paul kann ins Schwimmbad gegangen sein. B: #Paul kann doch (auch) nicht ins Schwimmbad gegangen sein. Paul can part (also) not to.the pool gone be

for an epistemic judgement such as What do you think? seems most appropriate (in contrast to something like What’s up? or What happened?). Thus, it seems that the question must relate to the epistemic status of the proposition. 3 As I mentioned above, I will not explore the reasons for this in this paper. Note, however, that in the presence of the focus-sensitive particle auch (‘also’) the accent on the negation plausibly is interpreted as a focus accent, where the alternative to the negative polarity is the positive polarity. The negation in this case is c-commanded by the focus particle. In a minimal variant where the negation directly precedes auch and thus c-commands it, see (i), the narrow scope reading is not available. (i)

Paul kann nicht auch ins Schwimmbad gegangen sein.

neg>mod, *mod>neg

234

sophie repp A: ‘Paul can have gone to the pool.’ B: ‘It is (also) possible that Paul has not gone to the pool, why are you not considering this, you should know.’ *neg>mod, #mod>neg

(6) illustrates that the disappearance of the wide scope reading is due to the combination of doch with accented nicht. Accenting a different element such as Schwimmbad (‘pool’) produces a felicitous wide scope reading of the negation. The accent on Schwimmbad here may indicate narrow focus on the DP or focus on the VP/vP/CP, the latter by focus projection (Selkirk 1984), as ins Schwimmbad (‘to the pool’) is the complement of the verbal head.4 (6)

A: B: A: B:

Paul kann ins Schwimmbad gegangen sein. Paul kann doch nicht ins schwimmbad gegangen sein. ‘Paul can have gone to the pool.’ ‘Paul CAN’T have gone to the pool, why are you not considering this, you okneg>mod, *mod>neg should know.’

Furthermore, if the sentence in (5B), i.e. the sentence that contains the modal particle doch and accented negation, occurs in a context different from (5A) narrow scope becomes coherent (again most easily attainable with added auch (‘also’)). This is illustrated in (7). Here the context contains the necessity modal müssen (‘must’). Note, however, that as in (5B), wide scope is unavailable with this intonation. (7)

A: Paul muss ins Schwimmbad gegangen sein. Paul must to.the pool gone be B: Paul kann doch (auch) nicht ins Schwimmbad gegangen sein. Paul can part (also) not to.the pool gone be A: ‘Paul must have gone to the pool.’ B: ‘It is also possible that Paul has not gone to the pool, why are you not considering this, you should know.’ *neg>mod, okmod>neg

This is the basic set of data that I will discuss in this paper. It will be extended with further examples as we go along. I will argue that the observed interaction of the modal particle doch and the negation in these data is due to the following. (i) The negative marker in the wide scope readings in (most of) the above examples denotes the CG-managing operator falsum, which basically says that the proposition it scopes over should not be in the CG. (ii) The modal particle doch denotes a CG-managing operator that instructs the hearer to retrieve this proposition from the CG. (iii) Since falsum and doch 4 The English paraphrase has the accent on the auxiliary, which expresses best the surprise/indignation that the German example conveys.

common ground management

235

give conflicting instructions with respect to CG management they cannot occur in the same utterance. As a consequence, the relevant examples do not have a wide scope, i.e. denial, reading. In the case of the incoherent narrow scope reading, the negation is ordinary propositional negation. Here, we shall see that the modal particle gives instructions with respect to CG management that are incompatible with the current discourse status of the proposition: it instructs the hearer to retrieve a proposition that is already currently being considered. To set the scene for the analysis of the data, the next section looks at the notion of CG, and at the negation and modal particles as CG-managing operators. 2. The CG and Common-Ground Managing Operators 2.1. The Notion of CG The original notion of the CG (Stalnaker 1974; Karttunen 1974; Lewis 1979) considers CG as the set of propositions which the participants in the conversation mutually agree to treat as true for the purposes of the exchange. Let us call CG under this definition CGt , t being a mnemonic for true. CGt has been a useful notion for the distinction between presuppositions vs. assertions as input requirements on CGt vs. update effects on CGt . CGt comprises the propositions that are pragmatically presupposed. In recent work on epistemic modals a distinction has been drawn between propositions that are taken to be true in an utterance situation, and those that the participants are aware of. Portner (2007), for instance, distinguishes between the CG and the common propositional space (CPS). The former is CGt , the latter is the set of propositions of which the participants in the conversation are mutually aware. CGt is thus a subset of CPS. Similarly, von Fintel & Gillies (2011) suggest that propositions can be ‘put into play’ without getting asserted, i.e. they are part of what would correspond to Portner’s (2007) CPS minus CGt . Recall from the introduction, however, that even ‘ordinary’ assertions cannot be assumed to add propositions to CGt directly. Assertions are up for agreement or disagreement by the addressee. In belief revision systems like Wassermann 2000 this is modelled as a distinction between provisional and actual beliefs. An interlocutor can entertain a provisional belief without being committed to it (yet).5

5 Also see e.g. Lascarides & Asher 2009 for the modelling of agreement, disputes and commitments in dialogue.

236

sophie repp

So according to these works, an extended notion of the CG, which is more inclusive than CGt , seems to be necessary. I suggest that the CG status of a proposition, as indicated by CG-managing operators, can also be part of the CG, e.g. if the interlocutors are not really committed to a proposition that proposition will be in the CG, and its status will be that of meeting low commitment by the interlocutors. I further assume that the CG does not include the proposition of the immediately preceding utterance. This is a consequence of the assumption that the default CG status of that proposition is that it is suggested to be added to the CG, that the speaker takes it to be true, and that the hearer’s position towards it can be the subject of the next utterance. Therefore it is not added ‘directly’ to CG. This addition happens only after the addressee has accepted the proposition. This acceptance may be signaled overtly if the addressee expresses his or her agreement. If he or she does not agree but neither expresses his or her disagreement the proposition will also become part of the CG. The term CG management was introduced by Krifka (2008) to distinguish inter alia the different kinds of contributions different kinds of speech acts make to the CG. Questions, for instance, do not introduce propositions that the interlocutors mutually agree to treat as true. They only have a CG-managing function and do not alter CGt directly: they signal to the interlocutor(s) how the CG should develop, and what the communicative goals of the interlocutors are. In a successful discourse, posing a question will influence the development of CGt since a cooperative interlocutor will react to a question and thereby change CGt in a way that is specified by the question. The CG-managing operators I am interested in are not speech act operators like Q or assert. They scope under such speech act operators, and indicate the CG status of a proposition. 2.2. Illocutionary Negation as the CG-Managing Operator falsum 2.2.1. Denials: The Classic View With Stenius (1967) the relation between a speech act and a proposition can be viewed as one between sentence mood (an illocutionary operator) and sentence radical (the proposition). I illustrate this for the negative sentence in (8), which, depending on the intonation, has the two readings in (8a) vs. (8b). (8a) is an assertion of a negative proposition. (8b) is a denial of a positive proposition. This reading might be viewed as containing what we might call illocutionary negation, i.e. a negation that operates on the level of the illocution. Note that illocutionary negation is not the same as speech act negation (or ‘denegation’) in the sense of I am not performing the act of

common ground management

237

X (see e.g. Cornides 1969; Dummett 1973), which in the scheme proposed in (8) would correspond to (8c). The proposal that I put forward for denials in this paper deviates from the structure in (8b), and from the one in (8c). This will be made precise further below. (8)

Max is not tall. a. Assert (¬Max is tall) b. Denial (Max is tall) c. ¬Assert (Max it tall)

Denials have been described as speech acts that remove (part of) previously introduced material from the CG (= CGt ) and perform a correction operation on contextual information (van der Sandt 1991). CGt in this view includes the directly preceding utterance. So, in (8b) the previously introduced material would be the proposition that Max is tall. Similarly, in (3) above, repeated below, the proposition that Paul can have gone to the pool is the proposition that is denied. The example illustrates that a typical characteristic of illocutionary negation in denials is that it carries stress (also see Horn 1989). (3)

A: Paul kann ins Schwimmbad gegangen sein. B: Paul kann nicht ins Schwimmbad gegangen sein. Paul can not to.the pool gone be A: ‘Paul can have gone to the pool.’ B: ‘Paul cannot have gone to the pool.’

neg>mod

A denial can also react to an utterance that contains a negation. In this case the denial would be a positive sentence (van der Sandt 1991). I shall come back to this issue in Section 5, where I discuss verum. As a speech act, a denial can target all sorts of aspects of a previous utterance: the proposition, presuppositions, implicatures, or formal aspects. The negation of non-propositional parts of an utterance has been called metalinguistic negation (Horn 1985, 1989; and others). Most theories of metalinguistic negation consider it a reflex of the negation taking scope over an entire utterance (cf. Jacobs 1982, Jacobs 1991; van der Sandt 1991; Carston 1996).6 For instance, according to van der Sandt (1991), an echo operator takes the sum of all information of a previous utterance φ−1 (including implicatures, presuppositions) as the propositional content of the utterance φ0 . This content can then be negated, which results in a denial reading. The information conveyed by φ−1 is removed from the CGt to avoid inconsistency. Note that in 6 Geurts (1998) suggests that the negation of implicatures and form aspects is different from the negation of presuppositions.

238

sophie repp

this account the negation is ordinary propositional negation. It is the echo operator that brings the entire content of the previous utterance under the scope of the negation.7 Although this is a very attractive view of denials I think that it misses out on certain parallels that can be found between the negation in denials and the negation in other types of speech acts, viz. certain types of negative polar questions. Some negative polar questions can be argued to involve illocutionary negation, see below, but the assumption of an echo operator is much less plausible for them. Furthermore, even for denials an analysis in terms of an echo operator cannot account for the whole range of data. We shall see that the discourse conditions for denials are rather intricate when it comes to propositions with scalar expressions. They cannot be accounted for by the presence of an echo operator. 2.2.2. Denials & Negative Polar Questions: The falsum Operator The negation in negative polar questions can come as outer negation, or inner negation (Ladd 1981). A negative polar question with outer negation double-checks p, see (9a). A negative polar question with inner negation double-checks ¬p, see (9b). In German this difference can be disambiguated by placing the particle auch (‘also’) below vs. above the negative marker (Repp 2006, 2009). In English the difference is reflected in the use of the positive polarity item too vs. the negative polarity item either (Romero & Han 2004). (9)

a. Schläft Peter nicht auch? sleeps Peter not also ‘Isn’t Peter sleeping too?’

outer negation

b. Schläft Peter auch nicht? sleeps Peter also not ‘Isn’t Peter sleeping either?’

inner negation

7 As indicated by the subscript, the ‘removal’ idea relates to CG . We would not want t to assume that denials erase all memory of the interlocutors of a previous discourse move. Approaching this issue from a discourse theoretical perspective, van Leusen (2004, 2007) makes a distinction between discourse description, and discourse meaning (= CGt ). The discourse description is a set of statements in a logical language that describes the syntactic, semantic, as well as (some) pragmatic relations characterising the discourse processed, and is modelled in a discourse tree. Van Leusen argues that every utterance—including denials and corrections—increments the discourse in a monotonic way—it leads to further specifications of the discourse tree. This means that statements are never retracted from a discourse description. Denials and corrections are nonmonotonic updates when it comes to discourse meaning in the sense that a statement that is corrected will no longer be entailed by the discourse meaning.

common ground management

239

Romero & Han (2004) suggest that these two ‘types’ of negation arise from the scopal interaction of the negation operator with the verum operator, which is a conversational epistemic operator. Outer negation scopes over verum whereas inner negation scopes below it. Repp (2006, 2009) argues that outer negation in questions is not a result of the negation taking wide scope over verum but that outer negation is a conversational epistemic operator itself. This operator, falsum, is the operator that also occurs in denials. Repp supports this view with evidence for the parallel behaviour of outer negation and negation in denials with respect to their syntax (e.g. the licensing of polarity-sensitive items; in German, the position of the negative marker with respect to focussed constituents), and with respect to their scopal behaviour in negative gapping sentences. A preliminary definition of falsum is given in (10) below.8 In denials the variable x corresponds to the speaker, in questions it corresponds to the addressee: (10) (to be revised)

⟦falsum⟧x = λp⟨s,t⟩ λw. ∀w′ ∈ Epix (w)[∀w′′ ∈ Convx (w′ )[p ∉ CGw′′ ]]

In all the worlds w′ that conform to x’s knowledge in w it holds that in all the worlds w′′ that conform to the conversational goals of x in w′ (according to the Maxims of Quantity and Quality), proposition p is not in the CG (CGt ).

Repp (2006, 2009) suggests to link the meaning of falsum to speech act theory by viewing epistemic speech act operators as expressing the degrees of strength of the sincerity conditions of a speech act. falsum expresses that there are zero9 degrees of strength for sincerely adding a proposition to the CG. This implies that p should not be part of the CG, according to the speaker. This proposal seems intuitive for denials if we consider them as assertions with zero degrees of strength of the sincerity conditions: a sincerity condition for assertions is that the speaker must believe that the proposition expressed is true. For questions this seems not so intuitive. A question

8 Repp (2009) makes a similar suggestion, which, however, is too weak to capture the discourse restrictions on denials. 9 In Vanderveken (1990), who made a first formal suggestion for degrees of strength, zero degrees of strength is the neutral degree of strength. Stronger degrees of strength are represented by positive integers, weaker degrees of strength by negative integers. I am taking zero here to be the lower boundary for degrees of strength. Epistemic speech act operators with higher degrees of strength are for instance German wohl (‘probably’), see Section 4, and English surely.

240

sophie repp

is a directive, and a directive expresses a desire. Intuitively it is not the case that in questions with outer negation the sincerity conditions for expressing a desire come with zero degrees of strength. However, the view that speech acts express attitudes like beliefs or desires is only one way of looking at them (cf. Harnish 2005). There is another view of speech acts, which is more compatible with the separation of illocutionary force and sincerity that is suggested in the present paper and was made explicit in the structure of utterances in (1) in the introductory section, repeated below for convenience: (1)

[illocutionary operator [CG-managing operator [proposition]]]

According to this other view speech acts change the commitments of speaker and/or hearer (Alston 2000, Krifka 2001, Searle 2001). For instance by sincerely making an assertion, a speaker commits him-/herself to the truth of the proposition expressed in the assertion.10 A question places the addressee under the commitment to answer, and lays down the form of the answer. From this point of view, the meaning contribution of falsum in polar questions can be described as follows. The speaker s wishes to double-check p, so s asks the addresse a whether or not in all the worlds that conform to a’s knowledge in w and fulfil all the conversational goals of a in w′ the proposition p is not in the CG (CGt ). This corresponds to enquiring about the degrees of strengths of the sincerity conditions for adding p to CG: the addressee is expected to determine whether or not there are zero degrees of strength for adding p to CG, see (11): (11)

a. Ist Paul nicht ins Schwimmbad gegangen? is Paul not to.the pool gone ‘Didn’t Paul go the pool?’ b. [Q [falsum [Paul went to the pool]]] = {There are zero degrees of strength for adding Paul went to the pool to CG; There are not zero degrees of strength for adding Paul went to the pool to CG}

Evidence for the proposal that questions of this kind indeed ‘ask’ for the degrees of strength comes from their answerhood conditions. When nega-

10 Searle (2001: 176): “In making an assertion we take responsibility for truth, sincerity and evidence. And such responsibilities, like commitments in general, have the upward direction of fit. These responsibilities are met only if the world is such that the utterance is true, the speaker is sincere, and the speaker has evidence for the assertion.”

common ground management

241

tive polar questions with outer negation in German are answered, and the addressee chooses for the option that p should be added to the CG, the answer particle doch (henceforth DOCH), which is homophonous with the modal particle doch in its accented form,11 must be used, rather than the neutral ja (‘yes’). This can be seen as an indication that the degrees of strength for sincerely adding p to CG were at issue. What degree of strength does DOCH indicate? If used in isolation, i.e. if used as a replacement for yes, it seems that DOCH indicates a high degree of strength. At first sight, this suggests that the paraphrase given in (11), that there are not zero degrees of strength for adding Paul went to the pool to CG, is too weak. (12) shows that it is not: it is perfectly possible to react to (11) with an utterance like (12): (12) Doch, ich glaube schon. Ich bin mir aber nicht sicher. part I believe indeed I am me but not sure ‘I believe he did but I am not sure.’

(12) signals that the speaker does not have enough evidence for sincerely adding Paul went to the pool to CG, i.e. there is a low degree of strength. This is compatible with the second paraphrase in (11), i.e. that there are non-zero degrees of strength. I suggest that DOCH expresses that there are non-zero degrees of strength. Depending on the context—depending on whether there are further mitigating comments or not—this can be a low, or a high degree of strength. For reasons of space I must leave open here the question of what the exact conditions are for one or the other. I conclude from this discussion that falsum scopes under Q but over the proposition, as suggested in the general formula in (1) above.12 This means that a speech act consists not only of an illocutionary operator and a proposition, as in Stenius (1967), but in addition there are CG-managing operators. On this view, denials are assertions with the falsum operator. We have just seen how the definition of falsum given in (10) can be fruitfully linked to the degrees of strength of the sincerity conditions for speech acts. The degrees of strength are part of the CG status of a proposition as described in the introductory section. In the previous section I suggested that such information must be part of the CG, where the notion of CG is extended to cover more than just the set of propositions held to be true by the interlocutors. With respect to the falsum operator in denials we

11 Note that accented doch has a different meaning from unaccented doch. Only the latter will be investigated in the remainder of this paper. 12 See Repp (2006) for further arguments from gapping that support this kind of structure.

242

sophie repp

may say that after a speaker has uttered a denial, the CG contains the information that the speaker is not committed to the truth of p due to a lack of sincerity—and therefore wishes to remove p from the CG. We may of course ask whether there is independent evidence for the suggestion that information on the degrees of strength of sincerity or similar speech-act relevant information indeed is part of the CG, or of a directly preceding utterance. I suggest that there is: Information of this kind can be called into question by an interlocutor independently of the truth of the ‘bare’ proposition (i.e. the proposition whose CG status has not been specified). Consider the following exchange: (13) Context: Speaker A does not have a watch, or access to some other time-telling gadget. Speaker B has a watch, and knows that speaker A does not know the time. It is 3 o’clock. A: It is 3 o’clock. B: That’s a very good guess.

Speaker B here implies that A cannot make an assertion with totally being committed to it—because A does not have access to precise information about the time. So B says that A is guessing, i.e. has a rather low degree of strength of sincerity for his or her assertion. This information is independent of the fact that the proposition expressed by A is true in this context. Similarly, consider the scenario in (14). It seems quite adequate for speaker B in this context to tell A that A was not sincerely making an assertion even though the proposition that A expressed is true in this context. (14) Context: Speaker A wants to fool speaker B about the time. A looks at his or her watch, which says that it is 10 o’clock, without realizing that the watch is actually an hour fast. B can see A’s watch but knows that it is in fact 9 o’clock. A: It is 9 o’clock. B: You’re lying—but you’re telling the truth.

A final example concerns the operator verum. Repp (2006, 2009) proposes that like falsum, verum is an operator that expresses the degrees of strength of the sincerity conditions. In the case of verum this is the maximum degree of strength. I follow this proposal here. Example (15) illustrates that the high degree of strength expressed by verum can be questioned by a second speaker even though that speaker shares the opinion of the first speaker with respect to the truth of the proposition at hand: (15) A: My friend Mary said that Peter is a zombie. B: He IS a zombie. C: Why are you so certain about this? I also think that he’s a zombie but we should check this to make absolutely sure.

common ground management

243

Again, this indicates that the sincerity of a speaker can be taken issue with independently of the truth of the proposition expressed. Let us move on to the discourse conditions of falsum. In my discussion of previous approaches to denials in Section 2.2.1 I mentioned that denials are reactions to information that has been expressed before: denials react to an immediately preceding utterance that proposes to add the proposition at issue to the set of propositions that are held to be true by both interlocutors. This condition on the discourse is not part of the meaning of falsum given in (10). It cannot be added innocently, though, because polar questions with outer negation seem to behave differently with respect to their discourse context. In polar questions with outer negation the speaker conveys a previous epistemic bias towards p and wishes to double-check that p is part of the CG. This is done by using falsum. The proposition p in this case, however, is not part of the CG but of the private assumptions of the speaker: polar questions with outer negation are felicitous with a(n immediately) preceding context that entails ¬p, or one that entails neither p nor ¬p, but not with one that entails p (Büring & Gunlogson 2000). So although falsum scopes over a positive proposition both in the case of denials and in the case of questions the contextual restrictions are different. I suggest that we capture this difference by making the context restrictions dependent on whether the variable x denotes the speaker or the addressee. Clause (i) in (16) captures the discourse conditions for declaratives (i.e. denials), clause (ii) captures the conditions for questions. Clause (ii) summarizes the generalization proposed by Büring & Gunlogson (2000). Clause (i) requires some further comments. I shall come to that presently. (16) ⟦falsum⟧x = λp⟨s,t⟩ λw. ∀w′ ∈ Epix (w)[∀w′′ ∈ Convx (w′ )[p ∉ CGw′′ ]] Discourse condition for utterance un with ⟦falsum⟧x (p):

(i) for x = speaker: un−1 entails strengthened p (i.e. with the meaning of p including its scalar implicatures) (ii) for x = addressee: un−1 does not entail p

Clause (i) of the above definition says that in a denial, the previous utterance must entail strengthened p. The motivation for this comes from examples like (17) and (18). (17) suggests that falsum can be used felicitously if the proposition in its scope is entailed by the previous utterance: smoking a cigar entails smoking. (17) A: Paul hat gestern Zigarre geraucht. Paul has yesterday cigar smoked

244

sophie repp B: Paul hat nicht geraucht. Paul has not smoked A: ‘Paul smoked a cigar yesterday.’ B: ‘Paul DIDn’t smoke.’

In (18), eating three biscuits entails eating two biscuits. Nevertheless, applying falsum to the proposition that Peter ate two biscuits is incoherent in the given context: (18) A: Peter ate three biscuits. B: Peter DIDn’t eat three biscuits. Bʹ: #Peter DIDn’t eat two biscuits.

