254 29 3MB
English Pages 470 Year 2012
Modality and Theory of Mind Elements across Languages
Trends in Linguistics Studies and Monographs 243
Editor
Volker Gast Founding Editor
Werner Winter Editorial Board
Walter Bisang Hans Henrich Hock Heiko Narrog Matthias Schlesewsky Niina Ning Zhang Editor responsible for this volume
Volker Gast
De Gruyter Mouton
Modality and Theory of Mind Elements across Languages edited by
Werner Abraham Elisabeth Leiss
De Gruyter Mouton
ISBN 978-3-11-027019-8 e-ISBN 978-3-11-027107-2 ISSN 1861-4302 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.dnb.de. ” 2012 Walter de Gruyter GmbH & Co. KG, Berlin/Boston Typesetting: RoyalStandard, Hong Kong Printing: Hubert & Co. GmbH & Co. KG, Göttingen 앝 Printed on acid-free paper 앪 Printed in Germany. www.degruyter.com
Preface The present collection goes back to the 42nd Annual Meeting of the Societas Linguistica Europaea on September 9–12, 2009, at Lisbon, Portugal. The conference hosted a workshop on patterns of overt modality, which was convened and organized by Werner Abraham and Elisabeth Leiss, the editors of this volume. We thank the organizers of the SLE conference and the SLE secretary, Bert Cornillie, for their excellent preparation and administration of both the conference and the workshop.
Contributors Werner Abraham Allg. Sprachwissenschaft Universita¨t Wien 1090 Sensengasse 3a LMU-Universita¨t Mu¨nchen Germanistische Linguistik Schellingstraße 3/RG D-80799 Mu¨nchen [email protected] Marco Coniglio Humboldt-Universita¨t zu Berlin Institut f. deutsche Sprache und Linguistik (SFB 632 ‘‘Informationsstruktur’’ Teilprojekt B4) Dorotheenstraße 24 10117 Berlin [email protected] Markus Egg Institut fu¨r Anglistik und Amerikanistik Humboldt-Universita¨t zu Berlin Unter den Linden 6 D-10099 Berlin [email protected] Yoshio Endo Kanda University of International Studies Graduate School of Language Science 1-4-1, Wakaba, Mihama-ku, Chiba-shi, Chiba, 261-0014 Japan [email protected] Helen de Hoop Department of Linguistics Radboud University Nijmegen P.O. Box 9103 6500 HD Nijmegen The Netherlands [email protected] Łukasz Je˛drzejowski Zentrum fu¨r Allg. Sprachwissenschaft Programmbereich 3 Schu¨tzenstr. 18, 10117 Berlin [email protected]
Elisabeth Leiss Ludwig-Maximilians-Universita¨t Mu¨nchen Germanistische Linguistik Schellingstraße 3/RG D-80799 Mu¨nchen [email protected] Jakob Mache´ Institut fu¨r Deutsche und Niederla¨ndische Philologie Freie Universita¨t Berlin Habelschwerdter Allee 45 14195 Berlin [email protected] Benjamin Meisnitzer Ludwig-Maximilians-Universita¨t Mu¨nchen Institut fu¨r Romanische Philologie Ludwigstr. 25 D-80539 Munich [email protected] Pierre-Yves Modicom U. Paris-Sorbonne, E´cole Normale Supe´rieure, 45 rue d’Ulm F-75230 Paris cedex 05 [email protected] Sonja Mu¨ller Universita¨t Bielefeld Fak. Linguistik und Literaturwissenschaft Germanistische Linguistik Universita¨tsstraße 25 D-33615 Bielefeld [email protected] Kees de Schepper Department of Linguistics Radboud University Nijmegen P.O. Box 9103 6500 HD Nijmegen The Netherlands [email protected]
Table of contents Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
v vii
Introduction: Theory of mind elements across languages . . . . . . . . . Werner Abraham and Elisabeth Leiss
1
Part I. The foundation: speaker and hearer deixis, shifter, and double displacement Epistemicity, evidentiality, and Theory of Mind (ToM). . . . . . . . . . Elisabeth Leiss Illocutive force is speaker and information source concern. What type of syntax does the representation of speaker deixis require? Templates vs. derivational structure?. . . . . . . . . . . . . . . . . . . . . . . . Werner Abraham
37
67
Exploring the Theory of Mind interface . . . . . . . . . . . . . . . . . . . . . Jakob Mache´
109
The distribution of knowledge in (un)acceptable questions. . . . . . . . Sonja Mu¨ller
147
Traces of Bu¨hler’s semiotic legacy in modern linguistics . . . . . . . . . Werner Abraham
211
Part II.
Instances of deixis and origo in sundry languages
Modal particles, speaker-hearer links, and illocutionary force . . . . . Marco Coniglio
253
Discourse particles at the semantics-pragmatics interface. . . . . . . . . Markus Egg
297
Modality in the Romance languages: Modal verbs and modal particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Benjamin Meisnitzer
335
x
Table of contents
The epistemological treatment of information and the interpersonal distribution of belief in language: German modal particles and the typological challenge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pierre-Yves Modicom
361
On mood, evidentiality, and person e¤ects . . . . . . . . . . . . . . . . . . . Kees de Schepper & Helen de Hoop
383
Illocutionary force and modal particle in the syntax of Japanese. . . Yoshio Endo
405
What is it that keeps the rein on quotative modals so tight? A cross-linguistic perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Łukasz Je˛drzejowski
425
General index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
455
Introduction: Theory of mind elements across languages. Traces of Bu¨hler’s legacy in modern linguistics1 Werner Abraham & Elisabeth Leiss Universita¨t Wien & Ludwig-Maximilians-Universita¨t Mu¨nchen 1. The architecture of modality: Modality solving the most basic presuppositions This volume, in a wider sense of topical relationship, is meant to close the distance between Bu¨hler’s fundamental ideas and modern thought in linguistics and the philosophy of language. Under the particular view that we pursue here, Bu¨hler’s work can be seen as a determined attempt to outline the universal function of modality. This view may surprise the reader. Modality can be characterized as the most complex functional category of all linguistic categories known to the human species. Grammatical modality is the functional category which is acquired later than all other functional categories. Thus, it is necessarily dependent upon, and colored by, the language-specific architecture of the earlier acquired functional categories such as aspect, tense, and mood, whose semantics serve as elementary building blocks for the construction of the exceptionally complex functional category of modality. In line with this, our starting point for a reassessment of Bu¨hler’s work will be the cross-linguistic diversity of modality. The aim will be to provide a unified picture of modality which explains the driving force (or illocutionary force) creating di¤erent cross-linguistic patterns of modality. The very search for the deeper sources of modality reveals that the linguistic architecture of modality largely depends on the development of the category of person. The category of person is defined as a potential 1. The present volume derives from papers prepared for the SLE-2009 Workshop Lisbon ‘‘Theory of mind or Relevance – approaches to linguistic description and communicative explanation’’ convened by Werner Abraham (ViennaMunich) and Elisabeth Leiss (Munich). The present introduction was jointly written under support from the German-British DFG-project ‘Un-Cartesian Linguistics’, project submitters E. Leiss/Munich and W. Hinzen/Durham, LE718/9-1.
2
Werner Abraham & Elisabeth Leiss
shifter (in the Jakobsonian sense). Its reference shifts with the origo of the speaker. It is of central importance to understand how shifters serve as building blocks for functional categories. Functional categories involve double displacement and the split-up of the speaker into multiple personalities or view points. As a consequence, one central aim in this Introduction will be to give evidence and theoretical support for the hypothesis that the development of Theory of Mind in language acquisition (as well as evolution) depends largely on the development of functional categories, especially those of aspect, tense, and mood, whose architectures are based on the speaking person, the observer or viewing person, and the absent third person as the object of communication. In complex functions of modality, i.e. in modal particles, the addressee is part of the intricate design on which the Theory of Mind operates. In other words, the development of ToM2 (or, as we would prefer to say the theory of ) depends on the development of language. Thus, basic linguistic categories such as aspect, tense, mood, and modality form the human mind. This is in contrast to the (non-Universalist’s) claim that language as langue or I(nternal)-language is essentially usage based (consequently, E(xternal)-language), and that ToM evolves as a social skill learned alongside extending linguistic usage (for the Universalist view and with ToM as background of her epistemic interest, see Papafragou 2002 and Papafragou et al. 2007). What is of central importance from a pragmatic point of view is that modality constitutes the utmost linguistic achievement in the creation of di¤erent viewpoints. The whole of modality in itself, of course, comprises a range of di¤erent means to express illocutionary FORCE, such as modal verbs, modal particles (common to all of Germanic – except for English – and Slavic), and modal adverbials. In order to understand what is common to them, what separates them, and what lies behind them, we finally have to investigate thoroughly the syntax of modality. The di¤erent layers of modality or illocutionary force are defined by the structural web of syntax, which specifies the functions (semantics and pragmatics) of di¤erent linguistic techniques of modality. An essential outcome of the syntax part will be that lexical modality is not on a par with grammatical modality. Another investigative aim will be to expand the syntactic operator of illocutionary FORCE into suboperators and to specify and define them in syntactic terms. This insight also involves that modality is excluded from syntactic domains where 2. Another current terminology is .
Introduction: Theory of mind elements across languages
3
illocutionary FORCE is inactive, such as in dependent clauses, insofar as they do not allow for truth assessment. Another important syntactic topic will be the disclosure of covert patterns of modality. Finally, it will be shown that processes of grammaticalization of modality are syntactically driven. The volume presented here aims to evidence that the syntax, semantics, and pragmatics of modality converge, thus allowing for a universally valid explanation of modality as a constitutive part of the specific, i.e. linguistic human cognition. In a narrower sense of topical relationship, the contributions in the present book will narrow down on those expressive means that encode Foreign Consciousness Alignment/FCA overtly: epistemic modal verbs, modal adverbials, and modal particles. Direct lexical expressions are not available – at least not in paradigmatic representation – in all languages alike. Russian, for example, as well as most of the Slavic languages, does not possess a paradigm of modal verbs anywhere similar to those available in Germanic. The Germanic languages appear to be those endowed with FCA-elements most richly possessing also a variety of modal particles, which we do not consider to be on a functional (i.e., FCA-triggering) par with modality elements of adverbial status. This latter distinction reveals most clearly what Jakobson meant to address with his notion of shifter; what comes closest to Bu¨hler’s origo instantiations (which, to be sure, address many more and other instantiations as well); and what covers Peirce’s considerations about Person Triangulation taken up later by Davidson (see Abraham 2012, in this volume). In each of these notional instantiations the speaker and his belief states are put to the disposal of the addressee’s appraisal, while simultaneously triggering the addressee’s positive appreciation of the speaker’s assumption. The syntactic consequences of this mind searching procedure are manifold. First of all, inclusion of FCA-triggering lexicals (such as modal particles in German(ic)) segregates embedded clauses in two types: those that accept such illocutive elements as opposed to those that make the embedded clause unreadable. This is essentially an insight we owe to Haegeman (2006) and Coniglio (2007, 2009, 2011, and this volume): Illocutive force is limited to such clauses that embody root structural properties. Given that subordinate clauses (ever since Hooper and Thompson 1973) have been taken not to be asserted but presupposed, the solid observation that there are embeddings with root properties such as to allow speaker-addressee implications is a novel linguistic insight in itself. The fact that implementations of modality are particular diagnostics for this syntactic division provides new and solid contextual results. This fundamental property of embedded
4
Werner Abraham & Elisabeth Leiss
structures expanding over di¤erent types of subordinate clauses (factive vs. non-factive complement clauses; adverbial clauses of event predication vs. logical adjunction; and clauses with restrictive vs. non-restrictive modification) yields new challenging analytic questions and paths of investigation (Abraham 2012b). 2. What are the main categories of ToM in adult speech? To what extent are these categories cognitive or rather linguistic in nature? 2.1. Conceptual issues The contributions presented in this volume discuss material instantiating linguistic speaker-hearer involvement and its consequences for the clausal and lexical analysis. As explained above this happens against the background of Bu¨hler, Peirce, Jakobson – and empiricist linguist-philosophers who built on the insights of the former ones. The contributions by Abraham, Leiss, and Mache´ set the tune for the entire volume by emphasizing that exploring the Theory of Mind interface highlights the fact that the internal organization of the speaker’s Theory of Mind has an impact on the design of linguistic items he uses. A typical adult speaker is aware of the fact that the knowledge of the hearer usually di¤ers from his own knowledge and that, in order not to divert the hearer’s interest in the communication, he has to fathom out a satisfactory common ground and invite the hearer to pursue the exchange on this terrain. Expressions such as thinking or believing as well as making sure of and inviting the hearer’s concurrence reflect this basic organization of the human mind. This can be achieved in various ways: overtly or covertly. Overt means are lexical predicates or predicate modifiers, modal adverbials such as infelicitously, unfortunately, probably as well as grammatical means such as modal verbs and modal particles as well as many other patterns. Covert means involve the embedded infinitival (this is / has to be done; cf. Bhatt 2006). In the strict meaning of the word, covertness may also be involved with modal verbs to the extent that the homonymy between root and epistemic meanings, at least in the Germanic languages and foremost in German and Dutch, hides a crucial opposition between root and epistemic modality. Homonymy, reaching far beyond the latter modal opposition, holds foremost for modal particles, which are so very characteristic of the Germanic and the Slavic as well as the South-East Asian languages – not, however, of the Romance languages and English. The present book invokes these phenomena sketched above. The first main section introduces the grammatical and pragmatic categorial distinc-
Introduction: Theory of mind elements across languages
5
tions that emanate from the theories developed by Bu¨hler and Jakobson (and which form the basis for a better understanding of linguistic typologies of Far East Asian languages like Japanese, Thai, and Chinese; see Abraham 2012a,b, 2011; Tanaka 2011). The second main chapter falling into two subchapters collects contributions which exemplify the latter categories of modality by applying them to various categorial representations. Moreover, it assembles papers which compare overt modality expressing languages with others that do not sport an identical inventory of overt modality expressives and discuss verbals in discourse. In Section 2 this is highlighted in more detail. Before doing that, however, let us briefly sketch what the two key notions, Shifting and (Double) Displacement, are about. We may also ask for this purpose: What are the main Theory of mind categories in adult speech? To what extent is this about language (as the root to cognition) and not only about cognition (as the root of language)? Take Davidson’s (2001) threefold distinction of mind awareness (see also Leiss 2009, 2011 this volume) as a point of departure. Davidson’s ‘Triangulation model’ involves di¤erent degrees of certainty concerning the truth assessment of the proposition attached to the three persons in a communicative act. The first person is the speaker, the second person the addressee, and the third person is located outside the dialogue. Each of these participants possesses a di¤erent set of belief and knowledge systems (‘possible worlds’). The success of the communication will crucially depend on the degree of overlap between the di¤erent belief and knowledge systems – ideally total common ground. In other words, we need to be after the process of achieving communicative intersubjectivity and hence objectivity. According to Sperber and Wilson (1986) and Davidson (2001), speaker and addressee continuously negotiate the meaning and the truth-values by means of language. The result of this potentially infinite process of truth assessment is objective – i.e., commonly and ideally publically shared – knowledge (according to Popper objectivity is an aim we can close in on, but never completely achieve). (1) Davidson’s mind awareness triangle / ‘Triangulation model’:
6
Werner Abraham & Elisabeth Leiss
‘‘Foreign/3rd persons’ knowledge awareness’’ or, as we prefer to call it, Foreign Consciousness Alignment/FCA (Theory of other minds), may also be understood as the specific process toward the achievement of ‘‘social/ objective/intersubjective knowledge’’ – i.e., some essential common ground that has been achieved through (near-)public inclusion in the assessment. Our claim is now that none of the ensuing lexical modality categories – evidential adverbials / evAdv, epistemic adverbials / epAdv – meets the processing criteria of (Double) Displacement in terms of FCA, except modal verbs (¼indirect displacement, i.e. displacement without person split) and MPs (¼transitive displacement, i.e. person split). What evAdv, and epAdv do in terms of displacement is separate the speaker from the direct source of knowledge acquisition – a process that is not covered by the Triangulation model, but by simple (Jakobsonian) shifting of the origo (as with Reichenbachian (e,)r < s for the Past tense or s < (e,)r for the Future tense. Consider also the temporal yesterday or local here, where the origo is not locatable until the locus of the speaker is manifested directly). This distinction is reflected by the feature bundles attached to the three modal categories discussed in (2)–(5). (2) a.
epistemic modal verbs/EMVs denote twofold illocutive deixis: one toward the source of knowledge, and the other toward the degree of certainty of the speaker’s knowledge state with respect to p. The source of knowledge scoping over p is ‘‘sourced out’’, or shifted, to some third person (left branch).
b.
Haider soll betrunken gewesen sein (p ¼ Haider ist betrunken gewesen) (‘Haider was drunk’)
c.
(3) a.
epistemic adverbials/epAdvs unfold but a simple deixis – one where the Source branch is missing – cf. (1) above. No source of knowledge is ‘‘outsourced’’/shifted (left branch missing), just the degree of certainty of a statement about p is somewhat diminished.
Introduction: Theory of mind elements across languages
b.
7
Haider war wahrscheinlich/vermutlich betrunken. (p ¼ Haider war betrunken) ‘Haider was probably/presumably drunk.’ (‘Haider was drunk’)
c.
(4) a.
evidential adverbials/evAdvs likewise unfold a simple deixis – one where speaker deixis is missing. The source of knowledge is separated from the speaker and shifted to some unindexed foreign source.
b.
Haider ist o¤ensichtlich betrunken gewesen. (p ¼ Haider ist betrunken gewesen) ‘Haider has obviously been drunk.’ (‘Haider was drunk’)
c.
(5) a.
b.
modal particles/MPs denote a di¤erent deixis in the following sense: The speaker makes an estimate about the knowledge awareness of the Addressee while letting the Addressee know about this estimate and giving him a chance to relativize, or correct, this estimate about p. In this sense, the feature/s under Sp-assessment is/are richer than those for EMVs according to the source of p, and another one according to the speaker’s assessment of p. [CG ¼ common ground] Haider war ja betrunken. (with p ¼ Haider war betrunken) (‘Haider was drunk’) ‘Haider was PART! drunk.’
8
Werner Abraham & Elisabeth Leiss
c.
d. mind awareness displacement in ‘Triangulation model’ ¼ Foreign Awareness Alignment/FCA: Alignment in the sense that Speaker aligns his knowledge in accordance with the Addressee’s reaction (double point arrow).
MPs unfold a twofold deixis in the sense of (4c): There is no outsourcing (shifting of the origo) of knowledge (left branch missing). However, there is displacement of the responsibility for the truth of p from the speaker to the addressee, and, by the same token, there is the search for a common ground (double point arrows). The general insight is that MPs serve the highest theory of mind mark (speaker deixis mark ¼ Speaker’s assessment of Addressee’s assessment), this to the extent that MPs – and only these as opposed to MVs and, even more so, ep/ev-Adverbials – induce a (silent, covert) intercommunication between Sp and Addr without giving full lexical expression to this exchange. No other category does this. And, what is interesting typologically, it is not clear how a language does this in the same direct-deictic, discourse prominent, way that does not possess the MP-category (see Abraham 1991a,b; 1995a,b; 2012b this volume; Zimmermann 2008: §§3.1, 3.2, and 3.3). In a way, the essence of shifting and displacement and, consequently, the distinction between the di¤erent modal categories in terms of feature separation lies in the requirement to locate the origo(s).
Introduction: Theory of mind elements across languages
9
2.2. Typological issues Both Speaker split (origo split) and Speaker displacement may play a major role in those languages that do not employ anaphoricity in terms of personal pronouns and the reflexive, but refer to addressees and thirds in terms of direct reference and assign degrees of truth reliability also in declaratives. This happens, e.g., in Japanese (Tanaka 2011; see also Abraham 2012a,b). As for the latter, rich inventories of signals for reassurance – as well as consensus-asking modi – are inserted in declaratives. Moreover, (the Indo-European) anaphors (pronouns) are replaced, on a general line, by classifying, or near-proper name, nominals specifying the relation of the speaker to the addressee or third person. This, then, will also capture possessives. In what follows we sketch two languages that display classes of modal particles similar to – and yet so di¤erent from – German(ic), except for Modern English. Thus, also Japanese discourse particles in sentence-final position serve as relaters between a speaker’s and a hearer’s mind. Spoken Japanese is characterized by frequent usage of these sentence-final particles as in (6) and (7) (particles emboldened).3 (6) Kyoo wa ii tenki desu ne. today TOPIC good weather is NE ‘Today the weather is good NE?’ (7) Matte yo. wait YO ‘Wait (for me) YO!’ A raw guess of the meaning of yo and ne is this: (6) seems to imply some kind of agreement between speaker and hearer regarding the proposition expressed, while (7) is something like its opposite as it expresses the speaker’s guess that its proposition is unknown information for the hearer. Obviously both yo and ne verbalize something regarding a speaker’s estimation of the hearer’s mind and thus are ToM and modality related. Japanese (Tanaka 2011, among others), much like Chinese (Li 2006), possesses a noticeable amount of such particles which instantly remind one of modality particles in languages like German, whose linguistic system is notoriously known for its semantic-pragmatic complexity. 3. This section on Japanese discourse particles is due to advice and help by Eric Schanz, student of Japanese and German Linguistics at the University of Munich.
10
Werner Abraham & Elisabeth Leiss
Japanese sentence-final particles are a heterogeneous word class regarding their semantics. Modal functions in the sense of the German (or Dutch) modal particles are only part of them. To distinguish them from the rest I define them as follows: In Japanese a sentence-final particle is a modal particle, if and only if it verbalizes relations between the speaker’s mind and his estimation about the hearer’s mind regarding the proposition uttered. To implement this subset of the Japanese sentence-final particles and its functions we need to consider the following subtasks. (8) Matching of minds (common ground ): This is information about whether the speaker expects the hearer’s mind to accord with the proposition. (9) Interaction between proposition and hearer’s mind: This represents the speaker’s expectation of how the hearer should deal with the proposition in regard to his own knowledge and experience. (10) Intensity of expectation: Combined with the expectation criterion mentioned above there is a certain degree of relevance pressure which is laid on the hearer regarding its intended fulfillment. (11) Weight of relevance: It expresses how important the proposition presented by the speaker is for the current topic. This list of criteria, together with some additional minor subcategories, enables the linguist to sketch an extensive profile of each Japanese modal sentence-final particle securing a better understanding of modality in Japanese and as a phenomenon of the languages of the world. Notice that there is a considerable common line between (8)–(11) and Kratzer’s (Kratzer 1981, here 1991: 649f.) theoretical modal design developed for modal verbs in Germanic. We relate Kratzer’s system to (10) and (11) above. Ad (8)–(11) – with (8)–(9) clearly not represented at all in Kratzer’s classification: Modal force: weak and strong obligatoricity, good and weak possibility, at least as good as possible, better possibility; possibly others. Modal base: circumstantial vs. epistemic (and possibly other bases such as the state of knowledge about certain sources, facts of a definite content). Ordering source: deontic (commitments, requirements), bouletic (wishes, desires), stereotypes etc. Kratzer’s approach is mono-categorial to the extent that it does not consider other modal categories such as discourse particles (modal particles in Germanic) and adverbials. Modal base and ordering source of modal
Introduction: Theory of mind elements across languages
11
verbs are not directly compatible with those of discourse particles, nor are they compatible between all modal verbs. For example, epistemic modal verbs may have ordering sources of an informative source (such as German soll, which may refer to knowledge of normal orders of events, reports of others, positions of belief ), while semantically near-synonyms (such as muss) refer solely to personal inference. Circumstantial modal bases refer to ordering sources such as laws, common social goals and agreements, plans. Like Japanese, Cantonese Chinese displays a rich system of markers of epistemic modality and evidentiality in clause final position.4 The majority of these particles is not lexically encoded and has a very fuzzy semantics. As this Chinese dialect is a tonal language, it is highly restricted in its sentence intonation expressing the speaker’s intention. Therefore, the intention, or attitude, towards the sentence is bundled in these particles and some of them apparently express epistemic modality (12) and evidentiality (13), whereas others simply mark questions (14) or function as focus particles (15). Depending on the analysis, the amount of sentence final particles in Cantonese number varies between 15 and 90. (12) keoi5 heoi3 Baa3lai2 gwa3 3Sgl. Go Paris MP ‘(I’m not sure, but I think) He is visiting Paris.’ (13) keoi5 heoi3 Baa3lai2 wo5 3Sgl. Go Paris MP a. ‘(Someone told me that) He is visiting Paris.’ b. wo3: ‘(I didn’t want to believe it, but) He is actually visiting Paris.’ (14) keoi5 heoi3 Baa3lai2 ma1 3Sgl. Go Paris MP a. ‘Is he visiting Paris?’ b. a1ma3: ‘He’s visiting Paris. (Don’t you know that?)’ (15) keoi5 heoi3 Baa3lai2 ze3 3Sgl. Go Paris MP a. ‘He is only visiting Paris.’ b. ze1: ‘(It’s not the end of the world) He is only visiting Paris!’ 4. This section on Mandarin discourse particles is due to advice and help by Jerra Busch, student of Chinese and General Linguistics at the University of Vienna.
12
Werner Abraham & Elisabeth Leiss
Although research has been conducted on Cantonese sentence final particles in several linguistic fields, so far no succinct e¤ort has been made to distinguish between the terms ‘sentence final’ and ‘modal’ particle. Indeed, they appear to be used interchangeably. The problem in classifying them is grounded in the distinction between intonation and tone and the wealth of combinations. Thus, particle wo3 serves an epistemic function (13b), whereas the wo-particle with a fifth tone expresses evidentiality (13a). The same holds for other particles: Even those which seem to have no potential of modality can produce some kind of epistemic modality in combination with other elements (14a,b) and (15a,b). Recent linguistic work (Law 2002; Li 2006; Abraham 2010a) has tried to constitute the basic elements of sentence final particles and ascribe certain semantics to them. 2.3. The deeper analysis of German modal particles: felicity conditions German(ic) modal particles/MPs are specific for the speech act property of the speaker about the proposition p: executed, for example, by means of halt, eben, schon, ja, wohl, nur, bloß, doch in an utterance like (16). (16) Paul soll Maria halt / eben / schon / ja / wohl / nur / bloß / doch zu einer Aussprache einladen. ‘‘Paul wants to invite Maria modal particles to sort out certain issues.’’ With (16) the speaker asserts that Paul invites Maria to sort out things with him and that he modifies his assertion such that the hearer, or addressee, (17) a.
b. c.
may understand the modification with wohl as an assertion that may be considered as being of lesser degree of assertable evidence (such as of reportive status); will use doch to encode that his assertion is not in line with the hearer/addressee’s assessment; by using the MP ja, the speaker places his statement on the basis of generally valid presuppositions shared by the hearer. Statements including the MP ja cannot be answers to pure wh-questions such as in the sequence: A: Wen hat Erin geheiratet? ‘‘Whom has Erin married?’’ B: *Erin hat ja Chris geheiratet ,,Erin has part Chris married.‘‘;
Introduction: Theory of mind elements across languages
d.
