Understanding Quotation 9783110240085, 9783110240054

Citation preview

Understanding Quotation

Mouton Series in Pragmatics 7

Editor Istvan Kecskes

Editorial Board Reinhard Blutner Universiteit van Amsterdam The Netherlands N. J. Enfield Max-Planck-Institute for Psycholinguistics Nijmegen The Netherlands

Ferenc Kiefer Hungarian Academy of Sciences Budapest Hungary Lluı´s Payrato´ University of Barcelona Spain

Raymond W. Gibbs University of California Santa Cruz USA

Franc¸ois Recanati Institut Jean-Nicod Paris France

Laurence R. Horn Yale University USA

John Searle University of California Berkeley USA

Boaz Keysar University of Chicago USA

De Gruyter Mouton

Deirdre Wilson University College London Great Britain

Understanding Quotation

edited by

Elke Brendel Jörg Meibauer Markus Steinbach

De Gruyter Mouton

ISBN 978-3-11-024005-4 e-ISBN 978-3-11-024008-5 ISSN 1864-6409 Library of Congress Cataloging-in-Publication Data Understanding quotation / ed. by Elke Brendel, Jörg Meibauer, Markus Steinbach. p. cm. ⫺ (Mouton series in pragmatics; 7) Includes bibliographical references and index. ISBN 978-3-11-024005-4 (alk. paper) 1. Quotation. 2. Semantics. 3. Pragmatics. I. Brendel, Elke. II. Meibauer, Jörg. III. Steinbach, Markus. PN171. Q6U53 2011 4011.43⫺dc22 2011009277

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. ” 2011 Walter de Gruyter GmbH & Co. KG, Berlin/New York Typesetting: PTP-Berlin Protago TEX-Production GmbH, Berlin Printing: Hubert & Co. GmbH & Co. KG, Göttingen ⬁ Printed on acid-free paper Printed in Germany www.degruyter.com

Contents Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contributors to this volume . . . . . . . . . . . . . . . . . . . . . . . .

vii ix

Exploring the Meaning of Quotation . . . . . . . . . . . . . . . . . . . Elke Brendel, Jörg Meibauer, and Markus Steinbach

1

Attitudes Toward Quotation . . . . . . . . . . . . . . . . . . . . . . . Barbara Abbott

35

Truth Value Talk Without Quotation . . . . . . . . . . . . . . . . . . . Manuel Bremer

47

Disquotational Indirect Reports in Focus . . . . . . . . . . . . . . . . . Sarah-Jane Conrad

59

Behabitive Reports . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cathrine Fabricius-Hansen and Kjell Johan Sæbø

85

Double-duty Quotation, Conventional Implicatures and What Is Said . . 107 Manuel García-Carpintero What Quotations Refer To . . . . . . . . . . . . . . . . . . . . . . . . 139 Mario Gómez-Torrente Quotation Marks and Kinds of Meaning. Arguments in Favor of a Pragmatic Account . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 Daniel Gutzmann and Erik Stei Quotation and Pictoriality . . . . . . . . . . . . . . . . . . . . . . . . . 195 Manfred Harth The Punctuation Theory of Quotation . . . . . . . . . . . . . . . . . . 209 Michael Johnson Misrepresenting Misrepresentation . . . . . . . . . . . . . . . . . . . . 231 Michael Johnson and Ernie Lepore Two Dogmas on Quotation . . . . . . . . . . . . . . . . . . . . . . . . 249 Jürgen Pafel Reporting and Quoting in Signed Discourse . . . . . . . . . . . . . . . 277 Josep Quer

vi

Contents

The Act of Quotation . . . . . . . . . . . . . . . . . . . . . . . . . . . 303 Paul Saka Axioms of Reference and Rules of Quotation . . . . . . . . . . . . . . 323 Savas L. Tsohatzidis Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331

Preface This volume is a collection of original papers on topical issues concerning various linguistic and philosophical aspects of quotations. Its aim is to advance the current debate among linguists and philosophers of language on the nature and function of quotation as a device of meta-linguistic discourse. Eight of the fourteen articles are versions of presentations given at the “International Conference on Quotation and Meaning” (ICQM), held in October 2007 at the Johannes Gutenberg-University Mainz, Germany. We gratefully acknowledge a Research Grant from the University of Mainz, as well as financial support from the Interdisciplinary Research Centre of Neurosciences (IFZN) at the University of Mainz, that enabled us to organize this conference. We would like to thank the authors for their contributions to this volume and for their cooperativeness during the reviewing and publication process. We are grateful to Sina Schade and Barbara Dienst in helping us with the preparation of the typescript and to Francey Russell for the English proofreading of our introduction to this volume. We would also like to thank the editors from Mouton De Gruyter, in particular Istvan Kecskes, for their patience and their support. Elke Brendel, Jörg Meibauer, and Markus Steinbach

Contributors to this volume Barbara Abbott Professor Emerita of Linguistics & Philosophy Michigan State University email: [email protected] Manuel Bremer Professor of Philosophy Universität Düsseldorf email: [email protected] Elke Brendel Professor of Philosophy Universität Bonn email: [email protected] Sarah-Jane Conrad Postdoctoral Research Assistant University of Berne email: [email protected] Cathrine Fabricius Hansen Professor of Linguistics ILOS, University of Oslo email: [email protected] Manuel García-Carpintero Professor of Philosophy Universitat de Barcelona email: [email protected]

x

Contributors to this volume

Mario Gomez-Torrente Research Professor Universidad Nacional Autonoma de Mexico email: [email protected] Daniel Gutzmann Research Assistant Goethe-Universität Frankfurt email: [email protected] Manfred Harth Postdoctoral Research Assistant Ludwig-Maximilians-Universität München email: [email protected] Michael Johnson PhD candidate The State University of New Jersey email: [email protected] Ernie Lepore Director of Cognitive Science Rutgers University email: [email protected] Jörg Meibauer Professor of German Linguistics Gutenberg-Universität Mainz email: [email protected] Jürgen Pafel Professor of German Linguistics Universität Stuttgart email: [email protected]

Contributors to this volume Josep Quer Research Professor of Sign Language Linguistics ICREA & Universitat Pompeu Fabra email: [email protected] Kjell Johan Sæbø Professor of Theoretical Linguistics ILOS, University of Oslo email: [email protected] Paul Saka Professor of Philosophy University of Texas email: [email protected] Erik Stei Research Assistant Universität Bonn email: [email protected] Markus Steinbach Professor of German Linguistics Georg-August-Universität Göttingen email: [email protected] Savas L. Tsohatzidis Professor of General Linguistics and the Philosophy of Language Aristotle University of Thessaloniki e-mail: [email protected]

xi

Exploring the Meaning of Quotation Elke Brendel, Jörg Meibauer, and Markus Steinbach

1.

Introduction

While quotation – a specific form of metarepresentation – is a complex and multifaceted phenomenon, this volume is mainly concerned with its philosophical and linguistic aspects. There is a long-standing tradition in philosophy of language focusing on metalinguistic (pure) quotation, such as (1a) below, when analyzing quotation. There is also the linguistic tradition of describing and explaining reported speech, be it direct (1b) or indirect (1c), as displayed by many different languages. (1)

a. b. c.

“Berlin” has six letters. Quine said: “Quotation has a certain anomalous feature”. Quine said that quotation has a certain anomalous feature.

Bringing together these two traditions, one of the aims of the book, is of crucial importance for the development of a comprehensive theory of quotation. Philosophers of language and linguists share a common interest, namely the exploration of meaning. Most researchers would agree that one cannot know a priori, whether a certain meaning-effect belongs to the realm of semantics, pragmatics or both. As will become clear in this volume, quotations have to do with both domains, but in a complex and intriguing way. The study of semantics and pragmatics in the philosophy of language and in linguistics will certainly profit from an in-depth exploration of the meaning of quotation. Likewise, analyses of quotations will benefit from new linguistic and philosophical approaches to the interaction of semantics and pragmatics. Starting with terminological and empirical issues, we deal in section 2 with nine puzzles about quotation, thereby addressing essential issues of quotation both in linguistics and in the philosophy of language. The discussion includes an overview of different kinds of quotation and quotation markers, as well as various theoretical approaches to different aspects of quotation. Section 3 gives a brief summary of the papers in this volume, which deal with various issues

2

Elke Brendel, Jörg Meibauer, and Markus Steinbach

addressed in the nine puzzles. The papers investigate quotations from a number of perspectives thus contributing to the complex picture of this fascinating phenomenon. 2.

Nine puzzles about quotation

In this introduction, we will examine the following nine puzzles about quotation in more detail. It will become clear that an integrative theory of quotation is a cornerstone of every theory about human cognitive metarepresentation. 1. Which kinds of quotation can be found across languages? 2. How can the different kinds of quotations be classified? 3. Which kinds of linguistic and paralinguistic markers can be found in quotations across languages? 4. How is the original utterance reconstructed? 5. Do quotation marks affect the grammaticality of sentences – and if so, in what way? 6. What impact do quotation theories have for the philosophy of language and logic? 7. How can an analysis of quotation help clarify the debate about the semantics/pragmatics interface? 8. What role does context play in determining the meaning of quotation? 9. How are the different kinds of quotation acquired and processed? 2.1. Puzzle 1: Which kinds of quotation can be found across languages? In the philosophical and linguistic literature, the prototypical and most discussed kind of quotation seems to be the direct quotation illustrated in (2). (2)

Lena said yesterday in the museum of modern art: “This piece of art is difficult to understand.”

In (2), the use of the quotation marks signals that the utterance “This piece of art is difficult to understand.” has been originally produced in a different context (i.e., yesterday in the museum of modern art) by a different speaker (i.e., Lena). Direct quotations are used to convey an utterance verbatim (with some flexibility, see section 2.4 below for more discussion). Since direct quotations are interpreted in the context of the original utterance, indexicals cannot be bound by the actual context. Moreover, direct quotations are referentially opaque and thus block substitution of extensionally equivalents salva veritate

Exploring the Meaning of Quotation

3

(Oedipus said: “I love Iokaste.” =/ Oedipus said: “I love my mother.”, for a more detailed discussion, see section 2.5 below). Indirect quotations (or indirect or reported speech/discourse) such as (3) are an alternative means to report an utterance that has been produced in a different context. Many languages have developed specific grammatical means to distinguish between direct and indirect quotations. English, for example, uses complementizers in sentence-initial position and sequence of tenses (see also section 2.3 below). (3)

Lena said that this piece of art is difficult to understand.

In indirect quotations, indexical expressions are usually interpreted in the actual context of utterance. Indirect quotations are not only used to report utterances but also express thoughts or beliefs (4a), something not always possible with direct quotations (4b). In this respect, indirect quotations are more flexible than direct quotations; however, indirect quotations also have some restrictions. Unlike direct quotations, indirect quotations cannot be used to report specific kinds of utterances such as “Hello” in (5a), they cannot be embedded under certain matrix predicates such as “cough” in (5b), and inherently speaker-related expressions such as “my damned car” in (5c) cannot be converted into indirect quotations (see d’Avis 2007 and section 2.4). In this respect, direct quotations are more flexible than indirect quotations. (4)

a. b.

Der Killer vermutete, dass er ihn kriege. #Der Killer vermutete: „Ich kriege dich!“ ‘The Killer supposed that he would get him.’

(5)

a. b. c.

Adam said, “Hello!” / *Adam said that hello! The killer coughed, “I will get you.” / *that he will get him. Adam said, “My damned car has a flat tire”. / #that his damned car has a flat tire.

Some researchers would object that indirect speech doesn’t qualify for the label quotation, because here, no mention of a linguistic representation is made (see de Brabanter 2010: 377). But not only is it important to come to grips with common and different properties of direct versus indirect speech, it is also important to explain the smooth transition between direct and indirect quotations (see section 2.3) and to account for a third kind of quotation, the so-called mixed quotation, that results from the combination of direct and indirect quotation. Mixed quotations such as (6) are indirect quotations that contain a directly quoted part in the indirectly reported utterance (usually a word or a constituent). Like direct quotations, mixed quotations are used to convey at least some part

4

Elke Brendel, Jörg Meibauer, and Markus Steinbach

of an utterance verbatim (again, with some flexibility). As opposed to direct quotations, the quoted expression is a proper part of the sentence. The quoted expression, “is difficult to understand,” which receives a standard interpretation in (6), is thus both used and mentioned. (6)

Lena said that this piece of art “is difficult to understand.”

For morphosyntactic reasons, deviations from the original utterance are possible, as is illustrated in the German example in (7) where the verb-final structure of the indirect quotation forces the finite verb soll, which is included in the direct quotation, to move from the second to the sentence-final position. In addition, mixed quotations permit bound indexicals like us in (8) (see Cappelen/Lepore 1997 and Cumming 2005). (7)

Original utterance: „Er soll nach Hause gehen.“ Sie sagte gestern zu ihm, dass er „nach Hause gehen soll“. ‘Yesterday, she said to him that he should go home.’

(8)

Their accord on this issue, he said, has proved “quite a surprise to both of us” / quite a surprise to both of them.

The philosophically most important kind of quotation is pure quotation as exemplified in (9). The analysis of pure (and direct) quotations has been quite important for philosophy of language, logic, formal semantics, and pragmatics (see section 2.6 below). Pure quotations share a particular and crucial property with direct quotations: in both kinds of quotation, the quoted expression is not used but only mentioned. A standard assumption of most theories of quotation is that pure quotations (just like direct quotations) are semantically singular terms and syntactically noun phrases (but see Pafel, this volume, for counterarguments against this standard view). (9)

“This piece of art is difficult to understand” is a sentence.

Pure quotations seem to display semantic inertia. The quotation does not denote (or refer to) to a referent in the world but to a linguistic property of the quoted expression itself (see section 2.6 below for more discussion). While the expression “Boston” in (10a) refers to the city Boston (i.e., the expression is used), the pure quotation ““Boston”” in (10b) does not refer to the city Boston but to a phonological property of the linguistic expression “Boston” (i.e., it is mentioned). (10)

a. b.

Boston is the capitol of Massachusetts. “Boston” is disyllabic.

Exploring the Meaning of Quotation

5

The examples (9)–(11) illustrate that any of the linguistic (phonological, graphematic, morphosyntactic, semantic, etc.) properties of the quoted expression can be referred to or exemplified by a pure quotation. Moreover, even nonlinguistic entities such as meaningless strings of letters in (11g) or symbols of formal logic (11h) can be quoted. Pure quotations are potentially ambiguous and their disambiguation depends on contextual pragmatic enrichment. (11)

a. b. c. d. e. f. g. h.

“Quotation” has 9 letters. “To quote” is a verb in infinitive form. “Quotation has a certain anomalous feature” is a sentence. “QUOTATION” is written in capital letters. “Es schneit” in German means “It snows” in English. “Living Dead” is an oxymoron. “Qxlourshpulparc” is a meaningless string of letters. “¬” is a negation symbol.

Traditional philosophical and linguistic analyses of quotations mainly focus on pure and direct quotations on the one hand, and on indirect quotations on the other. Recent discussions of quotations have directed the attention to socalled scare quotations like (12) (see Klockow 1980, Predelli 2003, and Gutzmann/Stei, this volume). Scare quotations equal mixed quotations insofar as the quoted expression is syntactically and semantically a proper part of the sentence, i.e., the quoted expression is not only mentioned but also used. Yet in contrast to mixed quotations, scare quotations do not necessarily refer to an original utterance as can be seen in the examples (13b–d) below. (12)

Lena, the philistine, said in disgust that “this piece of art” is difficult to understand.

In scare quotes, the quoted expression gets two interpretations. In addition to the standard interpretation, it receives an (contextually driven) interpretation, which again requires pragmatic enrichment. In this respect, scare quotations are similar to pure quotations. Such an interpretation could be, for example, (i) a certain distancing from someone else’s wording as in (13a), (ii) disagreement with the quoted expression as in (12) and (13b), (iii) a sarcastic/ironic/euphemistic connotation as in (13c), or (iv) the indication of a shift in language/idiolect as in (13d). (13)

a. b.

Lena, the philistine, said in disgust that “this piece of shit” is difficult to understand. His new “theory” is difficult to understand.

6

Elke Brendel, Jörg Meibauer, and Markus Steinbach

c. d.

Before becoming a movie star, he was an actor in some “adult films”. In the US people use “cell phones” when calling on mobile phones.

The last kind of quotation has only been described recently: Emphatic quotations (or greengrocer’s quotations) like (14) are used to emphasize an expression. Although some (prescriptive) speakers consider emphatic quotations to be incorrect, they are frequently found in situations that do not provide other means of emphasis. In emphatic quotations, like in mixed and scare quotations, the quoted expression is a proper part of the sentence and like in pure and scare quotations, the quotation marks trigger pragmatic enrichment. Emphatic quotes yield the specific pragmatic interpretation of emphasis. (14)

This piece of art is “ingenious”. (Written by an admirer of the artist’s work)

With the exception of indirect quotation, in all quotational uses we have so far discussed, quotation marks play a significant role, and one could assume that quotation has to do with the use of quotation marks. This, however, is not correct, firstly, because quotation marks are used in scare quoting and in noncitational, emphatic uses, and secondly, there seem to exist quotational uses without any quotation marks. Therefore, in order to be able to develop a comprehensive theory of quotation, we suggest a very broad notion of quotation, one that comprises all the aforementioned types (and possibly more), without laying too much emphasis on either the mention/use issue or the use of quotation marks (which are absent in spoken language anyway). What is truly important and unique, is the genuine metarepresentational property of these types. Let us finally mention two further kinds of quotations without quotation marks, that are frequently found in narratives: in free indirect and free direct quotations (or experienced speech), the change of perspective is neither indicated by quotation marks nor by grammatical, lexical, or other graphematic means, as can be seen in the free direct quotation in (15), which contains an abrupt context shift after the third sentence (see Coulmas 1986a and FabriciusHansen 2002). The reader must infer that the directly quoted sentences “You can’t measure them; all you can see is behaviour. So why forget about minds and just study behaviour instead?” were (or could have been) uttered by most psychologists these days. (15)

The reason I’m interested in language is because I’m interested in mind. That is an unfashionable position. Most psychologists these days will

Exploring the Meaning of Quotation

7

tell you that minds can’t be studied scientifically. You can’t measure them; all you can see is behaviour. So why forget about minds and just study behaviour instead? In addition to the lack of any indicators of quotation, a mismatch between grammatical form (third person) and point of view, which can lead to a fusion of points of view (narrator and hero), is typical for free indirect quotations as is illustrated in (16ab). (16)

a. b.

Had she not always thought that this piece of art is difficult to understand? Max was feeling remorseful. He shouldn’t have spoken to them so harshly. He would have to apologize to them next time he saw them.

2.2. Puzzle 2: How can the different kinds of quotations be classified? We can organize the different kinds of quotation discussed in the previous section into two basic classifications: (i) quotations that refer to an original utterance and those that do not refer to an original utterance; (ii) quotations with quotation marks and those without quotation marks. Quotations with quotation marks can be further subdivided into quotations where the quoted expression is only mentioned and quotations where it is not only mentioned but also used. The correlation between the reference to an original utterance, the use of quotation marks, and the use/mention-distinction is summarized in table 1 below. As can be seen in this table, reference to an original utterance does not depend on the use of quotation marks, nor do quotation marks necessarily indicate the reference to an original utterance. Likewise, the use/mention-distinction does not correlate with the reference to an original utterance and vice versa. Note that direct and pure quotations can be subsumed under Recanati’s (2001) closed quotations and mixed, scare and emphatic quotations under his open quotations. Table 1. Quotation marks and reported speech refers to original utterance quotation marks

no quotation marks

mentioned used and mentioned

does not refer to original utterance

direct quotation

pure quotation scare quotation mixed quotation emphatic quotation indirect quotation

free (in)direct quotation

8

Elke Brendel, Jörg Meibauer, and Markus Steinbach

Figure 1. Kinds of quotation marks

The correlation between the different kinds of quotation is systematically illustrated in figure 1 above. We already mentioned that direct quotations seem to be the prototypical kind of quotation with the following three properties: (i) quotation marks are used as a graphematic device to mark the quoted expression, (ii) the quoted expression refers to an original utterance, and (iii) the quoted expression is only mentioned. Mixed and pure quotations are both very similar to direct quotations. In mixed quotations, the quoted expression also refers to an original utterance and in pure quotations the quoted expression is also only mentioned. However, mixed and pure quotations also differ from direct quotations. In mixed quotations, the quoted expression is not only mentioned but also used (i.e., arrow Œ in figure 1). Pure quotations differ in two respects from direct quotations: (i) The quoted expression does not refer to an original utterance and (ii) pragmatic enrichment is necessary to disambiguate pure quotations (i.e., arrow ). Scare quotations differ even more from direct quotations. They combine two essential properties of mixed and pure quotations: as in mixed quotations, the quoted expression is used and like in pure quotations, pragmatic enrichment is involved in the interpretation of the quoted expression (i.e. arrows Ž and {). Scare quotations thus build on both mixed and pure quotations. Functionally, scare quotations are like pure quotations, since the interpretation of the quotation depends on pragmatic enrichment. Formally, they are very similar to mixed quotations. This can be seen in example (12) above, here repeated as (17a), where the directly quoted expression “this piece of art” is a proper part of the embedded sentence (the indirect quotation). Like mixed quotations, scare quotations can refer to an original utterance. Hence, scare quotations seem to have developed directly from mixed quotations and sentences like (17a) are

Exploring the Meaning of Quotation

9

ambiguous between an interpretation as a mixed or a scare quotation. Every use of quotation marks in indirect quotations may trigger an additional pragmatic effect typical for scare quotations since the use of quotation marks in (17a) produces a marked structure compared to the indirect quotation without quotation marks in (17b). Note again that scare quotations, unlike mixed quotations, do not necessarily refer to an original utterance. (17)

a. b.

Lena, the philistine, said in disgust that “this piece of art” is difficult to understand. Lena, the philistine, said in disgust that this piece of art is difficult to understand.

Emphatic quotations seem to relate to scare quotations (i.e., arrow ). Like in scare quotations, the quoted expression is not only mentioned but also used. Moreover, in both scare and emphatic quotations the quotation marks trigger an additional pragmatic effect. Note finally, that indirect and free (in)direct quotations are directly related to direct quotations since they involve reference to an original utterance (i.e., arrow ‘). Unlike in direct quotations, the quoted expression is not marked by quotation marks. 2.3. Puzzle 3: Which kinds of linguistic and paralinguistic markers can be found in quotations across languages? Numerous linguistic and paralinguistic markers are designed to signal that a content is quoted. In spoken language, prosodic markers like a rising or falling voice, accentuation, and/or pauses may draw attention to the quotation. Gestures, like the air quotes, or mimics, like raising eyebrows, are further means for signaling quotation. In written language, graphematic markers, such as quotation marks of different shape, the relation between the colon and quotation marks, and other typographic devices indicate quotations. While it is usually denied that there is an exact correspondence between phonology and punctuation, there nevertheless may exist analogies that should be investigated (see Steinhauer 2002). When we focus on quotation marks, which have been intensively discussed in the past, we note that there is a good deal of variation in their distribution: “When an expression is enclosed within quotation marks inside a larger matrix sentence we need to consider the distribution of punctuation marks within the quotation itself and in the matrix sentence. This is a matter on which there is a good deal of variation, firstly between AmE and BrE, and secondly, within BrE (and other non-American varieties), between different publishing houses.” (Nunberg et al. 2002: 1754)

10

Elke Brendel, Jörg Meibauer, and Markus Steinbach

Most languages display a set of grammatical markers of quotation. Furthermore, many languages have grammatical means to distinguish between different kinds of direct and indirect quotations. A complete survey of such grammatical markers is still lacking, but there are numerous attempts at pointing out the diversity of grammatical markers indicating quotation in the languages of the world (see Coulmas 1986ab, Li 1986, Güldemann/Roncador 2002, and Bucali´c 2007). For instance, German uses complementizers, verb-position, and subjunctive forms in indirect quotations. (18)

a.

b. c. d.

Ken sagte, dass er das nicht machen könne. Ken said that he that not make can. SUBJ ‘Ken said that he couldn’t do this.’ Ken sagte, dass er das nicht machen kann. Ken sagte, er könne das nicht machen. Ken sagte, er kann das nicht machen.

In (18ab), we find the complementizer “dass” and, accordingly, the finite verb in last position. In comparison, (18cd) lack a complementizer and the finite verb is in second position. Both types of embedding are grammatical, but it is hard to tell what exactly is the functional difference. The examples show, for instance, different degrees of integration: whereas (18a) is the structurally most integrated sentence, (18d) is the least integrated one (see Reis 1997, Truckenbrodt 2006, and Steinbach 2007). While in (18a) and (18c), quotation is marked by the subjunctive (i.e. “könne”), we find the indicative (i.e. “kann”) in (18b) and (18d). By way of prescription, the subjunctive is required for indirect quotation, but in everyday conversation the indicative is mostly preferred (see Schecker 2002, Fabricius-Hansen 2002 and Zifonun/Hoffmann/Strecker et al. 1997). Moreover, there is variation between the indicative, the subjunctive of the present and the preterit tense (19a), and the “würde”-construction as in (19b). (19)

a. b.

Ken sagte, dass er müde ist/sei/wäre. Ken said that he tired is/be SUBJ.PRES./be SUBJ.PRET. Ken sagte, dass er spielte/spielen würde. Ken said that he played IND./SUBJ. PRET./play INF.+ würde

In addition, there is no clear-cut dividing line between the different forms of direct and indirect quotation. Embedded roots without pronouns, indexicals and subjunctive (i.e. the indirect quotation in (18d) without the pronoun “er”) are in German structurally identical to directly quoted declarative sentences. This

Exploring the Meaning of Quotation

11

is illustrated in (20ab). The same holds true for the corresponding parenthetical constructions in (21ab). (20)

a.

b.

(21)

a. b.

Ken sagte, die Polizei kann das nicht machen. Ken said the police can that not make ‘Ken said that the police couldn’t do this.’ Ken sagte: „Die Polizei kann das nicht machen.“ Ken said the police can that not make ‘Ken said: “The police can’t do this”’ Die Polizei, sagte Ken, kann das nicht machen. The police said Ken can that not make „Die Polizei“, sagte Ken, „kann das nicht machen.“ The police said Ken can that not make ‘Ken said that the police couldn’t do this.’

This sketch gives a first impression of the complexity of grammatical mechanisms invoked in quotation. Note that there are many more, for instance systems of the sequence of tenses (see Li 1986 and Fabricius-Hansen 1989) or quotative particles, like the Slovenian “da”. On the lexical side, there are verbs of saying (verba dicendi) that select for certain types of complements. Moreover, there may be lexical markers of quotation, for example constructions like “in quotation marks,” and “I’m like/he’s like” (see Golato 2000), “bla, bla, bla” or “quote … unquote” as in the first line of the Magnetic Field’s song “I Don’t Believe You”: “So you quote love unquote me.” In sum, then, this puzzle cannot easily be solved, for lack of a comprehensive survey of the quotation marking systems in the languages of the world, and for lack of a theory of the general grammatical and lexical architecture of quotation. Let us finally turn to sign languages, which are very interesting not only from a typological point of view. Sign languages have a modality-specific means of quotation at their disposal, which is called role shift or shifted reference (see Lillo-Martin 1995, Quer 2005 and this volume, and Herrmann/Steinbach 2007). Role shift is not simply the sign language equivalent of direct quotation in spoken languages but combines properties of both direct and indirect quotation. The visual-gestural modality of sign languages offers the unique possibility of shifting into the role and adopting the perspective of the quoted person, typically by shifting the body position slightly, breaking eye contact with the addressee and changing the position of the head. These so-called nonmanual features marking role shift occur simultaneously with manual signs and scope over the whole embedded clause, as is illustrated by the examples in (22) from German Sign

12

Elke Brendel, Jörg Meibauer, and Markus Steinbach

Language (DGS). “< >” stands for the non-manuals indicating role shift, “ix” for the pointing sign index, and “y/n” for the non-manuals of yes/no-questions in DGS. The subscripts “3a” and “3b” indicate points in the signing space in front of the signers body that are used to link discourse referents. 3a3b ____________y/n

(22)

a.

peter ix3a mary ix3b ask : ix2 sad ix2 ‘Peter asked Mary whether she is sad’ ‘Peter asked Mary: “Are you sad?” ’

b.

peter ix3a mary ix3b say : tomorrow 1 help2 ‘Peter said to Mary that he will help her tomorrow’ ‘Peter said to Mary: “I will help you tomorrow.”’

3a3b

The shifted context is indicated by the local indices “3a” and “3b.” Hence, both examples in (22) trigger a context shift leading to the interpretation that the embedded sentence has been uttered in a context different from the actual context with Peter (i.e., 3a) being the speaker and Mary (i.e. 3b) being the addressee of the reported utterance. Consequently, indexicals such as ix2 (‘you’) in (22a) and agreement verbs such as 1 help2 in (22b) are interpreted in the shifted context (for a detailed discussion of the interpretation of indexicals in role shift see Quer, this volume). 2.4. Puzzle 4: How is the original utterance reconstructed? In direct, indirect, and mixed quotation, the quotation is related to a prior, authentic utterance. This utterance is called the original utterance. It is obvious that the quotation, in principle, serves to reconstruct this original utterance, but this happens in a way that usually is full of adjustments for different purposes. The question is, then, how and to what extent it is possible to reconstruct the original utterance. It is clear that direct quotation, indirect quotation, and mixed quotation behave differently in this respect. In general, there is trust that direct quotation renders the original utterance in an optimal way, while indirect quotation not only requires some alternations (e.g., the aforementioned subjunctive in German) but is usually considered a way of rendering an original utterance that is more tolerant of divergences from the original. The pure possibility to select from the original utterance, as is the case with mixed quotation, is another means of undermining the strong connection between the original utterance and the quotation. To develop a theory on how these possibilities may be used is still a task for the future.

Exploring the Meaning of Quotation

13

When we speak of an original utterance that is somehow represented in a quotation, we must bear in mind that there are quotations of imaginary utterances or thoughts. Because there is metarepresentation involved, the parlance of an “original” appears not to be superfluous, though of course, it does not make sense to speak of an original “utterance” here. The heaviest attack on the idea of verbatim reproduction of the original utterance stems from Clark/Gerrig (1990). Interestingly, they focus on direct quotation and argue that the assumption of verbatim reproduction (AVR) is problematic. Several aspects are pointed out: • • • • •

First, even the basic notion of verbatim reproduction is unclear. Second, no speaker is able to perfectly memorize an original utterance. Third, repairs show speaker’s uncertainty in this respect. Fourth, the assumption of verbatim reproduction excludes translations. Fifth, it denies the typical selectivity of reproductions, for instance, with respect to translations, change of dialects, sociolects/register, and idiolects, information structure, style, citation of noise, impossible demonstrations (e.g., a speaking finger), free indirect quotations, quoted thoughts, etc. • Sixth, we find integration of nonverbal elements, e.g. rude gestures or noises. AVR, from the point of view of Clark/Gerrig (1990: 800) is nothing but a prejudice that stems from the academic tradition. While these considerations show that the AVR may be taken cum grano salis, the assumption that one has access to the original utterance still grounds every theory of quotation. Even in everyday language, there is a good deal of reflection on the question of what someone said in reality, and to what extent the original utterance is misrepresented. Moreover, the citation practices in the academic world, in the political sphere, in jurisprudence and in the media show that the verbatim assumption or the need to reconstruct the original utterance is important. A critique of Clark/Gerrig (1990) and Wade/Clark (1993) is put forward by Johnson/Lepore in this volume. From the linguist’s point of view, it appears that inherently speaker related aspects of the original utterance cannot be converted into an indirect quotation (see d’Avis 2007 and section 2.1). These aspects include certain types of adverbial sentences, hedged performatives, rhetorical and echo questions, exclamatives and expressives, interjections, onomatopoetic expressions, vocatives, and discourse particles. Every theory of quotation has to explain why this is so.

14

Elke Brendel, Jörg Meibauer, and Markus Steinbach

2.5. Puzzle 5: Do quotation marks affect the grammaticality of sentences – and if so, in what way? The question as to what extent the use of quotation marks affects the grammaticality of a sentence or merely the orthographic or academic norms has an empirical as well as a theoretical dimension (for theoretical consequences, see also the following two sections). In standard theories of quotation, which attribute a fundamental grammatical and semantic (referential) function to quotation marks, quotational sentences without quotation marks like (23b) are ungrammatical and gibberish. Syntactically, the preposition “on,” unlike the quoted expression ““on”” in (23a), cannot be the subject of this sentence and semantically, the preposition “on” cannot be interpreted as the argument of the metalinguistic predicate “is a preposition”. According to such theories, the prevalent practice of quoting without quotation marks should be condemned. By contrast, pragmatic theories that only attribute a certain kind of pragmatic function to quotation marks argue that mentioning without quotation marks is a prevalent linguistic phenomenon that needs to be accounted for by every conclusive theory of quotation (see Klockow 1980, Washington 2002, Saka 1998, Meibauer 2007, and Gutzmann/Stei, this volume). Consequently, sentences like (23b) are perfectly grammatical and meaningful. (23)

a. b.

“On” is a preposition. On is a preposition.

Since quotation marks seem to have an impact on truth-conditions, pragmatic accounts are forced to argue that quotation marks are neither a necessary nor a sufficient grammatical device, but merely a practical pragmatic indicator of disambiguation. According to Gómez-Torrente (2003: 131), for example, “omitting quotation marks is also appropriate, only less informative or more susceptible of undesired interpretation.” Gutzmann/Stei (this volume) follow GómezTorrente and analyze quotation marks as a pragmatic device of triggering conversational implicatures. Likewise, Saka (1998, this volume) argues that quotation marks do not contribute to the grammaticality of a sentence since their only function is to signal mentioning. Although quotation marks are, of course, necessary devices in academic contexts and in formal languages, a strict connection between quotation marks and mentioning seems to be too strong for natural languages (see, for example, Klockow 1980: 53f. and Saka 1998). Quotational expressions without quotation marks can be frequently found in natural languages especially if the sentential context clearly favors the quotational use of expressions such as “cats” in (24a), “if” in (24b), and “Terebi” in (24c).

Exploring the Meaning of Quotation

(24)

a. b. c.

15

Cats is a noun. Scientists study the if. Terebi bedeutet “television” … ‘Terebi means “television”…’

The distribution of quotation marks seems to depend on the register and the kind of text. In addition, quotation marks are specific graphematic devices, which are part of the system of written language, and the constraints on the use of quotation marks are only acquired relatively late at school. Consequently, quotation marks are not a proper part of the core grammar (see also section 2.9).

2.6. Puzzle 6: What impact do quotation theories have for the philosophy of language and logic? Quotation is a fundamental tool of meta-linguistic discourse that has many significant implications for the semantics and pragmatics of natural and formal languages. Understanding quotation is thus of utmost importance for the philosophy of language and logic. In the first half of the twentieth century, quotation was primarily analyzed as a special semantic phenomenon of an innerlinguistic reference device. In section 2.1, we already mentioned that by means of quotation linguistic expressions are not used in order to relate to their outerlinguistic referents. Rather, quotation is a device of “turning language on itself” (see Cappelen/Lepore 2007: 1) in which linguistic expressions are mentioned, as in the following examples: (25)

a. b. c. d.

“Boston” has six letters. “Quote” rhymes with “goat”. “Quotation has a certain anomalous feature” is a sentence. “Living dead” is an oxymoron.

It is because the expressions in sentences (25a–d) are mentioned by means of quotation that those sentences can be semantically evaluated as true. So for example, ““Boston”” in sentence (25a) does not, of course, refer to the city of Boston, but to the linguistic expression “Boston” – and the linguistic property of having six letters can therefore be (truly) attributed to this expression. To account for the apparent fundamental difference between using and mentioning an expression, many philosophers insisted on an accurate employment of quotation marks. To confuse the use and mention of an expression was considered to be an indictable offense for any logician and philosopher of language (see, for example, Quine 1940: §4).

16

Elke Brendel, Jörg Meibauer, and Markus Steinbach

A fatal interdependency of a simultaneous use and mentioning of linguistic expressions also gave substance to certain self-referential logical and semantic paradoxes. For example, the notorious liar paradox emerged from a selfreferential sentence p which expresses its own falsehood. Since “p” is synonymous with ““p” is false,” and since “p” is apparently true if and only if p is the case, the contradiction that “p” is true if and only if “p” is false can be derived. Alfred Tarski has famously argued that in order to avoid self-referential paradoxes such as the liar paradox, we have to give up the assumption of a semantically closed language (see Tarski 1935). This means, in particular, that a consistent and materially adequate formal language L must not contain any quotation names of its own expressions. The quotation names of the expressions of L are part of the meta-language of L – a different and expressively richer language than L in which, in particular, the expressions of L are mentioned. The expressions of L’s meta-language are in turn part of an even richer meta-meta-language of L – and so on. As a consequence, a sentence attributing truth or falsehood to a sentence of L is a sentence of the meta-language of L and cannot be formulated within the object-language L itself. But in the liar paradox, a sentence p attributes falsehood to itself, i.e., p is used in an object-language and is at the same time meta-linguistically mentioned. So p violates the strict Tarskian hierarchy of different language-levels and does therefore not count as a well-formed sentence of a semantically open formal language. The discussion of self-referential paradoxes and, in particular, Tarski’s highly influential resolution of the liar paradox in formal languages motivated logicians and philosophers of language to focus initially on pure quotations and their applications in formal languages as a means of mentioning linguistic expressions. To be sure, the possibility of mentioning linguistic expressions of a formal language L and the possibility of formulating self-referential propositions via quotation within L does not necessarily lead to an inconsistency of L. In the aftermath of Tarski’s work on self-referential paradoxes logicians have therefore intensively explored the interrelations between self-reference, truth and contradiction. Although the study of pure quotation in formal logic has resulted in elaborated quotational logics for formal languages in which certain forms of selfreference are consistently expressible (see, for example, Smullyan 1957, 1987, Blau 1988 and 2008, Harth 2002, and Brendel 1993), a formal treatment of pure quotations in logical systems could not cope with the complexity of the different kinds of quotation in natural languages. In particular, the so-called proper name theory of quotation which was attributed to Tarski and Quine, and which forms the basis of many systems of quotational logic, has some serious shortcomings

Exploring the Meaning of Quotation

17

when applied to natural languages. According to this theory, a quotation, i.e., the quoted material together with the quotation marks, is treated as a simple, unstructured singular term – much like a proper name – and refers as a whole to the quoted linguistic expression. It is one of the advantages of the proper name theory that it accounts for the apparent referential opacity of pure quotations, i.e., co-referential expressions within quotation cannot be substituted for each other salva veritate. In this respect pure quotation is indeed very similar to proper names. So for example, substituting “the capital of Massachusetts” for “Boston” in (25a) would yield the false sentence (26): (26)

“The capital of Massachusetts” has six letters.

But treating quotations as referentially opaque expressions also creates problems for the proper name theory. In particular, the use of deictic expressions within quotation marks or anaphoric expressions referring to some parts of the quoted material indicates that quotations cannot always be understood as unstructured singular terms (see, for example, Partee 1973 and Seymour 1996). Furthermore, the proper name theory cannot accommodate the simultaneous use and mention of the quoted material in mixed quotations. In mixed quotations, but also in scare quotes and in emphatic quotes, quotations do not even seem to have the grammatical status of a singular term (see, for example, Cappelen/Lepore 1997: 437f.). Philosophers have also objected that the proper name theory cannot account for the apparent close, non-arbitrary relationship between a quotation and its referent (see, for example, Davidson 1984: 81f. and GarcíaCarpintero 1994: 254f.). It is also obvious that quotations do not capture their referents in a way in which names typically capture their referents (see, for example, Goméz-Torrente, this volume). In contrast to proper names, a quotation somehow points to its referent, i.e., to the linguistic expression inside the quotation marks. A quotation has, so to speak, its referent built in. Because of the above mentioned (and other) shortcomings of the proper name theory and because of a growing interest in quotation in the philosophy of ordinary language, the focus of the philosophical debate on quotation has been shifted to other theories of quotation in the last three decades. According to Donald Davidson’s highly influential demonstrative theory of quotation, the quotation marks (and not the whole quotation) function as a certain referring device (see Davidson 1984). They may be read as “the expression a token of which is here” (Davidson 1984: 90). The quoted material is then interpreted as an “inscription,” i.e., as the demonstrated token of an expression to which the quotation marks point. So, for example, the sentence (27)

“Alice swooned” is a sentence.

18

Elke Brendel, Jörg Meibauer, and Markus Steinbach

could be rewritten as (28)

Alice swooned. The expression of which this is a token is a sentence. (see Davidson 1984: 90)

A prominent further development and refinement of Davidson’s demonstrative theory is Manuel García-Carpintero’s deferred ostension theory of quotation (see García-Carpintero 1994, 2004 and this volume). According to GarcíaCarpintero’s theory, a quotation is an “ostensive sign.” As in Davidson’s account, the sole bearers of reference are the quotation marks, which are interpreted as a certain kind of demonstrative. The quoted material functions as a demonstrated index which helps to determine the referent together with the relevant contextually dependent speaker’s intention. The deferred ostension theory of quotation allows, in particular, to refer with a quotation not only to expression-types, but also to all different kinds of entities contextually related to the relevant token used, as for example, to certain graphic or phonetic types. One prominent opponent to the demonstrative theory of quotation is the socalled identity theory of quotation (see, for example, Frege 1982 and Washington 1992) or, as Cappelen/Lepore call it, the use theory of quotation (see Cappelen/Lepore 2007: ch. 8). According to the identity/use theory, it is neither the whole quotation nor the quotation marks that function as the bearer of reference. Rather, it is the quoted expression itself that does the referring. For example, in sentence (25a), it is “Boston”, i.e., the quoted material inside the quotation marks of the quotation ““Boston”” that refers to the expression “Boston.” So, in a way, the expression “Boston” refers to itself. In contrast to the demonstrative theory, in which the quotation marks play the fundamental role of reference, according to the identity theory, the quotation marks are merely marks that help to indicate that the quoted material is not used with its normal extension. For example, the quotation marks in (25a) indicate that “Boston” does not refer to the city in Massachusetts. One prominent recent theory of quotation that bears on some ideas of the identity/use theory is Paul Saka’s disambiguation theory of quotation (see, for example, Saka 1998). According to Saka, quotation marks do not contribute to the grammaticality of a sentence, neither in written nor in spoken language. They simply signal mentioning. Saka and other defenders of an identity/use theory criticize demonstrative theories for not adequately accounting for the prevalent phenomenon of mentioning an expression without using quotation marks. Although Saka’s view contrasts sharply with a Davidsonian theory with regard to the semantic function of the quotation marks, he also contends, like

Exploring the Meaning of Quotation

19

García-Carpintero, that quotations are ambiguous or indeterminate and that the speaker’s intention helps to identify the referent of a quotation. One concern put forward against both demonstrative theories and identity/use theories is that the reference relation in a quotation could become completely arbitrary. If the referent in a quotation depends solely on the speaker’s intention or on some other contextually relevant feature, it seems that there are no restrictions to what quotations could refer (see, for example Gómez-Torrente, this volume or Cappelen/Lepore 2007: 71f., and García-Carpintero’s reply to this objection in García-Carpintero, this volume). The main challenges and problems that are currently discussed with regard to quotation theories are questions concerning the status of quotation marks, questions concerning the context-dependency of quotations, and questions concerning the role semantic and pragmatic aspects play in quotations.

2.7. Puzzle 7: How can an analysis of quotation help clarify the debate about the semantics/pragmatics interface? The prevailing discussion about whether quotation is a (primarily) semantic or pragmatic phenomenon exhibits many similarities with the debate about the semantic-pragmatic divide with respect to other linguistic phenomena. Quotation is therefore, as Cappelen and Lepore put it, “an excellent testing ground for a general theory of the interface between semantic and non-semantic content” (Cappelen/Lepore 2007: 8). In particular, the growing interest in understanding open quotations has encouraged the development of pragmatic accounts of quotation. There is, for example, an ongoing debate about whether mixed quotations have to be analyzed in semantic or pragmatic terms. Many endorse a theory according to which mixed quotations (or the quotation marks in mixed quotations) do not function as a device for referring to linguistic expressions. They contend that mixed quotations are not semantically on a par with pure or direct quotations (see, for example, Recanati 2001, Reimer 2005, and Stainton 1999). According to Recanati, the quoted report in a mixed quotation has the same semantic content than it would have without quotation marks (see Recanati 2001: 660). The information that the speaker wants to communicate in a mixed quotation, namely that somebody uses certain words, is transferred via purely pragmatic mechanisms. Of course, a pragmatic account that adheres to the semantic redundancy of quotation marks in mixed quotations faces some problems, as Cappelen and Lepore have pointed out (see Cappelen/Lepore 2007: 54ff.). In particular, defenders of pragmatic accounts have to address apparent difficulties with indexicals in mixed quotation (29a), with the quotation of nonsense in

20

Elke Brendel, Jörg Meibauer, and Markus Steinbach

mixed quotation (29b), and with the quotation of foreign expressions in mixed quotation (29c). (29)

a. b. c.

When Markus saw the new Mercedes, he said that this is “my new car.” Nicola said that Kant is a “philosopher”. Kant said that all knowledge begins with “Erfahrung.”

In these cases, it seems that simply disquoting these quotes would not yield adequate results, i.e., would yield an unintended interpretation of the indexical, or an ungrammatical sentence. A special form of open quotations that was long neglected in the philosophical discussions about quotation but has recently received more attention (see, for example, Predelli 2003 and de Brabanter 2005) is the scare quotation. Whereas it is still disputed whether pure, direct, and mixed quotation can be principally analyzed in semantic terms, it seems that a full understanding of scare quotation definitely calls for a certain pragmatic account. As already mentioned in section 2.1, in a scare quotation, the quoted material receives an additional interpretation, such as a certain disassociation of the speaker from the content of the quoted expression. This additional interpretation seems to be driven by a pragmatic mechanism. Stefano Predelli has argued that scare quotes function as so-called attachment triggers. Besides its literal semantic content, which Predelli calls the message, a sentence with scare quotes also conventionally implicates an additional element, a so-called attachment that is triggered by the scare quotes and which accounts, for example, for a sarcastic or ironic interpretation. As with other linguistic expressions triggering conventional implicatures, both the message and the attachment contribute to the meaning of the sentence. Daniel Gutzmann and Erik Stei have raised some objections against Predelli’s view of analyzing scare quotes along the lines of conventional implicatures (see Gutzmann/Stei, this volume). Which attachment is triggered does not seem to be conventionally associated with the scare quotes. Whether scare quotes indicate a disagreement with the quoted material, or indicate sarcasm, irony or euphemism, or indicate a use of an idiolect etc., is context dependent. Furthermore, in some cases of utterances with scare quotes the “triggered attachment” seems to be the decisive communicated meaning. This is true, in particular, for cases of scare quotes indicating irony where the message and the attachment even contradict each other. Gutzmann and Stei champion a pragmatic theory in which scare quotes – but also other forms of quotation – contribute to the meaning of an utterance in which they occur by means of conversational implicatures. According to Gutzmann and Stei, quotation marks indicate that the sentence

Exploring the Meaning of Quotation

21

receives a context dependent and (via conversational principles) calculable interpretation that deviates from its literal interpretation – or at least, as it is the case in emphatic quotations, adds a certain emphasis to the quoted expression. The mentioned tendencies in focusing on various open quotations have shed some new light on the discussion about the semantics and pragmatics of quotation. Whether they can pave the way for a universal theory of quotation that can deal with both closed and open quotation, or whether quotation is a heterogeneous phenomenon that cannot be treated within a unified semantic and pragmatic framework, remains to be seen. 2.8. Puzzle 8: What role does context play in determining the meaning of quotation? Every act of direct, indirect, and mixed quotation involves at least two contexts, namely the context of the original utterance, and the context of the quotation. In a way, the latter makes part of the former accessible. This becomes most clear when context shift is studied. In most languages, indexical expressions in indirect quotations are interpreted in the actual context. Consequently, indexical expressions must be adapted to the new context (see Plank 1986). (30)

a. b.

Yesterday, Ken said: “Tomorrow, I will be there.” Yesterday, Ken said that today, he would be here.

In these examples, three shifts occur: temporal deixis: tomorrow > today; personal deixis: I > he, local deixis: there > here. In some languages, indexical expressions are (or can be) interpreted in the embedded context. Consequently, indexical expressions are not adapted to the new context. Slave is famous in this respect (Rice 1986 and Quer, this volume). (31)

John ?aranila séhdi John 2sg.home.go 3sg.said.1sg ‘John said to me: “You go home.”’ ‘John said to me, that I go home.’

Example (31) illustrates that in Slave, indexical expressions are bound by the most local speaker and hearer. Since quotations refer to an original utterance situation, the relevant aspects of that situation should be accessible for the quoting person: these are not only speaker and addressee, time and place of utterance, background of the discourse, etc., but also implicature, presupposition, speech act, reference, deixis, (in)definiteness, information structure, etc. Moreover, the quotational act itself

22

Elke Brendel, Jörg Meibauer, and Markus Steinbach

has a certain status as a specific illocution (not just a simple assertion) and a status in discourse (e.g., in news reports). Often contextual aspects (such as background information, etc.) are needed in order to decide about the conveyed meaning of a quotation utterance. Thus, the difference between emphatic quotes, scare quotes, and mixed quotes depends on the context of utterance: (32)

a. b. c.

Peter is a “real friend.” (He is always there when I need him.) Peter is a “real friend.” (I really hate him. He is having an affair with my wife.) Peter is a “real friend.” (That is how Paul described Peter to me.)

Thus, the respective context in brackets makes clear that (32a) is to be understood as an emphatic quote, while (32b) is a scare quote, and (32c) is a mixed quote. If one follows the ostension theory of Saka (1998), it may be argued that “quote marks […] serve to select from the multiple ostensions that are produced whenever any expression is uttered; they act to constrain pragmatic ambiguity or indeterminacy” (Saka 1998: 113). For instance, in the pure quotations in (33a–c), different uses of “run” are highlighted: (33)

a. b. c.

“Run” is used in the third-person plural but not singular. “Run” refers to run, runs, ran, running. “Run” consists of three letters.

Despite its prominence in pragmatic theory, not much progress has been made towards the development of a general theory of context. Clearly quotation is a phenomenon that will be crucial for such developments. 2.9. Puzzle 9: How are the different kinds of quotation acquired and processed? As should have become clear from the previous section, quotation is a highly complex phenomenon with many different linguistic and philosophical facets. In this final section we want to point out some aspects that are interesting for research on the acquisition and processing of quotations. Since quotation involves mentioning and metarepresentation, a competent use of quotation requires a certain advanced stage of cognitive development. In addition, many kinds of quotation can be used competently only if the graphematic, prosodic, morphosyntactic, semantic, and pragmatic categories briefly described in the previous sections have been acquired. Hickmann (1993: 72), for example, re-

Exploring the Meaning of Quotation

23

ports with regards to narrated films that “4-year-old-children had two clearly identifiable modes: the re-enacting and the descriptive modes. […] The 7-and 10-year-old children, as well as the adults, mostly relied on the reporting modes. However, the 7- and 10-year-olds clearly preferred the direct reporting mode, whereas the adults relied as frequently on the direct and the indirect reporting modes.” Recall from section 2.3 above that German indirect quotations can optionally be marked by the subjunctive. Moreover, in section 2.8 we mentioned that in indirect quotations indexical expressions are usually bound by the actual context of utterance. A pilot study at Mainz University on the acquisition of subjunctive mode and temporal indexicals in indirect quotations shows, however, an interesting difference between 11-year-old and 15-year-old pupils (see Strotkötter 2008). Only the older pupils frequently use the subjunctive as a marker of indirect quotation. By contrast, the younger pupils mainly use the unmarked indicative mode in indirect speech. In section 2.3 we mentioned that German, like many other languages, uses two different kinds of embedded clauses in indirect quotations. Interestingly, the 15-year-old pupils prefer to use the subjunctive in indirect quotations without a complementizer, i.e. in embedded sentences that have the same syntactic structure as main clauses. In sentence-final embedded clauses introduced by a complementizer, most of the older pupils do not use subjunctive but only indicative forms. Turning to temporal indexicals in indirect quotations, the younger children show a preference for the interpretation of temporal indexicals in the original context of utterance, whereas the older children interpreted the indexical expressions in the actual context of utterance. An especially interesting question for every theory of quotation is, of course, the acquisition and processing of quotation marks. Bredel (2004) investigates the acquisition of quotation marks by German children. Children start to use quotation marks in direct quotations by the age of 8. At this age, only about 50% of all possible quotation marks are used. In addition, opening quotation marks are used more securely than closing quotation marks. The beginning of the quoted expression seems to be more prominent than its end. Another interesting feature is the use of punctuation markers in quotations. Whereas children do not make many mistakes in the general placement of commas, only 17% of commas in direct quotations are placed correctly. Other uses of quotation marks are only rarely found in texts of children. Although the first scare quotations are already produced by 11-year-old children, a regular use of scare quotations is not evidenced before the age of 15. Quotation in general and quotation marks in particular are also interesting examples from the perspective of language comprehension. In recent years,

24

Elke Brendel, Jörg Meibauer, and Markus Steinbach

there is a lively debate on an old issue, namely the proper distinction between semantics and pragmatics. The background is the classical Gricean distinction between what is said on the one hand and what is implicated on the other. But how exactly the boundaries are to be drawn is still disputed. Some linguists and philosophers assume that underdetermined propositional structures need pragmatic enrichment. Others defend a minimalist view on truth conditions (for an overview see Meibauer 2006). From a linguist’s point of view, there has, however, always been a regrettable lack of empirical data in an otherwise sophisticated debate. Recently, a new strand of research has come in under the name of experimental pragmatics (see Noveck/Sperber 2004, Sauerland/Yatsushiro 2010 and Meibauer/Steinbach 2011). Experimental pragmatics is the attempt to collect experimental data on pragmatic issues by using psycholinguistic and neurolinguistic methods. One promising aspect of this research is that competing theories on the interface between semantics and pragmatics may be directly tested. The discussion in the previous sections has made clear that the different kinds of quotation investigated in this book are a perfect empirical and theoretical test ground for experimental studies investigating this interface. All dimensions discussed in section 2.2 may affect the processing of quotations: quotations in which the quoted material is only mentioned may differ from quotations in which it is both mentioned and used, quotations without quotation marks may differ from quotations with quotation marks, quotations that refer to an original utterance may differ from quotations that do not refer to an original utterance, and finally quotations that involve pragmatic enrichment (or trigger a conversational implicature) may differ from quotations that do not involve pragmatic enrichment. Moreover, almost all theories of quotations in the philosophy of language and in linguistics make predictions about the interaction of semantics and pragmatics. Many recent accounts of quotations are, for example, versions either of a demonstrative theory or an identity/use theory, or versions that combine elements of both. An exception is Cappelen’s and Lepore’s recent account of a minimal theory of quotation in which quotation is interpreted as a semantic and context insensitive phenomenon (see Cappelen/Lepore 2007). Hence, these rivaling theories provide a good theoretical basis for experimental studies of different kinds quotations.

Exploring the Meaning of Quotation

3.

25

Content of this book

This volume is a collection of original papers on topical issues concerning various linguistic and philosophical aspects of quotations. The articles take up many of the topics discussed in the previous sections. In her paper Attitudes toward Quotation, Barbara Abbott argues against quotational analyses of propositional attitudes proposed by W.V.O. Quine and Donald Davidson. She claims that the success of such an analysis is based on the assumption that languageless creatures do not have propositional attitudes. Since this assumption is dubious and contestable, Abbott suggests dismissing quotational analyses of propositional attitudes wholesale. As a consequence, she no longer sees any motivation for embracing a demonstrative theory of quotation. Whatever quotation is, it is, according to Abbott, not demonstrative in character. It is the main goal of Manuel Bremer’s paper to explore the possibilities of a quotation free semantics of formal languages. He shows that a theory of paraconsistent logics, along the lines of Graham Priest’s “dialethist” account, provides the semantic resources for Truth Value Talk without Quotation. Such a paraconsistent logic, which allows some contradictions to be true is based on a semantically closed language and contains, in particular, its own semantic expressions. Bremer argues that these expressions need not be modeled in the traditional way as predicates attributing semantic values to sentences by means of quotation, but can be formalized as certain (bivalent) operators. Sarah-Jane Conrad in her contribution Disquotational Indirect Report in Focus focuses on the semantics/pragmatics divide, as conceived of by Kent Bach in a number of papers. In Bach’s approach the so-called disquotational indirect reports (DIR) play an important role because they serve as a means to identify the content alias ‘what is said’ of an utterance. Conrad shows that this conception plays a major role in minimalist as well as contextualist approaches to the semantics/pragmatics divide. In their contribution Behabitive Reports Cathrine Fabricius-Hansen and Kjell Johann Sæbø investigate a specific family of illocutionary verbs, so-called “verbs of judging” or “behabitives” such as criticize or praise. They illustrate a curious ambivalence regarding the role played by their complement, which can be an object or a content of judgment. From a semantic viewpoint, these verbs select two different propositional arguments, while from a syntactic viewpoint, only one argument can be expressed in a sentence (beside the subject). According to Fabricius-Hansen and Sæbø, the argument that is not expressed in syntax is nonetheless understood to be there as an implicit semantic argument. It ei-

26

Elke Brendel, Jörg Meibauer, and Markus Steinbach

ther receives a definite (in the object case) or an indefinite (in the content case) interpretation. Manuel García-Carpintero deals with Double-Duty Quotation, Conventional Implications and What is Said. His analysis is based on his earlier deferred ostension view of quotation, and is contrasted with the minimal theory of Cappelen/Lepore (2007). For García-Carpintero, the concept of double-duty quotation is a way of understanding mixed quotations (alias open quotation or impure quotation). In his analysis, a distinction between the asserted content and the conventionally implicated content is made, such that double-duty quotations are cases of conventional implicatures. In his paper What Quotations Refer To, Mario Gómez-Torrente critically examines García-Carpintero’s deferred ostension view and Cappelen’s and Lepore’s minimal theory of quotation and develops his own account in which quotations are used in a purely referential way. He argues, in particular, that GarcíaCarpintero’s account allows quotation to have a too extensive flexibility and semantic versatility. Against Cappelen and Lepore’s account he subscribes to the thesis of a semantic uniformity of quotation. According to Gómez-Torrente, a quotation semantically refers to a unique abstract expression type, and pragmatic conditions of successful conveyed speaker reference account for a possible difference between speaker and semantic reference. In their paper Quotation Marks and Kinds of Meaning, Daniel Gutzmann and Erik Stei argue for a pragmatic account of quotation. According to Gutzmann and Stei, quotation marks are best understood as certain devices for generating conversational implicatures. They argue that all general features typically ascribed to conversational implicatures, such as context dependency, calculability, and cancelability, apply to quotational sentences. With their pragmatic theory of quotation, Gutzmann and Stei strive for the ambitious goal to achieve a universal theory that accounts for both open and closed quotation. In his contribution “Quotation and Pictoriality”, Manfred Harth defends a special version of an identity theory of quotation in which the pictoriality of a quotation is explained via a certain exemplification relation between a demonstrated token and its type. Harth claims that in comparison with the demonstrative theory, his exemplification theory is the more natural and intuitive theory of quotation. According to Harth, the exemplification theory provides, in particular, a more satisfying explanation of the dispensability of quotation marks and of a certain context-insensitivity of quotational sentences than its alternative theory. In The Punctuation Theory of Quotation, the author Michael Johnson argues that direct and mixed quotation are best analysed pragmatically. The contribution of quotation marks may not be reduced to the status of presuppositions,

Exploring the Meaning of Quotation

27

conversational implicatures, or conventional implicatures. The punctuation theory claims that punctuation marks deliver entailments sui generis. A solution to the main problems for pragmatic theories, namely those that arise in connection with indexicality, is sketched. Misrepresenting Misrepresentation, by Ernie Lepore and Michael Johnson, is an attack on the demonstration theory, a theory that analyses quotations as forms of demonstrations, as put forward by Herbert Clark and collaborators (Clark/Gerrig 1990, Wade/Clark 1993). There is, according to the authors, theoretical and empirical evidence against the demonstration theory. The authors argue instead for a version of a wording theory, understood as a theory demanding that the truth conditions of a direct or mixed quotation require that the quotational content has been said. Jürgen Pafel in Two Dogmas on Quotation, scrutinizes two standard assumptions of most theories of quotation: (i) pure quotations are syntactically noun phrases and semantically singular terms (i.e., dogma one), and (ii) pure and direct quotation are more or less on a par with each other syntactically and semantically (i.e. dogma two). Pafel shows in detail that syntactically, pure quotations definitely have the distribution of nouns, not of noun phrases. Semantically, their distribution suggests a predicate view of pure quotations, as nouns are usually predicates. In addition, pure quotations clearly differ from direct quotations in a significant manner in syntax and semantics: pure quotations are nouns, while direct quotations are sentences. Consequently, Pafel analyzes pure quotations as predicates whose meaning is to have a certain shape, with the shape being that which the pure quotation presents in between the quotations marks. Direct quotations, on the other hand, are propositions denoted by the quoted sentence q if q is uttered in context C. These two different uses of quotation marks nevertheless constitute a family of uses that resemble each other. Hence, quotation marks are analyzed a case of polysemy. In his contribution Reporting and Quoting in Signed Discourse, Josep Quer offers an overview of the mechanisms used in sign languages to report someone’s utterances, thoughts, or actions, which are commonly known as role shifts. Despite being able to resort to embedding under attitude predicates for indirect reported discourse, sign languages have a genuine strategy that flags the reported segment with an array of non-manual markers, which anchor it to the reported subject, as well as with a displacement in the referential framework for indexicals. Relying on Catalan Sign Language (LSC) data, he shows that despite the surface appearance of direct quotation, in LSC role shift used for direct and indirect discourse can actually be distinguished: the lexical markers introducing it, the syntactic position of the reported clause, and, most interest-

28

Elke Brendel, Jörg Meibauer, and Markus Steinbach

ingly, the (non)-shifting option for indexicals signal the two types of quoted discourse. Elaborating on his earlier account (Saka 1998), Paul Saka discusses in his contribution The Act of Quotation several versions of the identity theory of quotation (“quotations refer”) and points out arguments for a speech-act version (“speakers refer”) of this approach. In particular, he develops and explains (natural) Speech-Act Theses such as Universal Identity, Quotation Identity, Quotatum Identity, and Partial Identity. Moreover, he argues that the mechanism of quotation is referentially indeterminate and that there is a multiplicity of senses connected with it. Savas L. Tsohatzidis argues that two prominent principles concerning the semantics and interpretation of quotation cannot be held conjointly without giving rise to inconsistencies. The first principle, the so-called Reference Thesis (RT), states that an adequate truth-theoretic semantics for English will include all substitution instances of a general schema, according to which the referent of a quoted proper name is identical with the disquoted proper name. The second principle, the so-called Quotation Thesis (QT), expresses the view that a quoted and a disquoted occurrence of a proper name do not have the same referent. Tsohatzidis claims that while QT entails, for example, the truth of the sentence “The referent of “Aristotle” is not Aristotle”, RT, however, yields the contradictory result that the sentence “The referent of “Aristotle” is Aristotle” is true. Since, according to Tsohatzisis, there are no non-question-begging resolutions of this problem in sight, he opts for rejecting at least one of the two principles. References Blau, Ulrich 1988 Blau, Ulrich 2008 Bredel, Ursula 2004

Brendel, Elke 1993

Die Logik der Anführung und Quasianführung. Erkenntnis 29: 227– 268. Die Logik der Unbestimmtheiten und Paradoxien. Heidelberg: Synchron. Die Didaktik der Gänsefüßchen. In Schriftspracherwerb und Orthographie, U. Bredel, G. Siebert-Ott and T. Thelen (eds.), 207–240. Hohengehren: Schneider. Partial worlds and paradox. Erkenntnis 39: 191–208.

Exploring the Meaning of Quotation

29

Brendel, Elke, Jörg Meibauer & Markus Steinbach (eds.) 2007 Zitat und Bedeutung. Hamburg: Buske. Bucali´c, Tomislav 2007 Ein typologischer Beitrag zu Formen der Redewiedergabe. In Zitat und Bedeutung, E. Brendel, J. Meibauer and M. Steinbach (eds.), 45– 66. Hamburg: Buske. Cappelen, Herman & Ernie Lepore 1997 Varieties of quotation. Mind 106: 429–450. Cappelen, Herman & Ernie Lepore 1999 Using, mentioning and quoting: A reply to Saka. Mind 108: 741–750. Cappelen, Herman & Ernie Lepore 2007 Language Turned on Itself. The Semantics and Pragmatics of Metalinguistic Discourse. Oxford: Oxford University Press. Clark, Herbert H. & Richard J. Gerrig 1990 Quotations as demonstrations. Language 66: 764–805. Coulmas, Florian 1986a Reported speech: Some general issues. In Direct and Indirect Speech, F. Coulmas (ed.), 1–28. Berlin: Mouton. Coulmas, Florian (ed.) 1986b Direct and Indirect Speech. Berlin: Mouton de Gruyter. Cumming, Samuel 2005 Two accounts of indexicals in mixed quotation. In Hybrid Quotation, P. de Brabanter (ed.), 77–88. Amsterdam: Benjamins. D’Avis, Franz J. 2005 Über Parenthesen. In Deutsche Syntax: Empirie und Theorie, F.J. d’Avis (ed.), 259–279. Göteborg: Acta Universitatis Gothoburgensis. D’Avis, Franz J. 2007 Zitat und Sprecherbezug. In Zitat und Bedeutung, E. Brendel, J. Meibauer and M. Steinbach (eds.), 67–88. Hamburg: Buske. Davidson, Donald 1984 Quotation. In Inquiries into Truth and Interpretation, D. Davidson, 79–92. Oxford: Oxford University Press. De Brabanter, Philippe (ed.) 2005 Hybrid Quotation. Amsterdam: Benjamins. De Brabanter, Philippe (ed.) 2010 Quotation. In The Pragmatics Encyclopedia, L. Cummings (ed.), 376–379. London: Routledge.

30

Elke Brendel, Jörg Meibauer, and Markus Steinbach

Fabricius-Hansen, Cathrine 1989 Tempus im indirekten Referat. In Tempus – Aspekt – Modus: Die lexikalischen und grammatischen Formen in germanischen Sprachen, W. Abraham and T. Janssen (eds.), 155–182. Tübingen: Niemeyer. Fabricius-Hansen, Cathrine 2002 Nicht-direktes Referat im Deutschen – Typologie und Abgrenzungsprobleme. In Modus, Modalverben, Modalpartikeln, C. FabriciusHansen, O. Leirbukt and O. Letnes (eds.), 6–29. Trier: WVT. Frege, Gottlob 1892

Über Sinn und Bedeutung. Zeitschrift für Philosophie und philosophische Kritik 100: 25–50. (English translation: On Sense and Reference. In Translations from the Philosophical Writings of Gottlob Frege, P. Geach and M. Black (eds.), 56–78. Oxford: Basil Blackwell, 3rd edn. 1980.)

García-Carpintero, Manuel 1994 Ostensive signs: Against the identity theory of quotation. Journal of Philosophy 91: 153–264. García-Carpintero, Manuel 2004 The deferred ostension theory of quotation. Noûs 38: 674–692. Golato, Andrea 2000

An innovative German quotative for reporting on embodied actions: Und ich so/und er so ‘and I’m like/and he’s like’. Journal of Pragmatics 32: 29–54.

Gómez-Torrente, Mario 2001 Quotation revisited. Philosophical Studies 102: 123–153. Güldemann, Tom & Manfred von Roncador (eds.) 2002 Reported Discourse. A Meeting Ground for Different Linguistic Domains. Amsterdam: Benjamins. Güldemann, Tom, Manfred von Roncador & Wim van der Wurff 2002 A comprehensive bibliography of reported discourse. In Reported Discourse. A Meeting Ground for Different Linguistic Domains, T. Güldemann and M. von Roncador (eds.), 365–415. Amsterdam: Benjamins. Harth, Manfred 2002

Anführung. Ein nicht-sprachliches Mittel der Sprache. Paderborn: Mentis.

Herrmann, Annika & Markus Steinbach 2007 Wenn ich nicht ich ist. Direkte und indirekte Rede in Gebärdensprache. In Zitat und Bedeutung, E. Brendel, J. Meibauer and M. Steinbach (eds.), 153–179. Hamburg: Buske.

Exploring the Meaning of Quotation

31

Hickmann, Maya 1993 The boundaries of reported speech in narrative discourse: Some developmental aspects. In Reflexive Language. Reported Speech and Metapragmatics, J.A. Lucy (ed.), 63–90. Cambridge: Cambridge University Press. Klockow, Reinhard 1980 Linguistik der Gänsefüßchen. Untersuchungen zum Gebrauch der Anführungszeichen im gegenwärtigen Deutsch. Frankfurt/M.: Haag und Herchen. Li, Charles N. 1986

Direct and indirect speech: A functional study. In Direct and Indirect Speech, F. Coulmas (ed.), 29–45. Berlin: Mouton de Gruyter.

Lillo-Martin, Diane 1995 The point of view predicate in American sign language. In Language, Gesture, and Space, K. Emmorey and J. Reilly (eds.), 155–170. Hillsdale, NJ: Erlbaum. Meibauer, Jörg 2006 Meibauer, Jörg 2007

Implicature. In Encyclopedia of Language and Linguistics. 2nd edition. Vol. V, K. Allen (ed.), 568–580. Oxford: Elsevier. Syngrapheme als pragmatische Indikatoren: Anführung und Auslassung. In Von der Pragmatik zur Grammatik, S. Döring and J. GeilfußWolfgang (eds.), 21–37. Leipzig: Universitätsverlag.

Meibauer, Jörg & Markus Steinbach (eds.) 2011 Experimental Pragmatics/Semantics. Amsterdam: John Benjamins. Nunberg, Geoffrey, Ted Briscoe & Rodney Huddleston 2002 Punctuation. In The Cambridge Grammar of the English Language, R. Huddleston and G.K. Pullum (eds.), 1723–1765. Cambridge: Cambridge University Press. Partee, Barbara 1973

Plank, Frans 1986

The syntax and semantics of quotations. In A Festschrift for Morris Halle, S. R. Anderson and P. Kiparsky (eds.), 410–418. New York: Holt, Rinehart and Winston. Über den Personenwechsel und den anderer deiktischer Kategorien in der wiedergegebenen Rede. Zeitschrift für germanistische Linguistik 14: 284–308.

Predelli, Stefano 2003 Scare quotes and their relation to other semantic issues. Linguistics and Philosophy 26: 1–28.

32

Elke Brendel, Jörg Meibauer, and Markus Steinbach

Quer, Josep 2005

Context shift and indexical variables in sign languages. In Proceedings of Semantics and Linguistics Theory 15, E. Georgala and J. Howell (eds.), 152–168. Ithaca, NY: CLC Publications.

Quine, Willard Van Orman 1940 Mathematical Logic. Cambridge, MA: Harvard University Press. Quine, Willard Van Orman 1970 Philosophy of Logic. Englewood Cliffs: Prentice Hall. Recanati, François 2000 Oratio Obliqua, Oratio Recta. An Essay on Metarepresentation. Cambridge, MA: MIT Press. Recanati, François 2001 Open Quotation. Mind 110: 637–687. Recanati, François 2004 Literal Meaning. Cambridge: Cambridge University Press. Reimer, Marga 2005

Reis, Marga 1995

Reis, Marga 1997

Rice, Keren 1986

Saka, Paul 1998

Too counter-intuitive to believe? Pragmatic accounts of mixed quotation. In Hybrid Quotation, P. de Brabanter (ed.), 167–186. Amsterdam: Benjamins. Extraction from verb-second clauses in German? In Extraction and Extraposition, U. Lutz and J. Pafel (eds.), 45–88. Amsterdam: Benjamins. Zum syntaktischen Status unselbständiger Verbzweit-Sätze. In Syntax im Fokus, C. Dürscheid, K.H. Ramers and M. Schwarz (eds.), 112– 144. Tübingen: Niemeyer. Some remarks on direct and indirect speech in Slave (Northern Athapaskan). In Direct and Indirect Speech, F. Coulmas (ed.), 47–76. Berlin: Mouton de Gruyter. Quotation and the use-mention distinction. Mind 107: 114–136.

Schecker, Michael 2002 Über den Konjunktiv in der indirekten Rede. In Redewiedergabe, Redeerwähnung. Formen und Funktionen des Zitierens und Reformulierens im Text, D. Baudot (ed.), 1–14. Tübingen: Stauffenburg. Seymour, Daniel 1996 Content and quotation. Rivista di Linguistica 8: 309–329.

Exploring the Meaning of Quotation

33

Smullyan, Raymond M. 1957 Languages in which self reference is possible. In: Journal of Symbolic Logic 22: 55–67. Smullyan, Raymond M. 1987 Quotation and self-reference. In Self-Reference, S.J. Bartlett and P. Suber (eds.), 122–144. Dordrecht: Martinus Nijhoff. Stainton, Robert 1999 Remarks on the syntax and semantics of mixed quotation. In Philosophy of Linguistics, K. Murasugu and R. Stainton (eds.), 259–278. Boulder: Westview. Steinbach, Markus 2007 Integrated parentheticals and assertional complements. In Parentheticals, N. Dehé and Y. Kavalova (eds.), 53–87. Amsterdam: Benjamins. Steinhauer, Karsten 2003 Electrophysiological correlates of prosody and punctuation. Brain and Language 86: 142–164. Strotkötter, Marlene 2008 Die Flexibilität der indirekten Rede. MA-thesis, University of Mainz. Tarski, Alfred 1935

Der Wahrheitsbegriff in den formalisierten Sprachen. Studia Philosophica 1: 261–405. (English translation: The concept of truth in formalized languages. In Logic, Semantics, Metamathematics, Tarski, A. (ed.), 152–278. Indianapolis: Hackett, 2nd edn.)

Truckenbrodt, Hubert 2006 On the semantic motivation of syntactic verb movement to C in German. Theoretical Linguistics 22: 257–306. Wade, Elisabeth & Herbert H. Clark 1993 Reproduction and demonstration in quotations. Journal of Memory and Language 32: 805–819. Washington, Corey 1992 The identity theory of quotation. The Journal of Philosophy 89: 582– 605. Zifonun, Gisela, Ludger Hofmann & Bruno Strecker et al. 1997 Grammatik der deutschen Sprache. Berlin: de Gruyter.

Attitudes Toward Quotation1 Barbara Abbott

1.

Introduction

As is well known, Frege (1892) argued that the sentential complements of propositional attitude predicates refer to propositions. W.V. Quine, who disdained intensional objects like propositions, briefly suggested instead an analysis of such complements crucially involving quotation (1956), and Donald Davidson took up and elaborated this suggestion in a number of papers (1969, 1975, 1979). The main purpose of this paper is to argue against quotational analyses of propositional attitudes, although I’ll suggest at the end that the result may have consequences for the analysis of quotation itself. In the second section below we will review Quine’s comments and Davidson’s development of them. In § 3 we look at considerations involving proper names which seem at first to go in favor of this kind of analysis, but which ultimately probably do not. In § 4 we turn to an argument against quotational analyses – the fact that it seems to deny languageless creatures2 propositional attitudes. The final section contains some concluding remarks. 2.

Quine’s and Davidson’s accounts of attitudes

Although Davidson’s analysis of propositional attitudes is based on the one Quine briefly suggested (but ultimately rejected) there are subtle differences between the two, so we need to look in turn at what each has had to say.

1. Thanks to Rich Hall and an anonymous reviewer for their helpful comments on a previous draft version. 2. By “languageless creature” I mean a creature who does not have a natural language like English or Vietnamese. Fodor (1975, 1987) has argued for an innate, unlearned, language of thought for both humans and nonhuman animals, but such a language (if it exists) will not be under consideration here. I’m grateful to Rich Hall for reminding me of its potential relevance.

36

Barbara Abbott

2.1. Quine’s view As noted in the introduction, Quine was wary (to say the least) about intensional objects like propositions, commenting that “there is a way of dodging the intensions which merits serious consideration. Instead of speaking of intensions we can speak of sentences, naming these by quotation. Instead of: w believes that ..... we may say: w believes-true ‘.....”’; and he remarked that “[a] similar shift can be made in the case of the other propositional attitudes, of course, …” (Quine 1956: 185). One problem immediately strikes one (perhaps especially if one is an animal lover), and in the very next paragraph Quine addressed it. This semantical reformulation is not, of course, intended to suggest that the subject of the propositional attitude speaks the language of the quotation, or any language. We may treat a mouse’s fear of a cat as his fearing true a certain English sentence. This is unnatural without being therefore wrong. It is a little like describing a prehistoric ocean current as clockwise. (Quine 1956: 186.)

We will return to this issue shortly. But to complete our review we must first take note of an additional complication, originally pointed out by Church (1950). Since any given phonological or orthographic string could, by chance, have two very different meanings in two different languages, the quotation analysis of propositional attitudes needs to leave a slot for a specification of the relevant language. Thus we need to replace (1) with (2). (1) (2)

w believes-true ‘.....’ w believes-true ‘.....’ in L

Quine expressed some distress at the need of the analysis to mention a particular language, not only because that does not seem to be a part of any ordinary statement of belief (or other propositional attitude), but also because of the illdefinedness of languages. His conclusion was thus anything but definitive: The propositional attitudes are dim affairs to begin with, and it is a pity to have to add obscurity to obscurity by bringing in language variables too. Only let it not be supposed that any clarity is gained by restituting the intensions. (Quine 1956: 187)

And in Word & Object, he opted instead for an “analysis” which deems a propositional attitude predicate and its complement to be one indissoluble unit (Quine 1960: 216).

Attitudes Toward Quotation

37

2.2. Davidson’s view Davidson was concerned that Quine’s ultimate solution to the problem of propositional attitude contexts would result in an unlimited number of distinct atomic expressions for a language, ruling out the possibility of a Tarskian truth definition. He thought Quine’s earlier suggestion, involving quotation, was superior, and he sought to modify it so as to as to circumvent the problems Quine had noted. Davidson proposed what has come to be called the “demonstrative” theory of quotation. On Davidson’s theory, quoted elements are not semantically a part of the sentences in which they appear. They stand instead as separate utterances, and the quotation marks which surround them are seen as a kind of demonstrative which points to that now distinct utterance. Thus his analysis of (3) would be something like (4) (cf. Davidson 1979: 91). (3) (4)

“Donald collapsed” is a sentence. Donald collapsed. The expression of which that is a token is a sentence.

Davidson (1969) applied this theory of quotation to contexts of indirect speech. His analysis of (5) is given, in abbreviated form, in (6). (5) (6)

Galileo said that the earth moves. Galileo said that. The earth moves.

We are to take the second element of (6) as an utterance rather than a sentence; that way, we don’t have to worry about the language issue – the utterer of (6) makes the language determinate. Of course Galileo didn’t make his famous statement in English, so we need to understand Galileo said that, as it occurs in (6), in a special way. On Davidson’s analysis it would abbreviate (7) (Davidson 1969: 105). (7)

∃x[Galileo’s utterance x and my next utterance make us samesayers]

Introducing the relation of samesaying would seem to introduce an intensional element into the analysis – something Davidson, following Quine, did not want to do. A footnote (note 14, added in the 1984 reissue of this paper) assures us that this is not the case. In the introductory paragraph of his 1969 paper, Davidson suggests that he is about to give, not only the correct analysis of indirect discourse, but also “an analysis that opens a lead to an analysis of psychological sentences generally (sentences about propositional attitudes, so-called) …” (93). However nothing

38

Barbara Abbott

more is said in this paper about exactly how the extension would go. Let us make an attempt to fill in some details. Consider (8): (8)

Louise would like to go out for a while.

As a first crude step, we might paraphrase (8) in Davidsonian fashion as in (9): (9)

Louise would like that. Louise goes out for a while.

This appears to put Louise in the liking relation to an utterance. How can we make sense of this? Somehow we must consider Louise would like that, in (9), to abbreviate something which yields the right meaning for (8). Possibly this would be something like (10). (10)

∃x[Louise’s utterance of x and my next utterance would make us samesayers, and x expresses Louise’s desire]

Of course we would have to consider Louise’s utterance of x as describing something hypothetical rather than actual. This is so not only because people can have all kinds of unexpressed desires, but also because Louise is a cat and can’t say anything as articulated as my utterance of Louise goes out for a while. This may not be exactly what Davidson had in mind, but since he did not spell that out we will have to make do. We turn now first to the potential (but probably not actual) argument in favor of this kind of analysis, and then the argument against. 3.

A potential argument in favor of quotation

As was alluded to in our introduction, Frege proposed a shift of reference in propositional attitude contexts, whereby expressions come to denote their customary senses rather than their customary referents. He did this, at least in part, in order to account for apparent failures of substitutivity – the fact that expressions which are ordinarily coreferential are not in general intersubstitutable in such contexts. (11) gives a standard example (from Linsky 1967: 74). (11)

a. b. c.

Oedipus wanted to marry Jocasta. Jocasta = Oedipus’s mother. Oedipus wanted to marry his mother.