The crucial difference between (17) and (18), I suggest, is that (18) contains a scalar expression whereas (17) does not. Geurts (1998) observes that scalar expressions under denial receive an ‘exactly’ reading. So, speaker Bʹ in (18) says that the proposition that Peter ate exactly two biscuits should not be in the CG. This is not entailed by un−1 . Therefore, using falsum in this discourse produces incoherence. Note that under propositional negation scalar expressions do not receive an ‘exactly’ reading. Rather, the upper boundary (not more than 2) is suppressed, just as with other downward entailing operators (e.g. Chierchia 2004; Geurts 2009). The meaning of the sentence in (19) entails not (19a) but (19b): (19) Peter did not eat 2 biscuits. a. Either Peter ate less than two biscuits or Peter ate more than two biscuits. b. Peter ate less than two biscuits.

According to the definition in (16), the ‘exactly’ reading in denials is part of the meaning of falsum: if falsum scopes over a proposition with a scalar expression, the meaning of this proposition is strengthened, i.e. the ‘at least’ reading is strengthened to the ‘exactly’ reading. This then explains the difference between the acceptability of the discourses in (17) and (18): (17B) has no scalar implicature. The strengthened meaning is identical to the nonstrengthened meaning, and this is entailed by un−1 . The discourse is coherent. In (18) the strengthened meaning of the proposition in the scope of falsum is that Peter ate exactly two biscuits, which is not entailed by un−1 .13 13 One might wonder what the status of presuppositions in these denial scenarios is. (i) shows that an accented negation, which I have assumed denotes falsum in declarative sentences, cannot occur in discourses where a presupposition is rejected. The appropriate intonation is the one where there is an accent on the finite auxiliary, i.e. an intonation contour that we find with verum focus (see section 5).

common ground management

245

2.3. Modal Particles as CG-Managing Operators Modal particles have been ascribed different functions on the speech act level. Most commonly they have been considered modifiers of illocution types, e.g. Thurmair 1989; Lindner 1991; Jacobs 1991; Waltereit 2001; Karagjosova 2004; Zimmermann 2004. The particular way in which an illocution type can be modified differs in these approaches. For instance, for ja (roughly ‘recall’), Jacobs (1991) suggests that it may change the illocutionary type from a general type (e.g. assertion) to a more specific one (e.g. j-assertion), which has additional meaning postulates. Waltereit (2001) and Karagjosova (2004) argue that modal particles change the preparatory conditions of a speech act. For example, in an assertion, which comes with the preparatory condition that it is not obvious to both speaker and hearer that the hearer knows p and does not need to be reminded of p (Searle 1969), the use of the particle ja indicates that the speaker thinks that the interlocutors share the conveyed information and that in the present context this information should be reactivated. This clashes with the above-mentioned preparatory condition, which can then be cancelled (Waltereit 2001). More generally, Thurmair (1989) says that modal particles have the function to integrate an utterance into the current context: they are used to point to common knowledge of the interlocutors, to assumptions and expectations of speaker and hearer with respect to a previous utterance, or with respect to the current utterance. Similarly, Karagjosova (2004) characterizes modal particles as conveying interaction-regulating instructions to the hearer by indicating the speaker’s view on the cognitive status of the propositional content of past and current utterances and beliefs. These meaning

(i)

A: Peter hat aufgehört zu rauchen. Peter has stopped to smoke B: #Peter hat nicht geraucht. Bʹ: Peter HAT nicht geraucht. Peter has not smoked A: ‘Peter has stopped smoking.’ B&Bʹ: ‘Peter did not smoke before.’

I do not have the space to discuss this here but observe that examples like (i) are different from the cases of presupposition denial that are normally discussed in the literature, e.g. (ii). In (ii) the entire previous utterance is rejected and then the reason for the rejection is given: viz. that the speaker of the denial takes issue with the presupposition: (ii)

A: Peter has stopped smoking. B: Peter HASN’T stopped smoking: he never DID smoke.

See e.g. Geurts (1998) for an account of presupposition denials of the sort in (ii).

246

sophie repp

contributions are contributions about what I called the CG status of a proposition above: modal particles are CG-managing operators that indicate the CG status of a proposition. For ja and doch (recall that the latter figured prominently in the set of data presented in Section 1) we can state that they are used by the speaker to instruct the hearer to retrieve from CG a proposition which is not currently being considered, i.e. a proposition that was not conveyed as part of the meaning of the directly preceding utterance.14 Doch comes with the additional meaning component that the current proposition is at odds with a proposition or an implicature that was suggested to be added to the CG as part of the previous utterance, or that was presupposed in that previous utterance.15,16 In (20) and (21) I give definitions of what I mean by retrieve, and specify the discourse restrictions for ja vs. doch in a way similar to the definition of falsum that I proposed in the previous section. These

14

Note that both doch and ja can actually occur discourse-initially, e.g.

(i)

Du hast ja einen Zettel auf dem Rücken kleben! you have part a note on the back stick ‘There’s a note stuck on your back!’

(ii)

Maria, du warst doch schon mal auf Teneriffa. Maria you were part already once on Tenerife ‘Mary, you’ve been to Tenerife before, haven’t you.’

For ja this is unproblematic since it is used to signal familiarity and the immediately preceding utterance is just restricted to not entailing, implicating or presupposing the proposition of the ja-utterance. This is trivially fulfilled in the discourse-initial case. For doch it is often assumed that it calls the hearer to consider the propositional content of the current utterance because there was evidence that s/he had not done that before (e.g. Thurmair 1989, Karagjosova 2004). A zero context is compatible with not considering that content. Karagjosova also points out that the discourse-initial use of doch provides a common basis for further discourse by activating common knowledge: the hearer is asked to synchronize his/her beliefs with that of the speaker. So in (ii) the speaker sets the scene for a conversation that is to be taken up after the confirmation by the hearer Mary—that she indeed has been to Tenerife—say, by inquiries about Tenerife. Note that without doch, this utterance would be very impolite (although there are options to make it polite without using doch, e.g. by placing an accent on du (‘you’)). 15 This is a simplification. For larger discourses, this condition should probably be extended to the immediately preceding discourse section and not just the previous utterance— which, however, still awaits detailed investigation. Since the discourses I am looking at contain only two utterances this simplification is harmless. Furthermore, note that ‘being at odds’ essentially means that from the previous discourse it can be inferred that ¬ p (see Egg 2013 [this volume]). 16 Also compare Karagjosova (2004), who discusses the fact that modal particles may react to implicatures, and Egg (2013) [this volume], who shows that utterances containing doch often are used to reject presuppositions of the previous utterance.

common ground management

247

definitions capture the suggestions in the literature that I just summarized. Note that clause (ii) of definition (20), which is part of the definition of the retrieval function (it makes sure that the proposition that is to be retrieved is not part of the immediately preceding utterance) is subsumed by clause (ii) of definition (21). So both ja and doch have this retrieval function. Still, I shall substantially revise the definition of doch in Section 3.1 (definition (28)) as the contrast relations that have been suggested for doch and that are captured by clause (21ii) are not entirely adequate. (20) ⟦ja⟧ = ⟦retrieve⟧ = λp⟨s,t⟩ . p

Discourse conditions for utterance un with the meaning ⟦retrieve⟧(p): (i) CG entails or implicates p

(ii) un−1 does not entail, presuppose or implicate p.

(21) to be revised

⟦doch⟧ = λp⟨s,t⟩ . p

Discourse conditions for utterance un with the meaning ⟦doch⟧(p): (i) CG entails or implicates p

(ii) un−1 entails, presupposes or implicates ¬p.

Before we move on to the combinatorial possibilities of different CG managing operators I would like to discuss very briefly a proposal from the literature according to which nicht (‘not’) can be a modal particle without negative force (see e.g. Helbig 1988, Thurmair 1989). In line with other literature, I will reject this proposal (cf. Meibauer 1990, 1994; Rosengren 1992). Modal, non-negative nicht is said to occur in polar questions and in whexclamatives.17 For polar questions with negation we saw above that they convey that the speaker has a certain bias with respect to the expected answer. For questions with outer negation, we saw that this pragmatic effect can be accounted for in terms of the CG managing operator falsum. For inner negation, we saw that the effect can be explained as one of the— meaningful—negation scoping under the verum operator. This kind of analysis receives strong support from the observation that negative polar

17 There are other contexts where nicht has been suggested to be semantically void, e.g. so-called ‘expletive negation’ in German before-sentences, and counterfactuals, see Schwarz & Bhatt 2006. Krifka (2010) offers an account for before-sentences where nicht has negative force. See Yoon 2011 for a general account of ‘expletive’ negation that takes the negation to be negative, i.e. non-void.

248

sophie repp

questions have very similar pragmatic effects in languages that do not have modal particles, such as English (Ladd 1981; Romero & Han 2004), or Spanish (Ibáñez 1972). So postulating the existence of a non-negative modal particle for German seems stipulative from the cross-linguistic perspective. Turning to wh-exclamatives18 we find that, again, the occurrence of an apparently non-negative negative marker is a cross-linguistic phenomenon. It has been reported e.g. for French (van der Wouden 1994; Muller 1991), Catalan (Espinal 2000; Castroviejo Miró 2006), Spanish (Espinal 2000) and Paduan (Portner & Zanuttini 2000). Again, this suggests that a non-negative modal particle analysis is not very appealing (cf. Meibauer 1990; Portner & Zanuttini 2000; Roguska 2008). Without engaging in this discussion any further I conclude that nicht never is a non-negative modal particle in German. 2.4. Combining Modal Particles Modal particles can be combined. Thurmair (1989) looks at potential combinations of a great number of German modal particles and describes two general types of restrictions. The first type are restrictions that individual particles have with respect to the sentence type they can occur in: In the great majority of cases a combination of particles is impossible if the individual particles cannot occur in the same sentence type.19 For instance the particle etwa (which expresses unexpectedness or undesirability) can only occur in polar questions, whereas halt (which signals that the current proposition is a plausible explanation for the previous proposition), can only occur in declaratives or imperatives. These two particles cannot be combined. The second type of restrictions are pragmatic restrictions. For instance, particles which signal that the current utterance is meant as an explanation for the previous utterance, such as eben (‘obviously’, ‘immutably’), halt (‘obviously’) and einfach (‘just’) cannot be combined with particles like eh and sowieso (both roughly meaning ‘anyway’), which restrict the relevance of the previous utterance to the point that they signal that it is irrelevant. Thus, these particles signal contradicting communicative functions of the current utterance. 18

(i)

Here is an example. Wen hat die nicht alles eingeladen! whom has she.dem not all invited ‘All the people she invited!’

19 As one exception Thurmair mentions denn, which cannot occur in declaratives on its own but in combination with doch and ja, it can.

common ground management

249

Let us turn to the retrieval particles doch and ja. A paper by Lindner (1991) on modal particles carries the title given in (22), which combines ja (‘retrieve’) and doch (‘retrieve & contrast’). (22) Wir sind ja doch alte Bekannte. we are part part old acquaintances Roughly: ‘We are old acquaintances, you must agree, mustn’t you, because you know it as well as I do.’

The fact that ja and doch can be combined follows from the definitions I gave in (20) and (21). The meaning contribution of the two particles is the same, and the discourse conditions ja imposes on its context are a subset of those that doch imposes on its context. The two particles may happily combine.20 With this much in hand let us examine the modalized negative sentences from the introduction. 3. Explaining the Data: Doch, Ja and Negation I will start my discussion with example (5) from the introduction, repeated below, where speaker B’s utterance contains an accented negative marker. Wide scope of the negation over the modal is not available in this case. Narrow scope is incoherent. (5)

A: Paul kann ins Schwimmbad gegangen sein. B: #Paul kann doch (auch) nicht ins Schwimmbad gegangen sein. Paul can part (also) not to.the pool gone be A: ‘Paul can have gone to the pool.’ B: ‘It is (also) possible that Paul has not gone to the pool, why are you not considering this, you should know.’ *neg > mod, #mod > neg

In the following I will first explore why narrow scope is incoherent in this example (Subsection 3.1). This will lead to a redefinition of the meaning of doch, for which I formulated a preliminary version in Subsection 2.2. Afterwards I will turn to the unavailability of the wide scope reading (Subsection 3.2). In Subsection 3.3 I will then investigate example (6) from the introduction, where the negation in speaker B’s utterance is not accented.

20 There is an ordering restriction: ja has to occur before doch. I have nothing to say about this here.

250

sophie repp 3.1. The Retrieval Requirement of Doch/Ja: The Unavailable Narrow Scope Reading of the Accented Negative Marker

The reason for the incoherence of the narrow scope reading in example (5), I suggest, is that it is implicated by (5A), see (23): (23) ∃w′ ∈ Epix (w) [Paul went to the pool in w′ ] implicates

¬∀w′ ∈ Epix (w) [Paul went to the pool in w′ ] ⇔ ∃w′ ∈ Epix (w) [Paul didn’t go to the pool in w′ ]

According to clause (ii) of the definition of doch in (21) above, the utterance preceding the doch-utterance must entail, presuppose or implicate ¬p, which in the case of (5) means that (5A) must entail, presuppose or implicate that it is not possible that Paul did not go to the pool. However, as we just saw (5A) implicates the opposite. A discourse condition for using doch is not met. Therefore the use of doch is not licensed. We might ask, of course, whether the incoherence of (5) indeed is due to the retrieval aspect, or whether it maybe is due to the apparently missing contrast. We can find out by comparing (5) to an example where there is contrast but no retrieval, and to an example where there is retrieval but no contrast. The former would be a comparison with an utterance where doch is replaced by aber (‘but’), which I will show to have the same contrast-indicating function as doch. The latter would be a comparison with an utterance where doch is replaced by ja, which we already know to signal retrieval but no contrast. I will start with aber (‘but’). (24) is a variant of (5) where the conjunction aber (‘but’) replaces doch. (24) is a felicitous discourse.21 (24) A: Paul kann ins Schwimmbad gegangen sein. B. Paul kann aber auch nicht ins Schwimmbad gegangen sein. Paul can but also not to.the pool gone be A: ‘Paul can have gone.’ B. ‘But it is also possible that Paul did not go to the pool.’ mod > neg

To make the comparison between doch and aber valid let us investigate the contrast relations expressed by these two elements in detail. The contrast expressed by aber is often thought to be that a coordination or utterance sequence q but p corresponds to: q and therefore ¬ p, but actually p (cf. Lakoff 1971). The ‘therefore’ part refers to inference patterns that are based

21

The accent on the negative marker is obligatory in this context.

common ground management

251

on world knowledge or linguistic context. This is quite similar to what Egg (2013) [this volume] suggests for doch, except that the direction of inference is reversed (p and therefore ¬ q), and p is not assumed to be in the CG already and therefore is not retrieved from it.22 This characterization of the meaning contribution of aber, which was suggested by Lakoff (1971) and much subsequent literature, is actually a special case of a more general formula where the inference relation between q and p in an utterance (sequence) q but p is not quite so direct. According to e.g. Anscombre & Ducrot (1983), Lang (1991), and Merin (1996), q is assumed to support a certain proposition in the background (typically an inference) whereas p is assumed to ‘speak against’ this proposition. This is captured in the following definition: (25) ⟦aber⟧ = λp⟨s,t⟩ . p

Discourse condition for utterance un with ⟦aber⟧(p): (i) un−1 defeasibly entails r

(ii) p defeasibly entails ¬r.

Let us apply this definition to the dialogues in (26). (26) i. A: Barbara leitet die Sitzung. Barbara chairs the meeting B: Aber Peter ist Abteilungschef. but Peter is head.of.department A: ‘Barbara is chairing the meeting.’ B: ‘But [Peter FOC] is head of department.’ ii. A: Peter leitet die Sitzung nicht. Peter chairs the meeting not B: Aber Peter ist Abteilungschef. but Peter is head.of.department A: ‘Peter is not chairing the meeting.’ B: ‘But Peter is [head of department Foc]’.

First consider speaker B’s utterance in (26i). The subject Peter is focusmarked here (small capitals mark a falling pitch accent), which indicates that all contextually available alternatives to Peter are not head of department. A contextually available alternative is Barbara. Why would speaker B want to focus-mark Peter and express the resulting meaning? After all, the 22 Egg’s (2013) [this volume] definition of doch: p is the doch-proposition, q is an antecedent proposition.

(i)

⟦doch⟧(p)(q) iff both p and p > ¬q are part of the common ground.

252

sophie repp

A-utterance says that Barbara chairs the meeting, and not that she is head of department, so there does not seem to be a proposition of the form x is head of the department where x ≠ Peter in the immediate context as suggested by the subject focus. Now, note that the exchange in (26i) would be somewhat incoherent without the conjunction aber. Further, recall that aber has the effect that the utterances it links are related by inference to a proposition r, i.e. a third proposition in the background. Let us assume that in a situation where (26i) can be uttered felicitously, there is a department rule like Meetings are chaired by the head of department. From this and the utterance of A it can be inferred that Barbara is head of department (= proposition r). B takes issue with r, which is why s/he uses aber. Peter is focus-marked as a direct consequence of the proposition r being in the immediate context: Peter in (26i.B) is a corrective focus with Barbara as the corrected alternative. (26ii) is similar. Let us assume the same department rule. From A’s utterance it can be inferred that Peter is not head of department—let us pretend we are in a guessing puzzle of the who-is-who kind where unforeseeable obstacles like Peter’s being ill are not an option. If Peter is not head of department he is a simple employee (= proposition r). This is the proposition speaker B takes issue with because it is incompatible with Peter being head of department. Again, the focus marking is a consequence of the proposition r—that Peter is a simple employee—being in the immediate context. Let us turn next to a minimal variant of the above data set, where instead of aber, doch occurs, viz. (27). (27) shows that the data pattern exactly the same. (27) i. A: Barbara leitet die Sitzung. Barbara chairs the meeting B: Peter ist doch Abteilungschef. Peter is part head.of.department A: ‘Barbara is chairing the meeting.’ B: ‘[Peter Foc] is head of department. You should know, why are you not considering this?’ ii. A: Peter leitet die Sitzung nicht. Peter chairs the meeting not B: Peter ist doch Abteilungschef. Peter is part head.of.department A: ‘Peter is not chairing the meeting.’ B: ‘Peter is [head of department Foc]. You should know, why are you not considering this’

I would like to argue that the reasoning for the doch-cases in (27) also is exactly the same as for the aber-cases in (26). I illustrate this for the example

common ground management

253

with subject focus. So assuming the same department rule as above it can be inferred from speaker A’s utterance that Peter is not head of department: if Barbara is chairing the meeting she must be head of department. B says that it is Peter that is head of department. The focus marking on the subject indicates that a proposition of the form x is head of the department is in the immediate context where x ≠ Peter. This is compatible with B’s utterance. We find that the contrast relations are exactly the same in the doch- and in the aber-examples. The difference between the two cases is that in the former there is the component of retrieval: by using doch speaker B indicates that the proposition that Peter is head of department is already in the CG. The aber-utterances do not have this meaning component. As a consequence of the above discussion I suggest that we amend our definition of doch in a way that captures the parallelism in contrast formation between aber and doch, viz. that there is a third proposition r that is relevant for the contrast (clauses (iii) and (iv) in the definition in (28)). The fact that doch also has a retrieval function is captured by the same two clauses that made up the discourse conditions for the particle ja (see definition (20i+ii)). (28) ⟦doch⟧ = λp⟨s,t⟩ . p

Discourse condition for utterance un with the meaning ⟦doch⟧(p): (i) CG entails or implicates p

(ii) un−1 does not entail, presuppose or implicate p (iii) un−1 defeasibly entails r (iv) p defeasibly entails ¬r.

With this let us turn back to example (5), for which I suggested that the narrow scope reading is not available because this reading is an implicature of the A-utterance and therefore need not (and cannot) be retrieved. The data in (24), the aber-case, vs. those in (5) tell us that merely uttering the implicature of the previous utterance is harmless in comparison to instructing the addressee to retrieve it. (5) is incoherent whereas (24) is not. As for the contrast involved in (5), I suggest that it concerns what we might call purposeguided hypothetical reasoning. Speaker A says that Peter might have gone to the pool (= ◇q). If the issue of the conversation is to get hold of Peter a reasonable reaction to Speaker A’s utterance would be to look at the pool to find Peter, i.e. to assume that Peter indeed has gone to the pool. We might express this with the proposition we will find Peter at the pool (◇q defeasibly entails r). This then contrasts with B’s suggestion that Peter might be somewhere else (◇¬q, which defeasibly entails ¬ r). This is the same in (5) and (24).

254

sophie repp

Further evidence for the suggestion that it is indeed the retrieval function that is at issue in (5) comes from another minimal variant of (5), viz. (29), where doch is replaced by ja, which has a retrieval function but does not signal contrast. (29), like (5), is incoherent.23,24 (29) A: Paul kann ins Schwimmbad gegangen sein. B. #Paul kann ja auch nicht ins Schwimmbad gegangen sein. Paul can part also not to.the pool gone be A: ‘Paul can have gone to the pool.’ B: ‘It is also possible that Paul DIDn’t go to the pool, recall’. mod>neg

So instructing the addressee by the use of doch or ja to retrieve a proposition from the CG is an inadequate instruction with respect to the CG status of that proposition if it is part (e.g. an implicature) of the current exchange (un−1 ). Finally observe that the narrow scope reading becomes available in the doch-sentence if there is no implicature in the context that is to be retrieved in the target sentence. This is illustrated with the same example in a different context, viz. example (7) from the introduction, repeated below. Again the negative marker is accented here, which is required for a narrow scope reading (but did not save the same example in the context of (5)): (7)

A: Paul muss ins Schwimmbad gegangen sein. Paul must to.the pool gone be B: Paul kann doch (auch) nicht ins Schwimmbad gegangen sein. Paul can part (also) not to.the pool gone be A: ‘Paul must have gone to the pool.’ B: ‘It is also possible that Paul has not gone to the pool, why are you not considering this, you should know.’ *neg>mod, okmod>neg

23 An anonymous reviewer debates this judgement for the ja-variant but notes that a legitimate reaction of speaker A would be to reject (29B) as a superfluous contribution to the conversation, which shows that it is not an appropriate move by speaker B. 24 An anonymous reviewer points out that there are examples which suggest that indicating the retrieval of an implicature by the use of ja is generally possible:

(i)

Einige Studenten haben die Klausur bestanden. Und da ja nicht alle Studenten die Klausur bestanden haben, muss es eine Nachklausur geben. ‘A few students did not pass the exam. And since not all the students passed there will be a second exam.’