13
by using the MP halt as well as eben, the speaker foregrounds a fact in his statement, which in a wider context is in fact backgrounded; e. by using nur ‘‘just’’ or bloß ‘‘only’’, respectively, he does not attribute a higher status of importance to his statement p in relation to expectations and valued weight linked to p; f. he wants to attribute to his statement with schon ‘‘already’’ the status of refutation and contradiction as against the addressee (see Egg (in this volume) in far more detail); g. by using ja ,,yes‘‘ in his statement p he concedes to the addressee’s reading of p a high status of the latter’s expectations and valuation of p. However, such a mono-layered analysis of modal/discourse particles has the following weaknesses: – The common ground/CG does not come to the fore as an analytical criterion. – A felicitous continuance of the discourse remains outside of the analytic description – and, thereby, of explanation. – One may rightfully ask whether mono-clausal MP-examples render the full meaning of MPs in dialogue sequences. – Do we not have to posit separate criteria accounting for the propositional semantics of MP clauses as opposed to dialogical sequences (for the simple reason that only in the latter clausal mode common ground conditions will bear out)? – See Abraham (1991b, 1995a) for a dialogical analysis accounting for the uses of the German MPs doch und DOCH (English approximately ‘‘though’’): A:
Er(¼B) hat doch RECHT!? vs. C: Er hat DOCH Recht! he has mp right he has mp right ‘‘He is right, isn’t he?’’ ‘‘He is right nevertheless!’’ B before A: Bush ist ein Kriegsverbrecher. ,,Bush is a war monger.‘‘ D1 before C: ##Bush ist ein Kriegsverbrecher. ,,Bush is a war monger.‘‘ D2 before C: C:
B hat Unrecht: Bush war kein Kriegsverbrecher. ,,B is not right: Bush was no war monger‘‘ Er hat DOCH Recht! Bush WAR ein Kriegsverbrecher. ,,He is right nevertheless! Bush WAS a war monger.‘‘
14
Werner Abraham & Elisabeth Leiss
– With reference to the dialogue above: with doch A seeks from B an at least partially valid CG: felicity condition 1. – In contrast, by using DOCH C rejects A’s attempt at reaching a partial CG. A not only posits his own assessment of p, but he also relinquishes any CG which includes D’s assessment. – See the mono-layered semantics of doch in (17b) above. A fully legitimate MP-analysis will also have to account for the dialogical usage as in D2 above. – An MP analysis including felicity conditions will have to be divided into two components of analysis: one of a mono-layer propositional semantics; and another one referring to the dialogical felicity conditions. 3. Theory of mind in linguistic typology, in discourse, and, yet more abstract, in terms of shifting and (double) displacing speaker and hearer in utterances of propositions Most generally, this volume o¤ers novel studies that lay the ground for an understanding of instantiations of speech modalization and speakerhearer’s modeling of communicative common ground and its processing in linguistic communication – a truly cognitive endeavor not undertaken at length and in depth before. The first main chapter ‘‘The foundation: speaker and hearer deixis, shifter, and double displacement’’ introduces to the volume. In Abraham’s ‘‘Illocutive force is speaker and information source concern. What type of syntax does the representation of speaker deixis require? Templates vs. structural derivation’’ the class of modal particles/MPs, a particular brand of weakly veridical adverbs in the Germanic languages except for Modern English, serve as a ToM-category katexochen in the sense that MPs encode both speaker/addressee deixis and source deixis (vide D. Davidson’s ‘Triangulation Model’; see also Coniglio 2009). In the light of the fact that MPs presuppose and modify the illocutionary force, root clauses are taken to be the best candidates for the insertion of MPs. Needless to say that not all MPs appear in all illocutive types: denn occurs only in true w- and yesno-interrogatives (not even in echo questions – see 18 below); ja occurs only in declaratives – see (19). (18) a. Wie siehst denn DU aus? . . . speaker-concerned question what look part you like b. Wie siehst du denn AUS? . . . true w-question c. *Du siehst denn WIE aus? . . . *echo question d. *Du siehst denn schlimm aus. . . . *declarative
Introduction: Theory of mind elements across languages
15
(19) a. Du siehst ja (SCHLIMM) aus. . . . speaker-concerned declarative you look part (bad) like b. *Wie siehst du ja AUS? . . . *true w-question c. *Siehst du ja SCHLIMM aus? . . . *yes-no question However, the question whether or not embedded sentences can host MPs is not well understood. See (20) yielding no clear clue. (20) a.
Mir scheint, dass er ja la¨ngst da sein mu¨sste. me deems that he part already here be should b. *Er leugnete, dass er ja la¨ngst da sein mu¨sste. he denied that he part already here be should c. Sie fragte, ob er denn kommen wolle. she asked if he part come would d. *Sie wusste nur allzu gut, ob er denn kommen wolle. she knew all too well if he part come wanted
(18)–(20) lead us to generalize that independent (root) clauses are MPcompatible by default as long as the speech act contingencies of the individual MPs are met. How about dependent clauses – do they principally disallow MPs? The minimal conclusion is that there are such dependent clauses allowing for MP-insertions. Consequently, we would have to impute to such dependent clauses the illocutive clause node ForceP. Other dependent clauses, however, will not possess their own illocution and, consequently, project no ForceP. See (21) (also Haegeman 2002; Coniglio 2009, in this volume). (21) a.
dependent core clauses/DCC: Subj –
b.
dependent peripheral
c.
independent clauses:
–
–
TopP FinP . . . *MP
clauses/DPC5:
Subj ForceP TopP FocusP TopP FinP . . . MP –
ForceP TopP FocusP TopP FinP . . .
MP
Notice the compatibility correlation between MP and ForceP. What is at stake is to separate (21a) and (21b) distributionally: There are MPcompatible dependent clauses – since possessing their own, independent 5. ‘‘Peripheral’’ embedding may be a special type of non-integrated subordination (see Reis & Wo¨llstein 2010).
16
Werner Abraham & Elisabeth Leiss
illocution which has not been pre-assigned by the governing independent clause; and there are MP-incompatible clauses which are not in the possession of independent illocutionary power. The paper attempts to give an answer to the question: What is the specific quality of dependent peripheral clauses such that they possess their own illocutionary power and, consequently, allow MPs to be merged in Spec,vP. These are the empirical prerequisites for a formal derivation which yields the basis for semantic interpretability: from surface to covert scope position. (22) MPs occur only in the middle field; they have to raise to the head of ForceP to enforce their illocutive power for the utterance of p. (23) Since the invariable linear order of MPs in combination falls into three classes motivated by their pre-particle categorial status, there must be three di¤erent landing sites in ForceP, i.e. MP1 < MP2 < MP3. (24) EMVs merge no lower than TP due to their finiteness-only status; they have to raise to ForceP to execute their illocutive power. DMVs merge in VP and raise due to agreement. (25) The hierarchical rise of speaker deixis features from E-adverbials to EMVs and further to MPs maps onto an equally hierarchically structured CP-expansion. Agreement checking for MPs for covert LF-movement should proceed on the basis of the feature assignments in hierarchically rising order: (26) for epistemic adverbials/epAdv: Sp-assessment , , (27) for evidential adverbials/evAdv: Source of p (28) for EMV, specifically epistemic sollen: Source of p as well as Sp-assessment , , (29) for MPs, specifically the MP2 ja: Sp-assessment , , as well as the split decision Recall that the feature bundle under ‘‘Sp-assessment’’ needs terminological tightening, let us say (partly following Bayer
Introduction: Theory of mind elements across languages
17
2008). This collapses speaker identification with 1st and 2nd persons vs. with 3rd person. Thus, ‘‘Sp-concern’’ summarizes and generalizes for the specific feature bundles. If we take MP-representatives of the linearly first class, MP1 for the derivative pattern, this is the covert derivation of denn and aber: (30) MP1 such as denn Interrog or aberDeclar : [ FinP/ForceP1 . . . Fino/Forceo [ . . . [ ParticleP denn/aber [VP . . .]]]] move under agree % [ FinP/ForceP1 . . . Fino/Forceo denn/aber [ . . . [ParticleP denn/aber [VP . . .]]]] The landing sites for MP2 and MP3 are correspondingly lower in Force/MP2 and ForceP/MP3. EMVs land in the somewhat lower positions of EvidMV or EMV licensed by their corresponding speaker-deixis features (without , the latter being characteristic only of MPs). And, finally, epistemic adverbials, EpAdv, land even lower due to their weaker Sp-deictic predicaments. Abraham strikes three thematic chords within Theory of Mind or, as he prefers to call it, Foreign Consciousness Alignment/FCA in the discussion of modality: First, illocutive force represents the speaker’s concern about the source of the information – this is a concern expressed by epistemic modal verbs and the shift of the origo (Bu¨hler’s I – here – now); second, illocutive force represents the speaker’s concern about the degree of contextual common ground between him and the addressee in the interest of a reasonable and informative communication typical of modal particles/ MPs; and, third, the question is to what extent the parameters of FCA apply to dependent clauses as much as to independent/root ones. The author pursues this line of arguing through the paradigm of German modal particles in some detail bringing to light the essential di¤erences between grammatical and lexical expressives of modality. He shows that the representation of speaker and hearer deixis requires a modern derivative syntax for the simple reason that the expressives in case, MPs, surface in clausal positions not identical with their semantic-illocutive operator functions. Crucially, the separation of the speaker, in his dependence upon Bu¨hler’s origo manifestations, and the grammatical subject pervades the entire discussion. Modal verbs/MVs appear to be less complex as compared to MPs since, in the former, the hearer obtains no participating role in the assessment of the truth. With epistemic modal verbs/EMV, truth assessment is located inside the speaker (‘intransitive assessment’). By contrast, truth assessment of MP-
18
Werner Abraham & Elisabeth Leiss
modified p is ‘transitive’ in the sense that the speaker negotiates the truth value of p with the hearer. Modal adverbials are still less complex against such FCA-backgrounds. The central aim of Leiss’ ‘‘Epistemicity, evidentiality, and Theory of Mind (ToM)’’ is to take a close look at the axiomatics on which theoretical claims about modality are based. The author’s approach is a universal one, however, not based on a Rationalist framework, but on the functional axiomatics adhered to by Roman Jakobson and Karl Bu¨hler. The central hypothesis of Leiss’ contribution is that the development of Theory of Mind/FCA depends on the development of linguistic categories, and not vice versa, thereby supporting Papafragou (2002) and Papafragou et al. (2007).6 The special focus is on modality, especially on epistemic modal verbs. Leiss emphasizes that the linguistic architecture of epistemicity depends largely on the development of the grammatical category of person, which is defined as a shifter in Jakobsonian terms. The reference of personal pronouns shifts with the origo of the speaker. It is of central importance to understand how shifters serve as building blocks for complex grammatical categories such as tense and modality. Both categories involve double displacement and the split-up of the speaker into multiple person viewpoints with di¤erent functions. The features of aspectual, temporal, modal, and modality related shifting are analytically outlined with respect to their common denominator as well as their respective distinctive qualities. The focus is on the grammatical categories of evidential and epistemic modality, which are described as functionally equivalent. However, this does not hold for lexical epistemics and lexical evidentials. Special emphasis is laid on the argument that lexical evidentiality and lexical epistemicity (expressed by evidential and epistemic adverbials) are not equivalent to their grammatical counterparts, the latter being more complex in terms of deixis and shifting qualities. In a general outlook, the psychological approach to ToM/FCA by Papafragou and the theoretical approach of Abraham & Leiss (2008, 2009) will be compared against the background of philosophical approaches to language and cognition, where language is made responsible for the development of all higher forms of human cognition (based especially on Davidson 2001 and Karl Popper 1995). In sum, Leiss attempts to show that a functional approach to modality derives from an Un-Cartesian, i.e. non-Rationalist design of Universal Grammar. ‘‘Exploring the Theory of Mind interface’’ by Mache´ posits that most languages exhibit quite a lot of expressions that enable the speaker to 6. For full bibliographical citations which cannot be found in the References of this Introduction see the respective References of the individual contributors.
Introduction: Theory of mind elements across languages
19
signal to other interlocutors to what extent (s)he is committed to the truth value of the sentence (s)he utters. Such expressions are often simply called epistemic modifiers. In most cases these types of epistemic modifiers have non-epistemic counterparts as in (31) and (32) (following Sweetser 1990: 70). (31) John came back, [because he loved her]. (32) John loved her, [because he came back]. In its original use because refers to real world, or event chain, causality. Accordingly, the adjunct clause in (31) provides the reason for the action expressed by the matrix clause. In its epistemic interpretation, however, as in (32), the adjunct clause displays the reasoning or premise that leads the speaker to assume the truth of the action denoted by the matrix clause. The matrix clause has to be considered as an assumption rather than an assertion. The epistemic because-sentence provides two types of information: First, it labels the modified clause as a mere assumption of the speaker or premise to the apothesis, and, second, it exhibits the reason that caused the speaker to assume the content of the matrix clause. There are a couple of other modifiers that work in similar fashion: modals (e.g., Sweetser 1990), conditionals (e.g., Kratzer 2012), or locative modifiers (Maienborn 2004: 162). According to Sweetser (1990), linguistic structures do not contribute any information on their own that reveal the speaker’s knowledge. Grammar does not refer to the beliefs and attitude of the speaker. Instead, Sweetser (1990: 11, 59, 146) suggests that the human mind is divided into two domains: on the one side, an external domain that processes information belonging to the socio-physical world and, on the other side, an internal one that is devoted to the epistemic world. According to Sweetser, each linguistic expression can be translated into an abstract image-schematic structure. Sweetser (1990) now claims that interpretations will change depending on the domain in which the image-schematic structure will be processed. Whereas the internal domain provides epistemic interpretations, the external one renders non-epistemic, i.e., root readings. Therefore, she concludes that meaning relationships cannot be understood independently of human cognitive structure, including the metaphorical and cultural aspects of that structure. Whereas Sweetser (1990) claims that epistemicity is merely a cognitive phenomenon that is external to grammar, the position in the present paper is that epistemicity is ubiquitous or pervasive in linguistic structures at least to some extent. First of all, epistemic modifiers seem to form an independent syntactic class: All of them permit the modification of stative
20
Werner Abraham & Elisabeth Leiss
predicates lacking a situation argument (individual level predicates) as well as clauses that refer to an event prior to evaluation time. This is on a par with Hacquard (2006) who showed that root modals are evaluated with respect to the embedded event. Mache´ shows that Sweetser’s (1990) insights can be recast in terms of argument structure. This enables one to dispose of Sweetser’s stipulated concept of distinct mental domains. Linguistic structures then involve specifications that indicate at the interface how they should be interpreted: epistemic or non-epistemic. Mache´ first draws a similar distinction as Haegeman on what she calls event-combining conditionals vs. premise conditionals: Only the first ones argue an event causality between the two clauses, whereas the latter involve a discourse argument clearly brought in by the speaker. The author then pursues the sweeping argument that epistemicity (¼shift of the speaker’s origo) is a ubiquitous factor in all sentential modi independent of extra signals to this extent. Clearly, this line of thinking, if pursued more radically, takes one to the position of Japanese grammarians: namely that the speaker’s truth assessment perforates all modi of thinking, i.e. even assertive (reminiscent again of the possible world semantics; see Abraham 2011a,b, this volume, on Bu¨hler and Japanese grammarians). Mu¨ller’s ‘‘The distribution of knowledge in (un)acceptable questions’’ provides a seminal discussion of what sentential modi require in terms of sincerity and felicity conditions. As in the case of Egg’s contribution, Mu¨ller is concerned with the fundamental steps that question raising presupposes, and with the question of exactly which components of uttering conditions of the speaker and reacting presuppositions on the side of the hearer may be responsible for failures. By basing questions on the set of possible answers and their type of speech act and content Mu¨ller makes an exemplary attempt at surveying all aspects of felicitous and failing question types employing the instrumentarium that was developed in the beginning of this volume: processes used and calibrated in the Theory of other minds / Foreign Consciousness Alignment as well as Bu¨hler-based shifting of the origo and (double) displacement or person split. In more detail, this contribution pursues the idea that the distribution of knowledge between discourse participants has an impact on the overall acceptability or the availability of certain readings of questions. In this study, di¤erent knowledge states get linguistically encoded by (non-)(anti-)factive matrix verbs in extractions from that-complements, (di¤erent types of ) modal verbs in complex and simpler questions, and in the occurrence of modal particles in that-complements embedded under verbs of saying in extraction constructions . A more general implication of the analysis is the claim
Introduction: Theory of mind elements across languages
21
that violating pragmatic rules in illocutionary acts through linguistic encoding can a¤ect the well-formedness and/or the acceptable interpretations of linguistic expressions. In pursuing all of this the author takes as a point of departure classical speech act theory and what it has attempted at formulating the conditions that are necessary and su‰cient for a particular illocutionary act to be successfully and sincerely performed. Searle (1969: 66) formulates (33.1, 2) below in order to account for the structure of the illocutionary act question. (33) Preparatory rule: 1. S does not know ‘the answer’, i.e. [. . .] does not know the information needed to complete the proposition truly [. . .]. 2. It is not obvious to both S and H that H will provide the information at that time without being asked. Sincerity rule: S wants this information. Essential rule: Counts as an attempt to elicit this information from H. Assuming the di¤erentiation of such types of rules when it comes to carrying out certain illocutionary acts during communication, Searle (1969: 63) points at the analogy of rules formulated for moves in games in order to control and guarantee their correct course: If we ask ourselves under what conditions a player could be said to move a knight correctly, we would find preparatory conditions such as that it must be his turn to move, as well as the essential condition stating the actual positions the knight can move to. There are even sincerity conditions for competitive games, such as that one does not cheat or attempt to ‘throw’ the game. Transferring this analogy back to (33), uttering a question during communication thus leads one to assuming that the adequate application of this move depends on the situations set by the di¤erent rules in (33). As becomes obvious from looking at the formulations in (33) in detail, the distribution of knowledge as far as the discourse participants’ knowledge states are concerned plays a crucial role in determining the language game’s rules on the adequate usage of a sentence carrying out this illocutionary act: It is assumed that the speaker does not know the answer, s/he presumes that the addressee knows it and poses the question in order to elicit this information from the addressee. Especially the assumption on the speaker’s knowledge state as formulated in the first part of the preparatory rule has also found its way into other formulations of the pragmatic conditions having to be met when a
22
Werner Abraham & Elisabeth Leiss
question is asked in discourse (cf. 34 taken from Bu¨ring 2003: 5; cf. also e.g. Reis 1990: 51; Caponigro and Sprouse 2007: 9). (34) Informativity: [. . .], don’t ask for known things! By focussing on this (pre)condition for a questioner’s adequate move in discourse when taking part in communication by uttering an ordinary (Caponigro and Sprouse 2007: 9) (i.e., information-seeking) question (cf. Searle’s 1969: 66; Reis 1990: 51), Mu¨ller looks at the interaction of question formation and the occurrence of certain linguistic material in the questions, namely (non-) (anti-)factive verbs, modal verbs as well as modal particles. Due to this linguistic material’s meaning components, a speaker who chooses it to appear in his utterance makes a statement on the assignment of knowledge concerning the persons participating in the current discourse which might (not) clash with the assumption concerning the questioner’s knowledge state as formulated e.g. in a condition such as (33.1) and as being expressed by him posing a question and, thereby, carrying out the question act. The overall idea pursued in Mu¨ller’s contribution is that violating the condition on the questioner’s knowledge state has a negative impact on the acceptability of questions, the linguistic material serving the purpose of manipulating the circumstances regarding the distribution of knowledge between speaker, addressee, and possibly third instances. On the one hand, this assumption is directly mirrored in a conceivable decrease of a question’s acceptability. On the other hand, it is shown that this assumption can explain why only certain readings are available for certain questions. While pursuing the aim to find evidence for the idea that a condition on illocutionary acts such as 1. in (33) does have an impact on the wellformedness of sentences intending to carry out this act, the study indirectly also checks assumptions made on the interpretation of the linguistic material made use of in order to test for this hypothesis. The second main chapter ‘‘Instances of deixis and origo in sundry languages’’ starts out with Coniglio who shows in ‘‘Modal particles, speakerhearer links, and illocutionary force’’ that German modal particles constituting a small class of adverbial elements (such as ja, wohl, schon, etc.) display peculiar properties that distinguish them from adverbs in a traditional sense. The author considers their main syntactic properties, i.e. their distribution in the Mittelfeld (‘middle field’) of root and subordinate clauses and the conditions under which they are licensed there. In particular, he argues that modal particles are adverbial elements which function as modifiers of illocutionary force and that they therefore presuppose the
Introduction: Theory of mind elements across languages
23
presence of illocutionary force in the sentence they occur in. Given that they are root phenomena, one would expect to find them only in root contexts (i.e., in full CPs). However, Coniglio shows that they can also be found in certain types of subordinate contexts. The author provides an explanation of this fact by taking into account some recent work by Haegeman (2002, 2004a,b, 2006) on the di¤erent structure of the left periphery of root and subordinate clauses. In particular, the author starts from Haegeman’s assumptions that certain types of subordinate clauses display a full CPdomain, thus being endowed with the projection ForceP, whereas other types of embedded clauses present a truncated structure of the left periphery and have therefore no projection encoding illocutionary force (represented by ForceP). The author claims that German modal particles can only occur in contexts which are endowed with the full structure of the CP, thus showing that modal particles are to be considered as fundamental criteria for detecting the independent illocutionary force of clauses irrespective of whether they are taken to be independent or embedded. Egg’s contribution, ‘‘Discourse particles at the semantics-pragmatics interface’’, is a follow-up and more generalized version of his Lisbon presentation ‘‘Discourse particles in rhetorical questions’’, which we will deal with briefly before going into his present article. As the title says the former deals with discourse (¼modal) particles in rhetorical questions (RQs). Both RQs and discourse particles crucially refer to the common ground, and, since discourse particles occur frequently in RQs, one would expect there to be some interaction on this point. The author not only shows that it is possible to account for the way in which RQs and discourse particles interact, but he even succeeds in explaining some incompatibilities between RQs and discourse particles in terms of conflicting ways of referring to the common ground. Consider (36) and (37) as possible ways of reacting to (35). (35) Wir haben keine Bewerbungen fu¨r die Stelle in Vechta bekommen. ‘We got no applications for the position in Vechta.’ (36) Da wu¨rde auch/eben niemand hinwollen. ‘No one would want to go there.’ (37) Wer wu¨rde da auch/*eben hinwollen? ‘Who would want to go there?’ Although (36) and (37) have the same discourse function (utterance of negative statement), the discourse particle eben may only occur in the syntactically declarative (36), but not in the RQ in (37). This is di¤erent
24
Werner Abraham & Elisabeth Leiss
for the discourse particle auch, which is compatible with both possible answers. Rhetorical questions/RQ are not ordinary, information-seeking questions. In cases like (37), this follows from the fact that the question is extremely general (the answer would be a list of all the people that would want to go to Vechta), clearly not a piece of information that the speaker wanted to get. This seems to violate felicity conditions for questions (Searle 1969), e.g. the sincerity condition (speakers want to have a specific piece of information; see 33 above). However, RQs are used in indirect speech acts, where a ‘direct’ speech act refers to a felicity condition of the intended speech act. This explains why they do not seek information: In such speech acts, questions are evaluated against the common ground. The intended speech act for (36) is a statement: A preparatory condition of a statement is that it is not obvious for the speaker that the hearer already knows what is being stated, and the speaker can refer to this condition with a rhetorical question. When an RQ is evaluated against the common ground, the hearer recognizes that only one element of the set of possible answers is compatible with the common ground, for wh-RQs like (37), the statement that negates the existence of a suitable entity for which the question holds, e.g., for (37): No one wants to go to Vechta in any world, i.e., statement (36). RQs thus emerge as a means of presenting a statement not as the speaker’s personal opinion, but as a consequence of the common ground, which explains their persuasive e¤ect (Egg 2007). In his present contribution, Egg, following Karagjosova (2003, 2004), assumes that the meaning of discourse particles can be described in terms of their interaction with the common ground (CG). Consider the pair (36) þ (37) with eben. Intuitively speaking, the first sentence expresses surprise, which is mitigated by pointing out that in the light of the second sentence, the first sentence points at something that is to be expected. Formally, this can be rendered as (38). (38) In a sequence ‘p - eben q’, ‘eben q’ indicates that Bp can be deducted (non-monotonically; see Asher and Lascarides 2003 for details) from the common ground CG, then CG is updated with q, and p can be deducted from CG A q. For (36) þ (37) with eben, this means that receiving no applications for the position in Vechta is strange, because the CG would suggest that there should be some. But as soon as one integrates the proposition that no one
Introduction: Theory of mind elements across languages
25
would want to go there into the CG, the CG suggests the contrary, viz., that there should be no applications. The account for the pair (36) þ (37) with auch is similar: The speaker of the first sentence thinks that the proposition uttered is not yet in the CG (otherwise he would not have uttered it), and the speaker of the second sentence then points out that the proposition of the second sentence is in the CG, which would allow deriving the proposition of the first sentence as part of the CG (see also Thurmair 1989; Weinrich 2005). Formally: (39) In a sequence ‘p - auch q’, the response ‘auch q’ points out that q and q > p are in the CG, which allows the deduction of p in the CG (‘>’ stands for defeasible implication). For the pair (36) þ (37) with auch this means that receiving no applications for the position in Vechta can be deducted from the CG, because it comprises the propositions that no one wants to go there, and that if you do not want to go to there you do not apply for a position there. This first proposition (that there are no applications) is then integrated in the CG. This formalization presupposes that the CG is not closed under deduction. As regards the interaction of RQs and their discourse particles, Egg considers what happens in the sequence (36) þ (37) with auch. The RQ indicates that the sole possible answer to (35) (that no one would want to go there) is in the CG. The e¤ect of auch then is to confirm this and to state that the defeasible implication from the answer to the first part of the sequence also is in the CG (i.e., that the fact that no one wants to go to Vechta normally entails that there will be no applications for a position in Vechta). This means that (36) can be derived from the CG; it is finally added to the CG. For the sequence (36) þ (37) with eben the RQ once again indicates that the only possible answer to (35) is in the CG. But this then clashes with the requirement of eben that the opposite of this answer can be deduced from the CG. The division of labour between RQs and discourse particles from a dynamic perspective is thus that RQs ‘test’ the CG while discourse particles update it. Egg’s contribution is an excellent in-detail demonstration of the important role that criteria of common ground (the knowledge state that speaker and hearer share or not share) play in order to arrive at an explanatory analysis of modal particles in German. Furthermore, it shows how important it is to divert one’s interest from merely investigating declarativesMPs – this to the extent that declaratives hide the assertive illocutive force, while other speech acts – in the author’s case rhetorical questions –
26
Werner Abraham & Elisabeth Leiss
a priori open the sentential structure to the input of the hearer and his distinct cooperation to reach a common ground (very much reminiscent of Zimmermann 2004, 2008). ‘‘Modality in the Romance languages: modal verbs and modal particles’’ by Meisnitzer takes a thorough look at modality and epistemicity in Romance languages. The author finds it surprising that if the grammar is written by a national author, modal verbs/MV, more than often, will not even be mentioned as a distinct lexical category. Likewise, if one compares such grammars, the amount of verbs listed as MV will vary significantly. In modern linguistics it is generally suggested, on the one hand, that MVs must provide an epistemic reading as well as a root interpretation, further that they have to be combinable with other modal verbs, that they must have a complete verbal paradigm regarding their inflection, and that, in general, they can be used as full verbs (polyfunctionality). On the other hand, when they are used in the infinitive form they cannot have an epistemic reading (Abraham 2008). This forms the basis for showing that Span./Port./Cat. poder, Fr. pouvoir; It. potere; Roum. a putea is the only real modal verb across the Romance languages. Moreover, the author considers Port. dever, Sp. deber, Cat. deure, Fr. devoir, It. dovere and Rum. a trebuie to be still caught in di¤erent stages of their grammaticalization process across the Romance languages. Regarding modal particles/MPs, Romance linguists generally argue that they are specific of German(ic), although there seems to be some evidence for grammaticalization processes leading to modal particles in Romance as well (see Waltereit 2006; Waltereit and Detges 2008) varying tremendously not only in quantity, but also in form. Bearing in mind that Romance languages have no structure which could be compared with the German (and Dutch or Scandinavian) ‘‘Mittelfeld’’ or its syntactic-semantic discursive function, the author tests whether there is any Romance adverb or MP which meets the defining criteria of this category: foremost, whether they meet polyfunctionality, whether they occur in specific positions in the sentence, whether they are limited to occur with specific illocutions, and, finally, whether or not they are restricted to root clauses (and to their illocutionary force). Modal particles and the specific polyfunctionality of modal verbs in German(ic) form the basis for this volume’s discussions and methods of analysis in modality. Meisnitzer investigates to what extent, quality- and quantitywise, the Romance languages share these German(ic) paradigms of modality. Detecting only a very small share his conclusion is that the pertinent lexemes may well be on their diachronic paths toward equal
Introduction: Theory of mind elements across languages
27
functions of modality, but that they lag behind in achieving this goal not only across the di¤erent Romance languages, but also as individual lexemes. ‘‘The epistemological treatment of information and the interpersonal distribution of belief in language: German modal particles and the typological challenge’’ by Modicom discusses the links between the di¤erent domains of the epistemological treatment of information in speech and discourse. It is based on a confrontation of concepts coined by Guentche´va and Landaburu (2007) with a set of German modal particles. According to Guentche´va and Landaburu, Theory of Mind is part of a wider domain of the ‘‘epistemological treatment of information’’ together with mirativity and evidentiality. The concepts created by typologists are used to describe the distribution of modal functions between a few German modal particles examined in pairs. Both components of each pair have the same value in terms of Theory of Mind. The analysis makes the case for an internalist, monosemantic description of the meaning of those particles. Their di¤erent values in context are postulated to be inferable from the combination of their core meanings activated together with other syntactic and illocutionary parameters. It is shown that the complex meaning of German modal particles is best accounted for in a triangular framework including Theory of Mind, mirativity, and evidentiality as the permanent parameters followed by the speaker to track the distribution of knowledge between himself and the hearer (i.e., their common ground). Modicom goes an empirical step beyond Meisnitzer in considering pairs of modal particles in German. His conclusion is that categories and methods developed in linguistic typology – such as mirativity, evidentiality, and Gricean maxims for establishing the required common ground – are adequate methodological means to apply to the formal analysis and semantic explanation to modality as evidenced by MPs. de Schepper and de Hoop’s ‘‘On mood, evidentiality, and person e¤ects’’ takes as a point of departure the core function of language, i.e. to describe the world in order to convey information from the speaker/writer to the hearer/reader. More to the point, this paper explores this interaction by investigating mood, or sentence marking, and person. Thus, (40) can be taken to describe the state of a¤airs such that (at least according to the speaker) there is a possibility that he (whoever he refers to) will try to put the key into this slot. (40) He might try to put the key into this slot. (41) You might try to put the key into this slot.