Although (11a) and (11b) are true, our first impression is that (11c), on at least one interpretation, is not. (If (11c) is true, why did Oedipus put his eyes out when he found out what he had done?) Frege’s analysis seems to save the day:

Attitudes Toward Quotation

39

if we assume that Jocasta and Oedipus’s mother mean different things (i.e. have different senses), then given a shift in reference following wanted to, they are no longer coreferential as they occur in (11a/c). This solution seems to assume that a proper name like Jocasta actually has a meaning, something Frege appeared to accept, albeit with reluctance, in his famous footnote 2 of “On sense and reference”. However Mill (1843), before Frege, and Kripke (1972), after him, argued convincingly that proper names do not have meanings, or Fregean senses. Well, maybe we can still manage with the example in (11), since even without a meaning Jocasta would not be equivalent to Oedipus’s mother, which undoubtedly does have a meaning. However there exist quite a few entities with more than one proper name, especially these days. (12) illustrates a problem parallel to that shown in (11). (12)

a. b. c.

Gillian knows that John Wayne was a famous actor. John Wayne = Marion Morrison. Gillian knows that Marion Morrison was a famous actor.

Here again, it seems that the first two sentences could be true while the third is not. But if proper names do not express a sense, Frege has not given us a reason why we cannot substitute one for another in (12a). A successful quotational analysis of the complements of propositional attitude verbs would seem to solve our problem. If propositional attitudes are (somehow) relations between individuals and sentences, or utterances, then any difference in sentence/utterance can be expected to have the potential to make a difference in the attitude attributed. Since the sentences following that in (12a) and (12c) are different, as would be any utterances of those sentences, we no longer expect (12a) to entail (12c) even given the equivalence in (12b). I called this a “potential” argument for a quotation analysis because of some considerations pointed out by Kripke (1979). Peter is an American lad who knows that Paderewski was a Polish statesman. He also has heard about Paderewski the famous musician, but thinks that these are two different people, and also has a firm belief that politicians never have musical talent. The crucial example is given in (13). (13)

Peter believes that Paderewski had musical talent.

The problem is that we seem to have a situation that is very similar to that which Gillian is in with respect to John Wayne/Marion Morrison, but we do not have two different names with which to express Peter’s diverging beliefs. The Davidsonian analysis of (13) is given in (14).

40

Barbara Abbott

(14)

Peter believes that. Paderewski had musical talent.

There is only a single utterance to express Peter’s divergent beliefs. If Peter’s situation is indeed the same as Gillian’s, then the quotational analysis has not been shown to provide a solution. 4.

An argument against

Let us turn now to the other argument, the one which goes against quotation analyses of propositional attitude contexts. It is the one to which Quine alluded when he reassured us that such analyses, or his version anyway, need not be taken so as to exclude languageless creatures from the realm of the believing and desiring. Should we accept this reassurance? Davidson himself suggests that under one natural construal of the Quinean line, we shouldn’t. That would be a construal such that to have a thought is “to have a disposition to utter certain sentences with appropriate force under given circumstances” (Davidson 1975: 167). However Davidson, apparently like Quine, concludes that this construal is not necessary. The paratactic analysis of the logical form of attributions of attitude can get along without the mimic-theory of utterance. When I say, ‘Jones believes that snow is white’ I describe Jones’s state of mind directly: it is indeed the state of mind someone is in who could honestly assert ‘Snow is white’ if he spoke English, but that may be a state a languageless creature could also be in. (Davidson 1975: 167.)

From Quine we have at least the analogy to describing prehistoric ocean currents; from Davidson only a disclaimer. So there are two issues here. One is whether or not we should accept these reassurances from Quine and Davidson, that their analyses of propositional attitudes as relations to sentences or utterances do not preclude the truthful attribution of such attitudes to languageless creatures. And the second is whether we should in fact exclude languageless creatures from such attributions – whether in fact one must know a language before one can have a belief or a desire. 4.1. Do quotational analyses rule out languageless creatures? Is describing a mouse’s fear of a cat as a relation to a sentence (or an utterance) really like describing a prehistoric ocean current as clockwise? I don’t think so. In the case of the clock and the ocean current it is clear that there is a certain direction of motion which is shared between the two (assuming the description is

Attitudes Toward Quotation

41

correct). The fact that the two things (clocks and the ocean current) do not exist at the same time is not relevant to that direction of motion. Once the direction of motion has been characterized in our present day terms we know what it is, and it would be the same for both clocks nowadays, and the ocean current prehistorically. Furthermore we could, of course, have conveyed the same information in different terms – e.g. by describing the current as one which goes northward on its western side and southward on the east. However if we accept Quine’s and Davidson’s disclaimers, then it is not at all clear what is shared between, on the one hand, the state of a mouse who fears the attack of a cat (or a human who fears the attack of a lion), and on the other, a sentence or utterance. Consider again Davidson’s remark from the quote above: When I say, ‘Jones believes that snow is white’ I describe Jones’s state of mind directly: it is indeed the state of mind someone is in who could honestly assert ‘Snow is white’ if he spoke English, but that may be a state a languageless creature could also be in.

What kind of state is this? If Jones a normal human being who is not an English speaker, we could almost certainly teach them enough English so that they could assert “Snow is white” if they felt that that were true. But if Jones is a cat or a mouse, we could not teach them English – it is not possible to satisfy the if clause of this analysis. In order to believe that a mouse or a cat could be in this very same state is to disassociate this state completely from any state which we might have thought Davidson was talking about – and in particular any state involving sentences or utterances of English. Indeed, we know nothing at all about this state. Consider a parallel: we speak about the weather using sentences of English. Extending the quotational analysis, we might analyze weather phenomena as properties of those very sentences of English, or utterances of them. What is it to be raining? It is to be such that an utterance of the sentence It is raining is true. In giving this kind of analysis we can individuate all the possible weather conditions we might need to. However the analysis does not really tell us anything about weather – it does not tell us what rain consists of, or what causes it to be raining or not raining. In the case of Quine’s ocean current we have a direction of motion which is shared by two things, so that when we are told (accurately and believably) that the current is clockwise we learn something about it. But in the case of the propositional attitudes we are not offered anything which is shared by sentences or utterances of English, on the one hand, and the minds of possibly languageless creatures on the other. Ultimately, it seems to me, if we actually do want to allow that languageless creatures might have propositional attitudes then we cannot learn anything

42

Barbara Abbott

about their attitudes, or our own, from the Quine-Davidson type of analysis. We would have to conclude that such analyses are empty, and convey nothing of interest. Thus it seems crucial to the success of the quotational approach to establish that languageless creatures cannot, in fact, have propositional attitudes. Perhaps it is because of this that Davidson is at some pains to argue that very point. 4.2. Can languageless creatures have beliefs and desires? In his paper “Thought and talk” Davidson gives us no very powerful direct reason to believe that our attributions of propositional attitudes to languageless creatures are incorrect, although he suggests that such attributions “smack of anthropomorphism” (1975: 155). Instead his argument is rather contorted and indirect. But before we consider it, let us consider on independent grounds whether or not there might be any languageless creatures to whom we would want to attribute thought. One kind of example is considered by Davidson himself – a dog, say, who chases a squirrel up a tree and then stays by that tree barking. Wouldn’t a natural explanation of this behavior involve attributing to the dog a desire to catch the squirrel, and a belief that the squirrel is in the tree? Davidson apparently does not find this kind of example convincing, although he offers nothing against it except that the description may be wrongly anthropomorphic. Let us move on, then, and consider some members of the corvid family. A study reported in 2002 that New Caledonian crows are able to bend wires to make hooks in order to retrieve food from a tube. And more recently there is even more striking evidence of reason. In one study, four captive rooks quickly learned to drop stones into a narrow container of water, in order to raise the level and bring a tasty worm floating on the surface to within reach of their beaks. (Orangutans have apparently shown similar intelligent behaviors.) And in another, New Caledonian crows showed they were able to first use a small stick to drag a medium-sized stick out of a tube, then use that stick to drag an even longer stick out of another tube, in order finally to use the longest stick to obtain food from yet another tube. (See Bower (2009) for a summary of these results, plus references to the original articles.) The idea that these crows are accomplishing all of this without a belief or desire in their heads is mind-boggling. We should also consider humans. On Davidson’s view, apparently, prelinguistic toddlers are not in a position to have propositional attitudes. And there are also cases like that of Genie – an abused child who was essentially prevented from hearing or producing speech until she was discovered, in 1970 at the age of 12. Genie never acquired English, although she did learn a number

Attitudes Toward Quotation

43

of English words subsequent to her release, and was able to communicate with them, enough to inform caretakers, e.g., that her father had punished her for making noise. (See Curtiss 1977, Rymer 1992.) Nevertheless it would seem that on Davidson’s view we would have to conclude that Genie did not (and does not) have memories, given her lack of a language. This is not plausible. Examples like these give us reason to believe that Davidson’s conclusion – that to hold a propositional attitude one must have a language – is not correct. So let us look at his argument, to see where the weakness might be. As noted above, this argument is rather contorted – John Searle describes it as “not as clear as it could be” (2002: 67). It seems to have roughly the structure given below (cf. Davidson 1975: 167–170). (15)

a. b. c.

To have a belief requires recognizing the difference between true and false beliefs. Recognizing the difference between true and false beliefs requires membership in a speech community. Ergo, to have a belief requires membership in a speech community.

It is the second premise of this argument which is weak. This premise is vaguely reminiscent of Wittgenstein’s argument against the possibility of a private language (cf. Wittgenstein 1953), as elucidated by Kripke (1982). The idea seems to be that the only way of knowing one has a false belief is through comparison with, or correction by, other members of one’s speech community: “the contrast between truth and error...can emerge only in the context of interpretation, which alone forces us to the idea of an objective, public truth” (Davidson 1975: 170). Searle’s very sensible response is that awareness of the falsity of a belief, or the unfulfilled nature of a desire, does not require membership in a speech community. So, consider again the dog who is barking up the tree after a squirrel. Soon, we may imagine, she notices (sees and smells) the squirrel running down the tree and into the neighbor’s yard. She stops barking up the tree and chases the squirrel into the neighbor’s yard. No speech community is required for the dog to change her beliefs – only feedback from the world in the form of perceptions which conflict with those beliefs (cf. Searle 2002: 67–69). 4.3. Alternatives to quotation An anonymous reviewer of this paper has pointed out that I haven’t said anything about alternatives to the quotational analysis of propositional attitudes. For my argument to hold water I need to show that there are alternatives that do, indeed, tell us something about what propositional attitudes are attitudes

44

Barbara Abbott

toward. Following Frege (and Mill), the major alternative to quotational analyses holds that the objects of propositional attitudes are propositions. However there are some differences of opinion about what propositions are. We have at least three live possibilities these days. One is that propositions are sets of possible worlds. Another is that they are sets of situations, where situations can be as big as possible worlds are usually conceived of as being, but can also be much smaller – as small as an entity possessing a property. A third possibility is that at least some propositions are structured objects which contain entities, and properties or relations holding among them. Any of these possibilities for propositions, I would like to argue, tells us more about what propositional attitudes are than the quotational analysis does. That is because on any of these views propositions are constituted by objects with which any sentient being can be in at least partial acquaintance. That is, the raw materials of propositions are states of affairs, or situations, or entities, properties, and relations, and these are things which one can make cognitive contact with without any necessity of a language. Of course it is possible, as the reviewer suggests, for a supporter of the quotational view to reply by pointing out that we may not actually have access to our propositional attitudes. And indeed that may be true, but the fact remains that our minds create those attitudes, and it is much more plausible to assume that they do it using elements which they are acquainted with, than that they do it using elements which they are not acquainted with. 5.

Concluding remarks

The main point of this paper has been to argue (a) that quotational analyses of propositional attitudes of the kind proposed by Quine and Davidson will only tell us something about the attitudes if languageless creatures don’t have them, and (b) that there is good reason to think that languageless creatures do indeed have propositional attitudes. If these results stand up they give us a good reason to abandon quotational analyses of propositional attitudes. And that in turn, I think, can tell us something about quotation itself: if it is not at the bottom of attributions of propositional attitudes, it becomes a less important, and less opaque, phenomenon. Let me try to spell out this point in a little more detail. The relative dates of two of the papers I’ve cited above – 1969 for “On saying that” and 1979 for “Quotation” – suggest that Davidson was led to his theory of quotation because of his quotational analysis of propositional attitudes. That is, it seems that Davidson first had the idea of reanalyzing the complementizer that in sentences

Attitudes Toward Quotation

45

about propositional attitudes as the demonstrative pronoun that. This allowed him to reformulate Quine’s quotational analysis of the attitudes in a way that obviates reference to a language, by having instead an utterance to refer to. And from there, it seems, Davidson was led to see quotation marks themselves as a different form of demonstrative pronoun, and to propose the demonstrative analysis of quotation in general. However, if Davidson’s quotational analysis of the attitudes is problematic, as I’ve tried to suggest, then that fact would remove his original motivation for the demonstrative analysis of quotation. And without that motivation there is very little left to support the demonstrative theory. Paul Saka (2006) has argued in detail against the demonstrative analysis of quotation, and though there are aspects of his paper that I disagree with, it is nevertheless very convincing in its refutation of Davidson’s theory. So one positive outcome, if the considerations in this paper are sustained, is an increase in clarity about quotation – it is not demonstrative in nature. And there is another positive outcome, which is that we do not require a theory of quotation to account for propositional attitude sentences. Without that additional burden, there is less that a theory of quotation has to do. References Bower, Bruce 2009 Tool use to crow about. Science News 176 (5): 5. Church, Alonzo 1950 On Carnap’s analysis of statements of assertion and belief. Analysis 10 (5): 97–99. Curtiss, Susan 1977 Genie: A Psycholinguistic Study of a Modern-day “Wild Child”. New York: Academic Press. Davidson, Donald 1969 On saying that. In Words and Objections: Essays on the Work of W.V. Quine, D. Davidson and J. Hintikka (eds.), 158–174. Dordrecht: Reidel. References are to the reprint in Davidson (1984), 93–108. Davidson, Donald 1975 Thought and talk. In Mind and Language, S. Guttenplan (ed.), 7– 23. Oxford: Oxford University Press. References are to the reprint in Davidson (1984), 155–170. Davidson, Donald 1979 Quotation. In Theory and Decision 11, Hal Berghel (ed.), 27–40. References are to the reprint in Davidson (1984), 79–92.

46

Barbara Abbott

Davidson, Donald 1984 Inquiries into Truth & Interpretation. Oxford: Clarendon Press. Fodor, Jerry A. 1975 The Language of Thought. Cambridge, MA: Harvard University Press. Fodor, Jerry A. 1987 Psychosemantics. Cambridge, MA: MIT Press. Frege, Gottlob 1892 On sense and reference. In Translations from the Philosophical Writings of Gottlob Frege, P. Geach and M. Black (eds.), 42–55. Oxford: Blackwell. Kripke, Saul 1972 Naming and necessity. In Semantics of Natural Language, D. Davidson and G. Harman (eds.), 253–355 and 763–769. Dordrecht: Reidel. (Reissued as a monograph, 1980, Cambridge, MA: Harvard University Press.) Kripke, Saul 1979 A puzzle about belief. In Meaning and Use, Avishai Margalit (ed.), 139–183. Dordrecht: D. Reidel. Kripke, Saul 1982 Wittgenstein on Rules and Private Language. Cambridge, MA: Harvard University Press. Mill, John Stuart 1843 A System of Logic, Ratiocinative and Inductive, Being a Connected View of the Principles of Evidence, and the Methods of Scientific Investigation. London: John W. Parker. Quine, W.V. 1956 Quantifiers and propositional attitudes. Journal of Philosophy 53: 177–187. Quine, W.V. 1960 Word and Object. Cambridge, MA: MIT Press. Rymer, Russ 1992 A silent childhood. The New Yorker, April 13 & 20. Saka, Paul 2006 The demonstrative and identity theories of quotation. Journal of Philosophy 103: 452–471. Searle, John R. 2002 Consciousness and Language. Cambridge: Cambridge University Press. Wittgenstein, Ludwig 1953 Philosophical Investigations. Trans. by G.E.M. Anscombe. Oxford: Basil Blackwell.

Truth Value Talk Without Quotation Manuel Bremer

1.

Semantic predicates, semantic operators

A side issue in the general philosophy of language and linguistics, but a central issue in formal semantics and meta-logic has been whether in case a language is able to express semantic properties (like truth) at all, these should be expressed by predicates or by operators. If a semantic property is expressed by a predicate the expression (typically an eternal sentence) having this property has to be quoted. If a semantic property is expressed by an operator which modifies a sentence to yield another sentence no quotation is needed. Unfortunately there are problems in the way of naively extending a language with semantic expressions. According to Tarski’s Theorem a language containing the expressive resources of arithmetic cannot – on pains of inconsistency – also contain its own truth predicate. According to theorems due to Montague and Thomason even languages below the expressive power of arithmetic which use truth operators will be inconsistent, once the truth operator obeys some simple axioms characteristic of truth (value) talk. These theorems seem to exclude a language able to deal with its own semantics, a truly semantic universal language. Natural languages, however, seem to be able to deal with their own semantics. And they seem to yield some antinomies like the Liar by way of their expressive power. The scenery changes once paraconsistent logics are used as tools in modelling semantics. In as much as antinomies/contradictions are allowed within paraconsistent semantics and logics one may reconsider the use of truth predicates and truth operators in one’s formal language and modelling. In relation to the other papers in this volume it might be helpful to be precise on the point to be established. The argument does not try to show that all quotation marks can be dropped in a language: talking about the syntactic properties of a word or sentence (like: “morpheme” is disyllabic) requires quotation devices (i.e. quotation marks or Gödel numbers). The argument does not even show that all semantic quotation marks can be dropped. The focus is on quotation in relation to assignments of “true”, “false” and other semantic values of

48

Manuel Bremer

sentences/statements. Given a Davidsonian approach to meaning and reference an account of truth provides a theory of the core semantic concepts, so having a quotation free theory of truths goes a long way towards a quotation free semantics. Nevertheless there may still be a need for quotation is semantics, especially if sub-sentential units (like proper names or predicates) are described semantically. The point of the whole exercise here in the paper is to see clearer where we need quotation marks and where we do not, which kind of logic allows for dropping some quotation marks and which do not.1 2.

Dialetheism

Dialetheism is the claim that some contradictions are true. For anyone trained in standard logic and raised in the belief that already in antiquity Aristotle settled once and for all that there is the Law of Non-Contradiction dialetheism sounds not just false, but bizarre. On the other hand people contradict each other quite often and a couple of theories have turned out to be inconsistent. Nevertheless the people who held inconsistent beliefs have not (at the time of holding these beliefs) believed just anything, as the standard rule of ex contradictione quodlibet would have it. Thus paraconsistent logics (logics that invalidate ex contradictione quodlibet and thus can tolerate even provable contradictions) have gained interest and lots of them are investigated and explored nowadays (cf. Bremer 2005). Dialetheism is strong paraconsistency in the sense that one cannot just tolerate some contradictions, but one should endorse some of them. This certainly needs argument. Ever since its arrival dialetheism has been met with the proverbial incredulous stare, not only because of the inconsistent ontology of Routley’s noneism (Routley 1980), but also with respect to the dialetheist’s claim that one can knowingly believe and assert contradictions. Priest in the paper introducing his “logic of paradox” LP (Priest 1979) admits that the thesis of dialetheism is a dialetheia itself. In his book In Contradiction (Priest 1987) he argues that one can avoid dialetheism being a dialetheia itself if one is prepared to give up contraposition for the conditional in Convention (T). The problem about contraposition is the following. Given Convention (T), say – for a start – with a truth predicate and some quotation device, we have for the Liar statement  ()

 is false.

1. I have to thank an anonymous referee for requesting these necessary clarifications.

Truth Value Talk Without Quotation

(T)

49

True ≡ 

The dialetheist need not endorse (TF)

¬True˛ ≡ False˛

But the dialetheist endorses for some ˛: (D)

True˛ ∧ True¬˛

Now, for a (first naïve) dialetheist  is true and false, thus, because  is false, ¬ is true, and false, because  is true. Given that  is true, we have, by Convention (T), that ¬. Now contraposition with respect to (T) gives us (T’)

¬ ≡ ¬True

But as the (first naïve) dialetheist endorses both  and ¬ we can detach both in (T) and in (T’) and get (’)

True ∧ ¬True

Thus the statement of the dialetheist that some contradiction is true turns out to be a contradiction itself! That may be much more than the dialetheist bargained for. Priest thus endorses giving up contraposition in Convention (T). Another option (taken up in so-called Adaptive Logics) is to restrict both Modus Ponens and Modus Tollens to detachment with consistent sentences (cf. Batens 1989, Priest 1991). Nevertheless Priest defends that one can believe and assert contradictions. Up to now criticism of dialetheism has focused on the problems what the status of dialetheism itself is and how it may be possible to believe knowingly contradictions. It is argued here that within dialetheism the resources are available to claim that dialetheism is true only (i.e. not false at the same time). Furthermore there may be occasions on which it is rational to believe and/or even assert contradictions, without thereby positioning oneself on a slippery slope towards an attitude of “anything goes”. Paraconsistent logics can level the distinction between object and metalanguage. A semantically closed language not only is able to talk about its own expressions, but does contain at the same time its semantic expressions. These semantic expressions need not be taken as predicates (like a truth predicate applying to the quotation of a sentence), but can be taken as operators instead. One arrives at a paraconsistent language/logic which allows truth value talk without previously quoting the sentences which are evaluated.

50

Manuel Bremer

3.

Universality and contradictions

The main motivation for dialetheism is universality as a feature of language and cognition. Universality means that we are aiming at – and supposedly capable of – a theory of, say, language in general, that is not just of this or that language or languages of this or that formal structure. And this theory is expressed in language, so that at least some language can be its own meta-language (with respect to all interesting properties of that language, semantics included). Universality means as well that we use fundamental concepts like denotation or true unrestrictedly. A language L is semantically closed if and only if L is syntactically closed and able to talk about its own semantics. The meanings of the terms of L can be given within L then. If a language is semantically closed it can not only talk about its own expressions (by suitable names or quotation marks), but it can also apply semantic properties to these terms, and even to the terms that express semantic properties: (1) (2) (3)

Sentence number (1) is grammatical. This sentence contains six meaningful words. The third sentence displayed in this list is true.

Here some formula refers to itself by a description/name occurring in that formula. An infamous example is the Liar: ()

 is false.

The Liar is a fixed point for the predicate “( ) is false” (or “not-true”), saying “I am false”. Now consider (): If  is true, then  is false, because the general term in  “( ) is false” should apply to the singular term “”. If  is false, then  is true, because  is just saying that it is false. So we get: (4)

True ≡ False

or given that we have a two-valued logic where “false” is just the opposite of “true”: (5)

True ≡ ¬True

So  is an antinomy. There are more harmless semantically self-referential sentences, like (2). The Liar is the basic case of bad semantic self-referentiality. It bears its badness on its sleeve. It is not hidden. Within a language that is seman-

Truth Value Talk Without Quotation

51

tically closed semantic self-reference may be hidden, however. Suppose today you only utter a single statement (6)

What the pope declares today is true.

and as a contingent matter of fact he only says that day: (7)

Everything N.N. says today is false.

Dialetheism claims that it cannot be prevented, since a natural language is a semantically closed language. Since, furthermore, the antinomy can be proved, it has to be true. So the dialetheist has to show three things: 1. The contradictions can be proven in a sound non-standard logic, if we use a semantically closed language. 2. We have to use a semantically closed language. 3. There is no satisfactory alternative to accepting the antinomies (i.e. the attempts to prevent them either fail or have consequences worse than dialetheism). I will skip the arguments for these three claims here (for the details cf. Bremer 2005, or Priest 1987), and focus rather on the dialetheist’s method of talking about truth and antinomies. Priest shows in all formal detail how a (paraconsistent) language can contain its own truth predicate (cf. Priest 2006a: 125–40). He also shows in more recent work (cf. Priest 1997, 1999) that a (paraconsistent) language can contain its own denotation function (as applicable, say, to proper names or descriptions). Thus, adding this to what is to come in the next paragraphs, even sub-sentential semantic quotation can be circumvented to that extent by using a paraconsistent logic! 4.

Truth operators

One may think that since naive semantics contains Convention (T) (T)

T ≡ 

and negation can have the standard truth table in many paraconsistent logics, a contradiction like () is said to be both true and false (non-true), and the negation of () is said to be both true and false (non-true). With contraposition and Convention (T) (T’)

¬ ≡ ¬T

52

Manuel Bremer

one arrives – as we have seen in § 1 – at the embarrassing result that T is itself both true and false. One may take this as a reductio ad absurdum of the dialetheist position. Talk of truth and unique commitment have to be reconsidered, therefore. Truth concerns what is the case whether we believe it or not. Belief concerns what we are willing to include in our inferring. What we believe we take into account in our reasoning (belief is cognitive). Generally, being provided with reasons for ˛ is seen as the basis for believing ˛, given that the reasons for some  incompatible with ˛ are not stronger. On a gullible approach to (perceptual) belief one believes every ˛ one has no reasons against. The best backing for a belief ˛ is a proof of ˛. Having reasons is superior to mere belief in the truth of ˛. Having no independent access to the (ultimate) truth of ˛ going with reasons is the rational way, whatever the (ultimate) truth value of ˛ is or turns out to be. Typically it is taken to be rational to assent to [to affirm] what one believes. Assertion is to assent to or to affirm what one believes. If one has a belief ˛ one also has the disposition to assert ˛. One does not need additional reasons to proceed from believing to asserting. On the other hand, asserting ˛ is done by a speaker confronting an audience (assertion is pragmatic). Asserting ˛ is done with a purpose in view of an audience, so that this purpose exceeds using ˛ in one’s processes of deliberation. As an (speech) act with some purpose asserting ˛ has to meet the basic felicity conditions of successful action plans, like the purpose being not achieved anyhow without my action and this specific action being fit to the purpose. Asserting contradictions seems to fail both conditions. Dialetheism as a thesis should be asserted as being only/just true (i.e. not being false at the same time). One should be able to say, without saying something false, that a true sentence/statement is true. One should be able to express the semantic properties of all sentences/statements (including the antinomies). Dialetheism claims that some contradictions are true. So we have some sentence  with , ¬, T, T¬, F, F¬ to start with. The reasons for this are that these contradictions are provable given some unassailable principles and structures in a semantically closed language. Now, these antinomies being true and being justified as true, by proving them, give all the reasons to believe that they are true and thus to believe them (themselves). So a dialetheist should believe (8) (9) (10)

The Liar is true. thus The Liar thus The Liar is false.

Truth Value Talk Without Quotation

53

Giving up believing what one has proven seems to be a desperate and ad hoc manoeuvre. So a dialetheist has inconsistent beliefs. She reasons using both T and F if necessary. Now, paraconsistent logics can level the distinction between object and meta-language. A semantically closed language not only is able to talk about its own expressions, but does contain at the same time its semantic expressions. These semantic expressions need not be taken as predicates (like a truth predicate applying to the quotation of a sentence), but can be taken as operators instead. One arrives at a paraconsistent language/logic which allows truth value talk without previously quoting the sentences which are evaluated. Let us consider six such operators: “˛” says that ˛ is true only, “∇˛” that ˛ is false only, “ ◦˛” says that ˛ is consistent (i.e. has only one truth value), “ •˛” says that ˛ is contradictory. We can then say – and these being just true – that the Liar is true, false, not simply true, not consistent, and so on. “T” and “F” are now understood as operators applying to formulas/sentences not quoted. These operators are bivalent themselves and are defined according to the table (‘0,1’ as truth value means that ˛ is an antinomy): ˛

¬˛

T˛

F˛

˛

∇˛

◦˛

•˛

0

1

0

1

0

1

1

0

1

0

1

0

1

0

1

0

0,1

0,1

1

1

0

0

0

1

Using these operators dialetheism can fulfil the traditional condition on any decent theory: that it claims to be just true (and not only as true as its negation). We see firstly: Dialetheism is thus no form of trivialism (that everything is true). The trivialist proposes (∀˛)(T˛ ∧ T¬˛) or (∀˛)(T˛ ∧ F˛). The dialetheist claims (∃˛)(T˛ ∧ F˛), but also (∃˛)(T˛ ∧ ¬T¬˛), and (∃˛)∇˛. And given some formal system some formulas can be exhibited having these properties (e.g., defining a bottom particle ⊥ with ∇⊥ being valid). can be defined as the top particle with T(˛ ∨ ¬˛), being true only. The bottom particle ⊥ can be defined as ∇(˛ ∨ ¬˛), being false only.2 To have and use the (T)-scheme at the same time as these operators we need some revisions in the logic of the conditional, like giving up on the unrestricted validity of Contraposition. The language contains formula (e.g. T˛ ∧ ∇˛) that 2. Note that, in contrast to even the intuitionist negation rules, ⊥ ≡ (˛ ∧ ¬˛) need not hold if ˛ is a dialetheia, since then T(˛ ∧ ¬˛), and ∇ is incompatible with T.

54

Manuel Bremer

can be evaluated only as being simply false. These formulas, of course, cannot be derived.3 To return to the semantic use of the truth operators: Saying T is thus simply true: T. This does not exclude that F is also simply true: F. Now it seems that saying of the Liar that the Liar is false is just what the Liar is saying: (11)

F ≡ 

Then we might have: (12)

FF

and this contradicts F! But to derive (12) we use either (13) (14)

F =  (F ≡ )

or

The equivalence thesis (14) may be wrong. And substitution of identicals is one of those inferences restricted to consistent objects (to which  does not belong). Even if (14) is not wrong deriving FF supposedly has to use some form of detachment, which again is restricted to consistent sentences (to which  does not belong).

3. We do not need the details of all these restrictions here. The reader has only to know the general idea of paraconsistent logics and the idea of “adaptive logics” (Batens 1989, 2000) to restrict some rules to consistent sentences (respectively to retract some supposed consequences if the rules to derive them employed, against the restrictions, some inconsistent sentences). A paraconsistent logic like Priest’s LP can be developed into an adaptive logic with a restricted form of Modus Ponens and Modus Tollens (Priest 1991). Within paraconsistent logics “logics of formal inconsistency” (Marcos 2005) employ consistency operators in the object language. Truth operators can then be added. Blending these approaches one can have an adaptive paraconsistent logic which combines the extensional and intuitive truth conditions of LP with the use of truth and consistency operators and restrictions on substituting identicals to consistent objects (cf. Bremer 2005: 224–37). Substituting identicals has to be restricted to avoid so-called hyper-contradictions (cf. Bremer 2005: 185– 97). As adaptive logics restrict detachment to consistent sentences (a special sort of consistent object) the adaptive use of substituting identicals restricts substituting to objects which have not contradictory properties. As a consequence of these restrictions hyper-contradictions no longer go through, as Strengthened Liar reasoning in these languages depends on either detaching or substituting with the antinomic sentences or their names.

Truth Value Talk Without Quotation

55

Let us take it, thus, that F can be believed and – being bivalent – can be asserted. Asserting T or F certainly fulfils some purpose, be it in explaining dialetheism or in arguing with opponents of dialetheism. Asserting ˛ [A˛] means claiming that ˛ is true. One may thus have: (15)

A˛ ≡ AT˛

Denying ˛ [D˛] cannot be understood by a dialetheist as affirming ∇˛ [A∇˛]. Denying ˛ would thus be incompatible with affirming ˛ (i.e. affirming T˛), as we supposedly have: (16)

A(˛ ∧ ) ≡ A˛ ∧ A

Thus (17)

AT˛ ∧ A∇˛ ≡ A(T˛ ∧ ∇˛)

which is even too much for a dialetheist, as T˛ and ∇˛ are incompatible. One needs a distinction then between affirming ∇˛ and affirming F˛ [T¬˛]. Let us take affirming F˛ as denial and affirming ∇˛ as rejection of ˛ [R˛]. A dialetheist may then both assert and deny a sentence ˛: (18)

A˛ ∧ D˛

As this is equivalent to (19)

AT˛ ∧ AF˛

[i.e. A(T˛ ∧ F˛)]

Whereas there are situations in which a dialetheist accepts both ˛ and ¬˛, or T˛ and F˛, there are no situations in which a dialetheist accepts and rejects ˛ at the same time. As the foregoing distinction shows there is, furthermore, one kind of contradiction that (even) a dialetheist cannot support: (20)

¬ (A˛ ∧ R˛) [i.e. ¬(AT˛ ∧ A∇˛),

[i.e. ¬A(T˛ ∧ ∇˛)]

since T˛ and ∇˛ are semantically incompatible. Another simple point is that no-one (including the dialetheist) can have pragmatic contradictions: Speech acts being bodily movements that either occur or do not, there is no pragmatic parallel to having it both ways, i.e. (21)

¬ (A˛ ∧ ¬A˛)

This instance of the accepted tautology ¬(˛ ∧ ¬˛) expresses not only a semantic exclusion the dialetheist accepts (and sometimes nevertheless supersedes), but

56

Manuel Bremer

the absence of the mysterious feat of asserting something and not doing it at the same time.4 Thus without sliding into mystery or being silenced one can be a dialetheist and claim some crucial antinomies to be true. Dialetheism itself is not a paradoxical statement, but the theory that fits the aspirations of a universally minded philosophy. A crucial ingredient is to drop the object-/meta-language distinction and use (bivalent) semantic operators. References Batens, Diderik 1989

Batens, Diderik 2000

Dynamic dialectical logics. In Paraconsistent Logic, G. Priest, R. Routley and J. Norman (eds.), 187–217. München: Philosophia Verlag. A survey of inconsistency-adaptive logics. In Frontiers of Paraconsistent Logic, D. Batens et al. (eds.), 49–73. Baldock: Research Study Press.

Bremer, Manuel 2005 An Introduction to Paraconsistent Logics. Bern et al.: Lang Publishers. Bremer, Manuel 2008 Conceptual Atomism and Justificationist Semantics. Bern et al.: Lang Publishers. Marcos, Joao 2005

Logics of Formal Inconsistency. PhD Thesis, Departamento de Filosofia do Instituto de Filosofia e Ciencias Humanas da Universidade Estadual de Campinas.

Priest, Graham 1979

The logic of paradox. Journal of Philosophical Logic 8: 219–241.

Priest, Graham 1987

In Contradiction. Dordrecht: Martinus Nijhof.

4. One may also consider introducing Rejection as a speech act sui generis (i.e. as not being defined by Assertion and truth operators). One may then have a – arguably – more proper pragmatic theory of the different attitudes and the distinctions between asserting the opposite, withholding one’s opinion and rejection. An outline of these ideas can be found in: Bremer 2008: 79–90).

Truth Value Talk Without Quotation Priest, Graham 1991

Minimal inconsistent LP. Studia Logica 50: 321–331.

Priest, Graham 1996

Everett’s trilogy. Mind 105: 631–648.

Priest, Graham 1997 Priest, Graham 1999

57

On a paradox of Hilbert and Bernays. Journal of Philosophical Logic 26: 45–56. Semantic closure, descriptions and non-triviality. Journal of Philosophical Logic 28: 549–558.

Priest, Graham 2006a

Doubt Truth to Be a Liar. Oxford: Clarendon Press.

Priest, Graham 2006b

In Contradiction. Second Edition. Oxford: Clarendon Press.

Priest, Graham, J. Beall & Bradley Armour-Garb (eds.) 2004 The Law of Non-Contradiction. Oxford: Clarendon Press. Routley, Richard 1980 Exploring Meinong’s Jungle and Beyond. Canberra: Research School of Social Sciences, Australian National University.

Disquotational Indirect Reports in Focus Sarah-Jane Conrad

1.

Reports and a traditional framework

What systematic and methodological role do disquotational indirect reports (DIR) or saying reports 1 play for a theory of language and communication? Traditionally, an intimate relationship was taken to hold between the wording of a sentence, its semantic, i.e. truth-evaluable or propositional content and what can be meant by the sentence when uttered.2 The analysis of the concept of saying was designed to reveal that these three aspects are interlinked by applying the method of disquotational indirect reporting on what has been said when a specific sentence was uttered.3 Consequently, DIRs were considered to be eminently important from a systematic point of view not only for a semantic theory, but for a theory of communication, too, as they would provide access to the propositional content of a sentence uttered and speaker meaning. All that was necessary in order to identify the respective contents was to reproduce accurately the original wording of the utterance of a syntactically well-formed sentence. Hence, DIRs were taken to have great advantages from a methodological point of view, too, as they more or less automatically identified the relevant contents. However, the systematic relationship between the wording of a sentence, the proposition expressed by it and what can be meant in uttering the sentence has been questioned in recent years. Saying reports on one and the same sentence can vary considerably and still appear to be adequate. Not all of these variations can be traced back to an indexical or a demonstrative element in the sentence reported on and must therefore be explained differently. For example, reports on the sentence (1) “She is ready”, when uttered by Paul, can take on the following forms: 1. The notions “saying reports”, “reports on what is said” and “disquotational indirect reports” are used synonymously. 2. See Grice (1989: 22ff. & 117ff.). 3. See Soames (1989: 394). For a critical discussion of the possibility of relating all three aspects to what is said, see Harth (2003, 2005).

60

Sarah-Jane Conrad

(DIR 1) (DIR 1 ) (DIR 1

) (DIR 1

)

Paul said that a female person is ready. Paul said that Jane is ready. Paul said that Jane is ready for something. Paul said that Jane is ready for the theatre.

Substitution of the personal pronoun “she” (see DIR 1) by the name of the person referred to (see DIR 1 ) is a very common manoeuvre when reporting on a sentence uttered containing a demonstrative element. In addition, reports often transcend the original wording of the sentence uttered by specifying, for example, that Jane is ready for something (see DIR 1

), or even indicating what exactly Jane is ready for (see DIR 1

). DIRs on sentence (2) “I had breakfast” uttered by Jane and reported by Paul, can vary in a similar manner, as the following possible reports illustrate: (DIR 2) (DIR 2 ) (DIR 2

)

Jane said that she had breakfast. Jane said that she had breakfast at some time. Jane said that she had breakfast today.

Besides reporting the wording of the sentence accurately and substituting the personal pronoun “I” by the indirect reflexive “she” (see DIR 2 ), reports may specify that Jane had breakfast at some time (see DIR 2 ) or indicate the time when Jane had breakfast (see DIR 2

). All of these reports claim to refer to what a speaker said in uttering sentence (1) or sentence (2). Therefore, they raise the question if any conclusion can be drawn from the observable variability concerning the methodological and systematic status of DIRs. Is it still possible to defend the traditional view that reports on what is said must simply reproduce the wording of a sentence and that they thereby identify its semantic content and what a speaker meant in uttering it? Different answers have been given to these questions. We can mainly distinguish three different approaches. Each of them explains the varying saying reports differently and draws still other conclusions concerning the status of DIRs. In what follows, I will scrutinize all three lines of answers critically in order to evaluate the correct systematic and methodological status of DIRs for the theory of language and communication.4

4. In the following chapters, I explore the role of DIRs for a theory of a particular language in use. Therefore, I will ignore the problem of translation even though it surely deserves more attention when considering DIRs in a broader framework.

Disquotational Indirect Reports in Focus

2.

61

Variable DIRs or the problem of semantic underdeterminacy

The first approach presented here is influenced by the so-called contextualist’s stance.5 According to it, some DIRs show that the above-sketched view of a direct relationship between the wording of a sentence, the proposition expressed by it and its immediate relationship to speaker meaning is misconceived and modifications concerning the concept of saying are necessary. First, a simple report of the literal wording of a syntactically well-formed sentence often fails to determine a full-blown propositional content. With regard to sentence (1), for example, one must specify what Jane is ready for (see DIR 1

) in order to determine a truth-evaluable content. Hence, being syntactically well-formed, as example 2 surely is, does not automatically guarantee that the utterance of the sentence in question is semantically complete, even though all the relevant references have been fixed and the meaning of the expressions was disambiguated. The linguistic meaning of sentence (1) and many other sentences rather appears to be semantically underdeterminate (SU): Their wording fails to provide all the information needed for specifying a proposition. Instead we must refer to what a speaker meant – which is, by definition, truth-evaluable – in uttering the sentence in order to determine the proposition expressed. Accordingly, speaker meaning appears as an additional and yet indispensable criterion of adequacy for correct DIRs, besides the wording of the sentence; at least as long as one assumes that saying needs to be propositional. (DIR 2) certainly specifies a propositional content, but not the content Paul probably wanted to express when uttering the sentence (2). Only (DIR 2

) seems to identify the proposition he actually had in mind. (DIR 1

) and (DIR 2

) show, first, that the linguistic meaning is in some cases simply too weak to determine any propositional content of a sentence uttered; and second, that it is no warrant for identifying speaker meaning properly. The traditional view therefore errs twice: First, when interlinking the concept of linguistic meaning, propositions and speaker meaning directly in the wording of a sentence and, second, when believing that faithfully representing the wording of the sentence together with knowledge of the relevant references is sufficient for an adequate DIR. Apparently, it is no longer possible to anchor the concept of saying directly and exclusively in the wording of the sentence. The variability of DIRs and their respective complements – often referred to under the label of explicatures 6 – rather suggest that speaker meaning provides 5. Robyn Carston (2002), François Recanati (2004, 2007) and Dan Sperber and Deirdre Wilson (1986) represent this view. 6. See Carston (2002: 116ff.).

62

Sarah-Jane Conrad

an extra point of reference against the background of which the meaning of a sentence has to be evaluated in order to determine the propositional content or what is said.7 Hence, speaker meaning is an additional criterion of adequacy for formulating accurate DIRs, independent from the wording of a sentence and its linguistic meaning – either because the wording of a sentence simply fails to specify a propositional content at all, or because it does not determine the proposition the speaker had in mind, which has to be identified, however, for a correct understanding of the sentence. Propositions can therefore no longer be considered to be the sole matter of semantics, but equally belong to the realm of pragmatics, as speaker meaning as well as linguistic meaning contribute to the determination of their content. Of course, the systematic and methodological status of DIRs is directly affected by this criticism, without, however, questioning saying reports as a useful tool altogether. Yet, no direct relationship between the wording of a sentence, what is meant and truth is taken to be discernable on mere syntactic and semantic grounds and the respective connections must therefore be interpreted differently. Formulating a correct DIR seems to involve necessarily reference to speaker meaning and thus we need to bring in some highly context-sensitive and throughout pragmatic information. At once, DIRs have changed their systematic role and they need not specify a mere semantic, but rather semanticpragmatic information, composed of the linguistic meaning of the sentence, the relevant references and what was meant by the sentence uttered.8 Conse7. Some theorists like Emma Borg (2004a, 2004b) hold that a semantic theory must not deliver truth-evaluable contents, but truth-conditions only. Consequently, the concept of saying must be considered within the framework of a truth-conditional semantics. Sentence (1) obviously fails to fully specify the truth-conditions and accordingly, an adequate report on what is said must introduce some additional information; reproducing the original wording of the sentence will not do. However, it is not necessary to refer to speaker meaning but only to the syntax of the word “ready” in order to determine the truth-conditions: after having made transparent the transitive character of the adjective (see DIR 1

), the truth-conditions will be fully specified and the sentence used can be taken to express a propositional content, presumably quite trivially true. 8. This conclusion is not approved without exception and two different reactions may be mentioned for the sake of completeness. Some say that even if one considers the problem of SU to be genuine and believes that a theory of semantics and saying has to operate with truth-evaluable contents, the variability of DIRs can still be explained as a mere linguistic phenomenon. They can be taken to show that our ordinary conception regarding the syntax or the logical form of numerous sentences is misguided and many words take on additional arguments which have to be complemented rel-

Disquotational Indirect Reports in Focus

63

quently, the procedure of specifying a correct DIR turns out to be a much more challenging task than the traditional view held. While defending their criticism against the traditional picture, contextualists must presume that specific types of DIRs accurately report on what a speaker said. However, this assumption turns out to be problematic for two reasons: a first problem is related to the contextualists’ claim that linguistic meaning is sometimes too weak to determine propositional content, while in other cases it fails to identify the speaker’s meaning properly; the second problem is connected with the pragmatic consequences of the approach suggested. To begin with the first problem, we may ask how to distinguish cases where an amendment is necessary on conceptual grounds in order to guarantee a propositional content from cases where speaker meaning merely determines the proposition necessary for a correct understanding of a sentence. DIRs do not allow distinguishing these differently motivated amendments. It becomes difficult to evaluate the effectiveness of linguistic meaning properly, because sentence meaning will in any case be judged against the background of speaker meaning.9 The exact contribution of linguistic meaning can no longer be interpreted independently. Now, one may agree that the suggested inclusion of speaker meaning seems very plausible for a theory of communication, but may still want to question if it is really necessary for a theory of language and semantics. All the more so as the thesis of semantic underdeterminacy can be and has been questioned with regard to sentence (1) and many other sentences, too.10 There is no general agreement on the question, which sentences actually suffer from the alleged weakness. Thus, the fact that a specific sentence must in any case be evaluated against the background of speaker meaning perhaps merely rests on some theoretical predilection and less so on substantial reasons as many exponents of the

ative to the context of utterance, as the so-called Indexicalism suggests; see Stanley (2000). Others take the variability of DIRs to show that our traditional semantic theory has to be elaborated as Semantic Relativism suggests; see e.g. Predelli (2005) or Recanati (2007). 9. Recanati (2004: 61ff.) assumes two distinct mechanisms for completing and expanding the literal meaning. Thus, he is able to keep the two forms of amendments apart on conceptual grounds. However, the troubling question concerning the role of speaker meaning still remains. 10. Cappelen and Lepore (2005) surely have defended this view, even though some of their remarks on that matter seem to be more reluctant; see Cappelen (2008: 269).

64

Sarah-Jane Conrad

so called Minimalism have objected while trying to restore the force of semantics.11 In addition, Contextualism itself is not a very stable position and while some believe that only a limited number of sentences is genuinely underdeterminate, others claim that the phenomenon is pervasive and affects all sentences of a language.12 Consequently, eventual differences between sentences (1) and (2) lapse and truth-conditions are taken to be totally entrenched within the field of pragmatics. If we want to settle the matter and examine to what extent the problem of semantic underdeterminacy affects a theory of language, it will be difficult to argue for either position, as the effectiveness of linguistic meaning can no longer be examined independently. DIRs or an appropriate surrogate appear to be desperately wanted for defending any semantic claim at all. In the light of all of these troubles contextualists encounter with regard to their semantic assumptions either from an internal or an external perspective, one is tempted to blame them for abusing variable DIRs when adapting the concept of saying to their theoretical predilections. The worry that DIRs merely show the theoretical assumptions the position is operating with is corroborated. The second problem results from the fact that speaker meaning is taken to be an indispensable and independent ingredient of the concept of saying and from the respective consequences that follow from this assumption for a theory of communication and pragmatics. If speaker meaning must necessarily be considered in order to formulate a correct DIR, one has reasons to doubt if it is still possible to distinguish what a speaker said from what a speaker merely meant in the sense of a Gricean conversational implicature.13 Both contents obviously have to be specified by referring to speaker meaning, and thus the concept of saying as well as the concept of conversationally implicating depend on the very same concept. Will it then still be possible to keep them apart? Consider the following conversational episode between (A) and (B) as they are getting ready to leave for the theatre together with Jane: (A) (B)

“What about Jane?” “Jane is ready” (Meaning thereby that Jane is ready for the theatre). (B’s implicature: Jane is getting impatient).

11. See, again, Cappelen and Lepore, Emma Borg already mentioned above, or Kent Bach, to whom I will return below. Minimalists typically hold that what is said is determinable on mere semantic grounds. 12. See Cappelen and Lepore (2005) for discussion. For a radical interpretation of the SU-thesis, see e.g. Travis (1985) or Bezuidenhout (2002). For a moderate version, see e.g. Carston (2002). 13. See Grice (1989: 26).

Disquotational Indirect Reports in Focus

65

(B) says to (A) that Jane is ready, meaning thereby that she is ready for the theatre. In uttering sentence (1), (B) wants to conversationally implicate that Jane is getting impatient and wants to leave as soon as possible. The amendments to the wording of the original utterance, namely that Jane is ready for the theatre, as well as the content of the conversational implicature depend both on correctly identifying what the speaker meant in uttering sentence (1). The above representation of the dialogue captures the two contents separately. Yet, as they both depend on speaker meaning, one may wonder why we cannot merge the two contents together and interpret them as two amendments on par to what the speaker literally expressed and thus siding them with what the speaker said: (A) (B)

What about Jane? Jane is (getting impatient as she is) ready (for the theatre).

An adequate representation of what (B) said would therefore be: (DIR 1

)

Paul said that Jane is getting impatient, as she is ready for the theatre.

We have no reason to privilege the first representation (B) over (B ) the second on mere conceptual grounds. In addition, most contextualists generally hold that speaker meaning and linguistic meaning are necessarily not symmetrical with regard to their respective contents and thus the question whether some specific formulation adequately represents what has been meant can always be disputed.14 To this we may add that any completion considered adequate with regard to what is said will appear to be inadequate with regard to the conversational implicatures and vice versa. In other words: because of the double role speaker meaning has to play in determining the two contents, they risk annulling each other mutually. Instead of progressing from one content to the other, we find ourselves trapped between two likewise adequate and inadequate specifications of what has been meant.15 It is therefore highly questionable if the concept of speaker 14. The principal asymmetry between speaker meaning and linguistic meaning and the rejection of the principle of expressibility following from this assumption is one of the main reasons why linguistic meaning is considered to be ineffective for determining propositions: Linguistic meaning is considered to be too coarse-grained and many believe that it must be necessarily conceptualized in this way in order to bring it into service for all the different applications of a word; see e.g. Sperber and Wilson (1986) or Carston (2002: 69f.), but Montminy (2010). 15. Carston (1988: 158) believes that the framework of relevance-theory avoids this conclusion, but I do not see how she can argue for her point.

66

Sarah-Jane Conrad

meaning is effective enough to individuate the two pragmatically determined contents separately and if the conceptual distinction between what is said and what is conversationally implicated can be defended.16 In fact, we risk blurring important theoretical distinctions. If this problem cannot be resolved, it will have a great impact on the systematic status of DIRs: Instead of specifying what a speaker said, DIRs might as well be taken to specify what she has conversationally implicated. Their acclaimed role for defending a specific theoretical claim with regard to the concept of saying appears now totally wrecked. Thus recurring to speaker meaning turns out to have many troublesome consequences the relevant theoretical distinctions are at risk of becoming brittle. By arguing that speaker meaning must be taken as additional point of references when specifying a correct DIR, any claim concerning the effective impact of linguistic meaning on what is said becomes speculative and turns the main claim that speaker meaning actually is necessary speculative. This simply shows that DIRs only reflect those theoretical assumptions, which were introduced beforehand by definition, and their evidential role is thus more than questionable. 3.

Variable DIRs or saying goes pragmatic

According to the first approach, eventual amendments to the wording of a sentence are in many cases indispensable either for being able to associate a propositional content with the utterance of a sentence at all, or for specifying the proposition the speaker actually meant when uttering a sentence. Instead of being the automatic output of a syntactically well-formed sentence with references being fixed, speaker meaning appears at once as an indispensable corrective which together with the original wording of the sentence must be taken into account as an additional and independent criterion of adequacy for DIRs. Yet, if we consider our daily practice in using saying reports, one may reasonably doubt that a correct explanation of the variability of DIRs actually stops there.17 As a matter of fact, DIRs vary to a much greater extent than the first approach admits. For example, one may report Paul’s utterance of sentence (3) “At around 11 p.m., I put on a white shirt, a blue suit, dark socks and my brown Bruno Magli shoes. I then got into a waiting limousine and drove off into heavy traffic to the airport, where I just made my midnight flight to Chicago” as follows:18 16. See François Recanati (1989: 315ff.) and Owen Greenhall (2006) for a similar critique. See also Clark and Gerrig (1990). 17. This account has been put forward by Cappelen and Lepore (1997, 1998, and 2007). See also Clark and Gerring (1990). 18. Cappelen and Lepore (1997: 283) originally put forward this example.

Disquotational Indirect Reports in Focus

(DIR 3)

67

Paul said that he dressed around 11 p.m., went to the airport and took the midnight flight to Chicago.

In the light of these extreme deviances between the wording of the sentence originally uttered and the report given on it, one may come to believe that DIRs actually aim at something quite different than identifying the proposition associated with the literal utterance of a sentence. (DIR 3) certainly does not look as if it serves to identify the propositional content of the sentence. It rather resumes to what communicative purpose a specific utterance, i.e. either the sentence originally uttered of the saying report itself has been made for. Hence, an adequate DIR needn’t necessarily represent the semantic or semantic-pragmatic properties related to a sentence. Consequently, the original wording of the sentence plays a secondary role for the determination of an adequate DIR while a correct identification of what a speaker meant in uttering the sentence is more or less all that matters. Of course, different DIRs may be adequate in different contexts, depending on innumerable non-semantic aspects.19 The relevant standards for an adequate DIR will be established only within the communicative setting and are in many cases only accessible to the persons being involved in the conversation. Hence, the criteria of adequacy for formulating a correct DIR turn out to be of an entirely pragmatic nature while semantic aspects really play quite an insignificant role.20 In one sense, the second approach pushes the insights of the contextualist’s stance to its extremes: What a speaker meant or what she wanted to communicate when uttering a specific sentence is now regarded as the sole point of reference for formulating an adequate saying report. DIRs therefore report exclusively on pragmatically determined contents and have thus changed their systematic orientation completely. The concept of saying itself turns out to be pragmatic throughout.21 Indirect speech […] is a device reports use for characterizing acts (utterances) performed by speakers. […] For this reason, we should not expect semanticists to be able to exploit the standard usage of the English locution ‘says that’ in exacting features of utterances that interest semanticists.22

Such a conclusion seems to be confirmed by the fact that on the one hand a DIR may correctly report the propositional content of a sentence uttered and 19. See Cappelen and Lepore (1997: 291 & 282ff.). 20. See Cappelen and Lepore (1997: 284f.). 21. See Cappelen and Lepore (2005). The speech act pluralism suggested by Cappelen and Lepore (2005) follows directly from this criticism. 22. Cappelen and Lepore (1997: 289).

68

Sarah-Jane Conrad

still be false. Depending on the context of utterance, the following report on sentence (4) “Two weeks are a fortnight”23 is possibly false: (DIR 4)

Paul said that two weeks are two weeks.

On the other hand, a non-literal report can be true, namely when the sentence uttered was meant ironically.24 By transferring DIRs to the realm of pragmatics, they no longer are a reliable point of reference for identifying the semantic content of a sentence. Nevertheless, they can be considered a first guideline to the communicative standards operative in a particular communicative setting and thus may provide a systematically useful tool for a theory of pragmatics. But once again: formulating a correct DIR appears to be a real challenge if one ignores the standards relevant for the setting and it is certainly everything but an automatic procedure the traditional view took it for. Therefore, it is questionable how reliable of a methodological tool saying reports really are. All the more as the distinction between what a speaker said and what he meant will be difficult to defend if the criticism put forward here is taken seriously. This surely indicates that the ordinary use of either notion perhaps should not serve as a point of reference from which any conclusions should be drawn from. If the contents specified by DIRs depend only vaguely upon the linguistic meaning of the sentence uttered and instead rely on the communicative purpose, one may ask how to describe the semantic content of the sentence uttered, in very much the same manner as we have before. Having lost DIRs as an illuminating tool for describing the semantic content of an utterance, any assumptions concerning it or the effectiveness of linguistic meaning in general will appear to be purely speculative. Given the differing interpretations presented by minimalists and contextualists concerning the role of linguistic meaning for specifying propositional content and its relation to speaker meaning, a conclusive decision in favour of either party is difficult to make. Neither can we argue that the linguistic meaning is effective enough, at least in most cases, to determine a propositional content of the sentence uttered, with reference being fixed, nor, that the influence of context is simply too pervasive and the possibility of interpreting the use of a specific word too multifarious in order to be captured by linguistic meaning and consequently, we must delegate the task of determining propositions to pragmatics.25 Both attempts so far introduced are thus confronted with similar problems regarding the semantic conclusions following 23. See Cappelen and Lepore (1997: 286). 24. See Cappelen and Lepore (1997: 286). 25. Eventual reservations against Minimalism put forward by Cappelen and Lepore (2005) who claim that sentence (1) is semantically determinate partially relate to

Disquotational Indirect Reports in Focus

69

from them. Hence, it will come as no surprise that the efforts for rehabilitating DIRs as a useful tool for identifying the semantic content came as a prompt reaction to these tendencies of pragmatising DIRs completely and substantiate the worries put forward above: “[…] MA is an eminently plausible and methodological valuable tool for semanticists: one that enables them to home in on genuinely semantic content.”26 Besides, if Cappelen and Lepore are serious about the exclusive pragmatic status of DIRs, one might wonder why they still refer to them for testing if an expression is context-sensitive.27 They presumably hold DIRs to be informative about specific semantic properties of an expression. Such an assumption clearly contradicts their pragmatic interpretation of DIRs and I do not see how they can resolve this problem, as they have simply lost DIRs as a means for dismantling the contextualists’ thesis that the context operates on a wide range of expressions.28 Hence, the biggest critics of the traditional picture concerning the systematic and methodological role of DIRs seem heavily to depend on them as a tool for arguing for their minimalistic stance, which is peculiar enough. By arguing for a pragmatic interpretation of DIRs, Cappelen and Lepore have lost a useful instrument for defending any semantic thesis. 4.

Nonvariable DIRs or rehabilitating saying for semantics

The variability of DIRs and possible conclusions drawn from it with regard to the systematic and methodological role of the instrument provokes a number of delicate problems and it becomes difficult to evaluate the correct relationship between semantic and pragmatic aspects of the content expressed by a sentence uttered. As has already been mentioned above, it seems very attractive to try to rehabilitate DIRs as a valuable tool and taking its semantic origins serious. This is exactly what Kent Bach suggests in his philosophical work. He attempts a two-fold rescue of a semantically robust concept of what is said, first, by challenging the claim that DIRs or the indirect quotation-test, as he prefers to call it, this problem. See some of the contributions in Preyer and Peter (2007) or Carston (2008: 335). 26. Reimer (1998: 599). “MA” is a shorthand for “mistaken assumption”, as Cappelen and Lepore (1997, 1998) call the thesis that a semantic theory should assign contents to utterances of sentences which specify what a speaker said in uttering the sentence. See also Richard (1998) for a critical discussion of Cappelen and Lepore’s position. Yet, I will not discuss their position for reasons that will become clearer in a little. 27. See Cappelen and Lepore (2005: 94ff.). 28. See also Cappelen and Lepore (2007: 48f.).

70

Sarah-Jane Conrad

in short the IQ-test,29 is useless for a semantic theory and, second, by claiming that alternative explanations can be given for possible variability.30 The IQ-test was originally designed to show that all lexical items which pass the test must be sided with what is said, as they are intuitively considered to form part of the semantic content of the sentence uttered.31 Yet, if an additionally specified content such as the amendments of sentences (1) and (2) in (DIR 1

) and (DIR 2

) can be cancelled by the speaker without contradiction, this shows the supplemented information to be superfluous from a syntactic and semantic point of view and that seems enough of a proof that it is not part of what is said.32 The very possibility of cancelling these amendments suggests that (DIR 1) and (DIR 2) actually do determine what a speaker said in uttering a sentence. From both of these observations one can conclude that DIRs obey the so-called Syntactic Correlation Constraint (SCC), which is very similar to what Grice suggested in relation to his concept of saying:33 He [H. Paul Grice; cjs] imposed the Syntactic Correlation Constraint, as I call it, according to which what is said must correspond to “the elements of [the sentence], their order, and their syntactic character” (1989, 87). So if any element of the content of an utterance, i.e., of what the speaker intends to convey, does not correspond to any element of the sentence being uttered, it is not part of what is said.34

Bach believes to have a reliable criterion of adequacy for correct DIRs at hand with SCC thus specified. SCC emphasises the close connection between the syntax of the sentence uttered and what a speaker said: the latter must be related to conceivable, syntactical constituents of the sentence in question and represent the wording of it adequately. If the sentence is syntactically wellformed and thus non-elliptic, the original wording of the sentence together with its linguistic meaning provides a solid guideline to the semantic content of the 29. See Bach (1994: 146ff.). 30. See e.g. Bach (1994: 141ff. & 149ff.). Bach presents his ideas over a series of different papers; see Bach (1999), (2001), (2004), and (2005). 31. Bach (1994: 144ff. & 1999: 337ff.) points out that words generating conventional implicatures according to Grice (1989: 25) pass the test and therefore form part of what is said. 32. See Bach (2000: 264) and (2001: 18f. & 27). 33. Bach (1994: 143) subjects Grice’s (1989: 87f. & 120) concept of saying to a second modification as he rejects the Gricean thesis, namely that saying entails meaning. Bach believes the two notions are conceptually distinct – contra a typical contextualist’s stance. I will return to this point in a little. 34. Bach (2001: 15; [original styles; cjs]).

Disquotational Indirect Reports in Focus

71

utterance, thereby allowing to determine what a speaker said in uttering a sentence. Even though Bach doesn’t consider his test to be outright decisive,35 it provides a more or less sure methodological means to account for what is said by simply projecting the lexical items and the syntax of the sentence. Of course, the wording of the sentence originally uttered cannot be simply reproduced and certain modifications may be indicated, concerning not only the word order but the vocabulary, too. However, the grammar and the linguistic meaning of the sentence will dictate any necessary modifications. Having thus combined SCC and DIR, we have again a solid criterion for a correct saying-report based on the mere wording of the sentence, granted that it is syntactically well-formed. SCC obviously restrains the number of valid DIRs. With regard to what is said, Bach additionally assumes that any context influence relevant must be under total linguistic control. He suggests distinguishing two ways in which the context can influence the utterance-content: First, Bach points out that fixing the reference of “she” in the sentence (1) “She is ready” is a delicate task.36 The meaning of the pronoun surely implies some referential constraint concerning the sort of individual it can be used for to pick out. Yet, in order to determine the relevant female person, one must additionally know the speaker’s referential intentions, and these are, according to Bach at least, part of the speaker’s communicative intentions and thus pragmatic.37 Determining the reference of “she” and many other demonstratives is therefore no longer a matter of semantics, as it requires genuine communication-based information, to which Bach refers to as the wide context. Hence, only (DIR 1) adequately reports on (1), while all other reports include pragmatic information.38 The case presents itself a little differently for sentence (2): In order to determine what a speaker said by “I had breakfast”, the referent of the pure indexical “I” must be identified. Bach suggests a parameter approach to pure indexicals,39 as the meaning of these words allows grasping the relevant context almost automatically, previously identified by a handful of parameters such as speaker, 35. Bach mentions a few exceptions which can, however, be neglected in this context; see Bach (1994: 137) and (2001: 17). 36. See Bach (2001: 22 & 29). 37. See Bach (1987: 175ff.) and (2005: 39ff.). 38. At one point, Bach writes that disambiguating is also a matter of communicative intentions and hence a merely pragmatic process (Bach, 1987: 180, fn. 8). Sentences often fail to determine not only singular propositions, but general propositions, too. Bach therefore suggests equating what is said with an open proposition (Bach, 2001: 33). 39. See Kaplan (1989a & 1989b) and Perry (2001).

72

Sarah-Jane Conrad

addressee, place and time.40 Bach refers to these parameters as the narrow context and claims them to be operative for all pure indexicals. They can all be assumed without further ado as they are related to the sheer fact of having uttered a sentence.41 For this reason Bach takes it to be unproblematic to consider them as part of the mechanisms for identifying what a speaker said.42 Consequently, (DIR 2) provides an adequate report of (2), at least if “I” in (2) is interpreted as referring directly to the female person Jane. If the correct formulation of DIRs depends solely on SCC as well as the narrow context, the semantic content of the sentence uttered and thus what is said remain totally unaffected by eventual pragmatic influences. Once more, saying is intimately related to linguistic or conventional meaning. In addition, specific properties of the acts of uttering a sentence must be taken into account, namely that the sentence has been uttered by somebody, addressing herself to a hearer at a specific time and a specific place. The information provided by the linguistic meaning, on the one hand, and what is said, on the other, are almost identical, differing only with regard to the role the narrow context plays for specifying what is said. That is why the original wording of the sentence uttered serves as a nearly exclusive point of reference for providing a correct DIR, reassuring again the easy application of the method. In both regards Bach’s DIRs very much remind us of the systematic and methodological status they traditionally had; or they are, at least, very much alike. Two differences between Bach’s and the traditional DIRs are worth noticing, however: As mentioned above, Bach takes (DIR 1) to be perfectly accurate, but believes that the sentence fails to express a full-blown propositional content. It merely expresses a propositional radical. According to Bach, the only thing that follows from this is that most minimalists and contextualists wrongly assume propositionalism to be a necessary feature of what is said.43 Moreover, he agrees with the contextualism’s criticisms that the sentence’s literal meaning often underdetermines what the speaker meant. But again this does not affect the concept of what is said according to Bach, nor will it threaten the semantic40. See Bach (2000: 217f.). 41. Besides, perhaps, the addressee: Identifying to whom an utterance is addressed seems to involve knowledge of the communicative intentions. The question to what extent these different parameters allow to specify all pure indexicals is controversially discussed, but can be ignored here. See however e.g. Carston (2008: 325). 42. See Bach (2005: 39). 43. See e.g. Bach (1994: 127) and (2001: 23f.). Bach’s criticism of Grice’s conventional implicatures suggests already that Bach rejects the idea of taking truth-evaluable contents as the sole focus of a semantic theory.

Disquotational Indirect Reports in Focus

73

pragmatic distinction. He rather accuses the traditional view on DIRs of having committed yet another mistake by assuming an intimate relationship between the wording of a sentence, its linguistic meaning, and speaker meaning. An error Contextualism continues to repeat stubbornly. Because of both of these claims, Bach defends a radical minimalistic stance for what is said, based more or less exclusively on the criterion of literalism and he takes any other claim to be bogus.44 These three facts together – the possibility of providing accurate DIRs on the sole basis of SCC without having to implement propositionalism and showing any expansion or completion to be unnecessary – prove that none of the additional specifications found in any other DIR but DIR 1 and DIR 2 are a necessary part of the semantic content of the uttered sentence, but belong to pragmatics.45 If we apply SCC together with DIR, it is not only necessary, but almost sufficient to reflect more or less faithfully the wording of the original sentence and interpret the reference of pure indexicals on the basis of the narrow content. Consequently, DIRs of one and the same sentence will not vary and saying reports are rehabilitated as the methodologically valuable and systematically important instrument they were traditionally taken for;46 minus the close connection between saying and truth and without any connection between saying and speaker meaning. This makes any adequate saying reports much less far-reaching than the traditional view assumed, but just as easy to handle. If DIRs are associated with SCC and still conform with some of our intuitions concerning the use of “said”, as Bach’s empirically based results suggest, nothing can be cashed in on the variability of saying reports. They merely show that the expression can be used improperly, just as any other expression can. In all of these cases, the word “to say” actually refers to what a speaker tried to communicate or meant by uttering a sentence. In order to capture the content of the utterance adequately, any expanded or complemented report should really take on the following form: (* DIR 1

) Paul meant that Jane is ready for the theatre. (* DIR 2

) Jane meant that she had breakfast today. Mechanisms originating not so much from semantics than from communication are involved here and the content identified must therefore be related to the communicative notion of what is meant. The speaker’s communicative inten44. See Bach (2006: 435). 45. See Bach (2004: 27). 46. See Bach (2001: 22). Of course, Bach’s talk of semantic content does not imply any reference to propositions.

74

Sarah-Jane Conrad

tions operate together with the principles of rational communication, identified by Grice and adopted by Bach in a slightly modified version, without, however, changing Grice’s basic insights.47 These communicative presumptions, as Bach calls them, allow the speaker to express more and maybe something entirely different from what the literal meaning of the sentence suggests. Grice calls these additional contents conversational implicatures, as is well-known.48 Bach basically agrees with Grice’s conception, but does not believe it to be exhaustive. He proposes complementing the original distinction between what is said and conversational implicature by a third category, namely the category of conversational impliciture.49 According to Bach, both of the above mentioned (* DIRs) specify the content of such a conversational impliciture, as they build onto the original content of what is said, without determining a new proposition, as we would expect in the case of implicatures.50 However, implicatures and implicitures both determine pragmatic contents as the concepts involved to specify the respective contents originate from communication.51 Bach takes this conceptual fact to prove that the literal meaning of a sentence can’t account for speaker meaning and that this is necessarily so. The difference is programmed right from the beginning due to insuperable conceptual differences. If Bach’s criticism is correct, traditionalists and contextualists simply impose misguided demands upon the literal meaning of a sentence by siding speaker’s meaning with what is said and ignoring their conceptual disparity, thereby blurring the semantic-pragmatic distinction for nothing. By entrenching what is said in the wording of a sentence, Bach can prevent this confusion by conceptually and factually individuating any type of content as either semantic or pragmatic. In sum, while Bach delivers with SCC and the narrow context a reliably applicable criterion for identifying the semantic content of a sentence uttered, his theory of communication provides the necessary information for other, non-literal uses of saying.52 47. See Bach and Harnish (1979: 12ff.) and Bach (2005: 39). 48. See Grice (1989: 26ff.). 49. See Bach (1994: 126). 50. Because Bach defends the so-called principle of expressibility, he does not encounter the troubles Contextualism has been accused for above, at least not with regard to saying. And the double role of speaker meaning is in principle describable by an adequate formulation of the two respective pragmatic contents. 51. See e.g. Bach (2005: 22f.). 52. Bach’s category of implicitures identifies the same content as the category of explicatures, introduced by contextualists. Nevertheless these two concepts are totally different, even though this point is often ignored; see e.g. Borg (2007: 348) or Wedgwood (2007). Bach’s semantic-pragmatic-distinction very much reminds us of a con-

Disquotational Indirect Reports in Focus

5.

75

Evaluating Bach’s IQ-test and his concept of saying

Bach claims that knowledge of the linguistic meaning, compositional rules and narrow context sufficiently determines what a speaker said in uttering a sentence. By defending a conception of DIRs thoroughly entrenched in the SCC, Bach underlines the importance of almost strictly reflecting the wording of the sentence uttered. This provides a reliable point of reference for formulating adequate DIRs, at least for syntactically well-formed sentences. Bach thereby replaces an intuitive concept of what is said by a technically refined one because he emphasizes that the route to what is said must run exclusively along syntactic trails. The concept of saying thus analysed is certainly a term of art, but all other approaches have been shown to operate with theoretically biased concepts of saying, too. Consequently, this cannot be taken to indicate any disadvantage.53 Yet, in order to evaluate how useful Bach’s concept of saying actually is, the two following questions have to be examined: Is it still possible to defend an effective semantic theory, even though linguistic meaning has been disconnected from truth and speaker meaning? And does his semantic provide a useful starting point for a theory of communication? The first answer requires that we know what type of content Bach’s saying reports actually refer to and if their contents deliver conclusive information concerning the syntax and the semantics of a particular language. A positive answer to the second question will prove the concept strong enough for being put into service for a theory of language in use.54 textualist’s stance as Carston (2008) correctly shows. Yet his concept of saying differs completely from a contextualist’s approach, which takes explicatures to be part and parcel of what is said. Consequently, semantic as well as pragmatic concepts are constitutive for saying according to Contextualism. These assumptions are defended, as mentioned above, on the basis that propositions in general and the propositions actually meant by the speaker in uttering the sentence must be accounted for when specifying what is said. This shows that the role of semantic content is assessed against the background of propositionalism as well as the use of sentences in communication. Contextualism generally assumes a close conceptual connection between meaning, truth and communication. Bach rejects this conceptual triad and thus advances a totally different concept of what is said. He denies any conceptual link between saying and truth, and saying and communication, and takes linguistic meaning to be more or less the sole component of saying. 53. See also Reimer (1998: 599) on this point. 54. This means that language is put to service for a theory of speech acts and what is said must therefore serve as enough of an input for illocutionary acts; see Bach and Harnish (1979).

76

Sarah-Jane Conrad

We may thus ask what type of semantic content Bach’s IQ-test actually identifies. It is worth repeating, once again, that Bach’s talk of semantic content does not imply any reference to a truth-evaluable content. But what type of contents do his saying reports then specify? SCC privileges the wording of a sentence and any report on what is said, correctly called so, must reflect the syntactic structure of the sentence accurately. Any modification regarding the word order is due to the syntax of the language in question. However, the semantic content in question surely is no simple representation of signs (spoken or written), but refers to the linguistic or conventional meaning of the expressions mentioned in the report, which, according to the rules of the language, build up to a uniform content. Because Bach distinguishes the role of narrow context and wide context and assumes only the first to be relevant for specifying the semantic content, questions concerning the unity of content arise, however. First, disambiguating ambiguous expressions is considered as a pragmatic process and instead of being able to indicate the relevant meaning of the sentence, we will often encounter a series of different word-meanings as part of the semantic content. Second, while the referent of pure indexicals can be determined, the reference of all other singular terms will be specified by pragmatics. The referents of a referring-expression will not be necessarily identified by the saying report, but only in some cases. Consequently, if we report a sentence like (5) “I and she had breakfast” the predicate “had breakfast” will in one case identify the property attributed to the speaker of the utterance, in the other case, however, it will merely identify the meaning of the word and a competent speaker of the language will only know that the predicate-expression must be predicated on the subject-expression. The semantic content related to what is said appears to be hybrid and very complex, listening word-meanings, sometimes of one and the same word as well as referents.55 Is such a hybrid content with regard to what is said really useful for a theory of language and can it be used as a starting point for a theory of communication? Criticism has been put forward with regard to both questions. On the one hand, propositionalism has been taken to be indispensable for explaining the productivity and systematicity of a theory of language; on the other hand, Bach’s concept of what is said has been considered as being too weak for a theory of communication, as it does not allow determining the respective communicative content of speech acts. Certainly, Bach’s semantic content does not correspond to any traditional assumptions concerning propositional content as we may distinguish a variety of different, possible interpretations of the semantic content related to what 55. Carston (2008: 327) takes the hybrid character of what is said à la Bach to be one of its great disadvantages and therefore rejects Bach’s concept of saying.

Disquotational Indirect Reports in Focus

77

is said, namely when the sentence is syntactically or semantically ambiguous. However, because the syntactical rules as well as the vocabulary can be taken to be familiar to the person reporting, everything necessary for providing possible interpretations of the sentence is present.56 If one could show that a systematic and productive theory of language can in principle not be defended on the basis of conventional meaning alone, Bach would have to either challenge this assumption or revise his own position, as linguistic meaning and syntax, both ingredients of Bach’s DIRs, then would depend on propositions.57 The problem is, however, that most of his opponents simply claim propositionalism to be necessary without supporting their assumptions by a substantial argument. So long as an argument is missing, Bach certainly need not respond to these objections.58 On the other hand, contextualists, too, need to operate with a concept of linguistic meaning of an equally uninformative status and yet still believe it possible to determine on its basis different types of communicative content, namely what is said, explicatures and implicatures.59 If contextualists can manage to operate with this minimal kind of information, there is no reason to assume that Bach fails. Taking propositional contents to be indispensable for saying must be argued for independently as, up to now, substantial reasons for this claim merely seem to reflect some specific theoretical decisions. By contrast, Bach’s own concept of saying is based on little theoretical assumptions, which cannot be easily rejected neither from a theoretical nor from a practical point of view. This can certainly be considered as one of the greatest advantages of Bach’s analysis, much more so, as his SCC-constrained DIRs responds to some of our everyday intuitions. It seems to determine a common denominator to which theory and common intuitions can both refer to and it provides some insights regarding the mechanisms of a particular language. There are no reasons to doubt that it may serve as an input for the communicative processes, too. 6.

Concluding remarks

The contemporary debate on saying focuses on the question if the concept can be interpreted in purely semantic or purely pragmatic terms, or if it contains semantic as well as pragmatic information. Most of the different approaches on 56. Compare also Grice’s (1989: 25) first interpretation of saying. 57. See e.g. Stanley and King (2005: 120 & 140). 58. Yet, I must admit that I am arguing here in some sense against my own intuitions. But without conclusive arguments, principled resistance against Bach’s is futile. 59. See especially Carston (2008).

78

Sarah-Jane Conrad

the market criticise the traditional embedding of saying in a triad of concepts: the wording of a sentence, propositionalism and speaker meaning. Saying reports and DIRs play an important role for the discussion. They are used for demonstrating some allegedly problematic aspects of the notion of saying or for corroborating a certain interpretation of the relationship between a sentence, its meaning, the proposition expressed and speaker meaning. All three approaches discussed here take the traditional analysis of the concept of saying to be wrong. Two of the presented approaches point to the fact that DIRs can vary across context and therefore suggest implementing saying in a pragmatic framework, either exclusively or at least partially. Such an interpretation encounters three major problems: First, because of the extensive variability of saying reports, it is questionable to what extent they can be used as an evidential basis for a theoretical claim. Second, the variability of DIRs risks resolving any substantial difference between saying and meaning, either on theoretical or on practical grounds. Finally, saying reports merely show what kind of pre-theoretical assumptions have been made and their methodological and systematic value seems therefore questionable, as well as the theoretical conclusions drawn from saying reports. Hence, possible constraints on any criteria of adequacy for a DIR do nothing but reflect the theoretical assumptions that have been made beforehand. Given the theoretical predilection, the number of acceptable saying reports will vary according to the theory at hand and DIRs will not provide any evidential prove for or against a specific claim. Saying itself appears to be really a theoretical concept, which can be used to respond to a variety of theoretical demands. However, any tendency to ‘pragmatise’ the concept of saying raises the question if it still is possible to make any assumptions concerning the semantic content of a sentence. This problem should not be underestimated, because the very possibility of substantiating any claim concerning semantic concepts like convention, rule-following, composition, on which an explanation of the systematicity and productivity of a language is based on, depends on the possibility of providing a positive answer to this question. The gap a possible pragmatic implementation of DIRs opens up for a semantic theory is thus considerable. Bach jumps into this breach with a modest theoretical concept of saying, which is discharged from all problematic context-influence and reflects a number of theoretical and intuitive assumptions and yet appears powerful enough for explaining the mechanisms of language and communication. Despite the fact that saying reports can and do vary, it is nevertheless always possible to represent accurately the wording or syntax and this fact seems to be resistant to any crit-

Disquotational Indirect Reports in Focus

79

icism.60 Hence, it seems difficult to contradict his claim that a syntactically well-formed sentence can be reproduced in an identical manner across different contexts.61 In addition, we can resume these different reports and form one collective report on them.62 Intercontextually valid DIRs as well as the possibility of collective reports point to a certain linguistic stability. Thus the respective reports promise to provide some instructive insights about the mechanisms operating within a particular language by showing which rules are valid for it and what a specific lexical item can mean.63 As minimal as the suggested criteria of adequacy may be, they at the same time introduce the maximal constraints we can put on the concept of saying. A controversy about a more bolstered concept can be avoided, a controversy which is at risk to turn into a dispute merely about terminology and theoretical predilection. It is therefore difficult to see why we should replace Bach’s concept by another, more ambitious one. All the more as the claim that Bach’s concept fails and is of no use for a theory of language or a theory of communication is still awaiting a substantial proof, to my knowledge. Most criticisms do not argue the point, but merely claim it to be so. As long as good arguments to the contrary are missing, the concept can still be taken to do the theoretical work it originally was designed for. Of course, one major problem yet remains unresolved, namely the question to what extent the linguistic meaning is effective enough for determining a propositional content of a sentence uttered. Bach’s theoretical framework surely does not provide a conclusive answer to this question and all other approaches have been shown to be problematic. In addition, it seems natural to accuse Bach’s DIRs to be totally uninformative on the syntax and the semantic of a language as they only reflect the knowledge any competent speaker of a language has. But maybe that is the most natural reading of the concept saying and SCC thus accurately reflects the role of saying reports. Perhaps any question concerning propositions must be clarified by using DIRs of another type, namely “It is true that” or “It is false that …”. It is possible that these will prove more valuable for a thicker concept of syntax and semantics the traditional view 60. See e.g. Carston (2008: 332f.). 61. See also MacFarlane (2007: 24). 62. For a recent, substantial theory of saying based mainly on the possibility of formulating certain DIRs, see Stojanovich (2008); she basically draws her attention to the so-called collectivity-test, i.e. the possibility of resuming two utterances of one and the same sentence into one report. Just like DIRs, collectivity-tests seem to show a certain linguistic stability to be present so that it seems quite plausible to draw on their basis some conclusions about the concept of what is said and thus about the semantic content of sentences uttered. 63. See Cappelen and Lepore (2007: 55 & 92f.).

80

Sarah-Jane Conrad

on DIRs was looking for. But examining this question would open up yet a new topic and thus can be put aside for the moment.64 References Bach, Kent 1987

Thought and Reference. Oxford: Clarendon Press.

Bach, Kent 1994

Conversational implicitures. Mind & Language 9: 124–162.