An important difference between this example and those discussed above is that it does not involve a turn of speaker. I suggest that it is a legitimate move on the side of the speaker to draw attention to the implicature of his/her own previous utterance as s/he cannot be sure that the addressee was attentive enough. Such an attention-directing move is inadequate for someone else’s utterance because that other person would have chosen a different utterance had s/he not been aware of the meaning component carried by the implicature. The different behaviour of ja in same-speaker utterance sequences can be captured by relativizing the meaning of ja to speakers similarly to the case of the falsum operator (definition (16)).

common ground management

255

(7A) does not implicate, presuppose or entail the proposition p = ◇¬q expressed in (7B). The discourse is coherent on a narrow scope reading. 3.2. Retrieval Conflicting with Denial: The Unavailable Wide Scope Reading of the Accented Negative Marker The next question to ask with respect to example (5) is why the wide scope reading is not available. (5)

A: Paul kann ins Schwimmbad gegangen sein. B: #Paul kann doch (auch) nicht ins Schwimmbad gegangen sein. Paul can part (also) not to.the pool gone be A: ‘Paul can have gone to the pool.’ B: ‘It is (also) possible that Paul has not gone to the pool, why are you not considering this, you should know.’ *neg > mod, #mod > neg

In the introductory section I suggested that the wide scope reading in this example is a denial reading, and in Section 2 I defined as assertions with the falsum operator. For (5) this means that we have two CG-managing operators: falsum and doch. falsum signals that the proposition it scopes over, that Paul can have gone to the pool, should be removed from CG, while doch instructs the addressee to retrieve this proposition from CG and keep it in CG. We observe that falsum and doch give opposing instructions with respect to CG management. They cannot occur in the same utterance. In addition to this conflict there also is the fact that doch instructs the addressee to retrieve a proposition that was just uttered beforehand. This violates one of the discourse conditions of doch. Note that all this also carries over to the particle ja, for which an example was given in the previous section, example (29B). 3.3. Doch and Propositional Negation: The Wide Scope Reading of the Unaccented Negative Marker Example (6) in the introduction, repeated below for convenience, contains the modal particle doch and an unaccented negative marker. The lack of accent suggests that the negative marker does not express falsum but propositional negation. Thus doch is the only CG-managing operator in (6). It retrieves the wide scope reading from CG. If we assume that the proposition r in the background is something like we will find Paul by the pool, as I suggested earlier, we find that from (6A) it can be defeasibly inferred that r, whereas from (6B) on a wide scope reading it can be defeasibly inferred that ¬r. This is an appropriate use of doch according to definition (28).

256 (6)

sophie repp A: B: A: B:

Paul kann ins Schwimmbad gegangen sein. Paul kann doch nicht ins schwimmbad gegangen sein. ‘Paul can have gone to the pool.’ ‘Paul CAN’T have gone to the pool, why are you not considering this, you okneg>mod, *mod>neg should know.’

Note as an aside that (6B) with the above intonation is infelicitous without doch in this context. This is expected given that it is only the polarity that is different in the two sentences. In a discourse that is coherent with respect to information-structural requirements, we therefore expect the polarity of the second sentence, i.e. the negation, to carry an accent: the negative polarity is the focus in this sentence. Curiously, the presence of doch seems to make the accent marking unnecessary. As a matter of fact, we already observed that an accent on the negation in this example actually results in infelicity (example (5)). As for the missing narrow scope reading of (6B), I suggest that it is not available because narrow scope must be indicated by an accent on the negative marker (see the introduction). 4. More Data: Other Modal Particles Let us look at some more modal particles to corroborate the present analysis. The examples in (30) show the available scope options for a negative marker interacting with modality in mini discourses like those above with one of the following modal particles: eben (‘evidently’, ‘immutably’), einfach (‘just’), ruhig (‘no worries’), schon (‘admittedly’), wohl (‘presumably’). These particles all can occur in declarative sentences, the sentence type under investigation here. They do not have retrieval or rejection functions. The behaviour of these particles in the discourses in (30) is quite diverse: one particle—ruhig—cannot combine at all with an accented negative marker independent of the context. Other particles can combine with an accented negative marker but the scope options are limited, which in turn might lead to incoherence in the context of (30A). Note that in general—and as we already saw above—the narrow scope reading, if available, really only is felicitous if auch (‘also’) is placed before the negative marker (which for reasons of readability is not indicated below). (30) A: A: B.i:

Paul kann ins Schwimmbad gegangen sein. ‘Paul can have gone to the pool.’ Paul kann eben nicht ins Schwimmbad gegangen sein. ‘No, Paul obviously cannot have gone to the pool.’ neg>mod, #mod>neg

common ground management

257

B.ii:

#Paul kann einfach nicht ins Schwimmbad gegangen sein. ‘Paul simply cannot have gone to the pool.’ *neg>mod, #mod>neg B.iii: *Paul kann ruhig nicht ins Schwimmbad gegangen sein. (no translation) *neg>mod, *mod>neg B.iv: #Paul kann schon nicht ins Schwimmbad gegangen sein. ‘Paul indeed may not have gone to the pool, but …’. *neg>mod, #mod>neg B.v: #Paul kann wohl nicht ins Schwimmbad gegangen sein. ‘I consider it likely that it is possible that Paul has not gone to the pool.’ *neg>mod, #/?mod>neg

Let us start with (30B.iii), with ruhig, which is unacceptable under any reading. The particle ruhig signals reassurance, and is known to occur only25 together with modals or imperatives (see e.g. Thurmair 1989; Grosz 2011; Schwager 2010). Importantly, ruhig cannot co-occur with epistemic modals, which is the prominent reading of the modal in (30B.iii) (due to the present perfect tense), which explains the unacceptability of this sentence. With a deontic reading (aided by present tense) (30B.iii) would receive a narrow scope reading (with ruhig indicating something like ‘I don’t care, whatever Paul did’), see (31): (31) Paul kann ruhig nicht ins Schwimmbad gehen. ‘Whatever! Paul can not go to the pool: I don’t care.’

The wide scope reading would be infelicitous in (31): the reassuring meaning of ruhig is misplaced in a denial. Amongst the other examples in (30) only (30B.i) with eben allows a wide scope reading, which in this case is a denial reading. Eben has been argued to do different things by different linguists. Most subscribe to the view that it signals immutability of the state-of-affairs the proposition denotes (e.g. Weydt 1969; Helbig 1988; Autenrieth 2002). Thurmair (1989) argues that eben signals that the proposed state-of-affairs is obvious. Importantly, eben has no retrieval function like doch or ja do.26 This suggests that it only contributes to the meaning of the current utterance. Since (30B.i) on a wide scope reading is a denial, I suggest that the speaker uses eben to express that there is an obvious reason for the rejection of the previous utterance.

25

For some exceptions and discussion, see Schwager 2010. According to Karagjosova (2004), eben like doch and ja signals familiarity and indicates that the inferential relation between the preceding and the current utterance is common knowledge. Yet, she takes eben to be indifferent with respect to the question whether these pieces of common knowledge are active beliefs or not. So I conclude that there is not actually a retrieval function associated with eben. 26

258

sophie repp

Before we turn to the narrow scope reading of the eben-example let us consider why (30B.ii) with einfach (‘just’, ‘simply’) does not have a wide scope reading. The particle einfach is described by Thurmair (1989) as signalling obviousness, similarly to eben, with the difference that with einfach this obviousness only holds for the speaker. Now, if the state-of-affairs denoted by the proposition is only obvious for the speaker it is not well-suited as a denial because it cannot be convincing for the addressee. This is why (30B.ii) is infelicitous with a wide scope reading. Turning to the narrow scope readings of the sentences with eben and einfach we find that they are incoherent in the given context. I suggest that the reason for this is the implicature conveyed by the contextual sentence (30A), viz. ¬□q ⇔ ◇¬q, which is the proposition expressed by the B-sentences. Expressly marking that proposition as obvious is redundant: the addressee is already aware of the implicature—otherwise s/he would not have chosen the possibility modal (by the Gricean maxims of quantity and quality). So pointing out the obviousness of this meaning component is inadequate. We observe that CG-managing operators in general seem to be sensitive to implicatures of the previous utterance, although the details differ. Recall that with doch and ja the retrieval of an implicature was not possible if the implicature was a component of the directly preceding utterance. The particle schon has often been described as partly accepting and partly rejecting the previous utterance (e.g. Thurmair 1991: 148). For instance, schon may signal that the speaker confirms the previous utterance but wishes to rule out a (defeasible) entailment of it. Egg (2013) [this volume] offers the following definition:27 (32) ⟦schon⟧ = λp0 λp−1 [p0 ∧ p−1 ∧ according to CG p−1 defeasibly entails : ¬p0 ]

Since accepting (parts of) a previous utterance is incompatible with the meaning contribution of falsum, the modal particle schon is impossible in a denial, and therefore the wide scope reading in (30B.iv) is unavailable: (33) *⟦schon⟧(30A)(30B.iv) = falsum(p) ∧ p ∧ p defeasibly entails: falsum(p) denial reading intended

27 Note again (cf. fn. 15) that the restriction to previous and current utterance is a gross simplification. Egg (2013) [this volume] points out that the propositions which are the arguments of schon often do not correspond to the propositions expressed by these utterances. Importantly, though, deviations from this simple correspondence are reserved for discourse-initial occurrences of utterances with modal particles, and for cases where the previous utterance does not provide a singular proposition (again non-declarative sentences). As I am only concerned with discourses consisting of two adjacent declaratives this simplification is harmless. The conclusion reached below—that the semantics in (32) is too restrictive—is independent of this simplification.

common ground management

259

The meaning of (30B.iv) with narrow scope of the negation in the context of (30A) is given in (34), where p0 = ◇¬q, and p−1 = ◇q. What we find is that the defeasible entailment is a contradiction for all contingent propositions. (34) *⟦schon⟧(30A)(30B.iv)

= ◇¬q ∧ ◇q ∧ ◇q defeasibly entailsCG ¬◇¬q = ◇¬q ∧ ◇q ∧ ◇q defeasibly entailsCG □q

This seems to be a good reason for the unacceptability of (30B.iv). Unfortunately, the picture is a bit more complicated. Consider the discourse in (35).28 The meaning of the schon-utterance is given in (36). Again, the defeasible entailment is a contradiction for all contingent propositions p. Yet the discourse is perfectly acceptable. (35) A: Peter ist klug. Peter is clever

B: Ja, Peter ist schon klug. yes Peter is part clever

(36) *⟦schon⟧(35A)(35B) = p ∧ p ∧ p defeasibly entailsCG ¬p

It is important to note that the way B’s utterance is to be understood in (35) is not just a repetition of (the meaning of) A’s utterance: a plausible reading of B’s utterance is that B is considering some characteristics of Peter that are not as positive as his cleverness but s/he concedes that Peter indeed is clever. So it seems that a proposition that is defeasibly entailed by the first speaker’s utterance—say that being clever defeasibly entails being friendly, competent, not lazy etc.—and that is picked up by schon in the second speaker’s utterance does not actually have to be made explicit in that utterance. It is possible for B to continue with a remark such as that Peter is lazy and therefore not suitable for a particular job. The semantics of schon in (32) is too restrictive in this respect. Since the aim of this paper is not to investigate the specific semantics of schon, I leave the question of how exactly this should be spelled out for further research. Let us nevertheless assume for the present purposes that the basic intuition behind (32) and earlier proposals along these lines is on the right track. If we assume that much, why is it that (35A–B) is felicitous whereas (30A–B.iv) is not? We observed that in the discourse (35A–B) speaker B confirmed A’s utterance and (silently) considered some propositions that did not follow from the linguistic form of A’s utterance. In (30A–B.iv) speaker B makes explicit

28

Thanks to an anonymous reviewer for a similar example.

260

sophie repp

and seems to (partly) confirm an implicature of speaker A’s utterance: viz. ¬□p ⇔ ◇¬p. Now recall our purpose-guided hypothetical reasoning example in Section 3.1: if the purpose of the conversation is to find Peter, uttering ◇ p rather than ◇¬ p might be taken to be a suggestion to go to the pool and look for Peter there. Subscribing to the implicature ◇¬ p would have quite different consequences for future actions of the interlocutors. Therefore I suggest that for the purposes of a conversation like this making explicit the implicature cannot be taken to be a confirmation (as would be required by schon), which is why this discourse is incoherent. Let us finally consider (30B.v) with wohl. According to Zimmermann (2004) the particle wohl signals a weak commitment to the proposition whose illocutionary type it modifies. Repp (2006, 2009) argues that wohl just like falsum is an epistemic illocutionary operator that signals the degrees of strength of the sincerity conditions of a speech act. In the case of wohl the degrees of strength are fairly high. Therefore a combination with falsum, which signals that the degrees of strength are zero is impossible. There is no denial reading available. The reasons for the absence of the narrow scope reading of (30B.v) are somewhat unclear to me and need further scrutiny. Thurmair (1989) suggests that wohl explicitly signals to the addressee that s/he might object to the current utterance, which in the context of (30A) would be incoherent. On the other hand, a combination of wohl with the epistemic possibility modal seems a bit odd even in the absence of negation: (37) ?Paul kann wohl ins Schwimmbad gegangen sein. ‘Paul, I assume, can have gone to the pool.’

I leave these questions for future research. Let me summarize the findings of this section. We found that modal particles influence the scopal interaction of a modal verb and an accented negative marker. I suggested that this—just like in the case of doch and ja—is due to the modal particles being CG-managing operators. The modal particles of this section do not have retrieval functions like doch or ja but they express the speaker’s stance towards the current utterance and its role in the discourse. This influences the discourse appropriateness of the proposition the particle scopes over. For instance, explicitly pointing out the obviousness of the truth of a proposition in a context where it is clear that the addressee subscribes to the truth of that proposition anyway, is not a cooperative move. We also saw that the modal particles that I investigated in this section interact with falsum in a way that is predicted by the analysis of falsum which I suggested in Section 2.2.2.

common ground management

261

5. verum So far I have concentrated on the discourse effects that an accent on the negative marker may have. I assumed that such an accent may indicate a denial reading, which I analyzed as an assertion with the operator falsum. In Section 2.2 I mentioned that it has been suggested (van der Sandt 1991) that denials need not involve a negation at all, i.e. a positive sentence might be viewed as a denial of a negative sentence. One way of signalling such a positive denial seems to be to place an accent on the finite verb: (38) A: Paul hasn’t gone to the pool. B: Paul has gone to the pool.

In this section I will show that matters are not quite so straightforward and that the discourse conditions for an accent on the finite verb are a little surprising from the point of view of positive denials. We shall see that this accent can be used both to express a denial and an affirmation of a previous utterance. To begin, consider (39), which is another intonational variant of our standard discourse example. In (39), the finite verb carries the main stress. This example seems to be interchangeable with the version where the negation is stressed, i.e. the denial version as in (6). The translation suggests that the effect is similar in English. Note that (39B) can only have a wide scope reading of the negation, no matter what the context is. (39) A: Paul kann ins Schwimmbad gegangen sein. B: Paul kann nicht ins Schwimmbad gegangen sein. Bʹ: Paul kann nicht ins Schwimmbad gegangen sein. (= (6B)) A: ‘Paul can have gone to the pool.’ B/Bʹ: ‘Paul can’t have gone to the pool.’

Further note that the apparent interchangeability of the accent on the finite verb vs. the negative marker is not dependent on the finite verb being a modal verb: (40) A: Paul ist ins Schwimmbad gegangen. Paul is to.the pool gone B: Paul ist nicht ins Schwimmbad gegangen. Bʹ: Paul ist nicht ins Schwimmbad gegangen. Paul is not to.the pool gone A: ‘Paul has gone to the pool.’ B/Bʹ: ‘Paul HASN’T gone to the pool.’

An accent on the auxiliary or on the finite verb has been argued to be able to signal different kinds of focus: focus on tense, mood, aspect, the

262

sophie repp

lexical contents, or so-called verum focus, which originally has been taken as focus on the truth of the proposition (Höhle 1988, 1992). Often verum focus is assumed to be the same as polarity focus. In some recent literature, however, verum is not thought to indicate focus, but to be an illocutionary or conversational operator (Romero & Han 2004; Repp 2009; Gutzmann & Castroviejo Miró 2011). I already mentioned this above. Romero & Han (2004) suggest that verum expresses that the speaker is certain that the proposition in the scope of verum should be in the CG. Repp (2009) links this to the degrees of strength of the sincerity conditions, and assumes that verum signals the maximum degree of strength. Let us take this as a working hypothesis for the moment. Note, however, that neither of these proposals formulates conditions on the discourse for the use of verum. Gutzmann & Castroviejo Miró (2011) do take into account some discourse conditions. I will come to that further below. The above data suggest that verum scoping over a negative proposition has the same effects as falsum in positive sentences, in other words verum(¬p) seems to be the same as falsum(p), and it is tempting to hypothesize that verum has the same discourse conditions as falsum. However, we shall see that this is not the case. As already mentioned above, verum cannot only be used to express a denial but it can also do the ‘opposite’: it can signal affirmation. This can be seen in examples like (41) where a speaker uses verum to confirm something that the interlocutor assumes as well. (41) is from a TV talk show that has been recorded and transcribed in the Fokusdatenbank (‘focus data base’), a spoken-language data base developed by Lohnstein & Stommel (2007–2010). (41) /DAS sollte /AUCH erwähnt werden wenn wir sagen wir ham die falsche /ZUwanderungspolitik ich stimme ihnen /ZU die /ist falsch a aber wir ham sehr /VIEle zuwanderer nur keine zuwanderungspoli/TIK S21: Phoenix DW9 [78] ‘THAT should be mentioned as WELL if we say that we have the wrong immiGRAtion politics, I aGREE with you, it is wrong, but we have very MAny immigrants but no immigration POlitics.’

Example (42) below is a constructed minimal variant where the accent is placed on the negation instead of on the finite verb, and the context is altered slightly to make the discourse plausible: the relevant discourse portion here contains negative polarity (we do not have the right immigration politics) instead of positive polarity as in the above example. The result is incoherent. This is expected under the analysis of falsum given in Section 2.2.2, according to which falsum(p) can only apply if the preceding

common ground management

263

utterance entails (strengthened) p. This is not given in (42). The preceding utterance entails ¬ p in this case. (42) #/DAS sollte /AUCH erwähnt werden wenn wir sagen wir ham nicht die richtige /ZUwanderungspolitik ich stimme ihnen /ZU die ist /nicht richtig a aber wir ham sehr /VIEle zuwanderer nur keine zuwanderungspoli/TIK

‘THAT should be mentioned as WELL if we say that we do not have the right immiGRAtion politics, I aGREE with you, it is not right, but we have very MAny immigrants but no immigration POlitics.’

These data suggest that verum imposes different conditions on the CG and the immediately preceding utterance than does falsum. Further evidence for this conclusion comes from examples like (43) below, where our standard example occurs in the context of an utterance that contains a necessity modal. In this discourse the two accentual patterns are not interchangeable: only the pattern with an accent on the finite verb, i.e. (43A–B), is admissible (I am only interested in the wide scope reading here).29 (43) A: Paul muss ins Schwimmbad gegangen sein. B: Paul kann nicht ins Schwimmbad gegangen sein. Bʹ: #Paul kann nicht ins Schwimmbad gegangen sein. Paul can not to.the pool gone be A: ‘Paul must have gone to the pool.’ B/Bʹ: ‘Paul can’t have gone to the pool.’

The fact that (43A–Bʹ) is incoherent follows from the definition of falsum: (43A) does not entail the strengthened proposition in the scope of falsum in (43Bʹ), i.e. (43A) does not entail Paul can have gone to the pool but it is not the case that he must have gone to the pool. Therefore falsum cannot occur in this context. verum in (43B), in contrast, scopes over the proposition it is not the case that Paul can have gone to the pool. This proposition is incompatible with the proposition expressed in (43A). We have a—felicitous—denial reading. Although verum and falsum behave differently in the above discourses they are alike in the restriction that they cannot occur in discourse-initial utterances. The fact that falsum is restricted in this way follows from our analysis since the discourse condition for the use of falsum in denials is that the previous utterance entails strengthened p. If there is no previous 29 One might wonder if we are still dealing with verum—after all, the lexical contents of the two modal verbs are different, so we might be dealing with focus on the lexical contents. However, this leaves the negation in (43B) unaccounted for: it is not the case that the two utterances only differ in the modal verb that is used. So I assume that we are not dealing with lexical focus.