28
Werner Abraham & Elisabeth Leiss
By contrast, (41) is read as a suggestion or hortative by the speaker conveyed to the addressee to put the key into the slot. Hence, the sentence is no longer meant to convey (objective) information about the world, but rather to instruct the addressee. This relates to another core function of language, namely to persuade, or to influence, another person’s behaviour or opinions. This is the interactive function of language (Searle 1969). The two persons who are necessarily present in face-to-face communication are the speaker and the addressee. First and second person are the most important, therefore, but at the same time the most trivial or redundant. Consequently, in many languages, pronouns can be omitted when they refer to first and second person subjects. Even in English, the addressee is understood but not expressed in imperative constructions (Close the door, please!) and sometimes in questions (Need any help?), while the speaker can remain implicit in diary language (Went to the market yesterday). Two types of sentences typically involve the addressee in an interactive way: interrogative and imperative sentences. Omotic languages are wellknown for their rich mood and modality marking on their verbs. In Sheko, one finds many sentence type (‘mood’) markers, yet the imperative as well as the interrogative lack mood/modal marking, while declaratives can be marked in di¤erent ways: as realis, irrealis, mirative, negative, viewpoint etc. The imperative as well as the interrogative are the types of utterances in which modal distinctions are the least required. In Dime, another Omotic language, interrogative clauses treat the second person subject as distinct from first and third person (2nd person marking –a´a´ versus the rest –ı´ ). Finally, in Kooreete, one finds a di¤erence in marking between declarative and interrogative/imperative for first and third person, but not for the second person. This paper demonstrates how the distribution and use of 1st and 2nd person pronouns – Bu¨hler’s main representatives of origo and its shifts and displacements – provides more insight into how the communicative context influences grammar. Although modal readings are known for their context-dependency (Kratzer 1981, 1991), the influence of person (1st, 2nd, 3rd) on the type of modal marking has not been systematically investigated before. The first main chapter closes with Endo’s ‘‘Illocutionary force and modal particle in the syntax of Japanese’’. The author takes as a point of departure Abraham (2011a, this volume), who shows that the core property of German MPs is to make an appeal to the addressee to cooperatively confirm or correct the belief assumptions signaled by the speaker, and that
Introduction: Theory of mind elements across languages
29
this property is captured by the derivation in which MPs move to the head of ForceP to enforce their illocutionary force. Along this line, Endo discusses movement properties of Japanese MPs from the perspective of comparative syntax. He first examines some basic properties of Japanese MPs and shows the derivation in which MPs move into ForceP. The author continues by examining some consequences of movement of MPs paying special attention to the semantic interpretations of adverbial clauses developed by Coniglio (2008). Finally, he brings in Miyagawa’s (2010) recent ideas on the EPP from the viewpoint of MPs. Endo’s contribution highlights a crucial di¤erence between interpersonal MPs (involving speaker and hearer) and speaker-oriented MPs that can be found in some syntactic properties related to topic and focus. Interpersonal MPs can su‰x to, and are compatible with, a focus particle like dake ‘only’, while speaker-oriented MPs are not compatible with, and may not su‰x to, such a focus particle. (42) a.
b.
John-ga kita-dake-yo/ne/sa. John-NOM came-only-MP ‘John only came.’ *John-ga kita-dake-wa/ze. John-NOM came-only-MP ‘John only came.’
The general pattern one observes with MPs is that interpersonal MPs involving both a speaker and a hearer tend to allow the quantified subject ‘all’ to scope under negation more easily than MPs expressing only the speaker’s mood (¼speaker-orientated MPs). What could be the second subsection of the 2nd main chapter, ‘‘Verbal epistemicity in discourse’’, features Je˛drzejowski’s ‘‘What is it that keeps the rein on quotative modals so tight? A cross-linguistic perspective’’. The author pursues the question how modal verbs with quotative meanings richly employed in German(ic) are typically translated into Slavic languages which do not possess a paradigm of personal modal/epistemic verbs. Modal particles, epistemic or quotative modal verbs, as well as epistemic adverbs, di¤er in their deictic feature distribution. Whereas quotative modals can denote a twofold deixis, epistemic and/or quotative adverbs cannot. Considering some German-Polish contrastive data, the paper shows to what extent these assumptions are confirmed or need to be challenged. By bearing in mind the speaker-hearer link and focusing especially on quotative markers such as German sollen ‘be said’ and wollen ‘claim’, it is shown
30
Werner Abraham & Elisabeth Leiss
that quotative modals and epistemic/evidential adverbs are not on a par with each other, even if the latter seem to rank higher in speaker/addressee deixis. The author points out that, as has been commonly assumed by many scholars, Polish has at its disposal the quotative verb miec´ ‘have’, which is supposed to function as an e‰cient counterpart to the German quotative sollen as a marker of hearsay (cf. Hansen 2001, 2004). However, it is demonstrated that this equivocation is confusing and risky, especially since miec´ is not as grammaticalized as its German equivalent and since it does not always pertain to a foreign instance or source institution in the same way as German sollen does. This contribution by Je˛drzejowski continues the issues raised by Meisnitzer and Modicom above, i.e., on the one hand, the empirical discussion of (near-)equivalents to German modal verbs and modal particles and, on the other hand, their paradigmatic assemblies and overall unifying properties as compared to languages that do not display the same rate of overt paradigmatic representation of these modalities. In essence the author’s discussion touches on covert means to express modality variants what German does with paradigmatically ordered lexical. Slavic hardly has any modal verbs to the extent that Germanic displays them polyfunctionally and with regular root vs. epistemic readings. This triggers the question how Polish, in the concrete case, has registered variants of modality in the verbal domain with any near-equivocation to Germanic and whether or not, and how, recent developments in the oral-only usage of the language fill this apparent gap. References Aboh, Enoch Olade´ 2006 Complementation in Samaraccan and Gungbe: The case of Ctype modal particles. Natural Language & Linguistic Theory 24 (1): 1–55. Abraham, Werner 1989 Vorbemerkungen zur Modalpartikelsyntax. Linguistische Berichte 118: 443–465. Abraham, Werner 1990a Die Grammatikalisierung von Auxiliar- und Modalverben. Beitra¨ge zur Geschichte der deutschen Sprache und Literatur 112: 200–208. Abraham, Werner 1990b Modalverben in der Germania. In: Marga Reis (ed.), Akten des VIII: Internationalen Germanistenkongresses. Sektion Kontrastive Syntax, 79–98. Mu¨nchen: Iudicium.
Introduction: Theory of mind elements across languages
31
Abraham, Werner 1991a Modalverben in der Germania. In: Eijiroˆ Iwasaki (ed.), Begegnung mit dem ,,Fremden‘‘: Grenzen – Traditionen – Vergleiche. Akten des VIII. Internationalen Germanisten-Kongresses in Tokio 1990, Bd. 4, 109–118. Mu¨nchen: Iudicium. Abraham, Werner 1991b Discourse particles in German: How does their illocutive force come about? Between a maximalistic and minimalistic position. In: Werner Abraham (ed.), Discourse particles. Descriptive and theoretical investigations on the logical, syntactic and pragmatic properties of discourse particles in German and English, 220– 272. (Pragmatics & Beyond, New Series 12.) Amsterdam: John Benjamins. Abraham, Werner 1991c The grammaticalization of the German modal particles. In: Elizabeth Closs Traugott & Bernd Heine (eds.), Approaches to grammaticalization, Volume 2, 331–380. Amsterdam: John Benjamins. Abraham, Werner 1995 Wieso stehen nicht alle Modalpartikel in allen Satzformen? Zeitschrift fu¨r deutsche Sprache 23 (2): 124–146. Abraham, Werner 2010 Illocutive force is speaker concern about the source of information and information sharing. Anchoring multiple speakeraddressee deixis in syntax. Paper and handout for the MIC conference Paris Nov. 18–19, 2010. Abraham, Werner 2010a Diskurspartikel zwischen Modalita¨t, Modus und Fremdbewusstseinsabgleich. In: Theo Harden & Elke Hentschel (hg.) 2010. 40 Jahre Partikelforschung, 33–77. Tu¨bingen: Stau¤enburg. Abraham, Werner 2011 Fremdbewusstseinsabgleich in Syntax und Semantik. Ms. Univer(to appear) sities of Vienna and Munich. Abraham, Werner 2012a (Inter)Subjektivierung oder Fremdbewusstseinsabgleich als synchrone und diachrone Wandelbegri¤e? Begri¤sscha¨rfung und theoretische Einordnung. In: Jahrbuch der Gesellschaft fu¨r Germanistische Sprachgesellschaft Band 3. Berlin: De Gruyter. Abraham, Werner 2012b Sprecherdeixis und Merkmaldistributionsdi¤erential deutscher Modalita¨tselemente. Zeitschrift fu¨r deutsche Sprache 40: 1–24. Abraham, Werner & Elisabeth Leiss (eds.) 2008 Modality-aspect interfaces: Implications and typological solutions. (Typological Studies in Language 79.) Amsterdam: John Benjamins.
32
Werner Abraham & Elisabeth Leiss
Abraham, Werner & Elisabeth Leiss (eds.) 2009 Modalita¨t. (Studien zur deutschen Grammatik 71). Tu¨bingen: Stau¤enburg. Asher, Nicholas and Alex Lascarides 2003 Logics of conversation. Cambridge University Press, Cambridge. Averintseva-Klisch, Maria 2009 Rechte Satzperipherie im Diskurs. Die NP-Rechtsversetzung im Deutschen. (Studien zur deutschen Grammatik 74). Tu¨bingen: Stau¤enburg. Bayer, Josef 2008 From modal particle to interrogative marker: A study of denn. Ms. University of Konstanz. Cinque, Guilelmo 1999 Adverbs and functional heads. A cross-linguistic perspective. New York: Oxford University Press. Coniglio, Marco 2009 Deutsche Modalpartikeln in Haupt- und Nebensa¨tzen. In: W. Abraham and E. Leiss (eds.), Modalita¨t: Epistemik und Evidentialita¨t bei Modalverb, Adverb, Modalpartikel und Modus, 191– 221. (Studien zur deutschen Grammatik 77.) Tu¨bingen: Stau¤enburg. Davidson, Donald 2004 Subjektiv, intersubjektiv, objektiv. Translated by Joachim Schulte. Frankfurt am Main: Suhrkamp (English original 2001: Subjective, Intersubjective, Objective). Diewald, Gabriele 1991 Deixis und Textsorten im Deutschen. (Reihe Germanistische Linguistik 118.) Tu¨bingen: Niemeyer. Egg, Markus 2009 Discourse particles in rhetorical questions. Talk SLE 2009 Annual Conference Lisbon. Embick, David 2004 On the structure of resultative participles in English. Linguistic Inquiry 35 (3): 355–392. Endo, Yoshio (this volume) Illocutionary force and modal particle in the syntax of Japanese. 2011 In: Werner Abraham & Elisabeth Leiss (eds.) Theory of mind elements across languages. Berlin: de Gruyter. Ernst, Thomas 2001 The syntax of adjuncts. (Cambridge Studies in Linguistics 96.) Cambridge: Cambridge University Press. Frey, Werner 2011 Peripheral adverbial clauses, their licensing, and the prefield in German. In: E. Breindl, G. Ferraresi & A. Volodina (hg.) Satzverknu¨pfungen. Zur Interaktion von Form, Bedeutung und
Introduction: Theory of mind elements across languages
33
Diskursfunktion, 41–78. [Linguistische Arbeiten 534]. Berlin: de Gruyter. Frey, Werner & Karin Pittner 1998 Zur Positionierung der Adverbiale im deutschen Mittelfeld. Linguistische Berichte 176: 489–534. Hacquard, Valentine 2006 Epistemics with Attitude. SALT 20XVIII. Haegeman, Liliane 2002 Anchoring to speaker, adverbial clauses, and the structure of CP. Georgetown University Working Papers in Linguistics 17: 109– 141. Haegeman, Liliane 2004a Topicalization, CLLD, and the left periphery. ZAS Papers in Linguistics 35: 157–192. (Proceedings of the dislocated elements Workshop, ZAS Berlin. November 2003, vol. 1.) Haegeman, Liliane 2004b The syntax of adverbial clauses and its consequences for topicalization. Antwerp Papers in Linguistics 107: 61–90. Haegeman, Liliane 2006 Conditionals, factives, and the left periphery. Lingua 116: 1651– 1669. Hinzen, Wolfram 2006 Mind design and minimal syntax. Oxford: Oxford University Press. Hinzen, Wolfram 2007 An essay on names and truth. Oxford: Oxford University Press. Jakobson, Roman 1971 Selected Writings II. World and Language. The Hague: Mouton. Karttunen, Lauri & Standley Peters 1979. Conventional implicature’, 1–56. In Ch.-U. Oh and D. A. Dinneen (eds.) Presupposition. Syntax and semantics 11. New York: Academic Press. Kempson, Ruth M. 1979. Presupposition, opacity, and ambiguity. In: Syntax and semantics. Presupposition, 283–298. New York: Academic Press. Kratzer, Angelika 1981 The notional category of modality. In: Hans-Ju¨rgen Eikmeyer and Hannes Rieser (eds.), Words, worlds, and contexts, 38–74. Berlin: de Gruyter. Kratzer, Angelika 1991 Modality. In: Arnim von Stechow & Dieter Wunderlich (eds.), Semantik: Ein internationales Handbuch zeitgeno¨ssischer Forschung, 639–650. Berlin/New York: de Gruyter.
34
Werner Abraham & Elisabeth Leiss
Kratzer, Angelika 1999 Beyond ‘‘oops’’ and ‘‘ouch’’. How descriptive and expressive meaning interact’, Paper presented at the Cornell Conference on Theories of Context Dependency, March 1999. Kratzer, Angelika 2012 Modals and conditionals. [Oxford Studies in Theoretical Linguistics]. Oxford: Oxford University Press. Lahousse, Karen 2010 Information structure and epistemic modality in adverbial clauses in French. Studies in Language 34/2: 298–326. Law, Ann 2002 Cantonese sentence-final particles and the CP domain. UCL Working Papers in Linguistics 14: 375–398. Leiss, Elisabeth 2008 The silent and aspect-driven patterns of deonticity and epistemicity. A chapter in diachronic typology. In: Werner Abraham and Elisabeth Leiss (eds.), Modality-aspect interfaces: Implications and typological solutions, 15–42. (Typological Studies in Language 79.) Amsterdam: John Benjamins. Leiss, Elisabeth 2009 Drei Spielarten der Epistemizita¨t, drei Spielarten der Evidentialita¨t und drei Spielarten des Wissens. In: Werner Abraham and Elisabeth Leiss (eds.) Modalita¨t, 3–24. (Studien zur deutschen Grammatik 77.) Tu¨bingen: Stau¤enburg. Lewis, David 1973 Counterfactuals. Cambridge, Mass.: Harvard University Press. Li, Boya 2006 Chinese final Particles and the syntax of the periphery. Ph.D. thesis Univ. Leyden, NL. Marantz, Alec 1997 No escape from syntax: Don’t try morphological analysis in the privacy of your own lexicon. In: Alexis Dimitriadis, Laura Siegel, Clarissa Surek-Clark and Alexander Williams (eds.), Proceedingsof the 21st Penn Linguistics Colloquium, 201–225. (U. Penn Working Papers in Linguistics 4.2.) Philadelphia: University of Pennsylvania. Meibauer, Jo¨rg 1994 Modaler Kontrast und konzeptuelle Verschiebung. Studien zur Syntax und Semantik deutscher Modalpartikeln. (Linguistische Arbeiten 314.) Tu¨bingen: Niemeyer. Nuyts, Jan 2000 Epistemic modality, language, and conceptualization. (Human cognitive processing 5.) Amsterdam: John Benjamins. Nuyts, Jan & Ann de Roeck 1997 Autism and meta-representation: The case of epistemic modality. European Journal of Disorders of Communication 32 (2): 113–137.
Introduction: Theory of mind elements across languages
35
Papafragou, Anna 2002 Modality and theory of mind. Perspectives from language development and autism. In: Sjef Barbiers, Frits Beukema and Wim van der Wur¤ (eds.), Modality and its interaction with the verbal system, 185–204. (Linguistik Aktuell/Linguistics Today 47.) Amsterdam/Philadelphia: John Benjamins. Papafragou, Anna, Li Peggy, Choi Youngon & Han Chung-hye 2007 Evidentiality in language and cognition. Cognition 103, 253–299. Popper, Karl 1995 Eine Welt der Propensita¨ten. Tu¨bingen: Mohr. Potts, Chris 2002 Expressive contrast as conventional implicature’. Paper presented at NELS 33, MIT, Cambridge, MA, November 2002. Potts, Chris 2005 Potts, Chris 2005. The logic of conventional implicatures. Oxford: OUP. Reis, Marga & Angelika Wo¨llstein 2010 Zur Grammatik (vor allem) konditionaler V1-Gefu¨ge im Deutschen. Zeitschrift fu¨r Sprachwissenschaft: 111–179. Rizzi, Luigi 1997 The fine structure of the left periphery. In: Liliane Haegeman (ed.), Elements of grammar. Handbook of Generative Grammar, 281–337. Dordrecht: Kluwer. Sperber, Dan & Deirdre Wilson 1986 Relevance. Communication and cognition. Oxford: Blackwell. Stalnaker, Robert 1974 Pragmatic presupposition. In Semantics and philosophy, ed. Milton K. Munitz & Peter K. Unger, 197–213. New York: New York University Press. Stalnaker, Robert 1978 Assertion. In Syntax and semantics 9: Pragmatics, ed. Peter Cole, 315–332. New York: Academic Press. Stechow, Arnim von 2004 Modalita¨t. Class lecture, Universita¨t Wien. Tanaka, Shin 2011 Deixis und Anaphorik. Referenzstrategien in Text, Satz und Wort. [Linguistik: Impulse und Tendenzen]. Berlin: de Gruyter. Thurmair, Maria 1989 Modalpartikeln und ihre Kombinationen. Tu¨bingen: Niemeyer. Velde, John te 2005 Deriving coordinate symmetries. A phase-based approach integrating Select, Merge, Copy, and Match. (Linguistik Aktuell/Linguistics Today 89.) Amsterdam/Philadelphia: John Benjamins. Weinrich, Harald 2005 Grammaire textuelle du franc¸ais Paris: Didier.
36
Werner Abraham & Elisabeth Leiss
Weydt, Harald 1969
Abto¨nungspartikel. Die deutschen Modalwo¨rter und ihre franzo¨sischen Entsprechungen. Bad Homburg: Gehlen. Wilson, Deirdre & Dan Sperber 1988 Mood and the analysis of non-declarative sentences. In Human agency. Language, duty and value, ed. Jonathan Dancy, J. M. E. Moravcsik and C. C. W. Taylor, 77–101. Stanford, Calif.: Stanford University Press. Zeijlstra, Hedde 2008 Modal concord is syntactic agreement. In Proceedings of SALT 17. CLC Publications, Cornell University. Zimmermann, Malte 2004 Zum Wohl: Diskurspartikeln als Satztypmodifikatoren. Linguistische Berichte 199: 1–35. Zimmermann, Malte 2008 Discourse particles in the left periphery. In: P. Cook, W. Frey, C. Maienborn & B. Shaer (eds.) Dislocated elements in discourse, 200–231. Oxford: Routledge.
Part I.
The foundation: speaker and hearer deixis, shifter, and double displacement
Epistemicity, Evidentiality, and Theory of Mind (ToM) Elisabeth Leiss Ludwig-Maximilians-Universita¨t Mu¨nchen Abstract The central aim of this paper is to lend theoretical support to the hypothesis of Papafragou (2000) and Papafragou et al. (2007) that the development of Theory of Mind in children depends on the development of linguistic categories, and not vice versa. The special focus will be on modality, especially on epistemic modal verbs. It will be shown that the linguistic architecture of epistemicity largely depends on the development of the grammatical categories of aspect, tense, mood, modality, and person. The category of person is elementary. It is defined as a Jakobsonian shifter: Its reference shifts with the origo of the speaker. It is of central importance to understand how shifters serve as building blocks for the construction of complex grammatical categories such as aspect, tense, mood, and modality. These grammatical categories are special insofar as they involve double displacement and the split-up of the speaker into a twofold origo with di¤erent locations and functions. The focus of this paper will be upon the grammatical categories of evidential and epistemic modality, which will be described as functionally equivalent. In this respect, special emphasis will be put on the argument that lexical evidentiality and lexical epistemicity (evidential and epistemic adverbials) are not equivalent at all to their respective grammatical counterparts, the latter being more complex in terms of shifting qualities. A large amount of literature maintains exactly the contrary: Lexical and grammatical means of modality are considered as being equivalent, whereas evidentiality and epistemicity are labeled as di¤erent types of modality. The weaknesses of these approaches will be clearly outlined. In a general outlook, the psychological approach to ToM by Papafragou and the theoretical approach of Abraham and Leiss (2007) and Abraham and Leiss (2009) will be compared with philosophical approaches to language and cognition, where language is made responsible for the development of all higher forms of human cognition (especially Donald Davidson [2001] 2004 and Karl Popper 1995).
40
Elisabeth Leiss
1. Central thesis The central thesis of this paper1 consists of two seemingly independent claims. The first is that the understanding and production of grammatical categories is closely related to the development of a Theory of Mind in children (following Papafragou 2000; Papafragou et al. 2007; Choi 2006). The development of a Theory of Mind (ToM) progresses on a par with the development of grammatical categories, the latter being responsible for catalyzing ToM. The strong claim put forward here is that the human faculty of ToM is language driven. There is good reason to suppose that the most complex grammatical category is modality, which is in itself divided into subtypes of modality of di¤erent phases of complexity. The second claim is that lexical means of coding modality are far from being equivalent with grammatical means. In other words, an epistemic adverbial such as certainly or necessarily does not compare with the epistemic modal verb must in terms of complexity. 2. Theory of Mind and Modality The English notion of Theory of Mind is well established, but nevertheless not self-explaining and therefore somewhat misleading. There is a way out, when we start with Carnap’s notion of ‘‘Eigenbewusstsein’’ versus ‘‘Fremdbewusstsein’’.2 According to Carnap ([1928] 1998), we have direct access only to our own mental representations. This access is meant by Carnap to be even more immediate than is the access to sensual data, which must be reconstructed via our mental representations. This reconstructive process bridges the abyss between our own mental representations and those of others, thus preparing the way to get to intersubjective and objective knowledge of the world. Real objective knowledge therefore necessarily contains a bit of real data, which are negotiated between different human minds. In Carnap’s words: Objective knowledge is in any 1. This paper was written in the framework of the German-British project ‘‘Un-Cartesian Linguistics’’, with project leaders E. Leiss/Munich and W. Hinzen/Durham, German DFG-title LE718/9-1. 2. Carnap ([1928] 1998 and [1928] 1966) speaks of ‘‘eigenpsychischen Pha¨nomenen’’ and ‘‘fremdpsychischen Vorga¨ngen’’, which might be translated by ‘‘mental representations in the self ’’ versus ‘‘mental representations in others’’. The first ones are directly accessible, even more immediately accessible than are sensual data, according to Carnap, whereas the latter have to be inferred.