Bach, Kent 1999

Bach, Kent 1999 Bach, Kent 2000 Bach, Kent 2001 Bach, Kent 2004 Bach, Kent 2005 Bach, Kent 2006

The semantics-pragmatics distinction: What it is and why it matters. In K. Turner (ed.), The Semantics/Pragmatics Interface from Different Points of View. Oxford: Elsevier, 65–84. The myth of conventional implicature. Linguistics & Philosophy 22: 337–366. Quantification, qualification and context. A reply to Stanley and Szabò. Mind & Language 15: 262–283. You don’t say? Synthese 128: 11–31. Minding the gap. In The Semantics/Pragmatics Distinction, M. Devitt, et al. (eds.), 27–43. Chicago: Chicago UP. Context ex machina. In Semantics versus Pragmatics, Z.G. Szabò (ed.), 15–44. Oxford: Oxford UP. The excluded middle: semantic minimalism without propositions. Philosophy & Phenomenological Research 73: 435–442.

Bach, Kent & Robert M. Harnish 1979 Linguistic Communication and Speech Acts. Cambridge: MIT Press. Bezuidenhout, Anne 2002 Truth-conditional pragmatics. Noûs 36: 105–134. Borg, Emma 2004a

Minimal Semantics. Oxford: Oxford UP.

64. Many thanks go to Silvan Imhof, Nathalie Loetscher and Klaus Petrus for their extremely helpful comments on previous versions of this paper and the editors of this volume, especially Jörg Meibauer.

Disquotational Indirect Reports in Focus Borg, Emma 2004b

Borg, Emma 2007

81

Saying what you mean: Unarticulated constituents and communication. In Ellipsis and Non-sentential Speech, R. Elugardo et al. (eds.), 237–262. Dordrecht: Kluwer. Minimalism versus contextualism in semantics. In Context-sensitivity and Semantic Minimalism, G. Preyer et al. (eds.), 339–359. Oxford: Oxford UP.

Cappelen, Herman 2008 Content relativism and semantic blindness. In Relative Truth, M. Garcìa-Carpintero et al. (eds.), 265–286. Oxford: Oxford UP. Cappelen, Herman & Ernie Lepore 1997 On an alleged connection between indirect speech and the theory of meaning. Mind & Language 12: 278–296. Cappelen, Herman & Ernie Lepore 1998 Reply to Richard and Reimer. Mind & Language 13: 617–621. Cappelen, Herman & Ernie Lepore 2005 Insensitive Semantics. Malden MA: Blackwell. Cappelen, Herman & Ernie Lepore 2007 Language Turned on Itself. Oxford: Oxford UP. Carston, Robyn 1988

Carston, Robyn 2002 Carston, Robyn 2008

Implicature, explicature, and truth-theoretic semantics. In Mental Representations: The Interface Between Language and Reality, R.M. Kempson (ed.), 155–181. Cambridge: Cambridge UP. Thoughts and Utterances. Oxford: Blackwell. Linguistic communication and the semantics/pragmatics distinction. Synthese 165: 321–345.

Clark, Herbert H. & Gerrig, Richard G. 1990 Quotations as demonstrations. Language 66, 764–805. Greenhall, Owen 2006 The Semantics/Pragmatics Distinction: A Defence of Grice. PhD Thesis. Oxford: New College. Grice, H. Paul 1989 Harth, Manfred 2003

Studies in the Way of Words. Cambridge MA: Harvard UP. Concepts of what is said. In Saying, Meaning, Implicating, G. Meggle et al. (eds.), 72–83. Leipzig: Leipziger Universitätsverlag.

82

Sarah-Jane Conrad

Harth, Manfred 2005 Kaplan, David 1989a Kaplan, David 1989b

Logische Form und Gesagtes. http://www.meaning.ch/content/view/ 91/. Demonstratives. In Themes from Kaplan, J. Almog et al. (eds.), 481– 563. Oxford: Oxford UP. Afterthoughts. In Themes from Kaplan, J. Almog et al. (eds.), 565– 614. Oxford: Oxford UP.

MacFarlane, John 2007 Semantic minimalism and nonindexical contextualism. In Contextsensitivity and Semantic Minimalism, G. Preyer et al. (eds.), 240–250. Oxford: Oxford University Press. Montminy, Martin 2010 Two contextualist fallacies. Synthese 173: 317–333. Perry, John 2001

Reference and Reflexivity. Stanford: CSLI.

Predelli, Stefano 2005 Contexts: Meaning, Truth, and the Use of Language. Oxford: Clarendon Press. Preyer, Gerhard & Georg Peter 2007 Context-sensitivity and Semantic Minimalism. Oxford: Oxford University Press. Recanati, François 1989 The pragmatics of what is said. Mind & Language 44: 295–329. Recanati, François 2004 Literal Meaning. Cambridge: Cambridge UP. Recanati, François 2007 Perspectival Thought. Oxford: Oxford UP. Reimer, Marga 1998 Richard, Mark 1998 Soames, Scott 1989

What is meant by ‘what is said’? A reply to Cappelen & Lepore. Mind & Language 13: 598–604. Semantic theory and indirect speech. Mind & Language 13: 605–616. Direct reference, propositional attitudes, and semantic content. Philosophical Topics 15: 47–87.

Sperber, Dan & Deirdre Wilson 1986 Relevance. Cambridge: Harvard UP.

Disquotational Indirect Reports in Focus Stanley, Jason 2000

83

Context and logical form. Linguistics and Philosophy 23: 391–434.

Stanley, Jason & Jeffrey C. King 2005 Semantics, pragmatics, and the role of semantic content. In Semantics versus Pragmatics, Z. G. Szabò (ed.), 111–164. Oxford: Oxford UP. Stojanovic, Isidora 2008 De Se Assertion. http://jeannicod.ccsd.cnrs.fr/ijn_00280484/en/. Travis, Charles 1985

On what is strictly speaking true. Canadian Journal of Philosophy 15: 187–229.

Wedgewood, Daniel 2007 Shared assumptions: semantic minimalism and relevance theory. Journal of Linguistics 43: 647–681.

Behabitive Reports Cathrine Fabricius-Hansen and Kjell Johan Sæbø

1.

The problem: Locution or illocution?

“Behabitives” (Austin 1962: 160), or “verbs of judging” (Fillmore 1971b: 273), like praise or criticize, are illocutionary verbs, reporting certain speech acts. But can they also be locutionary verbs, quotation verbs, reporting speech? The data give conflicting answers. Judging from (1), they cannot; this sentence does not entail the sentence coming from it by substituting said for criticized and carrying out the other, syntactic, changes made necessary by that substitution, (3i). It entails (3ii), a sentence resulting from substituting said and including a general negative moral judgment predicate, act wrongly. But then, so does (2), a sentence which includes that predicate already; this sentence does entail the result of substituting said for criticized and carrying out the corollary changes. If (2) were to pattern with (1), it should entail (3iii), which it does not. In fact, (1) and (2) seem to mean more or less the same. (1)

John criticized Herod for marrying Herodias.

(2)

John criticized Herod for acting wrongly in marrying Herodias.

(3)

(i) John said that Herod had married Herodias. (ii) John said that Herod acted wrongly in marrying Herodias. (iii) John said that Herod acted wrongly in acting wrongly in marrying Herodias.

Furthermore, judging from (1) the verb criticize is a factive verb, triggering the presupposition corresponding to the proposition expressed by the gerund phrase (its empty subject controlled by the DP complement) – that Herod had married Herodias; but judging from (2) this is not so: this sentence does not presuppose that Herod acted wrongly in marrying Herodias. This confusing state of affairs is no accident of English. In German, say, the verb kritisieren, which typically takes a clausal, dass ‘that’ complement, exhibits the same ambivalence in regard to the role played by the proposition expressed by that clause: it can be what the matrix subject somehow uttered a

86

Cathrine Fabricius-Hansen and Kjell Johan Sæbø

criticism of, as in (4), or it can be what the matrix subject uttered as a criticism, as in (5). (4)

Einstein kritisierte, dass Bohr die Kausalität aufgegeben hatte. ‘Einstein criticized Bohr for having abandoned causality.’

(5)

Einstein kritisierte, dass Bohr voreilig die Kausalität aufgegeben habe. ‘Einstein criticized Bohr for having prematurely abandoned causality.’

In (4), what Einstein uttered is not specified, but whatever it was it was understood as a criticism of the fact that Bohr had abandoned causality, while in (5), what Einstein uttered was that Bohr had prematurely abandoned causality, – and this is understood as a criticism of some familiar fact (in this case probably the fact that Bohr had abandoned causality). Note that in the latter sentence only, the finite verb in the complement clause is in the present subjunctive – a fact we will return to in due course. Note also that provided (5) does report a criticism of the fact that Bohr had abandoned causality, (4) and (5) seem to have a common entailment like (6): (6)

Einstein said/wrote that Bohr acted wrongly in abandoning causality.

Here we reencounter the kind of general negative moral judgment predicate that we saw in (2) and in (3ii), and we see that the reason that (1) and (2) meant more or less the same is that the moral judgment predicate in (2) was so general; the negatively flavoured adverb voreilig ‘prematurely’ is much more specific (as we will see, the relevant locutions come in a wide array of forms and flavours). We will be saying that in one case, exemplified by (1) and (4), the complement proposition is the object of judgment, and that in the other case, exemplified by (2) and (5), it is the content of judgment. Let us refer to the two cases as the O case (for “object”) and the C case (for “content”), respectively. Evidently, in the O case, to criticize ˛ for doing P is to say something conveying disapproval of P(˛), whereas in the C case, it is to say P(˛), thereby conveying disapproval of something ˛ has done or is doing. Only in this case is the verb a quotation verb, reporting a locution as well as an illocution. What we have been loosely referring to as a negative moral judgment predicate we will term the evaluative element the presence of which in the complement clause is, as it appears, necessary (maybe even sufficient) for the C case. We are faced with a seeming ambiguity: a verb criticize 1 which is not a verbum dicendi and another verb criticize 2 which is. However, there are clearly strong semantic ties between the two, and the double-naturedness seems systematic, across verbs and across languages; the challenge, then, lies in recon-

Behabitive Reports

87

ciling these observations through an analysis that predicts the (il)locutionary alternation in the behabitive attitude verb in a systematic way. To that end, we need a clear understanding of the two cases and the constraints they are subject to. The rest of the paper is structured as follows. In Section 2, we address the various syntactic environments behabitive verbs occur in (including quotation marks), noting the constraints they put on one or the other interpretation and trying to explain those constraints. In Section 3, we turn to the semantic requirements posed by one or the other interpretation, exploring, in particular, the evaluative element characteristic of the content case interpretation. In Section 4, we develop a formal analysis, relating the two interpretations to a unitary format and motivating their distinctive properties. Section 5 brings concluding remarks.

2.

Object or content of judgment: Syntactic properties

The Object–Content ambivalence in the propositional argument of behabitives cuts across several constructions, but there are also constructions that definitely favour one or the other of the cases. In particular, the content case correlates with certain grammatical characteristics of quotations. In this section, we will explore this terrain, mapping the syntactic features and environments that are necessary or sufficient for this case to obtain. 2.1. Janus-faced abstract DP arguments Note, first, that both in English and in German, a verb like criticize/ kritisieren can take a single, proposition-type DP argument, usually a definite description built around a deadjectival or deverbal noun, and that even then, the C case is a viable alternative beside the O case, as witnessed by (7a–b) and (8a–b): (7)

a. b.

(8)

a.

The principal addressed the student body and criticized the content of our show. The principal addressed the student body and criticized the immoral content of our show. Der Trainer kritisierte das Deckungsverhalten seiner the coach criticized the coveringconduct hisGEN Truppe. troop ‘The coach criticized his team’s defense work.’

88

Cathrine Fabricius-Hansen and Kjell Johan Sæbø

b.

Er kritisierte das zu lasche Deckungsverhalten seiner he criticized the too lax coveringconduct hisGEN Truppe. troop ‘He criticized his team’s too lax defense work.’

From (7a), we can conclude that the principal said something to the effect that our show should have had a different content from what it had, but from (7b) we can conclude that (s)he said that our show had an immoral content – which is one way of saying that it should have had another content. In other words, the valency frame where the verb selects for a DP with a propositional content fails to discriminate between an object and a content role for that content. 2.2. Embedded V2 clauses in German: C only In German, there are two ways to construct a declarative complement clause: as a subordinate, verb-final clause introduced by the subjunction dass ‘that’, or as a verb-second clause without a complementizer. The C role is compatible with both realizations, as illustrated in (9a-b):1 (9)

a.

b.

Platon kritisierte, dass die Rhetorik unmoralisch ist. Plato criticized that the rhetoric immoral is ‘Plato criticized rhetoric as immoral.’ Platon kritisierte, die Rhetorik ist unmoralisch. Plato criticized the rhetoric is immoral ‘Plato criticized rhetoric as immoral.’

But the O role is only compatible with the subordinate clause realization: (10)

a.

Platon kritisierte, dass die Sophisten Rhetorik lehrten. Plato criticized that the Sophists rhetoric taught ‘Plato criticized the Sophists for teaching rhetoric.’ b. ??Platon kritisierte, die Sophisten lehrten Rhetorik. Plato criticized the Sophists taught rhetoric

It is possible to make sense of (10b), but only if teaching rhetoric can safely be taken to imply something negative; the object of criticism is then left implicit. What is the reason that only the C role is compatible with the V2 realization? Truckenbrodt (2006), building on Reis (1997) and Gärtner (2002), develops a 1. Note that a subjunctive finite verb form is a default in these cases; we turn to that in 2.3.

Behabitive Reports

89

theory of V2 clauses predicting a range of semantic restrictions on embedding predicates, inter alia, that they cannot be factive. If case O kritisieren is factive, its incompatibility with embedded V2 clauses is thus immediately accounted for. 2.3. Subjunctive in German complement clauses: C only In German, indirect quotation is normally marked by a subjunctive verb form (the Referatkonjunktiv).2 Thus the C case examples in (9) could equally well, or would probably, be formulated as in (11): (11)

a.

b.

Platon kritisierte, dass die Rhetorik unmoralisch Plato criticized that the rhetoric immoral sei / wäre. issubj-pres/past ‘Plato criticized rhetoric as immoral.’ unmoralisch. Platon kritisierte, die Rhetorik sei / wäre Plato criticized the rhetoric issubj-pres/past immoral ‘Plato criticized rhetoric as immoral.’

A close look at (9) reveals that the tense is a relative present even though the mood is indicative – again, this is typical of reported speech. In (10), however, we find an absolute past (or a past subjunctive substitution for a nondistinct relative present subjunctive); so the two cases correlate with distinct forms of the finite verb in the indicative as well. The subjunctive enhances the contrast: irrespectively of verb-last or verb-second, the O case is incompatible with it. (12)

a. ??Platon kritisierte, Plato criticized würden. would b. ??Platon kritisierte, Plato criticized

dass die Sophisten Rhetorik lehren that the Sophists rhetoric teach

die Sophisten würden Rhetorik lehren. the Sophists would rhetoric teach

Note that there is a slightly different reason for this than for the V2 constraint: while the latter rests on the resistance of V2 to factivity, the former is grounded on the presupposition that the content of the clause be uttered, a presupposition

2. A present or present perfect tense Referatkonjunktiv is mandatory in writing; in speech, past or past perfect subjunctive forms predominate in the same function.

90

Cathrine Fabricius-Hansen and Kjell Johan Sæbø

the C case can comply with but not the O case (for a discussion of the relevant features of the Referatkonjunktiv see Fabricius-Hansen and Sæbø 2004). 2.4. German correlative complement clauses: O only Conversely, only the O case is compatible with a pronominal (es ‘it’) expletive, or Korrelat ‘correlate’, for the complement clause. (13)

a.

Platon hat es kritisiert, dass die Sophisten Rhetorik Plato has it criticized that the Sophists rhetoric lehrten. taught ‘Plato criticized the Sophists for teaching rhetoric.’ b. ??Platon hat es kritisiert, dass die Rhetorik unmoralisch Plato has it criticized that the rhetoric immoral ist. is

Though there is considerable literature on the syntactic status of the correlate (e.g. Büring and Hartmann 1995, Müller 1995, Vikner 1995, or Sudhoff 2004), the semantic picture is much less developed. One of the two “crucial questions … not … touched on” by Sudhoff (2004) is whether the classification of matrix verbs … corresponds to certain semantic properties of the verbs, i.e. whether it is possible to predict the (in)compatibility with correlate-es from their meaning.

This question, however, “cannot be answered in a straightforward way. Several proposals have been made … , none of which seem to be entirely conclusive” (Sudhoff 2004: 23). Our intuition is that es signals that the complement clause it correlates with is not quite discourse-new. It agrees with factives like bedauern ‘regret’ (the strings hat es bedauert, dass and hat bedauert, dass are equally frequent) but it disagrees with verbs of saying like behaupten ‘claim’ (except in contexts such as niemand hat es behauptet, dass ‘noone has claimed that’); between these two extremes there are many verbs, including verbs of saying, where es seems to imply that the proposition expressed by the correlative clause is already under consideration in the discourse. If this is more or less correct, it suggests that case C judgment verbs resist the correlate because the content of judgment is in essential part discourse-new.

Behabitive Reports

91

2.5. Complement clause or DP + predicative phrase What we have (in Section 1) called the “evaluative element” which has to be present in a C case behabitive complement clause – for instance, act wrongly – can have a separate mode of expression in a small clause predicate introduced by a particle such as English as or German als. Consider (14): (14)

Die Ärzte the doctors Medikament drug

kritisieren criticize jetzt auf now on

(es) als unethisch, dass das it as unethical that the den Markt gebracht wird. the market brought becomes

One would think that once the evaluative element – here, unethical – has been safely put away, so to speak, in another phrase, the propositional complement is always the pure object of criticism, an objective fact, as in (14) indeed it is. However, this does not seem to be the case; we do encounter cases like (15):3 (15)

Die Ärzte kritisieren (es) als unethisch, dass das the doctors criticize it as unethical that the Medikament zu einem Zeitpunkt auf den Markt gebracht drug to a timepoint on the market brought werde, wo die Nebenwirkungen noch nicht becomessubj-pres where the sideeffects yet not genügend bekannt seien. sufficiently known aresubj-pres ‘Doctors criticize the decision to put the drug on the market at a time when side effects are not yet sufficiently known as unethical.’

It is perhaps not so clear from the English paraphrase, but the subjunctives in the complement clause show that we are dealing with a C case (the extra evaluative element, over and above unethisch ‘unethical’, residing in the adverb (nicht) genügend ‘(not) sufficiently’). We can imagine the doctors issuing the statement (16) and being reported to have done so with (15): (16)

The side effects of this drug are not yet sufficiently known. Putting it on the market now is therefore unethical.

3. One may ask why, apparently contradicting the claim made in Section 2.4, the correlate es is possible here; the answer seems to be that – independently of the behabitive verb – a predicate like unethisch requires a subject correlate when the subject clause is extraposed.

92

Cathrine Fabricius-Hansen and Kjell Johan Sæbø

In English, the “subject” of a small clause with a predicate introduced by as is mostly a DP, and often a DP that appears to denote a type e entity, as in (17): (17)

Opponents criticize the building as too high.

We wrote in Section 1 that to criticize ˛ for doing P is in the C case to say P(˛) and thereby convey one’s disapproval of some action of ˛. More generally, we can say that to criticize ˛ for – or as – P is (in the C case) to say P(˛) and thereby convey one’s disapproval of some fact about ˛. But even if we do, what is the relevant fact in a case like (17), about, here, the building? It seems clear to us that criticism of a person or thing is not the conveyance of disapproval of that person or thing per se. The disapproval one ascribes to a person when one describes that person as criticizing something is generally a counterfactual desire, so if that something is a type e entity, an individual, what the person disapproves of, wishing it were otherwise, is the fact that the entity has a certain property or set of properties. More specifically, the relevant fact about the building in cases like (17) or (18) is, in our view, that it has the relevant properties it actually has, and here the relevant properties seem to be properties pertaining to height, the relevant fact being simply that the building is as high as it (actually) is. In (18), the relevant properties are those pertaining to aesthetic appearance. (18)

Opponents criticize the building as an eyesore.

We will return to this issue when we develop our formal analysis in Section 4. To sum up, the constructions involving a small clause predicate introduced by a(l)s are consistently C case insofar as that predicate is an evaluative element, characteristic of the C case. This does not, however, exclude that a propositional participant appearing in the role of small clause subject is also a C case actant. To criticize someone/something ˛ as P is to say that ˛ is P and thereby convey one’s opinion that ˛ should not have all the properties s(h)e or it actually has, – and particularly not the ones that pertain to P. 2.6. Quotation mixed and pure Across languages, “mixed quotations”, where a proper part of a clause is marked as direct quotation (cf. e.g. Cappelen and Lepore 2005, Geurts and Maier 2005) show a very clear preference for the C case. (19)

The State Department has criticized Iraqi President Saddam Hussein for using “very brutal means” to suppress civil unrest.

Behabitive Reports

(20)

93

Seehofer lobte, dass Merkel “Großartiges” erreicht Seehofer praised that Merkel “grand” achieved habe. havesubj-pres ‘Seehofer praised Merkel for achieving “great results”.’

Is this clear preference an absolute correlation? Although in theory, there may be other reasons for inserting a piece of direct quotation in a judgment report than to represent some evaluative element, we find that in practice, it is difficult not to interpret such an insertion as something that helps constitute a criticism. (21)

a.

My boss has criticized me for not publishing anything for “a decade”.

(21a) hardly has a reading entailing that my boss uttered something extraneous conveying dismay at my not publishing anything for what she called a decade; maybe she uttered the whole of (21b), but the first half of this two-sentence text suffices to convey dismay at my not publishing anything for ten years. (21)

b.

You have not published anything for a decade. That is unfortunate.

To test properly whether the O case is at all compatible with a piece of quotation in the propositional argument, we could insert one in a German “O case only” construction (cf. Section 2.4): (22)

Sie kritisierte es, dass das she criticized it that the für Schülerinnen für “nicht for schoolgirls for “not vereinbar” hält. compatible” holds

Ministerium ein Kopftuchverbot ministry a hijab ban mit der Landesverfassung with the stateconstitution

‘She criticized the Ministry for ruling that a hijab ban for schoolgirls is “incompatible with the State Constitution”.’ It is fully possible to make sense of this sentence, but the natural way to do so is to ultimately attribute the quote not to the criticizer but to the criticizee, – while also, it would seem, inferring that the criticizer is quoting the criticizee. In conclusion, it seems as if a behabitive report containing a piece of quotation attributable to the agent of the report alone can only be interpreted as reporting her locution, the content of, here, criticism.

94

Cathrine Fabricius-Hansen and Kjell Johan Sæbø

Be this as it may, when the judgment report is directly quotational in toto, as in the German sentence (23), there can be no doubt that a content of judgment is being reported: (23)

“Das war taktisch völlig falsch”, kritisierte der Coach. “that was tactically totally wrong” criticized the coach

(24)

“Ist das kinderfreundlich?” kritisierte er. “is that childrenfriendly?” criticized he

The object of criticism is in both (23) and (24) a proposition referred to by the anaphoric pronoun das ‘that’ (in the context of (23), it is the fact that the team played long balls in the second half of the match, in the context of (24), it is the fact that the TV sets were mounted high on the walls of the hotel rooms). Disapproval of these facts is reported to have been conveyed by, primarily, the adjective phrase völlig falsch ‘totally wrong” and the adjective kinderfreundlich ‘child-friendly’, respectively – but note that in (24), this positive predicate is, essentially, literally called into question by the interrogative. 3.

Object or content of judgment: Semantic properties

The Object–Content alternation in the propositional argument of behabitives has some semantic correlates; in the lexical material making up the argument; first and foremost, the presence of a negative (for verbs like criticize) or positive (for verbs like praise) evaluative element is a condition for the C case to obtain. In this section, we explore the various forms this element can take, and we ask whether a given argument can evince a C case / O case ambiguity, depending on how a certain element is interpreted. We also investigate the ways in which the object of judgment can be expressed in or inferred from a C case argument, distinguishing between a “big” and a “small” C case; in examples of the latter type, the propositional argument practically consists in the evaluative element (from here on also: the EE) (to this we turn in Section 3.3). 3.1. Evaluative elements (EEs): Clear and less clear cases The EEs encountered so far – more exactly, from Section 1 through Section 2.5 – are relatively easy to identify: act wrongly, voreilig ‘prematurely’, immoral, the degree modifier zu ‘too’, unethisch ‘unethical’, nicht genügend ‘not sufficiently’, the degree modifier too, an eyesore. In the cases of mixed quotation considered in Section 2.6, quotation marks help identify the EEs as very brutal, a decade, and Großartiges ‘something great’, and in the examples of direct

Behabitive Reports

95

quotation, the EEs are völlig falsch ‘totally wrong’ and kinderfreundlich ‘child friendly’. The degree modifier too is a particularly clear case, in more than one regard.4 It is clearly evaluative, as it is (or necessarily interacts with) a (possibly covert) modal dependent on a normative ordering source, and it has a nonevaluative, “objective” counterpart in the locution as … as … (actually) … , so that the corresponding O case is closely comparable to the C case: (25)

a. b.

Dostoevsky criticized the book for having too many psychological details. Dostoevsky criticized the book for having as many psychological details as it did.

As noted in Section 1 in connection with (1) and (2), the C case example and the O case example, sharing much of their lexical material and their syntax, have roughly the same meaning here. As we will see in Section 3.2, however, it is far from always so easy to detach the EE from the propositional argument, reconstructing the O case proposition from the rest. Nor is it always easy to decide whether a given element is evaluative, or evaluative in the sense of inducing a C case reading of the propositional complement. Consider the predicate being a Socialist – as (26) suggests, it can be evaluative and induce a C case reading, but as (27) suggests, it can also be nonevaluative and form part of an O case complement – and only the broader context, plus world knowledge and cultural knowledge, can give a definitive answer. (26) (27)

Obama has been criticized for being a Socialist. Luxemburg was not only criticized for being a Socialist – she was assassinated for being one.

As is obvious to anyone familiar with U.S. politics from 2008 onwards, (26) is not intended to presuppose that Obama is a Socialist, but just as obviously, at least to anyone faintly familiar with early 20th century European history, (27) is intended to presuppose that Luxemburg was a Socialist. Many EEs can be classified as so-called Predicates of personal taste (cf. Lasersohn 2005 and Stephenson 2007), like eyesore in (18) or hässlich ‘ugly’ in (28). (28)

“Ich finde Noldes “Nadja” hässlich”, kritisiert Lea. “I find Nolde’s “Nadja” ugly”, criticizes Lea.

4. See Meier (2003) for a thorough treatment of this modifier and its dual enough.

96

Cathrine Fabricius-Hansen and Kjell Johan Sæbø

Such predicates are said to depend, for their intension or extension, on a judge, where some think of the judge as a semantic index while others regard it as a contextual parameter. On the basis of “subjective” attitude verbs like finden ‘find’ in (28), Sæbø (2009) argues that the judge index or parameter is relevant not only for Predicates of personal taste but also for, in particular, expressions like the modal modifier too, a prominent EE in C case reports of criticism. The emerging picture is that EEs inducing C case readings span a wide range, from relatively ‘hard-wired’ (second-order) negative judgment predicates like wrongly and immoral through modal expressions like too and insufficient and personal taste predicates like ugly and eyesore to ‘labile’ elements like Socialist where evaluativity is context- and culture-dependent, as well as expressions like a decade, where so-called Expressive meaning (cf. Potts 2007) may play a role, insofar as an attitude is conveyed by a particular lexeme signalling something over and above its contribution to the content at issue or to truth conditions. Note that on the one hand, intersubjectivity does not appear to be necessary – the speaker of (26) may herself not associate anything negative with Socialist – but on the other hand, it seems necessary, if nothing else (such as German V2 or subjunctive, a predicative phrase, quotation mixed or pure) excludes the O case, for speaker and hearer to be aware that the subject of the report is likely to associate something negative with the element at issue. 3.2. “De Re” readings of O case VPs Sometimes, the speaker chooses to represent the object of criticism in a guise that the subject would not subscribe to. The proposition the subject is reported to convey her / his disapproval of is not the proposition resulting from a regular, de dicto reading of the embedded (gerund) clause, but one resulting from what seems to be a de re reading of that clause. (29) is a case in point. By contrast, the clause in (30) is probably to be interpreted in the customary, de dicto way. (29) (30)

The Senator is criticizing Obama for keeping his top domestic promise. The reporters are being criticized for keeping their promise to protect their source.

Regarding (29), however, the proposition the Senator is reported to be conveying her disapproval of is not the proposition that Obama is keeping his top domestic promise (she might not be aware that he is) but – for example – the proposition that he is reforming the health care system. In the actual world, these two VPs – reforming the health care system and keeping his top domestic promise – are coextensional, they are two descriptions of one action.

Behabitive Reports

97

There would seem to be two main ways to model this intuition. • In analogy with the treatment of de re readings of nominals offered, e.g., by von Fintel and Heim (2009), index the VP keeping … promise (VP1 ) with a variable w0 for the actual world so that this proposition ensues: { w : Obama is in w the agent of the event that VP1 describes in w0 }. • In analogy with the treatment of de re readings offered by Maier (2006), substitute the context-determined, coextensional VP reforming … system (VP2 ) and read that de dicto so that this proposition ensues: { w : Obama is in w the agent of the event that VP2 describes in w }. As far as nominals (mostly, definite and indefinite descriptions) are concerned, the first option is standard in natural language semantics, while the second, ‘relational’ type of analysis (VP2 would be the ‘acquaintance relation’ that the subject has to the event) is more at home in philosophy of language. This second option appears to us decidedly more adequate for the case at hand. But this approach has been criticized as being noncompositional (von Stechow and Zimmermann 2004), at least in regard to the belief contexts to which it is commonly applied. However, it might just be possible to emulate a relational theory of de re here in a compositional way, exploiting the fact that VPs like keep his top domestic promise or do exactly the same thing she has just now done as occurring in (31) are in a sense second-order predicates, ‘criterion predicates’ according to Kearns (2003), ‘abstract predicates’ according to Sæbø (2008). (31)

She criticized her husband for doing exactly the same thing she has just now done.

Sæbø (2008) provides an analysis of abstract predicates as having an argument slot for a “concrete”, first-order predicate, e.g., reform the health care system. In line with that analysis, we can assume that a predicate like keep …promise (=  ) may presuppose a predicate P such that P is his top domestic promise: [[]]v = e : P|Kv (P). Pv (e) (notation: e: presupposition. content ; v for the ‘actual’ world ) The de re interpretation of the overt, abstract predicate  thus reduces to this: K exhausts its role in restricting the presupposed covert, concrete predicate P, extensionally; once P is embedded in a context like criticize … for, it is read de dicto. P mimics the ‘acquaintance relation’ in a relational analysis of de re, and the point made by Maier (2006) that this relation is anaphoric is mirrored

98

Cathrine Fabricius-Hansen and Kjell Johan Sæbø

in that P is presupposed. – In the context of the present paper, however, this suggestion must remain on a speculative note. 3.3. “Big” and “small” C case Propositions Sometimes, the object proposition can easily be reconstructed from the content proposition since it is expressed by a part of what expresses that proposition. Notably, this is the case when, as in (2), repeated here, a specificational in or by gerund phrase modifies an E(valuative) E(lement) predicate. (2)

John criticized Herod for acting wrongly in marrying Herodias.

The object proposition is obviously the fact that Herod has married Herodias. Another case in point is the too locution, discussed in Section 3.1 and exemplified here by (32). (32)

My mother criticizes me for becoming pregnant again too soon.

The object proposition is the fact that I have become pregnant again as early as I have. However, the O case is not always predictable from the C case; as often as not, the object proposition is not explicitly, only implicitly present in the C report. In fact, this correlates with the measure of generality in the EE predicate: if it is very general, as in (2), it is difficult to leave the object proposition out: (33)

? John criticized Herod for acting wrongly.

But if it is more specific, as in (34a), the specificational phrase can be left out: (34)

a. b.

Bush condemned the media for harming national security by revealing details of the program. Bush condemned the media for harming national security.

It cannot be left out in just any context, of course, only in a context providing a specification of the object proposition; we need to have been informed by what salient action the media were condemned for harming national security. Recall from Section 3.2 that predicates like keep a promise or break a promise have been described as ‘criterion predicates’ (Kearns 2003) or ‘abstract predicates’ (Sæbø 2008), as predicates of other predicates rather than predicates of events. Now act wrongly is a typical criterion predicate, and conversely, keep or break a promise would be a good EE in a C case behabitive report. Sæbø (2008) uses ‘abstract predicate’ as a cover term for criterion predicates on the

Behabitive Reports

99

one hand and ‘manner-neutral causatives’ – such as harm national security – on the other. In fact, the following generalization seems to be justified: when the object fact is left implicit, the EE is an abstract, second-order predicate; and conversely, whenever the EE is an abstract predicate, the object fact can be added in the form of a by or in gerund phrase modifier. According to Sæbø (2008), the task of such a modifier is to specify the more concrete predicate that the abstract predicate alludes to and predicates over. In the present context, its task is to supply the objective fact when in a C case the main predicate is an abstract – criterion or manner-neutral causative – negative evaluative predicate. But it is often left to the context to supply the objective fact. (34b) was a case in point, (35) is another. (35)

Southern women were criticized for jeopardizing their reputations.

The object of criticism – what Southern women did such that according to some, in so doing they were jeopardizing their reputations – is missing and must be retrieved from the context of utterance.5 In other words, it acts as an anaphoric presupposition, and it is natural to connect this behavior to the presupposition of factivity that the behabitive triggers concerning the O case. In our opinion, these observations hold a key to understanding the oscillatory behavior of behabitives and to a precise analysis. Both the object, O case fact and the content, C case proposition can be left unexpressed in the construction – one is if the other is not. Now if the O(bject) case fact is left out, it gets a definite interpretation, while if the C(ontent) case proposition is not expressed, it receives an indefinite interpretation, – and this correlates with the factivity of the behabitive verb with respect to the O case and the O case only. 4.

Analysis: One slot, two roles

Behabitives consistently assign two different propositional actants, O and C, although for syntactic reasons, only one can be expressed in any given sentence: there cannot at the same time be, in English, two for gerund phrase arguments, or, in German, two dass clause arguments. Thus in regard to proposition-type arguments, these verbs are semantically binary while syntactically unary. As we have seen, various formats and environments constrain the alternation, discouraging or even excluding one of the two from appearing in a construction. Some of these constraints should come out as consequences of a formal analysis. 5. In the actual context, they served as nurses in the Civil War.

100

Cathrine Fabricius-Hansen and Kjell Johan Sæbø

But first of all, the analysis must make a connection between the two roles so that (1) there is no reason to assume a systematic ambiguity in behabitives, and (2) the distinct semantic properties of the two roles are sufficiently motivated. Such a connection will have to consist in a precise conception of what we have loosely described as “saying something, thereby conveying one’s (dis)approval of something else”, where the “something” is of course the content proposition (for convenience: ) and the “something else” is the object fact (). Our first move is to model disapproval of a fact  in terms of a modal notion, defining that if you disapprove of , then in view of your ideals,  should not have been the case (N for necessity, f for the modal base and g for the ordering source in the theory of Kratzer (1981); ignoring the world index here): (36)

N f ,g ¬ where f yields the relevant facts and g yields your ideals (you disapprove of : in view of your ideals, non- is a necessity)

Approval, relevant to positive behabitives like praise, we may model as N f ,g . Our next step is to define your conveyance of this by claiming a proposition . Here, we may be simplifying somewhat, but this is at any rate our proposal: (37)

 ( f ,g ) (y) ∧ f ,g ⇒ N f ,g ¬ (you convey your disapproval of  by claiming : you claim entails that in view of your ideals, non- is a necessity)

and

 (for dicendi) stands for a verb like claim, y is you, ⇒ is for strict implication (though that may be too strict), and the conversational background parameters f and g are superscripted to to take care of the many (maybe all) cases where this proposition is (itself) modalized. As before, g yields your ideals. The next step is to build a behabitive from this. For simplicity, we first consider the German kritisieren as embedding a dass ‘that’ clause, – and the O case: (38)

kritisieren* = ye : . ∃  (

f ,g )

(y) ∧

f ,g

⇒ N f ,g ¬

Relevant examples are (4) and (10a). The pattern can be transferred to English criticize (˛ for P) by adding ˛ and substituting P for  and P(˛) for . (4)

Einstein kritisierte, dass Bohr die Kausalität aufgegeben hatte. ‘Einstein criticized Bohr for having abandoned causality.’

Behabitive Reports

101

(38) is of course only adequate for the O case; the content proposition is introduced by existential quantification. It would seem that we need a distinct C case definition, one that might look like (39): (39)

kritisieren* =  ye : ∃ | .  (

f ,g )

(y) ∧

f ,g

⇒ N f ,g ¬

As in (38), the object proposition  is a factive presupposition (recall from 3.2 the notation: presupposition. content ); here, however, since it is not an explicit but an implicit argument, it is, in line with the general analysis of zero arguments developed by Sæbø (1996), introduced in the presupposition and so acquires an interpretation as a zero anaphoric pronoun.6 How about joining the two definitions, (38) and (39), into one? From a semantic viewpoint, the verb has two propositional arguments, while from a syntactic viewpoint, it has one and only one obligatory argument beside the subject, in the sense that there must be one and only one clausal complement present, – and both arguments are optional as it is unspecified which is expressed. This much must be stipulated: exactly one clausal complement is mandatory. For the rest of the story, general principles about unexpressed arguments apply: Arguments can be zero; a zero argument can have an indefinite or a definite interpretation; if it occurs in a presupposition triggered by the verb, it tends to have a definite interpretation, if not, it tends not to, as shown by Sæbø (1996). With these considerations in mind, let us amalgamate the two cases, the O case and the C case, into one: (40)

kritisieren* =  ye : .  (

f,g)

(y) ∧

f,g

⇒ N f , g ¬

If the  argument is saturated, the argument is prevented from being so; it is then, in accordance with the general principle just cited, introduced at the level of assertion and interpreted as an indefinite propositional referent; if, conversely, the  argument stays unsaturated, this argument is, again in accordance with the general principle, introduced at the level of presupposition and interpreted as a definite propositional referent; then the argument has to be saturated. What remains unconventional – and slightly problematic – in connection with (40) is the assumption that the first argument can remain unsaturated although a proposition, say p, is sister to the verb. General principles cannot tell us how – whether (1)  is closed off prior to the merge (maybe an empty node is sister to the verb instead of p), regardless of p’s aptitude for the  or 6. For perspicuity, we employ a mixture of DRT and, say, Update Logic notation here; the quantifier ∃ may be interpreted as introducing a discourse referent.

102

Cathrine Fabricius-Hansen and Kjell Johan Sæbø

role (in this case two readings are composed, one more, one less reasonable), or whether (2) it is only closed off in the merge, upon inferring that p is not apt for the  role; p is then passed on to serve as argument for the  function. We see (dis)advantages with both strategies, but we would opt for the former. This means that the semantic composition of the verb and the propositional argument that comes to the surface proceeds in two stages: First, the verb merges with a node which can be empty, 0, / so that no proposition is sister to the verb, or nonempty; second, this merge merges with a node which is empty if and only if the first node was nonempty. The pattern in (40) can be transferred to the English valency frame criticize ˛ for … (or the corresponding German valency frame) by adding ˛, substituting P for  and Q for  , and replacing  by P(˛) and by Q(˛). As for the minimal valency frame illustrated in (41), with a type e complement and nothing more, we note that when both the P and the Q argument remain unsaturated, the interpretation that ensues is, or tends to be, indefinite on both counts, so to speak: even the object fact is not interpreted anaphorically. The context may specify what Anne criticizes Bill for, but that is hardly required. (41)

Anne criticizes Bill.

This, inter alia, sets criticize apart from blame, where a missing for complement does have an anaphoric interpretation. While we have no ready explanation for this, the way we would analyze the case is, as suggested in Section 2.5, to regard as the object fact the proposition that Bill has the properties he actually has: (42)

∃ v ( f , g ) (a) ∧ f , g ⇒ N f , g ¬w ({P| Pw (b)} = {P| Pv (b)}) (Anne says something entailing that in view of her ideals, Bill should not have had all the properties he in fact has)

The entailment that Bill should not have had all the properties he has is very weak, of course, the proposition that he has all the properties he has being so very strong; this reflects the highly underspecified nature of the criticism if all that is known about the object fact is a type e participant. The unitary definition (40) seems to us to fulfill the criteria for a satisfactory analysis set up at the beginning of this section: it makes a connection between the two roles so that (1) there is no reason to assume a systematic ambiguity in behabitives, and (2) the distinct semantic properties of the two roles – one an objective fact, the other an evaluative claim – are sufficiently motivated. In addition, the definition goes a long way toward explaining the constraints on this or the other argument role imposed by certain formats and environments,

Behabitive Reports

103

as described in Section 2. In particular, the fact that a V2 complement clause is incompatible with the O case can be traced to the factive presupposition that concerns this case, and the fact that a reportive subjunctive clause entails the C case can be tied to the fact that this case satisfies, through the  predication, the presupposition associated with the Referatkonjunktiv. 5.

Concluding remarks

We leave several issues unresolved. Most importantly, our suggestions fall far short of doing justice to the full range of relevant facts, in three dimensions: • the spectrum of constructions, some of which – e.g., those discussed in Section 2.5 – by their very nature exclude one of the two argument roles encoded in (40) • the spectrum of verbs of judging, which, as shown by Fillmore (1971a), display a variance in various ways and most of which would require an augmentation of (40) • the spectrum of languages: we have considered two, closely related ones; it is anybody’s guess to what extent the languages of the world exhibit a lexical item definable by (40) Moreover, there are some subtler aspects of the semantics of a verb like criticize which we have glossed over, foremost the aspect of agentivity – or responsibility. There is an implication that somebody is responsible for the fact, and if you criticize ˛ for something, you place that responsibility on ˛ or, if ˛ is inanimate but an artefact, as in (17) or (18), on the one responsible for it. Our definition (40), however, predicts (mutatis mutandis) that (44) can be inferred from (43): (43) (44)

Camilla says that ManU should have won. Camilla criticizes ManU for not winning.

(43) entails that Camilla says something that – in view of her ideals – entails that N f , g ¬ where g yields her ideals and  is that ManU did not win. But (44) means something more, which our analysis is not fine-tuned enough to capture.7 If it were amplified to say something about agency, this would offer a way to treat the German valency frame seen in (45), where there are both a DP and a finite clause complement, providing a role for the otherwise redundant DP: 7. A German sentence Camilla kritisiert, dass ManU nicht gewonnen hat does not necessarily ascribe to Camilla the attribution of responsibility to ManU but maybe to the referee.

104 (45)

Cathrine Fabricius-Hansen and Kjell Johan Sæbø

Niemand kritisiert Sie, dass Sie erst jetzt darüber nobody criticizes you that you only now itabout schreiben. write ‘Noone criticizes you for not writing about it earlier.’

That all said, it needs to be emphasized that to the best of our knowledge, the ambivalent, Janus-faced nature of verbs of judging, behabitives, has not been brought to light in earlier work, and that this object/content oscillation has interesting implications in at least two directions. First, for a theory of subjective or expressive meaning, or of so-called personal taste, the ‘evaluative elements’ that are constitutive of the ‘content case’ form a class that can inform the theory from an independent angle. Second, a theory of quotation must be equipped to deal with cases where the quotational status of a proposition is in no way explicitly marked but must be inferred from a variety of sources. References Austin, John L. 1962

How to Do Things with Words. Oxford: Oxford University Press.

Büring, Daniel & Katharina Hartmann 1995 All right! In On Extraction and Extraposition in German, U. Lutz and J. Pafel (eds.), 179–211. Amsterdam: Benjamins. Brandt, Margareta et al. 1992 Satztyp, Satzmodus und Illokution. In Satz und Illokution, I. Rosengren (ed.), 1–90. Tübingen: Niemeyer. Cappelen, Herman & Ernie Lepore 2005 Varieties of quotation revisited. In Hybrid Quotations, P. de Brabanter (ed.), 51–75. Amsterdam: John Benjamins. Fabricius-Hansen, Cathrine & Kjell Johan Sæbø 2004 In a mediative mood: The semantics of the German reportive subjunctive. Natural Language Semantics 12: 213–257. Fillmore, Charles 1971a Types of lexical information. In Semantics. An interdisciplinary reader in philosophy, linguistics and psychology, D. Steinberg and L. Jakobovits (eds.), 370–392. Cambridge University Press. Fillmore, Charles 1971b Verbs of judging: An exercise in semantic description. In Studies in linguistic semantics, C. J. Fillmore and D. T. Langendoen (eds.), 273– 290. New York: Holt, Rinehart and Winston.

Behabitive Reports

105

von Fintel, Kai & Irene Heim 2009 Intensional Semantic. Lecture notes, MIT. Gärtner, Hans-Martin 2002 On the force of V2 declaratives. Theoretical Linguistics 28: 33–42. Geurts, Bart & Emar Maier 2005 Quotation in context. In Hybrid Quotations, P. de Brabanter (ed.), 109–128. Amsterdam: Benjamins. Kearns, Kate 2003

Durative achievements and individual-level predicates on events. Linguistics and Philosophy 26: 595–635.

Kratzer, Angelika 1981 The notional category of modality. In Words, Worlds and Contexts, H.-J. Eikmeyer and H. Rieser (eds.), 38–74. Berlin: Walter de Gruyter. Lasersohn, Peter 2005 Context dependence, disagreement, and predicates of personal taste. Linguistics and Philosophy 28: 643–686. Maier, Emar 2006 Meier, Cécile 2003 Müller, Gereon 1995

Potts, Chris 2007 Reis, Marga 1997

Belief in Context: Towards a unified semantics of de re and de se attitude reports. PhD diss., University of Nijmegen. The meaning of too, enough, and so … that. Natural Language Semantics 11: 69–107. On extraposition and successive cyclicity. In On Extraction and Extraposition in German, U. Lutz and J. Pafel (eds.), 213–243. Amsterdam: Benjamins. The expressive dimension. Theoretical Linguistics 33: 165–197. Zum syntaktischen Status unselbständiger Verbzweit-Sätze. In Sprache im Fokus, C. Dürscheid, K. H. Ramers and M. Schwarz (eds.), 121–144. Tübingen: Niemeyer.

Sæbø, Kjell Johan 1996 Anaphoric presuppositions and zero anaphora. Linguistics and Philosophy 19: 187–209. Sæbø, Kjell Johan 2008 The structure of criterion predicates. In Event Structures in Linguistic Form and Interpretation, J. Dölling, T. Heyde-Zybatow and M. Schäfer (eds.), 127–147. Berlin: Walter de Gruyter.

106

Cathrine Fabricius-Hansen and Kjell Johan Sæbø

Sæbø, Kjell Johan 2009 Judgment ascriptions. Linguistics and Philosophy 32: 327–352. von Stechow, Arnim & Ede Zimmermann 2004 A problem for a compositional account of de re attitudes. In Reference and Quantification: The Partee Effect, G. Carlson and F.J. Pelletier (eds.), 207–228. Stanford: CSLI Press. Stephenson, Tamina 2007 Judge dependence, epistemic modals, and predicates of personal taste. Linguistics and Philosophy 30: 487–525. Sudhoff, Stefan 2004

On the Syntactic Status of German es as a Correlate of Embedded Clauses. Manuscript, University of Leipzig.

Truckenbrodt, Hubert 2006 On the semantic motivation of syntactic verb movement to C in German. Theoretical Linguistics 32: 257–306. Vikner, Sten 1995

Verb Movement and Expletive Subjects in the Germanic Languages. Oxford: Oxford University Press.

Double-duty Quotation, Conventional Implicatures and What Is Said* Manuel García-Carpintero

1.

Introduction

In my previous work on quotation (1994, 2004 and 2005 – this paper partly overlaps with and partly overrides the last of these three publications) I have defended a version of Davidson’s (1979) Demonstrative Theory, which I call the Deferred Ostension theory (DO henceforth) given its affinities with Nunberg’s (1993) general views on demonstratives. With respect to what Recanati (2001) calls open quotation and I, following Davidson, call double-duty quotation (because the descriptive suggestions of the latter expression are, I think, more theoretically appropriate to the nature of the phenomenon than those of Recanati’s), as in (1) below (from the New York Review of Books, May 27 2004, p. 10), I argued that it is entirely consistent with DO: (1)

Saddam Hussein did not pose an “immediate threat to the security of our people,” as Secretary of Defense Rumsfeld said.

In what follows, I will outline the main aspects of DO and the account it suggests for double-duty quotation. To do so, I will appeal to an account of conventional implicatures. Although I will mention at some points rival accounts of quotation, I will mostly refrain from making fully-fledged comparisons or otherwise arguing for what I in fact think is the case, that is, that DO provides the theoretically more systematic and simpler account of the evidence; to a large extent, I have already argued for this in the previous papers. Here I will just take up * Financial support for my work was provided by the DGI, Spanish Government, research project HUM2006-08236 and Consolider-Ingenio project CSD2009-00056; through the award ICREA Academia for excellence in research, 2008, funded by the Generalitat de Catalunya; and by the European Community’s Seventh Framework Programme FP7/2007-2013 under grant agreement no. 238128. Thanks to José Luis Prades and an anonymous referee for helpful discussion of some topics in this review, and to Michael Maudsley for the grammatical revision.

108

Manuel García-Carpintero

Cappelen and Lepore’s (2007) recent proposals on quotation and double-duty quotation. 2.

The deferred ostension view of quotation

On a plausible, Reichenbach-inspired account of indexicals, there are tokenreflexive rules conventionally associated with demonstrative types, which distinguish semantically, say, ‘he’ from ‘you’. In virtue of them, when a competent speaker uses a token demonstrative, by default he thereby manifests an intention to refer to an entity of a given conventionally indicated type (male, in the case of ‘he’), in virtue of the fact that it stands in some contextually specific existential relation (Peirce’s term) to the token he has produced, made manifest by the speaker’s communicative referential intentions.2 With demonstratives, sensible manifestation of those intentions will typically require the speaker to carry out a further case of some standardized intentional action (a demonstration: a certain pointing gesture, say) to instantiate the contextually specific existential relation determining the referent. This is because, in contrast to so-called pure indexicals, in typical contexts including the demonstrative there are several objects satisfying the conventionally indicated type. In some expedient cases, a unique such individual is already manifest, no demonstration being needed; this is in general the case with pure indexicals. When a successful demonstration occurs, it picks out an object, the demonstratum, typically coinciding with the referent. In cases of deferred ostension, the demonstratum is not the referent, because it does not satisfy the conventionally indicated type; it is rather an index (Nunberg 1993), helping to determine the referent in ways not unlike those in which typically the token determines it: in virtue of further contextually specific existential relations, properly manifested by the speaker. By default, the referent (in successful cases) is the demonstrative’s contribution to the proposition constituting the main speech act it helps to convey. The mutually known descriptive condition that must also exist in successful cases rather contributes to characterizing a presupposed propo-

2. Most semanticists, following Kaplan (1989), think that there are philosophical problems with this mention of particulars in a semantic theory, and prefer a more abstract characterization in which the theory quantifies over types in context, as opposed to tokens. In previous work, I (1998) have argued that the alleged advantages of this are illusory, and the alleged problems non-existent. Although we should assume a relative abstract conception of tokens for semantic purposes, it is crucial that referents are determined through existential relations with particulars.

Double-duty Quotation, Conventional Implicatures and What Is Said

109

sition, by specifying the content of the presupposed speaker’s communicative referential intention. All this seems to suitably fit the way quotation works in natural languages according to DO, a version of the demonstrative theory. Consider a written utterance of the sentence (2): (2)

‘Barcelona’ is disyllabic.

According to DO, quotation marks are the linguistic bearers of reference, functioning like a demonstrative; the quoted material merely plays the role of a demonstrated index.3 The referent is obtained through some contextually suggested relation in which it stands to the quoted material; in the default case the relation will be: …instantiates the linguistic expression __, but there are other possibilities determined by speaker’s intentions.4 In this way, DO accounts for the fact that we do not merely refer with quotations to expression-types, but also to other entities related in some way to the relevant token we use: features exhibited by the token distinct from those constituting its linguistic type, features exhibited by other tokens of the same type but not by the one actually used (as when, by using a graphic token, we refer to its phonetic type), other related tokens, poems or songs including the instantiated types as in (3), and so on and so forth.5 (3)

Almost all English schoolboys used to know by heart ‘The curfew tolls the knell of parting day’, but not anymore; now they rather know ‘Imagine’.

3. I will henceforth use quoted material to refer to what is inside the outermost quotation-marks – Barcelona in (2) – reserving ‘quotation’ for the whole constituted by it and the surrounding quotation-marks. 4. Predelli (2008) provides a useful formalization of a demonstrative account. In part to deal with issues raised by Cappelen and Lepore (2007) that I will be discussing later, he argues for a distinction between the character and the meaning of quotationmarks, taken as demonstrative expressions, which I do not find required. Of course, the distinction is always there, in that character is a function – an abstract mathematical representation of a semantic property. However, to the extent that the function represents the semantic instruction whose default case has been described in the main text, I do not think the distinction is needed – in particular, as I argue below, it is not needed to deal with Cappelen and Lepore’s arguments. 5. García-Carpintero (1994: 261) provides more examples; cf. also Cappelen and Lepore (2007: ch. 7).

110

Manuel García-Carpintero

On the contrasting Fregean Identity theory (IT), when an expression is referred to by means of quotation the quoted material itself is the linguistic referring expression. Quotation-marks are not needed; when they are used, they serve to make clearer the shift in syntactic and semantic properties effected on the quoted material by its occupying that linguistic context: whatever its usual syntactic function, the quoted material functions as a singular term; whatever its usual semantic function, in that linguistic environment the quoted material refers to itself. In this way, as Washington (1992) emphasized, IT accounts for those cases (particularly, although not only, in spoken language) where no form of quotation marks is used. DO entails that those sentences (say, a spoken version of ‘cat’ has three letters without any special intonation), missing an expression playing the syntactic role of subject, should sound syntactically defective in languages like English which lack the “pro-drop” feature. They are not perceived to be so, however; the presence of the quoted material is enough for speakers to feel no syntactic solecism. In presenting the two views, I have also indicated what I take to be their main strengths. They also constitute the main problems for the rival view, which typically they deal with by appealing to the semantics-pragmatics divide. Thus, for instance, the defender of DO will argue that, while the semantic referent of the subject in spoken utterances of (2) without quotation-marks is a city, and therefore what is said in those utterances is false, typically the speaker’s referent will be an expression-type, and the thereby conversationally implicated content true.6 The defender of IT, on the other hand, will typically argue that, while the semantic referents of quotations are always the very same expression-types they include as parts, their speaker’s referents may well be on many occasions other related entities like those previously mentioned – poems and so forth. I would like to emphasize for later use a point I have been careful to make in presenting the two views, namely, that they differ as regards what they count as the linguistic referring device in quotation. DO does not deny that the quoted material counts also as a referring sign, according to ordinary intuitive conceptions of reference. For we intuitively describe indexes in regular cases of deferred ostension as referring to whatever they help refer to. Moreover, as we just saw, in those cases in which no quotation marks are used, according to DO they are the vehicles of speaker’s reference. The real difference between DO and IT lies in that, according to the former view, in the strict sense of referring device in which only tokens of linguistic types that have a referring function in the system of a natural language, it is tokens of quotation marks that are the referring devices in quotation; only they convey semantic reference. According 6. See Cappelen and Lepore (1999: 742) and (2007: 35–43), and my (2004).

Double-duty Quotation, Conventional Implicatures and What Is Said

111

to IT, it is rather the token of the quoted type that is the referring device, in that strict sense. If, as happens in (2), the quoted type is itself conventionally a referring device, there is a systematic ambiguity involved. Quotation marks, or alternative contextual resources, help to disambiguate. At first sight, quotation marks look much more like punctuation marks with a disambiguating role of this kind than like fully-fledged linguistic referring devices. The impression is even stronger if it is kept in mind that devices like italicization and, in spoken language, some sorts of intonation are (in my view, at least) among the different shapes that quotation marks can conventionally adopt.7 I share the feeling. But no theoretically compelling argument against DO can be based on it. For the assumption behind such an argument should be that only word-like expressions – lexemes – have semantic roles. But, of course, we indicate, say, co-reference not only with lexemes (anaphoric expressions), but also by using tokens of the same type, and the relation being of the same type is not a lexeme; we indicate thematic roles with lexemes (declensional inflection), but also by means of syntactical relations that are not lexemes at all; we indicate focus by means of intonation, and so on and so forth. And what we indicate in all these variegated ways are crucial semantic features of natural languages.8 Nonetheless, there are important objections to DO. Some by Gómez-Torrente (2001) are among the most ingenious I have recently seen. I have offered replies to them in my (2004). All in all, as I argue there, I still think that DO is the best contender in the field. I will not rehearse here the reasons. I will just mention what I take to be the main problem for Reimer’s (1996) version of IT, which adopts aspects of DO; and then I will discuss Cappelen and Lepore (2007)

7. I understand that linguistic expressions are abstract entities, which might have, as it were, different physical embodiments. In this context, I am assuming that italicization and certain patterns of intonation are just two among the different devices for quotation, in addition to the many varieties of graphic marks that languages use. Perhaps this is a good point to note that in my view there is no non-theoretical way of telling what quotation is; it is for theoretical accounts of quotation to characterize the phenomenon. Of course, in pre-theoretical terms we can point to paradigm examples of the phenomenon we want to account for, such as (1) and (2), mentioning the pre-theoretical data that must be explained, including the facts we have already mentioned and those that will come out later, such as the “picturing” character of quotation. 8. I do not assume that all semantic devices alternative to lexemes conform a natural class of their own. The only relevant class is that of expressions with a semantic role, be they lexemes or not.

112

Manuel García-Carpintero

recent proposal (‘LTOI’ henceforth), which has some affinities with GómezTorrente’s views. Reimer’s view shares with orthodox versions of IT the claim that the quoted material is the linguistic referring expression. Her version of IT differs from previous ones concerning the semantic subcategory of singular terms inside which this linguistic expression referring to itself is to be included: according to her proposal, the quoted material is a demonstrative; on her view, quotation marks play the role of a demonstration. This accounts for their absence – this occurs when the referential intentions of the speaker are sufficiently clear so that a demonstration is not required, and also for the fact that quotation marks are more needed in writing – because then context does not make the speaker’s intentions so perspicuous (Reimer 1996: 140). However, this account of quotation does not fit the way demonstratives work in natural language. Two factors are needed to account for the semantics of indexicals and demonstratives: a linguistic rule associating distinct conventionally indicated types with distinct linguistic types (one, say, for ‘he’ and another for ‘you’), accounting for the semantic commonalities among different uses, and a truth-conditional import assigned to their tokens, or contextualized types, accounting for the potential differences in truth-conditional import. Given that Reimer’s proposal is a version of IT, on her view there cannot be anything playing the role of the first element common to all cases of mention; there cannot be a common rule associated with, say, the quoting expressions in (2) and (4): (4)

‘Hombre’ is disyllabic.

On any identity view, the referring expressions in these two cases are different. Therefore, there is not a type instantiated in all cases of quotation, with which a common linguistic rule is associated. Given Reimer’s account, whenever tokens of different expression-types are used for mentioning, different demonstrativetypes are also involved; otherwise, it would not be correct to say that the token of ‘Barcelona’ in an utterance of (2) lacking quotation marks is a linguistic expression that, by default, refers to itself, and the view would not be a version of IT. According to Reimer’s view, thus, every natural language includes as many demonstrative-types as it includes expression-types. Indeed, in view of the fact that we can use for indexical mention graphic and spoken material belonging to other languages, as in (4) – or even belonging to no used language – Reimer’s view entails that types they instantiate are further demonstrative expressions in English. I do not think this is correct. The introduction of expressive devices like quotation marks reflects the perception that a common semantic procedure ap-

Double-duty Quotation, Conventional Implicatures and What Is Said

113

plies in all cases, which disregards the type of the quoted material: whenever any expression is quoted, a common token-reflexive “picturing” rule operates. Whenever ‘you’ is used, the same indexical rule operates, one different from the indexical rule operating whenever ‘he’ is used, associated with a specific conventionally indicated type, being the addressee. Reimer’s proposal has it that, analogously, whenever ‘Barcelona’ is quoted, the same indexical rule is operating, one different from the rule operating when ‘hombre’ is quoted, a rule which might then be associated with a specific conventionally indicated type. This is not the case. The same indexical rule, associated with a unique conventionally indicated type – something like being an expression, in a broad understanding of that notion, a determinable to be determined in context – operates when we quote any expression whatsoever, even when we quote material that does not instantiate any expression-type of our language. The models we have for indexicals suggest that the indexical rule operating here is one associated with a type that is present no matter what the quoted type is; quotation-marks are the obvious candidate.9 I will further elaborate on the comparative merits of DO by comparing it with the recent proposal in Cappelen and Lepore’s (2007) new book on the topic. Here they abandon their former Davidsonian view, in favour of a certain Minimal Theory (MT henceforth). They appear to have been impressed by one of the objections to Davidsonian theories in Gómez-Torrente (2001) that I alluded to before,10 and to have built MT in part as way of accommodating it. However, as I will show, their remaining sympathies towards Davidsonian accounts compel them to modify Gómez-Torrente’s view in ways that impair it considerably; in addition, for a book-length discussion of the topic (the first in print, they proudly declare: op. cit., 16), their presentation of MT leaves matters unclear at crucial points. The criticism by Gómez-Torrente to which Cappelen and Lepore appear to be sensitive appeals to the intuitive truth of disquotational schemas for quotations such as the one this instantiates: (5)

‘ ‘Socrates’ ’ stands for (refers to, denotes) ‘Socrates’.

The problem that (5) poses to Davidsonian theories such as DO is as follows. According to the Davidsonian, it is just quotation marks that are linguistic expressions, susceptible of being in the standing for or referring to relation; whole quotations are not, because they include what in fact is a mere extra-linguistic 9. I take this to be the core problem for Reimer’s proposal. Caplan (2001) presents more elaborated objections. 10. This is just my personal impression; they in fact do not acknowledge it.

114

Manuel García-Carpintero

index that helps determining the referent. The expression quoted by the grammatical subject of (5) is a whole quotation, the opening quote followed by ‘Socrates’ followed by the closing quote. But that, according to the Davidsonian account, is not a term being in the standing for relation to anything – only part of it, the quotation marks, are; and, all by themselves, out-of-context, they do not refer to ‘Socrates’ or to anything else.11 Cappelen and Lepore appear to have been impressed by this argument, so much as to make a schema close to the one that (5) exemplifies (I will comment on the differences presently) the main tenet of their non-Davidsonian new theory MT: (QS)

‘ ‘e’ ’ quotes ‘e’.

MT “is the view that QS is the semantic rule for quotation. It serves as the full semantic treatment for quotation expressions. (…) QS is the fundamental axiom schema governing the semantics of quotation expressions” (LTOI: 124). As we will presently see, the appeal to this “rule” does not make everything about MT as crystal clear as it should be. The reader may have noticed an outstanding difference between QS and the schema that (5) instantiates: QS studiously avoids mentioning the reference relation, invoking instead a sui-generis quoting relation. Cappelen and Lepore (ibid.) note this difference, but do not justify it; they note also a second difference, which perhaps may in part account for the first. Gómez-Torrente (2005: 128) presents the likes of (5) as immediately accounted for by the Tarskian theory of quotation he defends, which characterizes pure quotation as governed by the following semantic Tarskian Principle: “by enclosing any expression 11. In my (2004) I provide a rejoinder that Gómez-Torrente (2001: 134) anticipates, although he (2005: 148) thinks that it puts the Davidsonian in a somewhat less attractive position vis-à-vis the Tarskian proposal he defends – about which I will say something later in the main text. The reply is that the intuitions on which the argument relies are not sensitive enough to the distinction between a properly linguistic expression, part of the expressive system of a particular natural language, and any old expressive resource, a mere sign; the quotation referred to by the grammatical subject of (5) is indeed such a sign, “referring” in an equally extended sense to ‘Socrates’. A theoretical account does not need to honor such intuitions; the only issue is whether, overall, it provides a better explanation of all relevant facts. (It is doubtful whether, as Gómez-Torrente’s rejoinder assumes, ceteris paribus an account would be better off if it captured this, in my view spurious, intuition, but ultimately that does not matter, because cetera are not paria in the full appraisal of the two accounts.) For reasons that I am rehearsing here, DO in my view fares better than its rivals, including Gómez-Torrente’s Tarskian account.

Double-duty Quotation, Conventional Implicatures and What Is Said

115

within quotation marks one gets a singular term – a quotation – that stands for (refers to, denotes) the enclosed expression”. Cappelen and Lepore mention a similar principle offered by Mark Richard when they present MT (LTOI: 124). They note, however, that these principles take the denoting items to be expressions; this, as we have seen, is as it should be, because the relations that Gómez-Torrente mentions, referring or denoting, relate linguistic expressions to other items. Unlike Gómez-Torrente, however, Cappelen and Lepore contend that it is not just expressions that appear between quotation marks in the quotations standing in the relation axiomatically characterized by MT, but, more generally, “quotable items”, including non-expressions. My guess is that they introduce the sui generis unexplained relation of quoting in their axiom QS in part because of this.12 Be this as it may, by relying on QS Cappelen and Lepore cannot apply Gómez-Torrente’s objection for to Davidsonian accounts. For they agree with their Davidsonian past selves that the first terms of the semantic relation that QS characterizes do not need to be linguistic expressions. Unlike GómezTorrente (who instead provides a pragmatic account of the data), they accept part of the evidence pointing to a demonstrative account, to wit, that quotations can refer to (“quote”, they would say) items which are not expressions of the language to which they belong (or indeed any other language – they contend that pictures can be quoted, LTOI: 23), and that quotation appears to manifest context-sensitivity. Now, if we substitute ‘quotes’ for ‘stands for’ in (5), the resulting claim is of course true even on Davidsonian accounts such as DO. I say that they accept that quotation “appears to manifest” context-sensitivity because, in sync with their tirades against contextualist views in their recent work about the semantics-pragmatics divide, Cappelen and Lepore reject that quotations are context-dependent. More specifically, they reject the following principle (LTOI: 68), which DO certainly endorses: (QCS)

Let S be a sentence with a quotation Q, containing no context-sensitive expressions other than possibly Q. Two utterances, u and u , of S can express different propositions because Q in u and in u quotes different items.

This denial notwithstanding, and at first sight paradoxically, they accept some of the data emphasized by demonstrative accounts such as DO that suggest the truth of QCS; for instance, they accept that, while in some context (6) might be 12. I guess that the other part of their motivation lies in the “syntactic chameleonism” that their treatment of mixed quotations ascribes to quotations, which I will discuss in the next section.

116

Manuel García-Carpintero

true, in more ordinary contexts it is rather its negation which is true (and this is a semantic, not merely a pragmatic phenomenon as writers such as GómezTorrente would have it), because at least one of the expressions (say, the one on the right-hand side of the identity sign) does not refer to the linguistic expression ‘Madrid’ but merely to its written manifestation in a specific font, Verdana (LTOI: 77–79): (6)

‘Madrid’ = ‘

’.

Cappelen and Lepore’s (LTOI: ch. 12) way of making their rejection of QCS consistent with their acceptance of this contextualist datum goes as follows. It is not necessarily linguistic expressions that do the quoting that QS characterizes. Quotations might include “quotable items” which, although “signs”, are not expressions. Quotations literally have those quotable items, some of them nonexpressions, as parts. Thus, to get the result that (6) is false, and its negation true, according to QS, the quotable items included in the quotations on one and the other side of the identity sign should differ; but this means that the quotations themselves differ, which is how the rejection of QCS can be upheld: when the quoted items differ, the quotations doing the quoting differ too. As a way of semantically accommodating part of the data favouring contextdependent, demonstrative proposals such as DO, while preserving the contextindependence of an account which relies on QS as the way of capturing what is essential to quotation, Cappelen and Lepore’s suggestion is more of an idiosyncratic change of topic than a theoretically illuminating proposal. QS is supposed to be integrated as an axiom in a semantic theory (LTOI: 130–131). It is, I take it, a common assumption that semantic theories deal with the properties of linguistic expressions. This is, as I emphasized before, for better or worse a crucial assumption behind the Davidsonian claim that only the quotation marks are expressions, akin to demonstratives. All writers I know of make this standard assumption, which linguists would of course make; as we have seen, Gómez-Torrente anti-Davidsonian argument crucially depends on it. On Cappelen and Lepore’s extraordinary view, however, the lexical items of natural languages on which semantic axioms such as QS range comprise in fact all kind of “signs”, including pictures and what have you.13 If we disregard this terminological idiosyncrasy, and interpret ‘expression’ in the same way as everybody else in this debate, Cappelen and Lepore’s view validates QCS – 13. Although the whole proposal is too underdeveloped to validate conjectures, the treatment of the “syntactic chameleonism” to be discussed in the next section might require that “quotable items” in general, including pictures, also have all kinds of syntactic categories.

Double-duty Quotation, Conventional Implicatures and What Is Said

117

unlike accounts truly relying only on a principle akin to QS, such as GómezTorrente’s Tarskian Principle, which do reject QCS and instead account for the intuition that (6) might be false in pragmatic terms. Let me develop this point a little more. QCS has a pre-theoretical interpretation, relative to which it is confirmed by examples such as (6); and a theoretical one that, depending on the theory, either embraces or rejects it. The pretheoretical interpretation appeals to pre-theoretical notions of what expressions are and what identity conditions they have. On this interpretation, ‘Madrid’ is the same quotation in a context in which (6) is true (because it refers to the linguistic expression ‘Madrid’), and in another in which the negation of (6) is true (because it refers to its written manifestation in Verdana). The pre-theoretical interpretation thus confirms QCS. On a theory of quotation that appeals only to Gómez-Torrente’s Tarskian Principle, semantically the two expressions have the same referent in both contexts, and, in general, QCS is false. On a Davidsonian theory, on the other hand, only the quotation marks are (demonstrative) expressions in both contexts, and they do have different referents; thus, QCS is true. Versions of the Identity Theory such as Saka’s or Reimer’s have the same result, with the expressions being the relevant tokens of the quoted material (which has the problematic consequences I highlighted before). Washington’s version of IT, closer to Frege’s I think, only accepts linguistic types as expressions, and then entails the falsity of QCS, exactly as Gómez-Torrente’s view does. The important point to notice is that Cappelen and Lepore in fact do not disagree with any of this. They just introduce a different, idiosyncratic understanding of what an item in the range of a linguistic theory is, relative to which they can declare that they join anti-contextualists in rejecting QCS. But if we understand ‘quotation’ in a non-idiosyncratic way, they are fully in agreement with the contextualist, Davidsonian diagnosis. Thus their rejection of QCS is merely verbal. In my previous work I provided examples showing that we can use quotations to refer not just to types distinct from linguistic expressions (such as, say, the Verdana written version of ‘Madrid’, in the case of the second occurrence of that expression in false utterances of (6)), but also to tokens. Thus, consider for example (7), about which Cappelen and Lepore say that they and their informants find impossible (or very difficult) to find a true reading (LTOI: 72): (7)

‘I’ tastes like peach.

In spite of the difficulties that Cappelen, Lepore and their informants report, it is not hard at all to find a context in which an utterance of (7) appears to be true; just imagine that we are speaking about the items in a bag of sweets in the shape

118

Manuel García-Carpintero

of letters. In fact, at other places Cappelen and Lepore are more prudent and do not deny the intuitive data (“we have no complaint against an audience being directed to a token with an utterance of a quotation expression” (LTOI: 71). However, although Cappelen and Lepore accept a semantic account of the former cases (as we have seen, what they end up providing is a terminologically idiosyncratic version of a contextualist account), they reject it when it comes to tokens, proposing instead to account for it by appealing to one of three familiar pragmatic strategies (LTOI: 76). I have difficulty in understanding the asymmetry. Firstly, if the arguments they provide against demonstrative accounts worked in the case of quotations referring to tokens, they would also apply to the cases of types other than the expression. Conversely, once one accepts as semantic the context-dependence they are prepared to grant (even if accounting for it by means of their terminological way of disowning contextdependence), it seems unprincipled not to incorporate the one they reject. Let us have a look at their arguments, because they will give the reader a fuller view of this new theory of quotation, before we move on to study double-duty quotations. The first argument is related to the one by Gómez-Torrente that we discussed above, but, of course, for the reasons I have given, cannot be exactly like it. They argue that demonstrative views accepting the wide-ranging context-dependence that DO assumes “cannot guarantee the truth of (dis)quotational sentences”, such as instances of QS like (8), because, on such views, it “should be on a par with” (9) (LTOI: 69): (8) (9)

‘ ‘Quine’ ’ quotes ‘Quine’. ‘that’ demonstrates that.

To begin with, unlike Gómez-Torrente’s argument, this one is very weak. There is a significant difference between (9) and (8), which demonstrative views such as DO highlight: instances of (8) always include adequate indexes to act as demonstrata, unlike instances of (9). As a result, what I characterized above as the default demonstrative rule for quotations can always safely operate in any context in which (8) is uttered; on that rule, a quotation refers to the linguistic expression instantiated by the demonstratum. On this interpretation, utterances of (8) are true.14 I think that this explains perfectly well the intuitions of an asymmetry between (8) and (9), and also the intuition that (8) is true “as a matter of meaning alone”, as they put it (LTOI: 70). It is certainly the case that,

14. Cp. the response to Cappelen and Lepore provided by Predelli (2008: 566) on behalf of the version of the demonstrative account that he formalizes there.

Double-duty Quotation, Conventional Implicatures and What Is Said

119

given DO, there are false instances of QS;15 (10) might be a case in point, in an appropriate context: (10)

‘ ‘Quine’ ’ quotes ‘

’.

However, we should not make too much of this; I bet that ordinary speakers would feel confused if we ask them whether an utterances of (10) is true (even if we substitute the more colloquial ‘refers to’ for Cappelen and Lepore’s technical ‘quotes’), and also that many would find it false if we set the context so that it is clear that the second quotation refers to the Verdana version of the expression. But, putting this aside, what makes their argument truly amazing is that, of course, Cappelen and Lepore agree that some utterances of (10) are false. As one would expect by now, they (LTOI: 153–4) resort to their terminological trick to argue that this does not contradict the truth of (8) “as a matter of meaning alone”, because in false utterances of (10) different quotations are at stake: the quotation that is part of the subject is not the same as the one used as object. But this, as shown above, is just the result of verbal stipulations. According to DO, there are false instances of QS, such as an utterance of (10) in the indicated context; and what determines whether we have a true or a false instance of QS is whatever determines the interpretation of demonstratives in context (in my own view, not just any old intention but the referential communicative intentions of the speaker). According to Cappelen and Lepore’s manner of speaking, all proper instances of QS are true; an utterance of (10) could either be an instance of QS, and then true, or false, and then not a true instance of QS; and what determines whether is one or the other is …well, I suppose the (not referential but) “quotational” communicative intentions of speakers. So, once again we just have an idiosyncratic terminology disguising the modest measure of philosophical illumination we have been given, relative to what can already be found elsewhere. I have argued that the anti-contextualist argument is not good, and that, in any case, a similar argument could be mounted against Cappelen and Lepore. Thus, it is very difficult to understand the asymmetry that they claim exists between the case of quotational reference to different types, in addition to the linguistic expression itself, and that of reference to other tokens. The argument that we have rehearsed does not work any better when we have in mind contextual variation due to reference of some quotation to a token, than when we have

15. Remember that what counts as an instance of QS depends on which theory of quotational expressions we appeal to; more about this in what follows.

120

Manuel García-Carpintero

contextual variation due to reference of some quotation to a type other than the linguistic type. Perhaps the ultimate reason for the asymmetry in Cappelen and Lepore’s treatment is that when it comes to tokens, only false pseudo-instances of QS can be found (the context previously suggested to find a true utterance of (7) can be easily invoked to obtain one). By definition of ‘token’, it cannot be the case that the quotation playing the subject role in an instance of QS contains the same token as that playing the object-role. Nonetheless, this does not provide a principled reason to reject that quotations might semantically refer to tokens, but merely a theory-driven one. One, moreover, driven by an unstable theory: if we accept Cappelen and Lepore’s reasons for context-dependence in quotation, context-dependence should be extended to embrace reference to tokens, which points to a Davidsonian account, or to a context-dependent version of IT; if we instead become persuaded of the importance of disquotational axioms for the semantics of quotations by their reasons against contextualism, then we should go for Gómez-Torrente’s Tarskian account, or for the context-independent version of IT. Two other arguments that they provide against demonstrative accounts have exactly the same problems, so we can go through them quickly. (1) Those accounts, they say, “over-generate”; they predict that instances of (11) can be true, exactly as instances of (12) can: (11) (12)

‘a’ =/ ‘a’. that =/ that.

In reply, we can make the same points as before: (i) the default rule explains why the schema feels as if its instances were true “as a matter of meaning”; (ii) there are exceptions, illustrated by contexts where (6) is false, but Cappelen and Lepore themselves can only validate the general truth of (11) by appealing to their terminological stipulation; if we instead use standard terminology they agree that (11) has counter-instances; (iii) to restrict counter-instances to cases involving reference to non-linguistic types is unprincipled. (2) Quotations, they argue, fail to pass two standard tests for context-sensitivity; firstly, the “Intercontext Disquotation Indirect Reporting Test”: If an occurrence of an expression e in a sentence S tends to block disquotational indirect reports (i.e., tends to render them false), then that’s evidence that e, and so S, is (semantically) context-sensitive (cf. LTOI: 73), and secondly the Collectivity Test: If a noun e is (semantically) context-sensitive (i.e. if it changes semantic value from one context of utterance to another, then on the basis of merely knowing there are two contexts of utterance, one in which ‘e is F’ is true and the other in which

Double-duty Quotation, Conventional Implicatures and What Is Said

121

‘e is G’ is true, we cannot automatically infer that there is a third context in which ‘e is F and G’ (cf. LTOI: 74). The three points I have made in reply to the over-generalization argument apply here as well, mutatis mutandis: (i) the appearance that quotations fail the two tests is accounted for by cases in which the default rule applies; (ii) there are cases in which quotations pass the two tests, as Cappelen and Lepore would acknowledge if we frame the cases in the standard terminology invoked by context-sensitive theories; (iii) to restrict an account otherwise accepting context-dependence (when stated in the standard terminology) to cases in which types, not tokens, are referred to by the quotations is unmotivated. I conclude this section by discussing Cappelen and Lepore’s fourth argument. Intuitively, unlike other referential devices, quotations have a “picturing” feature; but, they argue, “[i]t is hard to see how a semantic construal of QCS” can account for this, for “[t]here is no requirement built in that secures any kind of picturing or hieroglyphic relationship between the quotation and its semantic value. Given that the speaker’s intention (together with other facts about what is contextually salient) determines the semantic value of a quotation, it also remains a mystery how we are able to go from a quotation to what it quotes” (LTOI: 71). MT, in contrast, allegedly explains this in that, by sticking to QS, it contends that the semantic value of a quotation is already contained in the quoting expression itself. Now, it is fair to say that, given the variety of quotable items their own account envisages, it is not descriptively adequate to represent this relation between quotation and quoted item as “picturing”; first and foremost, a quotation can hardly be said to “picture” the embedded linguistic expression, as I pointed out in my (1994). I guess this is why Cappelen and Lepore, given their penchant for terminological redescription, characterize the datum as one of proximity rather than “picturing” (LTOI: 71). Now, the very fact that they themselves have to resort to redescription shows that, although there is a fact to be accounted for here – one suggested by the talk of pictures or hieroglyphs – this talk cannot be taken at face value. In my own view, as I explained in my (2004), the best account is provided by the fact that quotations are demonstrations that incorporate their index or demonstratum, so that, in paradigmatic cases where the default rule applies, mere linguistic knowledge is enough to identify the referent. In this section we have considered how four different theories of quotation (IT, DO, Gómez-Torrente’s Tarskian proposal (TT), and Cappelen and Lepore’s related MT) handle the data of (i) the apparent context-sensitivity of quotation, (ii) quotation without quotes, (iii) the apparent truth of sentences like (5) – including a recursively generated one with an additional pair of quotes – and

122

Manuel García-Carpintero

(iv) the apparent pictoricity of quotation. IT and TT provide straightforward accounts of (iii), and explain (iv) in terms of containment or plain identity between referring expression and referent. IT explains straightforwardly (ii), while TT and DO appeal to pragmatics here. DO provides a straightforward account of (i), while IT and TT should appeal to pragmatics here. I have made my preferences clear, but it is ultimately for the reader to evaluate the pros and cons of the relevant theoretical choices. MT, we have seen, is a theory combining features of TT and DO, which is sometimes disguised because of terminological idiosyncrasies. In this section we have not found any reason to prefer it to its models. 3.

Double-duty quotation

Cappelen and Lepore (1997), following Davidson, called attention to uses of quotation marks (mixed quotation, in their terms) as in (1), repeated below, which combine direct and indirect discourse. They argued that a proper account of those cases requires deploying together a Davidsonian demonstrative treatment of quotation, and a Davidsonian paratactic treatment of indirect discourse. In their more recent work (2007), they have also abandoned this view, in ways we are about to examine. (1)

Saddam Hussein did not pose an “immediate threat to the security of our people,” as Secretary of Defense Rumsfeld said.