264

sophie repp

utterance that condition is violated. For verum, Gutzmann & Castroviejo Miró (2011) suggest that the operator removes all questions resolved by the utterance it is part of from the partially ordered set that specifies the currently discussable issues, i.e. the question(s) under discussion. This presupposes that there is such a current question under discussion. For our standard discourse, exemplified e.g. in (43A–B), we would have to assume that whether it is possible that Paul went to the pool should have been the current question under discussion. The context, however, was about the necessity of Paul having gone to the pool. This suggests that (43A) and (43B) do not answer the same question-under-discussion. It seems then that the felicity of verum in this discourse remains unaccounted for. Nevertheless, we shall see that the notion of question under discussion is necessary to understand the discourse restrictions of verum. In the following I will demonstrate that both expressing affirmation and expressing a denial in an utterance with verum are only possible in specific discourse circumstances, which differ for the two types of replies. To appreciate the (un)acceptability of the discourses to be discussed note that I give the judgements for an intonation pattern where in the sentence with verum only the finite verb is marked with a falling accent and the remainder of the sentence is deaccented. Some of the judgements will change if there is another accent, for instance on an object DP (e.g. in (49) further below). I will come back to that. As before, I restrict my discussion to declaratives— both as context and as target utterance.30 (44) below is the ‘mirror image’ of (43): in (44) utterance A contains the possibility modal and utterance B contains the necessity modal. Neither an accent on the finite verb nor an accent on the negative marker seems to produce a coherent discourse (again, I am only interested in the wide scope reading). For falsum this result is expected. For verum we observe that utterance B is not a denial of the previous utterance: the proposition expressed by A (◇ p) is compatible with the proposition verum scopes over in B (¬□ p). So B cannot be used as a rejection. Yet, neither does B seem to work as an affirmation. (44) A: Paul kann ins Schwimmbad gegangen sein. B: #Paul muss nicht ins Schwimmbad gegangen sein. Bʹ: #Paul muss nicht ins Schwimmbad gegangen sein. Paul can not to.the pool gone be A: ‘Paul can have gone to the pool.’ B/Bʹ: ‘It is not the case that Paul must have gone to the pool.’ 30

See e.g. Gutzmann & Castroviejo Miró 2011 for questions and imperatives.

wide scope

common ground management

265

Let us look at this in greater detail and start with the conditions for expressing an affirmation by using verum. We shall see that utterance A must (defeasibly) entail B but that there are further discourse conditions to license the use of verum. Consider (45), where A entails the proposition under the scope of verum in B, but not vice versa. This discourse is only felicitous if the question whether Paul had fruit yesterday (i.e. the weaker proposition) was the question under discussion already before the utterance of A. B can be uttered to express that in view of the fact presented in A, that Paul had a banana yesterday, the proposition that Paul had fruit yesterday—which had been at issue earlier on—according to speaker B should be added to the CG because the sincerity conditions come with a maximal degree of strength. Without this kind of context (45) is incoherent. (45) A: Paul hat gestern eine Banane gegessen. Paul has yesterday a banana eaten B: #Genau. Paul hat gestern Obst gegessen. precisely Paul has yesterday fruit eaten A: ‘Paul had a banana yesterday.’ B: ‘Precisely. Paul did have fruit yesterday.’

[Only felicitous if [Q(Paul had fruit yesterday)] is at issue before A’s utterance.]

If A’s utterance does not (defeasibly) entail the proposition expressed by B, e.g. if B’s utterance expresses a stronger proposition than A as in (46) the discourse is incoherent in any context, i.e. even if Paul having had a banana yesterday was at issue before. (46) A: Paul had fruit yesterday. B: #Precisely. Paul did have a banana yesterday.

This is expected if we take into account overall discourse coherence: A’s utterance cannot be taken as an argument to support B’s intention (fed by a maximal degree of strength of sincerity) to add to the CG the proposition that Paul had a banana yesterday, which was at issue before: eating fruit does not (defeasibly) entail eating a banana. More generally, we can make sense of the discourse requirement on the use of verum, viz. that the proposition in the scope of verum should have been at issue even before the preceding utterance if we consider that a simple agreement with the interlocutor’s proposal to add a proposition to the CG could be signaled by silence or a simple yes (see Section 2.1). Pointing out that there is a particularly high degree of strength for adding a proposition to the CG must be motivated—the proposition must have been

266

sophie repp

up for negotiation before, or for a longer stretch of discourse. This is further supported by the fact that verum cannot be used to affirm the proposition that was uttered by the previous speaker if not uttered in a much more complex discourse (the ‘well’ in the English translation comment in (47) indicates that it is not enough that the question whether Paul had a banana is answered by A). (47) A: Paul had a banana. B: #(Yes.) Paul did have a banana.

[Only felicitous if [Q(Paul had a banana)] is at issue ‘well’ before A’s utterance.

Let us turn to verum in denials. In (48) verum scopes over the negation of an entailment of A’s utterance. This is a stronger proposition than the negation of the proposition expressed by A would be, i.e. the proposition expressed by B entails the negation of A. The result is a denial of A by B. (48) A: Paul hat gestern eine Banane gegessen. Paul has yesterday a banana eaten B: Paul hat gestern kein Obst gegessen. Paul has yesterday no fruit eaten A: ‘Paul had a banana yesterday.’ B: ‘Paul didn’t have any fruit yesterday.’

In (49) the scenario is exactly the same: verum scopes over the negation of an entailment of A. This time the proposition verum scopes over is positive. Curiously, the result is incoherent unless the question whether Paul had a banana yesterday was at issue before A’s utterance.31 (49) A: Paul hat gestern kein Obst gegessen. Paul has yesterday no fruit eaten B: #Paul hat gestern eine Banane gegessen. Paul has yesterday a banana eaten A: ‘Paul did not have any fruit yesterday.’ B: ‘Paul did have a banana yesterday.’

[Only felicitous if [Q(Paul had a banana yesterday)] is at issue before A’s utterance.]

31 Note that the fact that fruit appears in utterance A and banana in utterance B in (49) follows from the intention to produce a denial reading here. If the two elements swapped places the discourse would be incoherent. Recall the related discussion in the context of affirmation (example (46) above).

common ground management

267

The difference in acceptability between (48) and (49) at first sight is quite mysterious32 and I do not have the space to explore it in sufficient depth here. However, I would like to point out that this difference finds an echo in deaccentuation patterns we observe in other contexts. Consider (50) and (51) below. Deaccentuation of everything apart from the elements in small capitals (which come with a rising or falling accent) in these examples is only possible if the more specific term in the object position (banana) occurs in a sentence preceding another sentence with a more general term in the object position (fruit). (50) a. Erst hat /Paul eine Banane gegessen und dann hat \Maria Obst first has Paul a banana eaten and then has Maria fruit gegessen. eaten. ‘First Paul ate a banana and then Maria ate fruit.’ b.*Erst hat /Paul Obst gegessen und dann hat \Maria eine Banane gegessen. ‘First Paul ate fruit and then Maria ate a banana.’ (51) a. Erst wollte /Paul keine Banane essen und dann wollte \Maria first wanted Paul no banana eat and then wanted Maria kein Obst essen. no fruit eat. ‘First Paul didn’t want to eat a banana, and then Maria didn’t want to eat fruit.’ b.*Erst wollte /Paul kein Obst essen und dann wollte \Maria keine Banane essen. ‘First Paul didn’t want to eat fruit, and then Maria didn’t want to eat a banana.’

32 For the affirmation context it does not make a difference whether verum scopes over a positive or a negative proposition:

(i)

A: Paul hat gestern kein Obst gegessen. Paul has yesterday no fruit eaten B: #Genau. Paul hat gestern keine Banane gegessen. precisely Paul has yesterday no banana eaten A: ‘Paul didn’t have any fruit yesterday.’ B: ‘Precisely. Paul didn’t have a banana yesterday.’

[Only felicitous if [Q(Paul didn’t have a banana yesterday)] is at issue before A’s utterance.

268

sophie repp

These patterns follow from theories of Givenness33 like Schwarzschild 1999, where (very roughly) a referent is given and can be deaccented if the utterance u it occurs in is entailed by a salient previous utterance (provided focus-marked phrases in u are replaced with existentially bound variables). For instance, in (50) [Mary]foc ate fruit is given in the context of Paul ate a banana (= (50A)), because the focus-marked Mary is replaced by the existentially closed variable x and ∃x. x ate fruit is entailed by (50A). (51) indicates that the same holds for negative sentences. With respect to our data in (48) and (49) we observe that (48) ‘follows’ the licensing pattern as far as the VPs are concerned: eating a banana entails eating fruit. This does not hold for (49). This suggests that we might have a prosodic reason for why (48) is acceptable but (49) is not. This is further corroborated by the observation that the discourse in (49) can be rendered felicitous without a wider context, i.e. behaves like (48), if there is a (strong) accent on banana: banana would be replaced by a variable that would get existentially bound. This then would be fine with the entailment on the VP level. On the sentence, or utterance level things are a bit more complicated. Utterance A in (48) does not entail utterance B. The two have opposing polarities. Furthermore, the negative marker in (48B) is not accented. Hence we might assume that the negation is not F-marked and cannot be replaced by a variable. Now, the negative marker in this example can be accented—if the finite verb is unaccented. Yet, this still does not mean that we necessarily deal with polarity focus here. I suggest that the result would actually be the falsum reading, which arises if there is an accented negative marker in an utterance that is incompatible with the previous utterance. The difference between falsum and polarity focus is that the former is a CG-managing operator whereas the latter ‘only’ indicates semantic alternatives. I assume that polarity focus occurs in answers to (positive) yes-no questions whereas in denial contexts we find falsum. With respect to (48) I propose that the fact that not both the negative marker and the finite verb can be accented in one utterance can easily be explained if we assume that falsum and verum indicate different, incompatible degrees of strength (zero vs. maximal). The givenness-related accentual patterns of discourses involving negation require more detailed investigation, which I cannot provide here. For the present purposes I assume that the reason for the difference between 33 The general role of givenness for the use of verum has been noted before, e.g. Gutzmann & Castroviejo Miró (2011). However, the details as specified in the main text have not been explored before.

common ground management

269

(48) and (49) is a prosodic one. In the former, the VP in A entails the VP in B, which is why we can have deaccentuation. In the latter, the VP in A does not entail the VP in B, which is why we cannot have deaccentuation. Summarizing the discussion of the discourse conditions of verum I propose the following (tentative) analysis of these conditions: verum signals the maximum degree of strength of the sincerity conditions and thereby expresses that the speaker is certain that the proposition in the scope of verum should be in the CG (i.e. basically following Romero & Han 2004; Repp 2006, 2009). In addition, the utterance containing verum is subject to a number of discourse conditions. Below I give the meaning of verum and a non-exhaustive list of discourse contexts where verum can occur, which is based on the observations that I made in the course of the discussion (recall that CG does not include un−1 ). A precise analysis of verum must await further research. (52) ⟦verum⟧x = λp⟨s,t⟩ λw. ∀w′ Epix (w)[∀w′′ ∈ Convx (w′ )[p ∈ CGw′′ ]] (cf. Romero & Han 2004) Some admissible discourse contexts for utterance un with ⟦verum⟧x (p):

(i) affirmation a. Q(p) ∈ CG b. un−1 entails p

(ii) denial: verum[p] a. Q(p) ∈ CG b. un−1 entails ¬p

(iii) denial: verum[¬p] a. Q(p) ∉ CG b. un−1 entails ¬p c. the accentual pattern of un is licensed by considerations of Givenness.

6. Summary I have argued in this paper that modal particles, falsum and verum are CG-managing operators. They indicate the CG status of a proposition by indicating whether or not the proposition is part of the CG, by indicating the interlocutors’ current stance towards the proposition, and by indicating how the CG should develop. The CG status of a proposition has consequences for the discourse appropriateness of the proposition, i.e. CG-managing operators place conditions on the discourse context. This may influence the availability of certain truth-conditional readings of a proposition, which I illustrated in detail for the interaction of the negation with epistemic modal

270

sophie repp

verbs in German. Furthermore, I demonstrated in what way the combination of CG-managing operators in an utterance depends on their specific meaning contribution. For instance, we saw that rejecting and retrieving a proposition at the same time is impossible (e.g. when combining falsum and doch), or that partly agreeing to the truth of a proposition and rejecting it at the same time is impossible (e.g. when combining falsum and schon). The discussion has shown that the assumption that there is a negative operator that does not operate on the level of the proposition but is a CG-managing operator (falsum) can explain a wide range of data and makes possible a unified analysis for certain parallel effects in declaratives and in questions, essentially corroborating earlier suggestions in Repp 2006, 2009. Furthermore, the discussion of verum has revealed that this operator is subject to discourse conditions that had been unknown so far and that deserve much closer scrutiny in future research. We also saw that falsum and verum sometimes overlap empirically in interesting ways, i.e. they seem to express more or less the same (e.g. a denial). Yet we observed that the two operators impose quite different conditions on their occurrence in discourse. I have taken up, and developed further, threads in the current discussion on the nature of the CG, which emphasizes the need for a wider notion of CG than the set of propositions that the interlocutors hold to be true. I have suggested that the CG status of a proposition should also be part of the CG, which opens up an interesting perspective on the interplay of truth-conditional and non-truth-conditional meaning. Acknowledgements This work was supported by the German Research Foundation DFG as part of the Collaborative Research Centre (SFB) 632 ‘Information Structure’ within project A2 at the Humboldt-Universität zu Berlin. Very special thanks go to Andreas Haida for numerous discussions and very helpful comments. I also thank Markus Egg, Patrick Grosz and Noor van Leusen for comments on, and discussion of, earlier versions of this paper, and Horst Lohnstein and Hildegard Stommel for making available to me their verum focus database. I also thank the audience at the DGfS workshop Expressives and other kinds of non-truth-conditional meaning at the 31st Annual Meeting of the German Society of Linguistics (DGfS).

common ground management

271

References Alston, William P. (2000): Illocutionary acts and sentence meanings. Cornell: Cornell University Press. Anscombre, Jean-Claude & Oswald Ducrot (1983): L’argumentation dans la langue. Bruxelles: Pierre Mardaga. Autenrieth, Tanja (2002): Heterosemie und Grammatikalisierung bei Modalpartikeln. Eine synchrone und diachrone Studie anhand von eben, halt, echt, einfach, schlicht und glatt. Tübingen: Niemeyer. Büring, Daniel & Christine Gunlogson (2000): “Aren’t positive and negative polar questions the same?” Ms. USCS/UCLA. Carston, Robyn (1996): “Metalinguistic negation and echoic use”. Journal of Pragmatics 25, 309–330. DOI: dx.doi.org/10.1016/0378-2166(94)00109-X. Castroviejo Miró, Elena (2006): Wh-exclamatives in Catalan. PhD thesis. Barcelona: University of Barcelona. Chierchia, Gennaro (2004): “Scalar implicatures, polarity phenomena and the syntax/pragmatics interface”. In: Belletti, Adriana, ed.: Structures and beyond. Oxford: Oxford University Press, 39–103. Cornides, Thomas (1969): “Der Widerruf von Befehlen”. Studium Generale 22, 1215–1263. Dummett, Michael (1973): Frege: Philosophy of language. Oxford: Duckworth. Egg, Markus (2013): “Discourse particles, common ground, and felicity conditions”. In: Gutzmann, Daniel & Hans-Martin Gärtner, eds.: Beyond expressives. Explorations in use-conditional meaning. Leiden: Brill, 125–149. [This volume]. Ehrich, Veronika (2001): “Was nicht müssen und nicht können (nicht) bedeuten können: Zum Skopus der Negation bei den Modalverben des Deutschen”. In: Müller, Reimar & Marga Reis, eds.: Modalität und Modalverben im Deutschen. Hamburg: Buske, 149–176. Espinal, M. Teresa (2000): “Expletive negation, negative concord and feature checking”. Catalan Working Papers in Linguistics 8, 47–69. URL: http://ddd.uab.cat/ record/2223?ln=en. von Fintel, Kai & Anthony Gillies (2011): “Might made right”. In: Egan, Andy & Brian Weatherson, eds.: Epistemic modality. Oxford: Oxford University Press, 108–130. Geurts, Bart (1998): “The mechanisms of denial”. Language 74, 274–307. URL: http:// www.jstor.org/stable/417868. ——— (2009): “Scalar implicature and local pragmatics”. Mind and Language 24, 51–79. DOI: dx.doi.org/10.1111/j.1468-0017.2008.01353.x. Grosz, Patrick (2011): “German particles, modality, and the semantics of imperatives”. Proceedings of NELS 39, 335–348. Gutzmann, Daniel & Elena Castroviejo Miró (2011): “The dimensions of verum”. In: Bonami, Olivier & Patricia Cabredo-Hofherr, eds.: Empirical Issues in Syntax and Semantics 8, 143–165. URL: http://www.cssp.cnrs.fr/eiss8/gutzmann-castroviejo -miro-eiss8.pdf. Harnish, Robert M. (2005): “Commitments and speech acts”. Philosophica 75, 11–41. URL: http://logica.ugent.be/philosophica/fulltexts/752.pdf. Helbig, Gerhard (1988): Lexikon deutscher Partikeln. Leipzig: Verlag Enzyklopädie.

272

sophie repp

Höhle, Tilman N. (1988): “VERUM-Fokus”. Sprache und Pragmatik 5, 2–7. ——— (1992): “Über Verum-Fokus im Deutschen”. In: Jacobs, Joachim, ed.: Informationsstruktur und Grammatik. Opladen: Westdeutscher Verlag, 112–141. Horn, Laurence R. (1985): “Metalinguistic negation and pragmatic ambiguity”. Language 61, 121–174. URL: http://www.jstor.org/stable/413423. ——— (1989): A natural history of negation. Chicago: University of Chicago Press. Ibáñez, Roberto (1972): Negation im Spanischen. München: Fink. Jacobs, Joachim (1982): Syntax und Semantik der Negation im Deutschen. München: Fink. ——— (1991): “On the semantics of modal particles”. In: Abraham, Werner, ed.: Discourse particles. Amsterdam: John Benjamins, 141–162. Karagjosova, Elena (2004): The meaning and function of German modal particles. PhD thesis. Saarbrücken: Universität des Saarlandes. Karttunen, Lauri (1974): “Presuppositions and linguistic context”. Theoretical Linguistics 1, 181–194. DOI: dx.doi.org/10.1515/thli.1974.1.1-3.181. Krifka, Manfred (2001): “For a structured meaning account of questions and answers”. In: Féry, Caroline & Wolfgang Sternefeld, eds.: Audiatur Vox Sapientiae. Berlin: Akademie-Verlag, 287–319. ——— (2008): “Basic notions of information structure”. Acta Linguistica Hungarica 55, 243–276. DOI: dx.doi.org/10.1556/ALing.55.2008.34.2. ——— (2010): “How to interpret ‘expletive’ negation under bevor in German”. In: Hanneforth, Thomas & Gisbert Fanselow, eds.: Language and logos. Studies in theoretical and computational linguistics. Berlin: Akademie Verlag, 214–236. Ladd, D. Robert (1981): “A first look at the semantics and pragmatics of negative questions and tag questions”. Proceedings of CLS 17, 164–171. Lakoff, Robin (1971): “If ’s, and’s and but’s about conjunction”. In: Fillmore, Charles J. & D. Terence Langendoen, eds.: Studies in linguistic semantics. New York: Holt, Rinehart & Winston, 114–149. Lang, Ewald (1991): “Koordinierende Konjunktionen”. In: von Stechow, Arnim & Dieter Wunderlich, eds.: Semantik. Ein internationales Handbuch zeitgenössischer Forschung. Berlin and New York: de Gruyter, 597–623. Lascarides, Alex & Nicholas Asher (2009): “Agreements, disputes and commitments in dialogue”. Journal of Semantics 26, 109–158. DOI: dx.doi.org/10.1093/jos/ ffn013. van Leusen, Noor (2007): Description grammar for discourse. PhD thesis. University of Nijmegen. ——— (2004): “Incompatibility: A diagnosis of correction”. Journal of Semantics 21, 415–442. DOI: dx.doi.org/10.1093/jos/21.4.415. Lewis, David (1979): “Scorekeeping in a language game”. Journal of Philosophical Logic 8, 339–359. DOI: dx.doi.org/10.1007/BF00258436. Lindner, Karin (1991): “‘Wir sind ja doch alte Bekannte’—the use of German ja and doch as modal particles”. In: Abraham, Werner, ed.: Discourse particles. Amsterdam: Benjamins, 303–328. Lohnstein, Horst & Hildegard Stommel (2007-2010): “Fokusdatenbank”. URL: http:// www.fokus-db.de/. Meibauer, Jörg (1990): “Sentence mood, lexical categorial filling and non-propositional nicht in German”. Linguistische Berichte 130, 441–465.

common ground management

273

——— (1994): Modaler Kontrast und konzeptuelle Verschiebung. Studien zur Syntax und Semantik deutscher Modalpartikeln. Tübingen: Niemeyer. Merin, Arthur (1996): Die Relevanz der Relevanz. Fallstudie zur formalen Semantik der englischen Konjunktion but. Habilitation thesis. University of Stuttgart. Muller, Claude (1991): La négation en français: Syntaxe, sémantique et éléments de comparison avec les autres langues romanes. PhD thesis. Genève: Genève. Öhlschläger, Günther (1989): Zur Syntax und Semantik der Modalverben des Deutschen. Tübingen: Narr. Penka, Doris (2010): Negative indefinites. Oxford: Oxford University Press. Penka, Doris & Arnim von Stechow (2001): “Negative Indefinita unter Modalverben”. In: Müller, Reimar & Marga Reis, eds.: Modalität und Modalverben im Deutschen. Hamburg: Buske, 263–286. Portner, Paul (2007): “Instructions for interpretation as separate performatives”. In: Schwabe, Kerstin & Susanne Winkler, eds.: On information structure, meaning and form. Amsterdam: John Benjamins, 407–426. Portner, Paul & Raffaella Zanuttini (2000): “The force of negation in wh-exclamatives and interrogatives”. In: Horn, Laurence R. & Yasuhiko Kato, eds.: Studies in negation and polarity. Oxford: Oxford University Press, 193–231. Repp, Sophie (2006): “¬(A&B). Gapping, negation and speech act operators”. Research on Language and Computation 4, 397–423. DOI: dx.doi.org/10.1007/s11168006-9022-y. ——— (2009): Negation in gapping. Oxford: Oxford University Press. Roguska, Magdalena (2008): Exklamation und Negation. Berlin: Logos. Romero, Maribel & Chung-hye Han (2004): “On negative yes/no questions”. Linguistics and Philosophy 27, 609–658. DOI: dx.doi.org/10.1023/B:LING.00000338 50.15705.9. Rosengren, Inger (1992): “Zur Grammatik und Pragmatik der Exklamation”. In: Rosengren, Inger, ed.: Satz und Illokution. Vol. I. Tübingen: Niemeyer, 263– 306. van der Sandt, Rob (1991): “Denial”. Proceedings of CLS 27 (Vol. 2), 331–344. Schwager, Magdalena (2010): “Modality and speech acts: troubled by German ruhig”. Aloni, Maria & Kathrin Schulz, eds.: Amsterdam Colloquium 2009. Heidelberg: Springer. Lecture Notes in Computer Science 6042, 416–425. DOI: dx.doi.org/ 10.1007/978-3-642-14287-1_42. Schwarz, Bernhard & Rajesh Bhatt (2006): “Light negation and polarity”. In: Zanuttini, Raffaella & Héctor Campos & Elena Herburger & Paul Portner, eds.: Crosslinguistic research in syntax and semantics. Negation, tense and clausal architecture. Washington, DC: Georgetown University Press, 199–220. Schwarzschild, Roger (1999): “GIVENness, avoidF and other constraints on the placement of accent”. Natural language semantics 7, 141–177. DOI: dx.doi.org/ 10.1023/A:1008370902407. Searle, John R. (1969): Speech acts. An essay in the philosophy of language. Cambridge: Cambridge University Press. ——— (2001): Rationality in action. Cambridge, MA: MIT Press. Selkirk, Elisabeth (1984): Phonology and syntax. The relation between sound and structure. Cambridge, MA: MIT Press. Stalnaker, Robert (1974): “Pragmatic presuppositions”. In: Munitz, Milton K. & Peter

274

sophie repp

Unger, eds.: Semantics and philosophy. New York: New York University Press, 197–213. Stenius, Eric (1967): “Mood and language game”. Synthese 17, 254–274. DOI: dx.doi. org/10.1007/BF00485030. Thurmair, Maria (1989): Modalpartikeln und ihre Kombinationen. Tübingen: Niemeyer. ——— (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. DOI: dx.doi.org/10.1515/mult.1991.10.1-2.17. Vanderveken, Daniel (1990): Meaning and speech acts. Volume I: Principles of language use. Volume II: Formal semantics of success and satisfaction. Cambridge: Cambridge University Press. Waltereit, Richard (2001): “Modal particles and their functional equivalents. A speech-act-theoretic approach”. Journal of Pragmatics 33, 1391–1417. DOI: dx.doi .org/10.1016/S0378-2166(00)00057-6. Wassermann, Renata (2000): Resource-bounded belief revision. PhD thesis. University of Amsterdam. Weydt, Harald (1969): Abtönungspartikel. Die deutschen Modalwörter und ihre französischen Entsprechungen. Bad Homburg: Gehlen. van der Wouden, Ton (1994): “Polarity and ‘illogical negation’”. In: Kanazawa, Makoto & Christopher Piñon, eds.: Dynamics, polarity, and quantification. Stanford: CSLI, 17–45. Yoon, Suwon (2011): “Not” in the mood: The syntax, semantics and pragmatics of evaluative negation. PhD thesis. Universiy of Chicago. Zeijlstra, Hedde (2007): “On the lexical status of negative indefinites in Dutch and German”. Ms. URL: http://ling.auf.net/lingBuzz/000566. Zimmermann, Malte (2004): “Discourse particles in the left periphery”. In: Shaer, Benjamin & Werner Frey & Claudia Maienborn, eds.: ZAS Papers in Linguistics 35, 543–566.