Epistemicity, Evidentiality, and Theory of Mind (ToM)
41
case ‘‘sachhaltig’’. For this reason, objective knowledge can principally be falsified or verified, this in stark contrast to pseudo-knowledge. Pseudoknowledge can in fact be multiplied, however it does not qualify as intersubjective or objective knowledge, for it is not rooted in reality and therefore not ‘‘sachhaltig’’, and it has not been negotiated between su‰ciently competent individuals. At this point, the function of modality comes into the picture. Modality can be defined as the linguistic means that makes it possible to negotiate the truth-values of mental representations. Truth-values, however, can be attributed to mental representations only if they have the format of propositions, i.e. of finite sentences. For this very reason, modality arises late in language acquisition (see Doitchinov 2001, 2007; Choi 2006). The acquisition of modality follows the acquisition of the full architecture of finite propositions, i.e. the establishment of a complete range of grammatical categories such as person, aspect, tense, and mood. Doitchinov’s empirical investigation into the acquisition of the lexical epistemic meaning of German vielleicht (English ‘perhaps’) and of the grammatical epistemic function ko¨nnen (English ‘may’) reveals that the lexical function is acquired earlier than the grammatical function (both functions are acquired between six and nine years of age). As long as children cannot perceive a functional di¤erence between grammatical and lexical means of modality, they are not able to understand the grammatical epistemic meaning of epistemic modal verbs. This paper attempts to give an account of the di¤erence in complexity between lexical and grammatical modality (see also Leiss 2009a). Essentially, modality adds illocutionary force to propositions. The illocutionary function of lexical modality di¤ers from grammatical modality insofar as in the latter double displacement is involved. Lexical modality lacks this double displacement feature. Grammatical modality in itself is subdivisible into di¤erent means of illocutionary fine-tuning. We need all these subtypes of modality to negotiate our mental representations of the world, which are of a genuine subjective nature, built on episodic experiences, which are real, but do nevertheless vary from person to person. This process of negotiating our mental representations is the essence of ‘‘foreign consciousness alignment’’,3 a notion coined by Abraham in 2009 in his introductory note to the Lisbon workshop on modality at the SLE conference (Abraham 2011a, this volume). Abraham was alluding by this notion to Carnap’s terminology rather than to the psychological notion of Theory of Mind. 3. The German notion is ‘‘Fremdbewusstseinsabgleich’’, which is Abraham’s translation of ‘‘Theory of Mind’’ into German (see Abraham 2011a, 2011c).
42
Elisabeth Leiss
Foreign consciousness alignment is synonymous with Theory of Mind insofar as it is also defined as the ability to assess other people’s beliefs (mental states) about a state of a¤airs. Besides, foreign consciousness alignment underlines the importance of negotiating our beliefs to arrive at common ground.
3. Relevance Theory and Foreign Consciousness Alignment (FCA) Sperber and Wilson (1986) repudiated the model of communication by Shannon and Weaver (1949) with the imperative words ‘‘thoughts do no travel!’’. These words paraphrase the fact that our mental representations cannot be simply loaded into the brains of others by the process of communication. In other words, mental representations cannot travel from brain to brain via strings of linguistic sounds. An easy way out of this problem has quite often been suggested by the statement that we human beings dispose of the same mental representations, which are said to be due to the fact that we live in a same world. Convincing as this statement may be at first sight, it is exactly the other way round. Humans do not live in a same world. Animals, however, do this. They share the same sensual and cognitive environment, which is guaranteed by instinct. Human beings, in contrast, are almost devoid of inborn instinct. Given that only shared instinct provides shared cognitive content, we have to explain how it is possible for humans to communicate in an e‰cient way. Instinct shapes the world of animals such that they live in a common environment. Humans, however, have to shape their world to quite a considerable degree by themselves. They do it by constructing objectivity via foreign consciousness alignment. To achieve this, truth-values have to be assigned. We must be aware of the fact that the assignment of truth-values depends on a specific format, which is the proposition. This format is a linguistic format. A proposition is some sort of linguistic software whose function is to construct mutual knowledge about the world. ‘‘Mutual knowledge’’ is a core notion of the Relevance Theory of Sperber and Wilson (1986). Mutual knowledge is achieved by a lifelong process of propositionally established consciousness alignment between speakers. Once mutual knowledge is achieved, it belongs to a common store of mental representations. For this very reason, speakers generally remain mute about presuppositions (common knowledge or ‘‘common ground’’). What we gain, when we communicate, is the construction of new common ground, the deconstruction of old common ground not being excluded. The Mutual Knowl-
Epistemicity, Evidentiality, and Theory of Mind (ToM)
43
edge Hypothesis of Sperber and Wilson (1986: 3–24) can be summarized as follows: 1. Mutual knowledge between speaker and hearer is a prerequisite for communication. 2. In order to communicate successfully (economically), we have to eliminate most of the information we want to convey. 3. What we must eliminate is shared knowledge and shared presuppositions about the world. This explains why the amount of bits of messages transported by linguistic sounds is quite poor, compared with the amount of structured mental representations evoked in the hearer. Linguistic signs turn the attention of the speaker to shared common ground not immediately present, or they signal deviations from presuppositions. Linguistic signs can attribute truthvalues to a restricted part of the common ground, not to the common ground as a whole. Thus, truth-values can be assigned to (1), whereas in (2) this is not the case: (1) A flock of unicorns survived in Northern Bavaria. (2) the survival of a flock of unicorns in Northern Bavaria In contrast to finite clauses, such as (1), noun phrases are nonfinite and as such cannot transport truth-values. It is important to note, that there are always islands of non-finiteness in finite sentences, in most cases made up of noun phrases: (3) The survival of a flock of unicorns in Northern Bavaria is good news. In (3) only the finite part of the proposition can be verified or falsified. We can verify or falsify whether it is good news or not to have unicorns in Northern Bavaria. In contrast to this, the noun phrase in subject position is presupposed as common ground, even when the survival of a flock of unicorns in Northern Bavaria is highly improbable. Presupposed in the examples above is also that unicorns live in flocks. In other words, islands of non-finiteness and non-veridicality are ubiquitous in our discourse. Some of them are materially present in sentences, such as noun phrases or parts of these phrases. Most of the islands of non-finiteness, however, are covert and therefore are never uttered. In other words, mutual knowledge or common ground can be defined linguistically as islands of nonfiniteness and non-veridicality, being overt (non-finite linguistic sequences) or covert (unexpressed presupposed knowledge). Whenever modality nego-
44
Elisabeth Leiss
tiates such shaky common ground it negotiates risks of knowledge assessment. This means that 1. in order to communicate in the sense of Relevance Theory, we first have to assess the mental representations of the hearer; 2. the assessment of mutual knowledge is a risky endeavor; 3. modality can be defined as a tool kit to lower the risks of assessing shared knowledge; 4. modal verbs and modal particles are techniques to negotiate shared knowledge.
4. Language and Foreign Consciousness Alignment: The ATM family Assessment of truth-values is characteristic of the human mind and of the human mind only. This cognitive faculty is not found in animals,4 not even in dogs, who are followers of the human culture and developed human-like social skills in many a respect (see Hare & Tomasello 2005),5 nor in primates, who share a lot of genetics with humans. Why is that? The reason might be that the function of truth-veridicality is built up systematically by linguistic means only, and not by some general, nonlinguistic cognitive endowment. Foreign consciousness alignment (FCA) or ToM is gradually built up by the development of grammatical categories. It is quite probable that not all grammatical categories are involved in this construction of FCA. Involved are aspect, tense, mood, and modality. In other words, what is involved is the so-called TMA-complex, which we label ATM-complex with respect to the hierarchically ordered complexity of the categories involved. It is well known that aspect is the first building block, triggering the construction of tense, which itself triggers mood. Modality (illocutionary force as established by modal verbs and modal particles) is acquired as soon as the acquisition of the ATM-complex is complete. The question arises, what is the common denominator of ATM and modality? What is the common feature of these grammatical categories, such that other categories do not have it? The answer is that ATM and modality establish a special kind of division of the speaker into di¤erent 4. See Brandt (2009), who gives an overview of the arguments and the literature on this topic. 5. Hare and Tomasello (2005) start from di¤erent axiomatics. They suppose that the social behavior of communication made the human species intelligent and created human culture.
Epistemicity, Evidentiality, and Theory of Mind (ToM)
45
viewpoints. This division takes place when essential presuppositions concerning the natural origo of the speaker are being violated.6 What do we understand by ‘‘natural origo’’? How is it defined? The notion was coined by Karl Bu¨hler ([1934] 1982, [1934] 1990), who defines it an assembly of three features: [[þhere], [þnow], [þI/me]]. This bundle of features characterizes the egocentric point of view of the speaker. We share this natural viewpoint with higher animals. In contrast to them, however, we are in a position to distance ourselves from this egocentric perspective. Aspect is the basic building block in the ATM-family. Aspect divides the speaker into two persons of di¤erent locations. One person is located at the ‘‘origo’’ of the speaker,7 which equals the feature [þhere]. This feature is presupposed. The second location of the speaker is not inside of the event (‘‘inner aspect’’/imperfectivity), but at distance of it (‘‘outer aspect’’/perfectivity). The second location is the marked one. It arises whenever the basic presupposition (the ‘‘here-origo’’) is violated. Emphasis is on the feature [þdistance], which is of spatial nature in the first place. Tense also involves the feature [þdistance], which is reinterpreted as temporal distance. Tense divides the speaker into two di¤erent persons, the actual speaker and the viewer, who are located at di¤erent time spots. The actual speaker is cotemporal with utterance time, whereas the viewer looks at the event from a location at reference time. The second temporal location is the marked one. It is activated whenever the origo-related presupposition that speaker and event are located in the now is violated. In other words, the speaker and the event do not share the same temporal location. This is the reason why a second viewpoint comes in. Mood also shares the feature [þdistance] with the categories of aspect and tense. Again, the speaker is divided into two persons. One of them is, again, the utterer of the proposition; the other person involved takes a stance at a distance from the uttered proposition. What we find is the dissolution of the origo-related presupposition that the speaker is standing
6. The proposed functional explanation of the ATM-complex was first outlined in Leiss (1994). 7. ‘‘Origo’’ refers to Karl Bu¨hler’s notion of ‘‘Origo’’, which is defined by the features [[þhere] [þnow] [þreal]]. See Bu¨hler (1990) and Bu¨hler (to appear, 2011b, second edition of Bu¨hler [1934] 1990), with a new preface by Werner Abraham. See also Leiss (1994) where the violation of these basic features is discussed as the foundation of the ATM-complex. Andersson (1989: 37) first proposed [þdistance] as unifying feature of the ATM-complex.
46
Elisabeth Leiss
behind what he says. In other words, the speaker distances himself from himself, which is formulated somewhat paradoxically. It is nevertheless the very essence of mood. The speaker does not believe in what s/he says: ‘‘It is not me who supports the propositional content’’. The main function of mood is irrealis. To sum up, the common denominator of the ATM-complex is the feature [þdistance]. The distinguishing features are [ehere], [enow], and [eme]. Thus, we have the following feature bundles: Table 1. The features of ATM Aspect:
[þdistance] [ehere]
Tense:
[þdistance] [enow]
Mood:
[þdistance] [eme/I]
There is a clear hierarchy concerning the dissolution of ATM-presuppositions: The modal origo represents the most robust presupposition. Therefore, mood is acquired later in language acquisition than tense or aspect. Mood is also used in discourse less frequently than aspect and tense. The function of grammatical categories in general is to signal to the addressee that his natural presuppositions concerning the origo of the speaker are to be replaced by the marked features, which are the minus-features. In sum, the characteristic of the ATM-categories is double displacement. Each of these categories implies that the speaker is split up into two viewpoints. The attention of the hearer thus has to be divided into two di¤erent (aspectual, temporal, or modal) locations. This is a complex cognitive process. It is important to understand that double displacement und divided attention is characteristic of the ATM-categories. Lexical means, such as adverbials, are devoid of the process of double displacement. There are lexical shifters, such as yesterday, but they do not compare with the grammatical function of past tense in terms of deictic complexity. There will be more ample discussion of this important point below. Before going into more detail it is important to see the main di¤erence between mood and modality, the latter being the main topic of this paper. Mood as a functional category pertains to the propositional level, whereas modality is illocutionary in nature. Modality (epistemic modal verbs, modal particles) is still more complex than mood in terms of double displacement. The illocutionary level of double displacement dissolves an
Epistemicity, Evidentiality, and Theory of Mind (ToM)
47
even more robust presupposition than does mood. Epistemic modal verbs8 split up the speaker into an evaluator and into the source of the propositional content. The presupposition violated in epistemic modals is that the speaker is the source of the information. It is not di‰cult to conceive that the presupposition that the speaker is the source of the information is even more basic than Bu¨hler’s origo. Beyond this, modal particles are used when, additionally to the speaker and his multiple modes of displacement, the addressee becomes involved in the negotiation of the truth-value. The speaker negotiates his assessment of the common ground of the hearer by introducing modal particles. The function of these particles is to signal to the hearer that his common ground has been assessed, but maybe not in complete concordance with him. Thus, the addressee is split up into the real hearer and the assessment of the hearer’s common ground by the speaker. The speaker opens up a slot for the hearer to evaluate the success of the speaker’s foreign consciousness alignment. It seems that this is the most complex form of double displacement conceivable. It overrides even epistemic modal verbs in complexity. In modal verbs, the speaker is split up into an evaluator and into the source of information. The hearer does not participate in the evaluation of the truth-value. The assessment is located inside the speaker (intransitive assessment, a notion introduced by Abraham in his introduction to the workshop at Lisbon). In modal particles, however, the assessment of the truth-value is transitive. The assessment of the truth-value is located in two di¤erent persons (see Abraham, this volume). When we discuss complexity in terms of double displacement (involving aspect, tense, mood, modality of modal verbs, and modality of modal particles), the question arises what to do with lexical ‘‘equivalents’’ of these categories. At close sight, however, lexical modality adverbials are not equivalent to epistemic modals, because they do not involve double displacement. This is an important argument against the position that grammatical and lexical means of encoding modality (or any other kind of grammatical category) share the same function. In the next section, the implications of this insight are outlined. What is the di¤erence between mood and modality? Some use modality (wide sense) as a label for the whole range of grammatical and lexical 8. Modal verbs in the narrow sense adopted here can have epistemic readings (this in line with Oehlschla¨ger 1989, who showed convincingly that modal verbs cannot be defined by formal features exclusively).
48
Elisabeth Leiss
means of expressing mood and modality (narrow sense). We adopt here the notion of modality in the narrow sense (which includes modal verbs and modal particles). In contrast to modality, mood is a functional category pertaining to the level of the proposition, whereas modality is illocutionary in nature. Again, the feature of [þdistance] is being reinterpreted. This time, the speaker splits up into the utterer (locative person) and into the evaluator of the proposition (illocutive person). In other words, modality establishes a viewpoint from outside the proposition. This amounts to the fact that a metalinguistic viewpoint is being established. Modality can be defined as a key opening the door to metalanguage. Metalinguistic abilities are found in the human species only. Considering that modality is the final stage of the ATM-complex, and looking at the ATM-architecture as a bundle of techniques to dissolve the natural viewpoint or origo, the very metalinguistic abilities have to be defined as language driven human faculty. This explains why metalanguage is characteristic of humans only. If it were some general cognitive capacity, it should be expected to emerge in some other animal as well. However, it does not, not even in dogs or higher primates. Modality, or illocutionary force, is a domain that has not been well subdivided until now, when it comes to functional details. The hypothesis probed here is that the illocutionary domain of the sentence is functionally as diverse as are the functional categories at the propositional level. Characteristic of modality in the Germanic languages are modal verbs and modal particles. Modal verbs and modal particles are not universal when we look at them from a formal point of view. However, when we begin to understand their functions, we are in a position to uncover functional equivalents in the intricate grammatical patterns of hitherto not well-described or well understood languages. Again, double displacement is essential when it comes to defining modal verbs and modal particles. In modal verbs, the speaker is split up into an evaluator (assessing person) and into the source of the information (informing person). In modal particles, the speaker is divided into his egocentric viewpoint and the viewpoint of the hearer. Modal verbs are probably less complex as compared to modal particles. In modal verbs, the hearer does not participate in the assessment of the truth value. In epistemic modal verbs, the assessment of truth-values is located inside the speaker (intransitive assessment, according to Abraham 2011a). In modal particles, in contrast, the assessment of the proposition is transitive. The speaker negotiates the truth-value with the hearer. The semantics of modal particles involves the hearer (see Abraham 2011a) in
Epistemicity, Evidentiality, and Theory of Mind (ToM)
49
the process of assessment. The speaker’s choice of modal particles signals ‘‘my assessment of your assessment of the truth values is such that x. Please interfere if I am wrong!’’. The di¤erent modal particles explicate the variable x of the speaker’s assessment of the hearer’s assessment. As English has lost its modal particles almost completely since the Middle English period (van Gelderen 2001),9 a German example is in place to illustrate the fine-grained semantics of di¤erent modal particles: (4) Einho¨rner sind ja ausgestorben. Unicorns are MP yes died out. ‘Unicorns have died out, haven’t they?’ ‘Unicorns have died out, and I suppose that you, the hearer, are totally aware of it also.’ (5) Einho¨rner Unicorns ‘Unicorns ‘Unicorns
sind doch ausgestorben. are MP doch died out. have died out, haven’t they? have died out, in contrast to what you appear to believe.’
Examples (4) and (5) show that question tags do not translate the whole meaning of the di¤erent modal particles chosen in the respective sentences. They translate the common feature of all modal particles in that they involve the speaker AND the addressee in the final assessment of truthvaluation. The German equivalent of the question tags in (4) and (5) would be ‘‘nicht wahr?’’ (‘not real?’). These lexical means modalize a proposition. They are transitive in the sense that they involve the hearer’s assessment of the truth-value. However, they are situated at a lower metalinguistic level. The speaker invites the hearer to give his consent or to refuse it. Nevertheless, the speaker does not reveal to the hearer what his own assessment of the hearer’s assessment of the truth values are. This is a significant di¤erence between modal particles and question tags that is not to be neglected or downplayed. An interesting question yet to solve is how the English language does it to encode this significantly more complex metalinguistic level of modalization. This point has yet to be explored.
9. Abraham (1995, 2005) attributes the lack of modal particles in specific languages to the lack of a middle field, which is characteristic of German syntax and the syntax of most of the other Germanic languages, except English. This explanation based on formal arguments does not account for the loss of the function of modal particles.
50
Elisabeth Leiss
Are there covert means of this type of modalization? What is conceivable nevertheless is that the qualitative di¤erence between lexical and grammatical means of modalization must be accounted for. Summarizing the arguments concerning the architecture of the ATMMcomplex,10 we see that the feature bundle of [[þdistance] [þdouble displacement]] defines this specific complex of grammatical categories. We begin to understand why exactly these grammatical categories cluster together. Each of these categories builds up a di¤erent kind of double displacement, with each belonging to a di¤erent syntactic level. In other words, there is a hierarchy of double displacement. The higher the level of double displacement, the more essential is the presupposition to be dissolved. Animals are prisoners of these presuppositions, while the human species is not. We can violate presuppositions and translate these violations into marked features that are communicable to other members of the human species. Is there any double displacement in animals? It seems that some primitive form of spatial deixis, however, to be sure, not space shifting nor double displacement in space (aspect), is documented for dogs and even wolves (Hare & Tomasello 2005). All reinterpreted and higher forms of double displacement (tense, mood) are missing with dogs. It might help to refer to this point in order to understand the di¤erent level of complexity concerning deixis, shifting, and double shifting. The di¤erence of this threefold levels of deictic complexity will be explained in the next section.
5. The di¤erence between lexical shifters and grammatical double displacement The functional abyss between grammar and lexicon, disregarded fervently by partisans of Construction Grammar and similar approaches, becomes visible as soon as we take into account the defining features of the ATMM-complex. The division of the speaker into two persons, i.e. the speaking person and the viewing person, belongs to grammar only. Adverbials are devoid of such a second viewpoint. To grasp this intricate but important di¤erence, an exemplary look at temporal adverbials and tense should be conclusive. An adverbial such as yesterday is a shifter (in the sense of Jakobson 1957). In order to compute the temporal location of
10. ‘‘ATMM’’ abbreviates ‘‘aspect, tense, mood, and modality’’.
Epistemicity, Evidentiality, and Theory of Mind (ToM)
51
the event spoken about, the hearer has to locate the origo of temporal deixis. The temporal location of the events shifts with the origo of the speaker. The adverbial yesterday points from this origo to the event and locates it. In contrast to temporal adverbials, tense is not deictic, but phoric. It displaces and splits the speaker’s origo. The speaker travels mentally into the past or into the future. Time travelling (in the sense of Evans 2004 or Tulving 2005) with help of the category of tense can be anaphoric [þpast] or cataphoric [þfuture] in quality (see Fig. 1 for the anaphoric function):
Figure 1. Anaphoric shifting function of past tense
The past tense splits the speaker into speaker and viewer. See (6–8). The temporal adverbials in (6) and (7) are not responsible for this split.11 There are two kinds of temporal adverbials, those with absolute reference as in (7) versus those with relative reference (6). (6) An earthquake devastated Japan yesterday. (7) An earthquake devastated Japan on March 11, 2011. (8) An earthquake devastated Japan. In (7), the computation of the origo of the speaker is not necessary to locate the event temporally. In (6), by contrast, the origo of the speaker must be contextually present for the hearer to be able to locate the event spoken about. In (6), the event is located from the speaker’s perspective. In narrative fiction, such as in a novel of Thomas Mann, the origo is not located in the actual present, but in the fictitious present, which is the viewer’s perspective. As soon as the past tense is used, both perspectives
11. Diewald (1991) characterizes shifters, such as yesterday, as elements with strong deixis, this in contrast to weak deixis in grammatical categories. Instead of weak deixis, the notion of ‘‘phoricity’’ is suggested here. Her work on modal verbs (Diewald 1999) is also based on the essential distinction between strong and weak deixis.
52
Elisabeth Leiss
emerge immediately. Temporal adverbials, however, always choose just one of these perspectives, never two at a time. In (8) the past establishes two viewpoints. In (6) the past also implements double displacement. The additional adverbial in (6), however, does nothing more than select one of the two possible viewpoints. In this example, it is the speaker’s origo. It is evident, that the function of temporal adverbials, even when they are shifters, must be kept apart from the function of tense. Tense is essentially phoric, whereas temporal adverbials are deictic. To understand the semantics of deictic adverbials, a look at personal pronouns helps. The first person singular refers to the speaking person. Its origo changes with every turn taking in conversation. The second person refers to the addressee. Its reference depends also on the speaker’s location. Finally, the third person refers to something or someone outside the dialogue. The reference of personal pronouns shifts with the person speaking. In other words, the semantics of shifters is extracted from the context in which they are used. The same holds for temporal adverbial shifters. They cannot refer independently from the temporal context established by tense. Table 2 presents an overview of the di¤erent features discussed: Table 2. Lexicon versus grammar in past time reference absolute temporal adverbial (11 March 2011)
[þlexicon] [–finite]
direct reference
compares with the referential function of proper names
relative temporal adverbial (yesterday)
[þlexicon] [–finite]
computed reference, simple displacement
deictic function
past
[þgrammar]12 [þfinite]
double displacement
phoric function
Lexical elements deploy their function without being necessarily embedded in finite propositions. This holds for absolute temporal adverbials, which have proper-name-like referential qualities. This holds also for relative temporal adverbials: Deixis (computing an origo and pointing at an event) is neither verifiable nor falsifiable. In contrast, tense is not conceivable outside of finite propositions. Of course, the deixis of adverbial
12. The feature [þgrammar] of course implies [–lexicon].
Epistemicity, Evidentiality, and Theory of Mind (ToM)
53
shifters is more than just pointing to an object or event. Gestures that point at some object are understood correctly by dogs (see Hare & Tomasello 2005). In this respect, dogs and, to a lesser extent, even wolves are more human-like in their behavior than great apes are. However, nobody suspects dogs to be able to compute shifters. Hare & Tomasello (2005) think of a process of co-evolution triggered by the same social behavior of wolves and humans. Dogs seem to grasp the concept of shared attention, which is paramount to building up a Theory of Mind. With respect to this important feature, dogs are more able (even more intelligent, I dare say) than great apes are, this without having an ‘exploding brain’. They profit from human culture, which is the result of endless linguistically initiated negotiations of knowledge, thus due to the human faculty of language.13 In other words, enculturation is profitable, even without being involved in the task of negotiating knowledge. We all make the experience that we profit parasitically from knowledge which is beyond our understanding. It seems that participation in language-based culture makes dogs more intelligent than apes, which do not have a long history of enculturation compared with dogs. However, nobody suspects dogs to be able to compute shifters. The understanding of shifters however relies on the mastering of pointing gestures. In conclusion, there are four levels contributing to the acquisition of knowledge and, maybe, intelligence: Pointing gestures (no displacement, but shared attention), deixis (simple displacement), and double displacement. All three techniques are linguistic techniques, gestures being the evolutionary oldest one. The forth level, the most complex of all these techniques is modality, which is the most sophisticated kind of double displacement. The next section will elaborate on the distinction between lexicon and grammar with respect to modality. The emphasis will be on the di¤erent functions of modality adverbials and grammatical modality.
6. Grammar vs. lexicon: Modal verbs versus modality adverbials The qualitative di¤erence between lexical and grammatical modality has been denied by the quite influential book of Nuyts (2000). Those who tend to follow Nuyts (2000) and similar approaches are convinced that modality is a non-linguistic (so-called ‘‘cognitive’’) concept that exists in
13. The latter point di¤ers from Hare and Tomasello 2005.
54
Elisabeth Leiss
our minds independently of the linguistic means to express it. This view is currently a very popular one inside and outside of science. It stems from insu‰cient in depth-reflection (see Leiss 2009b for alternative axiomatics). The view of modality as a non-linguistic concept forces the belief that grammatical and lexical means are equivalent. This is, however, by no means the case. There is unequivocal evidence that we have to consider di¤erent levels of complexity when we compare grammar and the lexicon. Where does this complexity come from? Since it is not part of our nonlinguistic cognitive apparatus, it must be linguistic in nature. In other words, we depend largely on language to build up di¤erent levels of double displacement. Without language, we would be imprisoned in our natural origo. We share this natural origo with animals such as dogs or apes. On a par with them, our basic presuppositions are prelinguistic and ‘‘cognitive’’ in the sense of non-linguistically established mental representations. All deviations, or better derivations, from this natural origo are human-specific, as is language. The techniques used to signal the escape from the cognitive prison of the natural origo are genuine linguistic techniques, documented for humans only. With this caveat in mind, we are prepared to look at grammatical epistemicity. The examples are taken from German, where the system of epistemicity is more transparent than in English.14 The proposed description can be taken as a starting point to investigate more opaque systems of modality. As Abraham (this volume) explores the system of modal particles in German in detail, modal verbs will be highlighted in this paper. Both papers emphasize the essential difference between lexicon and grammar. Grammatical epistemicity in German is closely linked to the grammatical category of person. Sentences (9), (10), and (11) are constructed with three di¤erent modals. Each modal implies a di¤erent constellation of viewpoints. The modal verbs producing these di¤erent viewpoints are mu¨ssen ‘must’, sollen ‘shall’, and wollen ‘will’. The English translations
14. Epistemic modals in English have lost almost completely their evidential component (in contrast to other Germanic languages; see Abraham 2002, de Hahn 2005a and 2005b). The epistemicity of English modals involves simple deixis, which is characteristic of adverbs, not of modal verbs. English modals are becoming more and lexical in nature. They still look like functional categories, but their architecture equals the complexity of adverbs. Therefore, English modal verbs are not the best model to understand the semantics of epistemic modality. Most of them have even lost deontic meaning, especially in American English (in more detail, Abraham 2002).