I agree with Recanati (2001: 660) that mixed quotation is but a particular case of a more general phenomenon, including also among other things scare quotes as (13):16 (13)

A ‘fortnight’ is a period of fourteen days

Recanati describes all those cases as open quotations. I follow Davidson in using ‘double-duty quotation’. It is descriptively apt; for, at an intuitive level of description, the phenomenon consists in that the quoted expressions are both mentioned – i.e., used in order to explicitly call attention to themselves – and used – i.e., used in their standard ways, that of mentioning entities such as period of times in the case of (13). And I prefer to avoid the connotations of Recanati’s term, suggestive of his own account, which I (2005) have criticized elsewhere.

16. Cappelen and Lepore prefer to put aside the discussion of scare quotes (LTOI: 16– 17).

Double-duty Quotation, Conventional Implicatures and What Is Said

123

Cappelen and Lepore (2007: 29) usefully classify theories of double-duty quotation in three groups, which we can descriptively label ‘use-only theories’, ‘dual theories’ and ‘mention-only theories’. Dual theories respect the initial intuition; according to them, “the semantics for a mixed quotation require that the quoted words be both used and mentioned” (LTOI: 29). Use-only theories “deny the initial intuition by saying that in mixed quotation, the semantic content doesn’t recognize quotation marks – at the semantic level the quotation marks are superfluous” (ibid). Finally, mention-only theories “deny the initial intuition by saying that in mixed quotation, the semantic content doesn’t imply that the quoted words are used – only quoted” (ibid). On the latter view, which is the one they advocate now (against their previous selves, proponents of a dual account), “no corresponding indirect report [of Secretary of Defense Rumsfeld, in the case of (1)] follows from the semantic content of [(1)]” (LTOI: 28–29). (As we will see later, it is not that clear that Cappelen and Lepore defend in fact a mention-only account.) Relative to Cappelen and Lepore’s classification, and in contrast to both their new mention-only and their previous dual proposals, my own account of double-duty quotation (which extends Predelli’s (2003) account of scare quotes) is a use-only theory;17 but I should emphasize that this is only relative to their own assumptions on the semantics-pragmatics divide, which – against what I find adequate – place conventional implicatures and lexically-based presuppositions into pragmatics. It will be convenient to examine this issue with more care, before discussing accounts of double-duty quotation.18 Consider first lexically based presuppositions. Without going into the details of how to articulate it into a precise, theoretical account, in my view a correct account of utterances such as (14) would distinguish an asserted content (14a) (the only content counted as semantic on truth-conditional views about the semantics-pragmatics divide such as Cappelen and Lepore’s) and a presupposed content (14b), both of which might well rely on context to fill up the constraints or indications conventionally associated with expression-types: 17. The view I propose is very close to those advanced by (in addition to the already mentioned Predelli (2003)) Geurts and Maier (2005), Gómez-Torrente (2005) and Potts (2007a), particularly the latter. Gómez-Torrente is very careful and merely suggests a view along the lines of the one I propose, while Geurts and Maier advocate a presuppositional account; I defend that the quotational contents in mixed quotations are conventional implicatures and not presuppositions – but Gómez-Torrente’s nuanced attitude is advisable in this area, where the kinds at stake are not very “natural”, to say the least. 18. Cf. García-Carpintero (2006) for elaboration.

124 (14)

Manuel García-Carpintero

It was not John who stole the bicycle. a. Asserted content: John did not steal the bicycle b. Conventionally presupposed content: Someone stole the bicycle

Consider now cases of what I take to be conventional implicatures, rather than lexically based presuppositions. I follow Potts (2005) in assuming that conventional implicatures differ from lexically based presuppositions only in being correctly used to provide new information – i.e., information which is not presented as already part of the “common ground”.19 As we will see, what they share, and what distinguishes both of them from asserted or purely “truthconditional” contents is their “projectibility” beyond operations such as negation, conditionalization, modalization or questioning; both the lexically based presuppositions and the conventional implicatures of a sentence are typically preserved when the sentence is negated, made the antecedent of a conditional, embedded in modal operators, interrogated, etc. Following Potts, I will use nonrestrictive relative clauses as an example of conventional implicature (the reader can replace it with implicatures generated by ‘but’, expressives or any other favourite): (15)

John, who lives in Barcelona, did not steal the bicycle. a. Asserted content: John did not steal the bicycle. b. Conventionally implicated content: John lives in Barcelona.

Although the important point for our purposes, with respect to Cappelen and Lepore’s classification of accounts of double-duty quotation, is that on the present view conventionally implicated contents are not part of the asserted contents, and therefore not part of what they count as properly semantic, truthconditional content (which, as I said, makes the present proposal a use-only one on their terminology, in contrast with the dual account their earlier selves advocated and the mention-only one they defend now), I would like to indicate the main reasons why, on what I take to be a theoretically more accurate account of the semantics-pragmatics divide, conventionally indicated contents should count as semantic, not pragmatic. The main assumption justifying this 19. Of course, presuppositions can also be informative in some contexts, as in the classic example: ‘I cannot go with you, I have to pick up my sister at the airport’. However, I assume that this is a pragmatic phenomenon: the speaker relies on the hearer to “accommodate” him; cf. von Fintel (2008). Conventional implicatures may provide new information without relying on accommodation processes; it is only that this potentially new information is somehow ancillary, subordinate or otherwise secondary to the content of the main speech act, as witnessed by the “projectibility” data.

Double-duty Quotation, Conventional Implicatures and What Is Said

125

is one I have defended elsewhere.20 As Lewis (1980, 23–24) puts it, a semantic theory is “one suited to play a certain role in a systematic restatement of our common knowledge about language …the detailed and parochial part – the part that would be different …if we were Japanese”. Now, our semantic competence includes knowledge of conventionally implicated contents (and lexically based presupposed contents), as manifested by two pieces of data. Firstly, semantic competence underwrites the validity of inferences such as the following:21 (16) (17)

It was not John who stole the bicycle. ∴ Someone stole the bicycle John, who lives in Barcelona, did not steal the bicycle. ∴ John lives in Barcelona.

Secondly, in accordance with their being conventionally generated, the contents in question cannot be cancelled: (18) (19)

It was not John who stole the bicycle, but it is not the case that someone stole it. John, who lives in Barcelona, did not steal the bicycle, but it is not the case that John lives in Barcelona.

Keeping thus in mind these differences in views with respect to the pragmaticssemantics divide, I move on now to present my use-only account of double-duty quotation. As I said, Predelli (2003) offers an account along these lines for a particular case of double-duty quotation – scare quotes – which I find entirely congenial; as he insists, one does not need to be particularly committal on the specifics of the framework for these purposes. Predelli usefully distinguishes the main asserted proposition in examples such as (15) on the sort of view we are canvassing, from the proposition that typically plays a subsidiary role relative to it, as, respectively, message and attachment. Now, compare (20) to (13), and (21) to (1): (20)

A fortnight is, by definition of that term, a period of fourteen days

20. Cf. García-Carpintero (2006: 2007). 21. Of course, much more needs to be said to justify my claims here than bringing to bear intuitions about examples. For instance, I would like to count the inference from ‘Some students passed the exam’ to ‘Not all students passed the exam’ as a pragmatic generalized conversational implicature, but the mere appeal to intuitions of validity does not suffice to justify the distinction. My purpose here is to express my views while gesturing at least in the direction of the kind of considerations that could be used to support them.

126 (21)

Manuel García-Carpintero

Saddam Hussein did not pose an – as Secretary of Defense Rumsfeld said – immediate threat to the security of our people

According to a view on which non-restrictive clauses such as those in (20) – which logicians typically abbreviate with the subscript ‘df’ attached to the defined expression – and (21) do not signify parts of the messages signified by the main sentences including them, but rather attachments, they express, respectively, (m)essages and (a)ttachments as follows: (20m ) (20a ) (21m ) (21a )

That a fortnight is a period of fourteen days That being a period of fourteen days defines ‘fortnight’ That Saddam Hussein did not pose an immediate threat to the security of our people That Secretary of Defense Rumsfeld said it by using a sentence including ‘immediate threat to the security of our people’

Of course, in (20) and (21) reference to the expressions mentioned in the metalinguistic attachments is made not by quotation marks, but by expressions whose referential character is not under dispute, such as the demonstrative ‘that expression’ and the conjunction ‘as’ – analogous in this respect to the adverb ‘so’ in the notorious ‘Giorgione was so-called because of his size’. The present proposal is that, by including quotation marks around expressions that are otherwise used in their standard ways (as required by the syntax of the sentences including them), (13) and (1) come to express, in addition to messages like (20m ) and (21m ) respectively, attachments sufficiently close in context to (20a ) and (20a ). The view is not that quotation marks in those cases conventionally implicate meanings exactly like those of the non-restrictive clauses in (20) and (21); for the latter typically have more precise conventional meanings than what the use of quotation marks in those cases conventionally implicates. The conventional implication of such uses of quotation marks is just that some or other metalinguistic attachment about the quoted material is conveyed; the context determines what approximately the attachment is.22 Thus, we can put the proposal in the following way:

22. See Predelli’s (2003: 16–17) for elaboration. Cf. also Gómez-Torrente (2005). I am not assuming a narrow notion of context, on which it consists of more or less objective features such as the speaker, the place and time of the speech, etc. I doubt that such a notion is adequate even to understand how pure indexicals such as ‘here’, ‘now’ or even ‘I’ work. I have in mind a wider notion, on which the communicative intentions of the speaker are a crucial part.

Double-duty Quotation, Conventional Implicatures and What Is Said

(13m ) (13a )

(13a ) (1m ) (1a )

(1a )

127

Asserted content: That a fortnight is a period of fourteen days Pragmatically enriched conventionally implicated content: That the word ‘fortnight’ has in English the meaning being a period of fourteen days. Conventionally implicated content: That the word ‘fortnight’ is used in a contextually salient way. Asserted content: Secretary of Defense Rumsfeld said that Saddam Hussein posed an immediate threat to the security of his people. Pragmatically enriched conventionally implicated content: Secretary of Defense Rumsfeld uttered the words “immediate threat to the security of our people”. Conventionally implicated content: some contextually salient subject uttered words of which “immediate threat to the security of our people” is a contextually adequate version.

The distinction between the purely conventionally implicated content and the pragmatically enriched one helps us to respond to an objection to proposals like mine by Reimer (2005), drawing on Tsohatzidis (1998). Reimer claims that the cancellability test gives the wrong results, suggesting that the relevant contents can be cancelled and therefore are purely pragmatic as opposed to conventionally generated, by means of examples such as the following: (22) (23)

Descartes said that man “is a thinking substance”, but he didn’t use those very words, as he wrote only in French and Latin. A “fortnight” is a period of fourteen days, but unfortunately nobody uses the word with that meaning anymore, although they should if English were used as in the old good days.

While examples like (22) and (23) show that the pragmatically originated part of pragmatically enriched contents can – understandably – be “cancelled”, they do not show at all that the more abstract conventionally implicated contents can; and, in fact, if we apply the cancellation tests to those more abstract contents, according to my own intuitions, we do get the result that they are conventionally associated with the expressions and cannot be cancelled.23 Given our discussion of different views about the semantics-pragmatics divide, we can distinguish two different kinds of use-only accounts; on one, any quotational content that mixed quotations such as (1) may convey is purely and straightforwardly pragmatic. This is the sort of view Reimer wants to push for, 23. Gómez-Torrente (2005: 136–137) defends a similar claim.

128

Manuel García-Carpintero

using the cancellation results in (22) and (23). In contrast, my own variety of a use-only account insists that part of the quotational content is conventional, and thus semantic in my own preferred sense. To defend the specifics of this view, as I have presented it, we have now two tasks. Firstly, we have to indicate the reasons for preferring this view to a dual one, such as Cappelen and Lepore’s earlier view. Second, we have to justify the characterization of the quotational content as a conventional implicature rather than a lexically based presupposition. Bach (1999: 340) has argued that conventional implicatures are part of “what is said”; his main argument invokes embeddability considerations, in the form of an IQ test: An element of a sentence contributes to what is said in an utterance 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. If we interpret “what is said” in terms of Cappelen and Lepore’s “truth-conditional content”, i.e., my “asserted content”, we could use an argument of this sort to argue for a dual account, on the basis of intuitions about cases such as (24), uttered in a context requiring strict standards of direct quotation: (24)

A. Lewis said in the NYB (2004, May 27, 10) that Secretary of Defense Rumsfeld said that Saddam Hussein posed an “immediate threat to the security of our people”.

However, as I (2007: 174–177) have argued in detail elsewhere, Bach’s “what is said” cannot be equated to Cappelen and Lepore’s “truth-conditional content”. And, in any case, the IQ test does not distinguish between asserted contents and contents conventionally implicated or presupposed, because ‘to say’ may well work like a “plug” for such contents;24 in fact, as I argue in the previously mentioned paper, Barker’s (2003) view, which uses the very same test in support of the view endorsed here that conventional implicatures are not part of truth-

24. Gómez-Torrente (2005: 136–138) concurs. Unlike Karttunen’s (1973) presupposition holes (negation, conidtionalization, modalization, questioning, see below in the main text for examples), a presupposition plug creates a context where secondary contents are not “projected” to the main sentence. Potts (2005, 2007b) shows that conventional implicatures such as those generated by appositives and non-restrictive clauses can escape plugs more easily than straightforward cases of presuppositions (distinguished by their being assumed to be part of the background). However, there are clear cases of plugged CI; cf. Amaral et al. (2007), and the examples below in the main text.

Double-duty Quotation, Conventional Implicatures and What Is Said

129

conditional content, is not in contradiction with Bach’s when the terminology is unified. I said before that truth-conditional contents differ from conventionally implicated or presupposed contents in the “projectibility” of the latter under Karttunen’s (1973) standard “holes”, negation, conditionalization and other embeddings. It is straightforward to find examples for what I take to be the conventionally implicated quotational content in the case of scare quotes, such as (13), but, given that Cappelen and Lepore (LTOI: 16–7) put aside those cases (although they are happy to extend to scare quoting what they say about mixed quotation, if it turns out that the phenomenon has sufficiently semantic features, cf. ibid., 17, under option 2), the controversial cases are those involving mixed quotation. The problem with them, however, is that they already embed the quotation (which on the present view is the conventional trigger of the implicature), into a “saying” operator – a well-known “plug” for secondary contents. Bach’s IQ test shows that presuppositions and in some cases conventional implicatures, on very natural readings, stay under the embedding, without being projected to the main clause; thus, as Bach (1999) points out, if Marv said “Shaq is huge and agile”, it does not feel correct to report him with “Marv said that Shaq is huge but agile”. In order to circumvent this obstacle, I propose to examine first cases in between mixed quotations and scare quotes, i.e., cases in which there is in the background a direct-discourse ascription to a speaker, but the utterance itself is not a saying-ascription. Thus, suppose we are speaking in a context in which we are assuming that Rumsfeld said: Saddam Hussein poses an immediate threat to the security of our people. In this context, (25)-(27) show that the quotational content can project out of negation, conditionalization, modalization and questioning; i.e., only the non-quotational content is negated (25), questioned (26), or remains in the antecedent (27): (25) (26)

(27)

Saddam Hussein did not pose an “immediate threat to the security of our people”. Did Saddam Hussein pose an “immediate threat to the security of our people”? No, as the evidence we have gathered conclusively establishes. Even if Saddam Hussein had posed an “immediate threat to the security of our people”, the strategy pursued to deal with it was badly planned and dreadfully executed.

It is obvious in these examples that, in the intended interpretation, the quotational content is outside the scope of the negation, the antecedent of the condi-

130

Manuel García-Carpintero

tional, the modal and the question. The second clause in (26) also shows that that content can be projected outside the scope of a propositional attitude operator, if (as seems reasonable) we take ‘the evidence establishes that’ to be such. Now, with this context in mind, I think it is easy to find an interpretation of (28) where the quotational content is also projected out of the saying-ascription: (28)

Nobody with first-hand knowledge of what was going on in Irak (no UN-inspector, for instance) ever said that Saddam posed an “immediate threat”.

Thus, I think we have good evidence that quotational contents behave like conventionally implicated contents, even in the case of mixed quotations, which (in addition to general explanatory considerations) favour a common account for those cases, cases of scare or emphasizing quotes, and the mixed cases we have seen in the previous examples; and there is a natural explanation for the more usual cases of mixed quotation, where quotational contents are “plugged”: the subject to whom the saying is ascribed is the most natural contextual candidate for being the user to whom the partial direct quotation is ascribed.25 Let us examine some additional reasons to distinguish quotational contents from asserted contents, which will also help to confirm them as conventional implicatures, rather than presuppositions (even while agreeing that, as I noted above, clear-cut joins in nature should not be expected in this area). Cases of what Yablo (2006) calls “non-catastrophic presupposition failure”, already known to Strawson, show that tests for presuppositions that probe intuitions about whether or not utterances have a truth-value when the candidate presupposed content fails are very unreliable; for instance, we may judge false ‘I had breakfast with the King of France’, and true ‘the King of France is identical to himself’. Von Fintel (2004: 271) proposes an alternative hey, wait a minute test; consider the following dialogues, with ‘#’ being an indication of impropriety: (29)

It was not John who stole the bicycle. a. # – Hey, wait a minute, I had no idea that John did not steal the bicycle. b. Hey, wait a minute, I had no idea that someone stole the bicycle.

Targeting the asserted content with the “hey, wait a minute” objection does not feel right, while objecting in that way to the presupposed content does. 25. Cp. again Gómez-Torrente (2005: 138–141). As Gómez-Torrente shows, although there are cases in which the quotational content remains inside the scope of a negation, these could very well be a case of Horn’s (1989) “metalinguistic” negation.

Double-duty Quotation, Conventional Implicatures and What Is Said

131

This tracks the fact that presuppositions are represented as part of the common ground, while asserted contents are represented as new information. Given this, and on the assumption that not being taken to be in the common ground is the main difference distinguishing CI from lexically based presuppositions, on the present hypothesis this test should not be adequate to distinguish quotational contents from asserted contents; and this seems to be corroborated by intuition: (30)

Secretary of Defense Rumsfeld said that Saddam Hussein posed an “immediate threat to the security of our people”. a. # – Hey, wait a minute, I had no idea that Secretary of Defense Rumsfeld said that Saddam Hussein posed an immediate threat to the security of his people. b. ?? – Hey, wait a minute, I had no idea that Secretary of Defense Rumsfeld used the word ‘threat’.

Now, Glanzberg’s (2005: §2) provides another test for presuppositions, a repair-obligatory test, which improves on Strawson’s “truth-value-gapjudgment” criterion. Unlike von Fintel’s, this does not target the fact that presuppositions are represented as common knowledge, but rather the fact that their contents are projected out of negations, conditionalizations and so on; given that this is common to CI and presuppositions, we should expect this test to distinguish quotational contents from asserted contents if the present view is correct. The repair-obligatory test appeals to the intuition that, unlike asserted contents, failed presuppositions require some repairs. Firstly, in echo-assessment; thus, if we know that the speaker is suffering a hallucination and there is no palm tree, we would not feel proper to reply to his question, “Is that palm tree about to fall?” with “No, that palm tree is not about to fall”, or with “Er …there is no palm tree, so I guess that palm tree is not about to fall”, unlike what would be the case if the presupposition did not fail; we would rather say “That palm tree is NOT going to fall – there is no palm tree”, or “Er …no …, there is no palm tree”. Secondly, in indirect speech report: in the previous context, we would not feel it proper to report, “George said that that palm tree is going to fall”; we would rather say, “George uttered ‘That palm tree is going to fall’, but there is no palm tree”. Conventional implicatures do appear to pass this test; thus, let us assume that John does not live in Barcelona: (31)

Did John, who lives in Barcelona, steal the bicycle? a. # No, John, who lives in Barcelona, did not steal the bicycle. b. ?? Er … John does not live in Barcelona, so I guess it is not the case that John, who lives in Barcelona, stole the bicycle.

132

Manuel García-Carpintero

c. d. e. f.

It is NOT the case that John, who lives in Barcelona, stole the bicycle – he does not live in Barcelona. Er … no … , John does not live in Barcelona. # George said that John, who lives in Barcelona, stole the bicycle. George uttered ‘John, who lives in Barcelona, stole the bicycle’, but John does not live in Barcelona.

As predicted, Glanzberg’s repair-obligatory test suggests that quotational contents are CI, rather than asserted contents. Because of the already mentioned problems with straightforward cases of mixed quotation, I appeal to mixed cases like (25)–(28) before, interpreted relative to the context described before, i.e., relative to a previous direct discourse ascription to Rumsfeld; (32c) and (32d) assume that Peter uttered (25): (32)

Did Saddam Hussein pose an “immediate threat to the security of our people”? a. ?? Er …no, Rumsfeld actually used the word ‘menace’, so I guess it is not the case that Saddam Hussein posed an “immediate threat to the security of our people”. b. Er … no … , in fact Rumsfeld actually used the word ‘menace’. c. # (Concerning Rumsfeld’s famous claim) Peter agreed that Saddam Hussein did not pose an “immediate threat to the security of our people”. d. (Concerning Rumsfeld’s famous claim) Peter uttered in agreement “Saddam Hussein posed an ‘immediate threat to the security of our people”’, but Rumsfeld had in fact said ‘menace’.

A sufficiently detailed development of the view here presented should provide a proper account of more difficult cases. Thus, in (33) – uttered in a context in which it is common knowledge that a terrible fight is being reported – it appears problematic to maintain that the expression inside the scare quotes is used in its standard way to convey the message, because it is only ironically used. Something similar applies to (34), this time because the expression inside the scare quotes does not belong to the language being used; and also to the expression in the mixed quotation in (35), now because the context relevant to interpret the quoted indexicals is not the context of the utterance (in fact, ‘our’ in (1) already illustrates this point). With (36), the problem lies rather with the metalinguistic attachment, in that the mentioned expressions do not belong to the language of the individual whose discourse is reported.

Double-duty Quotation, Conventional Implicatures and What Is Said

(33) (34) (35) (36)

133

The ‘debate’ resulted in three cracked heads and two broken noses (Predelli 2003) Nicola said that Alice is a “philtosopher” (Cappelen & Lepore 1997) Like Luther, Lucian Freud seems to attest that “Here I stand, I can do no other” Descartes said that man “is a thinking substance” (Tsohatzidis 1998)

From the perspective of the present view, the reason why we resort to the illustrated procedures is clear. In these cases, we are putting forward two propositions, in both cases with the help of conventional, systematic resources provided by our language: a metalinguistic, quotational one plays a secondary, ancillary role with respect to the main speech act, while another, non-quotational, provides the content for that main act. In cases like (33)-(36) there is a prima facie conflict between the expressive requirements adequate to perform the first role, and those more adequate to perform the second; the examples show that under those circumstances we yield to the former requirements, at least when it is relatively easy for the audience to infer from it the expressions required for the second role. There is some systematicity in the way we carry out this “choice”, explored by Cumming (2005) and Geurts and Maier (2005). Cappelen and Lepore (2007) invoke this fact in support of their mention-only view. However, we have by now evidence enough in favour of the view that the quotational content is not part of the truth-conditional, semantic content. On the other hand, the fact that what on the present view are secondary, ancillary contents have a conventional basis in the system of our language makes this form of use-only view compatible with the systematicity in the data; as I said, the present account is close to Geurts and Maier (2005) and Potts (2007a), both of which are provided as part of theories that emphasize the compositional, systematic basis for lexically-based presuppositions and conventional implicatures. To conclude, I will briefly discuss Cappelen and Lepore’s (2007) recent, allegedly mention-only theory. First of all one must say that, for a work whose authors proudly declare it to be “the first book ever written on the topic of quotation and metalinguistic representation more generally” (LTOI: 16), the account itself is very sketchy; as the authors acknowledge, we are just given a few “observations” which are not “substitutes for a detailed analysis” (LTOI: 141). These observations include the rather surprising contention that quotations (of which, as the reader remembers, non-expressions might be part) can be, as it were, syntactically recruited to perform the syntactic role corresponding to any syntactic position that they might occupy in a mixed quotation; that is to say, any syntactic position at all. In addition to this, we get a very sketchy

134

Manuel García-Carpintero

description of the “logical form” corresponding to a particular example, and some gesturing to the consequences of semantic axioms that we are not given at all. It is very difficult with just these indications to obtain an adequate idea to appraise the theory. I will thus put my main objection to their mention-only account in the form of a dilemma. Remember that, according to Cappelen and Lepore’s characterization, mention-only views “deny the initial intuition by saying that in mixed quotation, the semantic content doesn’t imply that the quoted words are used – only quoted” (ibid); “no corresponding indirect report [of Secretary of Defense Rumsfeld, in the case of (1)] follows from the semantic content of [(1)]” (LTOI: 28–9). The dilemma I want to put to them depends on two possible ways of interpreting these remarks. Rumsfeld might have uttered something like (37), or rather something like (38): (37) (38)

Bush told me that he thought that Saddam Hussein poses an immediate threat to the security of our people. Saddam Hussein does not pose any immediate threat to the security of our people.

On the first way of interpreting Cappelen and Lepore’s indications of what a mention-only account claims, it follows that, in both cases, (1) would be true. For, on that interpretation, (1) only claims that Rumsfeld said something concerning Saddam Hussein and concerning posing something, using in so doing the words ‘an immediate threat to the security of our people’; and this is made true not only by his uttering (38), but also (37). I think that this prediction is intuitively unacceptable, and also that no amount of sensible “alleviating pragmatics” can change the fact. On its most straightforward literal, semantic interpretation (1) is false if what Rumsfeld uttered was (37), as he might well complain to the newspaper if that was the case. In personal communication, Cappelen agreed with this, and provided an alternative interpretation of the just quoted remarks and of the suggested semantic. This alternative interpretation leads to the second horn of the dilemma. On the alternative interpretation, the semantic truth-conditions of (1) do entail an indirect-discourse report, only one a little less specific than the one provided by use-only and dual accounts. This indirect report ascribes to Rumsfeld his having said a proposition which, using the structured-proposition jargon, consist of the semantic value (in context) of ‘Saddam Hussein’, the semantic value in context of ‘poses’, the semantic value of past tense, and, as it were, a gap corresponding to the position that an expression of the syntactic category to which the expression ‘an immediate threat to the security of our people’

Double-duty Quotation, Conventional Implicatures and What Is Said

135

belongs. Gappy propositions have been invoked by Millians to deal with the semantics of sentences including singular terms without referents.26 On this second interpretation of the remarks quoted above, which Cappelen said that he preferred in personal communication, the account has it that the truth-conditions of mixed quotations entail indirect discourse ascriptions with gappy contents. This avoids the intuitively outrageous consequence of the first interpretation; if Rumsfeld had in fact uttered (37), (1) would turn out to be false. The problem for it is this. Given that mention-only accounts, thus understood, do entail indirect-discourse ascriptions, as much as use-only and dual accounts do, it is difficult to justify their now distinguishing claim that those ascriptions have merely gappy contents, as opposed to the full-fledged contents we get by replacing the gap, in our example, by the semantic value (in context) of ‘an immediate threat to the security of our people’. It is true that, in order to obtain that semantic value, delicate operations should be performed on some parts of the quoted material (in our example, on the indexical ‘our people’). But speakers do perform those operations, whether they happen in their “semantic module” or rather in their “pragmatic module”. And in any case LTOI does not provide, as far as I can see, any specific argument in support of ascribing to truth-conditional, semantic content merely the indirect report with the gappy content, as opposed to that with the corresponding full one. 4.

Conclusion

In this paper I have mostly discussed Cappelen and Lepore’s new minimalist proposals about quotation, introducing my preferred DO view as a useful foil. I have explored first their allegedly non-Davidsonian, anti-contextualist views about pure quotation, and then their new views on mixed quotation. I have complained in the first place that their proposals are not presented as perspicuously as they should be; and in the second place that, when we have a clearer picture of what appears to be the favoured account, the differences with their previous proposals and others already in the literature are not as great as they claim. References Amaral, Patricia, Craige Roberts & Allyn E. Smith 2007 Review of The Logic of Conventional Implicatures by Chris Potts. Linguistics and Philosophy 30(6): 707–749.

26. Braun (2005: sections 1–5) provides an excellent discussion.

136

Manuel García-Carpintero

Bach, Kent 1987 Bach, Kent 1999

Thought and Reference. Oxford: Clarendon Press. The myth of conventional implicatures. Linguistics and Philosophy 22: 327–366.

Barker, Stephen 2003 Truth and conventional implicature. Mind 112: 1–33. Braun, David 2005

Empty names, fictional names, mythical names. Noûs 39: 596–631.

Caplan, Ben 2001

Quotation and demonstration. Philosophical Studies 111: 69–80.

Cappelen, Herman & Ernie Lepore 1997 Varieties of quotation. Mind 106: 429–450. Cappelen, Herman & Ernie Lepore 2007 Language Turned on Itself. Oxford: Oxford University Press. Cumming, Sam 2005

Two accounts of indexicals in mixed quotation. Belgian Journal of Linguistics 17: 77–88.

García-Carpintero, Manuel 1994 Ostensive signs: Against the identity theory of quotation. Journal of Philosophy 91: 253–264. García-Carpintero, Manuel 1998 Indexicals as token-reflexives. Mind 107: 529–563. García-Carpintero, Manuel 2000 A presuppositional account of reference-fixing. Journal of Philosophy 97 (3): 109–147. García-Carpintero, Manuel 2001 Gricean rational reconstructions and the semantics/pragmatics distinction. Synthese 128: 93–131. García-Carpintero, Manuel 2004 The deferred ostension theory of quotation. Noûs 38 (4): 674–692. García-Carpintero, Manuel 2005 Double-duty quotation: The deferred ostension account. Belgian Journal of Linguistics 17: 89–106. García-Carpintero, Manuel 2006 Recanati on the semantics/pragmatics distinction. Crítica 38: 35–68. García-Carpintero, Manuel 2007 Bivalence and what is said. Dialectica 61: 167–90.

Double-duty Quotation, Conventional Implicatures and What Is Said

137

Geurts, Bart & Emar Maier 2005 Quotation in context. Belgian Journal of Linguistics 17: 109–128. Glanzberg, Michael 2005 Presuppositions, truth-values, and expressing propositions. In Contextualism in Philosophy: Knowledge, Meaning and Truth, G. Preyer and G. Peter (eds.), 349–396. Oxford: Oxford University Press. Gómez Torrente, Mario 2001 Quotation revisited. Philosophical Studies 102: 123–153. Gómez Torrente, Mario 2005 Remarks on impure quotation. Belgian Journal of Linguistics 17: 129–147. Horn, Laurence R 1989 A Natural History of Negation. Chicago: University of Chicago Press. (Reissued 2001 by CSLI.) Kaplan, David 1989

Demonstratives. In Themes from Kaplan, J. Almog, J. Perry and H. Wettstein (eds.), 481–563. Oxford: Oxford University Press.

Karttunen, Lauri 1973 Presuppositions and compound sentences. Linguistic Inquiry 4(2): 169–193. Lewis, David 1980

Index, context, and content. In Philosophy and Grammar, S. Kanger and S. Öhman (eds.), 79–100. Dordrecht: Reidel. (Also in Papers in Philosophical Logic, D. Lewis, 21–44. Cambridge: Cambridge University Press.)

Nunberg, Geoffrey 1993 Indexicals and deixis. Linguistics and Philosophy 16: 1–43. Potts, Christopher 2005 The Logic of Conventional Implicatures. Oxford: OUP. Potts, Christopher 2007a The dimensions of quotation. In Direct Compositionality, C. Barker and P. Jacobson (eds), 405–431. Oxford: Oxford U.P. Potts, Christopher 2007b Into the conventional-implicature dimensions. Philosophy Compass 2: 665–679. Predelli, Stefano 2003 Scare quotes and their relation to other semantic issues. Linguistics and Philosophy 26: 1–28.

138

Manuel García-Carpintero

Predelli, Stefano 2008 The demonstrative theory of quotation. Linguistics and Philosophy 31: 555–572. Recanati, François 2001 Open quotation. Mind 110: 637–687. Reimer, Marga 1996 Reimer, Marga 2005

Quotation marks: Demonstratives or demonstrations?. Analysis 56: 131–141. Too counter-intuitive to believe? Pragmatic accounts of mixed quotation. Belgian Journal of Linguistics 17: 167–86.

Tsohatzidis, Savas 1998 The hybrid theory of mixed quotation. Mind 107: 661–664. von Fintel, Kai 2004

von Fintel, Kai 2008

Would you believe it? The king of France is back! (Presuppositions and truth-value intuitions). In Descriptions and Beyond, M. Reimer and A. Bezuidenhout (eds.), 269–296. Oxford: Oxford University Press. What is presupposition accommodation, again? Philosophical Perspectives 22: 137–170.

Washington, Corey 1992 The identity theory of quotation. Journal of Philosophy 89: 582–605. Yablo, Stephen 2006

Non-catastrophic presupposition failure. In Content and Modality: Themes from the Philosophy of Robert Stalnaker, J. Thomson and A. Byrne (eds.), 164–190. Oxford: Clarendon Press.

What Quotations Refer To* Mario Gómez-Torrente

Abstract. When quotations are used with a purely referential purpose, they are mostly used with the purpose of referring to expressions, in the sense of rather abstract expression types. However, in many cases purely referential quotations are used with the purpose of referring to things other than very abstract expression types, such as boldface types, sounds, particular tokens, etc. The paper deals with the question of what mechanism underlies the possibility of successfully referring to different things and kinds of things with one and the same quotation. I defend the view that a quotation has as its semantic reference a certain very abstract expression type, and that the possibility of referring to other things by means of it is to be explained as a pragmatic phenomenon of felicitously conveyed speaker reference.

1.

Introduction

When quotations are used with a purely referential purpose, they are mostly used with the purpose of referring to expressions, in the sense of rather abstract expression types. In typical utterances of the following sentences, the quotations they contain will be used with a purely referential purpose, and the referential purpose in question will be the purpose of referring to rather abstract expression types: (1) (2) (3)

Use “Velázquez”, not “Velásquez”. Does “hiss” appear in Shakespeare’s plays? “Batman” is a composite expression.

In utterances of (1), ‘ “Velázquez” ’ will typically be used with the purpose of directing the addressees to use generally the expression “Velázquez” instead of the expression “Velásquez” – not, e.g., to write a particular token of * This paper was written during a research stay at the Faculty of Philosophy of the University of Groningen. Conversations I had there with Allard Tamminga were very helpful. I also thank an anonymous referee for useful comments. Research supported by DGAPA-UNAM (IN401909-3) and by the Spanish MICINN (FFI2008-04263).

140

Mario Gómez-Torrente

“Velázquez” or to write always “Velázquez” in Times New Roman type, etc. In utterances of (2), the question that will typically be asked is the question whether the expression “hiss” appears in Shakespeare’s plays – not, e.g., whether some particular token of “hiss”, or some token in Times New Roman type, etc., so appears (in some particular printed token of the plays). In utterances of (3), similarly, the intended reference will typically be to the expression “Batman”, not to particular tokens of it or to its shape in Times New Roman type, etc. However, as emphasized in a number of places in the recent literature on quotation (see especially García-Carpintero 1994 and Cappelen and Lepore 2007: ch. 7), in many cases quotations used with a purely referential purpose are used with the purpose of referring to things other than very abstract expression types, as will be the case with some utterances of the following sentences: (4) (5) (6)

“Velázquez” is a Times New Roman type. “Hiss” is a hissing sound. Please paint “Batman” in black.

In utterances of (4), ‘ “Velázquez” ’ will typically be used with the purpose of referring to the Times New Roman type of “Velázquez”, which is certainly less abstract than the abstract type “Velázquez”; in utterances of (5), ‘ “Hiss” ’ will typically be used with the purpose of referring to a sound (or a sound type); in utterances of (6), ‘ “Batman” ’ will in some cases be used with the purpose of referring to a specific token of “Batman”, e.g. a token written in big wood letters above the entrance of a cinema theater. These referring purposes will typically be carried out successfully, in the sense that utterers of (4)-(6) will typically successfully convey to their audiences their intentions of referring to the appropriate things. What mechanism underlies the possibility of successfully referring to different things and kinds of things with what is presumably one and the same quotation? In this paper I will defend the view that a quotation has as its semantic reference a certain very abstract expression type, and that the possibility of referring to other things by means of it is to be explained as a pragmatic phenomenon of felicitously conveyed speaker reference. The paper thus provides some considerations intended to answer the second of the three main semantic questions about pure quotation that I mentioned in earlier work (GómezTorrente 2001, following a similar threefold distinction in Washington 1992: 584): the first was the question of what part of a quotation has a referring role, the mentioned second question was the question of what is the reference of that referring part, and the third question was the question of how that reference is fixed. In that earlier work I considered mostly the third and (especially) the first

What Quotations Refer To

141

of these questions. Let me begin here with a very brief account of what I take to be the right answers to the first and third questions (sections 2 and 3), so that it can become eventually clear how the three answers defended here fit together. Section 4 will develop the favored answer to the second question, the question that gives its title to this paper.

2.

What does the referring

The referring part of a quotation has been claimed to be the quotation itself (see e.g. Recanati 2001, Cappelen and Lepore 2007, Predelli 2008), the quoted expression (see e.g. Washington 1992 – perhaps interpreting correctly Frege 1980 – , Reimer 1996, Saka 1998), and the quotation marks (see e.g. Davidson 1984, Bennett 1988, García-Carpintero 1994, 2004). If we take seriously the disquotational character of quotation, it follows that only the quotation itself can be the referring part. It is normally taken for granted that things such as (7) are true: (7)

‘ “Socrates” ’ refers to “Socrates”.

But the truth of (7) is not compatible with the referring part of a quotation being either the quoted expression or the quotation marks. On any view, (7) says that ‘ “Socrates” ’ is a referring expression whose reference is “Socrates”. But if the quoted expression did the referring, then ‘ “Socrates” ’, as it appears in ‘ ‘ “Socrates” ’ ’ (the apparent subject of (7)), must refer to itself, not to “Socrates”. And if the quotes did the referring, then ‘ “Socrates” ’, as it appears in ‘ ‘ “Socrates” ’ ’, does not refer at all, let alone to “Socrates”. On the other hand, the claim that the referring part of a quotation is the quotation itself is compatible with the truth of (7): being a quotation, ‘ “Socrates” ’, as it appears in ‘ ‘ “Socrates” ’ ’, refers to something, and what it refers to is presumably the expression inside its quotes, namely “Socrates”. (These points are developed in Gómez-Torrente 2001.)

3.

How the referring is accomplished

While in earlier work I criticized some answers to this question, I gave only very sketchy consideration to the positive matter of what the right answer should be. A more extended discussion (both of negative and positive considerations) will again have to wait. However, my main views can be summarized as follows.

142

Mario Gómez-Torrente

Quotations are not typical names, and do not get their referents in the way typical names get their referents. Typical names presumably get their referents via individual acts of baptism, accomplished with the help of individual acts of ostension or descriptive stipulation, and they presumably keep those referents under certain sorts of transmission from speaker to speaker even in the absence of knowledge of identificatory ostensive or descriptive material. But quotations don’t get their referents via individual acts of baptism of any sort, or via transmissions of the mentioned sort. They obviously get their referents via some general convention or stipulation, explicit or implicit acquaintance with which puts someone in the position of using and understanding all quotations (at least in principle). However, the fact that quotations don’t get their referents by typical naming doesn’t imply that quotations are not like typical names in other respects, and it doesn’t imply that a quotation should get its referent via some particular equivalent description or demonstration. In some cases, one can effect acts of wholesale reference fixing for large, and even infinite, classes of expressions that otherwise behave like typical names and which presumably are not equivalent with corresponding particular descriptions or demonstratives. For example, the convention once existed in some parts of Spain (and probably still exists somewhere) of naming a child of a couple with the name of the (first) saint for that day in the calendar of Catholic saints (plus the first last name of the father and the first last name of the mother, in that order). The Lagadonian scientists Gulliver finds in his travels adopt the cumbersome but otherwise conceivable convention of using every thing as a name of itself. The viability of these general rules doesn’t imply that the names of the Spaniards named according to the mentioned convention, or the Lagadonian names, are really descriptions or demonstratives in disguise. Quotations presumably get their reference via a general convention very similar to the Lagadonian convention, namely this: (GCQ) Let the quotation of an expression refer to the expression itself. Unlike the old convention for naming Spanish children and the Lagadonian convention, this convention was perhaps not made explicit before philosophical or linguistic reflection on it was conducted. But in being merely implicit in linguistic written practice, it’s not different from other conventions about how the content of many words gets determined. The use of quotation can be seen as essentially Lagadonian. But it is strictly speaking only quasi-Lagadonian, because the expression that is adopted as referring expression is the whole quotation, including the quotation

What Quotations Refer To

143

marks.1 This results, of course, from an interest in avoiding potential confusions between normal uses of expressions and what, were it not for the quotation marks, would be pure autonymic (Lagadonian) uses. So quotations differ also from typical names in that they have discernible morphological structure, derived from this clarificatory, or “punctuatory”, nature of the quotation marks. But again the fact that quotations are morphologically complex, unlike typical names,2 doesn’t imply that they are more complex than typical names in any very interesting semantic sense. In particular, their reference need not be fixed by some compositional process that takes as inputs the referents of their components, or of components of descriptive or demonstrative phrases abbreviated by the quotation, etc. A quotation presumably refers directly to its referent, and does so in virtue of a general convention such as (GCQ). Some remarks on one view according to which the reference of quotations is fixed by individual equivalent demonstrative phrases will be given in the next section. 4.

What quotations refer to

As I announced, I will defend the view, which I take to be the standard one, that a quotation has as its unique semantic reference a certain very abstract expression type of the quoted token. This traditional view has come under attack mainly in the Davidsonian account of García-Carpintero (1994, 2004), especially designed to deal with the possibility of referring to different things with quotations, and according to which a quotation (in fact, its quotation marks; see section 1 above) is a special kind of demonstrative phrase which can get very different semantic referents depending on the context. The standard view has also been rejected in the recent non-Davidsonian account of Cappelen and 1. Perhaps a restricted Lagadonian convention is sometimes at work when the things named are noun phrases; maybe people sometimes use as a name of a noun phase the noun phrase itself (without quotation marks). It is perhaps only as restricted to some expressions that the Lagadonian idea is not too cumbersome to implement. 2. Though perhaps not unlike all names. Names like “Elizabeth II”, “Sammy Davis Jr”, “Washington DC”, etc., have discernible morphological structure: they have been originally formed by the composition of some expression that already named something with some particle intended to distinguish the new intended bearer from the old one(s). But just as with quotations, it is doubtful that they are more complex than names without discernible morphological structure in any very interesting semantic sense.

144

Mario Gómez-Torrente

Lepore (2007). I will first offer some brief critical considerations on GarcíaCarpintero’s account. Then I will turn to criticisms of the Cappelen and Lepore account. I will end by explaining and defending my own view. 4.1. García-Carpintero’s account According to a Davidsonian account, a particular use of a quotation involves a demonstration of some sort in which the thing demonstrated is the token involved in that particular use of the quotation. In García-Carpintero’s version, a quotation is what he calls an “ostensive sign”, something that “consists of a demonstrative (‘this’), a category term (‘sound’) – which may be merely implicit in the context – and a demonstration (the act of pointing, or directing one’s glance) toward a certain token, a certain physical thing [in the case of quotation, this is the relevant token of the quoted expression]” (1994: 258). The referent of the quotation (strictly speaking, of the quotation marks) is obtained via a mechanism of deferred ostension with the help of the contextually provided category term and the demonstrated token. For example, with a typical written utterance of (5), the utterer semantically expresses the same content as he would with “This sound is a hissing sound” said pointing to the expression “hiss”, and García-Carpintero takes the semantic referent of “this sound” in such a case to be the sound (type) by means of which the expression “hiss” is normally pronounced in English. I have already expressed a number of worries about the implications of this proposal regarding the first and third questions about quotation above (see section 1 above and Gómez-Torrente 2001). But even leaving aside those worries for present purposes, I would have a complaint about its specific implications for the question of what is the reference of a quotation. The complaint is, essentially, that García-Carpintero’s proposal seems to attribute excessive referential possibilities to quotations. The excessive nature of the proposal can be seen, I think, from the fact that it implies that, as a semantic matter of course, a quotation could be used to refer to anything appropriately related to its token by means of a (possibly implicit) category term, just as a phrase of the form “This X” could so refer. However, it seems clear that quotations are not so versatile in this respect as phrases of the form “This X”. One can point to this token: Velázquez and say “This man was a great painter”, easily managing to convey and conceivably semantically expressing the true proposition that this man Velázquez was a

What Quotations Refer To

145

great painter.3 However, it seems one can’t utter the following token sentence: “Velázquez” was a great painter and successfully convey to one’s audience the intended true proposition that Velázquez was a great painter,4 or at least one can’t do this without adding complicated peculiarities to the contextual setting.5 Also, if I point to the same (normal type) token of “Velázquez” above and utter This boldface type is a very dark boldface, I think it’s clear that I can easily convey and perhaps semantically express the (let’s suppose, true) proposition that the boldface version of Times New Roman “Velázquez” is a very dark boldface. However, if I utter the following token: (8)

“Velázquez” is a very dark boldface,

3. A referee wonders if it’s so clear that such an utterance of “This man was a great painter” would be felicitous, on the grounds that the utterer is not pointing at a person but at a written expression. Neither I nor any of my informants sense any infelicity, however. In fact, such an utterance would straightforwardly exploit the uncontroversial and uncontroversially effective mechanism of deferred ostension that García-Carpintero’s account relies on. 4. A referee suggests that such an utterance of ‘ “Velázquez” was a great painter’ sounds unacceptable, because the predicate “was a great painter” is usually applied to the unquoted expression. If the referee is right and the utterance is simply unacceptable, then so much the worse for García-Carpintero’s account, on which it should be just as acceptable as a suitable utterance of “This man was a great painter”. My own view is that the utterance is not unacceptable in any syntactic or semantic sense. Its oddness is due to the fact that it cannot be used to convey anything sensible except in complicated contextual settings. 5. In some uncomplicated contextual settings (for example, we can imagine that it has just been discovered that the painter’s real name was not “Velázquez”), the quotation marks in ‘ “Velázquez” was a great painter’ could be used as “scare” quotes, and ‘ “Velázquez” ’ would refer to the painter (whatever his real name) in a moderately clear sense. But I take these contexts to be irrelevant for a defense of GarcíaCarpintero’s account. In them, the quotation is not really purely referential. And in cases where the speaker wanted to refer to the painter by means of ‘ “Velázquez” ’, and do so by means of a purely referential use of the quotation marks, he could explicitly avoid the “scare” interpretation by openly explaining to the participants in the conversation that he did not mean the quotation marks to signal “distancing” of any sort. My intuition is that in such a case the addressees will be unable to grasp the speaker’s purpose of conveying the proposition that Velázquez was a great painter.

146

Mario Gómez-Torrente

I don’t think I can easily manage even to convey that proposition, as I probably require a fairly complicated contextual setting to do so.6 One simple explanation for these differences and difficulties is that with an utterance of (8) one actually cannot semantically express that proposition. What an utterance of (8) semantically expresses is presumably false without qualification. I take such examples to constitute strong indication that the quotation ‘ “Velázquez” ’ is not an ostensive sign in García-Carpintero’s sense. For if it were such a thing, we would at least have a feeling that in some unremarkable contextual setting it (or its quotation marks) could refer to Velázquez the man or to the boldface version of Times New Roman “Velázquez”. But I don’t think we do, so I surmise that García-Carpintero’s account attributes to quotations a semantic versatility that they don’t possess. 4.2. The Cappelen & Lepore account Cappelen & Lepore (2007: ch. 12) propose a novel theory of the reference of quotations. This is based on a view on the first and third questions about quotation that is in some ways congenial with my view summarized above. According to Cappelen and Lepore, referential quotation is essentially governed by the principle that all the instances of the following disquotational schema (which gives the abstract form of (7)) are true: ‘ “e” ’ refers to “e” (where “e” is replaceable by any quotable item, for an undetermined set of “quotable items”).7 This principle, together with the principle that “for any quotable item e, if a quotation expression Q refers to e, then e is contained in Q”,8 provide all there is to the semantics of the marks, on Cappelen and Lepore’s view. However, the Cappelen and Lepore apparatus for explaining the variability phenomena exemplified by the abovementioned utterances of (4)-(6) seems to me inadequate. On their view, typical utterances of (4) have a grammatical subject that refers to the Times New Roman type of “Velázquez” and hence 6. A referee describes one such contextual setting: a number of fonts are being discussed, and every font is represented by the written name of a different painter. 7. The schema Cappelen and Lepore actually use substitutes “quotes” for “refers”, because they intend it to govern both pure (referential) and impure uses of quotations. The principle in the text presumably governs purely referential uses. 8. Compare the notion of “interiority” in Gómez-Torrente (2001), where the principle is proposed that a quotation refers to something just in case that something constitutes the interior of the quotation.

What Quotations Refer To

147

are true, while other utterances of (4) have a grammatical subject that does not refer to a particular font type (some have a grammatical subject that refers to a more abstract type shared with normal type tokens of “Velázquez”), and hence are false. However, this doesn’t happen because all these utterances have one and the same grammatical subject that gets a different semantic reference as a function of variations in contextual determinants. It happens, according to Cappelen & Lepore, because the first type of utterances have one grammatical subject, containing the Times New Roman type of “Velázquez” (and presumably not containing a more abstract type), and the second type of utterances have another grammatical subject, containing a different, more abstract type of object, the word “Velázquez” (and perhaps not containing the Times New Roman type). To use also one of their examples (originally from García-Carpintero), consider: (9)

“Madrid” = “Madrid”

and (10)

“Madrid” =/ “Madrid”

Cappelen and Lepore say: first, imagine [(9)] tokened in a context C where the speaker is typing a rather obvious identity, but mid-sentence some strange formatting key is unintentionally pressed on the computer and switches font from Times Roman to Verdana (and where the font makes no difference to the writer at all – she doesn’t know what font she’s writing in, doesn’t know it changed mid-sentence, and wouldn’t have cared if she did notice). Intuition is that the resulting utterance is true. Next imagine [(9)] tokened in a context C’ where the writer is intending to bring to her audience’s attention the differences between the Times Roman and Verdana. She might utter [(9)] as a test case to her audience. In this context it’s easy to get the intuition that [(9)] is false and [(10)] true. (2007: 152)

Cappelen and Lepore’s explanation of these intuitions is again that “the subject of [(10)] is not the same as the subject of [(9)]” (2007: 153). The subject of (typical utterances of) (10) contains, in Cappelen and Lepore’s terminology, a mere sign (what I would rather call a mere expression), while the subject of (typical utterances of) (9) contains, again in their terminology, an expression (what I would rather call a word, or a meaningful expression). The relevant difference, unlike in García-Carpintero’s proposal, is not relevantly semantic, but effectively syntactic. While the main problem with García-Carpintero’s proposal was that it is excessive, the main problem with the Cappelen and Lepore proposal is that, unless

148

Mario Gómez-Torrente

it’s too eccentric, it’s insufficient. The proposal gives no satisfactory explanation of the apparent possibility of referring to a sound with an utterance of (5), or to a token different from the uttered one with an utterance of (6). For the only reason we are given of why, say, (9) and (10) (typically) have different subjects is that on some occasions a mere expression will be syntactically contained in the quotation that is the grammatical subject, while on other occasions it will be a meaningful expression that will be syntactically contained in the quotation. Even if the idea of containment were correct in a sense in which both mere expressions and meaningful expressions can be contained in quotations, I don’t see how the idea can be made sense of for sounds or for tokens different from the token contained in the uttered quotation. There is, for example, no literal sense in which a sound is syntactically contained in the grammatical subject of (5), or the big token of “Batman” is syntactically contained in the grammatical subject of (6). And if Cappelen and Lepore want to hold that a sound, or a token different from the uttered one, etc. are contained in the quotations that are the grammatical subjects of (5), (6), etc., they will need a theory of how “contains” can get (as it comes out of their mouths) as many different meanings (or contextually variable extensions) as one would need in order to explain the possibility of referring to all the different kinds of things one seems to be able to refer to with utterances of (5), (6), etc. And then they will be essentially back where they began. They will just have reduced the explanation of the mystery of variable quotational reference to another analogous (but even more mysterious) mystery. (It may also sound obscure at first sight to say that a quotation (even a quotation type) contains a meaningful expression (an object endowed with meaning), while it certainly doesn’t sound odd at first sight to say that it contains a mere expression (type). How could a meaning be contained in an expression? However, I think that there is a perfectly literal sense of “containment” on which it stands for a relation which obtains between meaningful expressions and quotations, induced by the relation of containment which obtains between the corresponding expressions and their quotations. Nothing similar can be said of the relation between a sound and the quotation of an expression typically pronounced by means of that sound, or of the relation between a token and a quotation of a different token of the same expression.) Note that, with a view to solving the variability problem posed by typical utterances of (4)–(6), Cappelen and Lepore cagily abstained from imposing any semantic constraint on the nature of the “quotable items”. But the idea that there is no such constraint provided by the semantics of the quotation marks is too eccentric. It seems clear that the meaning of the quotation marks does not contemplate the quotation of things that we do not take to be graphical items, such as tables, or elephants (or sounds), even if we can conceivably manage to put

What Quotations Refer To

149

them between quotation marks. It seems natural to postulate that something about the semantics of the marks constrains the nature of the quotable items, and the refusal to introduce some such restriction (so as to get an explanation of the variability phenomena) makes the Cappelen and Lepore account markedly suspicious. 4.3. A syntactically stable, semantically uniform, pragmatically variable account Let me then state what the basic ideas of my own proposal are. This will be followed by an explanation of how these ideas imply diagnoses and solutions to the problems we just saw for the García-Carpintero and Cappelen and Lepore proposals. Against Cappelen and Lepore, I subscribe to the standard (and natural, I think) view that what appear to be different tokens of the same quotation expression type, used in different sentences, are indeed syntactically identical tokens – in a word, that there is syntactic stability across different uses of a quotation. I also adopt the standard thesis of semantic uniformity: (an utterance of) a quotation always semantically refers to a unique thing. But what thing is this? I take a quotation to refer to the linguistically relevant graphical expression type to which the quoted expression token belongs, and consequently I think that when (GCQ) is implicitly grasped by a competent user of written language, the expressions that that user thinks about will be in typical cases linguistically relevant graphical expression types. The “linguistically relevant graphical expression type” of a written token is the type it has in common with all the tokens that have a shape sufficiently similar to be used for the same typical linguistic purpose (generally, the purpose of expressing some current or potential reference or meaning, by taking advantage of a particular linguistic convention which could be made explicit with the help of that token). Thus, the Times New Roman type of the tokens of “Velázquez” in this paper, their Verdana type, their type in Davidson’s handwriting, etc. are all less inclusive types than their linguistically relevant graphical expression type, which is the type they have in common with all the tokens of a shape that counts as sufficiently similar for the purpose of naming Velázquez taking advantage of the convention that originally gave the great painter his last name. (See Bennett 1988: 403, for a notion close to that of a linguistically relevant graphical expression type, though Bennett follows Davidson in thinking that it’s the quotes that do the referring.) Note that there is a corresponding linguistically relevant graphical expression type even in cases where there is no typical linguistic purpose currently

150

Mario Gómez-Torrente

served by the quoted expression token, e.g. cases where a new expression is introduced, cases where a currently meaningless proper part of a currently meaningful expression is quoted, cases when plain nonsense is quoted, etc. In these cases I think that the reference will typically be to the type of tokens with a shape sufficiently similar to that of the quoted token for them to be usable for the purpose of expressing one same potential reference or meaning (by taking advantage of a particular linguistic convention which could have been formulated with the help of the quoted token). For example, if someone utters the sentence token Let’s use “” to refer to the back side of the earlobe, there may be some vagueness involved as to what type ‘ “” ’ refers to, but presumably we will take it to refer to the type under which fall the tokens that could be used by normal humans in normal communicative exchanges in such a way that they could successfully talk about the back side of the earlobe by taking advantage of the convention just introduced. (This is of course quite vague and anthropocentric, in that these purposes will be dictated by the typical perceptual, communicative and other abilities of humans, but I don’t think any important objection to the current proposal will essentially depend on the fact that it appeals to vagueness and anthropocentricity.) I think that linguistically relevant graphical expression types are in most cases the objects that language users intend to refer to with quotations, as perhaps already made plausible by examples such as (1)–(3), which would appear to be more standard than examples such as (4)–(6). I also think that, in the absence of prominent contextual indications, the user of a quotation will be taken to refer to the linguistically relevant graphical expression type of the quoted token. In order to test this idea, we should try to find some quotational sentences lacking “internal” indicators (provided by the standard meanings of their constituent expressions) about what specific sort of thing is meant by the quotation; and we should consider what would be their default interpretation when uttered out of the blue, with barely any contextual clues. For example, we cannot rely on a consideration of something like Jones wrote “linguistically”, for “wrote” suggests that what is being referred to is something of a graphical nature. Nor can we appeal to something like “Linguistically” is an adverb,

What Quotations Refer To

151

for “is an adverb” suggests that what is being referred to is an object endowed with meaning. Nor can we appeal to something like “Cuckoo” is onomatopoeic, for “is onomatopoeic” suggests that what is being referred to is a sound (or an object that in some sense comprises a sound). Rather, we need something like Jones likes “linguistically” or Jones likes “ ”, for presumably “likes” does not suggest that what is liked is either a graphical item or a meaningful entity or a sound, to the exclusion of the other possibilities. And I think that we would by default take out of the blue utterances of those last two sentences to mean that Jones likes the linguistically relevant graphical expression types of the quoted tokens, and not some token or sound. I take all this to provide evidence that the objects conventionally referred to in quotation are linguistically relevant graphical expression types. As far as I can tell, if a quotation does have a unique semantic reference, then the only serious alternative to the idea that this reference is the linguistically relevant graphical expression type of the quoted expression is the idea that the reference is (in those cases in which it exists) the word or expression-cummeaning of which the linguistically relevant graphical expression type of the quoted expression presumably constitutes a part. However, a consideration that goes against this alternative idea is that expressions that don’t codify, or are not known to codify words seem to be perfectly quotable (as in some of the examples just given). This suggests that the cases in which reference to words or meaningful expressions is successfully accomplished by use of quotations are secondary with respect to the uses in which quotations are used to refer to linguistically relevant graphical expression types.9 9. The proposal advanced here would not be substantively altered if some convincing additional consideration suggested that quotation marks are mildly ambiguous (or polysemous), having a primary meaning on which the quotations formed with them refer to linguistically relevant graphical expression types, and a secondary meaning on which the quotations formed with them refer to the quoted word (when it exists). But of course, the proposal is in part designed so as to avoid postulating an implausible strong ambiguity of referential quotation marks, on which a referential quotation would have many different referents as a matter of semantic fact. On the other hand, I think that quotation marks quite generally, as opposed to purely referential

152

Mario Gómez-Torrente

If different uses of what appears to be a single quotation expression type are indeed syntactically stable and semantically uniform, how are we to explain the apparent possibility of successfully referring to different kinds of things by means of utterances of the same quotation expression type? In my view (already hinted at, but just hinted at, in Gómez-Torrente 2001), this is to be explained by means of the pragmatic mechanism of speaker reference. Speaker reference is to be contrasted with semantic reference. The semantic reference of a referring phrase (possibly taken with respect to a given occasion of use) is the object (if any) that is determined to be the referent (with respect to the given occasion of use) by general conventions of the language to which the phrase belongs (plus contextual determinants specific to the occasion of use). The semantic reference of a referring phrase is determined basically independently from the specific intentions of the user of the phrase (save for intentions which may be needed to determine the contextual content of items such as indexicals appearing in the phrase, and the like). On the other hand, a speaker’s reference for a referring phrase, on a given occasion of use of that phrase by that speaker, is an object the speaker wants to talk about by means of the phrase in that occasion of use, and that the speaker believes can be successfully talked about by means of that referring phrase in that occasion of use. (This notion is closely related to the notion of speaker reference introduced in Kripke 1977. It differs from this notion, however, in that Kripke stipulates that the speaker must believe, of an object that is a speaker’s referent, that it fulfills the conditions for being the semantic reference of the referring phrase; cf. Kripke 1977: 264. Kripke himself, however, considers debilitating the notion so as to include cases in which the speaker does not believe of her speaker’s reference that it is the semantic reference; cf. ibid.: 273, n. 22.) Here are some illustrations of the distinction in cases in which semantic reference and relevant speaker reference differ. The examples are familiar from the literature. “Two people see Smith in the distance and mistake him for Jones. They have a brief colloquy: ‘What is Jones doing?’ ‘Raking the leaves.”’ (Kripke 1977: 263) The semantic reference of “Jones” is uncontroversially quotation marks, are evidently ambiguous. Non-purely referential quotation marks are governed by semantic rules substantively different from those governing purely referential quotation marks; and among non-purely referential quotation marks, the marks as used in, say, so-called “mixed quotation” are clearly governed by semantic rules substantively different from those governing “scare” quotation marks. Some authors in the recent literature on quotation, however, have sought Procrustean semantically unified accounts of all types of uses of quotation marks. On this set of issues, and for an account of what all types of quotation do have in common, see Gómez-Torrente (2005).

What Quotations Refer To

153

Jones, if any is. However, the two participants in the colloquy have (unbeknown to them) Smith as a speaker reference for “Jones”. Next, the same two people visit the madhouse, where they see Peters dressed as Napoleon and shouting “I’m Napoleon!” (compare Kripke 1977: 273, n. 22). They have a new colloquy: “Napoleon is a loud shouter.” “Yes, he is.”. Again the semantic reference of “Napoleon” is uncontroversially Napoleon, if any is. But the two colloquy participants have (now known to them) Peters as their speaker reference for “Napoleon”. Finally, imagine that Peters flees the madhouse, and orders a ham sandwich at the first restaurant he comes across. A waiter takes the order but must go off duty soon. When the waiter is about to leave, the sandwich is still not out of the kitchen, so he passes the order on to a new waiter, saying “The ham sandwich is sitting at table 5.” “OK”, says the new waiter. Here the semantic reference of “The ham sandwich”, if any, is uncontroversially the ham sandwich already prepared on the kitchen table. But the two participants in the conversation have once more (and well known to them) Peters as a speaker reference for “The ham sandwich” (compare Nunberg 1979: 149). Note that we have here cases in which speakers manage successfully to talk about, and communicate thoughts concerning, objects that, on any reasonably natural view, are not the semantic referents of the referring phrases they use for those purposes. In the first case the speakers manage to talk about Smith, in the second case they manage to talk about the madman Peters, and in the third case they manage to talk about the madman turned restaurant customer Peters. In all these cases the success of communication seems to be straightforwardly explainable by the satisfaction of a couple of regulatory, pragmatically sufficient conditions for successful speaker reference. First, there is some connection prominently present in the speakers’ thoughts between the semantic reference and the relevant speaker reference: Smith just looks like Jones (and is thought to be Jones); Peters the madman is known to believe of himself that he is Napoleon; Peters the restaurant customer is known to have ordered the ham sandwich. Second, it’s clear to the participants in the conversations that the connection in question is meant to be exploited by the speaker in order to communicate something about the relevant object. These two conditions seem to be pretty much all that is needed to explain the fact that the relevant speakers are able to refer successfully to the things they intuitively manage to refer to. Note also that in none of these cases is it necessary, or even natural, to postulate that a demonstration is effected or that the reference is obtained by a mechanism of deferred ostension involving the use of a demonstrative abbreviated by the used phrases, in the way García-Carpintero postulated for the quotational case. I think it’s fairly clear that, provided the semantic reference of a quotation is indeed its linguistically relevant graphical expression type, one can see typ-

154

Mario Gómez-Torrente

ical utterances of (4)–(6) as cases where the relevant speaker reference differs from semantic reference but the communication of thoughts about the relevant speaker reference can be successfully accomplished by virtue of the satisfaction of the two mentioned pragmatic conditions. Consider first a typical utterance of (4) (‘ “Velázquez” is a Times New Roman type’) in a written exchange, e.g. an office memo exchange or e-mail exchange. The fact that Times New Roman “Velázquez” and “Velázquez” are connected, and specifically that the uttered token of “Velázquez” is a token of Times New Roman “Velázquez”, will certainly be prominently present in the conversational participants’ thoughts. And it will be clear to them that the connection in question is meant to be exploited by the speaker in order to communicate something about Times New Roman “Velázquez”, to a great extent precisely because what the speaker wants to communicate is in part that what he is talking about is a Times New Roman type. Or consider the case of a typical utterance of (5) (‘“Hiss” is a hissing sound’) in a similar written exchange. The fact that the expression “hiss” is normally pronounced by means of a certain sound will certainly be prominent for the speakers. And it will be clear to them that the connection in question is meant to be exploited by the utterer in order to communicate something about that sound, to a great extent precisely because what the speaker wants to communicate is in part that what he is talking about is a sound. Or, finally, consider the case of the imagined utterance of (6) (‘Please paint “Batman” in black’). The fact that the expression “Batman” has a nearby huge token will be salient in the conversational participants’ thoughts. And it will be clear to them that this connection is being exploited by the speaker in order to communicate something about that token, to some extent because what the speaker wants to communicate is in part that he wants something to be painted, something which necessarily will have to be a physical thing that he is reasonably interested in having painted, such as the token in question, etc. Typical utterances of (10) provide other examples of how the mechanism of speaker reference can be invoked to explain variability phenomena. In a typical utterance of (10) (‘“Madrid” =/ “Madrid” ’) the fact that Times New Roman “Madrid” and the general expression “Madrid” are connected, and specifically the fact that the first uttered token of “Madrid” is also a token of Times New Roman “Madrid”, will be prominent in the conversational participants’ thoughts. It will similarly be obvious and salient to the speakers that Verdana “Madrid” and the general expression “Madrid” are connected, specifically because the second tokened token of “Madrid” is also a token of Verdana “Madrid”. And it will be clear to the participants that the connections in question are meant to be exploited by the speaker in order to communicate something about Times New Roman “Madrid” and Verdana “Madrid”, specifically that Times New Roman

What Quotations Refer To

155

“Madrid” is not the same as Verdana “Madrid”, to a great extent precisely because what the speaker wants to communicate is in part that a certain couple of things are different. (9), on the other hand, semantically expresses the trivially true proposition that the linguistically relevant graphical expression type “Madrid” is self-identical, but it can be used, e.g. in the test context described by Cappelen and Lepore, in order to convey the false proposition that Times New Roman “Madrid” is the same as Verdana “Madrid”. That this falsehood can easily be conveyed in that context is again plausibly explained simply because in that utterance of (9) the facts that the first tokened token of “Madrid” is a token of Times New Roman “Madrid”, and that the second tokened token of “Madrid” is also a token of Verdana “Madrid”, will be prominent in the examinees’ thoughts. And it will be clear to them (or at least to those among them who are neither too dumb nor too clever) that these facts are being exploited by the speaker-tester in order to enunciate something (false) about Times New Roman “Madrid” and Verdana “Madrid”, specifically that Times New Roman “Madrid” is the same as Verdana “Madrid”. A great deal of the motivating force behind the present account comes from the apparently clear fact that it successfully passes the well-known test devised by Kripke for the decisiveness of alleged counterexamples against a linguistic proposal. Here the (arguably standard) linguistic proposal is the proposal that a quotation is syntactically stable and semantically uniform, and the alleged counterexamples are provided by (appropriate utterances of) (4)–(6), (9) and (10). Kripke’s test, let’s recall, is this: If someone alleges that a certain linguistic phenomenon in English is a counterexample to a given analysis, consider a hypothetical language which (as much as possible) is like English except that the analysis is stipulated to be correct. Imagine such a hypothetical language introduced into a community and spoken by it. If the phenomenon in question would still arise in a community that spoke such a hypothetical language (which may not be English), then the fact that it arises in English cannot disprove the hypothesis that the analysis is correct for English. (1977: 265)

That the traditional account defended here passes this test with respect to the alleged counterexamples that concern us is clear because it is clear (I think) that, even if we decided to speak a new language which was exactly like English except for containing the explicit stipulation that its quotations are syntactically uniform and always semantically refer to linguistically relevant graphical expression types, the possibility of intuitively referring to different things with quotations would still arise. For example, we could still use (5) in a similar utterance and successfully convey the proposition that the sound “hiss” is a hissing

156

Mario Gómez-Torrente

sound. Further, the mechanism of speaker reference would roughly but satisfactorily explain how this success could be achieved even in the presence of an explicit stipulation that ‘ “hiss” ’ always semantically refers to a linguistically relevant graphical expression type. Within the present account we can offer a plausible diagnosis of what’s going on in the examples that caused problems for García-Carpintero’s account, and also provide a reasonable way out of the dilemma facing the Cappelen and Lepore account. Recall the difficulties involved in conveying by means of the token “Velázquez” was a great painter the true proposition that Velázquez was a great painter, and how they seemed to show that García-Carpintero’s account implied for the quotation ‘“Velázquez”’ a role it could not perform. Observe that while it will usually be clear and prominent to someone who reads such a token that there is a connection between the relevant token of “Velázquez” (and of ‘“Velázquez” ’) and Velázquez, there will be serious obstacles to the conversational transparency of the speaker’s intention to exploit that connection in order to say something about Velázquez: we would not typically think of Velázquez if we saw the quotation ‘“Velázquez” ’, since, for one thing, we would tacitly assume that if the speaker had had the intention of referring to Velázquez, he would have just written “Velázquez”, saving himself time and work. There is here a failure of the condition that it should be obvious to the conversational participants that the speaker means to exploit the prominent connection between semantic reference and relevant speaker reference in order to communicate something about the relevant speaker reference. On the present account, this is enough to explain where García-Carpintero’s account goes excessive. Consider also the case of the token of (8) (‘“Velázquez” is a very dark boldface’) and the difficulties involved in conveying by means of it the true proposition that the boldface version of (Times New Roman) “Velázquez” is a very dark boldface. It will usually be clear to someone who reads such a token that there is a connection between (the relevant token of) “Velázquez” (and (of) ‘“Velázquez” ’) and the boldface version of “Velázquez”. But first, this connection will presumably not be prominent, given that the relevant token is not boldface. (Note that if someone points to a preexisting token of normal type “Velázquez” by means of a demonstration coupled with an utterance of “This boldface type is a very dark boldface”, the connection between the abstract type and the boldface type will of course be prominent, regardless of the exact mechanism by which “This boldface type” acquires a (semantic or speaker) refer-

What Quotations Refer To

157

ence.) And second, presumably largely because of the lack of prominence, there will be serious obstacles to the conversational transparency of the speaker’s intention to exploit that connection in order to communicate something about the boldface version of “Velázquez”: we would not typically think of the boldface “Velázquez” if we just saw the quotation ‘“Velázquez” ’ in normal type. (Also, we would tacitly assume that if the speaker had had the intention of referring to boldface “Velázquez”, he would have just produced an appropriate utterance with a boldface token such as “Velázquez”, making his intention clearer. Note that someone who points to a preexisting token of normal type “Velázquez” by means of “This boldface type” need not be assumed to have been the author of the inscription, and this will also help remove the obstacles to considering him as referring to boldface “Velázquez”.) The main problem for the Cappelen and Lepore account was that it is either too eccentric or it is insufficient. By appealing to their idea of syntactic instability, one can explain the apparent possibility of reference to sounds, tokens other than the uttered one, etc., only through the assumption that sounds, tokens other than the uttered one, etc., are “contained” in some quotations. But this is either literally false or based on an eccentric sense of “contained” requiring further explanation. By contrast, the present proposal at no point is eccentric, in the sense that it doesn’t invoke the quotability of things other than expressions: the proposal is compatible with the possibility that just expressions are quotable, which is in fact the only possibility licensed by the postulated principle (GCQ) from section 2. However, this doesn’t make the proposal insufficient for explaining the variability phenomena, as I hope to have shown in my discussion of how it explains the success for communicative purposes of the mechanism of speaker reference in the case of the relevant utterances of (4)–(6), (9) and (10). One must note, in particular, that the mechanism of successful speaker reference is a general mechanism based on pragmatic principles which plausibly apply frequently outside the restricted quarter of quotation (as shown by the examples of Smith-Jones, Peters-Napoleon, and Peters-the ham sandwich), and thus that the sufficiency of these principles for explaining cases where semantic reference and relevant speaker reference come apart presumably should come as no surprise when the principles are applied (under the assumption of semantic uniformity) to the case of quotation. As far as I know, the one objection that has been explicitly made to pragmatic accounts of phenomena of variable quotational reference is the objection, made by Cappelen and Lepore, that there is no principled reason for choosing which thing a quotation refers to (from among all the things speakers seem to be able to refer to by means of quotations, but especially, I suppose, from among all the graphical types of which the quoted token is a token). However, their

158

Mario Gómez-Torrente

argument for this objection involves a puzzling gap. The argument is just the following: Suppose we try to argue, for example, that one of [(9)] or [(10)] is false. If so, which one? And what could be the reason for choosing it and not its counterpart as the false one? Suppose we choose [(9)] as the false one on the basis of its quotable items being physically type-distinct. That could render every identity between quotational expressions false. Or, suppose we try to argue that, contrary to intuition, a spoken indirect disquotational report of [(10)] is as a matter of fact true. The extreme implausibility of this view makes such a position exceedingly unattractive. Similarly, suppose we try to argue that [(9)] is false. If this were so, should we conclude that [(11)] (with two quotation expressions in the same font but with one in normal print and the other in bold print) and infinitely many like identities are also false? [(11)]

“Madrid” = “Madrid”

And what about [(12)] (with two quotation expressions written in the same font but in different font sizes) and infinitely many like identities of different font sizes? Are they also all false? [(12)]

“Madrid” = “Madrid”.

Our inability to make a reasonable non-arbitrary choice here indicates how the pragmatic strategies fail in explaining this part of the variability data. (Cappelen and Lepore 2007: 79)

Well, of course I agree that it seems somewhat absurd to think that any difference between tokens of an expression is going to make all identities of the sort mentioned by Cappelen and Lepore false – though I myself don’t think that this consideration is fully decisive. However, even if the fact that this is absurd at first sight is conceded and taken to be decisive for the intended purpose, I fail to see how we are supposed to become convinced that it’s also arbitrary or equally intuitively absurd or unprincipled to think that it’s (10) that should be false (and that (9), (11) and (12) should all be true). As far as I can tell, Cappelen and Lepore say nothing to justify the idea that nothing principled favors this option, which is obviously left open by what they say explicitly. Perhaps they would just add that it seems somewhat absurd to think that no difference between tokens of an expression is going to make any identities of the sort mentioned by them false; but I myself don’t see much of an absurdity in this. In fact, there is at least one powerful positive argument that the option of taking (9) as true and (10) as false is the more principled one. The argument is this. It is fairly widely agreed that, ceteris paribus, an account of a linguistic phenomenon that postulates multiple syntactic or semantic senses for an ex-

What Quotations Refer To

159

pression is less satisfactory than an account of the same phenomenon that does not postulate a multiplicity of this sort and explains the phenomenon on the basis of a single syntactic or semantic sense for the expression plus general linguistic principles that are known or widely believed to apply elsewhere. As I have attempted to argue, I think that an extremely good case can be made that phenomena of apparent variable quotational reference can be explained without remainder under the assumption of a single reference for a quotation, by appealing to general principles that are (I also think) widely believed to apply in other instances of use of referring expressions. Assuming this is conceded, it follows that an account that postulates a single reference and syntax for a given quotation is to be preferred to an account that accepts multiple syntactic or semantic senses,10 and hence that the option of taking (9) to express semantically a true proposition is more principled than the option of taking (10) to express semantically a true proposition.11 References Bennett, Jonathan 1988 Quotation. Noûs 22: 399–418. Cappelen, Herman & Ernie Lepore 2007 Language Turned On Itself: The Semantics and Pragmatics of Metalinguistic Discourse. New York: Oxford University Press. Davidson, Donald 1984 Quotation. In Inquiries into Truth and Interpretation, D. Davidson, 79–92. New York: Oxford University Press. Frege, Gottlob 1980

On sense and meaning. In Translations from the Philosophical Writings of Gottlob Frege, 3rd edn, P. Geach and M. Black (eds.), 56–78. Oxford: Blackwell.

García-Carpintero, Manuel 1994 Ostensive signs: Against the identity theory of quotation. Journal of Philosophy 91: 253–264.

10. Furthermore, we also saw how one can independently motivate the particular proposal made above as to what this single reference probably is. 11. I think that considerable general but indirect support for the present proposal is also provided by the fact that, being arguably the standard idea about quotational reference, it passes Kripke’s demanding test for the decisiveness of alleged counterexamples to a linguistic theory (see above).

160

Mario Gómez-Torrente

García-Carpintero, Manuel 2004 The deferred ostension theory of quotation. Noûs 38: 674–692. Gómez-Torrente, Mario 2001 Quotation revisited. Philosophical Studies 102: 123–153. Gómez-Torrente, Mario 2005 Remarks on impure quotation. In Hybrid Quotations, P. De Brabanter (ed.), 129–51. Amsterdam: John Benjamins. Kripke, Saul 1977

Speaker’s reference and semantic reference. Midwest Studies in Philosophy 2: 255–276.

Nunberg, Geoffrey 1979 The non-uniqueness of semantic solutions: Polysemy. Linguistics and Philosophy 3: 143–184. Predelli, Stefano 2008 The Demonstrative Theory of Quotation. Linguistics and Philosophy 31: 555–572. Recanati, Francois 2001 Open Quotation. Mind 110: 637–687. Reimer, Marga 1996 Saka, Paul 1998

Quotation Marks: Demonstratives or Demonstrations? Analysis 56, 131–142. Quotation and the Use-Mention Distinction. Mind 107: 113–135.

Washington, Corey 1992 The Identity Theory of Quotation. Journal of Philosophy 89: 582– 605.

Quotation Marks and Kinds of Meaning. Arguments in Favor of a Pragmatic Account1 Daniel Gutzmann and Erik Stei

1.

“Varieties” of quotation

Traditionally, quotation was considered a semantic phenomenon. The origin of this view can be traced back to the proper name theory advocated by Alfred Tarski (1983) and W.V.O. Quine (1951) who postulated a close connection between truth-conditions and quotation marks in the ambit of formal logic. Various theorists attacked their remarks by somewhat unfairly applying it to natural language (cf. Davidson 1984, Washington 1992, Cappelen and Lepore 1997, Saka 1998), but these critics also more or less maintained the focus on semantics. We think it useful to reconsider this position. François Recanati recently proposed a rather pragmatic analysis, according to which at least some kinds of quotation are “a matter of what people do with words” (Recanati 2001: 637). In this paper, we want to focus on the reasons for developing a pragmatic account of quotation marks. We suggest that conversational implicature is the most suitable level of analysis for the contribution of quotation marks as it allows including a broader variety of quotation than other theories, namely the following types: (1)

a. b.

“This theory is difficult to understand” is a sentence. pure quotation (PQ) Alice says, “This theory is difficult to understand.” direct quotation (DQ)

1. We thank all participants of the International Conference on Quotation and Meaning (ICQM) at the University of Mainz, especially Manuel García Carpintero, Bart Geurts, Mikhail Kissine, and Ede Zimmermann for their valuable comments. Elke Brendel, Philippe De Brabanter, Jörg Meibauer, Paul Saka and Markus Steinbach commented on an earlier version of this paper which clarified many points and helped to considerably improve the paper. All remaining errors and shortcomings are our own.

162

Daniel Gutzmann and Erik Stei

c. d. e.

Alice says that this theory “is difficult to understand”. mixed quotation (MQ) This “theory” is difficult to understand. scare quotes (SQ) (headline of today’s newspaper) Alice’s “new” theory! emphatic quotes (EQ)2

Let us consider an ordinary English speaker and her interpretations of the examples given in (1). What influence does each quotation have on the meaning of the sentence in which it occurs? Here are some intuitive answers: In (1a), the QM signal that the quoted words are mentioned rather than used. In (1b) and (1c), they suggest that Alice uttered the quoted words. The reader of (1d) would infer that the alleged theory does not deserve the label theory. The type of quotation represented in (1e) seems to put emphasis on the novelty of the theory. Note, however, that these “intuitive” interpretations are less obvious than it may seem at first sight. In fact, as we will argue, the correct interpretation of a sentence containing quotation marks is highly context-dependent, and the same quotational sentence may convey different speaker meanings, depending on the conversational background of the utterance. Nonetheless, in the rest of this paper we will use the labels as introduced above to indicate that a specific reading of the quotation is to be considered, although we will not always explicitly describe a specific context yielding that reading. Please consider this only a matter of economy in description. Some terminological remarks: The constructions we are dealing with consist of an expression and a pair of quotation marks enclosing it. We will call these constructions “quotational constructions” or just “quotations”. The expressions between the quotation marks may be called “quoted material”, “quoted expression” or, if you like, “quotatum”. Obviously, quotation marks will be called “quotation marks” or abbreviated “QM”. We will refer to a sentence or an utterance containing a quotation with “quotational sentence” or “quotational utterance”. We will sometimes speak of “utterances” or “utterance meanings” of quotations. This is to be understood as referring to written utterances as QM are a phenomenon of written language. The strategy of this paper is twofold: First, we carry out a systematic investigation of the question of what specific kind of meaning quotation marks contribute to the overall meaning of an utterance. We consider the following 2. This type of quotation is often labeled “greengrocer’s quotes”, but we would like to propose a more functional and less elitist term.