BIASED POLAR QUESTIONS IN ENGLISH AND JAPANESE

Yasutada Sudo

1. Introduction The non-truth-conditional (or ‘use-conditional’) aspects of the meanings of declarative sentences, e.g. presuppositions, conventional implicatures, etc., have been extensively discussed in formal semantics and pragmatics, but in my opinion, those of question meanings are much less understood. This paper focuses on one particular use-conditional aspect of question meanings: biases associated with different kinds of polar (or yes/no) questions (henceforth PQs). In the standard analysis of questions in formal semantics (Groenendijk & Stokhof 1984; Hamblin 1958, 1973; Karttunen 1977; Krifka 2001), what is being asked by a question is analyzed in relation to the truth conditions of its (possible) answers, under the slogan “Knowing the meaning of a question is knowing its (possible) answers”. Therefore, according to this idea, the truth-conditional aspect of the meaning of a question is simply the set of the truth-conditional meanings of the (possible) answers to the question. However, questions may also convey information regarding the questioner’s bias towards a particular answer, which is not amenable to an analysis solely based on the meanings of possible answers (Asher & Reese 2007; Ladd 1981; Reese 2007; Romero & Han 2004; van Rooy & Šafářová 2003). In this paper I focus on inferences regarding biases conveyed by various types of PQs in two historically unrelated languages, English and Japanese. My main objective in this paper is to propose and motivate a novel feature-based description system that is fine-grained enough to characterize various flavors of biases that different forms of PQs encode in natural languages. The organization of the paper is as follows. I will first review in Section 2 Ladd’s (1981) dichotomy of positively and negatively biased PQs in English, and claim that although correct, it is not detailed enough to characterize the intricacy of the biases involved in various kinds of PQs in English. Building on Büring & Gunlogson’s (2000) analysis, I submit that two major types of bias should be recognized, each of which can be positive or negative, and put forward a feature-based theory, which will be shown to successfully describe

276

yasutada sudo

the biases of basic kinds of PQs in English. I will then present in Section 3 further empirical support for the proposed system from Japanese PQs, some of which have biases that are not observed with simple forms of PQs in English. Section 4 concludes the paper. 2. Two Kinds of Bias in English PQs In English, and perhaps in all languages, distinct forms of PQs may have the same truth-conditional meaning, or in other words, they can be used to ask about the same thing. For example, consider the following pair of sentences. (1)

a. Did John come to the party? b. Didn’t John come to the party?

What these questions are asking about is exactly the same, namely whether or not John came to the party. However an interpretive difference between them is glaring. Simply put, negative polar questions with inverted negation (henceforth NPQs) such as (1b) obligatorily carry some information about the questioner’s bias towards one of the answers that positive polar questions (henceforth PPQs) like (1a) do not. Because what is being asked by the two PQs is identical, this difference should lie in a non-truth-conditional aspect of their meanings.1 In this section, I will closely examine inferences involved in basic kinds of PQs in English such as those in (1), and claim that two qualitatively distinct kinds of bias must be recognized. I will first introduce Ladd’s (1981) important classification of NPQs, and then motivate my own view that is more complex than Ladd’s (1981). 2.1. ON- and IN-NPQs Ladd (1981) makes the important observation that there are two readings of NPQs, which he calls Outside-Negation (ON) reading and Inside-Negation (IN) reading. The following examples modeled after Ladd’s (1981) demonstrate that the exact same NPQ can have different flavors of bias.

1 English has other means of forming biased PQs such as tag questions, epistemic really, strong NPIs and marked intonation (Asher & Reese 2007; Ladd 1981; Romero & Han 2004; Reese 2007). I put these aside in this paper, hoping that they are amenable to the analysis put forward here. Also, I do not discuss negative PQs with in situ negation, which appear to be similar to PPQs with respect to the kind of bias they are associated with (see below). See also Asher & Reese (2007) for relevant discussion on negative PQs with in situ negation.

biased polar questions in english and japanese (2)

277

ON reading A: You guys must be starving. You want to go get something to eat? B: Yeah, isn’t there a vegetarian restaurant around here? Moosewood or something like that?

(3)

IN reading A: I’d like to take you guys out to dinner while I’m here. B: But there’s not really any place to go in Hyde Park. A: Oh, really, isn’t there a vegetarian restaurant around here?

Here are intuitive characterizations of the meanings of the NPQs in these examples. In (2), the NPQ suggests that the speaker thinks there is probably a vegetarian restaurant around where she is, and is wondering whether this expectation is correct or not. On the other hand, the string identical NPQ in (3) implies that contrary to the speaker’s initial expectation, she now thinks that there might not be a vegetarian restaurant, and asks for a confirmation of this negative supposition. In a nutshell, in (2), the speaker is biased toward the positive answer, while in (3), toward the negative answer. Ladd (1981) furthermore observes that Positive Polarity Items (PPIs) and Negative Polarity Items (NPIs) disambiguate these readings: when an NPQ contains a PPI, it only has an ON reading and when it contains an NPI, it only has an IN reading. This is demonstrated by the following examples with a PPI too and an NPI either respectively. (4)

a. Didn’t John come to the party too? → John probably came to the party. b. Didn’t John come to the party either? → (I thought he did but) John might not have come to the party.

This observation strongly suggests that the ON and IN readings of NPQs are distinguished in grammar rather than in pure pragmatics. This surely begs the question of how the two readings are compositionally derived (cf. Romero & Han 2004), but in my opinion, the nature of the biases should be further clarified before going into a compositional semantic analysis. Thus, in the present paper, I will be mostly concerned with how to formally characterize biases encoded in various forms of PQs. In particular, I will show that the mere dichotomy of positive vs. negative biases is too coarse to capture the intricate inferences involved in biased PQs. Instead, I put forward a feature-based theory of biases. Building on Büring & Gunlogson’s (2000) analysis, I propose that there are two qualitatively distinct types of bias, which I call epistemic bias and evidential bias. Precise definitions of these notions will be provided below, but roughly put,

278

yasutada sudo

an epistemic bias is about the speaker’s private belief/expectation, while an evidential bias has to do with evidence available in the current conversational context and has to do with a belief that could in principle be shared among the discourse participants. In what follows, I will demonstrate with concrete examples that these two notions are necessary and also sufficient to characterize the biases that are obligatorily associated with three kinds of English PQs, PPQ, ON-NPQ and IN-NPQ. 2.2. PPQs and Evidential Bias Let us start with PPQs. As Büring & Gunlogson (2000) observe, they are associated with a bias regarding evidence present in the conversational context, an inference that I call an evidential bias. As an illustration, consider the following example adapted from Büring & Gunlogson 2000. (5)

[Context: My officemate enters the windowless computer room wearing a dripping wet raincoat.] What’s the weather like out there? a. #Is it sunny? b. Is it raining?

In the context of this example, it can be inferred from the wet raincoat that it is likely to be raining. Therefore the negative answer (“No, it’s not sunny”) is more likely for (5a) and the positive answer (“Yes, it’s raining”) is more likely for (5b).2 The infelicity of (5a) indicates that when evidence suggesting the negative answer (henceforth negative evidence) is available in the context, the questioner is not entitled to ask a PPQ. On the other hand, (5b) shows that a PPQ is compatible with the presence of evidence for the positive answer (henceforth positive evidence). Incidentally, PPQs are fine in neutral contexts where no evidence favoring either answer is available in the conversational context. This is shown in (6), which is originally due to Büring & Gunlogson (2000). (6)

[Context: We’re talking long-distance on the phone.] What’s the weather like out there? a. Is it sunny? b. Is it raining?

2 What I mean by ‘positive’ and ‘negative’ answers is the following. For a PPQ of the form ‘p?’, its positive answer is the one that entails p and its negative answer is the one that entails ¬p. Likewise, for a negative PQ of the form ‘¬p’, its positive answer entails p and its negative answer entails ¬p.

biased polar questions in english and japanese

279

Ladd’s (1981) dichotomy of IN- and ON-NPQs is evidently not enough to capture this type of bias regarding what kind of evidence is available. Here I introduce the notion that has to do with incompatibility with positive/negative evidence, which I dub evidential bias (–). (7)

Evidential Bias (–) If a PQ is incompatible with ‘contextual evidence’ for the positive (resp. negative) answer, the PQ is said to carry a [–positive] (resp. [–negative]) evidential bias.

I adopt here the notion of contextual evidence due to Büring & Gunlogson (2000): (8)

Contextual Evidence Evidence that has just become mutually available to the participants in the current discourse situation.

It should be stressed that an evidential bias is about evidence that is in principle available to all the participants in the current situation, rather than something that is only accessible to the speaker.3 The observation made above shows that English PPQs carry [–negative] evidential bias, and hence are incompatible with contextual evidence for the negative answer. It should be emphasized that the importance of contextual evidence for the characterization of biased PQs has already been pointed out by Büring & Gunlogson (2000), and in fact the descriptive generalizations proposed in this section about evidential bias can be regarded as a restatement of their proposal with new terminology. The innovative aspect of my analysis lies in the claim that there is another flavor of bias involved in PQs, to which I now turn.

3 An anonymous reviewer kindly provided the following example as a case where the relevant evidence is not mutually available, but the same contrast as (6) still remains (cf. Clark & Marshall 1981).

(i)

Context: You find a letter by a funding agency on the work desk of your colleague, and quickly glancing at it you cannot help spotting words like happy, grant sum, etc. Your colleague has been away and is just coming back. You do not know that the letter was placed while he was away.] a. Is it good news? b. #Is it bad news?

Although it is true that the evidence is yet not accessed by the hearer in the given context, notice that it is readily accessible to her. I take this contrast to be showing that the mere availability of the evidence is enough for it to count as contextual evidence, and it does not have to be consciously recognized by all the participants.

280

yasutada sudo 2.3. ON-NPQs and Epistemic Bias

In this subsection I will examine ON-NPQs with the notion of evidential bias proposed above, and claim that a different notion of bias is necessary to capture the intricacy of inferences associated with ON-NPQs. Recall that an ON-NPQ strongly suggests that the speaker is inclined toward the positive answer, as shown by (2), repeated here: (2)

A: You guys must be starving. You want to go get something to eat? B: Yeah, isn’t there a vegetarian restaurant around here? Moosewood or something like that?

What kind of evidential bias does an ON-NPQ have? As the following examples show, ON-NPQs are incompatible with positive evidence. In these examples, the positive polarity item too forces the ON-NPQ construal of the NPQ, whose additive presupposition is satisfied in the given context. (9)

[Context: For a psychological experiment, we are looking for some lefthanded subjects. We have asked some of our friends, but only Mary was lefthanded so far. To my surprise, John is using a pencil with his left hand.] a. #Isn’t John left-handed too? b. Isn’t John right-handed too?

In (9) that John is using a pencil with his left hand is positive evidence for (9a), while it is negative evidence for (9b). The infelicity of (9a) indicates that an ON-NPQ is incompatible with positive evidence, or in our terms, has a [–positive] evidential bias. For the sake of completeness, consider the neutral context given in (10) where no evidence favoring either answer is present. The ON-NPQ is just fine in such a situation. (10) [Context: We just learned that Mary is left-handed, and are wondering who else is. I think John, who is not here, is probably left-handed too, but I am not sure.] Isn’t John left-handed too?

Thus, ON-NPQs are associated with a [–positive] evidential bias. However, I would like to point out that the [–positive] evidential bias is just one kind of inferences that ON-NPQs are obligatorily associated with. More specifically, ON-NPQs also typically imply that the speaker has or had an expectation compatible with the positive answer. Conversely put, in a context where the speaker does not have such a positive expectation, an ON-NPQ becomes infelicitous. This is demonstrated by (11).4 4 I thank an anonymous reviewer for pointing out the inadequacies of the example in an earlier draft.

biased polar questions in english and japanese (11)

281

[Context: We just learned that Mary is left-handed, and are wondering who else is. Given its rarity, I believe that Mary is the only left-handed person among us, so I think it’s very likely that John, who is not around, is righthanded.] #Isn’t John left-handed too?

Just as in (10), there is no contextual evidence about John’s dominant hand, as he is not present in the context. The crucial difference between (10) and (11) is the speaker’s expectation: in (11), the speaker thinks that the negative answer (‘John isn’t left-handed’) is probably true, and the ON-NPQ is infelicitous, while in (10), the speaker has an expectation for the positive answer (‘John is left-handed’). More generally, an ON-NPQ of the form ‘Isn’t p?’ necessarily implies a positive expectation on the speaker’s part that p. This is obviously a separate bias from the [–positive] evidential bias, which is an obligatory component of the meaning of an ON-NPQ. Thus, an ON-NPQ carries two different kinds of bias at the same time. One way to look at them is that the two biases of an ON-NPQ express the conflict between the speaker’s belief/expectation that p, and the lack of mutually available evidence supporting p. Importantly, the speaker’s positive expectation is obligatorily present with ON-NPQs but not with PPQs. This is illustrated by the contrast below. (12) [Context: You told me that you went to the party yesterday. I have absolutely no idea who else did.] a. Did John go to the party too? b. #Didn’t John go to the party too?

Here, the context makes it clear that the speaker’s epistemic state is neutral with respect to who came to the party, and hence is incompatible with a positive expectation. The infelicity of (12b) indicates that it is an obligatory component of the meaning of ON-NPQ. Furthermore the felicity of (12a) shows that it is not a bias that all PQs are associated with. Thus I conclude that ON-NPQs necessarily carry not only a [–positive] evidential bias, but also a positive bias based on the speaker’s belief or expectation. I call this type of bias epistemic bias. (13) Epistemic Bias If a PQ carries an implication compatible with the positive (resp. negative) answer based on what the speaker believes, the PQ is said to carry positive (resp. negative) epistemic bias.

The main difference between evidential and epistemic bias lies in how the inference arises: Evidential bias is about contextual information available

282

yasutada sudo

to all conversational participants, and hence is inherently public, while epistemic bias is rooted in the speaker’s private beliefs, and need not be shared by other conversational participants. One might wonder if evidential bias implies epistemic bias, as it is certainly conceivable that having some contextual evidence influences what the speaker believes.5 I suggest here that the epistemic bias does not have to be based on the speaker’s current belief state, but can be relative to what she believed before acquiring the contextual evidence. This is indeed the case in (9) above. It should also be noted that the ‘modal flavor’ of the epistemic bias is not always relative to the speaker’s belief per se. For example, there are ON-NPQs that imply a positive expectation stemming from the norm/rules (deontic) or what the speaker desires (bouletic), rather than what the speaker believes to be true. Examples of these cases are given in (14), which are taken from Asher & Reese (2007) who attribute them to Huddleston & Pullum (2002). (14) a. Deontic Aren’t you ashamed of yourselves? b. Bouletic Don’t you like it?

(14a) can be felicitously uttered in a context in which the addressees broke some rule. In such a context, the speaker’s expectation that they should be ashamed of themselves is most naturally taken to be relative to the norms of the society, rather than what she thinks is true. Similarly, one can ask the question in (14b) expecting that the addressee likes the referent of it, without holding beliefs about the addressee’s preference. In this case, the expectation is based on what the speaker wants, rather than what they believe. Thus an epistemic bias is not strictly specified for the modal flavor, and can be epistemic, deontic or bouletic. Rather, what is essential about an epistemic bias is that it is based on the speaker’s internal and private state, rather than something publicly shared by the conversational participants. To summarize the discussion so far, we have introduced two crucial notions, evidential and epistemic bias, and we described the biases of English PPQs and ON-NPQs as follows: (15)

Evidential Bias Epistemic Bias PPQ –negative ON-NPQ –positive

5

none positive

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

biased polar questions in english and japanese

283

2.4. IN-NPQs In the final part of the present section we will examine IN-NPQs. Recall that as Ladd (1981) points out, an IN-NPQ suggests that the speaker expects the negative answer to be true. The example in (3) is repeated here. (3)

A: I’d like to take you guys out to dinner while I’m here. B: But there’s not really any place to go in Hyde Park. A: Oh, really, isn’t there a vegetarian restaurant around here?

I will demonstrate below that the bias involved in an IN-NPQ can basically be characterized by the same two parameters, evidential and epistemic bias, but IN-NPQs are associated with a stronger notion of evidential bias than [–positive/negative]. That is, they are not only compatible with the presence of negative evidence but also require it. Firstly observe the contrast in (16). (16) [Context: Bill is right-handed and Mary is left-handed. We’re wondering who else is lefty. John is using a pen with his right hand in front of us.] a. #Isn’t John right-handed either? b. Isn’t John left-handed either?

The fact that John is using a pen with his right hand is positive evidence for (16a), but negative evidence for (16b). This contrast shows that an INNPQ is incompatible with positive evidence but compatible with negative evidence. Furthermore, IN-NPQs are infelicitous in the absence of any contextual evidence, demonstrated by the following example with a neutral context. (17) [Context: In the same context as (16), I think that I have seen Chris, who is not around right now, use a pen with his right hand.] #Isn’t Chris left-handed either?

Recall that ON-NPQs are only incompatible with contextual evidence for the positive answer and are perfectly felicitous in neutral contexts where no contextual evidence supporting either of the answers is available (cf. (10)). On the other hand, IN-NPQs require contextual evidence for the negative answer and are therefore infelicitous in neutral contexts, as in (17). To capture this, I introduce the notion of evidential bias (+): (18) Evidential Bias (+) If a PQ requires contextual evidence for the positive (resp. negative) answer, the PQ is said to carry a [+positive] (resp. [+negative]) evidential bias.

Notice that [+positive] is not equivalent to [–negative]. [+positive] means that positive evidence must be present and [–negative] means that negative evidence must not be present.