Epistemicity, Evidentiality, and Theory of Mind (ToM)
55
provide a first approximate view of the meaning of these constructions. The first glosses refer partly to lexicals in order to paraphrase the di¤erent e¤ects of the German modals. The lexical paraphrases in the first glosses are yet far away from revealing any systematicity. The paraphrases in the second glosses are, in contrast, systematic, yet also far away from providing the full meaning of the epistemic modals used. (9)
Sie muss schon mal in Lissabon gewesen sein. ‘She must have been to Lisbon before.’ ‘It is highly probable that she has been to Lisbon before’
(10) Sie will schon mal in Lissabon gewesen sein. ‘She pretends to have been to Lisbon before.’ ‘It is highly improbable that she has been to Lisbon before’. (11) Sie soll schon mal in Lissabon gewesen sein. ‘She is said to have been to Lisbon before.’ ‘It is quite probable that she has been to Lisbon before’ The first paraphrase tries to translate the source of evidence; the second one captures the probability e¤ect of the selected source. Involved are different degrees of certainty concerning the truth assessment of the propositions. Involved are also three di¤erent persons. It is very tempting to relate these persons to the first, second, and third person of the grammatical category of person. The first person then would be the speaker, the second person the addressee, and the third person would be located outside the dialogue. This model is quite attractive, as it can be related to Davidson’s modeling of the process how we achieve intersubjectivity and hence objectivity. According to Davidson ([2001] 2004), speaker and addressee continuously negotiate the meaning and the truth-values by means of language. The result of this potentially infinite process of truth assessment is objective knowledge (in the sense of Popper, according to whom objectivity is an aim we can approximate but never completely achieve; see also Abraham, this volume). Table 3 summarizes the distinctive features of the three German epistemic modals mu¨ssen, wollen, and sollen. It would be very tempting to include the addressee as a source of evidence. The advantage would be that it would fit best with Davidson’s triangulation model, where speakers and addressees achieve intersubjective knowledge in an endless process of communication, which is objective knowledge in the sense of Davidson (2004) and Popper (1995). It is, however, not possible to relate the propositional subject to the hearer. The second person involved is not the viewpoint of the addressee, but the view-
56
Elisabeth Leiss
Table 3. the assessment (inference) is based on evidence gained by the speaker himself source of information: speaker (1st person) ad (9) muss
assessor of information: speaker (1st person) The speaker splits up into two di¤erent persons. Speaker-based epistemicity: Inference is based on direct evidence (‘‘Eigenbewusstsein’’ in the sense of Carnap) The assessment is based on evidence gained by the propositional subject source of information: propositional subject
ad (10) will
assessor of information: speaker (1st person) The speaker is split up into speaker and viewer (propositional subject) Epistemicity is based on the propositional subject (indirect evidence) the assessment is based on evidence provided by some third person source of information: person(s) or instance outside the speakeraddressee-constellation (3rd person).
ad (11) soll
The speaker is split up into speaker and viewer (subject outside the speaker-hearer constellation): 2nd person assessor of information: speaker (1st person) Epistemicity is based on hearsay (3rd person)
point of the propositional subject. The hearer remains completely outside of this constellation. He enters the scene, however, with the illocutionary function of modal particles, which are still more complex than epistemic model verbs (see Abraham, this volume, who treats the speaker-addresseeconstellation in modal particles in full detail). When the propositional subject is not really the second person (the addressee), we are well advised to adjust our terminology. In order to illustrate the di¤erent person concepts involved in the assessment of the truth-values, the following notions are introduced:
Epistemicity, Evidentiality, and Theory of Mind (ToM)
57
Table 4. Three sources of evidence in epistemic modal verbs
muss
Source: locutionary subject (‘‘Eigenbewusstsein’’ – one’s own consciousness) Assessment: locutionary subject Probability: high15
will
Source: propositional subject (‘‘Fremdbewusstsein’’ – foreign consciousness) Assessment: locutionary subject Probability: low
soll
Source: illocutionary subject (‘‘intersubjective consciousness’’) Assessment: locutionary subject Probability: medium
Grammatically encoded epistemicity, as by epistemic modals, necessarily involves the source of assessment. Thus, grammatical epistemicity includes evidentiality. The same holds for grammatical evidentiality.16 Additionally to the source of evidence, assessment is always part of the meaning of evidentials. This does not hold for lexical epistemic adverbials, nor does it hold for lexical evidential adverbials. The reason is that adverbials do not split up into speaker and viewer (assessor and source). This amounts to the insight that lexical modality is either epistemic or evidential. Sometimes the meaning of an evidential adverbial skips into epistemic meaning, but its reading is evidential or epistemic, never both. Table 5 modifies Table 3 and details Table 4 with respect to the three di¤erent sources of evidentiality: 15. Kratzer (1991: 643–645) with her ‘‘graded notion of modality’’ does not distinguish between lexical and grammatical modality. From a systematic point of view, the material she presents is too heterogeneous to explore epistemicity in a principled way. 16. Whether grammatical evidentiality and grammatical epistemicity encode the same function or not (as Aikhenvald 2004 insists on) is matter of much controversy. Aikhenvald (2004) admits that evidentiality generally involves inferred assessment of truth-values. She insists, however, that assessment is not the core meaning of evidentials. Parallel, in the literature on epistemic modals the source of the assessment is not considered relevant. Maybe, we are just confronted with di¤erent traditions of description; maybe, there is a serious di¤erence. From a functional and universal perspective, the first option would be more attractive, as also Joseph (2004) summarized convincingly.
58
Elisabeth Leiss
Table 5. Locutionary, propositional, and illocutionary subjects as sources of evidence
ad (9) muss
ad (10) will
ad (11) soll
locutionary subject (L)
¼ source of evidence
propositional subject (P)
A source of evidence
illocutionary subject (I)
A source of evidence
evidential function: Source of evidence is based on one’s own consciousness epistemic function: Signaling high probability locutionary subject (L)
A source of evidence
propositional subject (P)
¼ source of evidence
illocutionary subject (I)
A source of evidence
evidential function: Low evidence, foreign consciousness as source of evidence epistemic function: Signaling low probability locutionary subject (L)
A source of evidence
propositional subject (P)
A source of evidence
illocutionary subject (I)
¼ source of evidence
evidential function: Quotative;17 intersubjective source of evidence epistemic function: Signaling medium probability
The high, low, and medium probabilities expressed by the German epistemic modal verbs mu¨ssen, wollen, and sollen are not semantically equivalent with the German modality adverbials certainly, certainly not, or probably. The reason is that the adverbials do not indicate the source of assessment. It is easy to see that double displacement is not part of the structure of lexical modality. The same holds for lexical evidentials such as obviously. This means that they indicate the source, but not the assessor, of the information. Lexical evidentials can be read as epistemics. In this case, however, the source of information is no longer part of their semantics. 7. Summary and outlook: Modality, language, and thought The three German modal verbs mu¨ssen, wollen, and sollen served as a succinct model to understand the function of modal verbs. The point of 17. Quotatives pertaining to mood and those pertaining to modality may have di¤erent functions. Modal quotatives do not necessarily share the feature of intersubjectivity.
Epistemicity, Evidentiality, and Theory of Mind (ToM)
59
departure was the Davidsonian triangle of communication, Relevance Theory of Sperber and Wilson (1986), Carnap’s Foreign Consciousness Alignment, and the Theory of Mind approach in psychology. To understand illocutionary force, we have to recur to the communicative situation, where the common ground between speakers is achieved and negotiated. We had to highlight the fact that the triangulation of the communicative situation, as proposed by Davidson, is fully expressed in modal particles only, not in modal verbs. There are obviously di¤erent levels of illocutionary force, which have still to be explored in detail. Further research should be invested into the architecture of modal verb systems. The reduced model proposed here does not fully account for the system of modal verbs in German. For instance, the proto-modal18 ko¨nnen (‘can’/‘may’) has been neglected in this paper. In Leiss (2009a), ko¨nnen is characterized as modal verb that neutralizes the distinctive features of the core modal verbs presented here. When we go back into the history of German modals, we see that at the very beginning of the grammaticalization of modal verbs in German, all modal verbs were able to express the meanings of the other modal verbs. They were excessively polysemous and synonymous at the same time. This seems to be characteristic of the early stages of the grammaticalization process of modal verbs. We learn from this that we should be careful not to attribute a apriori specific semantic features to each of the modal verbs present in a language. We have to take into account the stage of grammaticalization that each modal lexeme is presently undergoing and we might consider the possibility that several modal verbs may concur for the same function. The simple model proposed here might serve as an instrument to uncover similar functions in languages that do not have modal verbs or modal particles (which are far more frequent than languages sharing them). The model proposed should also be of some help to understand better the languages that do have modal verbs and modal particles and their di¤erent stages of grammaticalization (consider the di¤erence between English and German; Abraham 2002). The emphasis on the feature of double displacement highlights the di¤erence in feature complexity of lexical versus grammatical modality. As lexicon and grammar become more and more intermingled in current theoretical approaches to languages, this insight is not a minor point. Yet, we have to explain why the assumption that lexical and grammatical means are equivalent is still so popular, even more popular than it was
18. For details on proto-modality, see Abraham (2009).
60
Elisabeth Leiss
in the past four centuries. This assumption is based on the axiomatics of Rationalism. Rationalists (Cartesian philosophy) depart from the axiom that concepts are pre-linguistic in nature (for more details, see Leiss 2009b). For this reason, Rationalism reduces language to a means of expression of ready-made thoughts. In other words, Rationalism believes that thoughts are ready in our brains and that they are expressed by language. This axiom has been adopted by most of the cognitivists, be it the adherents of Chomsky (Cartesian Linguistics) or those of Piaget. The di¤erence between both rationalist language theories concerns just the point whether the language faculty is considered to be inborn or not. Both sides are Rationalists in that they are united in the conviction that thoughts are pre-linguistic concepts expressed by language. For this reason, the innate property of language is postulated to be devoid of content. Innate language in the sense of Cartesian Linguistics is reduced to rules, which are defined via regularity. Therefore, irregular grammatical elements are expulsed from the domain of grammar and moved into the lexicon. This is the very reason why the domains of the lexicon and of grammar become more and more confused even in rationalist universal approaches. There is no longer a clear delineation between the function of the lexicon and the function of grammar. This is the reason why Cartesian Universal Grammar is so vulnerable by attacks from the side of Construction Grammar. Goldberg (2006) attacks exactly at this front line, when she posits all linguistic rules to be nothing more but non-rule-governed lexical constructions (see Leiss 2009c). Both, the Chomsky-side and the Goldberg-side, overestimate or underestimate rules and regularity in language. They ignore that lexical semantics is no less rule governed than is grammatical semantics. They also ignore that irregular and conventionalized linguistic material is present in grammar as well as in the lexicon. An alternative to Rationalism is Un-Cartesian Universal Grammar. The approach of Un-Cartesian Linguistics adopted here claims that the function of lexical semantics and the function of grammar have to be clearly separated and must therefore be defined without recourse to the feature of regularity or non-regularity. One major aim in the project of Un-Cartesian Linguistics is to define the respective functions of grammar and lexicon in a way to avoid the cul-de-sac of the lexicalism versus rulegoverned innatism debate. According to Un-Cartesian Linguistics, language does not express thoughts, but imprints thoughts into our mind, i.e. mental representations and mental processes such as the feature of double displacement found in humans only. In other words, a big part of human cognition and human thinking is language based. The possibility to acquire intersubjective knowledge, which is the basic requirement for
Epistemicity, Evidentiality, and Theory of Mind (ToM)
61
the development of culture, is the result of a linguistic technique that we have labeled ‘‘ATMM-complex’’. The language-based approach to cognition presented here is, of course, a universalist approach to language. It is a functional approach to Universal Grammar in the sense of Jakobson and of Shaumyan (2006). We have tried to make evident that there is no general cognitive content common to lexical and grammatical means of modality. Lexical and grammatical means of modality di¤er in function. Their functions are the result of techniques that are purely linguistic in nature. They cannot be derived from some mysterious general cognitive function. References Abraham, Werner 1995 Wieso stehen nicht alle Modalpartikel in allen Satzformen. Zeitschrift fu¨r deutsche Sprache 23: 124–146. Abraham, Werner 2002 Modal verbs: Epistemics in German and English. In: Sjef Barbiers, Frits Beukema and Wim van der Wur¤ (eds.), Modality in interaction with the verbal system, 19–50. (Linguistik Aktuell/Linguistics Today 47.) Amsterdam/Philadelphia: Benjamins. Abraham,Werner 2005 Deutsche Syntax im Sprachenvergleich. Grundlegung einer typologischen Syntax des Deutschen. 2., verbesserte und erweiterte Auflage. (Studien zur deutschen Grammatik 41.) Tu¨bingen: Stau¤enburg. Abraham, Werner 2009 Die Urmasse von Modalita¨t und ihre Ausgliederung. Modalpartikeln anhand von Modalverben, Modalpartikeln und Modus. Was ist das Gemeinsame, was das Trennende, und was steckt dahinter. In: Werner Abraham and Elisabeth Leiss (eds.), Modalita¨t. Epistemik und Evidentialita¨t bei Modalverb, Adverb, Modalpartikel und Modus, 251–302. (Studien zur deutschen Grammatik 77.) Tu¨bingen: Stau¤enburg. Abraham, Werner 2011a, Illocutive force is speaker and information source concern. What this volume type of syntax does the representation of speaker deixis require? Templates versus derivational structure? Abraham, Werner 2011b Preface: Traces of Bu¨hler’s semiotic legacy in Modern Linguistics. In: Karl Bu¨hler: Theory of language. The representational function of language, xiii–xlvii. Transl. D.F. Goodwin. Amsterdam: Benjamins.
62
Elisabeth Leiss
Abraham, Werner to appear, Fremdbewusstseinsabgleich in Syntax und Semantik. In: Zeit2011c schrift fu¨r Sprachwissenschaft. Abraham, Werner and Elisabeth Leiss (eds.) 2007 Modality – aspect interfaces. Implications and typological solutions. (Typological Studies in Language 79.). Amsterdam/Philadelphia: Benjamins. Abraham, Werner and Elisabeth Leiss (eds.) 2009 Modalita¨t. Epistemik und Evidentialita¨t bei Modalverb, Adverb, Modalpartikel und Modus. (Studien zur deutschen Grammatik 77.) Tu¨bingen: Stau¤enburg. Aikhenvald, Alexandra Y. and R. M. W. Dixon (eds.) 2004 Studies in evidentiality. (Typological Studies in Language 54.) Amsterdam/Philadelphia: Benjamins. Aikhenvald, Alexandra Y. 2004 Evidentiality in a typological perspective. In: Alexandra Y. Aikhenvald and R. M. W. Dixon (eds.), Studies in evidentiality, 1–31. (Typological Studies in Language 54.) Amsterdam/ Philadelphia: Benjamins. Andersson, Sven-Gunnar 1989 Zur Interaktion von Temporalita¨t, Modalita¨t, Aspektualita¨t und Aktionsart bei den nichtfuturischen Tempora im Deutschen, Englischen und Schwedischen. In: Werner Abraham and Theo Janssen (eds.), Tempus – Aspekt – Modus. Die lexikalischen und grammatischen Formen in den germanischen Sprachen, 27–49. (Linguistische Arbeiten 237.) Tu¨bingen: Niemeyer. Brandt, Reinhard 2009 Ko¨nnen Tiere denken? Ein Beitrag zur Tierphilosophie. (edition unseld 17.) Frankfurt am Main: Suhrkamp. Bu¨hler, Karl [1934] 1982 Sprachtheorie. (Ullstein Taschenbuch; 1159.) Stuttgart/New York: Gustav Fischer. Bu¨hler, Karl [1934] 2011 Theory of language. The representational function of language. Transl. D.F. Goodwin. Amsterdam: Benjamins (German original [1934]: Sprachtheorie). 2nd revised version. Carnap, Rudolf [1928] 1966 Scheinprobleme in der Philosophie. Das Fremdpsychische und der Realismusstreit. Nachwort von Gu¨nther Patzig. Frankfurt am Main: Suhrkamp [Gedruckt nach der 2. Auflage von: Carnap, Rudolf: Der logische Aufbau der Welt. Erstauflage 1928]. Carnap, Rudolf [1928] 1998 Der logische Aufbau der Welt. (Philosophische Bibliothek 514.) Hamburg: Meiner.
Epistemicity, Evidentiality, and Theory of Mind (ToM) Choi, Soonja 2006
63
Acquisition of modality. In: William Frawley (ed.), The expression of modality, 141–172. Berlin/New York: Mouton de Gruyter. Davidson, Donald [2001] 2004 Subjektiv, intersubjektiv, objektiv. Frankfurt am Main: Suhrkamp (English original [2001]: Subjective, Intersubjective, Objective). Diewald, Gabriele 1991 Deixis und Textsorten im Deutschen. (Tu¨binger Arbeiten zur Germanistischen Linguistik 118.) Tu¨bingen: Niemeyer. Diewald, Gabriele 1999 Die Modalverben im Deutschen. Grammatikalisierung und Polyfunktionalita¨t. (Reihe Germanistische Linguistik 208.) Tu¨bingen: Niemeyer. Doitchinov, Serge 2001 ,,Es kann sein, dass der Junge nach Hause gegangen ist‘‘ – Zum Erstspracherwerb von ko¨nnen in epistemischer Lesart. In: Reimar Mu¨ller and Marga Reis (eds.), Modalita¨t und Modalverben im Deutschen, 111–134 (Linguistische Berichte, Sonderheft 9.) Hamburg: Buske. Doitchinov, Serge 2007 Modalverben in der Kindersprache. Kognitive und linguistische Voraussetzungen fu¨r den Erwerb von epistemischem ko¨nnen. (Studia Grammatica 67.) Berlin: Akademie Verlag. Evans, Vyvyan 2004 The structure of time: Language, meaning, and temporal cognition. Amsterdam/Philadelphia: Benjamins. Frawley, William (ed.) 2006 The expression of modality. Berlin/New York: Mouton de Gruyter. Gelderen, Elly van 2001 The syntax of mood particles in the history of English. Folia Linguistica Historica 22: 301–332. Goldberg, Adele E. 2006 Constructions at work. The nature of generalization in language. Oxford: Oxford University Press. Haan, Ferdinand de 2005a Semantic distinctions of evidentiality. In: Martin Haspelmath et al. (eds.), The world atlas of language structures, 314–317. Oxford: Oxford University Press. Haan, Ferdinand de 2005b Coding of evidentiality. In: Martin Haspelmath et al. (eds.), The world atlas of language structures, 318–321. Oxford: Oxford University Press.
64
Elisabeth Leiss
Haan, Ferdinand de 2006 Typological approaches to modality. In: William Frawley (ed.), The expression of modality, 27–70. Berlin/New York: Mouton de Gruyter. Hare, Brian and Michael Tomasello 2005 Human-like social skills in dogs. Trends in Cognitive Sciences 9: 439–444. Haspelmath, Martin, Matthews S. Dryer, David Gil and Bernard Comrie (eds.) 2005 The world atlas of language structures. Oxford: Oxford University Press. Jakobson, Roman 1957 Shifters, Verbal Categories, and the Russian Verb. Ms., Harvard University. Joseph, Brian D. 2004 Evidentials. Summation, questions, prospects. In: Alexandra Y. Aikhenvald and R. M. W. Dixon (eds.), Studies in evidentiality, 307–327. (Typological Studies in Language 54.) Amsterdam/ Philadelphia: Benjamins. Kratzer, Angelika 1991 Modality. In: Arnim von Stechow and Dieter Wunderlich (eds.), Semantik. Ein internationales Handbuch zeitgeno¨ssischer Forschung, 639–650. (Handbu¨cher zur Sprach- und Kommunikationswissenschaft 6.) Berlin/New York: de Gruyter. Leiss, Elisabeth 1994 Markiertheitszunahme als natu¨rliches Prinzip grammatischer Organisation. In: Klaus-Michael Ko¨pcke (ed.), Funktionale Untersuchungen zur deutschen Nominal- und Verbalmorphologie, 149– 160. Tu¨bingen: Niemeyer. Leiss, Elisabeth 2009a Drei Spielarten der Epistemizita¨t, drei Spielarten der Evidentialita¨t und drei Spielarten des Wissens. In: Werner Abraham and Elisabeth Leiss (eds.), Modalita¨t. Epistemik und Evidentialita¨t bei Modalverb, Adverb, Modalpartikel und Modus, 3–24. (Studien zur deutschen Grammatik 77.) Tu¨bingen: Stau¤enburg. Leiss, Elisabeth 2009b Sprachphilosophie. Berlin/New York: de Gruyter. Leiss, Elisabeth 2009c Konstruktionsgrammatik versus Universalgrammatik. In: Wieland Eins and Friederike Schmo¨e (eds.), Wie wir sprechen und wie wir schreiben: Festschrift fu¨r Helmut Glu¨ck zum 60. Geburtstag, 17– 28. Wiesbaden: Harrassowitz. Nuyts, Jan 2000 Epistemic modality, language and conceptualization. (Human Cognitive Processing 5.) Amsterdam/Philadelphia: Benjamins.
Epistemicity, Evidentiality, and Theory of Mind (ToM)
65
Oehlschla¨ger, Gu¨nther 1989 Zur Syntax und Semantik der Modalverben im Deutschen. (Linguistische Arbeiten 144.) Tu¨bingen: Niemeyer. Papafragou, Anna 2000 Modality: Issues in the semantics-pragmatics interface. (Current Research in the Semantics/Pragmatics Interface 6.) Amsterdam: Elsevier. Papafragou, Anna, Peggy Li, Youngon Choi and Chung-hye Han 2007 Evidentiality in language and cognition. Cognition 103: 253–299. Popper, Karl 1995 Eine Welt der Propensita¨ten. Tu¨bingen: Mohr. Shaumyan, Sebastian 2006 Signs, mind, and reality. A theory of language as the folk model of the world. (Advances in Consciousness Research 65.) Amsterdam/ Philadelphia: Benjamins. Shannon, Claude Elwood and Warren Weaver 1949 The mathematical theory of communication. Urbana, Ill.: University of Illinois Press. Sperber, Deirdre and Wilson, Dan 1986 Relevance. Communication and cognition. Oxford: Blackwell. Tulving, Endel 2005 Episodic memory and autonoesis. Uniquely human? In: Herbert S. Terrace and Janet Metcalfe (eds.), The missing link in cognition. Origins if self-reflexive consciousness, 3–56. Oxford: Oxford University Press.
Illocutive force is speaker and information source concern. What type of syntax does the representation of speaker deixis require? Templates vs. derivational structure? Werner Abraham Universita¨t Wien & Ludwig-Maximilians-Universita¨t Mu¨nchen ‘‘Language [is] a behavior emanating from the human mind.’’ (Nuyts 2000: 333)
1. Foreword: The non-empirical regime of the templatists or superficialists It is interesting to see, and it is certainly devastating for the meaningful concerted e¤ort in modern linguistics to solve the existing and well-known riddles, that di¤erent camps of linguistic thinking ignore one another systematically irrespective of empirical issues that appear to be addressed, or even solved, under the other flag and that one’s own methodology would seem to miss sorely. However, the mutual ignorance is by no means symmetrical. And the methodological virtue is on the side of those thinkers and schools of thinking whose analytic inventories and mechanisms serve deeper analytic purposes than those of others. While adherents to the more analytic camp will usually resort to descriptive work in order to exhaust the range of data to the full, the camps resorting to modules and layers of generalization and (if at all, in the first place) derivation in more superficial, non-formalizing terms will disregard results of the first type – this for the more than plausible reason that scrutinizing formal language usually requires familiarity and determined attempts at sacrificing time and energy to acquire such formalisms and their methodological backgrounds. Thus, one may justifiably say that missing empirical results on the side of the lesser formalists, or, in the nomenclature that I prefer, ‘‘superficialists’’, is due to their unwillingness to entertain the e¤ort and hard work to come to terms with work in the formal camp. One need not only think of modern syntax, semantics, morphology, and phonology, e.g. in the spirit of Generativism. Any semantic logic of the first order, let alone one of higher orders, would seem to likewise deter superficialist linguists.
68
Werner Abraham
The same holds for morphology or phonology: What are the superficialist correspondents to Natural Morphology (a` la Stampe 1972, Donegan and Stampe 2008; Dressler et al. 1987), what are those of generative-inspired approaches to phonology (optimality approaches a` la Prince and Smolensky 2004), where is the lexicalist link between syntax and the lexicon (a` la Marantz 1997; Embick 2004)? Notice, for example, that gradient categorization is a declared major issue in just any superficialist methodology – but not a single representative has made herself master of OT-mechanisms to give substance to her own claim to this empirical descriptive e¤ect.1 The present paper presents a case in point. Nuyts (2000) goes great lengths to argue that epistemic adverbs serve no other purpose than modal verbs (he is silent about modal particles, which likewise anchor to speaker information categories), which fizzle out speaker concern (scoping over the total proposition). While such epistemic adverbials as probably, to the best of my knowledge, certainly, among a variety of others, as well as analytic declaratives such as it is probable/certain that p etc. refer to speaker’s information and his knowledge status, both modal verbs and modal particles go beyond this layer of speaker concern in that they render access to the source of speaker’s knowledge status. Diewald (in her book from 1991, 1. Nuyts’ approach to categorization is remarkable. He claims to have attained epistemic categories by assembling semantic examples of adverbials stopping short, however, of the category of modal particles. Modal verbs are claimed not to di¤er at all in their epistemic quality from epistemic adverbials (Nuyts 2000: 55f.; 98 ¤.). Subclassification with respect to semantic scope (‘‘scoping out’’/‘‘scoping in’’ as in Frey and Pittner 1998 for German or Ernst 2001 for English, or Kratzer 1999, Zimmermann 2004/2008, and Grosz 2009, 2010, distinguishing MP-semantics, on the scoping basis, from conventional implicature), from which empirically important constraints follow, lacks completely. The author’s claim that universal semantic concepts exist before language comes to the fore is breath-taking – and telling in the following sense: How would humans be able to refer to their prelinguistic conceptualizations and knowledge except by linguistic means (Hinzen 2006, 2007)? In other words, the fundamental philosophical debate between Nominalists and Realists is suspended – a fundament without which any school of linguistic thinking is solidly and reliably debased. ‘‘The idea that thoughts can be generated in a language-less mental nirvana, and then get somehow ‘translated’ into language, is a philosophical myth that arose with the Cartesian rationalist tradition.’’ (W. Hinzen Sept. 12, 2008 on the internet, in the context of the discussion of reviews of Chomsky’s Minimalist syntax; see also Hinzen 2006, 2007). Notice that the ‘‘Nirvana assumption’’ mentioned above is in the center of the methodological underpinnings of Construction Grammar.