Quotation Marks and Kinds of Meaning

163

kinds of meaning: literal meaning (§ 2.1), conventional implicature (§ 2.2), presupposition (§ 2.3), and conversational implicature (§ 2.4). We present arguments in favor of a pragmatic analysis of quotation marks, claiming that the notion of conversational implicature seems to be the most promising alternative: All general features of this kind of meaning are met by quotational constructions. Nonetheless, an approach based on conversational implicatures faces some problems when taking direct and pure quotations into account, namely effects on truth-conditions and, allegedly, on grammaticality. Thus, our second aim is to propose acceptable solutions to these criticisms in § 3. Finally, in § 4, we consider how a radical pragmatic account of quotation could be integrated into a Neo-Gricean architecture of the semantics/pragmatics-interface.

2.

Kinds of meaning

It is difficult to find clear cut criteria in order to determine the kind of meaning contributed by a certain expression, and although there may be some dispute regarding how exactly to construe the taxonomy of meaning, we assume at least the following four basic kinds of meaning which we think are more or less uncontroversial: (i) literal meaning: the basic semantic meaning of an expression as encoded in the lexicon, (ii) conventional implicatures: conventionally conveyed meaning that is independent of the ordinary propositional content, (iii) presuppositions: aspects of meaning that are rationally presupposed to hold in order to evaluate an utterance, and (iv) conversational implicatures: further context-dependent, pragmatically inferred aspects of utterance meaning. In the following subsections, we apply some procedures to test which of these categories fits the meaning conveyed by quotation marks best. 2.1. Literal meaning If the meaning contributed by quotation marks were literal content, i.e., semantic content as encoded in the lexicon, it should be calculable compositionally from the meaning of the quoted expression and the meaning of quotation marks. Although this is not impossible a priori, it is obvious that this would have to include accounting for at least some contextual aspects. To see why, consider the following example: (2)

Charles G. Taylor is “innocent”.

Obviously it is difficult to grasp the exact meaning conveyed by an utterance of (2). This is due to a kind of contextual underdetermination of what is meant

164

Daniel Gutzmann and Erik Stei

by the QM, as in different contexts the utterance may convey another speaker meaning: (3) (4)

a. b. a.

(5)

b. a.

b.

Written in a discourse on human rights. Charles G. Taylor is “innocent”. A: What did the The Association for the Legal Defense Of Charles G. Taylor say about Liberia’s ex-dictator? B: Charles G. Taylor is “innocent”. Written on a poster of The Association for the Legal Defense Of Charles G. Taylor, a group lobbying for the acquittal of Liberia’s ex-dictator.3 Charles G. Taylor is “innocent”.

Depending on the utterance context, an utterance of (2) will receive different interpretations. In case it is read in the context of (3a), the quotation will be interpreted as scare quotes (SQ), while it will be understood as mixed quotation (MQ) if it is written in response to (4a). Finally, following (5a), we will end up with emphatic quotes (EQ). This shows that a simple-minded treatment of quotation as literal content does not lead us far since the interpretation of a quotational construction seems to depend crucially on extra-linguistic context. However, context dependency does not imply that quotation cannot contribute literal meaning after all. It only shows that if we wanted to treat quotation this way, we would have to get contextual information into the literal meaning. One way to achieve this is to analyze quotation by means of indexicals with reference to the context itself.4 This would probably result in claiming that a quotation literally means some property p provided by a context C where p is some kind of alternative meaning to the literal meaning of the quoted expression. Thus, we would get something along the following lines: (6)

‘fresh’  p ∈ C such that p is an alternative meaning for fresh in C.

This analysis is similar to the one proposed by Geurts and Maier (2005), who analyze quotation as triggering a presupposition of an utterance event in which some speaker used the quoted words to express some predicate p.

3. This is, in fact, the context in which we came upon this example. 4. This, however, is not the sense in which indexicality is employed in the demonstrative theory (Davidson 1984, Cappelen and Lepore 1997), according to which the QM are demonstratives referring to quoted material which is not, “from a semantical point of view, part of the sentence at all” (Davidson 1984: 90).

Quotation Marks and Kinds of Meaning

165

Although at first sight (6) seems to be an option if one wanted to treat quotation as contributing literal meaning, we reject it for two reasons: First, (6) somehow misses the crucial point about the context dependency of quotation. It is not always the case that context provides us with a unique alternative meaning for a quotation. The meaning of a quotation cannot be derived as straightforwardly as the meaning of pure indexicals like the famous first person pronoun. It is rather akin to demonstratives like that, insofar as speaker intention seems to play a crucial role in determining the meaning of a quotational construction. This leads us to our second objection: The view sketched by (6) is committed to the claim that quotation is always truth-conditionally relevant. But, as we will argue in 2.4, there are cases in which quotation does not contribute to the truth-conditions of the quotational utterance. 2.2. Conventional implicature It is tempting to read Predelli’s (2003) account as a proposal to analyze quotation marks as conveying conventional implicatures. In this paper, Predelli examines the notion of scare quotes, a phenomenon often neglected in the literature on quotation.5 Thanks to his paper, SQ received more attention in the discussion (cf. the contributions in De Brabanter 2005). Predelli analyzes scare quotes by means of a distinction between message and attachment, which is quite uncommon terminology, but we do not want to speculate about any possible reasons for the introduction of these terms. However, one of the hallmarks of Predelli’s proposal is to analyze utterances containing scare quotes as expressing two distinct propositions.6 Following Bach’s investigation of alleged conventional implicature devices, this means to reject the “one sentence, one proposition” (OSOP) assumption (Bach 1999: 350): (OSOP)

One sentence, one proposition. Every indicative sentence expresses exactly one proposition.

According to Bach (1999), sentences containing expression like but and therefore express two propositions which are not conjunct. We can give the following semantic translation for but borrowing Potts’ (2005) notation:

5. A seminal investigation was carried out by Reinhard Klockow (1976, 1978), especially in his comprehensive dissertation (Klockow 1980). 6. For some reason or other, Predelli (2003) does not speak of propositions, but this term seems to be fundamental in order to make sense of his view.

166

Daniel Gutzmann and Erik Stei

 (7)

but  PQx.

P(x) ∧ Q(x) , Gy[P(y) → ¬Q(y)]



The first part of the formula – P(x) ∧ Q(x) – corresponds to the conjunctive function of the meaning of but while the second part – Gy[P(y) → ¬Q(y)] – captures the adversative aspect. If the function in (7) is applied to its arguments, it will yield two propositions which are (and this is crucial) distinct and have to be interpreted independently of each other.7 In some sense, such an approach is therefore 2-dimensional. For illustration, consider the following example: (8)

a. b. c. d. e.

Mary is a linguist but a genius. P(x): linguist(mary) Q(x): genius(mary) message: linguist(mary) ∧ genius(mary) attachment: Gy[linguist(y) → ¬genius(y)]

In (8), but is applied to the propositions that Mary is a linguist (8b) and that Mary is a genius (8c). The first dimension of meaning concerning but, which emerges when the function in (7) is fed with these two propositions, is the ordinary coordination given in (8d). The second proposition would be (8e) which expresses the (alleged) contrast between being a linguist and being a genius. Thus, an interpreted sentence containing but receives a truth value tuple instead of a single truth value (cf. Potts 2005: 50). Predelli labels the first proposition expressed by an utterance like (8a) the message and the second one the attachment. Both propositions conjoined are called the semantic profile of an utterance. The semantic profile of an utterance containing scare quotes expresses two distinct propositions: (i) the message, which equals the ordinary proposition that would be expressed by the utterance without any quotation, and (ii) the attachment, that is contributed by the scare quotes. Following this analysis, the message of an utterance of (9a) would be (9b), while the attachment could be expressed by (9c):8 (9)

a. b.

The “debate” resulted in three cracked heads and two broken noses. message: The debate resulted in three cracked heads and two broken noses.

7. Note that the two propositions share the propositions P and Q as their arguments. That is the reason why Bach (1999: 352) calls expressions like but “preservative operators”. 8. This example was brought up by Predelli (2003: 3).

Quotation Marks and Kinds of Meaning

c.

167

attachment: It was not a debate but rather a brawl.

There are a number of problems with this view. The first is that, contrary to but and therefore, the contribution of SQ to the semantic profile does not seem to be definite. While the semantics of but is relatively straightforward – we need a conjunctive and an adversative part – this is not the case for scare quotes. SQ could indicate (i) that the speaker is not fully committed to all aspects of the quoted expression like in (9a), (ii) that she uses the expression ironically like in example (3), or (iii) that she wants to apologize for her use of a somehow deviant expression (e.g. a dialectal expression). The second and even more serious problem concerns the attempt to analyze scare quotes by means of the message and attachment theory in general. Recall that both message and attachment contribute to the semantic profile of an utterance, that is they are both communicated semantically by the speaker. But consider the following SQ sentence (10a) with its message (10b) and its attachment (10c): (10)

a. b. c.

Peter’s bagels are “fresh”. (SQ) message: Peter’s bagels are fresh. attachment: Peter’s bagels are not fresh.

If both the message and attachment associated with an utterance were part of its semantic profile, the semantic profile would yield a contradiction. Of course, the writer of (10a) does not communicate both (10b) and (10c). But although Predelli suggests that in some contexts, only the attachment is at issue, as in his example We watch ‘color’ TV, you, on the other hand, watch ‘colour’ TV (Predelli 2003: 22), he does not deliver an explanation of how it should be possible for the attachment to ‘overwrite’ the message. However, such a mechanism seems to be at work in cases like (10): The attachment not only gets priority over the message, but it is all that is communicated. Even if such processes are not always present in scare quotes, in many cases they are. Actually, if we take a closer look at the logical form of message and attachment in (9), Predelli’s strategy will yield a contradiction as well:9 (11)

a. b. c.

The “debate” resulted in three cracked heads and two broken noses. message: x[debate(x) ∧ result-in (broken-nose)(x)] attachment: x[¬debate(x) ∧ brawl(x)]

9. As a matter of convenience, we ignore tense.

168

Daniel Gutzmann and Erik Stei

The fact that no strict 2-dimensional theory of quotation allows for the needed interaction between the two dimensions of meaning in order to account for examples like (11) provides prima facie arguments against such a treatment.10 Another serious problem of 2-dimensionalism is the scoping behavior of conventional implicatures (CIs), at least if understood along Potts’ lines.11 According to Potts (2005), CIs project out of the scope of every higher operator. We do not have to care about the logic behind this here, but to get an idea of it, consider the following cases of conventional implicature borrowed from Potts (2005): (12)

a. b.

(13)

a. b.

(14)

a. b.

The damn Republicans deserve public support. The Republicans deserve public support, The speaker does not like the Republicans Bush says that the damn Republicans deserve public support. Bush says that the Republicans deserve public support, The speaker does not like the Republicans Do the damn Republicans deserve public support? Do the Republicans deserve public support?, The speaker does not like the Republicans

According to Potts’ (2005) theory, a sentence containing a CI conveys two propositions: the ordinary proposition as expressed by the sentence (this is called the at-issue content in Potts 2005) plus the CI as a second proposition. However, the CI proposition is always taken out of the semantic parse tree once it is calculated. Therefore, no operator is ever able to take scope over it.12 Hence the CI proposition that the speaker dislikes the republicans always jumps out of the semantic structure of the rest of the utterance, regardless of whether it is embedded under an attitude verb like in (13) or a question like in (14). In contrast to Predelli’s account of SQ, Potts’ (2005) theory is very explicit with regards to the scoping behavior of CIs. This makes it easy for us to test whether it is possible to analyze quotational constructions as CIs. We only have to check whether the meaning contributed by a quotation also scopes out of the sentence or not. First, consider scare quotes:

10. For further arguments against 2-dimensional theories of quotation, cf. Geurts and Maier (2005). 11. Since Predelli (2003) does not comment on the behavior of embedded scare quotes, we will consider Pott’s (2005) explicit approach here. 12. This effect is achieved by a particular definition of the semantic types and a semantic rule called CI application. Cf. the appendix in Potts (2005) for the logical details.

Quotation Marks and Kinds of Meaning

(15)

a. b. c.

(16)

169

Mary believes that Peter’s bagels are “fresh”. (SQ) Mary believes that Peter’s bagels are fresh, Peter’s bagels are not fresh Mary believes that Peter’s bagels are not fresh.

a. Are Peter’s bagels “fresh”? (SQ) b. # Are Peter’s bagels fresh?, Peter’s bagels are not fresh

In case of SQ, the meaning contributed by the quotation can be interpreted as having jumped out of the sentence, at least in the case of the belief report in (15). However, an embedded interpretation like (15c) is also available. But even worse for a Pottsean CI analysis of quotation, the unembedded reading is not available for SQ in (16b) at all. Next, have a look at mixed quotation: (17)

a.

Mary believes that Coffee & Bagels said that Peter’s bagels are “fresh”. (MQ) b. # Mary believes that Coffee & Bagels said that Peter’s bagels are fresh, Coffee & Bagels used the word fresh.

(18)

a. Did Coffee & Bagels say that Peter’s bagels are “fresh”? (MQ) b. # Did Coffee & Bagels say that Peter’s bagels are fresh?, Coffee & Bagels used the word fresh

Regarding mixed quotation, the results are even more obvious. If the contribution of the quotation escaped from the scope of the operators, both (17) and (18) would entail that Coffee & Bagels in fact used the words “fresh”. This is clearly not what an utterance of (17) or (18) means. Because of these problems, we think that an analysis of quotation along the lines of Potts (2005) is not promising at all. Not allowing for interaction and wrong predictions are two serious objections against treating quotations as contributing conventional implicatures. A look at the characteristic features of conventional implicatures will reveal another one. The list in (19) summarizes the most important features of conventional implicatures (cf. Potts 2005: 11 and § 2.5). (19)

a. b. c. d. e.

Conventional implicatures are independent of what is said. Conventional implicatures are speaker oriented. Conventional implicatures are detachable. Conventional implicatures are not context sensitive. Conventional implicatures yield multidimensional contents.

We will not discuss the whole list in detail, but we have already shown in § 2.1 that quotations are highly context sensitive. Thus (19d) is not met. But even if

170

Daniel Gutzmann and Erik Stei

we fix the meaning for the sake of argument as a SQ interpretation, (19e) will lead into trouble while (19b) is generally not fulfilled. Furthermore, the meaning of a quotation seems to be nondetachable in contrast to (19c): one can get an ironic reading by using italics instead of QM or, given that the context is rich enough, even without any special marking device. 2.3. Presupposition Following Geurts (1999: 6), we think that “there is a perfectly adequate procedure for testing if something is a presupposition”, which is based on the projection behavior of presuppositions and may be called ‘Projection Test Battery’ (PTB) (Geurts 1999: 6). The PTB consists of three stages, and if an expression suspected of triggering a presupposition passes all three stages, that is strong evidence for this expression being a presupposition trigger. However, the first stage is sufficient for our concerns.13 (PTB)

The Projection Test Battery for presuppositions (Stage 1) Let {} be a sentence containing an expression, triggering the inference that  is true. Stage 1 Check if sentences of the following form would normally imply that  is true: (i) not {} (ii) it is possible that {} (iii) a believes that {} (iv) if {} then (v) either {} or

It should be obvious that the usual suspects like definite descriptions (e.g. the notorious the present king of France) and factive predicates (e.g. to regret, to know) all pass the first stage of the PTB. In the history of linguistics it “may be hard to find any class of paralinguistic inferences that have not been dubbed ‘presuppositional’ at some time”, as Geurts (1999: 6) notes, quotation being no exception. Bart Geurts himself presents a presuppositional analysis together with Emar Maier (Geurts and Maier 2005). Thus, it might be interesting to examine whether the Geurts & Maier analysis is in accordance with PTB.

13. For the whole procedure, cf. Geurts (1999: 6–8).

Quotation Marks and Kinds of Meaning

171

According to Geurts and Maier (2005), a quotation triggers the presupposition that there is an individual x, an event e, and a property Q, such that x expresses (E) Q by means of fresh in e:14 (20)

“fresh”  xeQ: Ee (x, Q, fresh)15

At least regarding MQ, this analysis seems to be plausible at first sight. The allusion to a speech-event distinct from the actual one seems to be the hallmark of mixed quotation. Hence, we prefer to count examples like (21) as mixed quotation as well, despite lacking a verbum dicendum. (21)

The world is “all that is the case”.

Since in such cases the specific source of the quotation is absent and has to be reconstructed from the context, we may call these special instances of MQ anaphoric mixed quotation (cf. Potts 2007).16 However, the fact that there is an allusion to another speech-event in MQ does not provide any evidence for the presuppositional account of quotation provided by Geurts and Maier (2005). First, even if an allusion to a previous speech-event plays a central role in MQ, this alone does not lead to the conclusion that this allusion is a presupposition. Second, it is not quite obvious how an analysis along the lines of (20) can account for SQ, not to speak of EQ. Therefore, we are tempted to reject an analysis like (20) for conceptual reasons. But to be sure, let us add some evidence by applying the PTB in order to get a more concise picture of whether quotational construction can be regarded as triggering a presupposition or not. Consider the case of MQ, which is the primary subject of (20): (22)

What did Alice say about Peter’s bagels? a. Alice said that Peter’s bagels are “fresh”. b. Alice did not say that Peter’s bagels are “fresh”. c. It is possible that Alice said that Peter’s bagels are “fresh”. d. Peter believes that Alice said that Peter’s bagels are “fresh”.

14. Geurts and Maier (2005) are rather liberal about the solution of the presupposition of x and e. The solution of the speaker variable x may be the concrete individual, but it may be institutional or generic as well. The same holds for the speech event e, which may be a concrete event or a habitual or genetic one. 15. As usual, “” stands for “presupposes”. 16. For philosophers, it is of course an easy exercise to figure out the source of the quotation in (21).

172

Daniel Gutzmann and Erik Stei

e. f.

If Alice said that Peter’s bagels are “fresh” then he would be very proud. Either Alice said that Peter’s bagels are “fresh” or Mary has got a new theory.

In (22a), there clearly is an allusion to a speech-event, e.g. Alice’s utterance containing the word fresh to express something about Peter’s bagels. But crucially, this allusion vanishes in each of the constructions in (22b)–(22d) which would normally allow presuppositions to pass through (accordingly, such constructions are called holes). Furthermore, if the allusion in MQ were in fact a presupposition, then its falsity should lead at least to shaky intuitions about the truth-conditions of the utterance. Consider the infamous case of The present king of France is bald. Since there is no king of France at the time of writing this paper, the presupposition that there actually is a present king of France is not fulfilled and hence, intuitions vary whether such a sentence is strictly speaking false or does not receive any truth value at all. Compare this to the MQ in (22a). Even if neither Alice nor anybody else ever uttered the word fresh to say anything about Peter’s bagels, we would not hesitate to assign a truth value to (22a), whichever it should be. Together with the ‘projecting behavior’ of the allusion this provides us with strong support for our rejection of an analysis of quotation as presuppositions following Geurts and Maier 2005. Furthermore, it is hard to see how their proposal could deal with SQ and the like. This does not mean, however, that a presuppositional account is mistaken per se. It could be the case that a presuppositional account similar to the one proposed in (6) yields better results. A presuppositional version of it, such as (23), differs from (20) as it does not rely on the need for a previous utterance event. It seems quite obvious that SQ and EQ do not necessarily presuppose a previous utterance; the only thing needed is an alternative meaning for the quoted material being accessible in context. Let us check how far such an approach could take us. (23)

“fresh”  p: p is an alternative meaning for fresh.

A proposal like that seems to be able to derive not only the MQ reading, but also that of SQ and EQ. Moreover, it seems to be more adequate than its sibling (6), because – since in this case we deal with presupposition instead of literal meaning – we can allow the needed pragmatic reasoning to play a crucial role in calculating the speaker intended meaning for each context. But if we check some further facts about presuppositions, it is obvious that this analysis does not hold either. A well-known fact about presuppositions is that they do not project out of certain constructions, which are therefore called

Quotation Marks and Kinds of Meaning

173

‘plugs’. Typical plugs are propositional attitude verbs like to imagine, to dream or speech report verbs like to say. Consider the following examples: (24)

a. b. c. d.

Peter met the king of France. Peter dreams that he met the king of France. Peter imagines that he met the king of France. Peter says that he met the king of France.

In sentences likes (24b)–(24d), the presupposition of (24a) (that there is a king of France) no longer holds. A common explanation for ‘plugging’ presuppositions is that they still hold – not in the global context of the whole utterance, but in the local context of the embedded sentence, i.e., in the world of Peter’s dream, imagination or utterance respectively. In the case of (24b) for instance, we could paraphrase the local satisfaction of the presupposition along the following lines: (25)

Peter dreams that he met the king of France.  Peter dreams that there is a king of France and that he met him.

Consider how, in contrast to this, the alleged presupposition of an alternative meaning for the quoted expression behaves if embedded under such plugs. (26)

a. b. c. d.

Peter’s bagels are “fresh”. Alice dreams that Peter’s bagels are “fresh”. Alice imagines that Peter’s bagels are “fresh”. Alice says that Peter’s bagels are “fresh”.

Here, it is crucial that the meaning conveyed by the quotation holds in the actual context, not in the embedded one. With (26b) for instance, we want to say something about Alice’s dream about Peter’s bagels, e.g. that Alice dreams that Peter’s bagels are not fresh at all (SQ reading), that Alice dreams that Peter’s bagels are what Coffee & Bagels called “fresh” (MQ reading), or that she dreams that they are very fresh (EQ reading). An interpretation of (26b) according to which the alleged presupposition of (26b) is satisfied locally, i.e., (27), is highly implausible. (27)

Alice dreams that Peter’s bagels are “fresh”.  Alice dreams that there is an alternative meaning p for fresh and that Peter’s bagels are p.

Furthermore, there is a perfect reading available for (26b) in which the writer wants to convey that Alice dreams that Peter’s bagels are fresh and that the

174

Daniel Gutzmann and Erik Stei

writer applies QM to distance herself from Alice’s dream for whatever reason. It all depends on the context in which (26b) is uttered. Hence, typical plugs do not plug QM. 2.4. Conversational implicature Although there is no set of necessary and jointly sufficient conditions for an aspect of what is communicated to count as conversational implicature, usually at least the following characteristic properties of conversational implicatures are mentioned (cf. Levinson 1983, Horn 2005, Meibauer 2006): (28)

a. b. c. d. e.

Conversational implicatures are context dependent. Conversational implicatures are nondetachable. Conversational implicatures are reinforceable. Conversational implicatures are calculable. Conversational implicatures are cancelable.

We have already argued for the fact that the interpretation of a quotation crucially depends on the context of utterance. The examples (2)–(5) showed that the very same quotational construction can receive different interpretations by contextual variations. An alternative way to establish the context sensitivity of MQ, SQ, and EQ is by means of the collective description test, proposed by Cappelen and Lepore (2005) to prove that some allegedly context sensitive expressions are not, in fact, context sensitive (cf. Cappelen and Lepore 2005: 99): (CDT) The collective description test If an expression e is context sensitive, solely on the basis of knowing that there are two contexts C1 and C2 in which an utterance containing Ae and an utterance containing Be are true respectively, it cannot be automatically inferred that there is a true utterance containing (A&B)e. However, the collective description test as formulated above is biased, because it only accounts for truth conditionally relevant contents. But of course, nontruth-conditionally relevant context sensitive aspects of an utterance may block a collective description as well. Therefore, we need a slightly reformulated version of Cappelen’s and Lepore’s test, to capture non-truth-conditional aspects of an utterance as well: (MCDT)

The modified collective description test If an expression e is context sensitive, on the basis of knowing that there are two contexts in which an utterance containing Ae and an utterance containing Be are appropriate respectively, it cannot be

Quotation Marks and Kinds of Meaning

175

automatically inferred that there is an appropriate utterance containing (A&B)e. To see how non-truth-conditional aspects of an utterance may block collective descriptions because of their context sensitivity, consider the following examples of conversational implicatures:17 (29)

a. b. c.

C1 : Mary has got three children. C2 : Joan has got three children. → / Mary and Joan have got three children.

Suppose that (29a) is uttered in a context C1 in which the exact number of Mary’s children is at issue, for instance in a context in which a birthday party is being planned. In such a context, (29a) will conversationally implicate that Mary has no more than three children. Suppose that in contrast (29b) is uttered in a different context C2 in which the minimum number of Joan’s kids is at issue, for example a context regarding tax purposes. In that context, an utterance of (29b) will not implicate that Joan has no more than three children. Thus, suppose Joan actually has five children. Of course, the utterance of (29a) in C1 will not only be true regarding its semantic content but also appropriate. The same holds for uttering (29b) in C2 . Crucially, the appropriateness of both utterances in their particular context does not guarantee that an utterance of (29c) will be appropriate. In a context in which the definite number of children is relevant like the birthday party context, uttering (29c) will be inappropriate or even misleading and could end up with two crying children. Note that an utterance of (29c) only allows parallel interpretations: Either (29c) communicates that Mary and Joan have exactly three children or that both have at least three children.18 Equipped with MCDT, we can now provide more evidence for the context sensitivity of our three quotational constructions. (30)

a. b. c.

C1 : Peter’s Bagels are “fresh”. (SQ) C2 : Mary’s Bagels are “fresh”. (EQ) → / Peter’s and Mary’s Bagels are “fresh”.

17. We are aware of the fact that there is some dispute concerning the meaning of numbers. However, for sake of illustration, allow us the view that three semantically means at least three while the upper bound at most three is provided by an implicature. 18. Of course, (29c) could mean that Mary and Joan have three children together, but this ambiguity does not matter for the point we want to make here.

176

Daniel Gutzmann and Erik Stei

Just as in the example discussed above, an utterance of (30a) might be appropriate in some context C1 and an utterance of (30b) might be appropriate in another context C2 . Suppose that in C1 the quotes receive a reading as scare quotes while in C2 , they receive a reading as emphatic quotes. Obviously, there is no context in which (30c) would be appropriate because either it would communicate that both Peter’s and Mary’s bagels are disgusting or that both Peter’s and Mary’s bagels are extraordinarily fresh. Analogous to the case of (29c), (30c) does not allow distinct interpretations for the quotations. Combining the evidence provided by the examples in (3)–(5), (30), and MCDT, we conclude that context sensitivity is a crucial property of the quotational varieties considered. According to (28b), conversational implicatures are nondetachable. Nondetachability is commonly understood along the following lines (Meibauer 2006: 570): (31)

If there is an expression X’ that shares meaning with expression X that triggers the implicature, the same implicature should arise.

As we have already argued at the end of § 2.2, the interpretations associated with EQ, SQ, and MQ are not restricted to the presence of quotation marks as there are also other possibilities to convey the same meaning without using quotation marks. For further illustration, imagine a context in which a colleague says, smugly, that his own work is going to become as important as the work of Quine. Furthermore, assume that we all know this and we all know that the real Quine is unfortunately not able to speak to us in person anymore. Now suppose you utter this (borrowed from Recanati 2001): (32)

Quine wants to speak us.

Given the described context, (32) is a perfectly adequate utterance and we can reasonably take you to be implicating that our snobbish colleague wants to speak us. Of course, if our context is not rich enough, maybe because we do not know certain facts about our colleague or about Quine, enclosing Quine in QM is necessary to make sure that (32) is not interpreted as conveying anything about the philosopher Quine. Given the possibility of omitting quotation marks while conveying the same meaning, and given the possibility of using other means than QM, we conclude that the particular interpretation of a quotation is nondetachable. That conversational implicatures are reinforceable (28c) means that they can be made explicit without redundancy. This also holds for the particular interpretations due to quotation marks. Making them explicit does not yield redundancy but rather elaborates the intended interpretation of the quotes:

Quotation Marks and Kinds of Meaning

(33)

a. b. c.

177

Peter’s bagels are always “fresh”. I get sick every time I eat them. Peter’s bagels are always “fresh”. Indeed, they are the freshest in town. Peter’s bagels are always “fresh”. At least, that’s what I’ve read in Coffee & Bagels.

Another feature associated with conversational implicature – usually labeled calculability – demands that the meaning conversationally implicated could, in principle, be calculated on the basis of certain language independent conversational maxims and a general cooperative principle.19 For our purposes, a Q-principle and an I-principle will be sufficient, although one might use some M-principle as well, if one prefered a less minimalist model. We will briefly recall the relevant principles: (CP)

Cooperative principle “Make your conversational contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged.” (Grice 1975: 45)

(Q)

Q-principle “Say as much as you can (given I).” (cf. Horn 1984: 13)

(I)

I-principle “Say no more than you must (given Q).” (cf. Horn 1984: 13)

For illustrative purposes, take McCawley’s (1978) paradigmatic and wellknown example: (34)

a. b.

Black Bart killed the sheriff. Black Bart caused the sheriff to die.

(34a) seems to be in perfect accordance with the two principles. In contrast, the expression used in (34b) is somehow marked and a competent speaker, committed to the I-principle, would not utter this sentence if she wanted to convey that Black Bart shot the sheriff (or killed him in another stereotypical way, e.g., by stabbing him with a knife). Rather, (34b) seems to convey that the sheriff was killed in a non-standard way, e.g., Black Bart caused the sheriff’s gun to backfire by stuffing it with cotton. If this is correct, then a reading of (34b) according to which Black Bart shot the sheriff is pragmatically blocked by (34a), as a reader would expect a competent writer to obey the I-principle and the writer knows that. Now, if a writer in fact utters (34b) the reader will expect her 19. The principles are developed in Grice (1975: 45–47).

178

Daniel Gutzmann and Erik Stei

to have reasons for this, which indicates the need for further pragmatic reasoning, i.e., the intended speaker meaning has to be fleshed out according to the context of the utterance. Quotation marks seem to produce a similar blocking effect: (35)

Peter’s new “theory” is difficult to understand.

The expression theory, we propose, is marked by the application of QM and therefore it has to receive a marked interpretation that deviates from the standard interpretation, as the latter is blocked by the standard expression without QM. As we have seen, viewing QM as contributing conventional meaning to an utterance is problematic. Let us, therefore, assume that there is no semantic contribution of QM at all (NSC). Thus, given (36a), this would yield (36b). A more detailed inference might look like this: (36)

a. b. c. d.

e. f.

The writer W wrote Peter’s “theory” is difficult to understand. (assumption) By (a) W said that Peter’s new theory is difficult to understand. (NSC) W marked the term theory by enclosing it in quotation marks. (36a) If W only meant that Peter’s new theory is difficult to understand, W would have violated (I) as she used a marked expression to convey an unmarked meaning. (I + Q) Given W is cooperative, she will not violate (I). (CP) W wanted to convey something more/other than that Peter’s theory is difficult to understand. (36d)+(36e)

So far, the inference that “theory” is to be interpreted in a non-standard way (in contrast to theory) is pretty much independent of the particular context of utterance. To flesh out the target of quotation (Klockow 1980) or the point of quotation, as Recanati (2001) calls it, i.e., in order to substantiate the intended interpretation, we need a second part of the inferential process. This is where the specific context of utterance comes into play. Let us continue our inference sketched in (36) against the background of a specific context: (37)

a.

b.

I know that W is a very critical scientist and has a very strict understanding of the term theory in scientific contexts. (assumption) W wrote (35) in a scientific context. (assumption)

Quotation Marks and Kinds of Meaning

c. d. e.

179

I can infer from this utterance that she does not take the proposal in question to be a theory at all. (36f) + (37a) + (37b) W knows that I can infer this from her use of QM and she has not stopped me from inferring that. (CP + I + Q) Therefore, by uttering (35), W implicates that Peter’s new proposal is an ersatz theory. (37c) + (37d)

Of course, this is only one possible interpretation of (35). In a different context the QM might indicate another implicature, different from the one inferred through (37), e.g., W might be referring to another scientist’s diction, which would render a MQ-reading instead of the SQ-reading processed in (37). This leaves only one of the features listed in (28) to be addressed, namely cancelability. As we shall see, the analysis is less straightforward in this case, which is why we leave it for the following section, in which some worries about the conversational implicature view are discussed. Before that, let us take a closer look at a feature that is often considered to be one of the most important characteristic of conversational implicatures, which is that they do not effect the truth-conditions expressed by the sentence. According to the standard view, although a conversational implicature may change the felicity of an utterance, even dramatically, it does not affect the truth or falsity of the sentence uttered. For instance, consider the classic example by Grice (1975: 51) himself: (38)

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

In this short dialogue, B’s answer is true iff there is a garage round the corner. However, if the conversational implicature(s) of B’s utterance, e.g. that the garage is open and has petrol to sell, were false, B’s answer may be infelicitous, misleading or very mean. However, even if you could not buy petrol at the garage, and even if B knows that, B’s utterance would still be true as long as there is a garage round the corner. Given that non-truth-conditionality is an important feature of conversational implicatures, we have to check whether MQ, SQ, and EQ share this feature. If these quotations are non-truth-conditional we would get more evidence against treating them as literal meaning as a side effect. Let us start with some obvious cases: (39)

Charles G. Taylor is “Innocent”. (EQ)

180

Daniel Gutzmann and Erik Stei

Emphatic quotes are clearly not truth-conditionally relevant. Whatever makes the utterance (39) true in the given context makes the same utterance without QM true. The contribution of the quotation does not affect truth-conditions but rather expresses some additional meaning, e.g. that the writer is really convinced of Taylor’s innocence. Instances of anaphoric mixed quotation alluding to a famous citation and scare quotes apologizing for a somehow deviant expression etc. are also obvious cases of non-truth conditional contributions of quotations. (40)

a. b.

Life has “certain anomalous features”. At her university, many “geeks” would be pleased to help Mary with her computer.

However, for other interpretations of quotational constructions like MQ and SQ with an ironic reading, it is harder to decide whether they go beyond the truth-conditions of the quotational utterance or whether they manage to affect them. (41)

a. b.

Alice said that Quine is “difficult to understand”. Ew! I hate to eat at Peter’s. His bagels are always “fresh”.

It is hard to think of any procedure that can establish the truth-conditionality of the contribution by the quotes in (41) without any doubts. Here is something we could try: It seems as if the meaning of the quotations could be embedded under some higher operators or contexts like modal operators, conditionals or comparatives. (42)

Modal operators: a. It is possible that/Maybe Alice said that Quine is “difficult to understand”. b. It is possible that/Maybe Peter’s bagels are “fresh”.

(43)

Conditionals: a. If Alice says that life has “certain anomalous features”, she must know Quine’s work very well. b. If Peter’s bagels are always “fresh”, we should better avoid eating them.

(44)

Comparatives: a. It is more educated to say that life has “certain anomalous features” than that life has “certain anomalous properties”. b. Eating “fresh” bagels is worse than eating tasty bagels.

Quotation Marks and Kinds of Meaning

181

Examples like these seem to be troublesome for our attempt to treat the contribution by SQ and MQ as being conversationally implicated. However, we think that they actually only seem to be a problem. Even if the “whatever-itis” conveyed by the quotational constructions in (42)–(44) affected the truthconditions of the utterance in which it occurred, this does not mean that it cannot be conversationally implicated after all. One of the hot topics in the recent debate concerning the semantics/pragmatics interface is the question whether conversational implicatures at least can contribute to the truth-conditions of an utterance, i.e., whether they can be interpreted locally.20 Crucially, the constructions regularly discussed with regards to these questions are just the ones which seem to give rise to a truth-conditional effect of the quotations in (42)– (44). In such constructions, ordinary run-of-the-mill implicatures seem to be truth-conditionally relevant, too.21 (45)

a. b. c. d.

If Alice ate some of Peter’s bagels, there are still some bagels left. If you drink three beers and drive home, you will be arrested by the police. To read some of Grice’s papers is not as good as reading all of his papers. Driving home and drinking three beers is better than drinking three beers and driving home.

We do not want to enter the discussion about pragmatic intrusion here. Regardless of whether such cases are indeed cases in which conversational implicatures affect truth-conditions or whether they could be explained by a more detailed investigation of the pragmatic mechanism at hand, the crucial point is that there are cases in which the non-truth-conditionality of conversational implicatures is not as obvious as it is usually supposed to be. This is all we need to make our argument that the examples in (42)–(44) are not a serious objection to our attempt to analyze the contribution of quotations as being conversationally implicated. 3.

Some worries about the conversational implicature view

There is no free lunch, least of all in theories about quotational constructions in natural language. Accordingly, the conversational implicature view (CIV) 20. For discussion, cf. e.g. Blutner (1998, 2000), Levinson (2000), Recanati (2003), Sauerland (2004), King and Stanley (2005), Russell (2006), Geurts (2011). 21. For a comprehensive overview and detailed discussion of the intrusive construction, cf. Levinson (2000).

182

Daniel Gutzmann and Erik Stei

raises some worries as well. We consider the most challenging ones in the following sections. 3.1. Cancelability problems In § 2.4, we have seen that the effects indicated by means of QM exhibit the most central features of conversational implicatures as spelled out in (28). However, one important feature has not been addressed so far: the cancelability of conversational implicatures. At first sight, this is where genuine counterexamples to CIV, like e.g., (46)–(48), can be found. (46)

a. b. c.

(47)

a. b. c.

(48)

a. b. c.

C1 : What did Coffee & Bagels say about Peter’s bagels? Well, Peter’s bagels are always “fresh”. +> Peter’s bagels are what Coffee & Bagels called “fresh”. ?Well, Peter’s bagels are always “fresh”. But I don’t want to convey that Coffee & Bagels referred to them as fresh. C2 : Urgh, every time I eat a bagel at Peter’s I get sick! Well, Peter’s bagels are always “fresh”. +> Peter’s bagels are not fresh at all. ?Well, Peter’s bagels are always “fresh”. But I don’t want to convey that they are not fresh at all. C3 : Peter’s bagels are delicious! They are the best in town. Well, Peter’s bagels are always “fresh”. +> Peter’s bagels are extraordinarily fresh. ?Well, Peter’s bagels are always “fresh”. But I don’t want to convey that they are extraordinarily fresh.

(46c)–(48c) suggest that the implicatures indicated by QM are not comfortably cancelable. But this is not necessarily problematic for CIV, as there are other bona fide cases of conversational implicatures that are not comfortably cancelable either, namely subtractive implicatures.22 Their characteristic feature is that the speaker wants to convey the implicature only. The semantic content of the sentence is more or less irrelevant. Canceling subtractive implicatures leads 22. We borrow the terminology for this distinction from Levinson (2000: 406, fn. 59). For example, consider relevance implicatures due to thematic switch: (i) a. C: What do you think of Professor Smith? b. The weather will be fine today. +> I do not want to answer this question. c. ?The weather will be fine today but I do not want to convey that I do not want to answer this question.

Quotation Marks and Kinds of Meaning

183

to odd utterances, as there remains no relevant message the speaker rationally wants to convey. With this in mind, let us take a closer look at (46)–(48). It seems as if there are two ways in which one might think of canceling the effect indicated by QM. Firstly, one might want to cancel the indication that there is a marked interpretation to be derived (cf. the first step of the proposed inference (36)): (49)

a. b. c.

Peter’s bagels are always “fresh”. +> W wanted to convey something more/other than that Peter’s bagels are always fresh. ??Peter’s bagels are always “fresh”, but I do not want to convey anything more or other than that they are always fresh.

This option appears to be the reason why canceling implicatures indicated by QM seems so problematic. Indicating the need to process a marked interpretation of the expression between QM and then trying to cancel that indication is certainly odd, but this oddity is due to the apparent disorder of the writer’s intentions. With recourse to the notion of subtractive implicatures however, this problem can be explained without dismissing CIV: Analogous to the idea that subtractive implicatures only convey the implicature, QM-marked expressions only convey a non-standard interpretation. Thus, using QM without intending to convey anything more or other than the ordinary stereotypical meaning of the quoted expression would violate the I-principle without cause, and hence give rise to the suspicion that the speaker has quit being cooperative. This interpretation of the apparent counterexamples provides us with something like use conditions for quotation marks: One should only use QM if something more or other than the standard interpretation of an expression without QM is to be communicated. The second option is that a writer (W) wants to cancel a specific implicature indicated by the use of QM without canceling the indication itself, i.e., W wants to make sure the reader infers the intended “writer meaning”. In this case, there is just the ordinary mechanism of cancelability at work, independent of any specific effects due to QM. For illustration, consider the following example: Given the context described in (47), W realizes that the reader is about to draw an unintended conclusion and thus W intends to cancel the implicature. At first sight, (50d) still seems a bit odd, but consider some sequel like (50e) and the oddity will be considerably reduced. (50)

a. b. c.

C2 : Urgh, every time I eat a bagel at Peter’s I get sick! Peter’s bagels are always “fresh”. +> Peter’s bagels are not fresh at all.

184

Daniel Gutzmann and Erik Stei

d. e.

?Peter’s bagels are always “fresh”, but I do not want to say that they are not fresh at all. It’s just that Coffee & Bagels said they are fresh.

Of course, according to this sequel the QM in (50b) still indicate the need to process an implicature, but it is certainly a different one, rather conveying an MQ reading. This suggests that it is possible to cancel a specific implicature, but not without giving rise to another one. Thus, even though the procedural meaning of QM – indicating that a non-standard interpretation has to be derived – cannot be canceled without giving rise to the suspicion that the writer disobeyed CP, the implicature itself can be overwritten by another one. Given, of course, that more information about the intentions of applying QM is provided. This process is constrained by the non-standard interpretations of an expression that are rationally admissible in a certain context of utterance keeping CP, I and Q in mind. To sum up: The effect produced by the application of QM meets the criteria of context-dependency, nondetachability, reinforceability and calculability. There are certain problems concerning cancelability but these can be explained away by clarifying the role that QM play in indicating the implicature: Firstly, since QM increase the markedness of an expression, they pragmatically block the stereotypical interpretation of that expression, i.e., for any expression they block the generalized I-implicature to the stereotypical interpretation. Secondly, in the particular context of utterance, the specific interpretation of the expression enclosed in QM has to be fleshed out by means of a conversational implicature. Although the marking function of QM cannot be canceled without oddity, this does not apply for the implicature itself. 3.2. Truth-conditional effects and grammaticality The second worry is usually formulated, roughly, as follows: “How can the contribution QM make to the overall meaning of the sentence in which they occur be akin to conversational implicatures, given that in direct and pure quotation QM seem to have an impact on truth-conditions? After all, it is a central property of implicatures that they are not subject to truth-conditional evaluation.” Here are some examples typically brought forward in order to substantiate that worry: (51)

Context shifts in direct quotation a. [Maryi :] Peterk said myi bagels are fresh. b. [Maryi :] Peterk said, “Myk bagels are fresh”.

Quotation Marks and Kinds of Meaning

(52)

185

Meta-linguistic predication (mentioning) in pure quotation a. Boston is populous. b. “Boston” is disyllabic. c. Boston is disyllabic.

It is obvious that in the two instances of (51) there is a reference shift of the personal pronoun my. The most natural readings of (51a) and (51b) are the ones indicated by the indices: in (51a) my refers to Mary, whereas in (51b) it refers to Peter. It is fairly common to attribute this shift to the QM used in (51b). In (52a) and (52c) Boston, it is said, refers to the capital of Massachusetts, whereas in (52b) it refers to the word Boston. According to a classical view on quotation – based on Quine’s (1951) and Tarski’s (1983) seminal texts – this effect is due to the QM. Quotation marks or other devices, it seems, are necessary for certain types of metalinguistic predication. In case they are “missing”, as in (52c), the metalinguistic predicate takes the capital of Massachusetts as an argument, yielding an ungrammatical sentence. How do these points affect CIV? The arguments usually proceed as follows: (ATC) Argument from truth-conditions a. If QM indicate the need to derive conversational implicatures, then they should not have any influence on truth-conditions. b. However, as examples like (51) show, the occurrence of QM does have an influence on the truth-conditions of a sentence. c. Therefore, it is false that they indicate the need to derive conversational implicatures. (AG)

Argument from grammaticality a. If QM indicate the need to derive conversational implicatures, then they should not have any influence on the grammaticality of a sentence. b. However, as examples like (52) show, the occurrence of QM does have an influence on the grammaticality of a sentence. c. Therefore, it is false that they indicate the need to derive conversational implicatures.

If these arguments were correct, CIV would lose a lot of its plausibility as a general thesis on quotation marks. Even if one accepted CIV for some types of quotation, e.g., MQ, SQ and EQ, one could reasonably reject it for DQ and PQ, thus embracing some kind of homonymy view on QM.23 We think this move 23. Cf. Gutzmann and Stei (2011) for a discussion (and rejection) of this position.

186

Daniel Gutzmann and Erik Stei

is premature as there are arguments against the plausibility of ATC as well as AG. In order to show this, we challenge the second premise of each argument, i.e., we reject (ATC b) and (AG b). Let us take a look at (ATC b): we neither want to deny that there is a contextshift in (51) nor that it has an influence on the truth-conditions of the sentence. What we deny is that the context-shift is due to the application of QM. (53) suggests that punctuation marks and capital letters can indicate the context-shift just as good as QM can: (53)

a. b. c.

[Maryi :] Peterk said myi bagels are fresh. [Maryi :] Peterk said: Myk bagels are fresh. [Maryi :] Peterk said – myk bagels are fresh.

As in (51), we think that the indices in (53b) and (53c) indicate the most natural reading, namely the one in which there is a reference shift of the pronoun my to Peter. As far as we know, no one defends the thesis that colons, dashes or capitals have a direct influence on the truth-conditions of a sentence. We propose to take the same stance towards QM. The same disconnection between context-shifts and quotation marks holds the other way round: There are quotational constructions without any contextshift. This is obvious in the case of scare quotes (54a) and emphatics quotes (54b). However, there are also instances of mixed quotation in which indexicals are interpreted with respect to the context of the reporting utterance and not with respect to the reported one. This can be illustrated by the examples (54c) and (54d) which are taken from Johnson (this volume). (54)

a.

b. c. d.

C: Bill has mistaken Jane as Peter’s sister. Some days later, Bill and Peter are waiting for Jane. When she is finally approaching, Peter utters to Bill: Look, “my sister” is coming. (SQ) This could be “your new car”! (EQ) In the words of Gandhi, “if I had no sense of humor, I would long ago have committed suicide.” (MQ) As Abraham Lincoln said to the people of his time, “always bear in mind that your own resolution to succeed is more important than any other.” (MQ)

These examples combined with the argument presented above, suggest that quotation marks are neither a necessary nor a sufficient condition for a shifted context. If this is correct, then ATC collapses as the strong connection between QM and context-shifts underlying its second premise does not hold.

Quotation Marks and Kinds of Meaning

187

In the case of AG our rejection of the second premise is slightly more determined. We are not convinced by the claim that (52c) is an ungrammatical sentence. Although it is, of course, not a well formed expression in formal languages – for which Quine’s and Tarski’s theories were originally developed – it seems that the connection between QM and mentioning postulated by (AG b) is too strong for natural languages. Mentioning very often occurs without QM (cf. Gutzmann & Stei 2011). Consider the following examples: (55)

a. b. c.

My Name is Peter. The expression cats is a noun. Cats is a noun. (Saka 1998)

(56)

Scientists study the if. (Weinreich 1970: 94, after Klockow 1980: 53)

(57)

Er wurde mit vielen Ahs und Ohs begrüßt. (Klockow 1980: 56) He was greeted with many ahs and ohs.

(58)

Terebi bedeutet “television” […] (Klockow 1980: 62) Terebi means “television” […]

Proponents of AG are committed to the rather contentious claim that all these utterances are ungrammatical. This does not seem very plausible, least of all in the very frequent case of utterances structurally equivalent to (55a). Thus, the examples suggest that AG is implausible as the strong connection between QM and mentioning underlying its second premise does not hold in natural language. 4.

Conversational implicature and semantic input

Both ATC and AG depend on very strong premises. If our criticism is correct, then these arguments need better support in order to convincingly rebut CIV. There is, however, a more serious worry: According to Grice 1975, the derivation of conversational implicatures by means of CP and the conversational maxims needs fully propositional input. But, given CIV, it is not clear what a proposition expressed by an utterance of (52b) or, to consult another example, of (59b) should look like. (59)

a. b. c.

A lobster is an animal. “Lobster” is a word. Lobster is a word.

Apparently, whoever follows the arguments presented in the last paragraphs is committed to the claim that the word lobster has the same extension in all its occurrences in (59). This seems to follow from the thesis that the specific refer-

188

Daniel Gutzmann and Erik Stei

ence is calculated by means of conversational implicature, i.e., in other words, on post-semantic level. But then, what exactly is the propositional content of lobster in (59b)? A qualified answer to that question depends on the way one understands the semantics/pragmatics distinction. Of course, we cannot give anything like a conclusive summary of the theoretical implementations of this distinction in this paper. For illustrative purposes, however, we will briefly sketch CIV against the background of Gricean and Neo-Gricean approaches. According to Grice’s overall agenda, there are only two levels of meaning: what is said roughly corresponds to truth conditionally relevant semantic parts of the meaning of an utterance – with indexicals fixed and ambiguities solved – and what is implicated corresponds to the pragmatic part not contributing to semantic truth-conditions. Semantics has explanatory priority over pragmatics as the former serves as input for the latter. In other words, what is said is the basis for inferring what is implicated. Thus, the Gricean interpretation of CIV seems to be committed to the claim that the proposition expressed by a sentence containing QM is equivalent to the one expressed by its QM free counterpart. The most natural interpretation seems to be that (59b) and (59c) expresses the proposition THAT LOBSTERanimal IS A WORD from which the actual reference to the word lobster is inferred by means of CP and the conversational maxims. An unpleasant consequence of this move would be that most instances of direct and pure quotation would be false, even the ones in which QM are in fact applied. Another interpretation could run as follows: the question of whether an expression refers to an entity or to a word has to be disambiguated alongside with other disambiguation processes and the reference fixing of indexicals, i.e., before what is said is determined. This, however, would render the effects indicated by QM pre-propositional. Thus, they could no longer be regarded as conversational implicatures. Apparently, against the background of a classical Gricean picture of the semantics/pragmatics distinction, CIV is not a very attractive position. Neo-Gricean accounts like Stephen Levinson’s (2000) offer a more complex interpretation of the interaction between semantics and pragmatics on different levels of meaning. He expands the classical Gricean picture in allowing for the notion of pragmatic intrusion, according to which, roughly, pragmatic processes may play a role in determining the propositional content of an utterance. The idea is, roughly, as follows: The output of compositional semantics and indexical pragmatics does not always have fully propositional content. In this case, pragmatic processes, labeled Gricean pragmatics 1, are needed in order to arrive at truth evaluable content in the first place. Only then, once we have what is said, the usual pragmatic processes (Gricean pragmatics 2)

Quotation Marks and Kinds of Meaning

189

can be inferred as usual. Thus, Gricean pragmatics 1 influences the proposition expressed while Gricean pragmatics 2 is post-propositional.24 This architecture allows for a variety of incorporations of CIV: (a) analogously to the “Gricean” analysis given above, QM contribute to Gricean pragmatics 2 only. But with Levinson’s framework at hand, there are two further options: QM operate (b) completely on the level of pragmatics 1, or (c) QM operate on both levels. Against the background of this setup CIV appears more promising. Particularly, the thesis of pragmatic processes contributing to what is said might be used to accommodate examples like (59). Option (b) could be combined with several proposals. First, there is the notion of local expansion developed by Kent Bach (1994). (60)

a. b.

California is a long state. California is a long name.

Bach contends that the difference between (60a) and (60b) “does not correspond to anything specific to the name California – one can use any expression to refer to that very expression,” and suggests that mentioning is best regarded “as involving a special sort of local expansion: an expression ‘E’ is used as short for ‘the expression “E”’. The hearer can recognize such a use when, for example, ‘E’ occurs in subject position and the predicate is not plausibly applicable to E.” (Bach 1994: 153) CIV could be regarded as a suggestion of what the process of recognizing such a use might look like. Although, of course, Bach refrains from calling this process a conversational implicature, coining the term impliciture instead, CIV should be able to transfer the spirit of his proposal to the NeoGricean picture. We could then say that the process of recognizing whether the particular reference of an expression is its regular extension or rather the word itself, is located on the level of pragmatics 1: it helps determining the propositional content of the utterance under consideration. Other possibilities include adopting Saka’s (1998) view according to which orthographic, phonological or syntactic properties are genuine parts of the intension of an expression, or to side Pafel (2007, this volume), who treats pure quotation as a word formation process (a “generalized conversion”). In all cases the contribution indicated by QM would be located at the level pragmatics 1 since in Levinson’s Neo-Gricean picture disambiguation between different intensions or different syntactic/morphological structures happens at this level. Option (c), according to which both levels of pragmatics are involved, could be spelled out in terms of the derivation process we proposed in § 2.4. While 24. Cf. Levinson (2000: 188) for an illustrative visualization.

190

Daniel Gutzmann and Erik Stei

the first inference (36) – i.e., the indication that something more or other than the standard interpretation of the quotatum is intended by writer W – is located at the level pragmatics 1, the fleshing out of the target of the quotation happens at pragmatics 2 in form of the second inference in (37). The latter could then uncontentiously be called a conversational implicature. Thus, in case one is willing to call processes occurring at the level of pragmatics 1 conversational implicatures, CIV might become a plausible view after all. 5.

Conclusion

The analysis of QM occurrences in natural language carried out in §§ 2.1–2.3 showed that it is problematic to file the contribution which QM make to the overall meaning of an utterance under literal meaning, conventional implicature or to presuppositions. In § 2.4 we suggested that the notion of conversational implicature is more promising. We tackled some worries about CIV in § 3 and then showed in § 4 how it might be integrated into frameworks in the Gricean tradition. Of course, many points remain to be spelled out in detail but, given the arguments presented above, a first modest claim we wish to make is that the view put forward here deserves closer consideration. The second is that, terminological issues aside, a pragmatic explanation is most apt to explain all occurrences of QM in natural language. References Bach, Kent 1994 Bach, Kent 1999

Conversational impliciture. Mind & Language 9(2): 124–162. The myth of conventional implicature. Linguistics and Philosophy 22: 327–366.

Blutner, Reinhard 1998 Lexical pragmatics. Journal of Semantics 15: 115– 162. Blutner, Reinhard 2000 Some aspects of optimality in natural language interpretation. Journal of Semantics 17(3): 189–216. Cappelen, Herman & Ernie Lepore 1997 Varieties of quotation. Mind & Language 106(423): 429–450. Cappelen, Herman & Ernie Lepore 2005 Insensitive Semantics. Oxford: Blackwell.

Quotation Marks and Kinds of Meaning

191

Davidson, Donald 1984 Quotation. In Inquiries into Truth and Interpretation, D. Davidson, 79–92. Oxford: Oxford University Press. De Brabanter, Philippe (ed.) 2005 Hybrid Quotations. Amsterdam: John Benjamins. Geurts, Bart 1999

Presuppositions and Pronouns. Oxford: Elsevier.

Geurts, Bart 2010

Quantity Implicature. Cambridge: Cambridge University Press.

Geurts, Bart & Emar Maier 2005 Quotation in Context. In Hybrid Quotations, P. de Brabanter (ed.), 109–128. Amsterdam: John Benjamins. Grice, H. Paul 1975

Logic and conversation. In Syntax and Semantics 3: Speech Acts, P. Cole & J. Morgan (eds.), 41–58. New York: Academic Press.

Gutzmann, Daniel & Erik Stei 2011 How quotation marks what people do with words. To appear in Journal of Pragmatics. Horn, Laurence 1984

Horn, Laurence 2005

Toward a new taxonomy for pragmatic inferencing: Q-based and R-based implicature. In Meaning, Form, and Use in Context: Linguistic Applications, D. Schiffrin, 11–42. Washington, DC: Georgetown University Press. Implicature. In Handbook of Pragmatics, L. Horn and G. Ward (eds.), 3–28. Oxford: Blackwell.

Johnson, Michael this vol. The punctuation theory of quotation. King, Jeffrey & Jason Stanley 2005 Semantics, pragmatics, and the role of semantic content. In Semantics vs. Pragmatics, Z.G. Szabó (ed.), 111–164. Oxford: Oxford University Press. Klockow, Reinhard 1976 Gänsefüßchen-Semantik. In Sprachtheorie und Pragmatik, H. Weber & H. Weydt (eds.), 235– 245. Tübingen: Niemeyer. Klockow, Reinhard 1978 Anführungszeichen, Norm und Abweichung. Linguistische Berichte 57: 14–24. Klockow, Reinhard 1980 Linguistik der Gänsefüßchen. Frankfurt am Main: Haag und Herchen.

192

Daniel Gutzmann and Erik Stei

Levinson, Stephen 1983 Pragmatics. Cambridge: Cambridge University Press. Levinson, Stephen 2000 Presumptive Meanings. Cambridge, MA: MIT Press. McCawley, James D. 1978 Conversational implicature and the lexicon. In Syntax and Semantics IX: Pragmatics, P. Cole (ed.), 245–259. New York: Academic Press. Meibauer, Jörg 2006 Pafel, Jürgen 2007

Pafel, Jürgen this volume

Implicature. In Encyclopedia of Language and Linguistics, volume 5, 2nd edition, K. Brown (ed.), 568–580. Oxford: Elsevier. Ein Essay mit dem Titel ‘On pure quotation’. In Zitat und Bedeutung, E. Brendel, J. Meibauer and M. Steinbach (eds.), 201–214. Hamburg: Buske. Two dogmas on quotation.

Potts, Christopher 2005 The Logic of Conventional Implicature. Oxford: Oxford University Press. Potts, Christopher 2007 The dimensions of quotation. In Direct Compositionality, C. Braker and P. Jacobson (eds.), 405–431. Oxford: Oxford University Press. Predelli, Stefano 2003 Scare quotes and their relation to other semantic issues. Linguistics and Philosophy 26: 1–28. Quine, Willard Van Orman 1951 Mathematical Logic. Cambridge, MA: Harvard University Press. Recanati, François 2001 Open quotation. Mind 110(439): 637–687. Recanati, François 2003 Embedded implicatures. Philosophical Perspectives 17: 1299–1332. Russell, Benjamin 2006 Against grammatical computation of scalar implicatures. Journal of Semantics 23: 361–382. Saka, Paul 1998 Sauerland, Uli 2004

Quotation and the use-mention distinction. Mind 107(435): 113–135. Scalar implicatures in complex sentences. Linguistics and Philosophy 27: 367–391.

Quotation Marks and Kinds of Meaning Tarski, Alfred 1983

193

Der Wahrheitsbegriff in den formalisierten Sprachen. In Logik-Texte. Kommentierte Auswahl zur Geschichte der modernen Logik, 3. erw. Auflage, K. Berka and L. Kreiser (eds.), 445–546. Berlin: AkademieVerlag.

Washington, Corey 1992 The identity theory of quotation. Journal of Philosophy 89: 582–605. Weinreich, Uriel 1970 Erkundungen zur Theorie der Semantik. Tübingen: Niemeyer.

Quotation and Pictoriality Manfred Harth

Abstract. Any theory of quotation has to account for the pictoriality of quotation, i.e. the fact that you literally can see or hear what a quotation refers to and that a quotation resembles its referent. Proponents of the Demonstrative Theory of Quotation claim that the Identity Theory is not able to account for this fact. After explicating the relevant notion of pictoriality as exemplification of the quoted expression (type) by one of its tokens, I shall argue that my version of this theory, the Exemplification Theory, can explain the pictoriality by showing that exemplification is the constitutive feature of quotational (self-) reference. Then it is argued that the Demonstrative Theory cannot leverage the double duty, which a quoted expression carries out in mixed quotation. Moreover, two objections to the Demonstrative Theory are put forward: the dispensability of quotation marks and the context-insensitivity of quoting sentences. In conclusion, the Exemplification Theory, on balance, is shown to be preferable to the Demonstrative Theory.

1.

Introduction

“A quotation is not a description, but a hieroglyph; it designates its object […] by picturing it” says Quine (1961: 26) – who often is said to hold a propername theory of quotation – and it is widely uncontested that pictoriality is a distinctive feature of quotation. Any theory of quotation has to account for it, although it is not quite clear what “picturing” exactly means in this context. Nevertheless, Davidson claims that the pictoriality of quotation disproves the so-called Identity Theory of Quotation, which he, ironically, calls the “picture theory”, according to which, the quoted expression refers to itself. Davidson and most of his followers further maintain that the Demonstrative Theory of Quotation, or some version of it, e.g. García-Carpintero’s Deferred Ostension Theory, is the only theory that is able to account for the pictoriality of quotation. According to the Demonstrative Theory, pictoriality is to be explained by the fact that a token of the quoted expression is exhibited or displayed. Its crucial point is that the quoted expression does not belong to the quoting sentence, i.e. the sentence by which an expression is quoted. The exhibited token is external to the quoting sentence and referred to by demonstrative reference via

196

Manfred Harth

quotation marks. These are disguised demonstrative descriptions, which are to be read as: (D)

the expression (type) of which this is a token.1

With respect to the Identity Theory, Davidson writes: The difficulty is this. What is wanted is an explanation of how quotation enables us to refer to expressions by picturing them. But on the present theory, quotation marks create a context in which expressions refer to themselves. How then does picturing feature in the theory? If an expression inside quotation marks refers to itself, the fact that it also pictures itself is simply a diverting irrelevancy. […] In brief, once the content of the quotation is assigned a standard linguistic role, the fact that it happens to resemble something has no more significance for semantics than onomatopoeia […] The picture theory suggests no way to bring an inscription or utterance into the picture. This could be done only by describing, naming, or pointing out the relevant token, and no machinery for the purpose has been introduced. (Davidson 1984: 84 and 85)

The machinery Davidson introduced by his paratactic theory is pointing out the exhibited token, i.e. demonstrative reference via quotation marks. The proponent of a demonstrative account has to pay a heavy price for this: it is the awkward fact that the quoted expression is, from a semantic point of view, no integral part of the sentence which, from a pre-theoretical point of view, comprises the quotation as a grammatical part. What Davidson gains for this price is a standard, or nearly standard, logical approach to quotation: first order predicate logic extended by demonstratives. Yet, I think the price is too heavy. There are consistent and comprehensive logical systems of quotation that do regard the quoted expression as an integral part of a quoting sentence.2 There is, from a logical point of view, simply no need – and finally no plausible chance – to confine to first order logic in analysing quotation and, therefore, to reduce quotation to demonstrative reference. So logic does not decide the matter and pictoriality is still the point at issue. For the question remains as to how picturing can feature in a theory that does not regard quotation as pointing out an exhibited token, but as self-reference of the quoted expression (type) – as self-reference apparently has nothing to do with picturing. The primary aim of this paper is to show how picturing does feature in the Identity Theory, or at least in my own version of it, and that picturing is not a “diverting irrelevancy”, but a constitutive feature of quotational (self-) reference. 1. There are also refined versions of that paraphrase (cf. Cappelen and Lepore 1997, 1999 and 2005), but nothing hinges on my choosing (D). 2. See e.g. Blau (1988), Harth (2002), Kubinski (1965), Wray (1987).

Quotation and Pictoriality

197

Within the general framework of the Identity Theory, I will deliver an explanation of how quotation enables us to refer to expressions by picturing them. For this I will first explain the notion of pictoriality as employed in the context of quotation. Then my version of the Identity Theory, the Exemplification Theory, is introduced, and it is shown how inscriptions or utterances feature in the theory. Finally, some difficulties for a demonstrative account are presented and it is argued that, on balance, the Exemplification Theory is the preferable account on quotation. 2.

Pictoriality

That quotations are pictorial, that they are more like pictures than descriptions, is an intuition that is expressed in the following pre-theoretic characterisations of quotation: You can literally see (or hear) what a quotation refers to – but you cannot literally see what a description refers to. A quotation resembles or looks like its object or referent – a description, usually, does not. Similar characterisations can be found in pre-philosophical thinking about ordinary pictures (at least some kind of pictures, e.g. portraits): You can literally see what a picture refers to, what it depicts. A picture resembles or looks like its object or referent. Yet, quotations are not pictures, at least not in the ordinary sense. The notion of pictoriality involved in quotation has to be understood in a broader sense. This sense, which is already suggested in Davidson’s paper and elaborated by García-Carpintero (1994, 2004), is to be explicated in terms of exemplification or ostension. Quotation is pictorial in the sense that the quoted expression (type) is exemplified by a sample, i.e. by one of its tokens. Exemplification is conceived of here as a kind of reference. It is, in a primary sense, to be regarded as an act of reference performed in the course of an assertion (or question etc.), that is, as an act of reference to a type by presenting one of its tokens. But, in a derivative sense, it can be conceived of as a relation between token and type. The relation is explicated as follows (cf. Harth 2002: 109ff): (E)

A token t exemplifies an expression e iff t is a token of e and a person s exhibits t and, by exhibiting it, s thereby refers to e.

198

Manfred Harth

The referent of an exemplification, so far, is regarded as an expression type, e.g. a letter, a string of letters, a word or a sentence, but we could easily broaden the definition (E) and define exemplification as reference to some feature instantiated by or associated with the exhibited token: its type, its shape, its phonetic form, its meaning etc. The expanded definition reads: (E*)

A token t exemplifies a feature f iff f is instantiated by or associated with t and a person s exhibits t and, by exhibiting it, s thereby refers to f.

The kind of picturing involved in quotation can now be explained by (P). (P)

A quotation pictures its object or referent in the sense that the quoted expression (type) or some other feature instantiated by or associated with the exhibited token of that expression is exemplified.

Any theory of quotation has to account for (P); and the Identity Theory is said to being not able to account for it, since the alleged kind of reference of the quoted expression to itself does not involve exemplification or picturing. Picturing is said to be simply irrelevant to (quotational) self-reference. In the following, I will argue that, according to the Identity Theory, adequately understood, exemplification is not irrelevant – quite the contrary: exemplification is the constitutive feature of quotational self-reference, since exemplification originates self-reference in the first place. Exemplification explains why and how an expression inside quotation marks refers to itself (or some feature associated with the exhibited token). Quotational self-reference is not, as Davidson supposed, a standard linguistic role; it is non-standard and created by or within a certain context – often indicated by quotation marks – in which linguistic expressions play a new and non-standard referential role originated from exemplification. 3.

The exemplification theory of quotation

The Identity Theory – as it has been introduced into the current debate by Corey Washington (1992) and is recently defended by Paul Saka (2006) – is as old as medieval logic. Quotation was then called suppositio materialis (see e.g. Bochenski 1956 and Kretzmann 1982). The theory says that an expression at a certain position does not have its standard referential role (suppositio formalis) – a name, e.g., does not refer to the bearer of the name – but has a non-standard role, according to which the quoted expression refers to itself. Quotation marks help indicating the non-standard role, which is an ad hoc in-

Quotation and Pictoriality

199

terpretation of the expression at that special position. The standard role of the quoted expression, that is, its (conventional) meaning is thereby not affected; nor are linguistic expressions regarded as systematically ambiguous. That is to say, the non-standard self-reference of the quoted expression is not a further conventional role; it temporally replaces the standard role in the special context of a quoting sentence. For this reason, to quote foreign expressions, or even non-linguistic signs, does not pose a problem for the Identity Theory.3 The Exemplification Theory accepts this general framework of the Identity Theory, but modifies it in two respects. (i) The first modification is related to the flexibility, or multiplicity, of quotation. The fact that quotation is flexible or multiple, in the sense that by quotation not just the quoted expression type can be referred to, but also other related features, has been put forward as an objection to the Identity Theory (see García-Carpintero 1994 and Saka 1998). Examples for this sort of flexibility of quotation are: “House” rhymes with “mouse” “Color” is the same (word) as “colour” “Hund” means “dog” Yet, the Exemplification Theory is more flexible than the (original) Identity Theory as regards the referent of a quotation. As already explicated in (E*), the referent of an exemplification by a token is not necessarily the expression type that is instantiated by the token, but it might be some other object or feature instantiated by or associated with the token: its shape, its phonetic type, its meaning etc. What actually is referred to by a quotation is determined in the context of the quoting sentence or its utterance. Due to this modification the Exemplification Theory is not affected by the argument from flexibility. (ii) The second modification regards the pictoriality of quotation, which by proponents of the Demonstrative Theory is said to disprove the Identity Theory. The modification accounts for the pictoriality, since the Exemplification Theory – as the name already suggests – explains quotational reference or selfreference, i.e. the non-standard role of a quoted expression, as exemplificational reference. That is to say, within an inscription or utterance of a quoting sentence a token of the quoted expression is exhibited at a special position within the sentence, today often inside quotation marks, and it is (contextually) interpreted as exemplifying its type (or some related feature). The exemplification of the expression by one of its tokens, i.e. the referential relation from token to type, 3. The argument from universality of quotation is put forward e.g. by Müller (1996) and it is also rebutted by Saka (2006).

200

Manfred Harth

from a logical point of view, has to be regarded as reference of the expression to itself. For considered within the limits of formal semantics or logic, i.e. considered on the level of linguistic types, the exemplification of the type by one of its tokens has to be modelled as reference of the type to itself, since formal logic deals with types and tokens cannot be taken into consideration. That is to say, the self-reference, which in the logical analysis is ascribed to the quoted expression, just plays the part of the exemplificational reference, which the token of the quoted expression makes in the communicative practice. Exemplification cannot be captured within the limits of formal semantics and logic, since on this level of explanation (or abstraction) we exclusively deal with types and exemplification is a relation a particular token bears to a type. This vertical relation, as it were, from token to type has to be modelled by a horizontal relation from type to itself (or some related abstract feature). In this sense, exemplification, or picturing, is not just a “diverting irrelevancy”, but it constitutes the quotational reference of the quoted expression to itself. If we seek a more or less “informal” explanation of quotational reference, the Exemplification Theory offers exemplificational reference by the token to its type. But in seeking a formal semantic account and a logical analysis we have to confine to a semantic relation a type can bear – which is, in the case of quotation, reference to itself (or some associated feature).4 The pictorial, i.e. exemplificational nature of quotation cannot fully be captured by a logical analysis in a formal language. This nature, as it is a non-linguistic, since pictorial means, is reserved to natural language, to speaking and writing, i.e. producing and perceiving tokens. Thus, the Exemplification Theory is indeed able to capture the basic Davidsonian insight that a token of the quoted expression is exhibited in order to refer to its type. The Davidsonian mistake simply is to take quotational reference to be demonstrative. So the difference between the Exemplification Theory and the Demonstrative Theory can be formulated thus: According to the Exemplification Theory, quotational reference is not demonstrative via quotation marks to an exhibited token external to the quoting sentence, but it is exemplificational via the exhibited token itself, which is an integral part of the quoting sentence. 4.

Double duty

Now it seems that the Exemplification Theory is the preferable theory of quotation, since it has the advantage that the quoted expression is a semantic and 4. For the logic of quotation within the framework of the Exemplification Theory see Harth (2002: 163–170 and 2004: 228–232).

Quotation and Pictoriality

201

logical part of the quoting sentence and is able to account for the pictoriality of quotation. Yet, in addition to the pictoriality challenge, another objection has been put forward, which I will rebut now. It is the argument from double duty, which says that in the logical form the Identity Theory is assigning to quoting sentences “no single component serves […] two functions; and so, [the Identity Theory] fails to explain how a token can have this double function” (Cappelen and Lepore 1997: 436). It is the double function a token has in so-called mixed quotations, as a token of “hieroglyph” has in sentence (1)

Quine says that a quotation is a “hieroglyph”

or a token of “Ghosttown” has in the jocular sentence (Saka 1998) (2)

“Ghosttown” has more letters in it than people.

It is claimed that the paratactic analysis is preferable to the Identity Theory’s analysis since, according to the Identity Theory, the double function gets lost, as it is split up and spread onto two occurrences of the quoted expression, as e.g. in the paraphrase (3)

“Ghosttown” has more letters in it than Ghosttown has people.

Yet, this argument from double duty is not convincing – for the following reason: A double function, in principle, cannot be captured by a single component of a logical formula. A double function within a logical formula just would contradict the mere point of logical analysis and formal language, which has to be rid of vagueness, ambiguity and also double function. Double duty of an expression is a feature of the expression’s surface structure. Yet the fact that a certain component serves two functions at the surface of an English sentence certainly is not a sufficient reason to exclude the component from the logical form, in order to preserve the double function, and so to opt for a paratactic analysis. In considering Saka’s example, (2)

“Ghosttown” has more letters in it than people,

an acceptable logical form is given on the basis of the paraphrase (3)

“Ghosttown” has more letters in it than Ghosttown has people.

True, the joke disappeared (but logical forms are never funny) and so did the double function of “Ghosttown”. Yet this is no reason to exclude the name “Ghosttown” from the logical form, in order to retain the double function, and to

202

Manfred Harth

prefer a paratactic paraphrase. Anyway, a suitable paratactic paraphrase of (2) is not (4), as one should expect, but has to be (4*): (4)

(4*)

Ghosttown The expression of which this is a token has more letters in it than people. Ghosttown The expression of which this is a token has more letters in it than the referent of the expression of which this is a token has people.

Prima facie, (4) should be the paratactic paraphrase of (2), but (4), in my judgement, is just nonsense, unlike (2), and not even funny. Yet, in paraphrase (4*), which has to be chosen instead of (4), the joke disappeared, too – and so did the (alleged) advantage of the Demonstrative Theory. Granted, the exhibited token of “Ghosttown” serves two times as a sample to which is demonstratively referred to. But “Ghosttown” is outside the quoting sentence, and it cannot be a suitable constraint on logical analysis that a double function of an expression should be maintained at the cost of an exclusion of the expression from the logical form. A reasonable constraint for logical analysis just demands that the surface structure is to be respected – as far as possible and relevant. But in respecting the surface structure of English the Exemplification Theory certainly is preferable to a paratactic account, anyway, since the logical form assigned by the Exemplification Theory comprises the quoted expression as an integral part, just like the original English quoting sentence does, and the paratactic paraphrase does not. In conclusion, the argument from double duty fails, and the Exemplification Theory is not in an inferior position – quite the contrary, there are at least two serious troubles with the Demonstrative Theory. 5.

Dispensability

Since about 200 years quotation marks (in the western world) are conventional devices in script. In speech there are usually no conventional counterparts to them – although intonation devices are sometimes employed, as e.g. a pause before and after the quoted expression – and in ancient languages, e.g. in Latin, (conventional) quotation marks are even totally absent. Quotations in speech and script lacking conventional marks have to be detected by sentential features or non-linguistic contextual cues. So a satisfactory theory of quotation should account for two related facts of dispensability: First, no special linguistic sign is necessary in order to quote or to signal quotation. Second, in a

Quotation and Pictoriality

203

language that is provided with conventional quotation marks one can dispense with them, at least in some cases, without being ungrammatical (cf. Saka 2006: 463f). The argument from dispensability against the Demonstrative Theory, which was originally stated by Washington (1992) and Saka (1998, 2006) and is based on the facts that quotation marks can be omitted in script without ungrammaticality and that speech lacks them at all, can be illustrated by asking: how would a medieval Demonstrative Theory for Latin look like? The medieval proponent of this theory would have to claim that the demonstrative description “the expression of which this is a token”, i.e. the quoting sentence’s subject, is in a sense implicit in quoting sentences like “Homo verbum est”. García-Carpintero hints at two ways as to how quotation marks could be taken to be implicit in sentences lacking them at their surface: “to consider obvious contextual clues as realizations of them; to explain their appearance by resorting to the mechanism of conversational implicature” (1994: 263). Regarding the first, he concedes that “to accept contextual clues as the realization of the subject of a sentence […] seems difficult to swallow” and concerning the second way he acknowledges that it “leaves the Davidsonian in a worse position than one occupied by the Fregean [i.e. the Identity Theory]” (1994: 264). Although the assumption of implicit quotation marks seems to be congenial to the Identity Theory, it might also be available for the Davidsonian, e.g. by resort to a Gricean mechanism: The omission of the demonstrative description might be explained on the basis of a superficial ungrammaticality, e.g. “ukqaj is nonsense”, or an obvious absurdity, irrelevancy or falsehood of the sentence in its non-quotation reading, e.g. “Tom has three characters” or “Madonna is in vogue”. Assuming that the speaker is demonstrating the relevant expression is the most appropriate way to take him or her as cooperative, i.e. as saying something at all or as telling something that is informative, relevant or held to be true. Nevertheless, the Gricean explanation is more suitable to the Identity Theory, since the Davidsonian has to explain the semantic exclusion of the quoting sentence’s seeming subject, the quoted expression, and the omission of the actual subject, the demonstrative description. The adherent of the Identity Theory, basically, has to explain a self-referential non-standard interpretation of the quoted expression, which also could be regarded as a sort of disambiguation (cf. Saka 1998 and 2006).5 So there is an advantage of the Identity Theory over the Demonstrative Theory 5. Though (ordinary) disambiguation would imply that the quotational reading is due to the (lexical) meaning of the quoted expression. This is rather implausible. On my view, the quotational reading is best regarded as a form of divergent interpretation sui generis – neither (ordinary) disambiguation nor (ordinary) implicature.

204

Manfred Harth

with regard to the dispensability of quotation marks, which is even granted by a proponent of the demonstrative account, viz. García-Carpintero (1994). 6.

Context-sensitivity

The second objection to the Demonstrative Theory regards context-sensitivity. Its origin is the fact that, according to the theory, the quoted expression does not belong to the sentence quoting it and the reference to it is demonstrative. According to English surface grammar, though, quotation marks and whole quotations are not context-sensitive expressions.6 A quoting sentence, with respect to the quotation, is true or false whether or not an inscription or utterance of the sentence exists. A quoting sentence (type) can be true, although the sentence never was or will be uttered or written. With this observation, problems arise for the Demonstrative Theory.7 Quoting sentences, paraphrased paratactically, are true only if an inscription or utterance of such a sentence, especially a token of the quoted expression, exists. For the paraphrase of quotation marks, as any demonstrative description, (semantically) presupposes the existence of a demonstrated object. This object is, in the case of quotation, the exhibited token (referred to by the demonstrative “this”). If there happens to be no object to be demonstrated and referred to by a demonstrative, a description comprising the demonstrative fails to refer and the sentence fails to be true (or false). But, as an example, any English sentence of the following form – or the proposition expressed by it – is true: (5)

“e” is identical to “e”

(5) is true of any expression e to be substituted for the letter “e”, not only true of those expressions of which a token exists. Moreover, the paratactic paraphrase of an identity statement of the form (5), viz. (5*)

The expression of which this is a token is identical to the expression of which this is a token. e e

is not analytically true, as opposed to the original English statement. 6. An expression is context-sensitive just in case it is part of its meaning that its semantic value, e.g. its referent, is only determined with respect to a particular utterance of the expression, i.e. with respect to speaker, time, place, communicative intentions etc. Paradigmatic cases are indexicals like “I” and “now”. 7. A similar objection has been stated by Seymour (1994) against a paratactic theory of direct and indirect discourse.

Quotation and Pictoriality

205

In defence, the Davidsonian might deny that sentences are the truth-value bearers and hold that only utterances of sentences, or assertions made by those utterances, can be evaluated as true or false. And the Davidsonian might point at the fact that any English quoting assertion does imply the existence of a token of the quoting sentence and the quoted expression, such that the truth or falsehood of any quoting sentence is guaranteed. However, this sort of implication is not the relevant sense of presupposition. The relevant sense is illustrated by the following example. “Peter found a dog” presupposes the existence of something found by Peter, whereas “Peter is looking for a dog” does not presuppose the existence of something Peter is said to look for. But “John is looking for this dog” does presuppose the existence of a demonstrated object (dog) on grounds of the meaning of the demonstrative “this”. The general claim that only utterances or assertions are truth-evaluable and that, therefore, an assertion made by a quoting sentence implies, or guarantees, the existence of a token of the quoted expression has nothing to do with presupposition. For further illustration have a look at the following two sentences: (6) (7)

This is a pronoun. “This” is a pronoun.

Sentence (6) is context-sensitive and presupposes a demonstrated object (token), but (7) is not context-sensitive. Instead of claiming that any utterance of (7) carries about with it a part of its own context of utterance, viz. a token of the word “this”, it is more intelligible to say that the token is part of the utterance of (7) itself and not part of the context of utterance. Hence, two problems for the Demonstrative Theory arise: (i) Paratactic paraphrases of quoting sentences presuppose the existence of a (demonstrated) token of the expression inside quotation marks, while English quoting sentences do not. (ii) The paratactic paraphrases of quoting identity statements are not analytic, whereas their original English counterparts are. An adherent of the Demonstrative Theory might counter that these objections are begging the question: If syntactical integration of the quoted expression within a quoting sentence is not already granted, the demonstrative analysis is still open to us, although it diverges from surface grammar. The contextinsensitivity of English quoting sentences is a misleading feature of the surface structure of English – misleading with respect to logic and semantics. The paratactic analysis reveals a presupposition that does not appear at the surface.