284

yasutada sudo

In addition to the [+positive] evidential bias, IN-NPQs are obligatorily associated with a positive epistemic bias. For example, (16) above necessarily implies that the questioner expects that John is lefty on some grounds. To put this in a different way, an IN-NPQ of the form ‘Isn’t p?’ expresses a conflict between the speaker’s expectation that p, and contextual evidence suggesting that ¬p. The table below summarizes our discussion of English PQs. (19)

Evidential Bias Epistemic Bias PPQ –negative ON-NPQ –positive IN-NPQ +negative

none positive positive

Evidently this table does not exhaust the possible combinations of the values of the two parameters. As mentioned in fn. 1, English has other means of creating biased questions, and I believe that they fill in at least some of the gaps in the paradigm. However, instead of discussing other forms of PQs in English, I will turn to Japanese PQs in the next section in order to demonstrate that the present system is capable of describing biases of various forms of PQs in historically and typologically diverse languages. This constitutes further evidence for the claim that biases of PQs in natural language come in two basic categories, evidential and epistemic bias, regardless of their syntactic and morphological forms. 3. Japanese PQs and Their Biases Japanese PQs can be formed from declarative sentences purely intonationally, i.e. with a rising intonation towards the end, or with the combination of a rising intonation and a question particle. There is a variety of question particles in this language, but in the present paper, I especially focus on two of them, -no and -desho, and contrast them with PQs without a particle. (20a) shows a declarative sentence whose PQ counterparts look like (20b). (20) a. Mary-ga kita. Mary-nom came ‘Mary came.’ b. Mary-ga kita-{∅, no, desho}? Mary-nom came-q ‘Did Mary come?’

biased polar questions in english and japanese

285

The three PQs in (20b) share the truth-conditional meaning in that they all ask about the same thing, namely whether or not Mary came. Yet native speakers perceive clear interpretive differences among them, which, I claim, stem from the biases these PQs are associated with. Before proceeding to the concrete data of biased PQs in Japanese, it should be noted that Japanese also has a distinction between two readings of negative PQs, just like English has ON- and IN-NPQs.6 Moreover, as Aihara (2009) points out, PPIs and NPIs tease apart the readings, just as in the case of English. Abusing Ladd’s (1981) terminology for English, I call them ON-NPQs (forced by PPIs) and IN-NPQs (forced by NPIs), and treat them separately in the following discussion. In the examples to follow, we use dareka ‘somebody’ and other wh-ka phrases as PPIs and daremo ‘nobody’ and other wh-mo phrases as NPIs.7 In the remainder of this section, I will closely examine PQs without a particle and those with -no and -desho in this order. Each subsection starts with a summary table of the biases of PPQ, ON-NPQ and IN-NPQ, followed by a set of data verifying them. 3.1. PQs without a Particle The following table summarizes the types of bias that Japanese PQs without a particle obligatorily carry. Notice that none of these combinations of the values appears in the table for English at the end of the previous section. (21) ∅

Evidential Bias

Epistemic Bias

PPQ –negative &–positive none ON-NPQ –negative positive IN-NPQ +negative none

As shown here, Japanese PPQs without a particle do not carry any epistemic bias, but have a different flavor of evidential bias from their English counterparts. That is, they are [–negative] and [–positive] at the same time. Notice that these two biases are not logically incompatible with each other. What

6 Given the morphological and syntactic differences between Japanese and English negative PQs, we cannot regard Japanese negative PQs as the same construction as English NPQs. In particular, the position of the negation morpheme is fixed in Japanese, at least on the surface, and the distinction between inverted and non-inverted negation does not arise. 7 Wh-mo is often called a negative concord item rather than an NPI, and semantically inherently negative. Following the literature, I will gloss it as nobody in the examples to follow. I thank an anonymous reviewer for asking for a clarification on this point.

286

yasutada sudo

this combination means is that the PQ is only felicitous in the absence of contextual evidence for either of the answers, i.e. they are only licensed in neutral contexts. This is illustrated by the following examples. Please note that the English translations of the Japanese examples in this section are only approximate since the biases involved are different. (22) a. Neutral Context [Context: We’re looking for a left-handed person. I’m wondering about John, who is not around.] John-wa hidarikiki? John-top lefty ‘Is John lefty?’ b. Negative Context [Context: My friend has just entered our windowless office wearing a dripping wet raincoat.] #ima hareteru? now sunny ‘Is it sunny now?’ c. Positive Context [Same context as (22b).] #ima ame futteru? now rain is.falling ‘Is it raining now?’

Now let us turn to ON-NPQs without a question particle. They are infelicitous in negative contexts but fine in neutral and positive contexts. Also they obligatorily carry positive epistemic bias. For example, (23a) and (23c) necessarily imply that the questioner wants to go to a good restaurant and to a Japanese restaurant respectively.8 (23) ON-NPQs without a Particle a. Neutral Context [Context: I am in Osnabrück for the first time. My friend Daniel might or might not have been to this city before.] doko-ka oisii resutoran sir-anai? where-ka good restaurant know-neg ‘Don’t you know some good restaurant?’

8 These are bouletic readings. Epistemic and deontic readings are also available in appropriate contexts, but relevant examples are omitted for the sake of space.

biased polar questions in english and japanese

287

b. Negative Context [Context: At a student meeting. A is the student representative and knows who will be present today. B is another student.]9 A: We are all here now. Shall we begin the meeting? B: #dare-ka hokani ko-nai? who-ka else come-neg ‘Isn’t someone else coming?’ c. Positive Context A: (Looking at a guidebook) There are all sorts of restaurants around here. B: doko-ka nihon-shoku nai? where-ka Japanese-food not.exist ‘Isn’t there some Japanese restaurant?’

Here doko-ka ‘where-ka’ is a PPI and forces the ON-NPQ construal of the negative PQs. IN-NPQs without a particle have [+negative] evidential bias and are only felicitous in negative contexts, just like English IN-NPQs. However, they differ from their English counterparts in that they are not associated with any epistemic bias. For example, (24b) below does not necessarily imply that the questioner thinks there should or should not be someone else coming. Rather, it is just asking whether the negative evidence (A’s utterance here) is true or not. In these examples, doko-mo ‘where-mo’ forces the IN-NPQ reading. (24) IN-NPQs without a Particle a. Neutral Context [Context: I am in Osnabrück for the first time. My friend Daniel might or might not have been to this city before.] #doko-mo oisii resutoran sir-anai? where-mo good restaurant know-neg ‘Don’t you know any good restaurant?’ b. Negative Context [Context: At a student meeting. A is the student representative and knows who will be present today. B is another student.] A: We are all here now. Shall we begin the meeting? B: dare-mo hokani ko-nai? who-mo else come-neg ‘Isn’t anyone else coming?’ 9

I thank an anonymous reviewer for clarifying the importance of a context for this example.

288

yasutada sudo c. Positive Context A: (Looking at a guidebook) There are all sorts of restaurants around here. B: #doko-mo nihon-shoku nai? where-mo Japanese-food not.exist ‘Isn’t there any Japanese restaurant?’

3.2. PQ-no The biases of PQs with the particle -no are summarized in the following table. (25) -No

Evidential Bias Epistemic Bias

PPQ-no +positive ON-NPQ-no none IN-NPQ-no +negative

none positive positive

Comparing this table with the table at the beginning of the previous subsection, one notices that adding a question particle changes the biases entirely. But it does not affect the truth-conditional meaning, so what is being asked stays the same. Generally, a question particle is void of truth-conditional import, but operates on a non-truth-conditional dimension of the meaning. I am unable to offer a single meaning for it in the present paper in a way that would capture its semantic contribution in a uniform way. With the question particle -no, PPQs require positive evidence, unlike PPQs without particles. This is illustrated by the following data. The following examples reuse the contexts from the previous subsection, and the PQs here are minimally different in that they have a question particle. (26) PPQs with -No a. Neutral Context [Context: We’re looking for a left-handed person. I’m wondering about John, who is not around.] #John-wa hidarikiki-na no? John-top lefty-cop q ‘Is John lefty?’ b. Negative Context [Context: My friend has just entered our windowless office wearing a dripping wet raincoat.] #ima hareteru no? now sunny q ‘Is it sunny now?’

biased polar questions in english and japanese

289

c. Positive Context [Same context as (26b).] ima ame futteru no? now rain is.falling q ‘Is it raining now?’

ON-NPQs are interesting in that they do not have any evidential bias but just positive epistemic bias. Thus, all of the following are felicitous but obligatorily associated with a positive expectation on the speaker’s part. (27) ON-NPQs with -No a. Neutral Context [Context: I am in Osnabrück for the first time. My friend Daniel might or might not have been to this city before.] doko-ka oisii resutoran sir-anai no? where-ka good restaurant know-neg q ‘Don’t you know some good restaurant?’ b. Negative Context [Context: At a student meeting. A is the student representative and knows who will be present today. B is another student.] A: We are all here now. Shall we begin the meeting? B: dare-ka hokani ko-nai no? who-ka else come-neg q ‘Isn’t someone else coming?’ c. Positive Context A: (Looking at a guidebook) There are all sorts of restaurants around here B: doko-ka nihon-shoku nai no? where-ka Japanese-food not.exist q ‘Isn’t there some Japanese restaurant?’

IN-NPQs with -no have the exact same flavor of evidential and epistemic bias as English IN-NPQs. Thus the English translations in (28) are largely accurate. (28) IN-NPQs with -No a. Neutral Context [Context: I am in Osnabrück for the first time. My friend Daniel might or might not have been to this city before.] #doko-mo oisii resutoran sir-anai no? where-mo good restaurant know-neg q ‘Don’t you know any good restaurant?’

290

yasutada sudo b. Negative Context [Context: At a student meeting. A is the student representative and knows who will be present today. B is another student.] A: We are all here now. Shall we begin the meeting? B: dare-mo hokani ko-nai no? who-mo else come-neg q ‘Isn’t anyone else coming?’ c. Positive Context A: (Looking at a guidebook) There are all sorts of restaurants around here. B: #doko-mo nihon-shoku nai no? where-mo Japanese-food not.exist q ‘Isn’t there any Japanese restaurant?’

3.3. PQ-desho Finally, let us turn to PQs with the particle -desho. (29) -Desho

Evidential Bias Epistemic Bias

PPQ-desho none ON-NPQ-desho –positive IN-NPQ-desho none

positive negative negative

The particle -desho changes the biases in a different way from -no. For example, PPQs with -desho carry strong positive epistemic bias, but no evidential bias. Thus they are perfectly felicitous regardless of the nature of the contextual evidence, and just imply that the speaker expects that the positive answer should be the case. For instance, with the PPQ in (30c) the questioner is trying to confirm his expectation that it is raining now, but since there is positive contextual evidence, the question sounds redundant to some extent. (30) PPQs with -Desho a. Neutral Context [Context: We’re looking for a left-handed person. I’m wondering about John who is not around.] John-wa hidarikiki desho? John-top lefty q ‘Is John lefty’ b. Negative Context [Context: My friend has just entered our windowless office wearing a dripping wet raincoat.]

biased polar questions in english and japanese

291

ima hareteru desho? now sunny q ‘Is it sunny now?’ c. Positive Context [Same context as (30b).] ima ame futteru desho? now rain is.falling q ‘Is it raining now?’

Unlike PPQs with -desho, ON-NPQs with -desho have evidential bias against positive evidence. They also carry strong negative epistemic bias. For instance, (31a) implies that the questioner thinks that Daniel probably does not know any good restaurant. The most natural reading of (31b) is a bouletic one, where the speaker wants no one else to show up, which is a negative expectation. Due to the availability of the negative contextual evidence, (31b) sounds redundant in some sense, but could be perfectly felicitously used to stress the expectation. (31) ON-NPQs with -Desho a. Neutral Context [Context: I am in Osnabrück for the first time. My friend Daniel might or might not have been to this city before.] doko-ka oisii resutoran sir-anai desho? where-ka good restaurant know-neg q ‘Don’t you know some good restaurant?’ b. Negative Context [Context: At a student meeting. A is the student representative and knows who will be present today. B is another student.] A: We are all here now. Shall we begin the meeting? B: dare-ka hokani ko-nai desho? who-ka else come-neg q ‘Isn’t someone else coming?’ c. Positive Context A: (Looking at a guidebook) There are all sorts of restaurants around here. B: #doko-ka nihon-shoku nai desho? where-ka Japanese-food not.exist q ‘Isn’t there some Japanese restaurant?’

Lastly, IN-NPQs with -desho are a mirror image of PPQs with -desho in that they do not have evidential bias but epistemic bias. All of the fol-

292

yasutada sudo

lowing suggest that the speaker thinks/hopes that the negative answer is true.10 (32) IN-NPQs with -Desho a. Neutral Context [Context: I am in Osnabrück for the first time. My friend Daniel might or might not have been to this city before.] doko-mo oisii resutoran sir-anai desho? where-mo good restaurant know-neg q ‘Don’t you know any good restaurant?’ b. Negative Context [Context: At a student meeting. A is the student representative and knows who will be present today. B is another student.] A: We are all here now. Shall we begin the meeting? B: dare-mo hokani ko-nai desho? who-mo else come-neg q ‘Isn’t anyone else coming?’ c. Positive Context A: (Looking at a guidebook) There are all sorts of restaurants around here. B: doko-mo nihon-shoku nai desho? where-mo Japanese-food not.exist q ‘Isn’t there any Japanese restaurant?’

3.4. Discussion The biases associated with the three kinds of PQs in Japanese are summarized in the following table. (33)

Evidential Bias

Epistemic Bias

PPQ-∅ ON-NPQ-∅ IN-NPQ-∅

–negative & –positive none –negative positive +negative none

PPQ-no ON-NPQ-no IN-NPQ-no

+positive none +negative

none positive positive

10 As an anonymous reviewer pointed out to me, the English NPQ in the context of (31a)/(32a) may be considered (almost) insulting. The exact same connotation can arise for the Japanese PQs. This is expected as the negative epistemic biases in these situations can be taken as reflecting the speaker’s rude attitude.

biased polar questions in english and japanese Evidential Bias PPQ-desho none ON-NPQ-desho –positive IN-NPQ-desho none

293

Epistemic Bias positive negative negative

As mentioned already, the question particles do not add anything to the truth-conditional meaning of the PQ, and operate exclusively on non-truthconditional dimensions of meaning, which I believe is a general feature of Japanese particles occurring in all types of sentences. Admittedly, I cannot offer a compositional analysis of the meanings of these question particles that uniformly accounts for the biases of the three kinds of PQs, but the complexity of the present phenomenon strongly suggests that at least part of the biases is grammatically encoded in the meanings of those particles, rather than derived from general pragmatic principles (contra Romero & Han 2004; van Rooy & Šafářová 2003 for English PQs; see Reese 2007 for a similar claim). 4. Conclusions The data presented in this paper suggest that all English PQs are biased in one way or other, as previously suggested by Büring & Gunlogson (2000), and furthermore that the same holds in Japanese. Despite the morpho-syntactic diversity of PQs in these languages, however, their biases are all characterized by the proposed feature-based system with two different types of bias, evidential bias and epistemic bias. These empirical observations naturally lead to several further questions. For example, why is it that all PQs are biased? In other words, why is there not a completely neutral question? One possible answer lies in the pragmatics of questions. As Ginzburg (1995) and van Rooy (2003) among others claim, asking a question demands the hearer to identify the questioner’s intention, which is highly context dependent. Given the context dependency, one can speculate that the evidential bias and/or epistemic bias of a PQ help the hearer infer what type of information the questioner is seeking for what reason. Another interesting open question is the distinction between ON-NPQ and IN-NPQ and the relevance of PPIs and NPIs. As Aihara (2009) observes and I replicated above, this distinction is not specific to English NPQs but can be observed in Japanese negative PQs. As the licensing of NPIs and PPIs is largely a grammatical issue, this suggests that it is not enough to just state

294

yasutada sudo

what biases PQs have, but ultimately it should be explained how the biases arise in a compositional fashion (cf. Romero & Han 2004). Lastly, I would like to mention that Japanese wh-questions have several options of question particles, just as PQs: without a particle, with no and with ka. dare-ga kita {∅, no, ka}? who-nom came q ‘Who came?’

Just as in the case of PQs, the three versions of the wh-question are truthconditionally equivalent, but their interpretive differences are prominent. This suggests that wh-questions, in addition to PQs, have some kind of bias. I leave it open for future research whether the analysis of PQs put forward in the present paper is applicable to wh-questions as well. Acknowledgements I am especially indebted to the two anonymous reviewers for detailed comments and highly helpful suggestions, which greatly improved the quality of this paper. I would also like to thank the audience of the workshop Expressives and Other Truth-Conditional Meaning as well as Eric McCready for discussions and comments. All remaining errors are solely my own. References Aihara, Masahiko (2009): “(Un)biased negative yes-no questions in Japanese”. Proceedings of NELS 38, 13–24. Asher, Nicholas & Brian Reese (2007): “Intonation and Discourse: Biased Questions”. Interdisciplinary Studies on Information Structure 8, 1–38. URL: http://opus.kobv. de/ubp/volltexte/2008/1939/. Büring, Daniel & Christine Gunlogson (2000): “Aren’t positive and negative polar questions the same?” Ms. USCS/UCLA. Clark, Herbert & Catherine Marshall (1981): “Definite reference and mutual knowledge”. In: Joshi, Aravind & Bonnie Webber & Ivan Sag, eds.: Elements of discourse understanding. Cambridge: Cambridge University Press, 10–63. Ginzburg, Jonathan (1995): “Interrogatives: Questions, facts and dialogue”. In: Lappin, Shalom, ed.: The handbook of contemporary semantic theory. Oxford: Blackwell, 385–422. Groenendijk, Jeroen & Martin Stokhof (1984): Studies in the semantics of questions and the pragmatics of answers. PhD thesis. University of Amsterdam. Hamblin, Charles Leonard (1958): “Questions”. The Australasian Journal of Philosophy 36, 159–168. DOI: dx.doi.org/10.1080/00048405885200211.

biased polar questions in english and japanese

295

——— (1973): “Questions in Montague grammar”. Foundations of Language 10, 41–53. URL: http://www.jstor.org/stable/25000703. Huddleston, Rodney & Geoffrey K. Pullum (2002): The Cambridge grammar of the English language. Cambridge: Cambridge University Press. Karttunen, Lauri (1977): “Syntax and semantics of questions”. Linguistics and Philosophy 1, 3–44. DOI: dx.doi.org/10.1007/BF00351935. Krifka, Manfred (2001): “For a structured meaning account of questions and answers”. In: Féry, Caroline & Wolfgang Sternefeld, eds.: Audiatur Vox Sapientiae. Berlin: Akademie-Verlag, 287–319. Ladd, D. Robert (1981): “A first look at the semantics and pragmatics of negative questions and tag questions”. Proceedings of CLS 17, 164–171. Reese, Brian (2007): Bias in Questions. PhD thesis. University of Texas at Austin. Romero, Maribel & Chung-hye Han (2004): “On negative yes/no questions”. Linguistics and Philosophy 27.5, 609–658. DOI: dx.doi.org/10.1023/B:LING.0000033 850.15705.9. van Rooy, Robert (2003): “Negative polarity items in questions: Strength as relevance”. Journal of Semantics 20, 239–273. DOI: dx.doi.org/10.1093/jos/20.3.239. van Rooy, Robert & Marie Šafářová (2003): “On polar questions”. Proceedings of SALT 13, 292–309. URL: http://elanguage.net/journals/salt/article/view/13.292/1827.

EXPRESSING SURPRISE BY PARTICLES

Henk Zeevat

1. Introduction Emotion in the semantics and pragmatics of language is a small subject. Substantial contributions are Kaplan’s (1999) work on Ouch and Potts’s (2005) work on conventional implicature. Especially in the latter work, there is the suggestion that the truth-conditional content of the expression is strictly separated from the emotional expression. My claim would be that this is not what one should expect. Emotions come with presuppositions of the object feared, trusted or loved: the object should exist, the subject must be aware of it and must have some knowledge of the object. The fact that one can be worked up about objects that one merely believes to exist is the special case, and is comparable to the disappearing of the presupposition in (1). It is the special case. (1)

John thinks that Mary has left him and he is glad that she has.

What is special about emotions compared to information is that the emotion by itself gives a motive for action with respect to the object, while information alone does not give that: it requires an additional emotion to be a motive for action. In surprise, one finds the same pattern. In order to be surprised, one needs information that conflicts with an expectation that the information is false. One finds this in lexical expressions of surprise. In (2) it is presupposed that Mary left John but it is an entailment of surprise that John did not expect Mary to leave him, an entailment inherited from the emotion. (2)

John is surprised that Mary left him.

The entailment is not a presupposition since it does not escape negation or maybe as in (3). In fact, the only point of denying surprise or suggesting surprise seems to be to deny the expectation or to suggest it. (3)

John was not surprised that Mary had left him. Maybe John was surprised that Mary had left him.

Surprise and non-surprise are central notions for linguistic communication. Like vision, language interpretation is characterised by a high degree

298

henk zeevat

of local ambiguity: the smallest building blocks (bits of sound signal corresponding to phonemes, morphemes and roots) have a dazzling number of ways in which they can contribute to the construction of an hypothesis about what the speaker wants to achieve with her production of the speech signal. As in vision, the integration of smaller parts into a larger part is the central process by which the local ambiguity is overcome and as in vision, different integration possibilities must be weighed against each other. One can argue about the precise mechanisms that play a role but not at the parameter with respect to which the combinations are weighed for comparison with each other: their respective probability in the context. Any unforced deviance from this principle would be irrational and would lead to avoidable misunderstanding. The process of solving all the ambiguities seems to go with a remarkable speed and—unless there are problems—people do not seem to be aware of the ambiguities at all. But they are clearly present in any account of speech recognition (the set of phonemes that is compatible with a certain segment of the speech signal), in natural language processing (the many thousands of possible analysis trees that a sophisticated large coverage grammar assigns to inputs of maybe 20 words already), in the word senses distinguished by lexicographers and in the many resolution possibilities postulated by discourse theorists and other pragmaticists. One can apply this almost directly to surprise. Here the context is the common ground as given before the interpretation of the clause at hand in which the speaker tries to convey surprising content, i.e. a combination available to the hearer which has a low a priori probability in the context. This content is vulnerable to competing combinations that may be less surprising. Adding the mirative marker is a remedy to this vulnerability: the less surprising combinations are now required to be surprising which they are not and the intended combination can come through and be seen to be surprising. From this perspective, a mirative marker helps in reducing misunderstanding and it seems natural that surprise marking draws material to it in linguistic evolution. It undoubtedly does: the markers range from superfluous tense forms to the rather transparent markers of this paper (from non-mirative and surprise-free sources with meanings like ready, flat, clean, pure, free of movement, complete, his body, first and one). There is little question that mirativity is an attractor. The aim of this paper is to investigate some grammaticalised expressions of surprise, four mirative particles of English: even, only, already and still. Typological work (DeLancey 1997; Malchukov 2003) has shown that it is

expressing surprise by particles

299

useful to have a feature of mirativity for the description of certain tenses and particles: part of their meaning is the expression of surprise, often specialised at a dimension and with an orientation. Only (and other exclusive particles like just or merely) expresses that the size of something is disappointingly small: one expected more. Similarly, even expresses that one expected less and not what the host sentence of even asserts to be the case. Already and still specialise in times: earlier than expected and later than expected. It is not claimed that the particles always determine the overall emotion of the utterance, merely that surprise is present in the interpretation. In (4), the utterance seems primarily to express admiration at John’s feat, but the only is there to deal with the rather weak common ground surprise generated by the audience seeming to assume that others could do this too. (4)

Only John can climb the Matterhorn twice in a day.