Illocutive force is speaker and information source concern
69
not listed in the bibliographical list of Nuyts 2000), referring to Jakobson’s (1971) far earlier insight and terminology (‘‘double shifters’’), has called this the ‘‘double deixis’’ of the two categories in question. The present paper undertakes it to anchor ‘‘double deixis’’ in syntax with the aim to underpin the principled deixis di¤erence of epistemic adverbials and modal particles – categories that Nuyts considers to be on a par. This endeavour requires utmost analyticity and empirically optimal generalization across the types of modal lexical in the two di¤erent – verbal and quasi-adverbial, i.e. particle – categories. In pursuing this aim, fundamental claims with respect to the serialization of adverbs and modal particles will be made against the background of investigations entertained by Cinque (1999) and Frey and Pittner (1998), on the one hand, and novel investigations into the classification of embedded sentences as by Haegeman (2006) and Coniglio (2011). The discussion will lead, in an implicative way, to issues such as: If the CP-expansion required for anchoring information about the speaker involves ForceP as the illocutive anchor category in languages like German and Dutch, in languages that lack the categorial double deixis of both modal verbs (MVs) and modal particles (MPs) to the extent that they do not feature either category, the illocutive category must be of a fundamentally di¤erent kind with respect to the pragmatic features that this functional category conveys. This very typological question requires an immensely analytical depth that no other syntax can provide but the generative variety. The fundamental point of departure for what will be said in the subsequent discussion is a dictum by Sperber and Wilson (1986: 4–5): Thoughts do not simply travel from Speaker to Addressee (as in Shannon-Weaver’s 1949 model). We don’t send out messages to addressees when we are engaged in linguistic intercourse. What we do instead is try to make sure what exists, on the part of the addressee, in terms of shared knowledge relevant to the present linguistic interaction and in terms of what is in need of correction on either part. A very telling illustration may be provided by the two German modal lexemes ja ‘yes-particle’ vs. eben ‘flatparticle’: While ja makes an appeal to the addressee’s full consent on the basis of shared encyclopedic knowledge about p, the contribution of eben, rather than appealing to a common knowledge horizon, derives such an appeal from what has been part of the previous discussion between speaker and addressee. We shall come back to this specific contribution of MPs in German in due course (see also Egg, this volume). And we will focus on the di¤erence between epistemic adverbials and MP for the simple reason that adverbials have often been held to serve the same purpose as MPs. This will be shown to be false. But there is a
70
Werner Abraham
further, much farther reaching consequence. If constructions are taken to be the base for semantic interpretation in the sense of idiomatic chunks, without any derivation between pre-particle modals and their homonymic MPs, then the model will not be able to explain any such interactions. In other words, the investigation of MPs will have a bearing on the empirical power of schools of linguistic thinking: e.g., that of Construction Grammar, which has no derivative mechanics at its disposition and takes constructions at their face value, and other, more developed, semantic-syntactic models.
2. Clause structure externally and internally – and the existence of a speaker-source deixis 2.1. A word on modal verbs It has often been claimed that it is not reasonable to analyze modal verbs with respect to their event structure status, since, from the perspective of logical semantics, they di¤er greatly from lexical verbs. Modal verbs and lexical verbs come from two di¤erent domains of grammar (Rothmayr 2009: 234): While lexical verbs select for arguments to form a proposition (i.e., to become fully saturated), modal verbs are operators that act on these propositions. In other words, modals, in contrast to lexical verbs, are propositional operators (as argued for by Lewis 1973; Kratzer 1981, 1991; von Stechow 2004 or Butler 2004, among others). According to the standard logical-semantic view, modal verbs are quantifiers over possible worlds which are interpreted with respect to two entities: a modal base or a conversational background that specifies what the rules are, and an ordering source that determines the order of the possible worlds. The present position is di¤erent for a very good reason: Unless we adopt a syntax-oriented semantics we shall fail to see the common ground on which both modal verbs and modal particles as illocutive operators are based, and why either category likewise involves ForceP as a syntactic clausal representation. In particular, logical semantics, to the best of my own insight, has nothing to say about di¤erent types of dependent clauses with exactly this one di¤erence: Whether or not they are compatible with MPs, thus betraying beyond doubt the presence or absence of the speakerdeixis anchor, ForceP. Clearly, this is a clause operator position. But minimalist syntax has a lot more to o¤er in terms of derivation and the interface between surface structure and Logical Form (LF) to safeguard semantic interpretation; see below in (12) and Section 3.
Illocutive force is speaker and information source concern
71
2.2. Back to modal particles Modal particles modify the illocutive power and quality of sentences decidedly, and their insertion is contingent upon certain illocutive clause types: Thus, denn ‘then’ merges only in interrogative roots, just as much as ja ‘yes’ does only in declaratives. For modal verbs this is less clear. But see (1), which shows that certain modalities are incompatible with certain person-speaker distributions. [Subscripts ‘‘?’’ and ‘‘!’’ indicate the class allocation constraints for the two modal particles.] (1) a.
b. (2) a.
b.
Wie siehst DU aus? how look part? you part out ‘What do you look like after all?’
. . . wh-question
*Wie siehst DU aus? how look part you part out aus! . . . emphatic exclamation DU siehst ja you look part! out ‘Why, you are you looking like someone!’ *DU siehst denn aus! you look part? out
The illocutive restrictions denn and ja are subject to in (1) and (2) invite the conclusion that modal particles are modifiers of the general illocution force, ForceP (or IllocutionP) scoping over the entire clause. In the specific case of (1), the general illocutive quality is , with denn subclassifying the interrogative force in terms of the speaker’s concern (following Bayer 2008, ) with the reasons for the addressee’s (du) anomalous looks. On the other hand, ja in (2) serves the speaker’s purpose of giving strong expression to his surprise or dismay at the addressee’s anomalous looks. Thus, such MPs have to have exponency in the respective syntactic function for clausal illocution, ForceP, as well as a specific feature representing the di¤erential property of denn as opposed to ja. Needless to say, in order to illocutively wide-scope over the entire sentence, the MP has to move to ForceP, the functional category licensing illocutive force for the entire sentence. See (3) for what has come to be called Rizzi’s CP-expansion (Rizzi 1997, 2001) and for the integration of MPs into a similar syntactic structure in Logical Form (LF) Abraham (1995a,b). (3) Fine structure of the left sentence periphery: ForceP/IllocP/SpeakerDeixisP >FocP >TopP >FinP/CP > TP >MP >VP >V
72
Werner Abraham
ForceP is where illocutive force sits. Since such force is speaker related and since speaker relations are triggered by any MP and epistemic modal verbs, this syntactic category is the one that will figure most centrally in this discussion. One can also say that this category hosts primary discourse relations and relations to the speaker’s encyclopedic knowledge. FinP contains fundamental temporal and finiteness information with respect to the core sentence. TopP hosts unhighlighted fronted elements, while FocP is the position where elements under contrastive stress land. The triggering device for derivative movement of the MP is the feature responsible for speaker deixis: (following a suggestion by Bayer 2008; see also Bayer and Obenauer 2011). ForceP is thus the function responsible for the illocutionary type (interrogative, declarative, imperative etc.), while the MPs subclassify these core illocutions in line with the specific properties of the individual MPs: In the case of denn, it is the speaker’s concern to make sure about the deeper reasons for addressee’s behavior or activity; and for ja it is the speaker’s intention to signal to the addressee that he has presupposed part of the meaning of the core sentence. Thus, is the feature of the MP in Spec,vP that drives the derivation by movement all the way up to ForceP. See (50) below for this derivation initiated and driven through by the feature . For modal verbs this is less clear. But consider (4)–(5), which show that modalities are incompatible with certain person-speaker distributions. (4) a.
Er soll in Gro¨nland gewesen sein. he shall.mv in Greenland been be ‘He is supposed to have been in Greenland.’
b.
Ich soll in G. gewesen sein. I shall.mv in G. been be ‘I am supposed to have been in Greenland.’
(5) a.
Er will in Gro¨nland gewesen sein. he will.mv in Greenland been be ‘He pretends to have been in Greenland.’
b.
*Ich will in G. gewesen sein. I will.mv in G. been be
I assume that epistemic (but not modal root) readings of MVs also have to move to ForceP to satisfy such speaker constraints as shown in (4)–(5) above. However, since the illocutive force of MVs is more complex than that of MPs, I will not consider their contribution in the remainder of
Illocutive force is speaker and information source concern
73
Figure 1. Genuine subordination of a conditional CP (for event conditionals see Haegeman 2002: 131; Coniglio 2011: 4)
this paper and stick to MPs. The determining factor for our argument is that ForceP, the seat of illocutive power and speaker anchor, is confined to root sentences. Truly embedded clauses do not carry independent illocutive power, and consequently, have no functional category ForceP and are thus no man’s land for MPs. It is easy to devise such structures where, despite a subjunctive Comp, no true embedding takes place; see Figures 1 and 2 and the respective illustrations in (6).
Figure 2. Associated clauses, CP1 and CP2 – no genuine subordination despite formal conditional (for premise conditionals see Haegeman 2002: 131; Coniglio 2011: 4)
Fig. 1 structures an if-event clause like (6ai), Fig. 2 analyses a premise if-clause like (6bii/iii). (6) a. b.
If it rains-p we will all get terribly wet-q (i) Wenn es regnet-p, werden wir alle schrecklich nass-q. If [as you say-r] it is going to rain this afternoon-p, why don’t we just stay at home and watch TV-q (ii) Wenn [wie du sagst-r] es heute nachmittag regnet-p – warum bleiben wir dann nicht einfach zuhause und schauen uns einen Videofilm an-q? ‘If [as you say] it will be raining this afternoon – why then don’t we stay back home?’ (iii) Wenn du sagst-r, dass es heute nachmittag regnet-p. Dann bleiben wir doch lieber zuhause und . . . !-q ‘If you say that it will be raining this afternoon. Then we had rather stay back home.’
The German translations in (6bii/iii) are telling distributions: p is a premise with respect to the associated q, neither of which stands in a clear
74
Werner Abraham
eventive chain to one another, thus allowing for non-embedding paraphrases and speech act specifics as (ii) and, even more telling, the coordination of p and q in (iii). By contrast, p and q in (6a) are members of a clear causally linked chain of events excluding coordinations as in (i) and (iii). The two clause structures are thus clearly di¤erent as between (7) and (8); see te Velde (2005) for parallel solutions. (7)
(8)
3. MP-compatibility and ForceP In the light of the fact that MPs presuppose and modify the illocutionary force of MPs (and epistemic modal verbs, EMVs), root clauses are taken to be the best candidates for the insertion of MPs. Needless to say that not every MP appears in every illocutive type: denn occurs only in formal wand yes-no-interrogatives (however, not in echo questions – see 9c below); ja occurs only in declaratives – see (10). Notice that (question) form and illocutive function do not converge: (9a,b) are questions by word order and wh-word inception, but they are emphatic exclamations by function. (9) a.
b.
Wie siehst denn DU aus? . . . speaker-concerned how look part? you out question-exclamation ‘What, for god’s sake, are you looking like!’ Wie siehst du denn AUS?! . . . w-question-exclamation how look you part? out ‘What, for god’s sake, are you looking like!’
c. *Du siehst denn WIE aus? you look part? out
. . . *echo question
d.
. . . *declarative
*Du siehst denn schlimm aus. you look part? bad out
Illocutive force is speaker and information source concern
(10) a. b.
75
Du siehst ja (SCHLIMM) aus! . . . speaker-concerned you look part! (bad) out declarative exclamation *Wie siehst du ja AUS? how look you part! out
. . . *w-question
c. *Siehst du ja SCHLIMM aus? . . . *yes-no question look you part! bad out However, the question of whether or not embedded sentences can host MPs is not well understood; see the following illustrations, (11a–d), which yield no clear clue. (11) a.
Mir scheint, dass er ja la¨ngst da sein mu¨sste. me seems that he part! already there be should ‘I believe he should have been here all along.’
b.
*Er leugnete, dass er ja la¨ngst da sein mu¨sste. he denied that he part! already there be should
c. Sie fragte, ob er denn kommen wolle. she asked whether he part? come would ‘She asked whether he would really like to come.’ d.
*Sie wusste nur allzu gut, ob er denn kommen wolle. she knew just all too well whether he part? come would
(9)–(10) lead us to generalize that independent (root) clauses are MPcompatible by default as long as the speech act contingencies of the individual MPs are met. How about dependent clauses – do they principally disallow MPs? The minimal conclusion given in (11a–d) is that there are such dependent clauses allowing for MP-insertions. Consequently, we would have to impute to such dependent clauses the illocutive clause node ForceP. Other dependent clauses, however, will not possess their own illocution and, consequently, project no ForceP. See (12) (following Haegeman 2006; see also Coniglio 2011). (12) a. dependent core clauses: Subj – – – TopP FinP . . . *MP b. dependent peripheral clauses: Subj ForceP TopP FocusP TopP FinP . . . MP c. independent clauses: – Force P TopP FocusP TopP FinP . . . MP
76
Werner Abraham
Notice the compatibility correlation between MP und ForceP. What is at stake is to separate (12a) and (12b) distributionally: There are MPcompatible dependent clauses – since possessing their own, independent illocution which has not been pre-assigned by the governing independent clause; and there are MP-incompatible clauses which are not in the possession of independent illocutionary power. The question is: What is the specific quality of dependent peripheral clauses such that they possess their own illocutionary power and, consequently, allow MPs to be merged in Spec,vP and raise to SpecForceP/SpecIllocP (Abraham 1995a,b; now also Zimmermann 2004/2008). 4. Clausal subclassification under the ForceP-criterion There are well-argued classifications for (12a) vs. (12b) – see Haegeman (2002, 2004a,b, 2006) and Coniglio (2011), among others. But generalizations as to why it is that MP-insertion is possible only in the (12b)-class of dependents are far from trivial and, much less, clearly predesigned. For example, non-factive embedded complement clauses tolerate MPs, but factive embeddings do not (Coniglio 2011: 9–11); see (13)–(15): The examples in (14) are non-factive complement clauses, those in (15), by contrast, are factives. (13) may be assumed to list factives – which are consequently expected to exclude MPs. But they don’t.2 (13) a.
b.
Leider verschlechtern sich die Forschungsleistungen wohl weiter. pity deteriorate refl the research results part further ‘It’s a pity: As appears the research results are further deteriorating.’ Leider werden sich die Forschungsleistungen pity will refl the research results wohl weiter verschlechtern. part further deteriorate ‘It’s a pity: The research results will to all appearances be further deteriorating.’
2. Zimmermann (2004, 2008) makes a fundamental distinction between wohl as a modal particle (or expressive lexical, as he calls it drawing on a terminology developed by Potts 2002) and other expressives such as parentheticals and curse attributes, e.g. this damn guy, on grounds of conventional implicatures: MPs are not conventional implicatures in that they are not part of the propositional content (they do not scope into p).
Illocutive force is speaker and information source concern
77
c. Der Rektor bedauerte, dass sich die Forschungsleistungen the president regretted that refl the research results wohl weiter verschlechtern. part further deteriorate ‘The president regretted that the research results would be likely to further deteriorate.’ d.
Wahrscheinlich verschlechtern sich die Forschungsleistungen probably deteriorate refl the research results wohl weiter. part further ‘Probably, the research results will deteriorate further.’
e. Es ist absolut wahrscheinlich, dass sich die Forschungsleistungen wohl weiter verschlechtern. (14) a.
Er glaubt, die Leistungen werden wohl schwa¨cher werden. he believes the results will part weaker become ‘He believes that the results will go down.’
b.
Es fa¨llt mir gerade ein, dass er doch noch nicht eingetro¤en it occurs to me that he part yet not arrived sein kann. be can ‘It just occurs to me that he cannot have arrived as yet.’
(15) a.
*Er ha¨lt fu¨r unwahr, dass sie he considers untrue that she ja/WOHL/doch an Selbstzweifeln leidet. part!/part!,?/part! from self-doubts su¤ers
b.
*Es stimmt nicht, dass sie ja/doch/DOCH unvorbereitet ist. it is correct not that she part!/part!,?/part! unprepared is
Why do (13a–e), obviously being factives, tolerate MPs? Note, first, that factivity boils down to a present evaluation of an eventive or state. The German present tense, notably if triggered even more clearly by weiter, always carries the ambiguity between simultaneous and futurate reference, as in (13a, c, d, e). Furthermore, verschlechtern ‘worsen’ is a prospective perfective enhancing the non-simultaneous reading. Thus, we summarize
78
Werner Abraham
that seemingly factive predications may be non-factive on the basis of a wide variety of lexical and grammatical predispositions. What stands solidly so far is the conclusion that independent illocutionary force is required for MPs to surface. And we have seen that there are dependent clauses that allow illocutions speaker-anchored MP insertions of the type in (12b). There is an interesting division between the core and peripheral dependent clauses. Local and temporal adverbial clauses as well as restrictive relative clauses do not allow for MP-illocutionary force, while causal and adversative as well as non-restrictive relative clauses do. Since it is often the case that local and temporal subjunctive lexicals lend themselves to either causal or adversative readings, the insertion of MPs may decide between the two clause classifications – and status with respect to ForceP inclusion. In what follows this will be illustrated and discussed.
5. Locals-temporals vs. causals-adversatives and the restrictive relative criterion Locals-temporals as well as restrictive relative clauses are core dependents of the structural configuration in (16a), while causals-adversatives and non-restrictive relatives classify as (16b). Correspondingly, ForceP-bare clauses disallow MPs. (16) a. *Er las vormittags, wa¨hrend Mutter ja einkaufen war. he read in the morning while mom part! errands ran . . . temporal reading-*MP b. Er faulenzte, wa¨hrend SIE ja/wohl arbeitete. he lazed while she part!/part!,? worked ‘He lazed around while she was working alright.’ . . . adversative reading-MP c. *Nachdem sie ja eingetro¤en war, putzte sie gleich. since she part!,? arrived was cleaned she immediately . . . temporal reading-*MP d. Nachdem sie ja jetzt eingetro¤en war, konnte er sie fragen. since she part! now arrived was could he her ask ‘Since she had arrived alright he could ask her.’ . . . causal reading-MP
Illocutive force is speaker and information source concern
79
e. *Es ist dort scho¨n, wo er ja/eben/doch aufgewachsen ist. it is there nice where he part!/part!/part!,? grown up is . . . local reading-*MP f. Sie hassen ihn, wo er doch dort aufgewachsen ist. is they hate him where he part! there grown up ‘They hate him while, in fact, he grew up there.’ . . . modal reading-MP Likewise, conditionals can be divided into those that map events as against those that map premises. Only the premise type allows for MPs. Indirect questions side with the latter. (17) a. *Er kommt sich gut vor, wenn er schon/ja/eben he considers himself good as soon as he part!/part!/part! angeben kann. show o¤ can . . . event conditional-*MP b. Er sollte still sein, wenn er schon/ja doch so viele he should quiet be as long as he part!/part! so many Vorrechte hat. prerogatives has . . . discourse conditional-MP ‘He should shut up as long as he has such prerogatives.’ c. Ob sie denn/wohl/*ja gelernt habe, wollte er wissen. if she part?/part!,?/part! studied had wanted he know ‘He wanted to know whether she had really studied.’ . . . indirect question-MP (17b) is not a conditional chain in the direct sense of an if-then-event, but, much rather, one where a meta-type of reasoning takes place. Note, also, that the participation of MPs in indirect clauses is subject to the same constraint as in direct questions. Both purposive and consecutive clauses are open to MPs. keiner sieht. (18) a. Versteck dich, damit dich ja/wohl/eben/halt hide not so that you part!/part!,?/part/part no one sees ‘Do not hide such that no one can find you alright.’ . . . purpose clause-MP b. Behalte es fu¨r so fu¨r dich, Keep it so to yourself dass dir ja/wohl/eben niemand etwas an kann. one can harm that you part?/part!,?/part! no ‘Keep it to yourself that no one is likely to harm you.’ . . . consecutive clause-MP
80
Werner Abraham
Finally, restrictive relative clauses are MP-incompatible, while nonrestrictive ones allow for MP qualification. (19) a.
b.
*DER Hans, der ja/doch dort dru¨ben wohnt, ist blind. the John who part!/part there over lives is blind . . . restrictive relative-MP Hans, der ja/doch dort dru¨ben wohnt, ist blind. ‘John who lives over there alright, is blind.’ . . . non-restrictive relative-MP
6. The deeper reason behind the discrimination by illocutionary force in dependent clauses It is an absolutely non-trivial question to ask what the reason is behind the two types of dependent clauses: The ones that have their own illocutive power and thereby license MPs, and the others that do not; see (20), which takes this to a minimal formulation. (20) a. b.
Core dependents: Subj – – – Fin Peripheral dependents: Subj Top Foc Force Fin
The question of whether or not MPs scope out of dependent clauses was first raised by Kratzer (1999), although not on a general line so as to involve all kinds of dependent clause types. The question is which deeper properties distinguish core and peripheral dependents such that only the latter has illocutive power in its own right. Following Haegeman (2004, 2006), Lahousse (2010), and Abraham (2012), I argue that it is only Force-bearing clauses that may encompass illocutive power in its own right and consequently host MPs. To come to terms with this problem let us see what we can say about the deeper distinction between the two types of relative clauses. Restrictive relatives execute a choice function on the head DP singling out the individual qualified by the relative clause. Their reference is di¤erent from that of the head DP, although the extension of the relative is implied by the extension of the head. By contrast, the reference of non-restrictive relatives is identical to that of the head. All the relative clause does is point out a quality that, by itself, co-qualifies with the head DP. Proper names such as Hans in (19a, b) usually denote one singular individual, but determiner words and accent may open reference to a set of individuals with an identical name as in (19a) and, thus, license restrictive attributive
Illocutive force is speaker and information source concern
81
modification; see the concomitant properties of the two relative clauses in Table 1. Table 1. Complementary distributions for the two relative types Relative
Foregrounding
Independent extension
Focus
Coordination
Parenthesis
Epistemic adverb
þRestricted
þ
þ
þ
–
–
–
–Restricted
–
–
–
þ
þ
þ
Consider the following illustrations. Restrictive relative clauses behave like foregrounded and focused attributes, while non-restrictive ones are characterized as appositive modifications of the head: backgrounded and unfocused. (21) a.
Der *ja dort DRU¨BEN wohnende Hans ist blind. the part! there over living John is blind ‘That John living over there alright is blind.’ . . . restrictive/foregrounded/focused/extension independent
b.
Der ja dort dru¨ben wohnende HANS ist blind. the part! living John is blind ‘John who is living over there alright is blind.’ . . . non-restrictive/backgrounded/unfocused (default focused)/ extension implied
c. Hans ist blind. Er wohnt wohl/ja dort dru¨ben. John is blind. He lives part!,?/part! there over ‘John is blind. He lives over there alright.’ . . . coordinability only for non-restrictives d.
Hans – er wohnt wohl dort dru¨ben – ist blind. John – he lives part!,? there over – is blind ‘John – he lives over there alright – is blind.’ . . . parenthesis only for non-restrictives
e. . . . , der den Schlu¨ssel wohl finden wird. . . . who the key part!,? find will ‘. . . who will find the keys alright.’ f.
. . . MP
. . . , der den Schlu¨ssel sehr wohl finden wird. . . . who the key very well find will . . . equivalent epistemic adverb
82
Werner Abraham
g.
. . . , der den Schlu¨ssel sehr wahrscheinlich finden wird. . . . who the key very probably find will ‘. . . who will very probably find the key.’ . . . near-equivalent epistemic adverb
German has lexical and grammatical means to disambiguate restrictive and non-restrictive relatives. Indefinite determiners such as (der)jenig- ‘the one’, ein solch- ‘such a’, DER/DIE/DAS ‘that one’, EIN- ‘one such’ are main clause correlates singling out individuals and, thus, triggering the restrictive reading. As for what appears to be a gradual transfer from MP-status to adverbial status in (21e–g) see already Thurmaier (1989: 140), also with respect to wohl ‘well’. In the absence of such choice function elements, relative clauses may be ambiguous between the two readings. This is most clearly brought out by optional MP insertion; see (22). (22) a.
Voyeurs, die ja nichts verpassen wollen, stellen sich an jedes Guckloch. voyeurs who PART! nothing miss want put themselves up at every eye hole ‘Voyeurs anyway not wanting to miss anything put themselves up at every eye hole.’ . . . non-restrictive
b.
Voyeurs, die ja nichts verpassen wollen, stellen sich an jedes Guckloch. voyeurs who focal-PART! nothing miss want put themselves up at every eye hole ‘Those voyeurs who do not want to miss anything under any circumstances put themselves up at every eye hole.’ . . . (non-)restrictive
c. Jene Voyeurs, die ja /*ja nichts verpassen wollen, stellen sich an jedes Guckloch. voyeurs who focal-PART! /*non-focal ja nothing miss want put themselves up at every eye hole ‘Those voyeurs who do not want to miss anything under any circumstances put themselves up at every eye hole.’ While in (22a–c) accent on the MP disambiguates, other MP insertions are less clear – and leave it open whether or not the clause possesses ForceP, i.e. illocutionary power, in its own right. [rMC ¼ restrictive correlate in the matrix clause; nrEC ¼ non-restrictive embedded clause].
Illocutive force is speaker and information source concern
83
(23) a.
Sie will ein solches/dasjenige Auto, das es hier she wants a such/that one car that it here ja/wohl gar nicht gibt. part!/part!,? at all not gives ‘She wants such a car that is not likely to be had here.’ . . . rMC, nrEC
b.
Sie will ein Auto, das es hier ja/wohl gar nicht gibt. she wants a car that it here part!/part!,? at all not gives ‘She wants a car that is not likely to be had here.’ . . . nrMC, nrEC
(24) a.
b.
*Es sollen nur DIE Tauben vernichtet werden, die it shall only those pigeons killed be which ja wohl den ganzen Dreck machen. PART! PART!,? provide all the shit Es sollen nur die TAUBEN vernichtet werden, die it shall only the focal-pigeons killed be which ja wohl den ganzen Dreck machen. PART! PART!,? provide all the shit ‘Just the pigeons should be killed that are responsible for the shit in the first place.’ . . . nr
c. Es sollen nur die Tauben vernichtet werden, die it shall only the pigeons killed be which ja wohl den ganzen Dreck machen. part! part!,? provide all the shit ‘Just the pigeons should be killed which are responsible for the shit in the first place.’ . . . (n)r (24c) is open with respect to restrictivity of the relative clause. The choice function criterion may also be triggered by specific interrogative contexts (as shown by Coniglio 2011: 22/fn. 36). (25) a. b.
Welchen Schlu¨ssel soll ich nehmen? ‘Which key should I take?’