206

Manfred Harth

Yet, this story is implausible, since presuppositions are features based on the meaning of certain expressions, therefore presuppositions either appear at the surface or do not exist. If a presupposition is assigned to a sentence while there is none in the sentence meaning (or the surface), then the analysis of the sentence is defective. Perhaps the Davidsonian might reply that logical analysis cannot respect any semantic feature of the original and that a logical form sometimes happens to show features that the original lacks, and vice versa. So, for the Davidsonian, the presupposition, or the context-sensitivity in general, might not be a serious problem, but just a minor blemish. Be that as it may, even if it is granted that these are not mortal problems for the Demonstrative Theory, they provide good reason to prefer an alternative lacking these problems – and the Exemplification Theory is such an alternative. Moreover, recall that, on the one hand, as regards the double function in mixed quotation the proponent of a paratactic account relies on the surface structure in order to argue against the Identity Theory, but, on the other hand, in defence of the Demonstrative Theory the Davidsonian lays claim to a misleading surface as regards the alleged presupposition and the semantic integration of the quoted expression. This is incoherent reasoning. 7.

Conclusion

On balance, the Exemplification Theory is to be favoured – for the following reasons: First, it is preferable to the Demonstrative Theory for at least three reasons: (a) The Exemplification Theory is the more natural approach. That is, it copes better with our pre-theoretical intuitions and with surface structure, whereas proponents of the Demonstrative Theory have to accuse surface grammar for misleading us. (b) Regarding the dispensability of quotation marks in speech, and even sometimes in script, and their total absence in ancient languages the Demonstrative Theory is in a much weaker position than the Exemplification Theory. (c) According to the Demonstrative Theory, the logical form of quoting sentences shows two features which the original English sentences lack: Quoting identity statements, i.e. sentences of the form (5), are not analytic and the paratactic paraphrases of quoting sentences presuppose that a token of the quoted expression exists, whereas the original English identity statements are analytic and English quoting sentences do not presuppose the existence of a token of the quoted expression.

Quotation and Pictoriality

207

Secondly, the Demonstrative Theory is in no respect preferable to the Exemplification Theory. Although the demonstrative approach is sometimes held to be the only theory that accounts for the pictoriality of quotation, the Exemplification Theory adjusts this alleged shortcoming of the Identity Theory. Thus the Demonstrative Theory does not prove to be preferable to the Exemplification Theory with respect to pictoriality. Not even regarding double duty is the Demonstrative Theory to be favoured. So to Davidson’s claim that “it’s hard to find an alternative to the [demonstrative] theory [Cappelen and Lepore] promote …” (1999, 101) I would like to reply: it might be hard, but here it is. References Bennett, Jonathan 1988 Quotation. Noûs 22: 399–418. Blau, Ulrich 1988

Die Logik der Anführung und Quasianführung. Erkenntnis 29: 227– 268.

Bochenski, Josef 1956 Formale Logik. Freiburg: Alber. Cappelen, Herman & Ernie Lepore 1997 Varieties of quotation. Mind 106: 429–450. Cappelen, Herman & Ernie Lepore 1999 Using, mentioning and quotes: A reply to Saka. Mind 108: 741–750. Cappelen, Herman & Ernie Lepore 2005 Varieties of quotation revisited. Belgian Journal of Linguistics 17: 51–75. Davidson, Donald 1984 Quotation. In Inquiries into Truth and Interpretation, D. Davidson, 79–92. Oxford: Oxford University Press. Davidson, Donald 1999 Reply to Herman Cappelen and Ernie Lepore. In Donald Davidson. Truth, Meaning and Knowledge, Zeglen, Ursula (ed.), 100–101. London: Routledge. García-Carpintero, Manuel 1994 Ostensive signs: Against the identity theory of quotation. The Journal of Philosophy 91: 253–264. García-Carpintero, Manuel 2004 The deferred ostension theory of quotation. Nous 38: 674–692.

208

Manfred Harth

García-Carpintero, Manuel 2005 Double duty quotation. Belgian Journal of Linguistics 17: 89–108. Harth, Manfred 2002 Harth, Manfred 2004

Anführung. Ein nicht-sprachliches Mittel der Sprache. Paderborn: Mentis-Verlag. Anführung: Mit Exemplifikation zur Denotation. Facta Philosophica 6: 223–248.

Kretzmann, Norman 1982 The Cambridge History of Later Medieval Philosophy. Cambridge: Cambridge University Press. Kubinski, Tadeusz 1965 Two kinds of quotation mark expressions in formalized languages. Studia Logica 17: 31–47. Müller, Olaf 1996

Zitierte Zeichenreihen. Erkenntnis 44: 279–304.

Quine, Willard Van Orman 1961 Mathematical Logic. 2nd edition. Cambridge, MA: Harvard University Press. Saka, Paul 1998

Quotation and the use-mention distinction. Mind 107: 113–135.

Saka, Paul 1999

Quotation: A reply to Cappelen & Lepore. Mind 108: 751–754.

Saka, Paul 2006

The demonstrative and identity theories of quotation. The Journal of Philosophy 103: 452–471.

Seymour, Michel 1994 Indirect discourse and quotation. Philosophical Studies 74: 1–38. Simchen, Ori 1999

Quotational mixing of use and mention. The Philosophical Quarterly 49: 325–336.

Steinbrenner, Jakob 2004 Zeichen über Zeichen. Heidelberg: Synchron. Washington, Corey 1992 The identity theory of quotation. The Journal of Philosophy 89: 582– 605. Wray, David 1987

Logic in quotes. Journal of Philosophical Logic 16: 77–110.

The Punctuation Theory of Quotation1 Michael Johnson

1.

Introduction

Quotation is a term which covers the phenomena exhibited in (0a–c) and possibly (0d): (0)

a. b. c. d.

“Dog” has three letters. “I love sandwiches,” said John. Ernestine said that she was going to “that concert with that guy in the band.” George is a “sanitation engineer.”

The use in (0a) is known as pure quotation; in (0b) direct quotation; in (0c) mixed quotation; and in (0d) scare quotation (if you will). Establishing an effective procedure for determining which of these three categories, if any, a particular word or phrase belongs to can be difficult; for instance, we can effect pure quotation without quotation marks, as with italics, intonation, or indentation; nor does every direct quotation follow or precede the words ‘say,’ ‘says,’ or ‘said’ – sometimes they follow ‘ask’ or ‘think’, and sometimes they appear bare, as in dialogue. I, like the literature, will assume that my reader has some intuitive grasp on what counts as what and continue from there. In this paper I will argue that the effects of direct and mixed quotation2 are a pragmatic phenomenon, rather than a semantic one. That is, two sentences that differ only in the presence or absence of direct or mixed quotation have the same literal semantic meaning, though they differ in what they convey. The first part of the paper will be concerned with arguing that a pragmatic treatment is 1. Thanks are due to Daniel Altshuler, Sam Cumming, Gabriel Greenberg, Ernie Lepore, Sarah Murray, Jennifer Nado, Chung-chieh Shan, the participants of RULing 2007, and the participants of the International Conference on Quotation and Meaning 2007. 2. The reader is cautioned that I neither wish nor intend my claims to extend to pure quotes, scare-quotes, greengrocer’s quotes, or any uses of quote marks that are not clearly direct or mixed uses.

210

Michael Johnson

to be preferred because no semantic treatment can adequately handle the data. In the second part of the paper, I shall consider the main obstacle to pragmatic accounts, indexicality, and suggest how the obstacle might be overcome. Then I shall turn to the specific form a pragmatic theory of direct and mixed quotation might take: in particular, I will argue that the entailments that arise from the use of quotation marks are sui generis, in that they are not presuppositions, conversational implicatures, or conventional implicatures. In the conclusion, I will speculate wildly on the nature of quotation. 2.

Semantic theories of quotation

A semantic theory of direct/mixed quotation is a theory that claims that quotation marks have their effect in the semantics – that they alter the literal meaning of sentences without such punctuation. Such theories of quotation naturally divide themselves into three types: pure-mention theories, dual use-mention theories, and pure-use theories. Here it will be argued that no such theory is adequate. 2.1. Pure-mention theories A pure-mention theory is a theory on which the quoted material in a direct/mixed context designates, denotes, refers to, names, or otherwise mentions words, expressions, shapes, sounds, symbols or the like and where furthermore this is the material’s only contribution to the literal semantic meaning of the sentence (and indeed its only contribution at all). Such theories are defended by Tarski (1956), Washington (1992), and Cappelen and Lepore (2007). On Tarski’s theory, the quotations are names for the word-types of the quoted material, on Washington’s their meaning is identical to the quoted material itself. Cappelen and Lepore (1997) argue that mixed quotation provides a stark counterexample to views like Tarski’s and Washington’s. Consider the sentence in (1): (1)

Ernestine said she was going to the concert “with that guy in the band.”

If Tarski is right, for instance, and the quotation “’with that guy in the band”’ is a name for the expression whose first constituent is ‘with’, followed by ‘that’, followed by ‘guy’, etc. then (1) should be uninterpretable. One way of seeing the point is by using a different name for the same expression. Let ‘Bill’ name the expression whose first constituent is ‘with’, followed by ‘that’, followed by ‘guy’, etc. Then (1) is equivalent in meaning, on Tarski’s theory, with (2): (2)

#Ernestine said she was going to the concert Bill.

The Punctuation Theory of Quotation

211

But (2) has no meaning – one can’t go to a concert Bill. However, the initial example (1) clearly did have a meaning. Hence Tarski’s account must be wrong. A similar objection applies to Washington’s account, as one can’t go to a concert the words surrounded by quote marks in (1). Cappelen and Lepore (2007) propose a more sophisticated pure-mention theory. The details aren’t essential here, but the basic motivation is to avoid the obvious flaws of Tarski-and-Washington-style views. The claim is that ‘say’ and other verba dicendi have three arguments: the agent doing the saying, the semantic content of the non-quoted material, and the expression-type of the quoted material. A view in the style of Cappelen and Lepore has it that a speaker S bears the SAY relation to a content C (possibly an empty content) and expression type E (possibly an empty expression type) if and only if S expressed a proposition part of whose content was C and the rest of whose content is expressed by E, as used by the speaker. The insight behind the sophistication is this: in the original problem case (1) for Tarski-style views, the expression type named by the quotation couldn’t “directly combine” with the meaning of the unquoted part of the mixed quotation. But on a Cappelen-and-Lepore-style view, the combination is indirect, via the lexical entry for ‘say’. The meaning of ‘Ernestine said C, E’ is true of a content C and expression type E iff Ernestine said C-combined-with-the-meaning-of-E.3 Although the Cappelen and Lepore view is ingenious, in its own right, I think it nevertheless succumbs to a second kind of flaw. In most compositional treatments of quantifiers, something like the following is true: if a quantifier-expression syntactically binds a pronoun or a trace, then the (semantic) values that the pronoun or trace range over are the selfsame semantic values that are in the domain of the meaning of the quantifier expression. In fewer words: if an expression X syntactically binds an expression Y, then the meaning of X semantically binds the meaning of Y. Now consider the examples in (3ab), where syntactic binding relationships have been marked by italics: (3)

a. b.

George explained that the regrettable collapse of the levees was such that “no one knows who caused it” Which houses did the FBI say they could “search t without warrant”?

3. Due to space constraints, this is an enormous oversimplification of the Cappelen and Lepore view (it can’t handle the quoting of non-constituents, or multiple mixed quotes, for example), but I don’t think it difficult to extend my criticism here to the actual account proffered.

212

Michael Johnson

Suppose now that the expression “no one knows who caused it” in (3a) names an expression-type. As such, it has no internal semantic structure, and in particular, the pronoun ‘it’ cannot be semantically bound by the quantifier phrase ‘the regrettable collapse of the levees.’ A parallel observation holds for (3b) However, the constructions in (3ab) involving, as they do, ‘such that’ and whmovement, respectively, require that ‘it’ and ‘t’, respectively, be syntactically bound by the italicized quantifier phrases (one can’t, for instance, say that Sam was such that Mary was happy). Thus Cappelen and Lepore’s (2007) view runs afoul of well and independently motivated compositional principles. For completeness’ sake, it should be pointed out that sometimes quoted quantifier phrases syntactically bind pronouns and yet the pronouns do not refer to the expression type of the quoted phrase, but rather range over its domain. Consider (4a–c): (4)

a. b. c.

John said that “the queen of each man’s heart” loves only herself. George explained that “the regrettable collapse of the levees” is such that no-one knows what caused it. Phil wanted to know “which bloomin’ idiot” he had to vote for t in the next election.

Again, we have three cases, each where syntactic binding must be occurring (else ‘herself’ couldn’t occur in (4a), the ‘such that’ relative couldn’t occur in (4b) and the trace couldn’t occur in (4c)), yet the values ‘herself’, ‘it’, and ‘t’ in the three examples clearly don’t co-refer with the expression types “the queen of each man’s heart”, “the regrettable collapse of the levees” and “which bloomin’ idiot”, but rather range over the domains of the semantic values of those expressions. This indicates, via the principle of syntactic/semantic binding articulated above, that the semantic value of the quoted expressions is not the expression type of the quoted material, but rather its ordinary semantic value. In conclusion, pure-mention theories must be false. Their common feature is that from the point of view of direct semantic composition, the meaning of the quoted material is irrelevant. But examples (1) and (2) demonstrated conclusively that the meaning of the quoted expression was indeed relevant to semantic composition, and examples (3) and (4) demonstrated that it was directly relevant – i.e. not merely relevant via the semantics of some lexical item like ‘say.’ Since, furthermore, the examples in (4) showed it necessary that the ordinary semantic values of the quoted expressions served to semantically bind the relevant pronouns, we reach our first conclusion: Conclusion 2.1: in direct and mixed quotation, the quoted words are not merely mentioned, but rather are at least used, often with their normal semantic values.

The Punctuation Theory of Quotation

213

2.2. Dual use-mention theories A dual use-mention theory is a theory on which the quoted material in a direct/mixed context is used with its normal semantic value, but simultaneously it also designates, denotes, refers to, names, or otherwise mentions words, expressions, shapes, sounds, symbols or the like. Under this heading I include multidimensional accounts, which view sentences containing direct or mixed quotation as contributing two propositions, one linguistic and another metalinguistic (e.g. Potts 2007). But I also include paratactic accounts, where the quoted words are not a part of the speech report itself, and have their normal semantic values, yet due to a demonstrative in the speech report the symbols (or their shapes, sounds, etc.) in the quoted material are or can be mentioned (see e.g. Davidson 1984a, Cappelen and Lepore 1997). In order to show that dual use-mention theories are false, I will argue that in direct and mixed quotation, the quoted material is not mentioned. The argument has to do with the licensing of various types of anaphora. Anaphors are devices that we use to “refer back” to objects, events, and propositions that have previously been introduced into the discourse. I assume that referring to an object, event, or proposition is (with the exception of well-known island constraints) sufficient to make that object, event, or proposition available for later anaphoric reference in the discourse (ceteris paribus, of course). Thus any theory that has it that the quoted words in direct and mixed quotation are referred to will make an empirical prediction, specifically that subsequent to an utterance of a direct or mixed quotation, the direct or mixed quoted words will be available for anaphoric reference later in the discourse. Consider first the case of pure quotation (an account of which is outside the scope of the present paper). It is beyond doubt that in pure quotation, the quoted words are mentioned. This is because one can attribute to ‘Alice’, for instance, the property of having 5 letters, the property of being a name, the property of being pronounced in a certain way, etc. all of which are properties of words. Thus we should expect, and indeed we find, that in pure quotation, the quoted material is available for later reference. For example: (5)

a. b. c.

‘Life is difficult to understand’ is a sentence. It contains an infinitival verb phrase. ‘In quotation, words are used autonomously’ makes very little sense. It could use some editing. Max said the word ‘ain’t’, which is on his teacher’s list of no-no words.

214

Michael Johnson

In both (5a–c), the italicized pronouns anaphorically refer to the italicized quoted material. (5a–c) thus represent perfectly coherent discourses. The case of direct and mixed quotation, however, provides a stark contrast. While dual use-mention theories would have it that the quoted words are mentioned, these words are nevertheless not available for later anaphoric reference. Consider (6a–c): (6)

a. b. c.

Alice said life “is difficult to understand.” #It contains an infinitival verb phrase. The Davidson text tells me that “in quotation, words are used autonomously.” #It could use some editing. #John asserted that we might be going, but he “ain’t”, which is on his teacher’s list of no-no words.

This is strong evidence that the quoted words in (6a-c) are not mentioned. It has been pointed out to me by many in conversation that perhaps in (6a-c) the ceteris paribus clause has been sprung, as it were, in: ceteris paribus, referring to an object, event, or proposition is sufficient to make that object, event, or proposition available for later anaphoric reference in the discourse. The argument goes that for instance, in (6a) there is a sharp subject matter change (from Alice and her ruminations to grammar) that “jars” the listener into finding the discourse incoherent. But I find this reasoning completely unpersuasive. According to the dual-use mention theorists, one of the propositions expressed by the first sentence in (6a) is something like: Alice uttered the phrase ‘is difficult to understand.’ Thus it is part of the dual use-mention account that there is no sudden change in topic: the first sentence of (6a) is just as much about words as it is about Alice’s ruminations. And yet while (7) is coherent, (6a) is not: (7)

Alice uttered the phrase ‘is difficult to understand.’ It contains an infinitival verb phrase.

I have also heard philosophers express concerns about the use of anaphora in these tests. Discourse anaphora is a tricky and not well-understood phenomenon, they say, so we ought not to draw conclusions so quickly from discourse judgments involving anaphora. But much the same point can be made with more run-of-the-mill philosophical devices. Take for instance quantification. If in ‘John told me x’ the variable x can take on the same values as x in ‘x is a sentence’ (namely, word types or tokens), then we’d expect to be able to say: (8)

#Everything John told me was a sentence.

The Punctuation Theory of Quotation

215

But one can’t. The reason is that one doesn’t tell another a sentence; one doesn’t assert sentences; one doesn’t learn sentences; one doesn’t ask sentences. But the verbs ‘tell’, ‘assert’, ‘learn’ and ‘ask’ all take direct and mixed quote complements. This is not evidence that really, after all, one does assert sentences. It is instead evidence that not all quotation involves reference to word types or tokens. Only this hypothesis explains why (8) is ungrammatical; only this hypothesis explains why (6a-c) are incoherent discourses4 . And so we reach our second conclusion: Conclusion 2.2a: In direct and mixed quotation, the quoted words are not mentioned. In conjunction with Conclusion 2.1, we can thus derive: Conclusion 2.2b: In direct and mixed quotation, the quoted words are not mentioned, but rather are used, often with their normal semantic values. 2.3. Pure-use semantic theories There are other semantic accounts of quotation that do not hold that in direct/ mixed quotation the quoted words are mentioned. The most natural such account, in light of my arguments in Section 1, takes the quote-marks themselves to express a monstrous operator that shifts the context of utterance (see Kaplan (1989) and Schlenker (2003) concerning the possibility of monsters in natural language). For example, suppose the indirect-discourse verb ‘say’ expresses a relation between a speaker S, a context C, and a proposition P: the relation that holds when S said in C that P. Taking C* to be the actual context of utterance, the “logical form” of (9) a. is thus (9) b., where ‘speaker-of(C*)’ is taken to be directly referential: (9)

a. b.

John said that I am happy. ∃C said(John, C, that speaker-of(C*) is happy)

In indirect discourse, the introduction of the new context variable does nothing. The story is different, however, for quotation marks. The function of quotation marks is essentially to replace occurrences of C* in P with C. The logical form of (10a) is thus (10b):

4. My use of anaphora and quantifier-binding in these arguments is inspired by Asher (1993), who uses them for similar effect (though on an unrelated topic). See particularly p33ff.

216 (10)

Michael Johnson

a. b.

John said “I am happy” ∃C said(John, C, that speaker-of(C) is happy)

(10b) is true iff there is a context C such that in C John said that he, John, was happy. This account gets the intuitively correct truth conditions and furthermore is compatible with all the arguments I considered in Section 1. (Much refined) versions of this account are defended in Bittner (2007) and Cumming (2005). Nevertheless, I think it is unworkable. First, the monster-account has no resources to account for the original data that motivated so many word-mention views of quotation. It should be completely impossible on the monster-account to quote nonsense in direct or mixed quotation. Yet this is possible, as we see in (11) (from Cappelen and Lepore 1997: 445): (11)

Nicola said that Alice is a “philtosopher.”

On the monster-account, the quotation marks in (11) should map the character of ‘philtosopher’ onto the intension of ‘philtosopher’ at the context of Nicola’s saying; but ‘philtosopher’ has no character, being nonsense. So (11) should be nonsense. But it seems to convey a proposition, and is a perfectly acceptable utterance. Additionally, the monster account cannot explain the inferences from (12a) and (13a) to (12b) and (13b), respectively. For the semantic value of (12a) is just the semantic value of ‘Alice said that life was difficult to understand’ on the reading where the embedded verb is simultaneous with the matrix verb. But this sentence clearly doesn’t entail (12b). (12) (13)

a. b. a. b.

Alice said “Life is difficult to understand”. A token of “Life is difficult to understand” was uttered. Alice said that life “is difficult to understand”. A token of “is difficult to understand” was uttered. [Cappelen and Lepore (1997)]

I’m inclined to think that something like the context-shifting operator posited by the monster account is responsible for the indexical shifting observed in cases like (10a). My doubt, however, is that this has anything essential to do with the presence of quotation marks in these examples. The quotation marks are doing something else – they are being used to pick out words, they are just not doing so via the compositional semantics of the languages that contain them. This view will be elaborated on and defended in the rest of the paper.

The Punctuation Theory of Quotation

3.

217

Toward a pragmatic theory of quotation

A pragmatic theory of direct/ mixed quotation, as I’m using the term, will be a theory that says (a) the presence or absence of quotation marks in a direct or mixed quote is irrelevant to the literal semantic meaning of the sentence and (b) the contribution of quotation marks in a direct or mixed quote takes place at the level of what-is-conveyed by the sentence. For instance, a pragmatic theory might maintain that direct and mixed quotations engender the presupposition that the quoted material was uttered by someone, or it might maintain that direct and mixed quotations contain a conventional implicature to the effect that the quoted material was uttered by someone. Ultimately, I will argue that neither of these initially tempting views can handle the data, but first things first: indexicality. 3.1. Quotation and indexicals I have almost nothing to say on the topic of indexical shifting (the phenomenon exhibited in (9a) and (10a) above) in quotation. It is quite clearly a problem for a pragmatic account, since shifting the referents of the indexicals in a sentence changes the meaning of the sentence, and a pragmatic account denies that using quote marks in direct or mixed quotations changes the meaning of the sentence. If the indexical-shifting phenomenon is to be accepted, then, it must simply be denied by the pragmatic theorist that quotation is the culprit. Perhaps, she can say, it is a monstrous operator that does the work – an operator that is not essentially tied to the quotation marks themselves. This is indeed a difficult story to swallow. For why does indexical shifting always co-occur with quotation, if they are essentially unrelated? And here is the kernel of my defense: the two do not always co-occur. Suppose that I am writing a paper. My friends review the first draft and send me comments. Later, one of them runs across me in the department lounge, and I seem rather harried. She asks whether the others had thought poorly of the paper. I reply with (14): (14)

No, quite the contrary. To quote Oscar Wilde, “Whenever people agree with me, I always feel I must be wrong.”

Here I am not saying that whenever people agree with Oscar Wilde, he always feels he must be wrong, though what I say entails that. Rather, I am saying that whenever people agree with me, I always feel I must be wrong. I am just using Oscar Wilde’s words to do it. This is why I can’t continue (14) with (15):

218 (15)

Michael Johnson

…??But whenever people agree with me, I feel I must be spectacularly right, and that sends me into a frenzy.

In addition to constructions involving ‘to quote X’, constructions involving ‘in the words of X’ and ‘like/ as X said’ exhibit the same pattern of shiftless indexicals, ‘I’ again in (16a) and ‘you’ in (16b): (16)

a.

b.

In the words of Gandhi, “if I had no sense of humor, I would long ago have committed suicide.” ??But of course, I have no sense of humor, and I would never commit suicide. As Abraham Lincoln said to the people of his time, “always bear in mind that your own resolution to succeed is more important than any other.” ??But your own resolution to succeed is not important at all.

What are we to make of these examples? If a pragmatic theory is right, their analysis is straightforward: the use of the quotation marks, say in (16a), entails that Gandhi uttered the quoted words; but the words themselves are used with their normal semantic values. If instead a semantic view were correct, it would be faced with what we may call The New Problem of Indexicality: if quotation marks have the semantic effect of shifting indexicals, why do they not do so in (14) and (16ab)? The New Problem is much more serious than the old problem (the problem for pragmatic theorists of why indexicals ever shift), for it is open to the pragmatic theorist to say that there is an indexical-shifting operator, essentially unrelated to quotation marks, that does the job; whereas it is patently absurd for the semantic theorist to say that quote-marks do the shifting, but in (14) and (16ab) there is an additional “unshifting” operator that undoes the effects of the quote marks. This is absurd because it fails to explain why anyone would go through the trouble of at once doing and un-doing the same thing. A final observation is worth making. Just as quote-marks can occur without the shifting of indexicals, so too can indexical shifting occur without quotemarks. This is the well-known, but as yet ill-understood phenomenon of free indirect discourse. Consider ‘now’ as it occurs in the second sentence of (17): (17)

Bruce turned around the corner. Now things were about to get real hot, real fast.

‘Now’ in (17) does not refer to the speaker’s utterance-time, but rather to the ‘now’ of Bruce’s perspective upon turning the corner. But the perspective-shift is not due to quotation.

The Punctuation Theory of Quotation

219

The argument then is this: free indirect discourse shows us that we need some device other than quotation to shift indexicals. Constructions involving ‘to quote X’, ‘in the words of X’, and ‘like/as X said’ show us that quotation itself need not shift indexicals. So it is no longer clearly the best explanation of why there are shifted indexicals under quotation marks to attribute the shifting role to quotation itself. Indeed, considered in tandem with the argument of Section 1, that there could be no tenable semantic account of quotation, it looks very plausible that quotation is not the culprit. And this opens the door to a pragmatic theory of direct and mixed quotation. 3.2. Presupposition and conventional implicature If the problem of indexicals can indeed be circumvented, it is reasonable to ask what the form of a pragmatic theory of quotation should look like. I want to start by considering, and rejecting, two possible accounts that seem to me to be gaining ground: on the one hand, taking the effects of direct and mixed quotation to be presuppositions, and on the other taking them to be conventional implicatures (see, for instance, Garcia-Carpintero (this volume) for a view along these lines). Conventional implicatures, first discussed by Grice and considerably revived by Potts (2005), are non-cancelable implicatures. They are not “conversational implicatures,” because they arise not from speakers’ reasoning with the conversational maxims, but instead arise from the conventionalized aspects of the lexical items or constructions that engender them. Potts’s main examples come from what he calls supplements (to include appositives and parentheticals) and expressives (such as epithets). Examples (18a) and (19a) illustrate an appositive and an epithet, respectively; (18b) and (19b) indicate their conventional implicatures; while (18c) and (19c) illustrate their non-cancelability: (18)

a.

⇒

b. c. a. b.

(19) ⇒

c.

John, who is Bill’s best friend, ran a 100-meter dash in under 10 seconds. John is Bill’s best friend. …#although, I should tell you, he isn’t Bill’s best friend. John won’t stop making that damn racket. The speaker takes a negative attitude toward the noise John’s making. …#although I enjoy the noise he’s making very much.

Several features of direct and mixed quotation make a treatment of those phenomena as issuing in conventional implicatures (CIs) seem plausible. First, as we have argued, quotation marks do not contribute to the literal semantic mean-

220

Michael Johnson

ing of a sentence. This is also a feature of CIs: Potts, who uses the term ‘at-issue meaning’ for what I’ve been calling ‘literal semantic meaning’, argues convincingly that “no lexical item contributes both an at-issue and a CI meaning” (p. 7). Similarly, just as CIs are non-cancelable, so too are the entailments of quotation marks: Alice said that life “is difficult to understand.” #But no-one ever uttered the words “life is difficult to understand.” But, I think, the analogies between CIs and quotation-entailments run out fairly quickly. Perhaps the most interesting aspect of CIs is that they “always project to the highest possible point” (Potts 2005: 12). That is, they never take scope under other operators and free variables within them are never bound by higher quantifiers. For instance (20a), which contains the modal operator ‘should’, entails (20b) but not (20c), and (21a), which contains the quantifier ‘no-one’, is incoherent, rather than entailing (21b): (20) ⇒ ⇒ (21)

a. b. c. a. b.

John should have eaten the sandwich, which was poisoned. The sandwich was poisoned. The sandwich should have been poisoned. #No-one, who was the king of France, met Bill. No-one was the king of France.5

The case of quotation stands in stark contrast. (22a), rather than entailing (22b), entails instead (22c); similarly, (23a) is coherent, does not entail (23b), but rather entails (23c): (22) ⇒ ⇒ (23) ⇒ ⇒

a. b. c. a. b. c.

Alice should have said “life is difficult to understand.” Someone uttered “life is difficult to understand.” Someone should have uttered “life is difficult to understand.” No-one said life “is difficult to understand.” Someone uttered “is difficult to understand.” No-one uttered “is difficult to understand.”

I think this makes short work for any CI story of quotation entailments. But several participants at the conference (ICQM) showed me that the reasoning to this point was too quick. Consider the discourses in which one might utter (23a). The first that come to mind are discourses where some speaker has affirmed that someone uttered “is difficult to understand,” and the speaker of (23a) is denying this, as in (24ab): 5. It should be noted that there is a reading of the relative clause in (21a) that is not an appositive, and thus issues in no CI, which results in a coherent interpretation for (21a). That reading also does not entail (21b).

The Punctuation Theory of Quotation

(24)

a. b.

221

A: Alice said life “is difficult to understand.” B: No-one said life “is difficult to understand.” Alice said life “is un-understandable” and Bill said it’s “a snap.”

If this were true for every context of utterance of (23a), it would follow that (23a) did in fact entail (23b), my assertion to the contrary notwithstanding. But the facts go the other way. Often when someone wishes to illustrate our ability to understand novel utterances, he or she produces one. For example: (25)

No-one has ever said “irate sloths dance in the miniscule jungles of Tucson,” but you nevertheless understand it.

Surely, if sentences like (25) entailed that the utterance in question was not novel, we’d be at a loss to make sense of how this could be an illustration of our ability to understand novel utterances. Similarly, the most likely contexts where someone would utter (22a) are those where neither Alice, nor anyone relevant in the context, has in fact said “life is difficult to understand.” Of course, the fact that (23a) does entail (23c), whatever else it may entail, is itself sufficient to falsify the CI account of quotation-entailments. But how, one might ask, could (23a) entail (23c), if (23a), which is embedded in (24b), can be uttered after (24a), which contains an utterance of “is difficult to understand”? The answer is: quantifier domain restriction. B in (24b) clearly means not to include A’s utterance in his domain of discourse. On much the same evidence, we can reject the idea that direct or mixed quotation of words presupposes that someone has used those words. Consider two standard tests for presupposition. First, presuppositions “project” out of negation, so that if Q presupposes that P, not-Q also presupposes that P. For example both (26a) and (26b) presuppose that John has a son: (26)

a. b.

John’s son is into sports. John’s son is not into sports.

Second, presuppositions project out of the antecedents of conditionals. That is, if Q presupposes P, then if Q, then R also presupposes that P. For instance, both (27a) and (27) presuppose that Josh was training for the triathlon: (27)

a. b.

Josh stopped training for the triathlon. If Josh stopped training for the triathlon, then he’ll be free to come to the reading group.

We see, again, a stark contrast with quotation. For while (28a) entails (28b), (28c) does not; and whereas (29a) entails (29b), (29c) does not:

222 (28)

(29)

Michael Johnson

a. b. c. a. b. c.

George told me “the levees are fine.” Someone uttered “the levees are fine.” George didn’t tell me “the levees are fine.” Josh said “I can’t run anymore.” Someone uttered “I can’t run anymore.” If Josh said “I can’t run anymore,” you should believe him.

In conclusion, the entailments that arise from direct and mixed quotation have been seen to be (a) not semantic entailments, (b) not conventional implicatures, and (c) not presuppositions. It’s worth adding that they’re not conversational implicatures either. For conversational implicatures are by their very nature cancelable, whereas the entailments that arise from the use of quotation marks are not. That is, (30a) is an incoherent discourse, because it denies the quotationentailment, although (30b) is not – Alice might have told me snow was white by implying it, or by pointing: (30)

a. b.

Alice told me “snow is white.” #But she didn’t utter “snow is white” or any of its translations into foreign languages. Alice told me snow was white. But she didn’t utter “snow is white” or any of its translations into foreign languages.

Thus direct and mixed quotation are strange linguistic specimens indeed. A proper account of their behavior must posit a new kind of entailment, one unknown to most linguistic theories. So be it! 3.3. Calculating quotation entailments Section 1.2 contains a single reference to Potts (2007). There, I include him in the camp of “dual use-mention theorists” and dismiss him, along with the rest, on the basis of data concerning anaphora and quantifiers. And though I think he insufficiently addresses such questions, my dismissal was far too quick. It’s time to take another look. Potts has it that simple sentences containing direct and mixed quotations contribute a pair of propositions, one for each “dimension” of the sentence’s meaning: the assertive dimension and the utterance dimension (the proposition conveyed by the presence of the quote marks, what we are calling the quotation entailment). We can, for the purposes of this paper, ignore whether Potts ultimately believes or is committed to both dimensions being “on a par” – that is, being both part of the literal semantic content. Instead, I want to consider Potts’s treatment of the second dimension as a (partial) theory of how to calculate quotation entailments. This will in turn lead us to a fuller theory, and

The Punctuation Theory of Quotation

223

in the conclusion I shall make some remarks as to the nature of these entailments. Potts (2007: 417) observes that (31a) does not entail (31b), but rather (31c) and also that (32a) does not entail (32b) but rather (32c): (31) ⇒ ⇒ (32) ⇒ ⇒

a. b. c. a. b. c.

Ernie believes that Alice said “life is difficult to understand.” Alice uttered “life is difficult to understand.” Ernie believes Alice uttered “life is difficult to understand.” Everyone said “life is difficult to understand.” Someone said “life is difficult to understand.” Everyone said “life is difficult to understand.”

The observation in (32) is analogous to that in (23) above. But this, combined with (22) and (31) suggests the following principle: (Q1)

a.

(Q1)

b.

Any clause ‘x Vs that p’ where V is a verbum dicendi and p contains the quoted phrase q and does not contain any verba dicendi itself, has as its local quotation entailment that x uttered q Any sentence ‘OP p’ where OP is a sentence operator and p has quotation entailment Q has as its local quotation entailment OP Q.

This principle is borne out, not only by the examples thus far considered, but also by (33): (33) ⇒ ⇒

a. b. c.

Either it’s not raining or George said “the levees are fine.” George uttered “the levees are fine.” Either it’s raining or George uttered “the levees are fine.”

However, this theory of quotation entailments faces two challenges. The first is advanced by Potts. He argues “negation can target one dimension of meaning, but not both of them” (Potts 2007: 425). That is, sentences of the form ‘NOT p,’ where p has a quotation entailment q, can have as their quotation entailment NOT q iff the literal semantic content of the sentence is p. This is in conflict with (Q1), because (Q1) calculates the quotation entailment of a sentence independently of the literal semantic content of the sentence. Potts’ reasons for holding this view, however, are a little baffling. They involve the idea, which is common, that metalinguistic negation has to do with quotation. Consider the sentence (34): (34)

Charlie didn’t call the POlice, he called the poLICE6 .

6. Potts (2007: 424) quoted from Horn (1989: 371).

224

Michael Johnson

If the antecedent of (34) negated the literal semantic content of ‘Charlie called the POlice’ (namely, that Charlie called the police), then (34) as a whole would be an unassertable contradiction. Instead, Potts proposes, the negation in the antecedent targets the “assertion” dimension only: the speaker of (34) isn’t prone to utter “POlice.” But, of course, the speaker of (34) could be indifferent to the two pronunciations and as such be attempting to utter a contradiction. The point is that there is no reading of (34) where the negation operator applies both to the proposition that the speaker is prone to utter “POlice” and to the proposition that Charlie called the police. To me, this is dangerous reasoning. It is dangerous because it proceeds in the absence of clear data about quotation itself. Consider (35): (35)

Jimi Hendrix did not say “excuse me while I kiss this guy.”

It is not clear at all that (35) doesn’t both entail that Jimi Hendrix did not say to excuse him while he kissed some guy and did not utter “excuse me while I kiss this guy.” Indeed, on Potts’s account, if (35) literally semantically asserts that Jimi Hendrix did not say to excuse him while he kissed some guy, then Hendrix did utter the words “excuse me while I kiss this guy.” That is at least counterintuitive. But the point can be made much more sharply. Consider an operator that does modify both dimensions, say ‘should.’ On Potts’s account, the entailment of (36a) is either the two propositions in (36b) or those in (36c): (36)

a. b. c.

You should not say aloud that “this play is boring.” You should say aloud that this play is boring and you should not utter aloud the words “this play is boring.” You should not say aloud that this play is boring and you should utter aloud the words “this play is boring.”

But this is absurd! No-one who utters (36a) means to convey either (36b) or (36c). Thus, (Q1) b. remains intact: quotation entailments embed under propositional attitude verbs (31), quantifiers (23, 32), modals (22), sentential connectives (29, 33), and negation (28, 35, 36). Nevertheless, I think (Q1) a.–b. are false. Consider the scenario in (37ab) [taken from Potts 2007: 420]: (37)

a. b.

Ellen: The Godfather II is a total snooze. Frank: Well, Pauline Kael said that this “total snooze” is a defining movement in American cinema.

The Punctuation Theory of Quotation

225

(Q1) a. predicts that (37b) entails Pauline Kael uttered “total snooze,” which it clearly doesn’t. And amending (Q1) a. to (Q2) doesn’t help: (Q2)

Any sentence ‘x Vs that p’ where V is a verbum dicendi and p contains the quoted phrase q and does not contain any verba dicendi itself, has as its quotation entailment that y uttered q, where y is a salient individual in the discourse.

The amended (Q2), in conjunction with (Q1) b., also makes false predictions. For consider (38ab): (38)

a. b.

Ellen: The professor told me that quotation didn’t engender contentional implicatures. Frank [later]: Ellen said the professor told her that quotation didn’t engender “contentional implicatures.”

By (Q2), we can suppose that the clause ‘the professor told her that quotation didn’t engender “contentional implicatures”’ has as its quotation entailment: Ellen uttered “contentional implicatures.” But by (Q1) b., since quotation entailments embed under all higher operators, (38b) has as its quotation entailment: Ellen said Ellen uttered “contentional implicatures.” Yet (38b) clearly doesn’t entail this. This is a problem, yes, but I think it points the way to the solution. Although quotation entailments scope only under some operators, it is predictable where they will scope. Consider (39): (39)

Sue told me that the president should have said he would “fight for the people.”

Example (39) and its entailments are intuitively true in two scenarios: either Sue uttered “fight for the people” in saying what the president should have said, namely, that he would fight for the people, as in (40a); or Sue said that the president should say he would fight for the people and that he should do so by using the words “fight for the people,” as in (40b): (40)

a. b.

Sue: The president should have said he would fight for the people. Sue: The president should have said he would “fight for the people.”

The data then appear to be this. If x in a clause of the form ‘x V’s that p’, where p has some direct or mixed quoted element q (and p does not contain any verba dicendi), is taken to be the utterer of q, then the proposition x uttered q

226

Michael Johnson

scopes under all higher operators the clause is embedded in (e.g. (22, 23, 28, 32, 33, 35, 36)). However, if some higher subject y of a verbum dicendi is instead understood to be the utterer of q, then the proposition y uttered q scopes under only those operators higher than the verbum dicendi y is the subject of (e.g. (38, 39)). In the limiting case, as in (37b), there is no y identified in the utterance and so the proposition z uttered q, for salient z in the discourse, takes scope over all operators in the sentence. We can turn these observations directly into a general theory for calculating quotation entailments. Suppose, for perspicacity’s sake, that each noun-phrase and each quoted string of words is indexed. When the indices on a noun-phrase and a quoted string of words match (are identical), this corresponds to the intuitive case where the referent of the noun phrase uttered the quoted string (in the local context, i.e. under the scope of higher operators). This device allows us to characterize the (global) quotation entailments for any sentence containing a non-pure (i.e. direct or mixed) quotation. In general, sentences containing direct or mixed quotations have the form ‘OP1 OP2 .. OPn p’ where at least one of the operators OPi is a verbum dicendi and p is the smallest clause containing some quote q. We want our definition to capture the following general entailment: (41) ⇒

a. b.

OP1 …OPn Sally17 OP’1 …OP’m …q17 … OP1 …OPn Sally17 uttered q17

That is, all the operators between Sally and the quote she uttered disappear, and all the operators higher than Sally remain, in the global quotation entailment. This fact is brought out strikingly in the following two pairs of entailments: (42)

a.

⇒ (43)

b. a.

⇒

b.

Sally17 said that Bill must not have said that if it was raining, then “we’re all doomed”17 . Sally uttered “we’re all doomed.” Bill must not have said that if it was raining, then Sally17 said “we’re all doomed”17 . Bill must not have said that if it was raining, then Sally uttered “we’re all doomed.”

Our definition will work as follows. A quote will introduce a property x[x uttered q]. Until this property is saturated by the individual represented as having uttered q, it will simply get “passed up” through the higher operators. Once saturated, however, it will accumulate all operators above it. The introduction clause is:

The Punctuation Theory of Quotation

(Q3)

a.

227

If p is the smallest clause which contains some non-pure quotation q (possibly = p) indexed i, it contributes the property xi [xi uttered q] as its local entailment.

Next, we state the conditions under which an unsaturated local entailment gets “passed up”: (Q3)

b.

(Q3)

c.

If ‘xj V’s that p’ is a clause with verbum dicendi V and p has the property xi [xi uttered q] as its local entailment, the local entailment of the whole is the local entailment of p, if i =/ j. If ‘OP p’ is a clause with operator OP (not a verbum dicendi) and p has the property xi [xi uttered q] as its local entailment, the local entailment of the whole is the local entailment of p.

Third, we must state the conditions under which the property gets saturated. These are of two types: either it is saturated by a co-indexed subject of a verbum dicendi in the sentence, or by some salient individual in context. (Q3)

d.

(Q3)

e.

If ‘xj V’s that p’ is a clause with verbum dicendi V and p has the property xi [xi uttered q] as its local entailment, the local entailment of the whole is xj V’s that p if i = j If p is a non-embedded clause with the property xi [xi uttered q] as its local entailment, then the global entailment of the whole is xj V’s that p if xj is salient in the discourse, else undefined.

Fourth, we must state the conditions under which a saturated local entailment scopes under operators: (Q3)

f.

If ‘OP p’ is a clause with operator OP (possibly a verbum dicendi) and p has a non-property local entailment e, the local entailment of the whole is OP e.

Finally, we identify the global entailment of the largest clause containing the quotation with its local entailment: (Q3)

g.

If p is a non-embedded clause with a non-property local entailment e, then its global entailment is e.

Unlike Potts’ scattered remarks on how quotations combine with higher operators, (Q3) a.–g. present a straightforward, general theory of the quotation entailments for each sentence. The use of quotation marks contributes an unsaturated property to the asserted dimension. From the bottom up, so long as the property remains unsaturated, it “projects” up over the higher operators in the

228

Michael Johnson

derivation. Once it is saturated by the argument of a verbum dicendi, the new proposition (saturated property) scopes under all remaining operators (if any). This is consistent with all the data thus far presented, and indeed all the data the present author is aware of. 4.

Conclusion

Thus far, direct and mixed quotation remain fairly mysterious phenomena. We have seen that they do not contribute to the literal semantic content of utterances containing them; and though it is possible to devise a theory that effectively calculates the entailments that direct and mixed quotation give rise to, as in (Q3) a.–g., these entailments are nevertheless mysterious because they are unlike well-researched entailments (presuppositions, conversational implicatures, etc.). In this conclusion, I want to briefly gesture at what I believe to be the nature of quotation. Let me first distinguish expressions from the signs that articulate them7 . I will denote simple expressions with Roman-alphabet signs in uppercase letters, for example: CAT, and complex expressions with simple expressions and brackets, for example: [CATS [LIKE MICE]]. One and the same expression may be articulated by many distinct signs. For instance, a certain sound pattern /kæt/, a certain sequence of letters ‘cat’, and a certain gesture (as in ASL), may all articulate CAT. CAT itself has no sound, though it may be articulated by objects that do – and perhaps, at different times or places, by objects that differ in their acoustic properties (say, before and after the great vowel shift). Signs and the sign-systems they are a part of can have structure not present in the expressions they articulate (e.g. pitch contours); additionally, expressions can have structure that is not preserved in the signs that articulate them (e.g. the hierarchical structure of [CATS [LIKE MICE]] is absent in the string ‘cats like mice’). To serve our purposes in communicating expressions, we may alter our sign systems to articulate more of the structure of those expressions. In writing, we have moved from the Roman-alphabet articulation ‘catslikemice’ of [CATS [LIKE MICE]] to the more transparent ‘cats like mice.’ We have introduced capitalization, to mark sentence boundaries, and commas, to preserve some (though clearly not all) of the structure of the expressions we articulate. These innovations have fallen under the head of ‘punctuation,’ though their articula-

7. My discussion here has been highly influenced by Ernie Lepore.

The Punctuation Theory of Quotation

229

tory role in writing is not substantially different from the conventionalized use of ‘cat’ to articulate CAT. And yet, there is so much more we could convey in writing systems than merely the features of the expressions we articulate. We could, for instance, “stamp” our written sentences with a day of the week: 1 through 7 for Monday through Sunday. So, for instance, I would write ‘Quotation is a mysterious phenomenon: 1.’ The ‘1’ would not articulate any expression whose content I was attempting to convey to you, but it instead would “directly” convey the information that what I wrote I wrote on Monday. Consider another, perhaps more useful device. A written sign with multiple quantifiers can be ambiguous. ‘Every boy likes some girl’ has two readings. We could introduce a convention of indexing quantifier phrases in the order of highest to lowest scope. Thus ‘Every boy2 likes some girl1 ’ would mean that there is some particular girl whom every boy likes. The indices here would not articulate any aspect of the expression [[EVERY BOY][LIKES [SOME GIRL]]], though it might indicate some aspect of the latter’s logical form. This convention would introduce rather intriguing entailments. Though it would in some sense be about the relative scope of quantifiers, it would not introduce a metalinguistic proposition into the discourse that would allow later anaphoric reference to those quantifiers. It would not presuppose that the audience already knew the relative scope of the quantifiers. It would not conversationally imply scoping relations (that is, it would not be cancelable). And what’s more, the entailment would take scope with respect to higher operators. Quite naturally, writing ‘John said every boy2 likes some girl1 ’ would entail that John said there is some particular girl whom every boy likes. I think quotation is just such a device. It does not articulate any part of the literal semantic content of the expressions we communicate. Rather it is part of the sign system that adds new meaning where there was none before. It tells us that certain words were said before. It is not presuppositional, because its point is adding information. It subserves our interest in accurate reporting. It is not conversational implicature, because it is wholly conventional in nature. Its mysteriousness lies only in the fact that our current sign systems are vastly impoverished with respect to the information they could communicate; quotation is our only sign-system device of conveying that does not operate via articulation (for written language – intonation contours are such a device in speech).

230

Michael Johnson

References Asher, Nicholas 1993 Reference to Abstract Objects in Discourse. Boston: Kluwer Academic Publishers. Barker, Chris & Jacobson, Pauline 2007 Direct Compositionality. New York: Oxford University Press. Bittner, Maria 2007

Online update: Temporal, modal, and de se anaphora in polysynthetic discourse. In Direct Compositionality, C. Barker and P. Jacobson (eds.), 363–404. New York: Oxford University Press.

Cappelen, Herman & Lepore, Ernie 1997 Varieties of quotation. Mind 106: 429–450. Cappelen, Herman & Lepore, Ernie 2007 Language Turned on Itself: The Semantics and Pragmatics of Metalinguistic Discourse. New York: Oxford University Press. Cumming, Sam 2005

Two accounts of indexicals in mixed quotation. Belgian Journal of Linguistics 17: 77–88.

Davidson, Donald 1984a Quotation. In Inquiries into Truth and Interpretation, D. Davidson, 79–92. New York: Oxford University Press. Davidson, Donald 1984b Inquiries into Truth and Interpretation. Oxford: Clarendon Press. García-Carpintero, Manuel this vol. Double-duty quotation, conventional implications and what is said. Kaplan, David 1989 Potts, Chris 2007

Demonstratives. In Themes from Kaplan, J. Almog et al. (eds.), 481– 564. New York: Oxford University Press. The dimensions of quotation. In Direct Compositionality, C. Barker and P. Jacobson (eds.), 405–432. New York. Oxford University Press.

Schlenker, Philippe 2003 A plea for monsters. Linguistics and Philosophy 26: 29–120. Tarski, Alfred 1956

The concept of truth in formalized languages. In Logic, Semantics, Metamathematics, A. Tarski (ed.), 152–278. Indianapolis: Hackett.

Washington, Corey 1992 The identity theory of quotation. The Journal of Philosophy 89: 582– 605.

Misrepresenting Misrepresentation Michael Johnson and Ernie Lepore

1.

Introduction

It’s hardly news that speakers often fail to produce verbatim direct reports. Clark and his collaborators (Wade and Clark 1993, W&C; Clark and Gerrig 1993, C&G) attempt to exploit this widespread foible in practice to expose and undermine what they believe is a deep-seated assumption about the semantics of direct quotation, viz., that one is true just in case it is a verbatim reproduction of the original speaker’s words. Accordingly, Clark denies that (1) can be true only if Joe uttered (2). (1) (2)

Joe said ‘There are cats in the garden.’ There are cats in the garden.

In place of so-called wording theories, Clark defends the view that direct quotations are referring expressions that stand in a ‘fundamentally different’ relation to their referents than ordinary referring expressions. They are demonstrations that depict their referents. It is surely correct that acceptable direct quotations often depart from verbatim reports. Does this establish that accurate direct quotation does not require replication? In what follows, we’ll critically examine the case against wording theories and the defense of the demonstration theory. In the end, we will reject the demonstration theory, defend a sort of wording theory, and conclude that Clark and collaborators erred because they conflated reporting with quoting – both of which are obviously involved in direct quotation.

2.

The case against wording theories

W&C begin by distinguishing direct from indirect quotation. On most of what they say we concur, viz., that the latter can be introduced with complementizer ‘that’ while the former cannot; that tenses of verbs embedded inside an indirect quotation are controlled by the tense of its main verb, but not so for verbs

232

Michael Johnson and Ernie Lepore

embedded in direct quotation.1 Accordingly, (3) might be an accurate indirect report of an utterance of (2). (3)

Joe said that there were felines in the garden.

Based on these differences and others, most theorists conclude that what distinguishes the two is that the semantics of indirect quotation requires only ‘the gist or content’ of the original be preserved, while direct quotation requires ‘the wording’ itself be reproduced; a direct quotation is true only if words2 within its quotes were tokened by the original speaker. W&C begin their critique of wording theories by observing ‘a gap between theory and evidence’ when it comes to attitudes about direct quotation. They say ‘folk and linguistic theories appear to assume that direct quotations are verbatim reproduction’ (W&C: 805), but this assumption is ‘questionable’ (W&C: 806) since speakers ‘are rarely in a position to reproduce the original speaker’s words verbatim, and they must know that’ (W&C: 806). And it doesn’t help, W&C argue, to retreat to a ‘weak wording theory’ that has speakers acknowledging ‘verbatim reproduction is unattainable’ (W&C: 806), but settling on a direct quotation being true just in case it approximates the original speaker’s words (where what counts as an acceptable approximation varies with context). Weak wording theories, they claim, are unable to account for many acceptable direct quotations, such as (4)–(8) (W&C: 806–807): (4) (5) (6) (7) (8)

You can’t say, ‘Well Daddy I didn’t hear you’. (non-existent utterances) And I thought ‘well a I’m not going to’ you know ‘produce any sort of functional gaffes’ (unspoken thoughts) Many people have come to me and said ‘Ed, why don’t you run for the Senate?’ (generic utterances no particular person produced) ‘Well met, captain’ he said, quietly, in German (translations) Senatorial hair, flaring over the eras in authoritative gray waves, says ‘put me on Nightline’. (objects)

Quotations of what one can’t say (4), what one thought but didn’t say (5), what many have said (6), what someone said, but in a different language (7), and what senatorial hair says (8), are all of necessity not reproductions or approximations 1. We disagree, however, with their claim that the semantic values of singular terms inside indirect quotation must be determined by the context of the report, whereas those in direct quotations are invariably determined by the context of the reportee. The singular term ‘I’ inside ‘No one said ‘I’m happy’ lacks a referent altogether. 2. And not its pitch or tone (if these are irrelevant to word individuation).

Misrepresenting Misrepresentation

233

of what some particular individual at some past time did say. This, W&C take as sufficient to refute even weak wording theories. The conclusion then is that direct quotation does not require replication – exact or approximate – and so, W&C reject all wording theories; in their place, following C&G, they offer a demonstration theory. 3.

Demonstrations, non-serious actions, depiction and DQ

Key to understanding Clark’s account of direct quotation is the notion of a demonstration. As C&G use ‘demonstration’ it does not include demonstrating an object with a demonstrative pronoun. Rather, it’s used in the ordinary sense of the word, ‘to illustrate’ or ‘to exemplify,’ the sense in which you can demonstrate (illustrate, exemplify) how someone walks, dances or utters a sentence (C&G: 764 fn 2.; cf. also Recanati 2001: 640).3 Demonstrations enable ‘others to experience what it is like to perceive the things depicted’ (C&G: 765). Alice can demonstrate for Bill how to tie one’s shoes by tying her own. This allows Bill to perceive what it is like to tie one’s shoes, which, combined with practice, may suffice for Bill to learn how to tie his shoes. Demonstrations, according to C&G, differ from descriptions inasmuch as only the former are ‘non-serious’ actions, and depict rather than describe (C&G: 766). The distinction between serious and non-serious actions is from Goffman (1974): actions that are serious ‘are said to be really or actually or literally occurring’ (Goffman 1974: 47). An action is non-serious when it is not really or actually or literally occurring. Talking on the telephone is serious; but pretending to talk on the telephone with a banana is not. Demonstrations, according to C&G (p. 766) are non-serious actions. C&G illustrate with Alice’s demonstrating George’s limp: ‘she isn’t,’ they say, “really or actually or literally’ limping.’4 Demonstrations are of what an event, state, process or object looks, sounds or feels like. In effect, ‘by depicting how a thing looks, sounds or feels, they can refer to the thing itself’ (C&G: 766). The heart of the demonstration theory is that direct quotations are demonstrations. Recanati (another demonstration theorist) outlines the view as fol3. There are many aspects of C&G’s account of demonstration that we will not take up here but nothing pertinent to our evaluation of their account of direct quotation will be omitted from our discussion. 4. We note in passing that being a non-serious action can’t be essential to demonstration. A might really tie her shoe to demonstrate how to do so; Alice might demonstrate George’s limp by actually limping; one might demonstrate fairness to one’s children by literally distributing benefits to them equitably.

234

Michael Johnson and Ernie Lepore

lows: ‘In quotation, what we demonstrate is a piece of verbal behavior – a way of speaking. We demonstrate it by producing an instance of that behavior, that is, by speaking in the relevant way’ (Recanati 2001: 661). To quote, on this view, is to depict or portray a way of speaking: one’s audience perceives the depiction as though (in part) they were perceiving the target of the depiction (i.e. what the depiction depicts): ‘quotations are intended to give the audience an experience of what it would be like in certain respects to experience the original event’ (W&C: 808). As with other sorts of depictions, such as representational painting or sculpture, successful interpretation involves direct perceptual experience of the depiction (C&G: 767). The referent of a quotation then – its semantic content, or what it contributes to the truth-conditions of sentences containing it – is the quotation’s target, or what it depicts. Direct quotations, say C&G, selectively depict their targets. To see what this means, consider a teacher, who, after watching you bowl, sets out to demonstrate your behavior (perhaps, to help you improve your game). His actions are intended to depict your roll, but of course only selected aspects of it. It needn’t matter whether you winced or smiled while bowling, or whether he uses an actual ball or whether his depiction is in slow motion, or whether he faces you while speaking. None of these ‘shortcomings’ belie his demonstration or depiction. Likewise, according to the demonstration theory, with direct quotation a reporter is trying to demonstrate selected aspects of a prior speech act. In different contexts different aspects of this speech act may matter more or less, contingent on what the reporter is trying to get across to his audience: e.g. accent, grammatical or phonological infelicities, spelling, font or color of the tokened letters. [W]hen Alice quotes George, she may depict the sentence he uttered. She can also depict his emotional state (excitement, fear, shyness), his accent (Brooklyn, Irish, Scots), his voice (raspy, nasal, whiny), and even the nonlinguistic aspects that accompanied his speech (gestures, frown, head angle) (C&G: 769)

In regards to counter-data (4)–(8) to wording theories, W&C say ‘speakers are perfectly capable of depicting selective aspects of events that have not yet occurred, other people’s thoughts, collective utterances, utterances in other languages, gestures, and much more’ (W&C: 807). So construed, in employing a direct quotation, the reporter is demonstrating/depicting, and not describing/replicating.5

5. For further discussion of the demonstration theory, see Cappelen and Lepore (2007: Ch. 8).

Misrepresenting Misrepresentation

4.

235

Experimental data contra wording theories and in support of the demonstration theory

W&C center their discussion around three experiments intended to test the empirical predictions of wording and demonstration theories. In the first experiment, W&C set out to test the accuracy predictions of wording theories. Since wording theories have it that the literal semantic content of ‘A said Q’ for some agent A and quotation Q, is that Q reproduces verbatim (or at least approximates) A’s original utterance then, given the hypothesis that subjects, all things considered, desire to avoid saying what is literally semantically false, we should expect subjects’ direct quotations to be accurate – or at least more accurate than their indirect quotations. In the experiment, 20 Stanford undergraduate students were asked to watch four videotaped scenes and subsequently to narrate the events in the scenes to a friend. The main finding of the experiment was that there was no statistically significant difference between the verbatim accuracy of direct quotations produced by the subjects as opposed to their indirect quotations. This, again with the assumption that avoiding literal falsity has lexical priority for students narrating a scene, is evidence that wording theories are themselves false. In the second experiment, W&C set out ‘to test the intuition that narrators do not feel compelled to quote verbatim, even when they can’ (p. 814). The relevance of this hypothesis is supposed to stem again from the fact that if one assumes a wording theory, and one assumes that everyone speaks the literal truth whenever they can, no matter their aims, then one should expect that narrators do feel compelled to quote verbatim when they can, this being the literal truth (according to wording theories). W&C found that among 16 students asked to memorize the dialogue in a scene, those 8 asked to narrate it as accurately as possible produced 99% accurate quotations, whereas the other 8 asked to entertain in their narrative (as opposed to be accurate), produced only 62% accurate quotations. We note in passing that this finding is explainable, given a wording theory, under the assumption that students who are asked to entertain may quite reasonably sacrifice the literal truth of what they say, given their instructed aim. On the other hand, if undergraduates never say what is literally false (and always avoid hyperbole, understatement, metaphor, etc.), even when the explicit aim of the conversation is entertainment, then wording theories are in trouble. In the final experiment, 10 students listened to others narrate a dialogue, and were asked to judge the extent to which their interlocutors were accurate and to which they were attempting to be accurate. Among the 10 students, 5 thought their interlocutor was ‘mostly trying to be accurate,’ and the other 5 thought their interlocutor was ‘trying very hard to be accurate,’ but more than half of

236

Michael Johnson and Ernie Lepore

them thought that their interlocutor was ‘not trying to get the exact wording’ (p. 818). What this seems to show is that, among some of the 10 students, trying to be accurate is compatible with not trying to get the exact wording. This is problematic, say W&C, because ‘[a]ccording to the wording theories…speakers should at least attempt, and be understood as attempting, to reproduce the wording of the original’ (p. 816). We are admittedly at a loss to see why this is so. A theory on which the literal semantic content of ‘There are a million bugs on my windshield’ is that there are a million bugs on the speaker’s windshield hardly has the implication that speakers should attempt, and should be understood as attempting, to only use that sentence when it is literally true. Similarly, just because a wording theory says that ‘A said Q’ for agent A and quote Q is true when Q reproduces A’s utterance, it doesn’t follow that one ought to say that sentence only if Q reproduces A’s utterance. The logic of ‘is’ and ‘ought’ is a little more complicated than that. We do find it interesting that a handful of undergraduates consider general accuracy to be compatible with inexact wording in direct quotations. But later (Part IV, Section 2.ii), we will argue that on a more straightforward test (deference to correction) wording theories get the right result. Even if W&C’s experiments fall short of demolishing wording theories, they go some way toward establishing as plausible the claim that when speakers directly quote, they selectively depict the original utterance. For instance, in the first experiment, W&C found that 47% of direct quotations (as opposed to only 3% of indirect ones) were marked by intonation shifts, either at the beginning or throughout. This indicates that speakers ‘dramatize’ their direct reports in order to allow their audience to perceptually experience in part what the target utterance was like. 5.

Evaluating the demonstration theory

Our criticisms of the demonstration theory divide into two sorts. First, we evaluate the claim that direct quotations are demonstrations, depictions, and nonserious actions. And then we evaluate the relevance of W&C’s experiments vis-à-vis the evaluation of wording theories.6

6. Cf. Chs. 7 and 8 of Cappellen and Lepore (2007) for other criticisms of the demonstration theory.

Misrepresenting Misrepresentation

237

5.1. Why direct quotations are not demonstrations It is important to distinguish between the uncontentious claim that while directly reporting another, a speaker might also depict his utterance,7 and C&G’s stronger claim that direct quotations are depictions (viz., they are non-serious demonstrations) that depict the semantic content of a direct quotation; and so, concurrent depictions (of gestures, accent, etc.) are part of, and indistinguishable from, the semantic content of a direct quotation. When Alice reproduces George’s Irish accent, her audience experiences in part what it is like to hear George speak, recognizes Alice’s intention to demonstrate George’s accent, and accordingly, correctly interprets the depiction. But this is emphatically not the same thing as saying that the literal content of Alice’s direct quotation is determined in any way by what she happens to be depicting. Just as you may depict someone’s facial expression while referring to him (but not quoting him), you may also depict his facial expression while quoting him. The stronger semantic claim, however, is incompatible with the view that the referent of a direct quotation is not what a speaker depicts (if anything) but rather some quotable item. Much of C&G’s evidence for treating direct quotations as depictions is analogical. Since direct quotations exhibit features that are characteristic of depictions, they conclude they are depictions. Depictions, for instance, are nonserious actions, and are selective in which aspects of what they depict that they depict. Quotations, it is argued, have all of these properties too. 5.1.1. Quotations as non-serious actions C&G (pp. 770–774) argue that direct quotations are non-serious actions. The sense in which they are deemed non-serious is that when Matt utters (9) he is not “really or actually or literally making a request” (C&G: 770), though he is quoting one: (9)

She says, ‘Well, I’d like to buy an ant.’

As already noted, demonstrations are not always (and perhaps not even typically) non-serious, but even if they were, it would still be difficult to take this line of reasoning seriously (no pun intended). Setting aside factive operators, every embedded sentence is ‘non-serious’ in this sense. Each of (10a-c) embeds ‘I’d like to buy an ant,’ yet in none is the speaker requesting to buy an ant.

7. Uncontentious, of course, insofar as one does not have general misgivings about depiction itself.

238 (10)

Michael Johnson and Ernie Lepore

a. b. c.

She says that I’d like to buy an ant. If I win the lottery, I’d like to buy an ant. It might come about that I’d like to buy an ant.

Embedded requests are not ‘patterned after’ unembedded ones, at least in the relevant sense: they do not derive their significance from unembedded uses. Rather, their significance is derived directly by means of a compositional semantics. It’s hard to see why direct quotation should be any different. In fact, it isn’t. If we ignore the irrelevant claim that the embedded sentence in (9) is not used to make a request (irrelevant because it only establishes that ‘says’ is not factive), we might still fruitfully ask whether Matt, in uttering (9), succeeds in discussing ants and events of buying, or whether he is using these words ‘nonseriously.’ Consider (11a-c) as follow-ups to (9): (11)

a. b. c.

She says, ‘Well, I’d like to buy an ant.’ And then she found one [ant] that she liked. She says, ‘Well, I’d like to buy an ant.’ And then she did [buy an ant]. She says, ‘Well, I’d like to buy an ant.’ And she did [want to buy one].

What (11a–c) suggest is that Matt, in uttering (9), talks about the state of desiring to buy an ant; about the activity of ant-buying; and about ants. If this is correct, then there’s some reason to assume that in direct quotation, the quoted words are used seriously. The present authors are divided on the question of whether the anaphora facts in (11a–c) show that the quoted words in a direct quotation are used (in the sense of ‘use’ as opposed to ‘mention’). But we do agree that in some broader sense of ‘use’ (11a-c) do show that the words are used seriously. What we mean can be seen by considering another example: (12) (13)

Which suspects t did the CIA claim they could ‘interrogate t without restraint’? John used to say of every womani he met, ‘someday I’ll marry heri .’

In these contexts, the higher quantifiers (‘which suspects’ and ‘every woman’) require a variable in the lower clause that they bind. It follows then that in direct and mixed quotations, the quoted words are used seriously enough that they can meet the minimal syntactic and semantic demands of binding. Whether they are used in a stronger sense is a question beyond the scope of this paper.

Misrepresenting Misrepresentation

239

The acceptability of (11a–c) may also be exploited in a general argument against the demonstration theory, even when issues of seriousness are put to the side. Since demonstrations don’t involve reference to aspects or elements of events, certain anaphoric reference to such aspects or elements after a demonstration is generally impossible, as is readily seen in (12ab): (12)

a. b.

You should really learn this. [George ostentatiously ties his shoe]. ??You can get one from Sally. Arnold is physically impaired. [George limps around]. ??Just like Arnold did yesterday.

If one weren’t merely demonstrating but were instead describing such events, the infelicitous continuations in (12ab) would be perfectly acceptable, as in (13ab): (13)

a. b.

You should really learn this. Watch me tie my shoe. You can get one from Sally. Arnold is physically impaired. I am limping around, just like Arnold did yesterday.

It would seem to follow that utterances of direct quotations are unlike demonstrations. The internal semantic structure of a direct quotation is much like that of any other embedded clause. C&G claim ‘the internal structure of a quotation is really the structure of what is being depicted, and that can range from the raging of a person to the racket of a machine’ (C&G: 772). But this gets things precisely backwards when we consider the facts surrounding anaphora: the internal structure of the direct quotation is the structure of the objects, events and properties described by the quoted sentence. 5.1.2. Quotations as selective depictions To reiterate, we are not claiming speakers never depict when they directly quote. But we are denying that utterances of direct quotations – certain bits of language spoken or written – are themselves demonstrations and that their literal semantic content – what they contribute to the propositions they semantically express – is what individuals depict when they use direct quotation. Demonstrations, recall, are alleged to be selective depictions (C&G: 775). When Fred demonstrates a Frisbee toss, he may depict the motions involved without depicting the speed required for the toss and without letting the Frisbee go (depicting the toss’s completion). Direct quotation seems similarly selective in that you may quote another in her own dialect or you may opt not to. But

240

Michael Johnson and Ernie Lepore

this provides no reason to infer direct quoting is selective depiction. Consider indirect speech in ordinary belief attribution. If Fred is reporting what George believes, and both Fred and his audience know George is unaware that Hesperus and Phosphorus are identical, Fred may or may not choose to use ‘Hesperus’ and only ‘Hesperus’ in reporting George’s beliefs that he [George] would express using ‘Hesperus.’ Fred may selectively choose which aspects of George’s dispositions he wants to preserve in his attributions. But none of this entails or even renders plausible the idea that standard indirect speech is depictive. Being selective need not involve being depictive.8 Second, an even deeper problem tells against direct quotations being depictions. Suppose Andrew visits the Ministry of Silly Walks. He wants to give his audience a feel for what goes on there. So he demonstrates several of the silly walks he observed on his visit. But he doesn’t want his audience to think madness rules at the Ministry. So Andrew says, ‘but no-one walked like this,’ and then he walks in a very peculiar fashion – much too silly, even for the Ministry. Question: what did Andrew depict? We might answer that he depicted an actual state – the state of no-one walking in a certain way. But that can’t be right: the audience does not experience through Andrew’s demonstration, what it’s like to perceive no-one walking that way; quite the opposite! We might instead answer that he depicted a non-actual event, something that could have happened. But who is the agent of this event? Is it ‘no-one’? And what is the time at which Andrew depicts the event as nonactually happening? Does his audience experience what it would be like to perceive the occurrence of a non-actual event? Can one experience that? Furthermore, which specific non-actual event, among the countless infinities of candidates, does Andrew depict? And why that one? The right answer to our questions, of course, is that Andrew depicted nothing. A tempting position is that Andrew depicted a type of walk and not a particular event instantiating that type. We urge resistance to this temptation, for it is impossible to depict types, as Bishop Berkeley knew long ago. One can paint a man, but one cannot paint the type: man. The man one paints will have a certain height; a certain eye-color; a certain amount of hair. This is not the type: man. One can also selectively depict, and instruct one’s audience to ig8. To be fair, C&G have many more analogies between direct quotation and selective depiction (e.g., in a depiction, the aspects selected for depiction are usually marked as such). We take it, however, that all of these analogies are equally problematic, because they are equally true of non-depictive speech, like indirect discourse (e.g., Fred may well mark his using ‘Hesperus’ in his reports of George to track George’s dispositions, say by focusing it).

Misrepresenting Misrepresentation

241

nore aspects of one’s painting as irrelevant. But this does not thereby elevate the painting to a painting of the type: man. For the instruction would have to be: ignore everything. What this is supposed to show is that not all demonstrations are depictions. This result doesn’t wreak havoc with current theories of demonstration, so far as we’re aware. But it does preclude quotations from having as their semantic value what they depict. This is because it would require that perfectly sensible, truth-evaluable reports of saying, such as ‘No-one said, ‘It’s raining,” would lack a semantic value. What no-one said is not a thing, and so, it can’t be depicted.9 5.2. Building a defense of wording theories 5.2.1. Direct quotation and replication Many of the data points Clark and his collaborators adduce against wording theories (both strong and weak) hit wide of the mark. Recall (4): (4)

You can’t say, ‘Well Daddy I didn’t hear you’.

The ‘problem’ for wording theories is supposed to be that (4) might be true, even though there is no prior utterance that (4) reproduces, verbatim or otherwise. This however is a non sequitur. We take it that a paradigm case of a strong wording theory is one that asserts the equivalence in (14): (14)

For any direct quotation Q, ‘John says Q’ is true if and only if Q is a verbatim reproduction of an utterance made by the referent of ‘John.’

By (14), the truth-conditions of ‘You say, ‘Well Daddy I didn’t hear you” are that the quotation ‘Well Daddy I didn’t hear you’ is a verbatim reproduction of an utterance made by the referent of the first occurrence of ‘you’ in (14). Call these truth-conditions P. Presumably then the truth conditions of a sentence composed from a negation + a weak deontic modal + P are that it’s not the case that P is allowed, i.e. that it’s not the case that the referent of ‘you’ is allowed to reproduce verbatim the quotation ‘Well Daddy I didn’t hear you.’ This clearly does not require that anyone reproduced the quote verbatim, contra Clark et al.

9. In fairness, Recanati’s version of the demonstration theory is explicit in assuming that negative cases (e.g. ‘No-one said ‘It’s raining”) don’t depict anything, and the semantic value of the quote is the quote itself (that is, he assimilates such cases to flat mention). We return to this view in Part IV, Section+2.i.

242

Michael Johnson and Ernie Lepore

A precisely similar point can be made regarding (6): (6)

Many people have come to me and said ‘Ed, why don’t you run for the Senate?’

The ‘problem’, again, is supposed to be that the truth of (6) doesn’t rest on the quotation “Ed, why don’t you run for the Senate?” reproducing verbatim the utterance of any particular individual. This however is not a problem for any wording theory we’re aware of. If the truth-conditions of ‘x said Q’ for some direct quotation Q are that Q is a verbatim reproduction of an utterance by the object assigned to the variable ‘x,’ then the truth-conditions of ‘Many x: x said Q’ are that Q is a verbatim reproduction of what many people (not any particular person) said. The general lesson is that wording theories do not require that every direct quotation, no matter how many operators and quantifiers it is embedded under, reproduces an actual utterance. This would make a hash of directly quoting what someone might have said, or what no-one said. That wording theories lack this requirement falls out of the basic wording theoretic analysis of direct quotation (e.g. (14)) plus a compositional semantics for such words as ‘not,’ ‘can,’ ‘many,’ and ‘no-one.’ In fact, it’s somewhat ironic that Clark et al lay cases like (4) at the doorstep of wording theories, since they do create troubles for the demonstration theory. Since there is no prior utterance that the quotation “Well Daddy I didn’t hear you” depicts, there’s strong reason to doubt that quotation involves depiction, as we saw last section. Other demonstration theorists haven’t solved the problem either. Recanati, considering a case similar to (4) (Recanati 2001: 11), is forced to the unhappy conclusion that direct quotations embedded under negation are cases of flat mention, where the target (depicted aspect) of the quotation is the quotation itself, or properties thereof. This conclusion is unhappy first because it obviously won’t work for mixed quotations (see Part V below) embedded under negation (a case Recanati doesn’t discuss), and second, because such direct quotations share none of the hallmarks of flat mention – for example, they, unlike standard cases of flat mention, introduce discourse referents: (15) (16)

John didn’t say ‘I like apples.’ I know this because he hates them. The sentence ‘I like apples’ contains three words. ???John really hates them.

Clark’s other complaints against wording theories seem equally misdirected. Consider again (8):

Misrepresenting Misrepresentation

(8)

243

Senatorial hair, flaring over the eras in authoritative gray waves, says ‘put me on Nightline’

Wording theories say: it is literally false that senatorial hair says ‘Put me on Nightline’ because senatorial hair can’t speak. Thus, if (8) is false, then wording theories actually make the right prediction. Are W&C saying it’s literally, semantically true that senatorial hair says that? If they are, then their theory is plainly false; if they’re not, then the datum is irrelevant, because it fails to support the demonstration theory over any other. What’s to be said about (5), where direct quotation appears after a verb of thinking, and no prior utterance has been made? (5)

And I thought ‘Well a I’m not going to’ you know ‘produce any sort of functional gaffes.’

Several options present themselves, depending on how one thinks about thinking. If English speakers sometimes think in English, a strong wording theorist may plausibly assert that (5) is true if, and only if, the quotation “well a I’m not going to produce any sort of functional gaffes” is a verbatim reproduction of a prior thought had by the referent of ‘I.’ Alternatively, if speakers don’t think in English, (5) and similar examples may be all taken for false – we, at least, don’t have a clear intuition concerning truth-conditions in such cases, and are content to let theory decide.10 The important point is that C&G and W&C don’t have a better view on offer. For them, quotations are depictions and as such, are supposed to enable ‘others to experience what it is like to perceive the things depicted’ (C&G,: 765). But we doubt that they honestly believe the quote in (5) enables anyone to experience what it’s like to perceive someone else’s thoughts. It takes more than quoting a bat’s thoughts to know what it’s like to be a bat. The upshot of the discussion so far is as follows: we are unimpressed by the ‘problem cases’ W&C and C&G present to motivate rejecting wording theories. None of the cases discussed so far present any problem for strong wording theories, much less for weak ones. Additionally, many of the cases (e.g. (5) and (6)), are problematic for demonstration theories, and thus, even if they motivated a rejection of wording theories, they wouldn’t motivate turning to a demonstration theory. The only legitimate concern for wording theories in their discussion, as we see it, is the existence of acceptable direct quotations that are not reproductions but are rather translations of the speech of others, such as (7): 10. A third option, if agents think in language, but not in natural language, is to assimilate quoting thoughts to what we say about translation, below.

244

Michael Johnson and Ernie Lepore

(7)

‘Well met, captain’ he said, quietly, in German.

(7) is problematic for strong wording theories, because “Well met, captain” does not reproduce verbatim any German utterance; and it is a problem for weak wording theories, because it does not approximate any German utterance. If Clark et al want to limit the terminology of ‘wording theories’ to just these two types, then wording theories must be rejected. Our own view is that (7) is true if, and only if, the referent of ‘he’ said the quotation ‘Well met, captain.’ And we further claim you can say ‘Well met, captain’ in German, and you do so by … well, saying it in German. We should say a little more by way of motivation. Speakers permit a wide variety of divergence from verbatim reports in acceptable direct quotation. Indeed, most transcripts, which are after all just a form of direct quotation, have speakers saying fully grammatical sentences, even though, as a matter of fact, normally what issues from our mouths are incomplete ungrammatical sentences. Our utterances are usually chock-full of gaps, false starts, missing words, mistaken inflections, etc. Yet the practice of fixing up recorded interviews is ubiquitous, at least, in the press. What this shows, we think, is that for someone to have said a quote is not a matter of that quote’s reproducing a verbatim utterance. If Ernie utters ‘The demonstration theory ain’t true,’ Michael can truly report him by uttering ‘Ernie said ‘The demonstration theory is not true.’ This is because Ernie did say that – he said it colloquially. Similarly, when we read in a history of philosophy book that Descartes said, ‘I think, therefore, I am,’ we don’t regard the book as full of falsehoods. Descartes did say that – he said it in Latin. What this report shows (barring some reservations below) is something about the conditions under which a speaker can stand in a saying relation to a quoted sentence. Some acceptable direct quotations reveal to us that a speaker can stand in a saying relation to a quoted sentence (or sentence fragment) even though he never uttered a token of that sentence (or sentence fragment); he might have instead uttered only a translation of it in another language or even an ill-formed version of it in the same language. Our view, then, is like a wording theory except we’ve replaced ‘reproduced verbatim’/ ‘approximated’ in the truth conditions of direct report with ‘said.’ That is, for any quote Q, ‘Mary said Q’ is true if and only if Mary said Q; i.e. if Q is something Mary said; i.e. if what Mary said was Q. We doubt we can provide a fully explicit analysis of ‘say’ any more forthcoming than we could one for ‘good.’ But there’s no reason to suppose saying involves depiction. It’s a lot closer to verbatim reproduction, except for permitting cleaning-up and translation. It’s saying.

Misrepresenting Misrepresentation

245

5.2.2. Deference to correction None of the above means that anything goes. There are limits and there are errors. We permit some degree of fixing up – if we are intending to attribute a direct quotation to a speaker of an English sentence, we more likely than not will eliminate grammatical infelicities, especially if their existence is irrelevant to the purposes at hand. However reporters invariably defer to correction. If your direct quotation ‘Al said, ‘Mauve is my favorite color” draws the reaction ‘I didn’t know Al knew the word ‘mauve”, you might detract and defer, saying instead, ‘Oh, what he really said is ‘That color is my favor color’ pointing at a patch of mauve. In effect, you are conceding your original report is inaccurate. This datum dominates. Speakers invariably defer when confronted with, say, recorded documents. They might resist conceding vis-à-vis corrected infelicities (unless they matter for the discourse at hand) but they will defer (concede error) for non-verbatim reports when the original is presented. This suggests strongly wording theories are correct. It also suggests W&C’s three experiments are irrelevant to the debate over the correct semantics of direct quotation. The fact that (W&C: 811) ‘the percentage of verbatim words for direct quotations is 38% … is extraordinarily low for a theory that assumes speakers are committing themselves to verbatim reproduction’ is more than balanced by a reporter’s willingness to defer when confronted with inaccuracy. No one, we predict, will infer from these data that speakers don’t care which words an original speaker used. 6.