(5) expresses anger at the proposal under the discussion and only reacts to the opposite opinion that the speaker must adopt this course of action. (5)

Only a fool would do this.

In Section 6, the prediction is formulated that if the particle is in the main clause and intonationally prominent, the main point of the utterance is to express the relevant kind of surprise, as in (6). (6)

John gave ONLY 5 euro.

My claim is that the above statement of the expression of surprise—including the dimension and the direction—suffices for fully understanding the semantic effects of these particles, and in fact leads to considerable improvement in the understanding of the use of these particles. They would thereby be a clear case against a full separation of the expressive dimension and the content.1 1 Potts (2005, 2007c) postulates for expressives that they always project the speaker’s emotion to the top level, by using a separate expressive dimension. As argued below, only, still and already allow trivialisation and the conflicting expectation can be attributed to anybody in the common ground (or its local version) in all four cases. That gives the possibility of no projection by trivialisation. The attribution of the surprise to others than the speaker also suggests a projection behaviour that is quite different from Potts. That does not mean that Potts is wrong for expressives, only that emotional expression is not always the speaker’s emotion or always meant to be projected. This suggests an account of emotional expression that is more continuous with epistemic expression, e.g. the updated update semantics from the conclusion, since otherwise it is harder to account for evolutionary change from a full expressive into the more grammaticalised means of emotional expression studied here.

300

henk zeevat

Section 2 tries to motivate a linguistically useful account of mirativity in terms of correcting weak expectations. Sections 3, 4 and 5 show where and how the proposed account of mirativity improves the semantic/pragmatic analysis of the four particles, 6 tries to deal with the question whether the analysis motivated in Section 2 does not lead to too much surprise in natural language. 2. How to Model Mirativity? The emotion of surprise is typically triggered by a subject encountering situations that are unexpected in a way that matters to the subject, i.e. in having a bearing on established plans and goals of the subject. The emotion is important in the interaction of humans with their environment since it forces a temporary lowering of confidence and forces a process of change not just to the beliefs but also to the plans and goals. The importance to the subject is hard to model, but expectation can and has been modeled in many ways. It is important to reject some particular ways of modeling expectation since they do not work well for this linguistic problem. One approach is in terms of subjective probabilities. An expectation in these terms would be a proposition for which the subjective probability that it is true is larger than 0.5. For the purpose of the linguistic analysis, this quickly runs into problems. Compare (7) (derived from Umbach 2005) (7)

Things have changed in the Smith family. Yesterday, not only Ronald did the shopping.

Here there must be two expectations: a prominent one (based on observation of the habits of the Smiths and underpinning the use of the negation) that Ronald does the shopping alone. But there is also another expectation, this time based on the normal habits of families like the Smiths and allowing only, that couples do the shopping together. Now with respect to Ronald did the shopping alone this would require a value that is both larger and smaller than 0.5. The same argument can be used for the proposal that expectations are beliefs. This would make Ronald’s lonely shopping last Saturday and its negation both beliefs, and thereby would make the speaker’s beliefs and the common ground between speaker and hearer inconsistent. The alternative is to take expectations to be propositions for which there is some evidence, but not necessarily enough to make them into full scale beliefs which could be the basis for action, even under the kind of pressure

expressing surprise by particles

301

which prevents proper investigation. It is then possible that p and ¬p can both be expectations. Proper beliefs must be expectations. Anything that is doubted (and in particular the negations of proper beliefs) cannot be an expectation. But beyond that, the possibility of the inference that p or the suggestion that p is enough for an expectation. Interestingly, this notion can be generalized from subjects to the common ground between subjects. The common ground is itself a model of the joint beliefs and so also puts a limit on the expectations. It moreover contains joint evidence, joint suggestions and is the joint basis of inference for the subjects entertaining it. While it would be wrong to attribute a strong emotional response to the violation of a weak expectation, violation of a weak expectation has a very similar functional role. Its truth could have opened the possibility of achieving certain goals, its falsity may have cut these hopes short and may cause further revisions. Such weak expectations seem to play a role in other areas. Adversative and confirmative particles where the status of the assumption of the falsity of the host (for the adversative case) or for the assumption of its truth (for the confirmative case) can range from being fully given to being a rather vague inference. Or in the area of the “presupposition” of negative sentences and of wh-questions and yes-no-questions. The observation is that negative sentences “presuppose” the truth of the corresponding positive proposition, that wh-questions require that it is given that there is some true answer to the question, and that even polar questions tend to have a bias to a positive answer. Such presuppositions can be common ground (making the negation a correction), the opinion of the interlocutor, the opinion of others and even inferences from other things that are given in the common ground. Expectations as very weak common ground commitments seem to be useful in these cases as they describe exactly the kind of objects that can meet the presuppositional requirement of these expressions, as in (8). (8)

John dreamt that he would not pass the exam and indeed he did not.

Indeed is a confirmation particle and it picks up what is suggested by John’s dream. Or in (9), where there is a polite suggestion that the audience does not propose to go there because there is no such restaurant, while this would be an excellent idea. The negative expectation is needed for the negative form of the question, but the speaker may have evidence that there is one. (9)

Isn’t there a vegetarian restaurant around here?

302

henk zeevat

There is a good deal of variation in the behaviour of these triggers of weak presupposition. What they have in common is that they can take antecedents that are not straight inhabitants of the common ground but are possible inhabitants of the common ground for which the common ground supplies some evidence. The evidence can be an inference from given material, but also suggestions and opinions, of the interlocutor or by a third person. They need to be possible inhabitants, because overt doubt or denial makes the antecedent unavailable. But within these shared characteristics, different triggers have a varying degree of tolerance (e.g. indeed is exceptional in allowing dreams). A second shared characteristic is that these triggers only marginally allow accommodation. A second dimension of variation is whether or not they allow the presuppositional requirement to be trivialised. This seems to be the case for the temporal mirative particles, only and for polar questions and negations. Even, indeed, adversative particles, wh-questions and additive particles do not allow this possibility.2 (10) a. Nobody ever thought that I ate the cake, and, for the record, I did not. b. ??Nobody ever expected anybody to have a dime, but who has a dime? c. Nobody ever expected anybody else to come, and—indeed—only John came. d. Nobody ever expected him to have left Paris already, and—to be sure—he is still in Paris. e. Nobody ever expected him to arrive later, and—in fact—he is already here. f. ??Nobody ever expected John would not come, but even John came. g. ??Nobody ever expected that Bill would pass and indeed he did. h. ??Nobody ever expected that anybody but Bill would come and Bill came too. i. ??Nobody ever expected that Bill would not come and he came however.

The possibility of trivialisation can—for the particles—be explained by a remaining marking function, after the expectation is removed. Even in the absence of an expectation, only still implies an exhaustive reading of the focus, still and already still have their aspectual meaning, the negation has a truth-conditional meaning and the polar question can be a proper question even if neither answer has been suggested. In contrast, even, indeed, however

2 This makes them more suitable than the two superweak elements not and only used in (7) for examples of conflicting expectations. After all, one of the expectations may be trivialised. Suppose, it is common ground that John never kisses anybody, but now we are visited by Anna who is so pretty that Bill predicts that everybody will kiss her, which turns out to be right: John kisses her. One can now say: Indeed even John kissed her. Or ask: Who did even John kiss?

expressing surprise by particles

303

and too do not have a remaining effect, if they do not link up to an earlier suggestion. And wh-questions seem inappropriate unless their existential presupposition is assumed, since they do not allow no as a proper answer. The formalisation of weak presupposition can be done without further complications along the lines of van der Sandt’s (1992)’s treatment of presupposition. In that treatment a presupposition trigger gives rise to an auxiliary context which is matched against the accessible material in the larger context. If successful this will lead to identifications of new material in the auxiliary context with given material. Otherwise, it is attempted to accommodate the presupposition by adding the content of the auxiliary context to an accessible context. For weak presupposition, there would be an extra possibility of looking into contexts originating in accessible contexts, under positive operators. If this is successful and the presupposition is consistent with the accessible context that is scanned, the presupposition is resolved and identifications take place. If there is no antecedent, in the case of weak triggers, interpretation fails without a possibility for accommodation. In the case of weak triggers with a trivialisation possibility (“superweak” triggers), the presupposition can be skipped if it is consistent with the local context. In the sequel, a treatment along these lines will be assumed. In the case of negations and questions, weak presupposition has a rational basis. It makes no sense to deny something for which there is no reason to think it may be true. Similarly, it makes no sense to ask whether something is the case, unless there is a reason to think it might be the case. And— as indicated above—wh-questions are inappropriate in the absence of the belief that they have a proper answer. For the particles however weak presupposition needs a grammaticalisation account. First of all, there is no other way in which a particle can come into being. Second, lexical versions of the particles seem to lack the possibility of accepting weak expectations as antecedents. The examples in (11) illustrate the difference.3 (11)

a. i.

(Harry thinks others than John came to the party, but nobody else seems to expect anybody else to have come.) ii. Only John came to the party. iii. (false) It is surprising that nobody else apart from John came to the party. iv. (true but inappropriate) Exclusively John came to the party.

3 The observation of differences between lexical and non-lexical markers of exclusivity is due to David Beaver (p.c.). Examples for additive particles are in Zeevat & Jasinskaja 2007.

304

henk zeevat b. i. ii. iii. c. i. ii. iii. iv. d. i. ii. iii. e. i. ii. iii. f. i. ii. iii.

(Harry and only Harry wasn’t expecting John to come.) Even John came to the party. (false) It is surprising that John came to the party. (Correcting someone who does not know about Bill’s traveling plans.) Bill is still in Paris. (false) It is surprising that Bill has not left Paris by now. (mildly inappropriate without yet) Bill hasn’t left Paris. (Correcting someone who does not know about Bill’s traveling plans.) Bill is already in Paris. (false) It is surprising that Bill is in Paris by now. (Harry dreamt that Bill is in Paris.) Bill is indeed in Paris. (false) As we knew already, Bill is in Paris. (I am going to Paris next week.) Bill is in Paris too. (partly false) In addition, Bill is in Paris.

For the particles, weak presupposition seems related to grammaticalisation, in particular movement to a more pragmatical meaning. Stalnaker (1978) observed that violation of the presuppositional requirements of too or even does not lead to undefinedness, but merely to pragmatic incorrectness. The loss of accommodation possibilities and the possibility of expectations to fulfill the presuppositional requirements seems to go together and could be described as a change of the presuppositional requirement to a constraint on the common ground, e.g. demanding that somebody in the common ground has positively addressed the yes-no question. The trivialisation of superweak presupposition can be seen as a further weakening: somebody might address that question positively. Trivialisation seems to work only if the trigger has another expressive role.4 The other features can then be projected to the interpretation while omitting the weak presupposition. This other role can be syntactic (as in expletives) but also semantic or pragmatic. All the above merely served to make one point: that mirativity can be formalised as a weak or superweak expectation that the host sentence is not true.

4 There is by itself nothing against empty particles and it has been suggested that they happen within temporal and cultural bounds (this has been suggested e.g. for German halt and gell). Such particles would however face an evolutionary problem.

expressing surprise by particles

305

In the case of only, the weak expectation would be stronger than the host in the sense that a larger quantity than that specified by the host have the property attributed by the host. In the case of even, it would be the weak expectation that in contrast to others the specified object does not have the specified property. In the case of already the weak expectation would be that the specified state starts later than the time given by the host, in the case of still, the weak presupposition is that the specified state would have ended before the time given by the host. The following is a more formal version of the above. Before the colon, there are the presuppositions, with weak and superweak expectations (superweak is weak allowing trivialisation) indicated with weak and superweak, after the colon the asserted contribution. For the four particles, that is nothing except in the case of only. host(φ) indicates that φ should be resolved to the material contributed by the host. (12) a. only host(P, x, P(x)), superweak(y, P(y), P(x), distinct(x, y)) : ¬P(y) b. even host(P, x, P(x)), weak(y, distinct(x, y), P(y), ¬P(x)) : P(x) c. already host(s, t, at(t, s)), superweak(after(t, s)) : d. still host(s, t, at(t, s)), superweak(after(s, t)) :

What they share in each case is the property of the host or the particle itself correcting the weak expectation. 3. Even The semantics for even proposed here does not contain scales.5 It merely requires that the object provided by the host is not expected to have the property P. Typically that means inferences like (13) are not captured in the semantics. (13) even x P y is less likely to have P than x Therefore not y P

5 Approaches with scales all go back to the pioneering Horn 1969. In the current approach an expectedness scale would be the reason underpinning the weak expectation. Speaker and hearer must assign lower subjective probability to x having P than to y having P.

306

henk zeevat

But one should take the mirative theory seriously. If it is the point of an even-sentence to express surprise at the fact that there are more Ps than expected, the competent speaker will choose the most surprising P or one of the most surprising Ps. This consideration would make the inference an implicature and this seems correct. It does not seem incorrect to support my view that the ILLC drinks will be well attended by saying that even Marie will be coming (unexpectedly alright) even if I know there will be even more unlikely attendants (perhaps my interlocutor is especially interested in meeting Marie). Second, the representation of even should show the similarity with additive particles like too and also, representable as as (14). (14) host(x, P, P(x)), weak(y, Py, distinct(x, y)) :

The even-sentences share with the additive particles the ability to take weak antecedents. (15) extended and varied from an example due to Heim (the two children A and B are talking on fixed telephones) shows this ability for both too and even. (In the latter case, B should be less likely to be in bed than A, in both cases, the parents’ opinion is unimportant). (15) a. A: My parents think I am in bed. B: My parents think I am in bed too. b. A: My parents think I am in bed. B: My parents think that even I am in bed.

Latin et and Greek kai are not just conjunctions but also additive particles6 and equivalents of even. (16) gives Shakespeare’s Latin translation, Plutarchus’s original version and English translations of Caesar’s last words. (16) a. Et tu Brute. ‘Even/also you Brutus’ b. Kai su teknon. ‘Even/also you my son’

But there are also cases with a pure additive meaning like (17) (second example from Giannakidou). (17) a. Et ego in Arcadia. ‘Me too in Arcadia.’ b. Fere kai fruta ‘Bring some fruit too’.

6 This is not unusual: Russian i is similar but more complicated. One would expect this to be a frequent typological situation.

expressing surprise by particles

307

That means that these additive mirative markers need an adapted representation, that allows the negative expectation to be skipped in the last resort on the lines of (18). (18) host(P, x, P(x)), weak(x, y, distinct(x, y), P(y)), superweak(¬P(x)) :

4. Only On many people’s view (Groenendijk & Stokhof 1984; Rooth 1992) the question answer pair (19) A: Who came? B: John.

entails that John was the only one who came. Yet it is also possible to say (20). (20) A: Who came? B: Only John.

The question is then what allows the seemingly superfluous only. The answer is mirativity. More people were expected to come. The other ones did not come. The mirative marker is omitted if there is no expectation to be corrected. The same point can be made with respect to focused constituents as in (21). (21) a. PAUL passed. b. Only PAUL passed.

If as in the view of many the focus by itself means that nobody apart from Paul passed, only is superfluous. Mirativity allows a non-superfluous meaning for the extra only. Umbach’s example (22) is an even more beautiful illustration of the same point. (22) a. Things have changed in the Smith family. Last Saturday, only Ronald did the shopping. b. Things have changed in the Smith family. Last Saturday, Ronald did the shopping.

In (22a), we infer that Ronald used to shop with his wife (the corrected weak presupposition is taken to identify the change), while in (22b), we infer that the identity of the Saturday shopper has changed: it used to be Ronald’s wife, now it is Ronald.

308

henk zeevat 4.1. The Interpretation of the Host

On the view of this paper, only expresses disappointment at the small size of a quantity expressed by its host. Utterances can be used to specify quantities. They do that if they are divided into a quantity wh-question and its answer and the speaker intends that the answer fully answers the question. It is standard to connect this with marked intonation: in such an utterance the answer is intonationally prominent. The important thing for only is that the host must specify a quantity that the speaker could be disappointed by. So the occurrence of only in an utterance is a sign that the host specifies a quantity. The placement of only and the intonation can provide further clues as to what in the sentence specifies the answer and consequently at what the quantity wh-question is and thereby to what quantity is specified. Since this is invariably so, only is not just a mirative particle but also a marker of utterances that specify a quantity. And it is possible and likely that it is sometimes just used in the latter role. Suppose, everybody expects John to be the only guest tonight and that A says (23). (23) If only John comes, than we will have enough to eat.

Without the only and prominent intonation on John, this seems a dangerous statement to make: others may be along after all and create a food shortage. With just prominent intonation there may be a contrast with a bigger eater than John, but it must somehow be ruled out that others come as well. These uses should be typical of only in subordinated position. It is not ruled out that there is an expression of surprise, but it will normally not be the point of the utterance. The host therefore has its own semantic contribution to make. It must first of all be an exhaustive settling of the quantity wh-question, the meaning it would have if the wh-question would be made explicitly and the host offered as an answer to it. Something like (24) if the host is Pa, comparable to the cleft or pseudocleft version of Pa. (24) z = the P : z = a

But this misrepresents the division in given and new coming from only. That a is part of the Ps is given and there is a mistaken expectation of a more inclusive exhaustive answer. The new information of the host, i.e. what it asserts is therefore just denying that anything apart from a is part of z. That gives (25).7 7

It would be more satisfactory if the host meaning was just an ordinary meaning that

expressing surprise by particles

309

(25) a, Pa : ∀z(distinct(a, z) → ¬Pz)

Merging that with the semantics for only gives a schematic version of the standard only-sentence in (26). (26) x, Px, superweak(x, y, P(x), P(y), distinct(x, y)) : ¬P(y), ∀z(distinct(x, z) → ¬Pz)

There is still a superfluous element here. The assertion of only, ¬Py is entailed by ∀z(distinct(x, z) → ¬Pz) (since y is distinct from x). So it is possible to reduce (26) further to (27). It is also possible to remove the assertive part from the semantics of only entirely and to consider it as just a normal anaphoric particle, like so many others. (27) x, Px, superweak(x, y, P(x), P(y), distinct(x, y)) : ∀z(distinct(x, z) → ¬Pz)

And this gives the final analysis of only-clauses.8

(28) Only John sneezed. j, sneeze(j), superweak(j, y, sneeze(j), sneeze(y), distinct(x, y)) : ∀z(distinct(z, j) → ¬sneeze(z))

If nobody else sneezed, and nobody even weakly suggested anybody else did, the weak presupposition trivialises and the only contribution is the assertion that nobody else did. Finally, the strong presupposition can be accommodated, giving the projection behaviour illustrated by (29). In all four cases, John sneezed is projected. the host could have in other stereotypical situations, but I have not been able to find such stereotypical situations. I am here following the lead of Kasper (1992) in defining what is asserted as what is stated minus what is presupposed, an idea that can be fruitfully applied to the “Konjunktiv II”, but also to generic sentences. This version of the semantics of the host is the one needed or so it seems. The more inclusive (23) does not give the right results for the negation of only-sentences. I am committed to defending that only disambiguates the host by two methods: (i) by requiring it to define a quantity and (ii) by marking part of that definition as given. 8 The claim is that the host is interpreted as an exhaustive answer to the topic question, that is given both as an object and as an answer to the topic question. This would be a normal reading of the host, forced by the fact that only both requires the host to specify a quantity and refers to an extended expected exhaustive answer to the topic question. It is this semantic contribution of the host that needs to be combined with the semantic contribution of only, the expected larger quantity denoted by the topic question. This is just downwards quantity mirativity combining with the only reading of the sentence that is compatible with downwards quantity mirativity, i.e. the one with that particular topic-focus division in which it is given that the focus meets the topic. So the semantics of only is exhausted by downwards quantity mirativity and the combination of the particle and the host is just conjunctive.

310

henk zeevat

(29) a. b. c. d.

Did only John sneeze? If only John sneezed, why worry? Not only John sneezed. Maybe only John sneezed.

The “scope” of only allows negative polarity items, which follows from our construction of the assertion part in the host semantics. (30) a. Only John ever sneezed. b. Only John lifted a finger.

And, finally, this semantics can explain non-cancellation. Gazdar (1979) lists a whole series of cancellation causes. Only one remains in full force, as can be seen from (31). (31) a. If John sneezed, not only John sneezed. b. If John sneezed, only John sneezed.

But these are precisely the anaphoric cases where Heim (1983), van der Sandt (1992) and Soames (1989) showed that Gazdar’s explanation is inadequate. Conflicts with the common ground or other implicatures do not seem to lead to cancelling. All four combinations in (32) are weird. (32) [John did not sneeze.] a. b. c. d.

??Did only John sneeze? ??But, if only John sneezed, why worry? ??But not only John sneezed. ??But maybe only John sneezed.

Compare this with the same pattern with a well-attested standard presupposition trigger as in (33) (in the two exceptions, the context makes the question or the maybe improper). (33) [John did not sneeze.] a. ??Does he regret sneezing? b. So if John regrets sneezing, I’ll eat my hat. c. So John does not regret sneezing. d. ??Maybe he regrets sneezing.

The explanation is the double nature of the presupposition of only-sentences: both a superweak and a standard presupposition. Superweak antecedents that are not possible inhabitants of the common ground are not possible, even under trivialisation9 and this is the case in (32). The super-

9

Trivialisation can be described as adding somebody who might think that p to the

expressing surprise by particles

311

weak presupposition cannot be treated in that case and it does not allow accommodation. It is predicted that the situation dramatically improves if global presupposition accommodation is blocked without destroying the weak antecedents. This happens if the speaker is overtly ignorant about the presupposition. This is enough to block global presupposition accommodation, and indeed we seem to get local accommodations here, with A’s contribution functioning as an antecedent for the weak presupposition (that John sneezed can now be added to the common ground, as in The cat is on the mat, but Bill does not believe it). (34) A: John seems to have sneezed. B. [I do not know whether John sneezed.] a. b. c. d.