– *(25c) . . . wh-question
Soll ich etwas mitnehmen? ‘Am I supposed to take along anything?’ – ok(25c) . . . yes-no question
84
Werner Abraham
c. Nimm den Schlu¨ssel, der wohl auf dem Tisch liegt. ‘Hold on to the key which PART!,? is lying on the table.’ . . . nr relative clause þ MP (25c) is an appropriate answer only to the question in (25b), not to that in (25a) – which confirms our conclusion drawn above about the choice functional criterion attesting restrictivity of the relative. Notice, however, that reference to, and the solution for, adnominal dependent clauses ((non-)restrictive relative clauses) and the contextual use of MPs does not settle the question raised at the outset of this section. There are also complement dependent clauses (factive as well as nonfactive) and adverbial clausal dependents (temporal, local, causal, adversative, concessive, etc.), to which the backgrounding solution holding, with some plausibility, for adnominal dependents cannot so plausibly be extended. We leave this for later treatment (see Abraham 2012, submitted). 7. Veridical semantics and MP-modality Veridicality as well as non- and antiveridicality, respectively, are concepts di¤erentially considering the intensional properties of predicates (in the widest sense, thus including word classes other than verbs). In other words, veridicality properties take into consideration individual anchors for truth evaluability. This also means that the modal licensing is anchored not only to predicative finiteness, but beyond that to the level of illocution and speech acts. This is equivalent to a higher intensional level – expressed by feature sets and their hierarchical structure as in the representations of (34)–(38) below. The veridicality of predicates such as glauben/believe, tra¨umen/dream, sagen/say, behaupten/maintain – also non-factives – only presupposes that the modalized proposition, p, is true in one of the accessible possible worlds, irrespective in which of the alternative ones – i.e., by no means in all worlds as assumed for normal truth (cf. the feature corresponding to normality in (34c) below. In other words, the veridicality with respect to the world model shared by the speaker or, under speaker shift, with respect to the corresponding world model shared by the subject, is disclosed. Since p is true in the model of normative worlds, MB(s), as well as in the individual world model associated with the subject, MB(su), it counts as a strong veridical predicate. Predicates such as wissen/know are factive and are consequently classified as strong veridical properties, together with aspectual, commissive and implicative predicates (Giannakidou 1998: 114f.). They exclude MPs in the complement clause. From this follows
Illocutive force is speaker and information source concern
85
that MP-worthy predicates classify as weak veridical predicates. In other words, MP-bearing predicates (irrespective of whether with syntactically matrix or subordinate status) underlie restrictions on their world accessibility where p is valid only in the embedded model. This holds for epistemics as well as worlds of dream and fiction.3 Let us consider which MPs are compatible with which weakly veridical modality operators. Modal particles/MPs specify the speech act of the speaker of the sentence (of the proposition) p: for example, by means of halt, eben, schon, ja, wohl, nur, bloß, doch in a sentence like (26). (26) Paul soll Maria {halt / eben / SCHON / ja / JA / wohl / nur / bloß / doch} zu einer Aussprache einladen. Paul shall Maria {MP} for a discussion invite ‘‘Paul is supposed to invite Maria for a discussion.’’ With (26) the speaker asserts that Paul is supposed to invite Maria for a discussion and that he modifies this assertion toward the hearer/addressee in such a way that, approximately, (27) a.
b. c. d.
e. f. g.
he considers his assertion modified by wohl as a statement less solidly truth-ascertainable (for example, as something but reported by others); with doch, the speaker’s statement is not compatible with beliefs and assumptions shared by the hearer; with ja, the speaker’s statement meets the presuppositions shared by the hearer; with halt and eben, the speaker’s statement highlights a fact, which, given the discoursive context, does not deserve to be highlighted; with nur or bloß, the speaker attributes to the content of his statement lesser validation or weaker truth probability; with SCHON, the speaker’s p contradicts the hearer’s statement; with JA the speaker, given the high expectations he has with resoect to the addressee’s positive confirmation of his statement p, attributes a high commissive staztus to his statement p.
3. Strongly intensional verbs are non-veridical ( just as well as weak ones). The strong class of intensional verbs includes wu¨nschen/wish, wollen/want. The corresponding licensing and anchor model is the Belief, or Assume, model that the su(bject) unfolds in terms of future realizations in the real world: MBfut(su) – which, quite plausibly, holds generally for directives: i.e., beyond wu¨nschen/wish, wollen/want aslo for raten/advise, vorschlagen/propose, bitten/ ask.
86
Werner Abraham
Following the tenets of dynamic semantics (Stalnaker 1978), we posit the model-semantic representations in (28) for the MP-assertions in (27a–g): here restricted to the MP wohl as well as the EMV sollen. (28) Assume/Belief model: valid for the MP wohl Be c ¼ a context (the so-called Kaplan parameter) A model MB(x) a M is a set of worlds associated with an individual x representing worlds compatible with what x assumes/believes. [B ¼ believe(/assume) state, RC ¼ report communication, W ¼ world, cg ¼ common ground, s ¼ speaker, h ¼ hearer / addressee, wo ¼ world in which the utterance takes place, f ¼ function for validating variables, whereunder time and location of the utterance, W(c) ¼ set of contexts, c ¼ context]. (29) Reportative model: valid for the EMV sollen Be c ¼ a context. A model MRC(x) a M is a set of worlds associated with an individual x, which x considers compatible with reported worlds. It is assumed that weak veridical predicates such as non-factives, modaladverbial subordinates, and appositive (non-restrictive) relative clauses scope structurally higher than truth-evaluable models of worlds. This is for the simple reason that their world model is constrained by assumptions and beliefs of an individual speaker, thus intensionally richer. The sets of intensional features holding for worlds of individual speakers (and hearers) are larger than those of the worlds holding for normally valid worlds (i.e. worlds with generally accepted accessible worlds, with a largely undisputed Common ground). Notice that this assumption is based on the distinction of two types of Common ground: one based on the ongoing negotiation between speaker and addressee – we may call this a dynamically developed Common ground/DCG; and another one based on a more static belief (and assumption) Common ground/BCG containing more static and discourseindependent belief models of accessible worlds. The first CG-notion, DCG, is process-oriented. The participants in a specific discourse display to each other to what extent they feel prepared to accept certain stages of CG to be commonly presupposable to justify a successful continuing discourse. The second notion, BCG, lies at the base of the first dynamic notion. It may, but need not, be linked, and in fact is often neglected, or not realized
Illocutive force is speaker and information source concern
87
by the discourse participants to be linked (and committing) to the dynamic one, DCG.4 The discussion above justifies that the following steps of relativized veridicality are brought into play for the semantic and pragmatic description of modal particles. (30) Relativized (non-)veridicality for MP: Be c ¼ a context. i. A propositional operator/Op is veridical i¤: [[Op p]] c ¼ 1 ! [[p]] ¼ 1 in some epistemic models M(x) a c; otherwise, Op is nonveridical. ii. A non-veridical operator Op is antiveridical, i¤: [[Op p]] c ¼ 1 ! [[p]] ¼ 0 in some epistemic models MB(x) a c. iii. Epistemic models are among others: Assume/Belief models, MB(x), Reported Conversation models, MRC(x). Compare (28) and (29). It appears that such considerations as to the weakly veridical potential of MPs and to their highly intensional feature status including speaker and hearer worlds justify the assumption that the syntactic status of MPs in ForceP is cartographically above that of modal adverbials, but on a par with EMV (as opposed to DMV, which, as on a par with fully lexical verbs, is lower in VP).5 Since surface-distributional di¤erences have only been investigated rarely, it appeared to be a necessary, and adequate, step to point out this semantically motivated representational option. 8. Strong MP-deixis vs. weaker adverbial deixis The question is to be raised whether it is true that epistemic adverbials have the same speaker deixis as MPs. Let us find out what exactly it is that MPs trigger in terms of speaker and source deixis and then compare the result with adverbials. MPs have a double deixis (in the sense of Jakobson’s shifter qualities; see Jakobson 1971: 130 ¤.): One identifying the source of p, the proposition, 4. This distinction of two types of Common ground reflects the more general and basic discussion represented, for example, by Fetzer and Fischer (2007), Fischer (2007) and Pittner (2007). 5. I take it that this is compatible with Coniglio’s assumption that MPs have IPstatus (Coniglio 2011).
88
Werner Abraham
and one identifying the speaker’s attitude towards p. The latter is what modality is about in its core meaning: the modal shade that the speaker attributes to p. MPs do this throughout their paradigms: i.e., through their person references, their number references, their tenses and aspectual expressive; see (31)¤. below. (31) a. b.
der den Schlu¨ssel wohl finden wird. who the key PART!,? find will
. . . MP
der den Schlu¨ssel sehr wohl finden wird. who the key very well find will . . . equivalent epistemic adverb
c. der den Schlu¨ssel sehr wahrscheinlich finden wird. who the key very probably find will . . . near-equivalent epistemic adverb d.
der den Schlu¨ssel wohl finden wird. who the key PART!,? find will
. . . MP
Note that the elative comparison in sehr wohl ‘very well’ would point at adverbial status. Now if this is really so we would expect certain restrictions in their combinatorial properties. Consider the following attempts including the adverbials leider ‘sorry(ly)’ and wahrscheinlich ‘probably’ as well as MPoid wohl ‘well’. How well do the three combine, in which order, and what can we deduce from this evidence? (32) a.
*der den Schlu¨ssel leider wohl nicht finden wird. who the key sorry well not find will . . . *modal epistemic adverb/Moepadv þ MPoid b. der den Schlu¨ssel wohl leider nicht finden wird. who the key well sorry not find will ‘who won’t find the key, I am sorry to say.’ . . . MPoid þ Moepadv c. der den Schlu¨ssel leider eben doch nicht finden wird. who the key well sorry not find will ‘who won’t find the key all the like, I am sorry to say.’ . . . Moepadv þ MP þ focal-MP d.
*der den Schlu¨ssel eben doc leider nicht finden wird. who the key PART! PART!,? sorry not find will . . . MPoid þ Moepadv
Illocutive force is speaker and information source concern
89
(33) a.
*der den Schlu¨ssel leider wahrscheinlich nicht finden wird. who the key sorry probably not find will . . . Moepadv þ epistemic adverb
b.
der den Schlu¨ssel wahrscheinlich leider nicht finden wird. who the key probably sorry not find will ‘who will probably not find the key, I am sorry to say.’ . . . epistemic adverb þ Moepadv
c. der den Schlu¨ssel eben leider doch nicht finden kann. who the key PART! sorry PART! not find will ‘who will not find the key all the same, I am sorry to say.’ d.
der den Schlu¨ssel eben doch who the key PART! PART! nicht finden kann. not find will ‘who will not find the key all the same, I am sorry to say.’
e. der den Schlu¨ssel wahrscheinlich who the key probably doch nicht finden kann. PART! not find will ‘who will probably not find the key, I am sorry to say.’ f.
der den Schlu¨ssel ??wahrscheinlich doch who the key probably PART! nicht finden kann. not find will ‘who will probably not find the key, I am sorry to say.’
From (32a–b) one may conclude that leider ‘sorry’ scopes over wohl, not, however, vice versa. Since MPs, being ForceP-heads, have widest sentential scope, wohl cannot be a true MP – which given (31b) is what we suspected in the first place. There are thus MPs of di¤erent categorial belongings: squishes in the case of wohl. This may have to do with a noncompleted grammaticalizing status as compared with most of the other MP lexemes. Notice also that while the categorial squish wohl has adverbial replacers such as wahrscheinlich ‘probably’ and vermutlich ‘presumably’, hardly any of the other MPs has one: eben, ja, schon, nur / bloß,
90
Werner Abraham
denn, etc. This would appear to point at a less far grammaticalized status of wohl in comparison to other MPs (counter to Zimmermann 2004/2008, who essentially treats wohl on a par with other MPs).6 One would be illadvised to base one’s MP generalizations on wohl alone. Furthermore, (33e–f ) show that epistemic adverbials scope below MPs (at least below eben) as well as leider.7 This is in line with Thurmaier’s (1989: 139f.) as well as Coniglio’s (2011: 22) conclusions that E-adverbials like wahrscheinlich ‘probably’ and vermutlich ‘presumably’ do not possess illocutive power in the same way as MPs: They modify the proposition, but they do not raise to the illocutive category of ForceP. Let us see what E-adverbials miss in comparison to MPs.8 In Minimalist terms, any (covert) raising movement to ForceP, the illocutive function, must be licensed by content features that have to be verified by complementary linguistic distributions and thus creating alignment between the source (merge) position and the eventual final position extending the actual scope relations. Let us make concrete these di¤erences for epistemic modal verbs, for Ep- and Ev(identical) adverbials, and for MPs. In essence, this is an exercise in Sperber-Wilson’s (1986) relevance dimension. Let us claim the following.9
6. It is interesting to see that other lexicals such as ruhig ‘quiet; peaceful’ appear to share the properties of MPs on all counts as Grosz (2009) has shown – albeit in combination with doch and bloß/nur to yield imperatives such as Mach das doch/nur/bloß ruhig! standing approximately for English ‘‘Do this all the same!’’ 7. As opposed to E-adverbials, leider ‘sorry’ seems to scope on a par with MPs such as eben ‘all along’. We leave it at that for the time being. What (33a–b) exactly mean in the context of our discussion is not entirely clear. 8. While, more than often, modal particles of the German type appear to be tough to distinguish from E-adverbials, clearly, one of the criteria characteristics of MPs is polyfunctionality such that a diachronic source lexical with clear lexical classificatory properties is homonymic to the MP, which is bare of lexical semantics. See, for example, Aboh 2006 on Saramaccan modal particles, which are not polyfunctional. Much rather, as Aboh claims, they appear to be adverbs. They translate as operators inducing obligation, injunction, and exhortation (Aboh 2006: 12f.). No epistemics (FCA [or ToM] operators, in the sense of the present article; WA), in any of the complement-dependent contexts, have been reported. 9. This section has gained much from discussions with Elisabeth Leiss.
Illocutive force is speaker and information source concern
(34) a.
b.
91
Epistemic modal verbs/EMVs denote a twofold deixis: According to the source of p, and another one according to the speaker’s assessment of p. Haider soll betrunken gewesen sein (with p ¼ Haider ist betrunken gewesen) (Haider was drunk) ‘Haider is supposed to have been drunk.’
c.
We take as a point of departure Davidson’s (2004) threefold distinction of mind awareness (see Leiss 2008, this volume) – see (35–38) on this basis projecting both speaker and addressee as well as foreignness as constituting properties of all epistemic categories described below and discussed so far. (35) Davidson’s ‘‘mind awareness triangle’’:
‘‘Foreign/3rd persons’ knowledge awareness’’ may also be understood as ‘‘social/objective/intersubjective knowledge’’. (36) a. b.
Epistemic adverbials/epAdvs unfold but a simple deixis – one where the Source branch of (34c) is missing. Haider ist wahrscheinlich/vermutlich betrunken gewesen. ‘Haider was probably/presumably drunk.’
10. The features are listed with redundancy, at this point of the discussion.
92
Werner Abraham
c.
(37) a. b.
Evidential adverbials/evAdvs likewise unfold a simple deixis – one where the speaker deixis is missing. Haider ist o¤ensichtlich betrunken gewesen. (with p ¼ Haider betrunken gewesen sein) ‘Haider was obviously drunk.’
c.
(38) a.
b.
Modal particles/MPs denote a di¤erent deixis in the following sense: The Speaker makes an estimate about the knowledge awareness of the Addressee while letting the Addressee know about this estimate and giving him a chance to relativize, or correct, this estimate about p. In this sense, the feature(s) under Sp-assessment are richer than those for EMV given according to the source of p, and another one according to the speaker’s assessment of p. Haider ist ja betrunken gewesen. (with p ¼ Haider betrunken gewesen sein) ‘Haider is supposed PART! to have been drunk.’
Illocutive force is speaker and information source concern
93
c.
Thus, MPs unfold a deep twofold deixis in the sense of (38c). The general insight is that MPs serve the highest theory of mind mark (speaker deixis mark ¼ speaker’s assessment of Addressee’s assessment), this to the extent that MPs – and only these as opposed to MVs and, even more so, Eadverbials – induce an intercommunication between Sp and Addr without giving full lexical expression to this exchange. No other category does this; and, what is interesting typologically, no language does this in the same direct-deictic, discourse prominent, way that does not possess the MPcategory (see Abraham 1991a,b; 1995a,b). In what follows we intend to show how grammatical persons as subjects to MV-clauses interface with the speaker/addressee deixis insights above. 9. Discourse functions/I(nformation)S(tructure), epistemicity, root phenomena in dependent clauses, and truth value assessability 9.1. The focal function of VS in French and its distribution in dependent peripheral clause/DPC vs. dependent core clause/DCC In Lahousse (2010), it is argued that, contrary to what is often assumed, embedded adverbial clauses have an information structure articulation independent from that of the main clause. More particularly, it is shown that the specific way in which information structure is expressed in adverbial clauses depends on the possibility vs. impossibility of epistemic qualification in the adverbial clause. The claim is based on new empirical evidence concerning the distribution of a clearly information structure-driven syntactic configuration: verb-subject/VS word order in French. The literature agrees more or less explicitly that the postverbal subject has to be interpreted as the new information focus (or part of the new information focus)
94
Werner Abraham
of the clause. On the other hand, the distribution of main clause phenomena shows that the internal IS articulation of adverbial clauses is not exactly the same as that of main clauses, to the extent that they do not occur in all types of adverbial clauses. For instance, it is well known that topicalization can occur in an although-clause as (39), but not in a temporal while-clause as (40): (39) The shape seemed to be looking through a book, although [what the book was] [Henry could not tell]. . . . DPC (40) *While [this book] [Mary was writing this time last year], her children were staying with her mother. . . . DCC (Haegeman 2006: 33) If IS plays exactly the same role in main and adverbial clauses, then it is hard to explain why certain clearly IS-driven phenomena do not occur in all of them alike. The purpose of this paper is to examine the internal IS status of adverbial clauses on the basis of the distribution of a clearly IS-driven syntactic configuration: verb-subject word order with its focus reading in French. Although Modern French is often described as a language with a generalized subject-verb word order, modulo certain conditions, the subject can follow the verb, as in (41). (41) Quand [verb arriva] [subject la tante], cela se fit tout naturellement. when arrived the aunt that Refl did very naturally ‘When the aunt arrived, that happened very naturally.’ . . . DCC (Sabatier, Frantext) In causal and concessive clauses VS can only appear if it is licensed by an additional factor. This is confirmed by (42), which do not contain a spatiotemporal topic or an explicit indication of the focal interpretation of the subject, and where VS is not acceptable or at least very marginal: (42) a.
Causal clause ??? Je dois partir puisque sont arrive´s les enfants. . . . DPC I have to-leave since have arrived the children ‘I have to leave since the children arrived.’
Illocutive force is speaker and information source concern
b.
95
Causal clause ??? Je dois partir parce que sont arrive´s les enfants. . . . DPC I have to-leave because have arrived the children ‘I have to leave because the children arrived.’
c. Concessive clause ??? Je dois partir bien que soient arrive´s les enfants. . . . DPC I have to-leave although have arrived the children ‘I have to leave although the children arrived.’ The examples of VS in causal and concessive clauses in (42) sharply contrast with similar examples of VS in temporal, comparative, purpose clauses and clauses beginning with sans que ‘without . . . -ing’. The examples in (43) indeed show that, in the latter types of adverbial clauses, VS is fully acceptable even in the absence of any additional factor: (43) a.
b.
Temporal clause Je dois partir quand arrivent les enfants. I have to-leave when arrived the children ‘I will leave when the children arrive.’
. . . DCC
Comparative clause Il pleurait comme le font les enfants. . . . DCC he wept in-the-same-way-as it do the children ‘He wept as children do.’
None of the adverbial clauses in these examples contain a spatio-temporal topic, and there is no anaphoric element between the verb and the postverbal subject. Moreover, the postverbal subject is not indefinite, does not contain a restrictive modification, and is not contrastively focused. In other words, VS is allowed in these adverbial clauses in the absence of a spatio-temporal topic or an indication of the focal interpretation of the subject. Table 2. Descriptive generalization on IS-criteria DPC
DCC
Causal clauses Concessive clauses
Temporal clauses (Comparative clauses Clauses with sans que ‘without . . . –ing’ Purpose clauses)
VS necessarily co-occurs with spatio-temporal topic or indication of the focal interpretation of the subject
VS is acceptable without additional factor favoring its appearance
96
Werner Abraham
10. Speaker dependency vs. subject dependency on MVs in Theory of Mind Let us check all three grammatical persons on their contribution to the epistemic readings of MVs/EMVs. One further variable is tense: Present and analytic preterit (‘‘perfect’’). ‘‘Depersonalization’’ (given # in column 4 of Table 3 below) means that the Speaker is able to truth-assess p embedded under MV even when identical with the I-subject superficially, this being the case if Sp reports about I (as in a dream) as a detached object. # signals that an E-reading is possible only under such I-ReferenceDetachment (Reporter-Origo as Reported reference). Reference is made again to Davidson’s (2004) awareness sketch in (35) above. Some plausible assumptions: Given that, in E-readings of MVs, the subject must not be in the possession of the knowledge frame presupposed for the speaker, and given furthermore that the E-reading collapses speech act time, tS, and event time, tE, the following generalizations appear to be warranted. (44) Epistemic Universal Hypothesis/EUH: If the speaker/truth value assessor and clausal subject collapse referentially, the E-reading is excluded. This is corroborated by 5, 8, 11, and 14 in Table 3 above. Yet, what is to be presupposed for (44) to hold is that a high degree of intersubjective assessment of the truth is provided. (45) Intersubjective truth assessment presupposition/ITAP: For EUH to hold, objectivity or a high degree of intersubjective truth assessibility has to exist. (45) is not warranted if speaker/truth assessor and clausal subject are referentially identical. It characterizes a discourse implausibility. (46) The distribution of person variables on epistemic readings (column 3 in Table 2): EUH, the universal hypothesis in (44), remains una¤ected by E-readings under first person reference (I-reference) in those cases where speaker and grammatical subject assessing evidence (‘‘Evid.’’) or hearsay (via third persons: ‘‘3rd (sg./)pl.’’) are not collapsed – in other words, where the truth assessor and the assessed are no longer coreferential under person detachment (Depersonalization) and, consequently, allow for a highly marked E-reading.
5
Evid./Sp
7
8
9
10
11
12
13
14
15
Sp A Subj
# 3rd pl.
Sp A Subj
Sp A Subj
Evid./3rd pl.
Sp A Subj
Sp ¼ Subj
*
Sp ¼ Subj
6
4
3
2
1
Sp A Subj
Sp A Subj
Truth assessment of E-reading lies on
Er will in Gro¨nland (gewesen) sein ‘He pretends to have been in G.’ Ich will in Gro¨nland (gewesen) sein ‘I pretend to have been in G.’ Du willst in Gro¨nland (gewesen) sein ‘You pretend to have been in G.’
Er mag in Gro¨nland (gewesen) sein ‘He may have been in G.’ Ich mag in Gro¨nland (gewesen) sein ‘I may have been in G.’ Du magst in Gro¨nland (gewesen) sein ‘You may have been in G.’
Er soll in Gro¨nland (gewesen) sein ‘He is supposed to (have) be in G.’ Ich soll in Gro¨nland (gewesen) sein ‘I am supposed to (have) be in G.’ Du sollst in Gro¨nland (gewesen) sein ‘You are supposed to (have) be in G.’
Er muss in Gro¨nland (gewesen) sein ‘He must (have) be(en) in G.’ Ich muss in Gro¨nland (gewesen) sein ‘I must (have) be(en) in G.’ Du musst in Gro¨nland (gewesen) sein ‘You must (have) be(en) in G.’
Er ist in Gro¨nland (gewesen) ‘He was in Greenland.’ Ich bin in Gro¨nland (gewesen) ‘I was in Greenland.’ Du bist in Gro¨nland (gewesen) ‘You were in Greenland.’
Person variable: 3rd, 1st, and 2nd person
þ
*
*
*
*
*
þ þ
*
*
*
þ
þ
þ
þ
þ
þ
þ
þ *
#
þ
þ
#
þ
þ
þ
þ
Root modal: Present
*
*
þ þ
*
*
Root modal: Perfect
#
þ
Ereading
Table 3. Root and epistemic MV-readings dependent upon person and tense variables
Depersonalization?
Depersonalization
Depersonalization
Depersonalization
E-reading possible (given #) only under
Illocutive force is speaker and information source concern
97
98
Werner Abraham
(47) The distribution of tense variables on root modality readings (columns 4 þ 5 in Table 2): Following observations by Abraham (1990a,b; 1991a,b; 2001a,b; 2005), root modals behave like perfectives. (Prospective) perfectives project a futurate tE – i.e., what holds time-referentially is tS/tR < tE. Any tE < tS/tR is incompatible – this, in turn, yields that its ungrammaticality is computable, unlike the present allowing for tS/tR < tE. (48) Safeguarding the truth value on E-readings (column 1 in Table 2): Safeguarding truth, which is crucial for any E-reading, lies preferably at the hands of 3rd sg./pl., and it is least warranted with the 1st sg. This is at the bottom of the unacceptable E-readings under 1st person, ‘‘I ’’, with each EMV. Where an E-reading is coe¨rced anyway, the respective I-reference shifts to that of the assessed I (cf. the #-cases among the E-readings). Furthermore, we discern scalar degrees of truth safeguarding: Highest on this scale is the speaker evidence (Sp ¼ truth assessor) encoded by MUSS ‘must’ – i.e., speaker evidence is the most reliable assessment comparable to that encoded by a declarative as in Table 3 –, followed by modification on the part of the social environment (‘‘3rd sg./pl.’’) as in the case of SOLL ‘shall’; the scale of modal force is terminated by MAG ‘may’ which leaves open alternatives.
11. Derivation: From surface to covert scope position These are the empirical prerequisites for a formal derivation which yields the basis for semantic interpretability. – MPs occur only in the middle field; they have to raise to the head of ForceP to enforce their illocutive power (suggested first by Abraham 1995a,b, 2010; more recently, Zimmermann 2004 / 2008) – Since the invariable linear order of MPs in combination falls into three classes (Thurmair 1989) motivated by their pre-particle categorial status (Abraham 1995a,b), there must be three di¤erent landing sites in ForceP, i.e. MP1 goal-oriented necessity (or: obligation) Some of these accounts refer explicitly to the semantic analysis proposed by Kratzer (1981, 1991), as for instance Butler (2003) does. One major shortcoming of Kratzer’s work is that it does not account for the semantic restrictions and idiosyncrasies of each single modal. Moreover, she does not explain why most of the modals can be used as circumstantial modifiers and as root modifiers. In order to o¤er a solution to these riddles, cartographic approaches were developed. But as it turned out, they are far to restrictive, they exclude a lot of combinations that occur quite regularly
Exploring the Theory of Mind interface
135
in natural language production. What is needed, then, is an account that is flexible enough to capture all the idiosyncrasies of each single modal verb.
5. Westmoreland (1998): modal operators vs. evidential markers Westmoreland (1998) argues that only root modals can be treated as modal necessity and possibility operators (k, y). Epistemic modals, however, are evidential markers that label a proposition as an assumption made by the speaker. If they were indeed modal operators, it should be possible to define them in terms of each other as in (43). As Westmoreland (1998, 8) illustrates, this does not seem to be the case, as the examples in (42) are not equivalent. (41) I just put the cat outside. . . (42) a. . . . so she can’t be under the table b. A . . . so she must [not be under the table] (43) Byp ¼ kBp (44) neg (can epistemic (p)) A mustepistemic neg(p)) It is a matter of debate to what extent epistemic modals can be negated: Butler (2003, 984) and Moscati (2006, 31) claim that this is possible, ¨ hlschla¨ger (1989, 207), Askedal (1997, 63), Diewald (1999, 84) and Drubig O (2001, 5) argue for the opposite. The crucial point is that it is not clear whether can’t in the example above is based on a practical possibility reading or an epistemic one. Since the criterium for epistemicity defended by Westmoreland (1998) and Ziegeler (2006) seems not to hold for these use of can, it is rather doubtful whether it really can be considered epistemic. But then, Westmoreland’s (1998) objection can be circumvented.