General methodological lessons for semantics

W&C seem to think that speaker behavior determines semantics. That speakers sometimes admit they are bad at verbatim direct quotation, they claim, supports their conclusion that it cannot be part of the semantics of direct quotation that it is verbatim. On the contrary, what is pertinent in determining the semantics of direct quotation is not what people do when they directly quote but rather how they respond to correction. What we find utterly relevant is that speakers invariably defer, e.g., upon re-hearing an original utterance. W&C say, ‘Speakers believe they are trying only to approximate the original speaker’s utterance as best they can.’ If that were right, the correct response to a charge of mistaken direct quotation would be ‘we are only trying to approximate’. We sincerely doubt anyone would offer this excuse for his failure. We end by identifying a serious deficiency in W&C’s discussion, namely, their failure to acknowledge the practice of mixed quotation (a practice we mentioned in passing in IV,2.i). The first systematic discussion of mixed quotation

246

Michael Johnson and Ernie Lepore

appears in Davidson (1984). Before this, the category of mixed quotation was mostly ignored. For the following twenty years, it was still relegated to footnotes. But during the last ten years or so, it has occupied front row center for anyone interested in quotation. Davidson’s famous example is ‘Quine said that quotation ‘has a certain anomalous feature’.’ This report quotes Quine by reporting what he said, but attributes to him only an utterance of the expression ‘has a certain anomalous feature’. This practice is ubiquitous, but W&C neglect the practice entirely. There are many reasons why we might opt to mixed quote another rather than directly or indirectly quote him. We often mixed quote another because (i) the reported utterance is too long to directly quote, but the reporter wants to ensure accuracy on certain key passages, (ii) certain passages in the original utterance were particularly well put (as in, Quine says that quotation ‘has a certain anomalous feature’), (iii) perhaps the words used by the original speaker were (potentially) offensive to an audience and the speaker wants to distance himself from them by indicating that they are the words of the individual being reported and not his own (as in ‘Vice President Dick Cheney on the Senate floor told Senator Patrick Leahy to go ‘fuck’ himself’), and (iv) the expressions being mixed quoted might be ungrammatical or a solecism and the speaker might be trying to indicate that he’s not responsible; or that he recognizes something is funny or odd about them (as in ‘Mary said that John is a wonderful ‘philtosopher’ ’ or ‘Howard said that he ‘ain’t gonna take it no more”). The same sort of point extends to mixed quoting foreign expressions (as in, ‘Mario said that he was ‘en casa’ when the murder occurred’). Since W&C, rather surprisingly, in their various experiments, do not control for mixed quotation, it’s unclear what differences if any these tests could establish between direct and indirect quotation. When they conclude (W&C: 811) that speakers are no more accurate with direct than with indirect quotation they are presuming they have identified legitimate cases of indirect quotation but their data are equally construable as mixed quotation. In fact, their method (see the ‘Method’ section, W&C: 809) was to mark spoken quotations as either direct or indirect, without the option of mixed quotation. This was in accord with how they judged a ‘novelist or copyeditor’ would insert the punctuation; yet we submit that the results would have been substantially different, had they allowed for mixed quotation and had they punctuated as a newspaper fact-checker would have. W&C are testing for something; but not for semantic properties of linguistic forms. That reporters sometimes employ ‘dramatized intonation…to depict annoyance, embarrassment, sarcasm’ shows that they were not engaged solely in the practice of direct quotation.

Misrepresenting Misrepresentation

7.

247

Conclusion

We argued against demonstration theorists that direct quotations are neither demonstrations nor selective depictions nor non-serious actions. We also argued that their requirement that true direct quotations depict prior speech cannot be squared with the fact that direct quotations can report what someone might have said or even what no-one said. We challenged the problematic cases they alleged against wording theories. However, a legitimate worry their discussion raises concerns direct reports of utterances translated into, or cleaned up in, the language of the direct report. These cannot be reproductions. We noted, that the saying relationship between a person and quote permits something less than a verbatim reproduction. We do not take this qualification to be an interesting compromise of wording theory, however, since the practice of invariable deference to correction is still in place. When challenged we push our direct reports further and further towards verbatim reproductions. This dominating fact about our practice of direct quotation renders, we conclude, that all of the experimental data advanced against wording theories is irrelevant. If that were not enough, we ended our discussion by complaining that Clark et al neglected to control for mixed quotation. Without some such control Clark runs the risk of contrasting apples with apples, since both direct and mixed quotation reports utilize bona fide quotation. References Cappelen, Herman & Ernie Lepore, 2007 Language Turned on itself. Oxford: Oxford University Press. Clark, Herbert H. & Richard J. Gerrig 1990 Quotations as demonstrations. Language 66: 764–805. Davidson, Donald 1984 Quotation. In Inquiries into Truth and Interpretation, D. Davidson, 79–92. Oxford: Oxford University Press. (Originally published in Theory and Decision 11 (1979): 27–40.) Goffman, Erving 1974 Frame Analysis. New York: Harper & Row. Malcolm, Janet 1990

The Journalist and The Murderer. New York: Vintage Books.

Recanati, Francois 2001. Open quotation. Mind, 110(436): 637–87.

248

Michael Johnson and Ernie Lepore

Wade, Elizabeth, & Herbert H. Clark 1993 Reproduction and demonstration in quotations. Journal of Memory and Language 32: 805–819.

Two Dogmas on Quotation∗ Jürgen Pafel

1.

Introduction

It seems quite obvious that pure quotations are singular terms. In sentences like ‘Dogma’ is a noun of Greek and Latin origin the subject of the sentence obviously is a singular term denoting a certain word. This use of pure quotations is a quite frequent one. Surprisingly, one should not conclude from this fact that pure quotations syntactically are noun phrases and semantically singular terms denoting linguistic expressions, as is done quite unanimously in the literature on quotation.1 Carefully looking at pure quotations’ syntactic distribution sug* I am especially indebted to Philippe De Brabanter, Elke Brendel, Fabian Dirscherl, Manuel García-Carpintero and Jörg Meibauer for interesting discussions in the context of my presentation with the same title at the ‘First International Conference on Quotation and Meaning’ in Mainz, 19 october 2007, and, last not least, to Markus Steinbach for his very helpful comments on the manuscript. 1. “[T]ake quotation marks: applied to any sort of expression, what they produce is a singular term (naming […] the expression inside)” (Quine 1960: 143). “It is clear that quotations are noun phrases, as they function as grammatical subjects” (Saka 1998: 119). “[Closed] quotation functions as a singular term within the mentioning sentence” (Recanati 2001: 651). “[T]he semantic function of [pure and direct quotation] […] is to be a singular term” (Cappelen and Lepore 2005: 60). “[P]ure quotations are always NPs” (Cappelen and Lepore 2007: 138). “[Pure quotation] can be understood as governed by a simple semantic principle, apparently first identified by Tarski (1933): by enclosing any expression within quotation marks one gets a singular term – a quotation – that stands for (refers to, denotes) the enclosed expression” (Gómez-Torrente 2005: 129). “On the demonstrative theory, neither the quotation as a whole (quotes plus filling) nor the filling alone is, except by accident, a singular term. The singular term is the quotation marks, which may be read ‘the expression a token of which is here”’ (Davidson 1979 [1984: 90]). “Still, we can retain the essentials of Davidson’s theory, while rejecting his treatment of the quotation marks as a singular term. The quotation marks merely indicate that the quoted words are being demonstrated. It is, I suggest, the demonstration itself which assumes the function of a singular term, in closed quotation” (Recanati 2001: 655).

250

Jürgen Pafel

gests something else: pure quotations definitely have the distribution of nouns, not of noun phrases. (Sentences like the one above should be compared to sentences where the subject noun phrase visibly contains a noun only, as is the case, for instance, in sentences with proper names: Quine was an extraordinary philosopher.) This distributional fact has semantic consequences: one cannot cope with the entire distribution of pure quotations if one takes them semantically as singular terms. Instead, their distribution suggests a predicate view on pure quotations, as nouns usually are predicates. Pure quotations, as we will see, are predicates whose meaning is to have a certain shape, with the shape being that which the pure quotation presents in between the quotations marks. This is a kind of self-reference as the meaning of such a noun makes reference to its shape and accounts for the picture-like character of pure quotations. There is a second, quite wide-spread assumption, a second ›dogma‹, which is not as obviously true as one might initially be tempted to think, namely, the assumption that pure and direct quotations are more or less on a par with each other, as far as syntax and semantics is concerned.2 In combination with the first dogma one gets the apparently obvious conclusion that direct quotes are noun phrases and singular terms denoting linguistic expressions.3 Carefully comparing the distribution of pure and direct quotations exhibits, as we will see, significant differences. Direct quotations have the distribution of clauses, not the distribution of noun phrases or nouns. This syntactic difference suggests a semantic difference. As (argument) clauses normally denote propositions, one should have conclusive evidence at hand for direct quotations differing from normal clauses, i.e., for direct quotations denoting linguistic expressions. On the contrary, there is evidence that they do denote propositions, which suggests an analysis of direct quotes as a special kind of definite description of propositions. The self-referential character of direct quotation is, exactly like the pure quotation’s, a consequence of its meaning, which makes reference to the quo-

2. See most of the quotes in footnote 1. 3. “Wenn man in der gewöhnlichen Weise Worte gebraucht, so ist das, wovon man sprechen will, deren Bedeutung. Es kann aber auch vorkommen, daß man von Worten selbst […] reden will. Jenes geschieht z.B., wenn man die Worte eines anderen in gerader Rede anführt. Die eigenen Worte bedeuten dann zunächst die Worte des anderen, und erst diese haben die gewöhnliche Bedeutung. Wir haben dann Zeichen von Zeichen. In der Schrift schließt man in diesem Falle die Wortbilder in Anführungszeichen ein. Es darf also ein in Anführungszeichen stehendes Wortbild nicht in der gewöhnlichen Bedeutung genommen werden” (Frege 1892: 28).

Two Dogmas on Quotation

251

tation’s shape. After looking at several kinds of quotations, the diverse uses of quotations marks present themselves as a case of polysemy.4 2.

The syntax of pure quotation

I begin with presenting some examples illustrating the variety of uses of pure quotations in languages like English, French, Polish and German. I confine myself to English and German examples. In (1) we probably have the most frequent use of pure quotations with the pure quotation being the one and only visible element of a noun phrase.5 (1)

a. b.

‘Dogma’ ist ein Nomen mit griechischem und lateinischem Ursprung. ‘Dogma’ is a noun of Greek and Latin origin.

In these cases the pure quotation behaves like a noun phrase even if it is a sentence which is quoted. See the English examples in (2) and (3). The quotation can fill the subject position of the clause (see (2a)), but the quotation cannot be extraposed (see (2b)). The subject clause in (3b), however, can be extraposed. (2) (3)

a. b. a. b.

‘That two plus two equals four’ is a dependent clause. *It is a dependent clause ‘that two plus two equals four’. That two plus two equals four is always true. It is always true that two plus two equals four.

Similarly, the pure quotation in German can be located sentence-internally (more precisely, in the ›mittelfeld‹ of the German sentence), but not at the right end of the sentence in an extraposed position (see (4a) vs. (4b,c)). A subject, or an argument, clause, however, can hardly be located sentence-internally. Its most natural position is the extraposed one (see (5)).6 4. For descriptive reasons, I use distinct quotation marks for the diverse kinds of quotation: ‘pure quotes’, “direct quotes”, »mixed quotes«, ›scare quotes‹. 5. There are languages where a pure quotation cannot stand alone as the subject of a sentence, where it must be introduced by a determiner. Swabian, a dialect in the south-west of Germany, is such a language. (i) *(des) ‘le’ dud dr Schwob fascht iberall drõhenge. the ‘le’ does the Swabian almost everywhere add ‘The Swabian can add ‘le’ to almost every word.’ 6. It is generally the case that German subject and object clauses prefer the extraposed position at the right end of a clause. Only under very special conditions can they be located sentence-internally in the mittelfeld. Being topical seems to be one decisive

252

Jürgen Pafel

(4)

a. b. c.

(5)

a. b. c.

Doch ist ‘Es regnet’ ein deutscher Satz. however is ‘it rains’ a German sentence *Doch ist ein deutscher Satz ‘Es regnet’. however is a German sentence ‘it rains’ *Da (es) ein deutscher Satz ist ‘Es regnet’, … as (it) a German sentence is ‘it rains’ ? Doch ist, dass es regnet, eine Tatsache.7 however is that it rains a fact Doch ist es eine Tatsache, dass es regnet. however is it a fact that it rains Da es eine Tatsache ist, dass es regnet, … as it a fact is that it rains

Apart from pure quotations being the one and only visible element of a noun phrase, there are uses of pure quotations with additional elements in the noun phrase. In (6) and (7), for instance, the pure quotation is located in the position where the noun is located in a noun phrase: determiner and adjective in front of the quotation, PP-attributes behind. The noun phrase which contains the pure quotation can be a referential (see (6)), a quantificational (see (7)) or a predicational one (see (9)). As seen in (8), pure quotations can be found in places where so-called narrow appositions are located. Pure quotations can even be part of compounds, see (9) and (10). (6)

(7)

Auf Balzacs Spazierstockgriff: Ich breche alle Hindernisse. Auf meinem: Mich brechen alle Hindernisse. Gemeinsam ist das ‘alle’. (Franz Kafka) ‘On the stock of Balzac’s walking-stick: “I surmount all obstacles.” On mine: “All obstacles surmount me.” They have the ‘all’ in common.’ a. In jedem seiner Sätze kommen mindestens zwei ‘natürlich’ vor. ‘In each one of his sentences, he uses ‘of course’ at least twice.’

condition as the following well-formed example of Webelhuth (1992: 103ff.) suggests (cf. Meinunger 2000: 183): (i) Weil ich, dass Hans krank ist, nicht glauben kann. as I that John ill is not believe can ‘As I can’t believe that John is ill.’ I conducted a litte corpus study with the result that, as for one edition of the weekly paper Die Zeit (7 august 2008), there was not one case with a (finite or non-finite) argument clause sentence-internally. All argument clauses were extraposed (the majority) or located at the left edge of the sentence. 7. It is difficult to determine the correct level of (5a)’s unacceptability.

Two Dogmas on Quotation

b. (8)

a.

(9)

b. a. b.

(10)

a.

b. c. d.

253

His speech abounded in many I think so’s. (Cf. Clark and Gerrig 1990: 771f.) Die Präposition ‘vor’ regiert den Dativ. ‘The preposition ‘vor’ governs the dative.’ The preposition ‘at’ selects a noun phrase as a complement. ‘Vor der Tür’ ist eine PP, eine ‘vor’-PP. “Vor der Tür’ is a PP, a ‘vor’ PP.’ ‘At the door’ is a PP, it is an ‘at’ PP. Und jetzt kommt wieder der ‘Keine Macht den Drogen’-Schmarrn. (Cf. Meibauer 2003) ‘And now there is this ‘No power to drugs’ nonsense again.’ The I-told-you-so phase of the Iraq invasion is thankfully ending. (New York Times) These are not ‘I really should’ radishes. (Clark and Gerrig 1990: 771f.) Er verwendete wieder das Herabsetzungs-‘du’. ‘He once again used the downgrading ‘du’.’

A closer look at the distributional syntax of pure quotations thus reveals that pure quotations have the distribution of nouns, but not the distribution of noun phrases. This suggests that pure quotations literally are nouns. Sentence (7b) is quite instructive in this respect: here the pure quotation is pluralized, which is a clear piece of morphological evidence for it being a noun. As for German, it is obvious from the choice of the article (das) that pure quotations are neuter nouns (see (6)). In English, the pronoun one can replace nouns as well as pure quotations (The first ‘all’ and the second one). If, as I will assume, pure quotations are nouns, then the syntactic properties, especially the syntactic category, of the quoted material is of no importance to the syntax of pure quotations (the sentences in (2) to (10) are ample evidence for this). Thus, the view of pure quotations as nouns copes with their entire distribution. The case where pure quotations are the one and only visible element of a noun phrase is a case where the noun phrase visibly only consists of a noun. This is nothing peculiar. Well-known uses of indefinite noun phrases as well as proper names are cases of the same type, a noun phrase visibly only consisting of a noun. Assuming a DP-analysis of noun phrases, I will assume a nonvisible determiner ([D ø]) as the head of the noun phrase whose complement is the pure quotation (an analogous assumption is quite common with respect to proper names and certain uses of indefinite noun phrases).

254 (11)

Jürgen Pafel

a. b. c. d.

[DP [D this] [N noun]] [DP [D ø] [N nouns]] [DP [D ø] [N Quine]] [DP [D ø] [N ‘dogma’]]

The distribution facts we just looked at throw doubt at least on the syntactic side of the assumption that pure quotations are noun phrases and singular terms. But the semantic side of the assumption becomes questionable as well. It seems obvious that in the cases where the pure quotation is visibly a proper part of the noun phrase – i.e., where there is an overt determiner, an adjective or a prepositional phrase accompanying the quotation – it cannot be a singular term. In such cases, the quotations should be what nouns generally are, namely, predicates. Thus, the wide-spread assumption in question cannot cope semantically with these cases, and leads to an asymmetry in the use of pure quotations. A symmetric account, in contrast, would be one which treats pure quotations syntactically as nouns and semantically as predicates throughout. One could call it ‘the predicate view of pure quotations’. If such an account could sensibly be developed it would be superior to the asymmetric account. This is what I will try to do in § 4 on the basis of Pafel (2007). There are striking similarities between pure quotations and proper nouns, syntactically as well as semantically. Both are parts of noun phrases, both can have the status of singular terms, but they can also be combined with determiners, adjectives, post-nominal PPs and can be narrow appositions as well as part of compounds. (12)

a. b. c. d.

Quine died in 2000. the Quine of Word and Object the philosopher Quine a Quine reader

Assuming that proper nouns are singular terms, we have difficulties in coping with their use as a proper part of a noun phrase. Instead, the distribution suggests a predicate view of proper nouns (cf. Burge 1973, Larson and Segal 1995, Katz 2001, Bach 2002, Elugardo 2002, Karnowski and Pafel 2005). We will come back to these similarities between pure quotations and proper nouns at the end of § 4.

Two Dogmas on Quotation

3.

255

The syntax of direct and mixed quotation

I have already alluded to the fact that in German, argument noun phrases and argument clauses have distinct distribution. Whereas argument noun phrases can be located in the mittelfeld and can be extraposed under special circumstances only (cf. (13)), it is just the other way around with respect to argument clauses: they are most naturally extraposed and can be placed sentence-internally under special circumstances only. (13)

a.

(14)

a. b.

Heine hat das gesagt. b. *Heine hat gesagt das. Heine has that said Heine has said that Heine hat gesagt, dass Moritz ihm der liebste sei. ? Heine hat, dass Moritz ihm der liebste sei, gesagt. ‘Heine said that he likes Moritz most.’

These facts provide us with a test concerning the syntactic category of direct quotations. The result is that the direct quotation of a sentence behaves syntactically like a sentence. It can be placed in positions where an argument clause can, thus it can be extraposed and topicalized (see (15), (16), (17)). Parentheticals (out of quote) can be inserted into direct quotations in the same way in which they can be inserted into ordinary sentences (see (18a,b) and (19a,b)). It is questionable whether direct quotations can be used sentence-internally at all (see (20)) – it might be possible, however, under special circumstances. (15) (16) (17)

(18) (19) (20)

a. b. a. b. a.

Heine hat gesagt “Moritz ist mir der liebste”. Heine said “I like Moritz most”. “Moritz ist mir der liebste” soll Heine gesagt haben. “I like Moritz most,” Heine is assumed to have said. Dass Heine gesagt haben soll “Moritz ist mir der liebste”, war mir neu. b. That Heine said “I like Moritz most” was new to me. a. “Moritz”, hat Heine gesagt, “ist mir der liebste.” b. “I,” said Heine, “like Moritz most”. a. “Moritz ist mir”, hat Heine gesagt, “der liebste.” b. “I like,” said Heine, “Moritz most”. ? Heine hat “Moritz ist mir der liebste” gesagt.8 Heine has “I like Moritz most” said

8. Comparable to (5a), it is difficult to determine the correct level of (20)’s unacceptability.

256

Jürgen Pafel

It is not possible to modify a direct quotation in the same way as a pure quotation by adding, for instance, a determiner and an adjective in front of the quotation or PP-attributes behind.9 (21)

a. *Heine hat gesagt das überraschende “Moritz ist mir der liebste”. b. *Heine said the surprising “I like Moritz most”.

Thus, direct quotations differ significantly from pure quotations with respect to syntax. Direct quotations seem to be syntactically transparent insofar as their syntactic category is the same as the category of the quoted material. Direct quotations seem to be sentences, not noun phrases. With totally different, for the most part, semantic considerations, Cappelen and Lepore (2007: 145f.) come to the same conclusion: “It may be that we cannot treat direct quotations as NPs.”10 They stipulate a quotation marker Q which adjoins to syntactic categories of any kind converting them into quotation expressions of the same syntactic category (ibid. 138). In direct quotation it adjoins to an expression of the category S converting it into a quotation expression of the same category. Such a move copes with the syntactic behavior of direct quotations (but see below for the application of this marker to mixed quotations). Are all direct quotations quotations of sentences? It might be so, as there is evidence that even direct quotations of elliptical constructions should be analyzed as quotations of elliptical sentences. Take a question-answer pair like Gehst du normalerweise vor Mitternacht ins Bett? In aller Regel ja ‘Do you go to bed before midnight? Usually I do so’ (a word-to-word translation of in aller Regel ja is ‘in every rule yes’). Reporting the answer we might construct the following sentences: (22)

a. b. c.

Sie hat mir geantwortet: “In aller Regel ja” “In aller Regel ja”, hat sie mir geantwortet. ? Sie hat mir “In aller Regel ja” geantwortet.

9. In languages like Swabian where a pure quotation must be accompanied by a determiner (cf. footnote 5), direct quotations come without determiner and cannot be combined with it. (i) De Bienzle hot gsagd (*des) “Mi geht des nix õ”. ‘Bienzle said “This is of no concern to me”.’ 10. Their argument, however, is unconvincing, as the argument could successfully be applied to pure quotations too and would show that pure quotations (being the one and only visible element of a noun phrase) could not be noun phrases. See below § 5.

Two Dogmas on Quotation

257

The elliptical answer exhibits the distribution of an argument clause. Assuming that it is an argument clause would immediately explain its distribution. See also the possibility of inserting parentheticals: (23)

“In aller Regel”, antwortete sie, “ja”.

But note that the one-word answer Ja ‘Yes’ behaves like a noun phrase in a speech report – we seem to have a pure-quotation use here: (24)

a. b. c.

? Sie

hat geantwortet ‘Ja’. ‘Ja’ hat sie geantwortet. Sie hat ‘Ja’ geantwortet. ‘She answered ‘Yes’.’

Mixed quotations are syntactically transparent too: they smoothly fit into the syntactic context as if there were no quotations. (25) (26)

Quine says that quotation »has a certain anomalous feature«. Er berichtet in seiner Autobiographie darüber, »daß auch in Berlin das gewaltige Ereignis« der Französischen Revolution »große Teilnahme erregte« und die »meisten jungen Männer sich offen zugunsten der Revolution« aussprachen, bis die Ermordung Ludwigs XVI. die Stimmung umschlagen ließ. (De Bruyn 2006: 17)11

Example (26) is especially illuminating in this respect. Delete the quotation marks and you get a well-formed sentence. But the linguistic material that is put in quotation marks does not always seem to be a well-formed constituent: Firstly, daß auch in Berlin das gewaltige Ereignis is not a sensible constituent of the sentence, as the attribute of Ereignis ('event') – namely, der Französischen Revolution ‘of the French revolution’ – is not part of the quotation (only categorial grammarians here might disagree). Secondly, meisten jungen Männer sich offen zugunsten der Revolution cannot be a constituent either, as the article die is not part of the quotation despite the fact that the article (die) and the adjective (meisten) are a lexicalized syntactic unit (comparable to many a in English with the opposite order of article and adjective). Compare Cumming (2005), who stresses the fact that mixed quotations can quote non-constituents.

11. Translation: “In his autobiography, he reported »that even in Berlin, the huge event« of the French revolution »aroused great sympathy« and many »a young men« spoke »frankly in favour of the revolution«, until the killing of Louis XVI radically changed the mood.”

258

Jürgen Pafel

Non-constituency in mixed quotation militates heavily against Cappelen and Lepore’s quotation marker Q, which they introduced in mixed quotations and transfered to direct quotations (see above), as Q adjoins to arbitrary ›constituents‹ of a sentence in their account. Scare quotes are syntactically transparent too. Take an ironic statement like He invited the ›best actress in the world‹, where the speaker conveys the meaning that someone other than himself takes the invited person for the best actress in the world. The scare quote fits smoothly into its syntactic context as if there were no quotation. The same is true with the well-known examples These are indeed ›subliminable‹ ads and I am not a ›philtosopher‹. The misshapen words behave like an ordinary adjective resp. noun, just as the intended words do. The facts looked at in this paragraph throw into doubt the syntactic part of the second dogma, namely, the assumption that pure and direct quotation are more or less on a par with each other, as far as syntax and semantics are concerned. The assumption that direct quotations are noun phrases cannot be upheld. In § 7 we will see reasons to doubt the semantic side of the second dogma. 4.

The predicate view of pure quotations

As pure quotations are nouns, one should assume that they semantically are predicates, as nouns usually are. This has the immediate consequence that all the accounts cannot be on the right track which assign pure quotations the semantics of a singular term.12 This has, in addition, the consequence that referential noun phrases whose one and only visible element is a pure quotation cannot be semantically unstructured, there must be an additional element which leads to the referential interpretation of the noun phrase. But let us look at the meaning of the pure-quotation predicate first. As in ‘Dogma’ is a noun of Greek and Latin origin the subject denotes a certain word, the pure-quotation predicate should denote a concept which this word satisfies. Which aspect of the word will probably be picked up by this concept? It seems obvious that it is the form aspect of the word, its shape, which is picked up. Now it is near at hand to assume that pure quotations are predicates whose meaning is to have a certain shape, with the shape being that, which the pure quotation presents, which is enclosed in quotation marks. Pure quotations are predicates which refer to a certain part of the quotation, namely the graphical

12. One of the latest versions of such an account is Cappelen and Lepore (2007: 124), who propose the schema ‘ ‘e’ ’ quotes ‘e’ as the semantic rule for quotation.

Two Dogmas on Quotation

259

entity enclosed in quotation marks. The pure quotation of the noun dogma, for instance, has the meaning displayed in (27). (27)

‘dogma’ means to have the shape

The letters ‘d’, ‘o’, ‘g’, ‘m’ and ‘a’ enclosed with angle brackets denote a sequence of letters, ›graphs‹. Note that, for instance, the difference between small und capital letters can be a relevant difference in shape – cf. He didn’t write ‘dogma’ on the wall, but ‘DoGmA’.13 How is reference to the shape achieved? I don’t think that we need demonstration here, as is often assumed following Davidson (1979). It might be enough to make use of what is generally part of lexical entries in the mental lexicon. A lexical entry minimally contains the idiosyncratic features of an expression, i.e., those of its features which cannot be predicted on the basis of general linguistic rules. Among those features there is at least a description of the phonetic or graphical elements the expression consists of and a description of its meaning. The crucial idea is that, as far as the lexical entry of pure quotations is concerned, the semantic description makes use of the phonetic or graphical description. The lexical entry for pure quotations in (28) consists of three parts: a graphical description (GRAPH), a syntactic (SYN) and a semantic one (SEM). (28)

Lexical entry for pure quotations with quotation marks: GRAPH ®˛© with (i) ˛ being a graphical entity and (ii) with ® and © staying for ‘ resp. ’, “ resp. ”, » resp. «, › resp. ‹, “ resp. ” or similiar things. SYN N SEM predicate(x,˛) x[x has shape ˛]

The graphical description describes the graphical form of the expression: an arbitrary graphical entity ˛ is enclosed with quotation marks. There is the syntactic information that the lexical element is a noun (as for German, we would add the gender feature neuter) and the semantic information that it is a twoplace predicate with the meaning x has shape ˛. The crucial point is that we make use of the variable ˛ – a variable for graphical entities – not only in the graphical description, but also in the semantic description of the lexical entry. 13. As for quotation in spoken language, word and sentence accent, syllable boundaries and other prosodic phenomena can be relevant beside segmental properties (cf. ‘KONstanz’ is the correct pronounciation of the name of the city in the south of Germany, not ‘KonSTANZ’).

260

Jürgen Pafel

Thereby, a pure-quote noun makes reference to one aspect of itself, its shape. (In § 5 we will modify this account a bit.) In working with such a lexical entry, nothing peculiar has to be stipulated. The fact that a variable is shared by different parts of a lexical entry (»structure sharing«) is something which is very common in theories which are working with elaborated lexical entries, theories like Head-driven Phrase Structure Grammar, for instance (see Pollard and Sag 1994: 19, Müller 2008: § 2.4). The lexical entry in (28) is not the entry of a certain word, but a template, i.e., a lexical entry for a class of words. Templates of this kind are well-known in morphology, especially in wordbased morphology. With this in mind, we could call the process of construing pure quotations a generalized conversion. This is even clearer in cases where pure quotation comes without quotation marks. In that case, a lexical entry like (29) is at hand (we could combine (28) and (29) into one entry by allowing in PHON/GRAPH that, with respect to ®˛ ©, ® and © can be left empty). (29)

Lexical entry for pure quotations without quotation marks: PHON/GRAPH ˛ SYN N SEM predicate(x,˛) x[x has shape ˛]

When a noun phrase visibly only contains a pure quotation, it is a singular term, a rigid designator. This is beyond doubt.14 However, if we are right, such a singular term is not semantically unstructured, it has a predicate as a proper part. We, therefore, need an additional semantic element that is responsible for the referential feature. This is the definite description operator () which is defined as follows. The operator in combination with a predicate (x[x]) is a referential term to denote the entity that is maximal with respect to F, the property of being the most salient entity which satisfies the predicate (i.e., the entity that has property F and has all entities as parts which have property F).15 14. “A designator d of an object x is rigid, if it designates x with respect to all possible worlds where x exists, and never designates an object other than x with respect to any possible world” (Kripke in a letter to Kaplan – see Kaplan 1989). The quotation in the embedded clause (i) denotes exactly the same as what the quotation in ‘Dogma’ is a noun of Greek and Latin origin denotes, namely, the lexem ‘dogma’. (i) It is possible that ‘dogma’ will no longer have religious connotations in two hundred years. 15. Using maximality in the semantics of definites goes back to Sharvy (1980). Maximality is especially apt to cope with plural definites.

Two Dogmas on Quotation

261

This is a referential approach to definite descriptions in opposition to a quantifier approach. The adequate analysis of definite descriptions is still an unsolved topic since the time of Frege and Russell, with linguists nowadays leaning towards a referential approach (see, e.g., Heim and Kratzer 1998) and philosophers to a quantifier approach (see, e.g., Neale 1990). Syntactically, I assume that noun phrases are DPs (cf. (11)) and that, as for noun phrases which visibly only contain a pure quotation, the description operator is the meaning of a determiner in syntax which has no counterpart in the phonetic structure of the sentence, i.e., an empty determiner ([D ø] in (11d)). Thus, in the sentence ‘Dogma’ is a noun of Greek and Latin origin the quotation syntactically is a DP (with an empty D and with ‘dogma’ being a noun) and semantically a definite description (the x which has the shape ) denoting the maximal, most salient thing which has the shape . As far as I can see, the predicate view of pure quotation is compatible with each of the uses of pure quotations at which we have been looking. What can a pure-quotation DP refer to? It can refer to things which have a certain shape, a shape which can be presented by graphical or phonetic or – as far as sign languages are concerned – gestural means. First and foremost, such DPs refer to various kinds of linguistic entities: to linguistic types as letters, sounds, chains of letters or sounds, word-forms, lexems, phrases, etc., and to linguistic tokens of various kinds. In (30), the quotation DPs refer to or quantify on concrete linguistic entities in space and time – take a situation where a handwritten letter is in focus. (30)

a. b.

No ‘t’ resembles the other ones. Compare the first ‘t’ with the second one.

In (31), a certain chain of letters is referred to, i.e., a certain linguistic type. (31)

‘Red’ is a sign that articulates one word in English and a different word in Norwegian. (Cappelen and Lepore 2007: 149)

In (32a), a word-form, and in (32b), a lexem is denoted: (32)

a. b.

‘Laughed’ is a past-tense form. ‘Laugh’ is a verb with regular inflection. (Vs. *‘Laughed’ is a verb with regular inflection.)

To summarize, there is a certain vagueness in what pure-quotation DPs refer to (or quantify on) despite the fact that pure-quotation predicates have a definite meaning (»having a certain shape«). This vagueness is resolved by the context which comes into play by the meaning of the description operator »the maximal

262

Jürgen Pafel

most salient thing which satisfies the predicate«. (See § 6 below on what pure quotations do not denote.) We have already seen that there are striking syntactic and semantic similarities between pure quotations and proper nouns. A quotational account of proper names is especially apt to explain these facts. In one version of such an account, proper names like N or the N are analyzed as rigid definite descriptions of the form »the most salient x which carries the name N« (Karnowski and Pafel 2005). The proper name Quine, for instance, gets the meaning »the most salient x which is called Quine«. The proper noun Quine is thus a predicate which applies to everything that is called Quine or carries the name Quine. In exactly the same way as pure quotations, proper nouns refer to their own shape: the semantic description of a proper noun’s lexical entry contains the structure »predicate(x,˛)« with ˛ being identical to the graphical or phonetic description of the proper noun (this structure gets the interpretation »x is called ˛«). 5.

On coreference, binding and quantifying-in in quotation

Before we can ask how well the predicate view fares with respect to the diverse adequacy conditions on theories of quotation which have been set up in the literature, it is necessary to look at coreference, binding and quantifying-in in quotation from the perspective of the predicate view. Despite the fact that the meaning of a pure quotation is no function of the meaning of the quoted material, the quoted material can make an entity salient that is referred to by elements outside of the quotation. (33a), for instance, can easily have a reading where the pronoun sie refers to Angela Merkel (cf. the examples in Partee 1973: 412, 417).16 (33)

a.

b.

Der Satz ‘Angela Merkel ist Bundeskanzlerin und Parteivorsitzende’ ist wahr, wenn sie beide Ämter inne hat. ‘The sentence ‘Angela Merkel is chancelor and party leader’ is true if she has both positions.’ Angela Merkeli ist Bundeskanzlerin und Parteivorsitzende, wenn siei beide Ämter inne hat. ‘Angela Merkel is chancelor and party leader if she has both positions.’

16. Cappelen and Lepore (2007) base their argument that direct quotations are sentences, not NPs, on similar considerations. As these considerations relate to direct and pure quotations alike, their argument concerning the sentential status of direct quotations cannot be convincing.

Two Dogmas on Quotation

263

Nevertheless, pure quotations are syntacto-semantic islands insofar as there are no systematic syntactic or semantic relations between the quoted material and the linguistic context of the pure quotation. There is, for instance, no binding relation in (33a) between Angela Merkel and the pronoun – whereas there is one in (33b). This is confirmed by the fact that quantifier binding is possible in (34b), but impossible in (34a): the quantifier cannot bind outside the scope of the quotation, i.e., (34a) cannot have the same reading as (34b). (34)

a.

b.

*Der Satz ‘Jederi ist Bundeskanzler und Parteivorsitzender’ ist wahr, wenn eri beide Ämter inne hat.17 ‘The sentence Everyone i is chancelor and party leader is true if hei has both positions.’ Jederi ist Bundeskanzler und Parteivorsitzender, wenn eri beide Ämter inne hat. ‘Everyonei is chancelor and party leader if hei has both positions.’

Compare the (im)possibility of reflexive binding in (35) (p.c. Markus Steinbach): (35)

a. b.

*‘Angela Merkeli ’ ist ein Eigenname und von sichi überzeugt. ‘Angela Merkel is a proper name and convinced of herself.’ Angela Merkeli ist Bundeskanzlerin und von sichi überzeugt. ‘Angela Merkel is chancelor and convinced of herself.’

The island character of pure quotations is confirmed by the absence of certain principle C violations. Coreference between the pronoun and Angela Merkel is possible in (36a), but not in (36b). (36)

a.

Sie hat, wenn der Satz ‘Angela Merkel ist Bundeskanzlerin und Parteivorsitzende’ wahr ist, beide Ämter inne.

17. It doesn’t seem to me that the sentence improves at all when the pronoun er is substituted by der, dieser or diese Person ‘this one’, ‘this person’. Even if it would improve, this would not be evidence for a binding relation as (ii) shows. (i) *Der Satz ‘Jeder ist Bundeskanzler und Parteivorsitzender’ ist wahr, wenn der (/dieser, diese Person) beide Ämter inne hat. ‘The sentence Everyone is chancelor and party leader is true if this one (/this person) has both positions.’ (ii) Jederi ist Bundeskanzler und Parteivorsitzender, wenn *deri (*dieseri , ?? diese Person ) beide Ämter inne hat. i ‘The sentence Everyone i is chancelor and party leader is true if this onei (/this personi ) has both positions.’

264

Jürgen Pafel

b.

‘She has both positions if the sentence Angela Merkel is chancelor and party leader is true.’ *Siei hat, wenn Angela Merkeli Bundeskanzlerin und Parteivorsitzende ist, beide Ämter inne. ‘She has both positions if Angela Merkel is chancelor and party leader.’

With respect to coreference and binding, pure, direct and mixed quotation behave similarly: The quoted material in direct and mixed quotes is semantically active in the same way as in pure quotations – the original examples in Partee (1973: 412, 417) were all cases of direct quotation. (37)

a.

b.

Auf dem Schild steht “Hier hat Goethe genächtigt”, obwohl er dort nur den Abend verbracht hat. ‘One can read on the sign “Goethe spent a night here”, although he only spent an evening there.’ Auf dem Schild steht, dass in diesem Haus »Goethe genächtigt« habe, obwohl er dort nur den Abend verbracht hat. ‘One can read on the sign that in this house »Goethe spent a night«, although he only spent an evening there.’

Certain principle C violations do not exist in direct and mixed quotations either – in (38) sie ‘she’ can refer to Angela Merkel: (38)

a. b.

Sie weiß, dass er gesagt hat “Die Merkel ist unfähig”. ‘She knows that he has said “Merkel is inept”.’ Sie weiß, dass er gesagt hat, dass »die Merkel unfähig« sei. ‘She knows that he has said that »Merkel is inept«.’

There are two candidates for quantifying into quotation. The first candidate is quantifier binding into direct quotation (cf. Johnson 2007): (39)

a. b.

Er hat bei jedem Bewerber gesagt “Den halte ich für unfähig”. ‘He said of every candidate “I take him to be inept”.’ Jedesmal, wenn er jemanden Neues trifft, sagt er ihm: “Du erinnerst mich an meinen Bruder”. ‘Every time he makes a new acquaintance, he tells him: “You remind me of my brother”.’

It depends on crucial details of the semantic analysis whether these might be true cases of quantifying into quotation with den in (39a) and du in (39b) being bound variables. According to the analysis of direct quotation I will present

Two Dogmas on Quotation

265

in § 7, these sentences won’t exemplify quantifying into quotation (see footnote 25). The second candidate for quantifying into quotation is certain cases of metalinguistic quantification: (40)

a.

b.

Ich hasse Aussagen der Form ‘Etwas muss so und so und nicht anders sein’. ‘I hate statements of the form Something must be such and such and can’t be another way.’ Ich hasse Aussagen wie ‘Das und das muss so und so und nicht anders sein’. ‘I hate statements like This or that must be such and such and can’t be another way.’

The statements referred to in these two sentences are statements of the form ‘ı must be  and can’t be another way’. Thus, the expressions etwas (‘something’), das und das (‘this or that’) and so und so (‘such and such’), which occur inside the quotation, are mere place-holders for linguistic material. This looks like quantifying into quotation as the following clumsy paraphrase of (40) exhibits: »I hate statements x which are such that there is an ı and a  such that x has the shape « Thus, the pure quotation ‘Etwas muss so und so und nicht anders sein’ has the meaning (41)

x[∃ı∃ Shape(x,)]

But even these examples of metalinguistic quantification need not be taken as true cases of quantifying into quotation. In the semantic description of a purequote noun according to the predicate view in § 4, the second argument of the predicate is an expression denoting a chain of phonetic or graphical elements (˛ in the templates (28) and (29)) – an expression taken from the phonetic/graphical description of the pure-quote noun. The sentences being discussed make it necessary to assume quantifiers which range on phonetic or graphical elements and bind into the second argument of the predicate.18 However, this is not really quantifying into quotation, as in the semantic description 18. Recently, Sudo (2008) presented a semantics for existential quantification into quotation. It seems that quantifiers introduced by this kind of metalinguistic quantification have the smallest possible scope with respect to quantifiers and negation (pace Sudo). (i) He has never given a statement of the form Something must be such and such and can’t be another way. (ii) He hasn’t given a statement of the form Something must be such and such and can’t be another way.

266

Jürgen Pafel

there are no quotations, but a special vocabulary for phonetic/graphical elements. These cases of metalinguistic quantification indicate, however, that the variable-sharing between the phonetic/graphical and the semantic description, which is an essential aspect of the predicate view in § 4, might be too simple of a procedure to account for all kinds of pure quotation. It seems to work in many, perhaps most, cases, but not in certain cases of metalinguistic quantification. To cope with them in a predicate view of pure quotation, the semantic description in the lexical entry for the pure-quotation template has to be supplemented by a function f which takes the phonetic/graphical description as an argument: (42)

predicate(x,f(˛))

In the most frequent case, the function maps a phonetic/graphical description onto itself. In metalinguistic quantification, however, it inserts variables and quantifiers on phonetic/graphical elements, i.e., if there are occurrences of expressions like so und so, das oder das or etwas in ˛, the function f can map ˛ onto ˛’ with ˛’ being identical to ˛ except that occurrences of the beforementioned expressions are substituted by variables on phonetic/graphical elements and that existential quantifiers are introduced which bind these variables.19 In English and German, place-holders occur in direct quotation also – see the following passage from Robert Musil’s Mann ohne Eigenschaften (begin of chapter 4), where dies oder das functions as a place-holder for linguistic material: (43)

[Wenn es aber Wirklichkeitssinn gibt, und niemand wird bezweifeln, daß er seine Daseinsberechtigung hat, dann muß es auch etwas geben, das man Möglichkeitssinn nennen kann.] Wer ihn besitzt, sagt beispielsweise nicht: Hier ist dies oder das gesehen, wird geschehen, muß geschehen.20

There is a convention in quoting to indicate unquoted material inside of quoted material. I could, for instance, quote the statement ‘Quine, who died in 2000, 19. In the system of my 2005 the existential quantifiers are put into the quantifier store of ˛’ and are assigned features (a-weak, r-weak) which make the quantifiers receive the smallest absolute and relative scope (cf. footnote 18). 20. Translation: “But if there is a sense of reality, and nobody will doubt that it has every right to exist, then there must be something in addition which one could call sense of possibility. He who owns it would, for instance, not say: This or that did happen, will happen, must happen here.”

Two Dogmas on Quotation

267

is one of the most important philosophers of the twentieth century’ by omitting the relative clause as in (44): (44)

He wrote, “Quine […] is one of the most important philosophers of the twentieth century”.

This sign for omission in quotation is a further place-holder for linguistic material and occurs in mixed quotation too: John said that you are an »… idiot« (with ‘…’ being a place-holder for, for instance, fucking); Er hat wörtlich gesagt, dass du »ein A…loch« wärst (‘He literally said that you are »an a…hole«’). To cope with metalinguistic quantification, there was no need to refer to substitutional quantification – it is objectual existential quantification on graphical or phonetic elements that we made use of. Brendel (2007: § 4) argues that substitutional quantification is necessary to capture what is special about sentences of the type ‘Pentasyllabic’ is pentasyllabic. In (45) the quoted ‘F’ should be interpreted as a substitutional variable bound by a substitutional existential quantifier: (45)

There are (property expressions) F such that ‘F’ is (/has the property) F.

We can, however, capture what is special about this type of sentences without substitutional quantification: (46)

There are (property expressions) F such that F has the property which F denotes.

‘Pentasyllabic’ has the property which ‘pentasyllabic’ denotes is a true instantiation of (46) and entails ‘Pentasyllabic’ is pentasyllabic – presupposing the disquotational schema »Something has the property which ‘F’ denotes iff it is F«. In contrast to their syntactic behaviour we looked at in §§ 2 and 3, we could observe in this paragraph that pure quotation, on the one hand, and direct and mixed quotation, on the other, behave very similarly with respect to coreference, binding and metalinguistic quantification.

6.

The adequacy of the predicate view

The predicate view of pure quotations satisfies many of the adequacy conditions for quotation accounts that have been set up in the literature (cf. Saka 1998: § 6, Harth 2002: § I.5, Steinbrenner 2004: 96f., Cappelen and Lepore 2007: § 3, Harth 2007: § 4). The predicate view, to begin with, makes it possible to explain

268

Jürgen Pafel

that understanding the mechanisms of quotation requires mastering a capacity to understand and generate a potential infinity of new quotations. In the predicate view, the competence in understanding and generating pure quotations consists essentially in possessing lexical templates like (28) and (29), which allow arbitrarily many uses. The predicate view is compatible with the fact that pure quotations can be used to introduce novel words, symbols and alphabets – but it must be presupposed that the graphical/phonetic description in (28) or (29) is rich enough to represent shapes of arbitrary form. Quotation is iterable: with (28), ‘dogma’ as well as ‘ ‘dogma’ ’ can be generated, the first one on the basis of dogma and the second one on the basis of ‘dogma’. The two quotations differ in meaning: ‘dogma’ means »x has shape « and ‘ ‘dogma’ ’ means »x has shape «. Quotation marks are not an essential element of quotation in the predicate view, see the template (29). The predicate view explains the opacity of pure quotations, i.e., the fact that in quotations neither coextensive nor synonymous expressions can generally be substituted for one another salva veritate. As two coextensive or synonymous expressions necessarily differ in shape, the pure quotations of two such expressions differ in meaning according to the predicate view and, thus, cannot generally be substituted for one another salva veritate. Let’s assume that bachelor and unmarried man are coextensive. Then (47) is true, but (48) is false as the predicates having the shape and having the shape are not coextensive. (47) (48)

Bachelor and unmarried man are coextensive. ‘Bachelor’ and ‘unmarried man’ are coextensive.

Yet the fact that pure quotations are opaque does not preclude that the quoted material has a meaning. The quoted material can, as we have seen in § 5, make an entity salient that is referred to by elements outside of the quotation. Ordinary objectual quantification into pure quotation is impossible and the predicate view of pure quotation is fully compatible with this fact. Chomsky is a famous linguist and a well-known political writer entails There is someone such that he is a famous linguist and a well-known political writer. However, ‘Chomsky’ is a proper name does not entail something like There is someone such that ‘he’ is a proper name (which in itself is not a comprehensible wellformed sentence due to the vacuous quantification). We have already seen that in diverse forms of quotation, metalinguistic quantification is possible and can be conceived as objectual quantification on linguistic entities.

Two Dogmas on Quotation

269

The picture-like character of pure quotation (“a quotation somehow pictures what it is about” Davidson 1979 [1984: 83]) is accounted for by the pure quotation’s semantics which makes reference to its shape. It is sometimes claimed that pure quotations are ›flexible‹ in the sense of being ambiguous to a high degree (see Saka 1998, Hart 2002, 2007). It is claimed that a pure-quotation DP can denote very different things depending on the predicate it is combined with. We have already seen in § 5 that pure-quotation DPs can denote various kinds of linguistic types and tokens. However, the influence of the predicate should not be exaggerated. With respect to (49a), for example, one might argue that the quotation does not denote a word form which has phonological, morphological, syntactic and semantic properties, but a sequence of phonemes only as the predicate features a phonological concept. Such a move is unnecessary. There are clear cases where the quotation refers to a word form (see (49c), where various linguistic properties are predicated of what is denoted by the pure quotation), and reference to word form is possible, perhaps even mandatory in (49a) if we consider the continuation in (49b). (49)

a. b. c.

‘Runs’ contains four phonemes. ‘Runs’ contains four phonemes and it is a verb in the third person singular. ‘Runs’ and ‘rennt’ are similiar with respect to the fact that they contain four phonemes, are verbs in the third person singular and denote a certain kind of bodily movement.

Even if a predicate features semantic aspects, a pure-quotation DP denotes a linguistic entity and not the meaning of the linguistic entity. See (50), where we obviously refer to words or lexems, not to meanings: (Samstag is the German translation of saturday) (50)

a. b. c.

‘Saturday’ and ‘Samstag’ are synonymous. The nouns ‘Saturday’ and ‘Samstag’ are synonymous. *The day that ‘Saturday’ denotes and the day that ‘Samstag’ denotes are synonymous.

When the linguist states The concept ‘Saturday’ is the concept ‘Samstag’, he makes use of a metonymic way of speaking: “The concept that ‘Saturday’ denotes is identical to the concept that ‘Samstag’ denotes.«

270 7.

Jürgen Pafel

A propositional view on direct quotation

With respect to coreference, binding and metalinguistic quantification, pure, direct and mixed quotation behave very similarly. With respect to syntax, pure, direct and mixed quotation differ in a significant manner. Pure quotations are nouns, direct quotations sentences, and mixed quotations constituents and nonconstituents of various types. It would come as a surprise if these syntactic differences should not have semantic consequences. If pure quotations are predicates, as nouns usually are, it is not probable that pure and direct quotations are semantically on a par with each other. Expressions of the sentence type usually are not predicates. Thus, the semantic part of the second dogma is very doubtful too. Nevertheless, the assumption (being a conclusion of the two dogmas) that direct quotes are noun phrases and singular terms denoting linguistic expressions might be true with respect to its semantic part, as it seems obvious that direct quotations denote linguistic expressions. Thus, pure quotation and direct quotation seem to have at least something in common, namely that direct quotes as well as noun phrases which visibly only contain a pure quotation denote linguistic expressions. Even this is not as obviously true as one is tempted to think. Two related facts point into this direction. Firstly, if you want to translate ‘It rains’ is an English sentence into German, you should not translate the pure quotation, otherwise you would get a false sentence as the translation of a true one (‘Es regnet’ ist ein englischer Satz). Even if the translation would not get a false sentence (as, e.g., the translation of ‘It rains’ consists of two words as ‘Es regnet’ besteht aus zwei Worten), it would not be faithful to what the sentence actually means. With direct quotations, we do not find this strict dependence on the quoted expressions. With direct quotations, it is possible and quite natural to translate the quotation together with the other parts of the speech report – see, e.g., Heine said, “I like Moritz most” as a translation of Heine hat gesagt “Moritz ist mir der liebste”. Reading the English sentence in an English biography on the German poet Heinrich Heine, one would not be inclined to think that Heine made his remark in English. You can observe this phenomenon in nearly every edition of a newspaper which reports on foreign affairs.21 Secondly, as for pure quo21. Compare Cappelen and Lepore (2007: § 5) on ›impure direct quotes‹, i.e., quotes which are translated and quotes which are fixed up (“omitted words filled in, grammatical infelicities repaired, repeated words eliminated, utterances of ‘hmm’, ‘aahh’, etc. eliminated”). Contrary to Cappelen and Lepore, I think that these phenomena have a »bearing on how we think about the semantics of quotation« (ibid. § 5.3), as they are evidence for a propositional view of direct quotation.

Two Dogmas on Quotation

271

tation, it is possible to insert linguistic material from a distinct language into a sentence without getting code mixing or code switching (see, e.g., ‘Es regnet’ is a well-formed German sentence). With direct quotation, we seem to get a mix or switch (Heine said, “Moritz ist mir der liebste”). Thus, it is questionable whether direct quotations really behave like singular terms that denote linguistic types or tokens. As direct quotations are sentences, it is actually near at hand to assume that they have the same type of denotation as ordinary that sentences, that is, that they denote propositions. They might additionally refer to linguistic entities, but basically, they have a propositional content – as that clauses and indirect quotations do.22 This (surprising) conclusion is supported by the fact that direct quotes can be used in attitude contexts (Heine thought “I like Moritz most”) – it is thus no longer necessary to claim that this is a metaphorical use of direct quotation – and by coordination data like the following, where a direct quote is coordinated with a that clause: (51)

Philosophiestudenten in aller Welt lernen: “Wovon man nicht sprechen kann, darüber muss man schweigen”. Und dass er ein kubistisches Haus baute, das Palais Wittgenstein in Wien. (Die Zeit, 24 july 2008, page 13)23

One way to implement such a propositional view on direct quotation is to analyze a speech report like S said “…” as follows: (52)

S said that p, with p being the propositions denoted by the sentence ‘…’ if ‘…’ is uttered in context C.

The direct quotation “…” itself can be analyzed as a definite description:24 22. The influential scepticism towards propositions, partially fueled by behavoristic inclinations, might be the reason that the propositional view is not an investigated option in the analysis of direct quotation. Confronted with the varieties of embedded clauses, the sceptic takes direct quotation as the obvious point of departure with apparently no commitment to propositions, views indirect quotation with direct quotation in mind and tries to get a grip on content clauses in attitude sentences with indirect quotation in mind (cf. Falkenberg 1998: § 11). 23. Translation: “Students of philosophy all over the world learn “One should be silent about what one cannot speak”. And that he built a cubist house, the Palais Wittgenstein in Vienna.” 24. Direct and indirect quotations are thus two different kinds of descriptions of propositions. (In my 1999 I have shown how the similiarities and differences between several kinds of declarative and interrogative clauses can be accounted for if we take them to be different kinds of definite descriptions of propositions.) The further development

272 (53)

Jürgen Pafel

p[p are the propositions denoted by ‘…’ if ‘…’ is uttered in context C]

The locution ‘uttered in context C’ might be cashed out as: ‘uttered by speaker S to addressee A at time T and location L with respect to salient entities E’.25 In direct quotation, indexicals are shifted: A token of the indexical I which is part of a direct quotation embedded in a speech report, for instance, does not refer to the one who utters the token, i.e., to the speech reporter, but to the person whose speech is reported. This indexical shift, which is an essential feature of direct quotation, is the consequence of the context-shift built into the semantics of direct quotation. Quite similar to pure quotation, direct quotation can be captured in a lexicalist way: (54)

Lexical template for direct quotation ® © ˛ PHON/ GRAPH with (i) ˛ being a phonetic or graphical entity and (ii) with ® and © staying for ‘ resp. ’, “ resp. ”, » resp. «, › resp. ‹, “ resp. ” or similiar things or being empty SYN S SEM p[p are the propositions denoted by f(˛) if it is uttered in context C]

Identical to the template for pure quotation, there is structure sharing between the phonetic/graphical and the semantic description in form of the variable ˛. It is this aspect which might be common to all kinds of quotation and might explain why pure quotation and the other kinds of quotation behave very similarly with respect to coreference, binding and metalinguistic quantification. Mixed of the propositional view should, of course, elucidate what exactly the ›propositions‹ are that are denoted by direct quotations, as not only statements, but also questions, commands, exclamations etc. can be directly quoted. The fact that direct quotations can be more sensitive to the actual wording of the original utterance than indirect quotations is, I suppose, a fact which should be accounted for by pragmatic means of a familiar Gricean sort (“if someone uses direct instead of indirect quotation, one is entitled to assume that he is faithful to the original utterance”). 25. A sentence like He said of every candidate “I take him to be inept” (cf. (39a)) could now be analyzed along the following lines: With respect to every candidate2 he1 uttered assertively the proposition that is denoted by ‘I take him to be inept’ if it is uttered by speaker S1 to addressee A at time T and location L with respect to the salient entity X2 . According to this analysis, there is no quantification into quotation in such a sentence.

Two Dogmas on Quotation

273

quotation must resemble direct quotation to some extent, as there can be indexical shift in mixed quotation too (see Cumming 2005, Cappelen and Lepore 2007: 29f.). In scare quoting we have structure sharing as well if, in a sentence like (55a), we analyze the noun phrase containing the scare quotation semantically as (55b): (55)

a. b.

He invited the ›best actress in the world‹. the x which is called ‘best actress in the world’ by B

The meaning of a scare quote can be described as x[x is called f(˛) by B] with ˛ being the phonetic/graphical description of the linguistic material in between the quotation marks. 8.

Beyond the two dogmas

There is ample evidence that throws doubt on the first dogma of quotation, i.e., the assumption that pure quotations are noun phrases and singular terms which denote linguistic expressions. With the evidence presented, it seems obvious that pure quotations are nouns. The predicate view of pure quotations is at least a very natural consequence of the view of pure quotations as nouns. As for the second dogma, i.e., the assumption that pure and direct quotation are more less on a par with each other syntactically and semantically, we have clearly seen that pure quotation and direct quotation differ in a significant manner in syntax. Pure quotations are nouns, while direct quotations are sentences. It does not come as a surprise that they differ semantically too. These differences paved the way to the propositional view of direct quotation. Although I stressed the differences between the kinds of quotation throughout this essay, we can see, nevertheless, that the different uses of quotation marks we looked at constitute a family of uses which resemble each other, that quotation marks are a case of polysemy. References Bach, Kent 2002 Brendel, Elke 2007

Giorgione was so-called because of his name. Philosophical Perspectives 16: 73–103. Zitat und Selbstbezug. In Zitat und Bedeutung, E. Brendel, J. Meibauer and M. Steinbach (eds.), 27–43. Hamburg: Buske.

274

Jürgen Pafel

Brendel, Elke, Jörg Meibauer and Markus Steinbach (eds.) 2007 Zitat und Bedeutung. Hamburg: Buske. [= Linguistische Berichte, Sonderheft 15] Burge, Tyler 1973

Reference and proper names. Journal of Philosophy 70: 553–576.

Cappelen, Herman and Ernie Lepore 2005 Varieties of quotation revisited. In Hybrid quotations, P. de Brabanter (ed.), 51–75. Amsterdam: Benjamins. Cappelen, Herman and Ernie Lepore 2007 Language turned on itself. The semantics and pragmatics of metalinguistic discourse. Oxford: Oxford University Press. Clark, Herbert H. and Richard J. Gerrig 1990 Quotations as demonstrations. Language 66: 764–805. Cumming, Samuel 2005 Two accounts of indexicals in mixed quotation. In Hybrid quotations, P. de Brabanter (ed.), 77–88. Amsterdam: Benjamins. De Brabanter, Philippe (ed.) 2005 Hybrid quotations. Amsterdam: Benjamins. [=Belgian Journal of Linguistics 17, 2003]. De Bruyn, Günter 2006 Als Poesie gut. Schicksale aus Berlins Kunstepoche 1786 bis 1807. Frankfurt/M.: Fischer. Davidson, Donald 1979 Quotations. Theory and Decision 11: 27–40. (Reprinted in: D. Davidson (ed.), Essays on truth and interpretation, 79–92. Oxford: Clarendon 1984.) Elugardo, Reinaldo 2002 The predicate view of proper names. In Logical form and language, G. Preyer and G. Peter (eds.), 467–503. Oxford: Clarendon. Falkenberg, Gabriel 1998 Sinn, Bedeutung, Intensionalität. Der Fregesche Weg. Tübingen: Mohr Siebeck. Frege, Gottlob 1892

Über Sinn und Bedeutung. Zeitschrift für Philosophie und philosophische Kritik, C: 25–50.

Gómez-Torrente, Mario 2005 Remarks on impure quotation. In Hybrid quotations, P. de Brabanter (ed.), 129–151. Amsterdam: Benjamins.

Two Dogmas on Quotation Harth, Manfred 2002 Harth, Manfred 2007

275

Anführung. Ein nicht-sprachliches Mittel der Sprache. Paderborn: Mentis. Die logische Form von Anführungssätzen. In: Zitat und Bedeutung, E. Brendel, J. Meibauer and M. Steinbach (eds.), 135–151. Hamburg: Buske.

Heim, Irene and Angelika Kratzer 1998 Semantics in generative grammar. Malden: Blackwell. Johnson, Michael 2007 The punctuation theory of quotation (or something like it). Paper presented at the ‘First International Conference on Quotation and Meaning’, 20 October 2007, Mainz. Kaplan, David 1989

Afterthoughts. In Themes from Kaplan, J. Almog et al. (eds.), 567– 614. Oxford: Oxford University Press.

Karnowski, Pawel and Jürgen Pafel 2005 Wie anders sind Eigennamen? Zeitschrift für Sprachwissenschaft 24: 45–66. Katz, Jerrold J. 2001

The end of millianism: Multiple bearers, improper names, and compositional meaning. Journal of Philosophy 98: 137–166.

Larson, Richard and Gabriel Segal 1995 Knowledge of meaning: an introduction to semantic theory. Cambridge, MA: MIT Press. Meibauer, Jörg 2003

Phrasenkomposita zwischen Wortsyntax und Lexikon. Zeitschrift für Sprachwissenschaft 22: 153–188.

Meinunger, André 2000 Syntactic aspects of topic and comment. Amsterdam: Benjamins. Müller, Stefan 2008 Neale, Stephen 1990 Pafel, Jürgen 1999 Pafel, Jürgen 2005

Head-driven phrase structure grammar. Eine Einführung. Tübingen: Stauffenburg. Descriptions. Cambridge, MA: MIT Press. Interrogative quantifiers within scope. Linguistics and Philosophy 22: 255–310. Quantifier scope in German. Amsterdam: Benjamins.

276

Jürgen Pafel

Pafel, Jürgen 2007

Ein Essay mit dem Titel On pure quotation. In Zitat und Bedeutung, E. Brendel, J. Meibauer and M. Steinbach (eds.), 201–214. Hamburg: Buske.

Partee, Barbara Hall 1973 The syntax and semantics of quotation. In A Festschrift for Morris Halle, S.R. Anderson and P. Kiparsky (eds.), 410–418. New York: Holt, Rinehart & Winston. Pollard, Carl and Ivan A. Sag 1994 Head-driven phrase structure grammar. Chicago: The University of Chicago Press. Quine, Willard van Orman 1960 Word and object. Cambridge, MA: MIT Press. Recanati, François 2001 Open quotation. Mind 110: 637–687. Saka, Paul 1998

Quotations and the use-mention distinction. Mind 107: 113–135.

Sharvy, Richard 1980 A more general theory of definite descriptions. The Philosophical Review 89: 607–624. Steinbrenner, Jakob 2004 Zeichen über Zeichen. Grundlagen einer Theorie der Metabezugnahme. Heidelberg: Synchron. Sudo, Yasutada 2008

Quantification into quotations: evidence from Japanese wh-doublets. In Proceedings of SuB12, A. Grønn (ed.), 613–627. Oslo: ILOS.

Webelhuth, Gerd 1992 Principles and parameters of syntactic saturation. New York: Oxford University Press.

Reporting and Quoting in Signed Discourse∗ Josep Quer

1.

Introduction

The linguistic resources displayed by sign languages (SLs) in order to reproduce or recreate someone else’s utterances or thoughts have remained absent from the syntactic and semantic research on reported discourse and quotation till very recently. This paper aims at contributing some fresh SL data to the discussion about the proper characterization of reported discourse contexts and to analyze some of their peculiarities within the broader perspective of (in)direct reports. Despite the apparent differences at the surface, it will be argued that the core distinctions of reported structures in signed discourse coincide with the ones we find across spoken languages. I will describe the formal mechanisms distinguishing direct and indirect discourse in the signed modality. Special attention will be devoted to the behaviour of indexicals in shifted contexts. I will defend and refine a unified treatment of quotational and non-quotational use of role shift in SLs, in line with Zucchi (2004) and other previous research. An analysis will be offered where a covert Point of View Operator will be posited and held responsible for the morphological and semantic properties of role shift constructions. One partial conclusion will be that the crosslinguistic validity of the “Shift-Together Constraint” for indexicals by Anand & Nevins (2004) might not be instantiated in the SLs examined.

* This chapter draws directly on the research reported in Quer (2005) and Quer & Frigola (2006). I would like to thank the audiences at SALT 15 (Los Angeles, April 2005), the Sign Language Workshop Signa Volant (Milan, June 2005) and the conference Text structure: form, meaning and processing (Göttingen, July 2010) for their valuable comments and criticism. Special thanks go to Annika Herrmann and Markus Steinbach. This work would have been impossible without the collaboration of my Deaf colleagues Santiago Frigola and Delfina Aliaga. The research was partly made possible by a grant awarded by the Spanish Ministry of Science and Innovation to Josep Quer (FFI2009-10492).

278

Josep Quer

The new data discussed comes mainly from Catalan Sign Language (LSC), the SL used by the Deaf Community in Catalonia.1 However, I also undertake limited crosslinguistic comparisons with published American Sign Language (ASL), German Sign Language (DGS), Lingua Italiana dei Segni (LIS) and Danish Sign Language (DSL) data. 2.

Role shift: Properties of reporting strategies in SLs

The grammatical phenomenon known as role shift (RS) (also called role taking, role switching, reference shift or in some instances constructed dialogue, as in Metzger 1995) in SLs is often identified as the equivalent of a direct discourse report or quotation in the visual-gestural modality. It is the genuine means these languages have in order to convey the utterances or thoughts ascribed to a discourse agent, and sometimes to reproduce or rather recreate the dialogue between two or more subjects in a displaced context. It mostly appears in narratives, but not exclusively. General characterizations and analyses of the phenomenon have been put forth by Engberg-Pedersen (1995), Lee et al. (1997), Poulin (1994), Poulin & Miller (1995), Lillo-Martin (1995, in press), Zucchi (2004), Quer (2005), Quer & Frigola (2006) and Herrmann & Steinbach (2007, 2009, 2010) among others. From a descriptive point of view, RS is typically flagged by two phenomena: (a) displacement of the referential loci associated with 1st and 2nd person discourse referents in signing space; (b) array of nonmanual markers coarticulated with the reported utterance. Starting with the latter, we observe that among the formal characteristics associated with role shift the following ones are the most remarkable, and recurrent: (a) slight body shift to the side associated with the author of the reported utterance;

Fig. 1. Body shift.

1. If not indicated otherwise, the examples appearing in this paper are from LSC.

Reporting and Quoting in Signed Discourse

279

(b) linguistic and affective facial expression associated to the author of the reported utterance;

Fig. 2. Facial expression.

(c) change in the position of the head;

Fig. 3. Head position.

(d) change in the direction of the eye gaze towards the alleged interlocutor in the reported context and, consequently, temporal interruption of eye contact with the actual interlocutor.

Fig. 4. Eye gaze.

These nonmanual markings are simultaneously coarticulated and coordinated with the manual material that is interpreted as the reported proposition. Not all of them are obligatory, and the nonmanual marking is often so subtle that it escapes nonnative signers. For DGS, Herrmann & Steinbach (2009, 2010) conclude from a corpus study that the only compulsory marker for RS is eye gaze and that the other markers can supplement it in an overlapping fashion

280

Josep Quer

depending on the narrative style of the signer. Although no statistical count on a corpus study has been carried out, for LSC the data analyzed point to the obligatoriness of eye gaze for minimal RS marking. The second main characteristic of RS is that the indexical referential loci are formally and interpretively displaced. Reference shift affects 1st and 2nd person pronouns (IX-1, IX-2),2 the corresponding possessive pronouns and the verbal agreement associated to them (in agreement verbs and in the agreement auxiliary). Therefore, within a role fragment, the 1st person pronoun typically does not refer to the actual signer but to another individual and the 2nd person one does not refer to the actual interlocutor but to the one in the reported context. In example (1), the pronoun IX-1 occurring in the role part is not interpreted as the actual utterer, but as the referent of the proper name MANEL. At the same time, the interpretation of subject and object agreement of the verb GIVE is not established with 1st and 2nd persons of the actual context of utterance, but with those of the shifted context (Fig. 3).3,4 (1)

role-i MANELi THINK IX-1i 1-GIVE-2 AT-ALL

‘Manel thinks that he won’t give me anything at all.’

2. Pronominal reference in the singular is typically realized in SLs as an index (glossed as IX) consisting in a pointing handshape that is oriented towards present referents (IX-1=author, IX-2=addressee, IX-3=[–author, –addressee] present referent). Nonpresent referents are localized in a locus of the signing space to which IX points. 3. I follow the usual glossing conventions in the SL literature, according to which manual signs are represented by the capitalized word corresponding to the translation of the sign. The scope of nonmanual markings is represented with a line that spreads over the manual material with which it is coarticulated. The relevant abbreviations for the purposes of this paper are the following ones: #-VERB-# (verb agreeing with subject and object; the number before the verb refers to the grammatical person of the former and the one after the verb refers to the latter); AGR (unbound agreement marker); eg (eyegaze); IXa (locative index poiting to locus a); IX-# (pronominal index; the number corresponds to person); hs (negative headshake); RS (role shift); t (topic marking); wh (wh marking); +++ (repetition of the sign). The referential indices i, j, etc. link the first person role in RS fragments to the intended author of the reported utterance. 4. Sometimes the role adopted does not correspond strictly speaking to an individual, but to an animal (personification) or a group of people or an entity (e.g. ‘The government think that…’). It is also possible that the reported subject is the actual signer, but in another context and maybe interacting with another illocutionary agent.

Reporting and Quoting in Signed Discourse

281

Fig. 5. 1-GIVE-2

Similarly, in the RS fragment occurring in the LSC example in (2), the 1st person pronoun does not refer to the actual signer but to the signer of the reported context, as we can observe: the pronominal sign IX-1 ‘I’ does not refer to the author of this example, but to the referent of JOAN, the individual to whom the thought is ascribed as subject of the verb THINK introducing the reported discourse. (2)

t RS-i IXa MADRID JOANi THINK IX-1i STUDY FINISH HERE MADRID

‘When he has in Madrid, Joan thought he would finish his studies there in Madrid.’

3.

Varieties of report in RS

The strategy with RS constitutes a much more genuine mechanism for SLs to report someone’s utterance or thought than regular indirect discourse of the kind we find in spoken languages like English, for instance. However, the latter is an existing option, as we can see in example (3) with regular embedding of a complement clause reporting on Anna’s utterance. It should be compared with the parallel case with RS in (4). If we put he nonmanual markings of RS aside for a moment, the main difference resides in the use of pronominals: the 1st person pronoun in (4) does not get interpreted in the actual context of utterance, but in the derived context of the report. The subject pronominals of the reports in (3) and (4) is contrasted in Figures 6 and 7. (3)

ANNAi 3-SAY-1 IX-3i FED-UP LOSE+++

‘Anna told me that she was fed up with losing so often.’ (4)

RS-i ANNAi 3-SAY-2 IX-1i FED-UP LOSE+++

‘Anna told you that she was fed up with losing so often.’

282

Josep Quer

Fig. 6. IX-3i in (3).

Fig. 7. IX-1i in (4).

Unlike in English indirect discourse, for instance, the default interpretation of indexical personal pronouns in the scope of RS is not determined by the utterance context but rather by the context of reported conversation.5 This is not a particular fact of RS in LSC, but it seems to be recurrent in the other SLs where the RS phenomenon has been attested, as in (5) from ASL (Lee et al. 1997). (5)

JOHNi

SAY

IX-1i

WANT

RS-i GO

(ASL)

‘John said: “I want to go.”’/ ‘John said that he wanted to go.’6 Lee et al. (1997) treat examples of this sort as instances of reported direct speech or direct quotation realized as two juxtaposed clauses. Although such cases do exist, we will see that RS is also realized in constructions where reported direct speech cannot be at play (cf. Lillo-Martin 1995, who provides empirical arguments for the embedded status of the reported clause in ASL). Characterizing RS fragments simply as direct speech is not accurate. On the one hand, LSC has explicit markers of direct quotes such as VOICE (Fig. 8), SAY1 SENTENCE (Fig. 9), AUTHOR+verb of saying (Fig. 10), DECLARE (Fig. 11), etc. These markers cannot introduce regular indirect discourse reports, i.e. non-RS structures. 5. This remark only makes sense under the narrow view that SL pronouns are pointing signs directed to referential locations. From a composite view of pronominal paradigms like Berenz’s (2002), both manual and nonmanual ingredients determine the form of the pronoun: in the RS version of the 1st person pronoun the pointing is equally directed towards the signer’s chest, but unlike in the non-RS counterpart, eye gaze is not directed towards the interlocutor, but rather towards the locus of the reported interlocutor. 6. The first translation is the one offered originally by Lee et al. (1997), but in order to remain neutral as to the direct/indirect character of RS, the indirect report version in English has been added.

Reporting and Quoting in Signed Discourse

Fig. 8. VOICE.

283

Fig. 9. SAY 1 SENTENCE.

Fig. 10. AUTHOR

Fig. 11. DECLARE

An instantiation of direct quote can be found in (6). RS-i (6)

ANNAi EXPLAIN SAY1 SENTENCE IX-1i BROTHER MAN 3-IGNORE-1

“Anna told me: ‘My brother ignores me.”’ On the other hand, RS segments can be introduced by propositional attitude predicates such as SAY (Fig. 12), REPLY (Fig. 13), THINK (Fig. 14), etc. (see for instance (1)–(3)).

284

Josep Quer

Fig. 12. SAY.

Fig. 13. REPLY.

Fig. 14. THINK

At face value, these RS introducers might seem to simply signal direct quotation. On closer inspection, though, we identify cases that cannot be classified as such due to the interpretation of the indexicals appearing in its domain. An instance of RS used in a non-direct quotation can be found in (7) from LSC. The crucial fact to note in this example is that the reported thought could not be a quotation with the intended meaning of the indexical HERE: uttered in Barcelona, HERE in (7) refers by default to Barcelona and not to the reported context the location parameter of which is Madrid, where Joan is argued to entertain the reported thought. (7)

t RS-i IXa MADRIDm MOMENT JOANi THINK IX-1i STUDY FINISH HEREb

‘When he was in Madrid, Joan thought he would finish his study in Barcelona.’ This kind of RS example must be contrasted with one involving direct quotation introduced by DECLARE, as in (8): given similar place and time coordinates as in the previous example, the reported discourse can only be interpreted as meaning that Joan said he would finish his study there last year. In this case, the time and place indexical coordinates do not have access to the context of utterance, arguably because RS has been overtly introduced by a direct quote marker. (8)

RS-i YEAR-LAST JOAN-i DECLARE IX-1i HERE YEAR-THIS STUDY FINISH

‘Last year Joan declared: ‘This year I’ll finish my study here’.’ On the basis of this evidence we can safely conclude that RS structures do not form a uniform class and that they actually serve the expression of both direct and non-direct reported discourse. A further property that distinguishes both cases is of syntactic nature: while direct quotes in LSC can be preposed (topi-

Reporting and Quoting in Signed Discourse

285

calized) vis-à-vis the introducing main sentence (9), ungrammaticality obtains if we try to do the same with a RS indirect discourse segment (10): (9)

RS-i IX-1 ANGRY AGR-2 IX-3c SENTENCE SAME IX-3a ANNAi 3a-SAY-3b IX-3b PEDRO

“‘I’m angry at you’ said Anna to Pedro.’ (10)

RS-i *IX-1 ANGRY AGR-2 YESTERDAY IX-3a ANNAi 3a-SAY-3b IX-3b PEDRO

‘Anna said to Pedro that she was angry at him.’ This brief characterisation of RS varieties allows us to distinguish between standard indirect discourse of the English type from non-direct reported discourse. In LSC the former is not marked by RS, while the latter is, but unlike direct reported discourse (direct quote), it allows for partial indexical access to the matrix context. For indexical interpretation, see section 5 below. However, further disctinctions have been made in the domain of reported discourse in SLs. As discussed in Zucchi (2004) for LIS, RS seems not to be exclusively restricted to quotational environments, being able to also appear outside the scope of an attitude predicate, as in (11): the main clause is not an attitude report in the usual sense, as it is not introduced by a reportive predicate such as ‘say’ or ‘think’, but it has the same surface properties as a direct quotation. In this example RS implies that the subject of the agreeing verb DONATE is coreferential with GIANNI (Zucchi 2004). RS-i

(11)

GIANNIi

ARRIVE

BOOK 1-DONATE-2

(LIS)

‘When Gianni arrives, he will give you the book as a present.’ The question can be raised whether this is an instance of quotative RS or not. As a matter of fact, the translation given does not convey the reportive component of the utterance. A more accurate rendering would be “When Gianni arrives, he will be like ‘I’ll give you the book as a present.”’, where the quotative component is clearly present. Some other examples instantiate the possibility for RS to occur independently (non-introduced RS), as in (12) from ASL. As in the previous case, no overt predicate introduces the report and only the attitude holder MOM is made explicit (Lillo-Martin 1995).

286

(12)

Josep Quer

MOMi IX-1i

RS-i BUSY

(ASL)

‘Mom’s like, I’m busy!’ A comparable case in LSC can be found under (13): the RS stretch has to be attributed to the author of the e-mail referred to in the first part of the utterance and the 2nd person pronoun is linked to the actual 1st person that reports having received the e-mail. (13)

t RS-i JOANi MAIL ELECTRONIC 3-SEND-1 IX-2 ALL GUILT IX-2

‘In an e-mail Joan sent to me, he was like ‘It’s all your fault.” The peculiarity these three examples of RS share is that the RS is not explicitly introduced. I would like to claim that what has been called “non-quotative RS” is plain RS lacking an explicit introducer. This is actually the default in extended narrative discourse, where illocutionary agents have been established and utterance/thought reports are regularly interspersed in the narration. Lillo-Martin (in press) suggests that the use of a 1st person pronoun in RS segments is what crucially distinguishes quotative from non-quotative uses, the latter reporting actions from the perspective of the agent, not utterances or thoughts. Although this matter deserves further examination, I would like to defend the idea (cf. Quer 2005) that such distinctions, if necessary at all, cannot be reduced to the presence or absence of an overt 1st person pronoun: as soon as RS is activated, the whole array of properties of a displaced context are present, irrespective of whether a 1st person pronoun is present or not.7 Beyond the lack of a lexical introducer, these instances of so-called nonquotative RS display the same properties with respect to pronominal interpretation, but also with respect to locative and temporal indexicals such as HERE or YEAR-THIS. Although more crosslinguistic work is needed on this specific group of data, they turn out to be crucial for the overall account of context shifting in SLs. Here I have put aside from the discussion cases of so-called constructed action (Metzger 1995), which frequently cooccurs with RS in narrative discourse. Within constructed action, the signer adopts the role of the referent in order to 7. An alternative possibility that needs to be explored is whether cases of so-called non-quotative RS (or constructued action in the sense of Lillo-Martin in press) share some core property with free indirect discourse (represented speech in Smith’s 2009 terms). However, I am skeptical about this link, as pronominal reference (and the phenomena related to it) in alleged non-quotative RS is identical to the one in quotative RS.

Reporting and Quoting in Signed Discourse

287

reproduce not his/her linguistic discourse, but his/her actions, postures or gestures in a more or less imitative fashion. This aspect is a poorly studied one, although it is found regularly in the descriptions of narrative techniques in sign languages (see Quinto-Pozos 2007). An example of this is presented in Figures (15)-(17), from Aesop’s fable “The lion and the mouse”: the lion’s roaring is first mimicked with a mouth gesture and facial expression (Fig. 15) and then reported with a classifier construction for the opening of the mouth (Fig. 16) and the lexical sign SOUND (Fig. 17).

Fig. 15. Roaring gesture.

Fig. 16. CL-open-mouth.

Fig. 17. SOUND.

These are the most complex cases to account for, as they require teasing apart what is gestural from what is linguistic in a RS segment. As a consequence of the language modality, both regularly coexist, either simultaneously or consecutively. Although sometimes the limiting line between the two sorts of elements is hard to draw and specific research is seriously needed, I would like to defend that it exists. The main conclusion of this brief characterisation of RS in LSs is that irrespective of whether RS is introduced or not by an attitude predicate, SLs display finer distinctions in RS between its use in direct quotes and non-direct reported discourse (next to indirect reported discourse without RS). In the next section

288

Josep Quer

we will try to put some of the attested facts on deictics into the perspective of indexical interpretation in language in general. 4.

Indexicals that shift

The received view on indexical expressions is that elements such as 1st and 2nd person pronouns, temporal and locative deictics are directly referential, following the basic approach of Kaplan (1989). This characterization has been summarized by Schlenker (2003: 29) in the Fixity Thesis reproduced under (14). (14)

Fixity Thesis (a corollary of Direct Reference) The semantic value of an indexical is fixed solely by the context of the actual speech act, and cannot be affected by any logical operators.

Although operators are in principle conceivable that could shift the context of evaluation of an indexical, Kaplan excludes this possibility and calls them ‘monsters’. At face value, this view seems to account quite accurately for indexical interpretation in a language like English. However, Schlenker (2003) argues that such monsters do exist and are instantiated in certain languages by attitude predicates. An example of such a shifted indexical would be represented in the following example in Amharic, where the 1st person in the scope of ‘say’ does not refer to the actual utterer but to John, the reported utterer (Schlenker 2003: 68): (15)

Situation: John says: ‘I am a hero’ jon j@gna n@-ññ yil-all John hero be.PF-1sO 3M.say-AUX.3M ‘Johni says that hei is a hero.’ (Lit.: ‘Johni says that Ii am a hero.’)

(Amharic)

From a crosslinguistic point of view, this is not an isolated case. Languages like Havyaka Kannada (Dravidian), for example, use the same set of pronouns for denoting actual and reported speech act participants. As a consequence, the reference of the embedded 1st person pronoun can be anchored to the reported context of utterance (16i) or to the matrix context (16ii) (Bhat 2004: 58): (16)

en-na ello:ru-de hogaïuttavu he: ¸li ra:ju enna-tre me.ACC all.EMPH praise that Raju me-with he:liddã tell.PERF

Reporting and Quoting in Signed Discourse

289

(i) Raju1 has told me2 : “Everybody praises me1 .” (ii) Raju1 has told me2 that everybody praises me2 . What this example shows is that the semantic value of the same linguistic expression, the1st person pronoun in the reported proposition, is not unambiguosly determined by the actual context of utterance. Recent work (Anand & Nevins 2004, Schlenker 2003, Speas 1999) has extensively shown that the Kaplanian analysis of indexicals in the scope of attitude reports is challenged empirically by languages like Amharic, Navajo, Slave or Zazaki, where 1st person pronouns embedded under a verb of saying, for instance, can corefer with the matrix clause subject (the attitude holder or reported agent), and not necessarily with the actual utterer, as happens in English. In view of this sort of facts, the definition of indexical expression should arguably be made more precise. According to Schlenker (2003), “an expression qualifies as indexical if its semantic value is determined by some feature of the context of utterance” (Schlenker 2003: 31). For instance, Amharic ‘I’ in (15) above qualifies as a strict indexical, as it must refer to the speaker of some context, although not necessarily the context of the actual speech act. It differs in that respect from logophoric pronouns, which are only grammatical in embedded contexts. Schlenker implements his proposal in an extensional semantics, where attitude verbs are quantifiers over contexts of thought or speech and may bind free context variables. The simplified representation of this view can be found under (17) for example (15): ci stands for the context of the reported speech act, and c∗ for the context of the actual utterance. (17)

SAY ci be-a-hero (agent(ci ), time(ci ), world(ci ))

Crosslinguistic variation in the shifting possibilities of indexicals is made dependent on whether the denotations of particular indexicals have free context variables or not. In the case under examination, Amharic ‘I’ would be lexically underspecified for its context variable (18b), as opposed to its English counterpart (18a): (18)

a. b.