Did only John sneeze? But, if only John sneezed, why worry? But not only John sneezed. But maybe only John sneezed.

The view on only-sentences only X P in this paper is not that the host P(X) (the prejacent) is presupposed, but that the prejacent is interpreted in such a way that it has a presupposition x, Px which is also part of the weak presupposition implementing only’s mirativity. This double nature explains why the presupposition has special features. 4.2. Formalisation The semantics for only-sentences (35) is schematic and requires filling in the details for the different semantic types to which only can apply and with respect to the context in which it occurs. The latter aspect is here relegated to the question that is assumed: the intended question to which the only-sentence gives a disappointing answer can be quite fine-grained and have many further pragmatic restrictions. The question given with the examples below is just an indication. These questions can be further specified by extra domain restrictions. (35) x, Px, superweak(x, y, Px, Py, distinct(x, y)) : ∀z(distinct(x, z) → ¬Pz)

In (36), the type is natural numbers and distinct comes out as “larger natural number”. This is because an event of three students coming is always also an event of two students coming. common ground. That creates the necessary antecedent if the common ground does not throw doubt on p.

312

henk zeevat

(36) Only THREE students came. How many students came? 3, 3 students came, superweak(m, 3 students came, m students came, m > 3) : ∀k(k > m → ¬k students came)

In the construction of the sentence as in (37) (a rather inaccessible reading, but compare Only a student came) distinctness must be a question of necessarily disjoint kinds. (37) Only three STUDENTS came. What kind of X did there come three of? students, 3 students came, superweak(K, 3 students came, 3 K came) : ∀K(distinct(K, students) → ¬3 K came)

In the next example (38), distinct is disjointness as for sets.

(38) Only THREE STUDENTS came. Who came? X, X came, superweak(X, Y, X came, Y came, distinct(X, Y)) : students(X), 3(X), ∀Z(disjoint(Z, X) → ¬Z came)

In (39), the events must be distinct by not sharing subevents (separate). (39) Only three students came. What happened? e, e = [students(X), 3(X), X came], superweak(e, e′ , happened(e), happened(e′ ), separate(e, e′ )) : ∀e′′ (separate(e, e′′ ) → ¬happened(e′′ ))

In (40), distinctness must be analogous to the case of numbers. Larger degrees of passing (getting As or Bs) must count as distinct, being nonentailed. (40) I only PASSED. How did I do at the exam? PASS, I PASSed, superweak(PASS, P, I Ped, I PASSed, superior(PASS, P)) : ∀Q(distinct(PASS, Q) → (¬I Qed))

In the last example (41) the objects are circumstances. These can be thought of as sets of basic propositions holding at the same time. A circumstance is distinct from another if they do not share basic propositions. (41) Only if you sing, I will stay. Under which circumstances will I stay? ◇[you sing, I will stay], superweak(p, distinct(p, you sing), ◇[p, I will stay], ◇[you sing, I will stay]) : ∀q(distinct(q, you sing) → ¬◇[q, I will stay])

The semantics for if p, q as ◇(p ∧ q) is unusual but not novel and has been pioneered by Kratzer (1979) and Sæbø (1985), in the last case even for only. It can be forced by operators like can, may and sometimes as in (42).

expressing surprise by particles

313

(42) a. If John comes, Mary may not want to come. b. If John was there, Mary sometimes visited. c. It can be slippery, if it gets cold.

It is however also possible to force this reading without an overt operator as in (43). (43) a. A: B: b. A: B:

Did Mary never visit if John was there? No, she did visit, if John was there. But sometimes only. When did Mary visit? Mary visited, if John was there. But not very often.

What these examples have in common is that the topic switches from the normal What happens if so and so? to Under what circumstances does so and so happen? If the analysis of only in this paper is on the right track, only forces a reading on its host where it answers a question provided by the host. Given the position of only in an only-if -sentence, this forces the alternative topic-focus structure and the alternative existential interpretation just like sometimes, can and may, or the environments of (43) do. On the view of Kratzer (1979) and Sæbø (1985), the operators outscope if, including an invisible one for the examples in (43). One does slightly better by assuming that if is a locative for hypothetical “locations”. It can then behave like other generic sentences with locatives. In (44), the first example is preferentially interpreted as universal, the second as existential. Operators may help to disambiguate next to projecting their own meaning, but the ambiguity is already there without them. (44) a. In Paris, John drinks beer. b. John drinks beer in Paris.

The much stronger universal interpretations of if-then-sentences would be due to the way one can know what goes on in hypothetical circumstances. Mathematical proof and causal and other rules can bring about much strength, while the not so strong standard universal readings rely on probabilistic reasoning based on multiple observations. 5. Already and Still The semantics of already and still 10 would consist of directed temporal mirativity, i.e. in the superweak presuppositions and in at(t, s), the semantics of

10

Already can also modify non-stative sentences and in English then stands in opposition

314

henk zeevat

the host. s is an occasion of some proposition being true and determines a maximal interval of moments of time at which this is the case. (45) a. still s, t s, superweak(t, s, end(s) < t)) : at(t, s) b. already s, t s, superweak(t, s, start(s) > t) : at(t, s) c. John is still in Paris s, superweak(t, s, end(s) < t) : s, in(s, j, p), at(t, s), now(t) d. John is already in Paris s, superweak(s, t, start(s) > t) : s, in(s, j, p), at(t, s), now(t)

It is moreover clear that start(s) and end(s) themselves have a strong lexical presupposition: the existence of s. In the van der Sandt treatment adopted in this paper, checking the weak presupposition cannot occur before the presuppositional requirements of the weak presupposition itself are fulfilled. That entails that s is a strong presupposition of both still- and already-sentences and that accommodation is necessary, if s is not already given. This can be brought out in our notation, as in (46). (46) a. still s, t s, superweak(s, t, end(s) < t)) : at(t, s) b. already s, t s, superweak(s, t, start(s) > t) : at(t, s) c. John is still in Paris s, superweak(s, t, end(s) < t) : s, in(s, j, p), at(t, s), now(t) d. John is already in Paris s, superweak(s, t, start(s) > t) : s, in(s, j, p), at(t, s), now(t)

The literature11 associates an anteriority presupposition with still, that s must also be true at a moment of time before t. With already, one can have the posteriority presupposition that there is already a moment in the future in which s must be true. This treatment has intuition on its side (though this would seem to be the same intuition as the one behind a mirative analysis) and the weak presuppositions proposed are their counterparts. A moment during s before end(s) will do for still, a moment during s after start(s) will do for already. There are three cases to consider: a strong antecedent for the weak presupposition, a weak antecedent for the weak presupposition and no presupposition resolution at all. to only as in John will arrive already/only tomorrow. The generalization to this other use of already and this use of only should be obvious. 11 A quite comprehensive treatment is Löbner 1989. A recent treatment of still is Greenberg 2009 building on Krifka 2000 and Ippolito 2007.

expressing surprise by particles

315

In the last case, the requirement on the common ground is that the weak presupposition is not false. For already, this means that it cannot be CG that start(s) ≤ t, for still that it cannot be CG that end(s) ≥ t. If s is given, the state s for already cannot be given as having started at t. And the state of s cannot be given as persevering up till t for still. So s is a potentially future state for already and a potentially past state for still. This is weaker, but good enough for the intuition. When there is a strong or weak CG antecedent, still and already mark a correction of the CG or of a suggestion in the CG. Now the common ground will at least suggest that the anteriority/posteriority presuppositions are true. The treatment in terms of anteriority/posteriority presuppositions comes however with a serious problem. The problem is that when s holds at a point of time t, it also has to hold in an open environment of t, i.e. at times before t and at times after t. This is the standard test for stative sentences and can be found in e.g. Moens & Steedman (there is also an inchoative reading of these sentences, but that is not the relevant one). Anteriority (and posteriority) therefore always follow when t is a point and if the facts licence an utterance of (47). (47) s at t John slept at three am.

Therefore in the same circumstances both of (48) are also true and licensed. (48) a. still s at t John still slept at three am. b. already s at t John already slept at three am.

But intuitively this is not the case at all. And the mirative theory explains this neatly and directly. Still is licensed by an expectation that s has terminated and already by an expectation that s is not going on yet. Without such expectations, there is no point to the use of the particles and their use is infelicitous. Using superweak rather than weak presuppositions undermines this point to some extent. The use of superweak is motivated by the emerging use of both already and still as aspect markers. If it is necessary to rule out the inchoative use of At t s, the use of still has a point, if it is necessary to mark perfectivity (of the state of not yet s) already has a point. In such cases, expectations can be absent. This would be normal grammaticalisation. See Fong 2003 for a similar but different treatment. (49) gives some relevant examples, the second from spoken British English.

316 (49) a. b. c. d.

henk zeevat If he was home at three, he has left early. If he was still home at three, he has left early. You eat? (Are you eating?) You eat already? (Have you eaten?)

The mirative theory also predicts that t cannot be new and indeterminate, as observed by Greenberg (2009). In that case, there could not be an expectation about s terminating before or starting after t. (50) gives two of Greenberg’s examples. In each case, the past tense in the bad example is not tied to any given moment of time. Given that the moment is unknown, it is hard to establish shared surprise that at that moment s is still ongoing or to jointly entertain the presupposition that s should have terminated before this unlocated t. (50) a. A: How’s John? B: Well, he is still unemployed (but we hope he will have a job soon). B: *Well, he was still unemployed (but now he has a job). B: Well, he was still unemployed last month (but now he has a job). b. A: There is an important meeting tonight. B: I won’t be able to be there. I am still ill, and I must rest. B: *I won’t be able to be there. I was still ill, and I must rest. B: I won’t be able to be there. I was still ill this morning.

6. The Scope of Mirativity for Particles It does not seem that an alternative treatment of mirativity will work for the phenomena discussed in this paper. If that is so, the formalisation proposed can be used as a definition of mirativity and it then turns out to be quite a large phenomenon in the world of particles comprising correction, concessivity and adversativity. Correction falls apart in several cases in (51). (51) a. p : ¬p b. P(x) : ¬Px, Py c. ¬p : p

Explicit concession would be (52) (52) though p, q weak(□(p → ¬q)) : q

And pure adversativity is (53). (53) nevertheless q weak(¬q) : q

expressing surprise by particles

317

If mirativity can be defined as correction of expectations, all of these will be mirative: but, though, however, nevertheless and even not. At first sight, this seems to conflict with intuition. But consider the cases where the mirative analysis of the particles considered in this paper seems most plausible. These are typically main clause occurrences in which the particle is intonationally prominent and where the main point seems to be the expression of surprise. Now the normal use of adversative particles seems to be stressless and sentence initial as conjunctions. But there are other cases, as in (54), where the expression of surprise seems to be at least part of the point. (54) a. b. c. d.

John came NEVERTHELESS. John did NOT bring the book. No, John IS coming, THOUGH Mary told him not to. Mary is not coming tonight, BUT she will come later this week.

The mirative analysis of the four particles turned out to make them similar to adversative markers and that suffices for the conceptual points made above. The fact that adversative markers are normally just playing a more subordinate role (plausibly that of helping to keep information storage consistent) does not mean that they do not have the potential of expressing surprise too. 7. Conclusions The mirative analysis seems to bring progress in each of the four cases. In the case of even, it brings out the close relationship of even to the additive particles, which seems to be more difficult in the scalar analyses. In the case of only, it essentially adds to the existing analysis of exhaustivity with a presupposed host, by explaining the difference with intonational and contextual means of marking exhaustivity and by explaining the special characteristics of the presupposition: it is not cancellable as other presuppositions, because it is also weakly presupposed. In the case of already and still, there is a direct explanation of the observations of Greenberg 2009 and a solution to the triviality problem. It turns out to be possible in all four cases to maintain the line that the particle means one thing and the host another (the host influences the meaning of the particle by supplying antecedents, the particle may disambiguate the host) and that the two meanings are just merged, as if they were separate utterances. This should be the ideal for particle semantics and should only be given up when the particle is a proper operator like not or maybe.

318

henk zeevat

There is an element of reduction to emotion in the cases considered. In even, the scalar aspect partly follows from the surprise, in only, the exhaustive interpretation follows from the need to be able to express quantity surprise, the weak presuppositions of already and still expressing surprise capture the intuition behind the proposed anteriority and posteriority presuppositions. Both the interaction of the host semantics with the particle semantics and the possibility of reduction to emotion support an updated version of update semantics where the information state (the common ground) does not just record facts, but also obligations and emotions and where the communicators try to establish joint emotional attitudes, obligations and facts at the same time. This would give a natural way of modelling these interactions and reductions. Consider (55). (55) Those damn farmers are blocking the road.

The epithet damn is not just an expression of negative evaluation of a given group of farmers, it is also an invitation to the hearer to join in with this negative evaluation. And the host sentence also gives an argument why moral indignation should be directed at these farmers: they are obstructing the traffic. There is also further specification of the content, namely the inference that they are doing this on purpose and not as a side effect of their normal agricultural activities. The analyses of particles in this paper therefore illustrate the general point that language interpretation is rather holistic and that our formal models of interpretation should reflect this, as in the updated update semantics. Acknowledgements This paper comes out of work on only which was reported in Zeevat 2009 (recapitulated here) and which started by reading parts of an old draft of Beaver & Clark 2008 and discussions with David Beaver about that draft. I am also grateful for the questions of audiences in Osnabrück and Amsterdam. And to the anonymous reviewers of this paper for many useful points.

expressing surprise by particles

319

References Beaver, David I. & Brady Clark (2008): Sense and sensitivity. How focus determines meaning. Oxford: Blackwell. DeLancey, Scott (1997): “Mirativity: The grammatical marking of unexpected information”. Linguistic Typology 1, 33–52. DOI: dx.doi.org/10.1515/lity.1997.1.1.33. Fong, Vivienne (2003): “Unmarked ‘already’: Aspectual expressions in two varieties of English”. Ms. Stanford University. Gazdar, Gerald (1979): Pragmatics. Implicature, presupposition and logical form. New York: Academic Press. Greenberg, Yael (2009): “Presupposition accommodation and informativity considerations with aspectual still”. Journal of Semantics 26, 49–86. DOI: dx.doi.org/ 10.1093/jos/ffn009. Groenendijk, Jeroen & Martin Stokhof (1984): Studies in the semantics of questions and the pragmatics of answers. PhD thesis. University of Amsterdam. Heim, Irene (1983): “On the projection problem for presuppositions”. Proceedings of WCCFL 2, 114–126. Horn, Laurence R. (1969): “A presuppositional analysis of only and even”. Proceedings of CLS 5, 98–107. Ippolito, Michaela (2007): “On the meaning of some focus sensitive particles”. Natural Language Semantics 15, 1–34. DOI: dx.doi.org/10.1007/s11050-007-9004-0. Kaplan, David (1999): “The meaning of ouch and oops. Explorations in the theory of meaning as use”. Ms. University of California: Los Angeles. Kasper, Walter (1992): “Presupposition, composition and simple subjunctives”. Journal of Semantics 9, 307–331. DOI: dx.doi.org/10.1093/jos/9.4.307. Kratzer, Angelika (1979): “Conditional necessity and possibility”. In: Bäuerle, Rainer & Urs Egli & Arnim von Stechow, eds.: Semantics from different points of view. Berlin: Springer, 117–147. Krifka, Manfred (2000): “Alternatives for aspectual particles. Semantics of still and already”. In: Proceedings of BLS 26, 401–412. Löbner, Sebastian (1989): “German schon – erst – noch: An integrated analysis”. Linguistics and Philosophy 12, 167–212. DOI: dx.doi.org/10.1007/BF00627659. Malchukov, Andrej (2003): “Towards a semantic typology of adversative and contrast marking”. Journal of Semantics 21, 177–198. DOI: dx.doi.org/10.1093/jos/ 21.2.177. Moens, Marc & Mark Steedman (1988): “Temporal ontology and temporal reference”. Computational Linguistics 14, 15–29. Potts, Christopher (2005): The logic of conventional implicatures. Oxford: Oxford University Press. ——— (2007c): “The expressive dimension”. Theoretical Linguistics 33, 165–197. DOI: dx.doi.org/10.1515/TL.2007.011. Rooth, Mats (1992): “A theory of focus interpretation”. Natural Language Semantics 1, 75–116. DOI: dx.doi.org/10.1007/BF02342617. Sæbø, Kjell Johan (1985): “Notwendige Bedingungen im Deutschen zur Semantik modalisierter Sätze”. Arbeitspapiere des SFB 99 der Universität Konstanz 108. van der Sandt, Rob (1992): “Presupposition projection as anaphora resolution”. Journal of Semantics 9, 333–377. DOI: dx.doi.org/10.1093/jos/9.4.333.

320

henk zeevat

Soames, Scott (1989): “Presupposition”. In: Gabbay, Dov & Franz Guenther, eds.: Handbook of philosophical logic. Vol. IV. Dordrecht: Reidel, 553–616. Stalnaker, Robert (1978): “Assertion”. In: Cole, Peter, ed.: Pragmatics. New York: Academic Press, 315–332. Umbach, Carla (2005): “Contrast and information structure. A focus-based analysis of but”. Linguistics 43, 207–232. DOI: dx.doi.org/10.1515/ling.2005.43.1.207. Zeevat, Henk (2009): “Only as a mirative particle”. Sprache und Datenverarbeitung 33, 179–196. Zeevat, Henk & Katja Jasinskaja (2007): “And as an additive particle”. In: Aurnague, Mixel & Kepa Korta & Jesús Maria Larrazabal, eds.: Language, representation and reasoning. Memorial volume to Isabel Gómez Txurruka. San Sebastian: University of the Basque Country Press, 315–340.

INDEX appositives, 21, 22, 78, 82, 87, 111, 164, 165, 168, modal particles, 11, 12, 26, 27, 34–36, 95–102, 211, 212 104–121, 125–130, 143, 146, 147, 168, 179, 187, assertion, 45, 64, 81, 82, 152, 163–165, 235, 236, 203, 208, 231–235, 241, 245–249, 255, 256, 239–242 258, 260 secondary, 164 doch, 95–101, 105–107, 109, 110, 113–120, at-issue content/meaning, 82, 89, 163–165, 125, 135, 152, 179, 232–234, 241, 246–255 168, 170, 171, 177, 208, 211–213, 215–219, halt, 179, 248, 304 224 ja, 11, 12, 22, 27, 34–36, 95–101, 105–110, 113, 116–120, 179, 187, 232, 241, 245–250, binding, 173, 188, 189 253–255, 257, 258, 260 wohl, 11, 95, 96, 98, 100, 101, 105–108, common ground (CG), 95, 98–101, 119–121, 111–114, 116–120, 168, 179, 187, 239, 256, 126, 129–136, 138, 139, 141, 143, 144, 231, 232, 260 234–247, 251, 253–255, 258–260, 262, 263, modals, 37, 38, 75, 76, 204, 211, 212, 231, 232, 265, 268–270, 298–302, 304, 310, 311, 315, 235, 249, 257, 258, 260, 261, 263, 264, 318 269 context shift, 39–42, 95, 97, 98, 101–121 morphology, use-conditional, 24–26 contextual judge, 41, 42, 49, 76, 78, 103, 104, 106, 107, 110, 112, 114, 117 negation, 9, 34–37, 141, 157, 178–180, 204, 211, 212, 231–241, 243, 244, 247, 249, 255, 256, datives, 14, 15, 171–188 259–263, 266, 268, 269, 276–278, 285, 297, denial, 35, 36, 45, 160, 211–214, 219, 233, 300–303, 309 235–239, 240–245, 255, 257, 258, 261–264, illocutionary, 231, 232, 236–238 266, 268–270 metalinguistic, 157, 179, 237 descriptive ineffability, 42–44, 46, 49, 78, 84, non-restrictive relative clauses, 21, 82, 164, 91, 161, 177, 178, 185, 189 167, 170 nondisplaceability, 16, 37–40, 48, 49, 76, 77, epithets, 4–6, 11, 12, 82, 97, 158, 164–168, 170, 91, 176 177, 179, 190, 318 evidentials, 104, 115–118, 120, 167, 184, 187, performatives, 9, 25, 44, 45, 49, 61, 63, 76, 189 78–87, 89–92, 164, 165 perspective dependence, 40–42, 49, 76, 78, immediacy, 9, 44–46, 49, 68, 80, 84 84, 91 implicatures presuppositions, 17, 37, 51, 97, 132, 134, 151, conventional, 97, 151–171, 177–179, 184, 156, 161, 190, 235, 244–246, 280, 297, 185, 189–191, 201–202, 212, 214–218, 301, 303–305, 307, 309–311, 313–316, 224, 227, 297 318 conversational, 131, 156, 159, 167, 246, 253, Principle B, 173, 188 254, 258, 260, 306 projection behavior, 189, 211, 299, 309–311 indexicals, 95, 97, 101–104, 117, 120 indirect speech, 37–40, 41, 97, 104–108, 117, question under discussion, 23, 264, 265 162 questions, 9, 13, 17, 18, 36, 45, 85, 101, 103, 104, interjections, 6, 7, 10, 22, 27, 29, 31, 39, 43, 47, 112–117, 128, 133–135, 137, 141, 143, 205, 209, 48, 79 214, 236, 238–241, 243, 247, 248, 264–266, interrogatives, 20, 36, 37, 45, 87, 89, 100, 114, 268, 270, 275–294, 301–304 135, 144 rhetorical, 114, 137, 138, 141–144

322 repeatability, 46, 47, 49, 81, 91 slurs, 7–10, 27, 30, 48, 166–168, 177 topic, 23, 24, 75, 154, 155, 202, 212, 220, 223, 226, 309, 313 topicalization, 23, 24

index use conditions, 10, 15, 17, 21, 23, 24, 27, 45, 48, 51, 79, 82, 85, 96, 128, 132–140, 143–145, 147, 188, 243, 247, 249–251, 253, 255, 261–263, 265, 269, 270 verum, 19–21, 231, 232, 239, 242, 261–270