English ‘I’: [[ I ]] = agent(c∗ ) Amharic ‘I’: [[ I ]] = agent(c), c an underspecified context variable

It is further argued by Schlenker that, unlike 1st and 2nd person pronouns, temporal adverbials in English such as two days ago shift optionally, as in example (19): only the shifted reading of the temporal expression would yield a felicitous result (19b), while the temporal expression the day before yesterday, which is

290

Josep Quer

indexical to the actual context, results in infelicity (19a). According to him, this constitutes an argument in favour of treating attitude verbs as quantifiers over contexts rather than as context-shifting modal operators that overwrite all the contextual variables (for the opposite view, though, see Anand & Nevins 2004, who treat this expression as anaphoric). (19)

John has told me repeatedly over the years: ‘I was sick two days ago.’ a. # John has told me repeatedly over the years that he was sick the day before yesterday. b. John has told me repeatedly over the years that he was sick two days ago.

On the other hand, under this account logophoric pronouns would be indexicals that can never be dependent on the actual context of utterance, as represented in (20). (20)

Logophoric pronoun: + indexical, −c∗

Mupun (Chadic) (Frajzyngier 1993) instatiates the case of 1st and 2nd person logophoric pronouns as characterized by Schlenker. They are always anchored in the derived context (21b)–(22b), as opposed to the non-logophoric ones (21a)– (22a), which refer to the actual discourse participants. (21)

a.

b.

(22)

a.

b.

wu sat n@ n-nas wur 3MSg say that beat.1Sg 3MSg ‘He said that I beat him.’ wu sat n@ di nas an 3MSg say that Log1MSg beat 1Sg ‘He1 said that he1 beat me.’ n-sat n-wur n@ wur ji 1Sg-say to-3Sg that 3Sg come ‘I told him1 that he2 should come.’ n-sat n-wur n@ gwar ji 1Sg-say to-3Sg that Log2 come ‘I told him1 that he1 should come.’

These are the essential features of the framework I am going to assume for the analysis of indexical behaviour in RS in SLs. However, nothing crucial hinges on this particular choice, and probably other approaches that can accommodate the phenomenon of indexical shift and the basic crosslinguistic facts related to it could be adopted.

Reporting and Quoting in Signed Discourse

5.

291

SL indexicals in role shift

5.1. Shifted second person reference The published data on RS in different SLs indicates that pronominal reference in reportive contexts behave much as Amharic 1st person pronoun, that is, it is not indexical to the actual utterer, but to the individual to whom the reported attitude is ascribed. Nevertheless, 2nd person reference unsurprisingly shows shifty behaviour as well. In the LSC examples under (23) and (24) this fact is illustrated through verbal agreement: 2nd person on the unbound agreement marker in (23) and on the agreeing lexical verb in (24).8 (23)

RS-i YESTERDAY ANNAi IX-3a 3a-TELL-3b PEDRO IX-1i ANGRY AGR-2 ‘Yesterday Anna told Pedro that she was angry at him.’

(24)

RS-i YESTERDAY ANNAi IX-3a 3a-TELL-1 IX-1i 1-HELP-2 ‘Yesterday Anna told me that she would help me.’

The occurrence in these examples (23) and (24) of 2nd person morphology linked to parameters of the embedded context and not to the actual context of utterance already makes clear that shifted reference of pronouns in RS is not limited to 1st person, as Zucchi’s (2004) seems to imply for LIS. 5.2. Non-shifted interpretation of indexicals in RS Despite the general tendency for indexical pronouns to shift reference within RS, non-shifted (or back-shifted) interpretations of 1st /2nd pronouns in the scope of RS have been also reported for SLs. One such example is (25), taken from Engberg-Pedersen (1995): in a RS stretch of discourse reporting her mother’s signing, the utterer uses 1st person pronouns (regular and possessive pronouns) to refer to herself, where a 2nd person would have apperared in direct discourse. This amounts to picking up the reference of the 1st person pronoun from the actual context of utterance within the domain of a reported context, as indicated by the RS nonmanuals (Engberg-Pedersen 1995). 8. Most of the SLs documented to date group their verbal lexical items in three main categories according to their behaviour with respect to agreement: (i) plain verbs, which do not agree; (ii) agreeing verbs, which display agreement with subject and/or object, and (iii) spatial verbs, which agree with their locative arguments (see Padden 1988). Some languages have additional means to show agreement with plain verbs, like agreement auxiliary predicates. The AGR sign in (23) is such a case in LSC.

292

Josep Quer RS-mother neg

(25)

IX-1 MOTHER FATHER HOME AGAIN / IX-1

(DSL)

‘… that my mother and father would go home again, not me.’ There are even more complex cases that involve pronouns which access both the reportive and the actual context: in the LSC sentence (26) the dual 1st person pronoun TWO-OF-US, the 2nd person included in the dual pronoun is ambiguous between the actual addressee of the utterance or the reported addressee. In the former case, we actually have an instance of “mixed” indexicality. However, due to the complexity derived from additional factors in plural pronouns, I leave such cases out of consideration here. (26)

RS-i ANNAi IX-3 3-TELL-2 TWO-OF-USi+2 WIN AT-LAST

‘Anna told you that the two of you had won at last.’ A legitimate question to ask at this point is whether we are actually dealing with the same series of pronouns in RS and non-RS contexts. One could in principle argue that the coarticulation of a manual pronoun with the specific nonmanual markings of RS is actually the realization of a distinct series of pronouns of the logophoric type. This position, though, cannot be sustained mainly for two reasons. First, the nonmanual morphology associated to RS marking is coarticulated with the whole stretch of reported discourse/thought, and not only with the pronominal form. Second, some indexicals other than pronouns can show the same shifty behaviour and that would mean that we systematically have two series of indexicals, one for those referring directly to the main utterance context and another one for those referring to the parameters of the shifted context. Actually, such an assumption would create an even bigger problem, since it entails that we should potentially have two parallel series of lexical items for the whole lexicon (one with RS nonmanual marking and another one without it). For these reasons, I reject this possibility. Let’s examine now some data making the second objection clearer. Although the RS marker extends over the whole reported proposition, not all indexicals need to be interpreted in the reported context. Locative indexicals like HERE do not shift obligatorily in the scope of an attitude predicate such as SAY, and the default interpretation is the one that links them to the spatiotemporal parameters of the main context of utterance. Consider the LSC in (7), repeated here as (27) for convenience: while the 1st person pronoun in the reported thought is interpreted as coreferential with the attitude holder JOAN, the locative indexical HERE refers most naturally to the context of utterance

Reporting and Quoting in Signed Discourse

293

(Barcelona), and not to the derived context where the locative parameter is explicitly fixed (Madrid). (27)

t RS-i IXa MADRIDm MOMENT JOANi THINK IX-1i STUDY FINISH HEREb

‘When he was in Madrid, Joan thought he would finish his study in Barcelona.’ However, if the indexical HERE is further specified, as in the minimally modified sentence (28) (=(2)), it can end up referring to the other location, Madrid. (28)

t RS-i IXa MADRID JOANi THINK IX-1i STUDY FINISH HERE MADRID

‘When he has in Madrid, Joan thought he would finish his studies there in Madrid.’ The fact that one and the same indexical (HERE) with associated RS morphology can receive both interpretations argues against the possible alternative mentioned above that would resort to two different series of indexical elements in the lexicon. Not all temporal and locative indexicals, though, behave in the same way. One of them, NOW in (29), does not permit shifted reference to the embedded context for some speakers. YEAR THIS in (30), though, seems to be able to access both contexts. This must be attributed to lexical differences between indexical expressions: in a subset of cases, we do find strict indexicality to the main context parameters, while for other indexicals both contexts are accessible. (29)

t RS-i LAST-YEAR JOANi THINK IX-1i STUDY FINISH NOW

’Last year, Joan thought he would finish his studies {now/%then}.’ (30)

t RS-i LAST-YEAR JOANi IX-3 THINK IX-1i STUDY FINISH YEAR THIS

’Last year, Joan thought he would finish his studies {this year/then-that year}.’ On the basis of Navajo data displaying Direct Discourse Complements, Speas (1999) argues for a split between the system determining deixis for person marking (functional) and the system determining deixis more generally (semantic). In view of the data discussed so far in this paper, we must conclude that such a clear-cut divide does not hold for the SLs at hand, despite the parallelism with some of the Navajo facts. What we have been able to show is that

294

Josep Quer

shifted reference of indexicals is not something specific to a particular context parameter. The proposals made for the set of data with shifting indexicals in certain spoken languages like Amharic could straightforwardly tackle the SL examples of RS such as (1)–(3) with an introducing reportive/attitude predicate, that is, instances of so-called quotative RS. However, the same analyses are faced with an additional problem in instances of alleged non-quotative RS like (11)–(13): no attitude verb is present in the structure in order to license the shifted reading of indexicals, a scenario which is explicitly excluded in Schlenker (2003: 69). In the next section I will sketch an approach that tries to solve the problem without giving up the insights of previous analyses. 6.

Proposal: Point of view operator

With the aim to provide a unified account of both introduced and unintroduced instances of RS, I follow the basic insight in Lillo-Martin (1995), where she proposes that a sentence like (31) (=(12)) involves a covert Point of View Predicate (POV). The relevant part of the structure is depicted in (32): the covert POV predicate selects an embedded CP and binds the operator in its Spec. In turn, this operator binds the 1st person pronoun in the RS complement (Lillo-Martin 1995). (31)

RS-i MOMi IX-1i BUSY

(ASL)

‘Mom’s like, I’m busy!’ IP

(32)

I’

NPi MOM

I

VP V POV

CP C’

Spec Opi

C

IP

IX-1i BUSY

Reporting and Quoting in Signed Discourse

295

Building on Lillo-Martin’s (1995) analysis, I suggest that the type of languages we are looking at instantiate a Point of View Operator (PVOp), rather than a Point of View Predicate. This operator materializes in RS nonmanual morphology and accounts for the attested shifted interpretations of indexicals in its scope (1st and 2nd person pronouns, time and locative indexicals). Spreading of nonmanuals over the c-command domain of an Operator has been argued to exist in ASL for other operators such as Wh, Q or Neg (Neidle et al. 2000), whether they are overt or covert: furrowed eyebrows associated with the Wh-Operator in (33) and headshake associated with the Neg Operator in (34) (Neidle et al. 2000). (33)

wh LOVE JOHN WHO

(ASL)

‘Who loves John?’ (34)

hs JOHN BUY HOUSE

‘John didn’t buy a house.’ Unlike Lillo-Martin, though, I argue that PVOp is not a covert reportive/attitude predicate taking a subordinate CP (the reported proposition), but a covert operator over contexts (à-la-Schlenker) sitting in a very high projection of the functional structure of the clause (cf. Cinque 1999 on the expanded left periphery of the clause). For simplicity, I will assume that this projection is CP in order to account for a number of related facts.9 The PVOp would occupy the head of this projection and determine the morphological and interpretive properties of the RS structure within its c-command domain. If such an analysis is on the right track, it is able to unify both the quotational and non-quotational instances of RS, as it does not link context shift to an overt attitude predicate. However, in cases of lexically introduced RS, it remains to be determined how the overt attitude predicate interacts with the empty PVOp, as both seem to fulfill the same function in the analysis. For a minimal variant of example (31) with lexically introduced RS, we would have to posit a structure like the one under (35). The extra assumption required would be that the two attitude operators compose semantically as a result of the incorporation of PVOp into the lexical verb. At this point, though, the details and consequences of such a proposal remain to be worked out. 9. With this I reject the proposal to adopt a Speech Act Phrase hosting POVOp that was mentioned in Quer (2005) on the basis of Speas (1999), Speas & Tenny (2003) and Speas(2004). For a critical evaluation of the latter, see Gärtner & Steinbach (2005). Nothing important hinges on this modification.

296

Josep Quer

(35)

IP NPi MOM

(ASL) I’

I

VP V SAY

CP Spec

C’ C°

PVOp

IP

IX-1i BUSY

In contrast to other analyses, the one sketched here does not reduce the issue of RS to the interpretation of 1st person pronoun (Zucchi 2004), since other contextual variables display a comparable pattern of behaviour. It offers strong confirmation for the idea that context variables (author and addressee, time, location) in a derived context can be unselectively bound independently from each other by an attitude operator or identified with the value of the parameters in the main context of utterance. 7.

Some consequences

7.1. Independent shift of indexicals On the basis of the SL data discussed here, one is forced to relativize the crosslinguistic validity of the “Shift-Together Constraint” (see (36)) proposed in Anand & Nevins (2004) for Slave and Zazaki (Anand & Nevins 2004). (36)

Shift-Together Constraint Shiftable indexicals must shift together.

As we have seen, when indexical pronouns (1st and 2nd person) shift under RS, locative and temporal indexicals can still refer to the actual context of utterance, which appears to instantiate direct deixis in their unmarked interpretation. In addition, under appropriate circumstances some of those locative and tempo-

Reporting and Quoting in Signed Discourse

297

ral indexicals can shift, too, with some exceptions that must be derived from their lexical specification. Moreover, it is not clear either that all pronouns in a sentence must shift together, as we find examples of shifted 1st person next to a non-shifted 2nd person in a reported context (see (26) above). Such cases deserve further investigation before they can be consistently incorporated into the general picture. 7.2. Quantifier binding of shifted pronouns Furthermore, additional support can be offered for Schlenker’s (2003) binding analysis of contextual variables with fresh evidence from SL showing quantifier bound readings of a shifted first person prononun within the scope of RS. The LSC instances of this are the following ones:

(37)

RS-i t eg:1 eg:front PUPIL ALLi THINKˆSEE.refl IX-1i INTELLIGENT SUPERLATIVE

‘Every pupil thinks that he is the most intelligent.’ (38)

RS-i NOONEi SAY IX-1i AGR-1 SCARED DARKNESS

‘Noone says he’s scared of darkness.’ A comparable case was independently observed for Abe by Koopman and Sportiche (1989): in (39) a 3rd person referential pronoun (akin to the behaviour of 1st and 2nd person pronouns) can be bound by the main quantificational NP subject. Their hypothesis is that this is made possible by the occurrence of the complementizer kO. Similarly, for the SL cases in (37)–(38) I would like to tentatively suggest that it is the covert PVOp what mediates in the bound reading of the embedded subject pronoun (Koopman & Sportiche 1989: 584). (39)

ni ye SE apoOUNi e hE kO nobody Neg said Comp he is handsome

Note in passing that the SL examples discussed in this subsection constitute strong support for the idea that pronominal indices are actual pronouns, and not just pointing/mostly gestural expressions, as has been often defended in the literature, most prominently by Liddell (see for instance Liddell 2000 for an overview of this position; in favour of the linguistic status of pronouns, with an argument from RS, see Meier 1990).10 10. This argument is more forceful if pronouns are bound by a negative quantifier in the main clause or by a second person plural subject. The LSC data elicited such as (38)

298

Josep Quer

7.3. Is PVOp a “monster”? As a consequence of the overall discussion of SL data in this paper, one can defend that Kaplanian indexical “monsters” do exist in SLs. The PVOp we find in SLs instantiate such a monster, as shiftable indexicals in its scope are not “directly referential” to the main context of utterance. Zucchi (2004) offers a unified analysis of the quotational and non-quotational RS in terms of the presupposition associated with the 1st person morphology in RS, namely that the 1st person is coindexed with another term in the discourse other than the utterer. However, this might be a simplification, because we have already seen that the referential shift in these constructions can also affect 2nd person pronouns/agreement, locative and time indexicals, and sometimes independently of each other when they cooccur. 8.

Concluding remarks

Indexical shift seems to be pervasive within what is known as RS, which is present in most if not all SLs described to a bigger o smaller degree. It has been mostly attested for pronominal reference shift (and corresponding verb agreement), but there is little discussion in the literature about other indexicals like temporal and locative deictic expressions. The phenomenon of indexical shift in reportive/attitude contexts described in less familiar spoken languages seems to be pervasive in SLs. The main points of the discussion can be summarized as follows: (i)

RS in SL has properties of both direct and non-direct reported discourse, as in several spoken languages. (ii) The interpretive and morphological properties of RS can be derived from a Point of View Operator. (iii) Quotational and non-quotational instances of RS can be accounted for in a unified fashion, as in Zucchi (2004). (iv) Indexicals in RS (1st 2nd person pronouns, locative and time indexicals) can all shift under appropriate conditions. This chapter constitutes just an attempt at addressing in SL the same kinds of questions raised in the semantics and pragmatics literature on indexical reference in reportive contexts. As we have seen, despite the effects of the visualgestural modality, some of the the “uncommon” facts described for certain seem to confirm this prediction so far. Thanks to Gennaro Chierchia for making this point.

Reporting and Quoting in Signed Discourse

299

spoken languages are replicated in SLs. This allows us to evaluate the empirical crosslinguistic validity of concrete aspects of recent proposals on indexical shifting. One question, though, that I have only hinted at in section 6.2 is the a priori possible modality effect in shifting reference in reportive contexts. As mentioned there, it has been claimed that SLs do not have actual pronouns and that what we have been calling pronouns here must be reduced, at least in part, to pointing or indicating gestures. Some of the phenomena discussed here obviously argue against such a position. Still, there remain quite important issues to be addressed in this domain, such as the relation between actual deixis and grammaticalized deixis in SLs, but these are questions that must await future work. References Anand, Pranav & Nevins, Andrew 2004 Shifty operators in changing contexts. In Proceedings of Semantics and Linguistic Theory 14, R. Young (ed.), 20–37. Ithaca: CLC Publications. Berenz, Norine 2002

Insights into person deixis. Sign Language & Linguistics 5: 203–227.

Bhat, D.N.S. 2004

Pronouns. Oxford: Oxford University Press.

Cinque, Guglielmo 1999 Adverbs and Functional Heads. A Cross-linguistic Perspective. Oxford: Oxford University Press. Engberg-Pedersen, Elisabeth 1995 Point of view expressed through shifters. In Language, Gesture, and Space, K. Emmorey and J.S. Reilly (eds.), 133–154. Hillsdalle, NJ: Lawrence Erlbaum Associates. Frajzyngier, Zygmunt 1993 A Grammar of Mupun. Berlin: Dietrich Reimer Verlag. Gärtner, Hans-Martin & Steinbach, Markus 2005 A skeptical note on the syntax of speech acts and point of view. In Form, Structure, and Grammar. A Festschrift Presented to Günther Grewendorf on Occasion of his 60th Birthday, P. Brandt and E. Fuß (eds.), 313–322. Berlin: Akademie Verlag. Herrmann, Annika & Steinbach, Markus 2007 Wenn ,ich‘ nicht ich ist: Redewiedergabe in Gebärdensprachen. In Zitat und Bedeutung. Linguistische Berichte Sonderhefte 15, E. Bren-

300

Josep Quer del, J. Meibauer and M. Steinbach (eds.), 153–179. Hamburg: Helmut Buske Verlag.

Herrmann, Annika & Steinbach, Markus 2009 Visible context shifts: Quotation in Sign Languages. Talk presented at the SLE 42nd Annual Meeting of the Universidade de Lisboa, Faculdade de Letras, Portugal: Global Languages Local Languages. Lisboa. Herrmann, Annika & Steinbach, Markus 2010 Eine neue Perspektive auf Role Shift in Deutscher Gebärdensprache (DGS). Perspektivwechsel als nichtmanuelles Kongruenzphänomen. Das Zeichen 84: 112–119. Kaplan, David 1989

Demonstratives. In Themes from Kaplan, J. Almog et al. (eds.), 481– 563. New York: Oxford University Press.

Koopman, Hilda & Sportiche, Dominique 1989 Pronouns, logical variables, and logophoricity in abe. Linguistic Inquiry 20: 555–588. Lee, Robert G. et al. 1997 Role shift in ASL: A syntactic look at direct speech. In Syntactic structure and discourse function: An examination of two constructions in ASL, C. Neidle, D. MacLaughlin and R. G. Lee (eds.), 24–45. Boston University (Rep. No. 4), American Sign Language Linguistic Research Project. Liddell, Scott K. 2000 Indicating verbs and pronouns: Pointing away from agreement. In The Signs of Language Revisited, K. Emmorey and H. Lane (eds.), 303– 320. Mahwah, NJ: Lawrence Erlbaum Associates. Lillo-Martin, Diane 1995 The point of view predicate in American sign language. In Language, Gesture & Space, K. Emmorey and J.S. Reilly (eds.), 155–170. Hillsdalle, NJ: Lawrence Erlbaum Associates. Lillo-Martin, Diane in press Utterance reports and constructed action in sign and spoken languages. In Sign Language. An International Handbook, R. Pfau, M. Steinbach and B. Woll (eds.). Berlin: Mouton. Meier, Richard P. 1990 Person deixis in American sign language. In Theoretical Issues in Sign Language Research, Vol. 1: Linguistics, S.D. Fischer and P. Siple (eds.), 175–190. Chicago: University of Chicago Press.

Reporting and Quoting in Signed Discourse

301

Metzger, Melanie 1995 Constructed dialogue and constructed action in American sign language. In Sociolinguistics in Deaf Communities, C. Lucas (ed.), 255– 271. Washington, D.C: Gallaudet University Press. Neidle, Carol et al. 2000 The Syntax of American Sign Language. Functional Categories and Hierarchical Structure. Cambridge, MA: MIT Press. Padden, Carol 1986

Padden, Carol 1988

Verbs and role-shifting in American sign language. In: Proceedings of the Fourth National Symposium on Sign Language Research and Teaching, C. Padden (ed.), 44–57. Silver Spring, MD: National Association of the Deaf. Interaction of Morphology and Syntax in American Sign Language. New York: Garland Publishing.

Poulin, Christine 1994 Null arguments and referential shift in American sign language. MIT Working Papers in Linguistics 23: 267–281. Poulin, Christine & Miller, Christopher 1995 On narrative discourse and point of view in Quebec sign language. In Language, Gesture & Space, K. Emmorey and J.S. Reilly (eds.), 117–131. Hillsdalle, NJ: Lawrence Erlbaum Associates. Quer, Josep 2005

Context shift and indexical variables in sign languages. In Proceedings from Semantics and Linguistic Theory 15, E. Georgala and J. Howell (eds.), 152–168. Ithaca, NY: CLC Publications.

Quer, Josep & Frigola, Santiago 2006 The workings of indexicals in role shift structures in Catalan sign language (LSC). Actes del 7è Congrés de Lingüística General, Universitat de Barcelona. CD-ROM. Quinto-Pozos, David 2007 Can constructed action be considered obligatory? Lingua 117(7): 1285–1314. Schlenker, Philippe 2003 A plea for monsters. Linguistics & Philosophy 26: 29–120. Smith, Carlota S. 2009 Accounting for subjectivity (point of view). In Text, Time and Context, R.P. Meier et al. (eds.), 371–393. Heidelberg: Springer.

302

Josep Quer

Speas, Margaret 2000 Person and point of view in Navajo verbs. In MIT Working Papers on Endangered and Less Familiar Languages, Vol. 1, A. Carnie, E. Jelinek and M. Willie (ed.). Cambridge, MA: MIT Press. Speas, Margaret 2004 Evidentiality, logophoricity and the syntactic representation of pragmatic features. Lingua 114: 255–276. Speas, Margaret & Tenny, Carol 2003 Configurational properties of point of view roles. In Asymmetry in Grammar, A.M. di Sciullo (ed.), 315–344. Amsterdam: John Benjamins. Zucchi, Alessandro 2004 Monsters in the visual mode? Unpublished manuscript, Università degli Studi di Milano.

The Act of Quotation Paul Saka

1.

Introduction

I focus on one approach to understanding quotation, the identity theory; I delineate varieties thereof; and I cite some considerations for favoring a speech-act version. Along the way we shall see how the study of quotation can illuminate the general conflict between speech-act semantics and formal semantics, and we shall see fresh arguments for insisting that the mechanism of quotation is referentially indeterminate. The literature on quotation can be confusing because a number of conflicting views have title to “the identity theory”. In this article my aim is to delineate different varieties of the identity theory, to cite some considerations for favoring a particular speech-act version, and to use the study of quotation to illuminate the general conflict between speech-act semantics and formal semantics. More specifically, § 2 characterizes the difference between formal and speechact theories of quotation, § 3 argues that formal theories are overly determinate, § 4 finds insight by taking a speech-act perspective, § 5 argues that, to the extent that formal theories of quotation might be true, they are grounded in speech-act theories, and § 6 suggests that the referent indeterminacy of quotation is most plausibly explained by a speech-act approach. Throughout, I clarify, correct, and add new defenses to the theory of quotation as sketched in Saka 1998. Before beginning, however, I wish to establish some terminology. By quotation, I mean that which consists of some quotatum (quoted matter) combined with some mark of quotation, for instance a flanking pair of double apostrophes. In the example below, the entire line makes up quotation (a) while the word by itself, excluding the punctuation marks, is (a)’s quoted matter, (b). (a)    “octopus”    (b)

304 2.

Paul Saka

Formal semantics versus speech-act theory

The idea of identity can be used, in a theory of quotation, in a number of different ways. Most especially, it is useful to distinguish between formalist identity theories and speech-act identity theories. For the formalist, the object of linguistic study is a linguistic form. It is a configuration of soundwaves, inkmarks, or some abstraction thereof; it is, in a word, a word (or more pedantically, a morpheme, word, phrase, sentence, or other reified linguistic entity). For the speech-act theorist, the object of linguistic study is a communicative event, implicating a speaker-hearer dyad, or at least a cognitive event, implicating an intentional agent in relation to words. According to the speech-act approach, either it is the speaker that refers or it is the act of expression that refers. According to the formal approach, words and other linguistic entities refer, i.e. the object of one’s expression refers. (For more on the indeterminacy between expression act and expression object, or between instance of expressing and that which is expressed, see Saka 2010.) Most frequently, however, “expression” is used in the object sense, and that is how I shall use it in the rest of this paper. I begin by cataloguing the logical space of formalist versions of the identity theory – those that are couched in terms of word reference or denotation. Some Formalist Theses (F1) universal identity: for any expression x, x denotes x. Thus, in the octopus example above, (a) denotes (a), and (b) denotes (b).1 (F2) quotation identity: for any quotation x, x denotes x. Thus (a) denotes (a), but (b) does not necessarily denote (b). (F3) quotatum identity: for any quoted matter x, x denotes x; i.e. expressions denote themselves when inside of quotation marks. Thus (b) denotes (b), though (a) does not necessarily denote (a). (F4) partial identity: for any quotation x, x denotes one part of x, namely the quotation denotes the quoted matter. Thus (a) denotes (b).

1. Does (a) really constitute a linguistic expression? There are two concerns here. One, it may be questioned whether the written form (a) corresponds to any spoken form, and hence whether it counts as language (Saka 2012). Here I shall simply assume that quotation marks are part of language. Two, demonstrative theorists deny that quotations form a linguistic unit. This is doubly irrelevant, however. Aside from the fact that (F1) does not purport to express the view of the demonstrative theory, in the intended sense of the term an expression need not be a unit: “I see you” contains within it the expression “I see”, even though the latter is not a syntactic constituent.

The Act of Quotation

305

For purposes of comparison, it is worth looking at some approaches that do not involve identity. According to the paratactic demonstrative theory, for any quoted matter x, x is denoted by a pair of quotation marks that are not only distinct from x but entirely disjoint from x, indeed not even in the same sentence as x (Davidson 1979, Garcia-Carpintero 1994, 2005, Cappelen and Lepore 1997, 2005, Washington and Biro 2001, Caplan 2002, Benbaji 2004, McCullagh 2007, Predelli 2008, Sorensen 2008). According to this view, (1) is equivalent to (1’): (1) (1’)

“Lobster” is a word. Lobster. That is a word.

Another non-identity approach is that of Searle (1969), for which quoted matter is not referred to (hence, for which there is no identity between quoted matter and its referring term). Then too, for the ellipsis theory of Wertheimer 1999, (2) is elliptical for (2’): (2) (2’)

“Chortle” is a verb. The word, “chortle”, is a verb.

On Wertheimer’s view the referring term is “the word” or some other implicit descriptor; hence there is no identity or even particular resemblance between it and its referent. But back to the identity theory. To my knowledge, no one advocates Universal Identity (F1) and no one advocates Quotation Identity (F2). One problem with Universal Identity is that, although it explains the semantic values of (a) and (b), it does not explain the semantic value of the quotation marks. In addition, (F1) makes the identity theory circular, for it explains quotation in terms of quotation, which we cannot understand until we already understand what is being explained. The same kind of circularity appears in Quotation Identity. Moreover, (F1) and (F2) fail to explain how the following can be true: “cat” consists of three characters. For they hold that what gets denoted by a quotation is the whole quotation itself, which in this case consists of five characters. (Sorensen 2008: 58 regards this as a problem for the identity theory tout court because he does not distinguish the straw theses [F1] and [F2] from the real theses [F3] and [F4].) Whereas (F1) and (F2) explain quotation in terms of quotation, the other identity theories non-circularly explain quotation in terms of quoted matter. For Quotatum Identity (F3), if an expression is inside of quotation marks then it denotes itself. Proponents of (F3) include Frege (1980), Tajtelbaum (1957), Washington (1992), Reimer (1996), Akiba (2005), and Harnish (2007). Al-

306

Paul Saka

though these works vary in terminology, they appear to share a commitment to formal semantics: quoted words “designate words of the other speaker,” says Frege (1980); “a quoted expression invariably mentions itself,” says Washington (1992: 605); “quotationally embedded expressions refer to themselves,” says Reimer (1996: 132). Partial Identity (F4) is even more popular, being backed by Belnap and Grover (1973), Haack (1974), Parsons (1982), Richard (1986), Ludwig and Ray (1998), Saka (1998), Pietroski (1999), Gomez-Torrente (2001), Botterell and Stainton (2005), and now Cappelen and Lepore (2007). For these writers, quotation marks signify a kind of identity function. But instead of acting on the semantic value of its argument x, mapping from |x| to |x|, quotation marks operate on the argument expression itself, mapping from x to x: “the result of writing an expression in quotation marks, has as value the quoted expression,” says Haack (1974: 288); “the customary reference of [quotation marks] is that function which maps each and every phrase of the language to itself,” says Parsons (1982: 320); quotation marks “denote a function from a symbol to itself,” say Botterell and Stainton (2005: 22); “semantically, a pair of quote marks is a concept or intension, quot, which ambiguously or indeterminately maps its argument expression x into some linguistic item saliently associated with x,” says Saka (1998: 127).2 The formalist theses have speech-act counterparts that are couched in terms of rules for speaker reference. Some Basic Speech-Act Theses (S1) universal identity: for any expression x, if a speaker S uses x then S may refer to x. Thus S may use (a) to refer to (a), and may use (b) to refer to (b). (S2) quotation identity: for any quotation x, if S uses x then S may refer to x. This rule licenses the use of (a) to refer to (a), but not the use of (b) to refer to (b). (S3)

quotatum identity: for any quoted matter x, if S uses x then S may refer to x. This rule licenses the use of (b) to refer to (b), but not the use of (a) to refer to (a).

2. My statement is qualified by context: on the same page I state that denotational or “referentialist semantics is by no means essential to my account. I myself prefer the anti-realist ‘internal semantics”’ that is found in conceptual-role and speech-act theories.

The Act of Quotation

(S4)

307

partial identity: for any quotation x, if S uses x then S may refer to one part of x, namely to the quoted matter. This rule licenses the use of (a) to refer to (b).

As I read it, the literature on quotation says very little about speaker actions except for Saka (1998, 2006) and, in opposition, C&L (2007: 85). Several writers do imply a speech-act perspective, but only so as to leave doubt (Washington 1992, Reimer 1996, Recanati 2001, Gutzmann and Stei, this volume, Johnson, this volume). As one example, Reimer (1996) describes quotations as demonstrations, which can be understood as either acts or objects. The act interpretation may seem more charitable, but the objectual or formal interpretation is the only one that seems to occur to Reimer’s chief critic, Caplan (2002). As another example, Washington (1992: 583) speaks of quotational versus standard uses. If he has a use theory of meaning in mind then he is indeed invoking speech-acts, broadly understood; but if he uses “use” as lexicographers do, then he is simply saying that expressions are polysemous between use and mention (to use “use” in yet another way). The use-theoretic or speech-act interpretation again may seem more charitable, but this is contra-indicated by Washington’s assertion that quoted matter refers to itself. Not only does he speak of word-reference rather than speaker-reference, he deliberately chooses Quotatum Identity over Partial Identity, which makes sense only to formalists (see § 3 below). Formalist and speech-act semantics are widely regarded as compatible (e.g. Strawson 2001), and this may explain the equivocation that I detect in the work of Washington and others. Such compatibilism is explicitly argued for by Lycan (1999: 25), who responds to the slogan “Guns don’t kill, people do” by insisting that, when someone kills by using a gun, “there is a perfectly good sense in which the gun [as well as the gunner] did kill”. By analogy, the speech-act thesis “words don’t refer, people do” would appear to be a contradiction in terms. For to say that words do not refer is just shorthand for saying that people do not use words to refer. Yet Lycan’s argument is unpersuasive. That there is a reasonably good sense, as far as casual purposes go, in which guns do kill and words do refer, does not mean that there is a perfectly good sense in which they do. After all, there is a sense, adequate for most conversational purposes, in which time is absolute, but this does not make the concept of absolute time scientifically correct. The formalist and speech-act approaches are logically different, and moreover the rhetorical and metaphorical framing of an issue has cognitive effects and political fall-outs (Lakoff 2004, Nunberg 2006). We can see this already in “guns don’t kill”, and we can see it in the case at hand, as I shall try to explain in the rest of the paper.

308 3.

Paul Saka

Formal semantics and faux precision

The formalist theses of Quotatum Identity and Partial Identity are distinct: in one case the quotatum does the denoting, and in the other it is the quotation as a whole that does. The counterpart speech-act theses, however, are indistinguishable: in both cases it is the speaker who does the referring. Granted, (S3) and (S4) differ in that one quantifies over quoted matter where the other quantifies over quotation; but because quotation and quoted matter have been defined so that they always co-occur (p. 303), it is impossible to use one without using the other. Thus, (S3) and (S4) amount to two formulations of the same claim; they coalesce into a single thesis. Therefore formal semantics and speech-act semantics do not merely say the same thing in two different ways. From the speech-act perspective, Quotatum Identity and Partial Identity (S3) and (S4) are both right. At the same time, their formalist counterparts (F3) and (F4), because they parcel out denotation in determinate ways when determinacy does not actually obtain, are both mistaken. 4.

Conventional speech-acts

If the formalist theses really were shorthand for their speech-act counterparts, or if the speech-act theses were expansions of the formalist theses, the two idioms could be exactly translated into each other. Yet I do not think that that is possible. Below I consider a couple of proposals for intertranslation, and I eventually find fault with both. (A)

bridge principle a: Expression x denotes object o ≡ for some S, if S utters x then S refers to o (“water” denotes water ≡ for some S, if S utters “water” then S refers to water).

Principle A combines with the basic speech-act theses (S1–S4) to yield the formalist theses (F1–F4). Principle A, however, is a straw bridge that cannot be crossed. It fails because a speaker may use a term metaphorically or sarcastically or “referentially”, to use Donnellan’s unfortunate terminology.3 In such

3. Donnellan distinguishes between “attributive” and “referential” uses of definite descriptions; yet definite descriptions refer when used attributively and need not literally refer to what they are intended to when used “referentially”.

The Act of Quotation

309

cases a speaker uses x to refer to o even when x does not denote o (a poet can use “water” to refer to life without “water” denoting life). What unites metaphor, sarcasm, and Donnellanan reference is that all count as non-literal uses. For this reason a better bridge would invoke convention. (B)

bridge principle b: x denotes o ≡ for some S there is a convention whereby, if S utters x, then S may refer to o.4

Although Principle B cannot get us to the formalist theses (F) from the basic speech-act theses (S), it can get us to (F) if we revise our basic speech-act theses so that they too invoke convention. Some Conventional Speech-Act Theses (S1’) universal identity: there is a convention whereby, for any expression x, if S uses x then S may refer to x. (S2’) quotation identity: there is a convention whereby, for any quotation x, if S uses x then S may refer to x. (S3’) quotatum identity: there is a convention whereby, for any quoted matter x, if S uses x then S may refer to x. (S4’)

partial identity: there is a convention whereby, for any quotation x, if S uses x then S may refer to one part of x, namely to the quoted matter.

As in the case of basic (S3) and (S4), I do not see any practical difference between conventions (S3’) and (S4’). On the formalist approach, we saw that Universal Identity and Quotation Identity lack explanatory power. The speech-act approach, however, facilitates our seeing that they are in even worse position than that. Universal Identity and Quotation Identity – of the conventional speech-act variety and hence also the formalist variety – are outright false, as there is no convention by which speakers utter an expression in reference to that very expression.

4. Even better yet, I would prefer bridge principle c: (C)

Expression x denotes object o, relative to convention or community c ≡ for those who believe S is party to c, if S utters x then S refers to o.

Principle C makes denotation and reference both relative, and I believe it brings us closer to the truth. My discussion will center on Principle B, however; it is simpler than C, less controversial, and yet adequate for present purposes.

310 5.

Paul Saka

Non-conventional speech-acts

Reference does not always require convention. You can refer to an object, for instance, by pointedly gazing at it. I say that gazing can count as an act of reference because referring is nearly the same as drawing attention. When your primary physician refers you to specialists, he or she directs your attention to them; when an attorney says “I refer you to code 27B”, he or she directs the court’s attention to 27B; when I say “what a magnificent piece”, while gazing at your vase, I refer to your vase rather than your painting on the wall, and I do so by drawing special attention to it. (Provisos regarding intention and mutual intention might be added, but they would not change the structure of my argument.) When reference happens without convention, I shall call it natural. Natural reference is possible because physical context guides plausible interpretation, normal human psychology can be presupposed, speakers can draw rational Gricean inferences, and so forth. In particular, when interlocutor A utters x in the presence of B, A normally makes x salient for B. By uttering x, therefore, A draws B’s attention to x; and to draw attention, as I said, is (near enough) to refer. Thus, instead of developing the basic speech-act theses (S1-S4) in terms of convention (S1’-S4’), we can develop them in terms of natural capacity. Some Natural Speech-Act Theses (S1”) universal identity: due to our cognitive nature, for any expression x, if S uses x then S may refer to x. (S2”) quotation identity: due to our cognitive nature, for any quotation x, if S uses x then S may refer to x. (S3”) quotatum identity: due to our cognitive nature, for any quoted matter x, if S uses x then S may refer to x. (S4”) partial identity: due to our cognitive nature, for any quotation x, if S uses x then S may refer to one part of x, namely to the quoted matter. So long as we are speaking of speech-acts rather than denotation, and so long as we are speaking of natural rather than conventional reference, I believe that Universal Identity is valid. I would go further, however, and say that reference need not be verbal, and need not rely on strict identity. Direct ostension and deferred ostension make for the ur-principle upon which the very possibility of language use is founded. (0)

ostensive loose identity: due to our cognitive nature, for any entity x, if S makes x sufficiently salient then S may refer to something

The Act of Quotation

311

recognizably related to x. For instance, when x is a linguistic token, S may use x to refer to a whole class of related tokens. Thus an utterance of “joie” makes perceptually salient the very token uttered (it becomes very salient for any French audience and for anyone else able to discern the phonology of French, less salient for others); and by deferred ostension it makes the word-type salient as well. At the same time the utterance makes joy salient, for any French audience, due to convention. The notion of salience, it should be clear, does not presuppose the existence of an external world: dragons can be salient to you, in the sense that they are actively on your mind, despite their irreality. Principle (0) thus supports at least two kinds of reference. (U) (M)

use: S uses expression x ≡ S produces a token of x, intending to refer to x’s conventional denotation. mention: S mentions x ≡ S produces a token of x, intending to refer to some linguistic feature or another that is related to x.

Formula (M) represents a revision of Saka (2005: 199), where mention is characterized in terms of S’s referring to “something associated with x other than its extension [denotation]”. This is mistaken because uttering “water” in reference to life is a use, not a mention, of “water”. The revision here stipulatively includes, as examples of linguistic features of “water”, the form /waDr/, the content water, and the lexeme ; and it excludes literal extension or denotation, metaphorical extensions, and part of speech.5 Principle (M) subsumes all of the (S”) theses as special cases. For instance, suppose x is a quotation. Then in most contexts, for most interpreters, x is recognizably related to its quoted matter, from which (S4”) follows. Consequently it is not always necessary to know the conventional use of quotation marks in order to understand quotation. To say that knowledge of convention is not always necessary, however, is not to say that conventions don’t exist; to assert (S4”) isn’t necessarily to deny (S4’). Indeed, natural reference may actually promote conventional reference. 5. (M) actually requires additional revision. If you imagine (i) printed so that the first “exuberant” was pink and the second “exuberant” purple, then the quotations in (i) would signify chromatic types; yet chromatic types aren’t linguistic features. (i) No instance of “exuberant” is the same as any instance of “exuberant”. Perhaps (M) should specify the features of x that are neither identical to x’s conventional denotation nor cognitively generated therefrom, but spelling all this out lies beyond my present purview.

312

Paul Saka

As Strawson (2001: 112) writes, “Because it has worked, it becomes established; and then it works because it is established.” Kripke (1977: 182) adds: “what was originally a mere speaker’s reference may, if it becomes habitual in a community, evolve into a semantic reference.” Such dual force of nature and convention can be seen in onomatopoeia. If you are talking to someone who has never before heard English, in the right context you can say “woof” in order to refer to the sound of a dog, just as they can say “rau” to convey the same thing. Comprehension will be less speedy and reliable, however, than if you and your interlocutor were party to the same standing conventions for dog sounds. Hence a full account of English would specify “woof” as a lexical entry but not “rau”. A similar story applies to quotation, I believe: mentioning is completely natural, and to signal mentioning there is a fairly narrow range of devices, delimited by the nature of their job, that have coalesced into culturally variable conventions: single apostrophes in British English, angle brackets in German, corner brackets in Japanese, etc. It is my thesis that these conventions of modern literary language, each in its own way, harness the capacity for natural reference. For American English, the convention distinguishes between simple and iterated quotation, and involves other complications as well, but the basic idea can be stated briefly. (Q)

quotation: there is a convention whereby, if S uses an opening double apostrophe, then S signals that S is commencing to rely on (M); and if S uses a closing double apostrophe, then S signals that S is ceasing to rely on (M).

The interpretation of quotation is thus a two-step process. First, seeing quotation marks around expression x and relying on conventional knowledge as represented by (Q), the interpreter recognizes that S is pointedly engaging in natural reference. Second, relying on pragmatic knowledge (M), context, and common sense, the interpreter infers what S is referring to and why. Convention (Q) comes close to (S3”) and (S4”), which raises the question of connecting it to formalist identity (F3) or (F4) via some principle like Bridge B. I am reluctant to think this is wise, for a variety of reasons. First, as mentioned in § 3, I believe the formalist identity theses are overly determinate. Second, (Q) appeals, via (M) and thence to (0), to recognizability or recognition, which is too subjective and indeterminate for the truth-conditional presuppositions of the formalist approaches that I have been considering. Third, Bridge B is symmetric, equating denotational facts with facts about conventional speaker reference. If (Q) is correct, however, speaker reference is ontologically and explanatorily prior to denotation. Whereas (Q) explains why speakers are able to use quota-

The Act of Quotation

313

tions to refer to quoted matter, the formalist theses come across as brute stipulations. Fourth, (Q)’s formulation, in terms of signaling rather than denotation, frees us from standard notions of compositionality. This is necessary, I believe, because standard compositional theories fall to the fact that quotation marks sometimes preserve the customary syntax of an expression (3) and sometimes do not (4). (3) (4)

She “chortled”. “Chortled” is a word made up by Lewis Carroll.

Quotations are thus analogous to communication whereby you signal to your spouse that you are nearly out of milk by ostentatiously displaying the bottle: you can do that – or not – while simultaneously using the bottle for its customary function, that of enabling you to pour milk into your cereal bowl. By contrast, when denoting expressions are combined, their independent meanings compose according to strict logic-like rules rather than intelligible association. In moving from (5) to (6), for instance, you do not retain the message in (5) and then add something else; you end up without referring at all to kissing your sister. (5) (6)

I kissed my sister. I kissed my sister’s friend.

I do not assert that formalist denotational semantics cannot account for the diverse syntactic reflexes of quotation, but every proposal that I have seen is either empirically inadequate, implausibly artificial, or both (Saka 2011). Be that as it may, (Q) is denied or at least neglected by proponents of (F3) while (M) is denied or at least neglected by proponents of (F4). The two need to be brought together, is my claim. The question has been raised, how does my account compare with the intriguing discourse-representation approach taken by Geurts and Maier (2005), henceforth G&M? Despite a shared reliance on context, the differences are considerable. First, G&M’s account is somewhat circular. For they analyze pure quotation (7) and mixed quotation (8) as follows: (7) (7’) (8) (8’)

“Boston” contains six letters. The name “Boston” contains six letters. George says Tony is his “bestest friend”. George says Tony is his [what George calls “bestest friend”].

314

Paul Saka

To understand quotation of either sort, G&M require that we first understand quotation. My account, by contrast, non-circularly explains quotation in terms of mentioning, and mentioning in terms of certain ostensive intentions. Second, G&M give two different accounts of quotation, one for the pure case (7) and one for the mixed (8). Why do quotation marks sometimes imply “what so-and-so calls” and sometimes not? They just do. My pragmatic account, in contrast, invokes the presence of a reasoning speaker who can use real-world knowledge to fill in the gaps of what is said. Third, G&M’s account is open to technical challenges. G&M ask us to suppose: (9) (10)

When in Santa Cruz, Peter orders “[ei]pricots” at the local market. When in Amherst, Peter orders “[æ]pricots” at the local market.

They then gloss (9) as: (9’)

Whenever Peter orders what he calls “[ei]pricots” or “[æ]pricots” at the local market in Santa Cruz, he orders what he calls “[ei]pricots”.

But what Peter calls “[ei]pricots” is the same thing as what he calls “[æ]pricots”, in which case (9’) is truth-conditionally equivalent to: (9”)

Whenever Peter orders what he calls “[ei]pricots” or “[æ]pricots” at the local market in Santa Cruz, he orders what he calls “[æ]pricots”.

Because (9”) does not accurately represent (9), I conclude that the formalistic, truth-conditional account of G&M is untenable. (It may be that G&M’s gloss [9’] does not express their actual technical commitments, but it’s impossible to be sure because their formal notations are variously opaque, mistaken, and incomplete. Their notations are confusing when variables seem to change, midformula, from bound to unbound; use-mention errors are committed when variables are in quotation marks; [9’] is never translated into G&M’s formalism; and so on.) To summarize, my account holds that a word can be deployed in two ways: it can be used, in which case it carries its customary reference, or it can be mentioned, in which case it is used to refer to itself or to something like itself. If it is mentioned, the speaker may add quotation marks in order to signal that it is being mentioned. One might go so far as to say that, in such a case, quotation marks are required by the conventions of some communities, though it would remain true that it is S’s use of the quoted word (plus or minus the quotation marks) that does the referring.

The Act of Quotation

6.

315

Referential multiplicity

Principle (M) allows for many uses of quotation. For instance, the quotation in (11) can be used to signify a linguistic type, for a token cannot appear in two places at once; the quotation in (12) can be used to signify a token, for indexical types do not have reference; and the quotations in (13) and (14) can be used to signify a formal structure and propositional content, respectively. (11) (12) (13) (14)

In her letter, “you” appears twice. In her letter, “you” refers to me. The sentence “I am sorry” is made of three words. “I am sorry”, in German, is [identical to] “Das tut mir Leid”.

The indeterminacy or multiplicity thesis – that quotations can variously signify types, tokens, spellings, meanings, and so forth – has been called unnecessary (e.g. Haack 1974: 282) and “extraordinary” (Cappelen and Lepore 1999: § 4). Arguments against it, however, are unpersuasive, as I explain below (i, iii, iv, vi); what’s more, there is positive support for the multiplicity thesis, above and beyond the prima facie case already made (ii, v). (i) Cappelen and Lepore, henceforth C&L, complain that the postulation of type-token ambiguity entails statements of the form “a” =/ “a” (C&L 1999: § 4 & 2007: 71). C&L regard this as absurd, but it is not. After all, some contextualists ascribe different truth-values to different tokens of the liar sentence . 

 is false.

To avoid deducing that  is both true and false, truth-contextualists such as Goldstein 2001 tell us that it is by recognizing that one deduction establishes one token of  as true, while a different deduction establishes some other token of  as false; hence (15)

“ is false” =/ “ is false”.

Regardless of whether truth-contextualists are right about ’s truth-values, it seems that they succeed in using the two quotations in (15) to refer to distinct tokens. (ii) If C&L are right that quotation lacks type-token ambiguity, then quotation would be unlike virtually all other count nominals, and unlike so much other language as well. For type-token polysemy is quite systematic: (16)

Turing turned the computer off. [computer token] Turing read about the invention of the computer. [computer as type]

316 (17) (18)

Paul Saka

The elephant is about to charge. [elephant token] The elephant is endangered. [elephant as type] Yesterday the dean woke up at 7 am. [waking-at-7 token] All semester long the dean woke up at 7 am. [waking-at-7 as type]

Why do “computer”, “elephant”, and “woke up at 7 am” all exhibit type-token ambiguity? It’s no coincidence, as the same pattern runs throughout English and other languages. It cannot be ascribed to the forms in (16–18), since the forms have nothing in common with each other. What (16–18) do have in common is that they are all used by intelligences that can mentally associate tokens and their types; and this suggests a prominent role in semantic theory for speaker/hearers, a role that is ignored or downplayed by formal approaches but recognized and emphasized by speech-act approaches. (iii) C&L (1999: § 4) suggest that speaker/hearers who understand (19) paraphrase it as (19’). (19) (19’)

“I” refers to me. The token of “I” in this sentence refers to me.

Of course speaker/hearers do not literally restate (19) as (19’); they do not say it out loud and they do not say it sotto voce. Evidently what C&L have in mind is that (19) means the same as (19’), or more pedantically that (19) means the same as (19”). (19”)

The token of “I” in sentence (19) refers to me.

But to acknowledge that a speaker uses (19) to express what (19”) does is to concede my point! (iv) C&L further suggest that the language community may simply be mistaken, due to metaphysical misapprehension, in thinking that its quotations refer to anything other than expression types. While their scenario is indeed conceivable, it is irrelevant. To begin with, ontological error does not impact meaning: speakers can use “‘chili”’ for the purpose of referring to the concept chili even if concepts don’t exist, just as physicists might use “element 120” in order to refer to element 120, even if element 120 doesn’t exist. And even were that not so, C&L’s appeal to ignorance hardly undermines the multiplicity thesis, for idle speculation has no evidentiary force. (v) While C&L energetically criticize the multiplicity thesis, I confess I do not completely understand their motivation. True, Davidson does state that quotations always refer to expression types as individuated by shape, but I never took this element of the demonstrative theory seriously, and indeed Garcia-

The Act of Quotation

317

Carpintero (1994: 259–261), defending the demonstrative theory, asks us not to. In fact, C&L’s rejection of the multiplicity thesis seems to go against their own work. Because C&L (1997: 216) place no semantic constraints on the sametokening relation, I take it that x tokens the same as y iff there is some contextually delimited type F such that x and y each belong to F. Consequently (20), given C&L’s analysis (20’), allows many readings. (20) (20’)

“Lobster” is a word. ∀x (x is a word if x tokens the same as that). Lobster

In (20’), “the same” indeterminately refers to any number of types: same wordform type, same word-class type, or even same token. In effect, from what I can tell C&L’s demonstrative analysis entails the multiplicity thesis. C&L (2007: § 12.2) admit that a quotation can refer to either an expression or a sign, a medium-specific representation of an expression. To reconcile this concession with their univocality thesis, they say that a quotation form may represent two syntactically distinct structures. However, to say that a single form, for instance “visiting relatives can be boring”, represents two syntactically distinct structures is exactly what it means to say that it is syntactically ambiguous. (vi) I need to repeat that my account treats quotation as referentially, not lexically, ambiguous. C&L claim that I regard (21) as grammatical and true. (21)

Runs is a noun.

Indeed I do regard this as grammatical, but that does not mean I think it true in the intended sense.6 Grammatical subjects needn’t be noun phrases and they needn’t be headed by lexical nouns. Grammatical subjects can take the form of verb phrases and wh-clauses: (22) (23)

Running twenty minutes every day is good exercise. Whoever called it necking was a poor judge of anatomy. [Groucho Marx]

Adjectives, verb phrases, and verb-particle compounds can serve as nouns: (24)

the rich, the old; a quick pick-me-up; the print-outs

If you use a non-noun to refer to something thing-like, you can often use it as if it’s nominal, so it’s no surprise that when you use a verb to refer to the thing 6. If you say (21) in reference to the “runs” of “Jo runs” then your statement is false, although if you refer to the “runs” of “dysentery victims suffer from the runs” then your statement is true.

318

Paul Saka

that is itself, as you do in (21), you can use it as a nominal (you can use it as a noun or a noun phrase, pace Pafel, this volume). The phenomenon of functional shift is widespread (Bauer 2005, Meibauer 2007), and it’s not true that positing it is “particularly unattractive, … requiring substantial changes in how we think about English syntax” (C&L 2007: 96). While many examples of functional shift have become conventionalized, the general phenomenon is driven by creative, yet rationally constrained, speaker agency, as illustrated by novel instances whose interpretations require background knowledge: (25)

The judges were Susan Boyled.

This fact suggests to me that the best explanation for the functional shift of citational quotation invokes speaker reference rather than denotation. (vii) That said, I follow Gomez-Torrente (2001: 128) in denying that the multiplicity thesis decisively adjudicates among larger theories of quotation. Denying the thesis is not integral to either the demonstrative approach or the minimal approach, nor is asserting it “central” to any broad alternative, pace C&L (1999: 741). It does, however, suggestively support a speech-act approach, if (ii) and (vi) are correct; and it matters in fixing on any detailed theory of quotation. 7.

Conclusion

According to the demonstrative theory, quotation marks refer; according to formalist identity theories, quotations refer; according to speech-act theories, speakers refer. I propose that speakers can refer naturally because of (M), which entails all of the natural speech-act theses (S”), and I propose that speakers can use quotation marks to refer by convention (Q). (Q) combines with (M) to entail loose-identity versions of Conventional Speech-Act Quotatum and Partial Identity (S3’/S4’), which entail Basic Speech-Act Quotatum and Partial Identity (S3/S4). The speech-act theses, however, neither reduce to nor expand to the formalist theses. Speech-act theses (S3) and (S4) are extensionally equivalent and so are not counterparts of formalist theses (F3) and (F4), which are distinct from each other (§ 3). We cannot derive (F1–F4) from (S1–S4) because Bridge Principle A is invalid, and we should not derive (F1–F2) from (S1’–S2’) because the latter are false (§ 4). We cannot derive theses (F) from (Q), under standard truth-conditional construal, both because truth is objective while recognizability is subjective and because the effects of quotation marks do not appear to be

The Act of Quotation

319

compositional (§ 5). Finally, no theory can explain quotation in terms of strict identity; similarity is all that is called for, thus lending credence to a pragmatic understanding of the phenomena (§ 6).7 References Akiba, Ken 2005

A unified theory of quotation. Pacific Philosophical Quarterly 86: 161–171.

Bauer, Laurie & Salvador Valera (eds.) 2005 Approaches to Conversion. Muenster: Waxmann. Belnap, Nuel & Dorothy Grover 1973 Quantifying in and out of quotes. In Truth, Syntax and Modality, H. Leblanc (ed.), 17–47. Amsterdam: Springer. (Reprinted in Prosentential Theory of Truth, D. Grover (ed.). Princeton: Princeton University Press.) Benbaji, Yitzhak 2004a A demonstrative analysis of ‘open quotation’. Mind and Language 19: 534–547. Benbaji, Yitzhak 2004b Using others’ words. Journal of Philosophical Research 29: 93–112. Botterell, Andrew & Robert Stainton 2005 Quotation. Critica 37.110: 3–33. Caplan, Ben 2002

Quotation and demonstration. Philosophical Studies 111: 69–80.

Cappelen, Herman & Ernie Lepore 1997 Varieties of quotation. Mind 106: 429–450. (References are to abridged reprint in Murasugi & Stainton.) Cappelen, Herman & Ernie Lepore 1999 Reply to Saka. Mind 108: 741–750. Cappelen, Herman & Ernie Lepore 2005 Varieties of quotation revisited. In Hybrid Quotations, P. de Brabanter (ed.), 51–75. Amsterdam: Benjamins. Cappelen, Herman & Ernie Lepore 2007 Language Turned on Itself. Oxford: Oxford University Press. 7. Portions of this paper were presented at the International Conference on Quotation and Meaning (Johannes Gutenberg University, Mainz, 2007). I am very grateful to the conference organizers, the audience, Jan Terje Faarlund, James Myers, and above all Daniel Gutzmann.

320

Paul Saka

Davidson, Donald 1979 Quotations. Theory and Decision 11: 27–40. (Reprinted in: D. Davidson (ed.), Essays on truth and interpretation, 79–92. Oxford: Clarendon 1984.) De Brabanter, Philippe (ed.) 2005 Hybrid Quotations. Amsterdam: Benjamins. [=Belgian Journal of Linguistics 17, 2003]. Frege, Gottlob 1980

On sense and meaning. In Translations from the Philosophical Writings of Gottlob Frege, 3rd edn, P. Geach and M. Black (eds.), 56–78. Oxford: Blackwell.

Garcia-Carpintero, Manuel 1994 Ostensive signs. Journal of Philosophy 91: 253–264. Garcia-Carpintero, Manuel 2005 Double-duty quotation. In Hybrid Quotations, P. de Brabanter (ed.), 89–108. Amsterdam: Benjamins. Geurts, Bart & Emar Maier 2005 Quotation in context. In Hybrid Quotations, P. de Brabanter (ed.), 109–128. Amsterdam: Benjamins. Goldstein, Laurence 2001 Reply to Weir. Analysis 61: 115–126. Gomez-Torrente, Mario 2001 Quotation revisited. Philosophical Studies 102: 123–153. Gutzmann, Daniel & Erik Stei this vol. Quotation marks and kinds of meaning. Haack, Susan 1974

Mentioning expressions. Logique et Analyse 17: 277–294.

Johnson, Michael this vol. The punctuation theory of quotation. Kripke, Saul 1977

Lakoff, George 2004

Speaker’s reference and semantic reference. Midwest Studies in Philosophy 2: 255–276. (Reprinted in Meaning and Truth, J.L. Garfield and M. Kiteley (eds.), 162–188. New York: Paragon House.) Don’t Think of an Elephant. White River Junction: Chelsea Green.

Ludwig, Kirk & Greg Ray 1998 Semantics for opaque contexts. Philosophical Perspectives 12: 141– 166. Lycan, William 1999

Philosophy of Language. New York: Routledge.

The Act of Quotation

321

McCullagh, Mark 2007 Understanding mixed quotation. Mind 116: 927–946. Meibauer, Jörg 2007

How marginal are phrasal compounds? Morphology 17: 233–259.

Murasugi, Kumiko & Robert Stainton (eds.) 1999 Philosophy and Linguistics. Boulder CO: Westview. Nunberg, Geoffrey 2006 Talking Right. New York: Public Affairs. Pafel, Juergen this vol.

Two dogmas on quotation.

Parsons, Terence 1982 What do quotation marks name? Philosophical Studies 42: 315–328. Pietroski, Paul 1999

Compositional quotation. In Philosophy and Linguistics, K. Murasugi and R. Stainton (eds.), 245–258. Boulder, CO: Westview.

Predelli, Stefano 2008 The demonstrative theory of quotation. Linguistics and Philosophy 31: 555–572. Recanati, Francois 2001 Open quotation. Mind 110: 637–687. Reimer, Marga 1996 Richard, Mark 1986 Saka, Paul 1998 Saka, Paul 2005 Saka, Paul 2006 Saka, Paul 2007 Saka, Paul 2009

Quotation marks: demonstratives or demonstration? Analysis 56: 131–141. Quotation, grammar, and opacity. Linguistics and Philosophy 9: 383– 403. Quotation and the use-mention distinction. Mind 107: 113–135. Quotational constructions. In Hybrid Quotations, Brabanter (ed.), 187–212. Amsterdam: Benjamins. Reprinted in Saka 2007. The demonstrative and identity theories of quotation. Journal of Philosophy 103.9: 452–71. Reprinted in Saka 2007. How to Think about Meaning. Dordrecht: Springer. The semantics of quotation and the nature of language. Ms., University of Houston.

322

Paul Saka

Saka, Paul 2010

Tensions in the theory of truth. Topoi 29.2: 126–135.

Saka, Paul 2011

Review of Cappelen and Lepore. Protosociology, forthcoming.

Saka, Paul 2012

Quotation and conceptions of language. Dialectica, forthcoming.

Searle, John 1969

Speech Acts. Cambridge: Cambridge University Press.

Sorensen, Roy 2008

Empty quotation. Analysis 68.1: 57–61.

Strawson, Peter F. 2001 Meaning and Truth. In Philosophy of Language, fourth edition, A.P. Martinich (ed.), 110–119. Oxford: Oxford University Press. Tajtelbaum, Al 1957

It is impossible to be told anyone’s name. Analysis 17: 52–53.

Washington, Corey 1992 Quotation. Journal of Philosophy 89: 582–605. Washington, Corey & John Biro 2001 A logically transparent approach to discourse reporting. Mind & Language 16: 146–172. Wertheimer, Roger 1999 Quotation apposition. Philosophical Quarterly 49: 514–519.

Axioms of Reference and Rules of Quotation Savas L. Tsohatzidis

1.

Introduction

My purpose in this paper is to argue that two fairly widely, and often conjointly, held views in contemporary philosophy of language – one concerning the structure of semantic theories, and the other the interpretation of quotation – cannot be consistently adhered to. As far as I shall here be able to tell, the prospects of a face-saving restoration of consistency do not look bright. 2.

The problem

On the basis of a variety of well-rehearsed reasons, many philosophers and linguists routinely assume that theories of meaning for natural languages should be finitely axiomatized theories of truth for those languages. And most of those who share that assumption also assume that, among the axioms that are required, and are available, for the construction of such theories, there are those that specify the referents of the non-complex singular terms, in particular the proper names, that natural languages commonly contain. Thus, in the simplest, homophonic, case, a truth-theory of English, expressed in English, and purporting to play the role of a semantic theory of English, is generally supposed to include a finite number of reference axioms such as (1)–(3) – often equivalently expressed in a quasi-English idiom that employs interpretation brackets and the equality sign –, (1) (2) (3)

The referent of “Aristotle” is Aristotle The referent of “Napoleon” is Napoleon The referent of “Darwin” is Darwin

each of which is a substitution-instance of the schema (A) in which a proper name occurs first quoted and then disquoted: (A)

The referent of “Npr ” is Npr

324

Savas L. Tsohatzidis

Let us call “Reference Thesis”, or RT, the thesis that an adequate truth-theoretic semantics for a natural language such as English will include, among its finite list of reference axioms, appropriate substitution instances of (A), for example (1)–(3).1 Now, it is equally, if not more, widely supposed that a quoted and a disquoted occurrence of a name do not have the same referent – that, for example, the quoted occurrence of the proper name in (4) does not have the same referent as the disquoted occurrence of the same proper name in (5), or that the quoted occurrence of the proper name in (6) does not have the same referent as the disquoted occurrence of the same proper name in (7): (4) (5) (6) (7)

“Quine” is written with five letters Quine has read five books “London” has two vowels London has two orchestras

For, it is pointed out, (4) is true just in case a certain name (and not a certain person bearing that name) is written with five letters (there is, in fact, no literal sense in which a person ‘is written with letters’), whereas (5) is true just in case a certain person (and not a certain personal name) has read five books (there is, in fact, no literal sense in which names ‘read books’); and similarly, (6) is true just in case a certain name (rather than a certain city bearing that name) has two vowels (there being no sense in which a city can ‘have vowels’), whereas (7) is true just in case a certain city (rather than a certain city name) has two orchestras (there being no sense in which a name can ‘have orchestras’). It follows, then, that, whatever else they might share, a quoted and a disquoted occurrence of a name do not have the same referent.

1. In the literature on truth-theoretic semantics, one sometimes encounters, instead of reference axioms in the style of (1)–(3), reference axioms in the style of (1a)–(3a): (1a) “Aristotle” refers to Aristotle (2a) “Napoleon” refers to Napoleon (3a) “Darwin” refers to Darwin For present purposes, that difference in style is inconsequential, since anyone taking (1a)–(3a) to be true can hardly avoid taking (1)–(3) to be true as well, and will certainly regard (1b)–(3b) as blatantly contradictory: (1b) “Aristotle” refers to Aristotle, but Aristotle is not the referent of “Aristotle” (2b) “Napoleon” refers to Napoleon, but Napoleon is not the referent of “Napoleon” (3b) “Darwin” refers to Darwin, but Darwin is not the referent of “Darwin”

Axioms of Reference and Rules of Quotation

325

Let us call “Quotation Thesis”, or QT, the thesis that a quoted and a disquoted occurrence of a name do not have the same referent2 . QT entails that every substitution instance of the schema (A’), in which the same proper name occurs first quoted and then disquoted, (A’)

The referent of “Npr ” is not Npr

is true. For, supposing that some substitution instance of (A’) is not true is supposing that the disquoted occurrence of a name that that instance contains does correctly specify the referent of the quoted occurrence of the same name that it contains. And that supposition is inconsistent with what QT asserts, namely, that a name’s quoted and disquoted occurrences do not have the same referent. In particular, QT entails that each of the following sentences is true: (1’) (2’) (3’)

The referent of “Aristotle” is not Aristotle The referent of “Napoleon” is not Napoleon The referent of “Darwin” is not Darwin

For, if these sentences were not true, and the disquoted occurrences of names in them were correctly specifying the referents of the corresponding quoted occurrences, it would follow that a name’s quoted and disquoted occurrences can have the same referent, which is just what QT denies. It is evident upon inspection that (1)–(3) and (1’)–(3’) contradict each other, and the same holds for any corresponding substitution instances of (A) and (A’). Unless, therefore, one was prepared to embrace each contradiction that results from conjoining instances of (A) with corresponding instances of (A’), one should reject either RT or QT (or, of course, both). The problem is that most philosophers and linguists with an interest in systematic semantic theorizing are tacitly or openly committed to both RT and QT. Thus, a typical truth-theoretic semanticist will begin his account of English by adopting reference axioms like (1)–(3) – or their quasi-English renditions employing interpretation brackets and the equality sign – and will insist, if appropriately prompted, that such axioms portray one of the fundamental ways in which ‘language connects with extra-linguistic reality’. To a naïve interlocutor who might wonder how such seemingly trivial axioms can achieve that goal, the theorist will reply that, despite appearances, the axioms are not trivial. For – the 2. That thesis evidently concerns only instances of quotation that, in terms familiar from recent literature on the topic, would be described as instances of “pure” or “closed” quotation (Cappelen and Lepore 1997, Recanati 2001). This is also the variety of quotation to which all examples of quotation discussed in the present paper belong.

326

Savas L. Tsohatzidis

theorist will explain –, the quoted and disquoted occurrences of names that appear in each of (1)–(3), no matter how alike they may sound, do not have the same referents, the former picking out linguistic entities (specifically, words) and the latter extra-linguistic ones (specifically, persons); there is, then, nothing trivial in what each axiom says, and nothing preventing it from portraying a connection between a linguistic entity, on the one hand, and an extra-linguistic one, on the other. In giving that reply, however, the theorist will be forgetting that if –as he claims– quoted and unquoted occurrences of names do not have the same referents, then what the axioms say, though indeed not trivial, is patently false: If a quoted and a disquoted occurrence of Aristotle’s name do not have the same referent, then a disquoted occurrence of that name cannot be correctly specifying the referent of its quoted occurrence, and so (1) must be false. If a quoted and a disquoted occurrence of Napoleon’s name do not have the same referent, then a disquoted occurrence of that name cannot be correctly specifying the referent of its quoted occurrence, and so (2) must be false. If a quoted and a disquoted occurrence of Darwin’s name do not have the same referent, then it is impossible for a disquoted occurrence of that name to be correctly specifying the referent of its quoted occurrence, and so (3) must be false. And if (1)–(3) have to be false, it is their contradictories in (1’)–(3’) that have to be true. Unless, therefore, the truth-theoretic semanticist were happy to simultaneously accept both the purported axioms in (1)–(3) and their contradictories in (1’)–(3’), he should recognize that there is something wrong with his procedures, and that at least one, and possibly both, of the theses RT and QT, which inform these procedures, should be rejected. 3.

A non-solution

It might be suggested that the above clash between RT and QT could be avoided by adopting a thesis to the effect of thesis (X ) below, in the hope that that thesis would ensure that corresponding instances of (A) and (A’) are not contradictory: (X ) When a sentence of the form (A’) is true or a sentence of the form (A) is false, there is a pair of invisible quotation marks surrounding the pair of visible quotation marks that the sentence contains; when, on the other hand, a sentence of the form (A’) is false or a sentence of the form (A) is true there is no pair of invisible quotation marks surrounding the pair of visible quotation marks that the sentence contains.

That suggestion would be ineffective, however. For one thing, one might simply refuse to take seriously the idea that invisible quotation marks can be freely supposed to spring into existence just when a particular combination of views

Axioms of Reference and Rules of Quotation

327

is threatened by the behaviour of visible quotation marks. For another thing, and more importantly, one would have very good reasons to deny that thesis (X ), even assuming it to be adopted, would have any power to prevent RT and QT from conflicting. For, clearly, (X ) could not possibly ensure the noncontradictoriness of corresponding instances of (A) and (A’) unless it were supplemented by a separate thesis specifying what, exactly, the semantic difference between single and double quotation marks (visible or invisible) is supposed to be. And, as far as I can see, the only available candidate for that role would be a thesis to the effect of thesis (Y ) below: (Y ) The result of placing a quotation-less expression inside a single pair of quotation marks is a singular term referring to that expression, whereas the result of placing a quotation-less expression inside two pairs of quotation marks is a singular term referring to the result of placing that expression inside a single pair of quotation marks.

Thesis (Y ), however, immediately generates an even more obvious clash between RT and QT. For, although it is consistent with QT, it is patently not consistent with RT. Recall that, according to RT, sentences like (1)–(3) – repeated below for convenience – are all true: (1) (2) (3)

The referent of “Aristotle” is Aristotle The referent of “Napoleon” is Napoleon The referent of “Darwin” is Darwin

If, however, the result of placing a quotation-less expression inside a single pair of quotation marks is a singular term referring to that expression, as thesis (Y ) states, then the quotations in (1)–(3) must be taken to refer to certain linguistic expressions, and (1)–(3) must consequently be taken to be falsely saying, of those linguistic expressions, that they are identical with certain extra-linguistic entities. In particular, accepting (Y ) requires accepting that (1) is true just in case the linguistic expression referred to by the quotation in (1) – which happens to be Aristotle’s name – is identical with Aristotle; it requires accepting that (2) is true just in case the linguistic expression referred to by the quotation in (2) – which happens to be Napoleon’s name – is identical with Napoleon; and it requires accepting that (3) is true just in case the linguistic expression referred to by the quotation in (3) – which happens to be Darwin’s name – is identical with Darwin. Since, however, it is evident that Aristotle’s name is not the same thing as Aristotle, Napoleon’s name is not the same thing as Napoleon, and Darwin’s name is not the same thing as Darwin, it follows from thesis (Y ) that, contrary to RT – though not contrary to QT –, (1)–(3) must all be false. Indeed, this even more obvious way in which RT and QT clash makes it even

328

Savas L. Tsohatzidis

more perplexing that so many philosophers and linguists insouciantly declare their allegiance to both RT and QT. 4.

Conclusion

It is easy enough to imagine under what conditions the problem here considered would fail to arise, but not easy at all to imagine what reasons could be offered in favour of thinking that these conditions are satisfied. The problem would fail to arise if the conjunction of the following three claims could be adequately defended: that an expression of the form “the referent of “Npr ”” refers sometimes to a word-type whose tokens have themselves a certain extra-linguistic referent and sometimes to that extra-linguistic referent itself (for example, that “the referent of “Aristotle”” refers sometimes to a word-type whose tokens refer to Aristotle, and sometimes to Aristotle himself); (ii) that this difference is a function of the fact that the expression “the referent of “Npr ”” is ambiguous between two sharply different (though perhaps metonymically related) senses, one of which enables it, given a context, to pick out a word-type as its referent, and the other of which enables it, given a context, to pick out the referent of tokens of that word-type as its referent; and finally, (iii) that when the expression “the referent of “Npr ”” occurs in a true sentence of the form (A) or in a false sentence of the form (A’), it has only the latter of these senses, whereas when it occurs in a false sentence of the form (A) or in a true sentence of the form (A’), it has only the former of these senses. (i)

Unfortunately, however, saying under what conditions a problem would fail to arise is not showing that any of these conditions actually obtains. And I can, at present, think of no good reason that one might offer for taking any of (i)–(iii) to be true – no reason, in other words, expect the reason that, if they were so taken, then no clash between RT and QT would have to be faced. Since adopting such a procedure in the present circumstances would obviously be questionbegging, the only options that I am currently entitled to regard as open are the unattractive ones previously indicated: rejecting RT, rejecting QT, or rejecting both RT and QT. At least, none of these is as unattractive as rejecting the law of non-contradiction.3 3. I am grateful to Elke Brendel for an exchange regarding an earlier draft. All errors are my own.

Axioms of Reference and Rules of Quotation

References Cappelen, Herman & Ernie Lepore 1997 Varieties of quotation. Mind 106: 429–450. Recanati, François 2001 Open quotation. Mind 110: 637–687.

329

Index

acquisition 22f. ambiguity 22, 86, 94, 100, 102, 111, 151, 175, 201, 315f. anaphora 213ff., 222, 238f. attachment 20, 125f., 132, 165ff.

double duty

behabitive 25, 85, 87, 91, 93f., 98ff., 102, 104

formal semantics 4, 47, 200, 303f., 306, 308 Frege, G. 38f., 44, 110, 117, 203

context-sensitivity 115, 120f., 204, 206 context shift 6, 12, 21, 184, 186, 216, 272, 286, 290, 295 contextualism 64, 72–75, 120 Davidson, D. 17f., 25, 35, 37–45, 48, 107, 113–117, 120, 122, 135, 195f., 198, 200, 203, 205–207, 316 deferred ostension theory 18, 107, 195 definite description 87, 170, 250, 260ff., 271, 308 deixis 21, 293, 296, 299 demonstration 13, 27, 108, 112, 121, 153, 156, 232ff., 241, 244, 247, 249, 259, 307 demonstratives 71, 107f., 112, 116, 119, 142, 164f., 196 demonstrative theory 17f., 24ff., 37, 45, 107, 109, 164, 195, 199f., 202– 207, 249, 304f., 316ff. depiction 233f., 236f., 239–244, 247 dialetheism 48–53, 55f. direct discourse 122, 129, 132, 278, 291, 293 disambiguation theory 18 dispensability 26, 195, 202ff., 206 disquotational indirect report 25, 59, 120

195, 200ff., 207

exemplification 26, 195, 197–200, 202, 206f. experimental pragmatics 24

greengrocer’s quotation 6 Grice, H. P. 24, 70, 74, 163, 187–190, 203, 310 identity theory 18, 26, 28, 110, 117, 195–199, 201, 203, 206f., 303, 304f. illocution 22, 25, 75, 85f., 280, 286 implicature 21, 74, 77, 124, 129, 175f., 179, 181–184, 203, 219, 225 – conventional 20, 26f., 70, 72, 107, 123f., 128-131, 133, 163, 165, 168f., 190, 210, 217, 219, 222 – conversational 14, 20, 24, 26f., 64f., 74, 125, 161, 163, 174, 176f., 179, 181f., 184f., 187–190, 203, 210, 219, 222, 228, 229 indeterminacy 22, 303f., 315 indexical 2–4, 10, 12, 19ff., 23, 27f., 71ff., 108, 112f., 126, 132, 135, 164f., 186, 188, 204, 217ff., 272, 277ff., 315 indirect discourse 27, 37, 122, 134f., 204, 215, 218f., 240, 277, 281f., 285f. interface 19, 24 judgment report kinds of meaning

93f. 163

332

Index

languageless creature 25, 35, 40ff., 44 liar paradox 16 locution 67, 85ff., 93, 95, 98, 272 logic of quotation 200 mention 3–9, 14–17, 24, 112,123f., 133–135, 142, 162, 187, 210–216, 222, 238, 241f., 249, 306f., 314 message 20, 125f., 132, 165ff., 183, 313 meta-language 16, 50, 53, 56 metalinguistic quantification 265–268, 270, 272 metarepresentation 1, 2, 6, 13, 22 minimalism 64, 68 minimal theory 24, 26, 113 non-manual 12, 27 non-serious action 233, 237, 247 ostension theory

18, 22, 107, 195

paraconsistent logic 25, 47–49, 51, 53f. pictoriality 26, 195ff., 199, 201, 207 pragmatic enrichment 5f., 8, 24 pragmatic indicator 14, 184 pragmatics 1, 4, 15, 21, 24, 62, 64, 68, 73, 76, 122f., 134, 188ff., 289 presupposition 21, 26, 85, 89, 97, 99, 101, 103, 123f., 128–131, 133, 163f., 170–173, 190, 205f., 210, 219, 221f., 228, 312 processing 22ff. proper name theory 16f., 161 propositional attitudes 25, 35ff., 39– 45 punctuation theory 26f. quantifying-in 262 Quine, W.V.O. 16, 25, 35ff., 40ff., 44f., 161, 176, 185, 187, 195 quotation – closed 7, 26, 249, 325

– direct 2–9, 11ff., 19, 23, 27, 92f., 128, 130, 161, 184, 209, 231–250, 255f., 258, 264, 266, 270ff. – double duty 26, 107–108, 118, 122–125 – emphatic 6f., 9, 21 – free direct 6 – free indirect 7, 13 – indirect 3–13, 21, 23, 89, 128, 231f., 246 – mixed 3ff., 7ff., 12, 17, 19ff., 26f., 115, 122f., 127, 129f., 132–135, 162, 164, 169, 171, 180, 186, 195, 201, 206, 209ff., 245ff., 255ff., 264, 270, 313 – open 7, 19ff., 26, 107, 122 – pure 1, 4f., 7f., 16f., 22, 27, 114, 135, 140, 161, 163, 184f., 188f., 209, 213f., 249ff., 313 – scare 5–9, 20, 23, 209, 273 quotation mark 1f., 6–9, 11, 14f., 17– 20, 23f., 26f., 37, 45, 109–117, 122f., 126, 161ff., 165, 176, 178, 185f., 195f., 198ff., 202–206, 304ff., 311f., 314, 318, 326f. replication 231, 233, 241 reported discourse 27, 277, 284f., 281, 287, 298f., 292 reported speech 1, 3, 7, 89, 288f. role shift 11f., 27, 277f., 280, 291 self-reference 16, 51, 196, 198ff., 250 semantics-pragmatics divide 25, 110, 115, 123f., 127, 188 semantics/pragmatics interface 2, 19, 163, 181 semantic reference 26, 110, 312 sign language 11, 27, 139f., 143, 147, 151–154, 156f., 261, 277f., 287 speaker reference 26, 139f., 152ff., 156f., 306f., 312, 318 speech act 21, 28, 52, 55f., 67, 75f., 85, 108, 124, 133, 234, 288f., 303f., 306–310, 316, 318

Index speech report 131, 173, 213, 257, 270ff., 295 subjective meaning 104 subjunctive 10, 12, 23, 86, 88f., 91, 96, 103 suppositio materialis 198

Tarski, A. 16, 37, 47, 114, 117, 120f., 187, 210f. truth operator 47, 51, 54, 56 truth predicate 47ff., 51, 53 truth-theoretic semantics 28, 324

333

types 18, 108ff., 112, 117, 119ff., 123, 168, 200, 240f., 311, 315ff. underdetermination 163 use/mention-distinction 7 use theory 18, 24, 307 verbs of judging 25, 85, 103f. verb second 88f. what is said 24ff., 59f., 60, 62, 64ff., 66, 110, 128, 169, 188f., 314 wording theory 27, 231f., 235f., 241f., 244, 247