6. A feature based account The preceding sections have reviewed di¤erent traditions of accounting for ambiguities of epistemic modifiers. One of the rather surprising results was that Cognitivst, Generativist and Formal Semantic accounts resembled each other to a much greater extent than was probably expected. One shortcoming of the former two accounts is that they were both too restrictive, in particular insofar as they could not cope with lexical idiosyncrasies
136
Jakob Mache´
such as particular ambiguities of ko¨nnen and wollen. By contrast, the approach developed by Kratzer (1978, 1981) has di‰culties to explain why not every modal is compatible with any type of conversational background (ordering source). Without any further stipulations, her proposal would predict that each modal should tolerate any kind of conversational background, i.e. any modal reading. Therefore, she needs to assume lexical restrictions operating on each specific modal. The account presented here tries to unify the diachronic plausibility of Sweetser (1990) and the descriptive power of Kratzer (1978, 1981). In order to capture idiosyncratic behaviour of lexical items, a lexicalist approach such as HPSG appears to be the most appropriate option. Building on the semantic analysis by Kratzer (1981, 1991), this section presents a sketch of an analysis on how lexical restrictions on the specific modal verbs can be integrated into a framework based upon possible world semantics. Note that according to Kratzer (1981, 1991) a modal operator consists of three elements: (i) a modal force: k (o), y (p), (ii) a modal base: circumstantial (objective facts) and epistemic (knowledge of the speaker) and (iii) an ordering source: bouletic, deontic, etc. . . . Following Koenig and Davis (2001, 19), Mu¨ller (2007, 380), Melnik (2010), a modal operator is integrated into the argument structure of its host, and hence, represented as an unary relation in the content attribute. As already proposed by Koenig and Davis (2001), we will assume that all of the subparts of the modal operator are attributes on their own. Being faithful to Kratzer (1981, 1991), the content-value of a modal could look like what is exemplified in (45): (45) ko¨nnen –ability control 2 modal source 6 soa 6 6 modal force 6 6 modal base 6 4 ordering source modal relation
verb (preliminary version) 3 l 1 7 ½ind event 7 7 existential 7 circumstantial 7 7 5 ability
Unlike in the cartographic approaches, modal verbs will be lexical categories of the type of a verb. This ensures unrestricted mutual selection of modal verbs accounting for the examples in (39) and (40). Whenever a root modal fails to embed another one, this will be for pragmatic reasons. As is (proto)typical of a verb, the content value of ko¨nnen contains the verbal arguments with semantic role specification, in this case a subject
Exploring the Theory of Mind interface
137
argument with the role type ‘modal source’, and an object argument realised by the type of state-of-a¤airs. Crucially, the referent of the modal source will be linked here to the subject on the subcat-list. The attribute modal base, however, is dispensable. Its main purpose is to distinguish between circumstantial and epistemic modality. First, this is once more an approach that fails to capture quotative modals – Kratzer (1991, 680) suggests that the modal base remains empty for quotative sollen. Second, this distinction can entirely be derived from the modal’s argument structure: While circumstantial modals are event modifiers, epistemic (and quotative) modals are clausal (or speech act) modifiers. Therefore, the approach outlined here will no longer take them into consideration. As with any other verb in the lexicon, modal verbs also exhibit idiosyncratic restrictions for argument structure, i.e. the number of arguments, their specific case, semantic roles and syntactic categories. What ordering source is possible for each modal verb is partially determined by its argument structure. An ordering source defined by the abilities of the subject requires a verb with an argument structure specified for: [þsubject argument], [– speech act modification], the assignment of [þaccusative] to its complement is optional. A bouletic ordering source typically requires the specification [þsubject argument], [– speech act modification], the assignment of [þaccusative] to its complement is also optional here, in addition the direct object can be represented as dass-clause.5 As it turns out, a small number of argument structure features are su‰cient to determine in large parts of the semantic properties of all possible modal interpretation. The most important features a [esubject argument], [espeech act modification] and [þaccusative]. – Dynamic modality (volition, ability): ! control verbs, event modifiers – Circumstantial modality (practical, deontic): ! raising verbs, event modifiers – Quotative modality: ! control (raising) verbs, speech act modifiers – Epistemic modality: ! raising verbs, speech act modifiers
5. Kratzer (1991) also attributes a bouletic ordering source to the raising verb sollen. Indeed, already Bech (1949) considered this verb as a verb of external volition. Since Kratzer’s account remains very sketchy, it does not necessarily contradict the assumptions made above. As it seems, the bouletic ordering source is not restricted to verbs that encode the source of volition as subject (subject control verbs), but is also compatible with verbs that involve a covert argument that corresponds to the source of volition.
138
Jakob Mache´
Figure 6. Type hierarchy verbal complements
More fine-grained distinctions (e.g. possibility vs. permission) can be obtained by employing Kratzer’s (1981) conversational backgrounds, in particular ordering sources. In order to account for the lexical ambiguity of modal modifiers, it is useful to draw on the concept of multiple inheritance. Assuming a basic entry for each modal, the respective interpretation can be derived from the interaction of two type hierarchies: One hierarchy provides the categories of the arguments (see Figure (6)), the other one specifies the type of modification, in particular the index specification of the state of a¤airs value soa (see figure (7)).
Figure 7. Type hierarchy modifier
By means of these two type hierarchies, all possible modal readings can be derived. Note that some of the combinations will fail for technical reasons. For example, a verb with a NPaccusative complement cannot inherit from speech act-modifier, since the latter requires an appropriate index value. As a consequence, these two chunks of structure will not unify. But since patterns like this do not occur in natural language, they are correctly predicted to be ruled out. Assuming a basic lexicon entry modal predicate that provides the attributes modal force and ordering source, the following lexicon entries may be derived:
Exploring the Theory of Mind interface
139
(46) ko¨nnen –ability transitive ( ! NPaccusative þ object-modifier) 2 3 modal source l 1 6 soa ½ind referential 7 6 7 4 modal source existential 5 modal relation (47) ko¨nnen –ability control ( ! control infinitive þ event modifier) 3 2 modal source l 1 6 soa ½ind event 7 7 6 4 modal force existential 5 modal relation (48) ko¨nnen –possibility raising ( ! raising infinitive þ event modifier) 3 2 modal source underspecified 7 6 soa ½ind event 7 6 5 4 modal force existential modal relation (49) ko¨nnen –epistemic possibility raising infinitive þ speech act modifier) 2 3 modal source underspecified 6 soa ½ind speech act 7 6 7 4 modal force 5 existential modal relation The analysis of wollen will roughly be the same, except that there is an additional pattern that remains to be accounted for: the quotative use. The derivation of this specific interpretation can easily be obtained. Quotative wollen inherits control infinitive from the hierarchy of verb types, and speech act modifier from the hierarchy of modifiers. Finally, we are able to turn to the diachronic dimension of modal ambiguity. As Lightfoot (1979, 1998) argues, language acquisition is the main locus of diachronic development. According to Green (2003), language acquisition basically has to be considered as the acquisition of new distinctions. Assuming that a learner just became aware that his target language has elements that modify other elements, he will first use those modifiers that are most easily processed, in correspondence with the observations made by Pienemann (1998, 42). After a while, he will find out that there are di¤erent ways to employ the modifier in natural language, thus assuming a new distinction, as for instance object modifier versus event modifier. Only after a further
140
Jakob Mache´
period has elapsed will he find out that there are more subtle options in the use of a modifier, as in speech act modification. This happens quite late in the development of a child, as the investigation by Doitchinov (2001, 2007) has shown. This is how type hierarchies become more and more distinctive. As soon as these hierarchies are acquired, they are available for inheritance. Items that are more abstract, such as speech act modifiers, will be acquired later. But what causes event modifiers to transform into speech act modifiers? At least two prerequisites have to be met. First of all, the type speech act modifier must be available in the learner’s hierarchy. Second, there need to be critical contexts where the communicative e¤ect of the event modifier interpretation and the speech modifier interpretation would be almost the same, as shown by Diewald (1999, 368). Then it might happen that under certain circumstances the event modifier interpretation is highly marked and that it can only be maintained when it is supported by pragmatic repair mechanisms. Being event modifiers, circumstantial modal verbs prototypically do not select for states. Nevertheless the subcategorisation of stative verbs is possible if they are reinterpretated as events, by means of a repair mechanism as developed by Maienborn (2003, 216). As soon as this mechanism of reinterpretation becomes too opaque for the language learners, they will opt for any alternative that is easier to process, see Mache´ (2008, 412) for more details. If there is a speech act interpretation and the event interpretation are synonymous in the relevant contexts, the language learner will reanalyse the circumstantial modal as an epistemic one. This corresponds exactly to Pienemann’s (1998, 5) findings, who showed that a learner only produces structural options that he is able to process. Since more abstract lexical items typically involve less argument structure, which also causes them to be less easily processable, historical developments can be considered as attrition of the integrity of the sign, as proposed by Lehmann (1995). The analysis of German modal verbs provided above can by and large be adapted for English. The situation is quite similar, except that English modals behave a little less idiosyncratically than their German counterparts. Epistemic adverbial clauses as discussed above can be analysed in more or less the same manner as modal verbs. They are even easier to handle, since they only involve two di¤erent patterns: event modification versus speech act modification. Since the proposal outlined here remains sketchy, it remains to be worked out in more detail.
Exploring the Theory of Mind interface
141
7. Implications for the Theory of Mind Interface The first central question that was raised at the beginning was to determine the nature of epistemic modification, and to find an account that neatly captures its essential properties. Comparing epistemic interpretations with their non-epistemic counterparts, it has been shown that the former can be described as speech act (or clausal) modifiers, while the latter are event modifiers. The account developed here is much more flexible than the cognitive perspective taken by Sweetser (1990) and Sweetser and Dancygier (2005), or cartographic approaches in the spirit of Cinque (1999), which are popular in the generative tradition. Sweetser (1990) assumes that each modal interpretation is processed in a di¤erent mental domain. Her approach therefore crucially hinges on the existence of these mental domains. Cartographic theories, on the other hand, rely on the existence of a considerable number of functional categories. Both concepts, functional categories and mental domains, are very vague notions that are di‰cult to operationalize, and moreover lack thourough empirical support. Therefore, we should regard them with suspicion. As it has been shown in this paper, very detailed and powerfull descriptions can be obtained by employing the formal semantic account provided by Kratzer (1981). Sweetser’s (1990) objections against frameworks that are based upon modal logic are not justified, since they rely on a misunderstanding of Kratzer’s framework. Moreover, it turned out that formal logic is an important tool for the description of natural languages. Whitout acknowledging this, Sweetser (1990) defines her Modal Force Dynamics in terms of modal logic. Still, one should avoid the overuse of formalisms – as has correctly been pointed out by Sperber and Wilson (1995). In a similar vein, Kratzer (1978) has shown that the main labour is done by pragmatics and the context of the utterance. However, in some aspects, Kratzer’s analysis turned out to be too unrestrictive. But, as pointed out here in much detail, restrictions on the ordering source can be derived from the argument structure of the relevant modifiers. In the end, it turned out that all of these di¤erent approaches share a lot of insights, yet the account presented here appears to be the most flexible one. What inferences can we draw from epistemic modifiers with respect to Theory of Mind? Corresponding to the three di¤erent approaches discussed here, there are three di¤erent answers. According to Cognitivist approaches, ToM is closely connected to two separate mental spaces: the epistemic domain and the speech act domain. From the perspective of a cartographist, ToM is a phenomenon that is reflected by a set of func-
142
Jakob Mache´
tional categories that originate in a high position in the clausal hierarchy. Yet, neither of the newly introduced notions is necessary for the stipulation of either an unknown number of mental spaces or an unknown number of functional categories. All phenomena related to epistemic modality can be explained in terms of speech act modification. By using speech act modifiers, the speaker is able to signal that s/he is aware that the modified speech act does not hold without restriction, that the assertive validity is relativised. Accordingly, grammar already partially encodes information that is relevant to ToM. Jakob Mache´ Institut fu¨r Deutsche und Niederla¨ndische Philologie Freie Universita¨t Berlin Habelschwerdter Allee 45 14 195 Berlin
References Abraham, Werner 2001 Modals: towards explaining the ‘epistemic non-finiteness gap’. In Reimar Mu¨ller and Marga Reis (eds.), Modalita¨t und Modalverben im Deutschen, pages 7–36, Hamburg: Buske. Abraham, Werner 2003 Canonic and non-canonic deliberations about epistemic modality: its emergence out of where? On an extended notion of grammatic(al)ization. In Germania et alia Digital volume dedicated to Hans den Besten, University of Amsterdam. Abraham, Werner 2008 Wie ist die historisch spezifische Ausgliederung der Modalita¨ten bei den Modalverben denkbar? Linguistische Berichte 214, 185– 213. Abraham, Werner and Leiss, Elisabeth 2008 Modality-Aspect interfaces. Implications and Typological Solutions, volume 79 of Typological Studies in Languages. Amsterdam: Benjamins. Abraham, Werner and Leiss, Elisabeth 2009 Modalita¨t. Epistemik und Evidentialita¨t bei Modalverb, Adverb, Modalpartikel und Modus, volume 77 of Studien zur deutschen Grammatik. Tu¨bingen: Stau¤enburg. Askedal, John Ole 1997 ‘Brauchen’ mit Infinitiv. Aspekte der Auxiliarisierung. Jahrbuch der ungarischen Germanistik pages 53–68.
Exploring the Theory of Mind interface Axel, Katrin 2001
Bech, Gunnar 1949
143
Althochdeutsche Modalverben als Anhebungsverben. In Marga Reis and Reimar Mu¨ller (eds.), Modalita¨t und Modalverben im Deutschen, pages 37–60, Hamburg: Buske.
Das Semantische System der Modalverben. Travaux du Cercle Linguistique de Copenhague VI, 3–46. Brennan, Virginia 1993 Root and Epistemic Modal Auxiliary Verbs in English. Ph.D. thesis, University of Massachusetts, Amherst. Brennan, Virginia 1997 Modalities. Ms. Vanderbilt University. Butler, Jonny 2003 A minimalist treatment of modality. Lingua 113, 967–996. Chomsky, Noam 2001 Derivation by phase. In Michael Kenstowicz (ed.), A life in language, pages 1–52, Cambridge, MA: MIT Press. Cinque, Guglielmo 1999 Adverbs and functional heads: a cross-linguistic perspective. Oxford studies in comparative syntax, New York: Oxford University Press. Diewald, Gabriele 1999 Die Modalverben im Deutschen: Grammatikalisierung und Polyfunktionalita¨t. Reihe Germanistische Linguistik, No. 208, Tu¨bingen: Niemeyer. Doitchinov, Serge 2001 ‘Es kann sein, daß der Junge ins Hausgegangen ist’ – Zum Erstspracherwerb von ‘ko¨nnen’ in epistemischer Lesart. In Reimar Mu¨ller and Marga Reis (eds.), Modalita¨t und Modalverben im Deutschen, pages 111–134, Hamburg: Buske. Doitchinov, Serge 2007 Modalverben in der Kindersprache. Kognitive und linguisitische Voraussetzungen fu¨r der Erwerb von epistemischen ‘ko¨nnen’. Studia grammatica, No. 67, Berlin: Akademie. Drubig, Hans Bernhard 2001 On the syntactic form of epistemic modality. Enc¸, Mu¨rvet 1996 Tense and Modality. In Shalom Lappin (ed.), The Handbook of Contemporary Semantic Theory, pages 345–358, Oxford: Blackwell. Erb, Marie Christine 2001 Finite Auxiliaries in German. Ph.D. thesis, Katholieke Universiteit Brabant.
144
Jakob Mache´
Favell, John H. 2004 Theory-of-Mind development: retrospect and prospect. MerillPalmer Quaterly 50(3), 274–290. Fritz, Gerd 1997 Historische Semantik der Modalverben. In Gerd Fritz and Thomas Gloning (eds.), Untersuchungen zur semantischen Entwicklungsgeschichte der Modalverben im Deutschen, Reihe Germanistische Linguistik, No. 187, pages 1–157, Tu¨bingen: Niemeyer. Green, Georgia 2003 Modelling grammar growth: Universal grammar without innate principles or parameters. Hofmann, Thomas Ronald 1976 Past Tense Replacement and the Modal System. In James McCawley (ed.), Notes from the Linguistic Underground, Syntax and semantics, No. 7, pages 86–99, New York: Academic Press. Jespersen, Otto 1917 Negation in English and other languages, volume 1 of Historiskfilologiske meddelser. København: Hoest. Koenig, Jean-Pierre and Davis, Anthony R. 2001 Sublexical modality and the structure of lexical semantic representations. Linguistics and Philosophy 24(1), 1–54. Kratzer, Angelika 1978 Semantik der Rede. Kontexttheorie – Modalwo¨rter – Konditionalsa¨tze. Ko¨nigstein: Scriptor. Kratzer, Angelika 1981 The Notional Category of Modality. In Hans Ju¨rgen Eikmeyer and Hannes Rieser (eds.), Words, Worlds and Contexts. New approaches in World Semantics, Berlin: Mouton de Gruyter. Kratzer, Angelika 1991 Modality. In Arnim von Stechow and Dieter Wunderlich (eds.), Semantik. Ein internationales Handbuch der zeitgeno¨ssischen Forschung, 639–650, Berlin: Mouton de Gruyter. Kratzer, Angelika 1995 Stage-level and individual-level predicates. In Gregory N. Carlson and F. J. Pelletier (eds.), The Generic book, pages 125–175, Chicago: University of Chicago Press. Lehmann, Christian 1995 Thoughts on Grammaticalization. Revised and expanded version, volume 1 of LINCOM Studies in Theoretical Linguistics. Mu¨nchen: Lincom Europa. Lightfoot, David 1979 Principles of diachronic syntax, volume 23 of Cambridge studies in linguistics. Cambridge: Cambridge University Press.
Exploring the Theory of Mind interface
145
Lightfoot, David 1998 Development of Language: Acquisition, Change and Evolution. Maryland lectures in language and cognition, No. 1, Malden: Blackwell. Lyons, John 1977 Semantics, volume 2. Cambridge: Cambridge University Press. Mache´, Jakob 2008 The autopsy of a modal – insights from the historical development of German. In Werner Abraham and Elisabeth Leiss (eds.), Modality-Aspect Interfaces. Implications and Typological Solutions, volume 79 of Typological Studies in Language, Amsterdam: Benjamins. Mache´, Jakob 2009 Das Wesen epistemischer Modalita¨t. In Werner Abraham and Elisabeth Leiss (eds.), Modalita¨t, Epistemik und Evidentialita¨t bei Modalverb, Adverb, Modalpartikel und Modus, volume 77 of Studien zur deutschen Grammatik, pages 25–55, Tu¨bingen: Stau¤enburg. Maienborn, Claudia 2003 Die logische Form von Kopulasa¨tzen. studia gramatica, No. 56, Berlin: Akademie Verlag. Maienborn, Claudia 2004 A pragmatic explanation of the stage level / individual level contrast in combination with locatives. In Brian Agbayabi, Vida Samiian and Benjamnin Tucker (eds.), Proceedings of the WECOL, volume 15, pages 158–170, Fresno: CSU. Melnik, Nurit 2010 Modal Predicates in Modern Hebrew. In Stefan Mu¨ller (ed.), The Proceedings of the 17th International Conference on HPSG, Stanford: CSLI Publications. Moscati, Vincenzo 2006 The scope of negation. Ph.D. thesis, Universita` di Siena. Mu¨ller, Stefan 2007 Head-Driven Phrase Structure Grammar. Eine Einfu¨hrung, volume 17 of Stau¤enburg Einfu¨hrungen. Tu¨bingen: Stau¤enburg. ¨ hlschla¨ger, Gu¨nther O 1989 Zur Syntax und Semantik der Modalverben, volume 144 of Linguistische Arbeiten. Tu¨bingen: Niemeyer. Pienemann, Manfred 1998 Language Processing and Second Language Development: Processability Theory, volume 15 of Studies in bilingualism. Amterdam: Benjamins. Pienemann, Manfred 2005 An Introduction to Processability Theory. In Manfred Pienemann (ed.), Cross-linguistic aspects of processability theory, volume 30 of Studies in bilingualism, Amsterdam: Benjamins.
146
Jakob Mache´
Reis, Marga 2001
Rizzi, Luigi 1997
Bilden Modalverben im Deutschen eine syntaktische Klasse? In Marga Reis and Reimar Mu¨ller (eds.), Modalita¨t und Modalverben im Deutschen, pages 239–262, Hamburg: Buske. The fine structure of the left periphery. In Liliane Haegeman (ed.), Elements of Grammar, pages 281–337, Amsterdam: Kluwer Academic Publishers.
Roberts, Ian G. 2003 Synactic change: a minimalist approach to grammaticalization, volume 100 of Cambridge studies in linguistics. Cambridge: University Press. Sperber, Dan and Wilson, Deirdre 1995 Relevance: communication and cognition. Cambridge, MA: Blackwell, second edition. Sweetser, Eve 1990 From etymology to pragmatics. Metaphorical and cultural aspects of semantic structure, volume 54 of Cambridge studies in linguistics. Cambridge: University Press. Sweetser, Eve and Dancygier, Barbara 2005 Mental spaces in grammar: conditional constructions. Cambridge: Cambridge University Press. Wegener, Heide 1993 weil -das hat schon seinen Grund. Zur Verbstellung in Kausalsa¨tzen mit ‘weil’ im Gegenwa¨rtigen Deutsch. Deutsche Sprache 21, 289–305. Welke, Klaus 1965 Untersuchungen zum System der Modalverben in der deutschen Sprache der Gegenwart, volume 10 of Schriften zur Phonetik, Sprachwissenschaft und Kommunikationsforschung. Berlin: Akademie Verlag. Westmoreland, Robert R. 1998 Information and intonation in natural language modality. Ph.D. thesis, University of Michigan, Michigan. White, Lydia 2003 Second Language Acquisition and Universal Grammar. Cambridge Textbooks in Linguistics, Cambridge University Press. Wurmbrand, Susi 2001 Infinitives. Restructuring and Clause Structure, volume 55 of Studies in Generative Grammar. Berlin: Mouton de Gruyter. Ziegeler, Debra 2006 Omnitemporal ‘will’. Language Sciences 28, 76–119.
The distribution of knowledge in (un)acceptable questions* Sonja Mu¨ller Universita¨t Bielefeld Abstract This paper pursues the idea that the distribution of knowledge between discourse participants has an impact on the overall acceptability or the availability of certain readings of questions. In this study, di¤erent knowledge states get linguistically encoded by (non-)(anti-)factive matrix verbs in extractions from that-complements, by (di¤erent types of ) modal verbs in complex and simpler questions and in the occurrence of modal particles in that-complements embedded under verbs of saying in extraction constructions. A more general implication of the analysis is the claim that violating pragmatic rules for illocutionary acts by linguistic encoding can a¤ect wellformedness and/or acceptable interpretations of linguistic expressions. 1. Introduction Classical speech act theory has attempted to formulate the conditions that are necessary and su‰cient for a particular illocutionary act to be successfully and sincerely performed. Searle (1969: 66) e.g. formulates (1) in order to account for the structure of the illocutionary act question. (1) Preparatory rule: 1. S does not know ‘the answer’, i.e. [. . .] does not know the information needed to complete the proposition truly [. . .]. 2. It is not obvious to both S and H that H will provide the information at that time without being asked. Sincerity rule: S wants this information. Essential rule: Counts as an attempt to elicit this information from H. * I am grateful to Verena Vogt for proof-reading this article. I would also like to thank the anonymous reviewer as well as Elisabeth Leiss and Werner Abraham for their suggestions, comments, and critical remarks.
148
Sonja Mu¨ller
Assuming the di¤erentiation of such types of rules when it comes to carrying out certain illocutionary acts during communication, Searle (1969: 63) points at the analogy to rules formulated for (moves in) games in order to control and guarantee their correct course: If we ask ourselves under what conditions a player could be said to move a knight correctly, we would find preparatory conditions such as that it must be his turn to move, as well as the essential condition stating the actual positions the knight can move to. There are even sincerity conditions for competitive games, such as that one does not cheat or attempt to ‘throw’ the game.
Transferring this analogy back to (1), uttering a question during communication thus leads to assuming that the adequate application of this move depends on the situations set by the di¤erent rules in (1). As becomes obvious from looking at the formulations in (1) in detail, the distribution of knowledge as far as the discourse participants’ knowledge states are concerned plays a crucial role in determining the language game’s rules on the adequate usage of a sentence carrying out this illocutionary act: It is assumed that the speaker does not know the answer, presumes that the addressee knows it and poses the question in order to elicit this information from the addressee. Especially the assumption on the speaker’s knowledge state as formulated in the first part of the preparatory rule has also found its way into other formulations of the pragmatic conditions having to be met when a question is asked in discourse (cf. e.g. (2) taken from Bu¨ring 2003: 5, cf. also e.g. Reis 1990: 51, Caponigro and Sprouse 2007: 9). (2) Informativity: [. . .], don’t ask for known things! By focussing on this (pre)condition for a questioner’s adequate move in discourse when taking part in communication by uttering an ordinary (Caponigro and Sprouse 2007: 9) (that is information-seeking) question (cf. also Searle’s 1969: 66 comment, Reis 1990: 51), this paper will look at the interaction of question formation and the occurrence of certain linguistic material in the questions, namely (non-)(anti-)factive verbs (cf. Section 3), modal verbs (cf. Section 4) as well as modal particles (cf. Section 5). Due to the meaning components expressed by this linguistic material, a speaker who chooses these linguistic expressions to appear in his utterance makes a statement about the assignment of knowledge concerning the persons participating in the current discourse which might (not) clash
The distribution of knowledge in (un)acceptable questions
149
with the assumptions about the questioner’s knowledge state as formulated e.g. in a condition such as 1. in (1) and as being expressed by the questioner posing a question and, thereby, carrying out the question act. Before studying this interaction, Section 2 will, however, first look at the role of questions in discourse in more detail than sketched here by referring to formal means of modelling discourse situations. The overall idea pursued in this paper is that violating the condition on the questioner’s knowledge state has a negative impact on the acceptability of questions, the linguistic material serving the purpose of manipulating the circumstances regarding the distribution of knowledge between speaker, addressee and possibly third instances. On the one hand, this assumption will be directly mirrored in a conceivable decrease of a question’s acceptability, on the other hand, it will be shown that this assumption can explain why only certain readings are available for certain questions. While pursuing the aim to find evidence for the idea that a condition on illocutionary acts such as 1. in (1) does have an impact on the well-formedness of sentences intending to carry out this act, the study will indirectly also check assumptions made on the interpretation of the linguistic material made use of in order to test for this hypothesis.
2. The role of questions in discourse Linguistic utterances obviously do not occur in isolation, but within larger contexts, that is certain persons (speaker, addressee) who are influenced by their views about the world in general and the topic of the conversation in particular participate. They occur at a place, at a certain time and are preceded by other utterances. Linguistic phenomena such as e.g. reference of deictic pronouns or coreference relations provide ample evidence for this assumption. Since Stalnaker (1978) who has studied the e¤ects that assertions have on a context, diverse assumptions about the formal representation of an utterance context have been made (cf. e.g. Giannakidou 1998, Roberts 1996, Bu¨ring 2003, Caponigro and Sprouse 2007, Gunlogson 2003, Bartels 1999, Portner 2005, Farkas and Bruce 2009). The central concept in the characterisation of a context in Stalnaker (1978) as well as in the other approaches mentioned here is the Common Ground (CG) which can be modelled as a set of propositions which represent the assumptions the discourse participants knowingly mutually agree on. On the one hand,
150
Sonja Mu¨ller
mutual belief can be achieved by tacit assumptions, on the other hand, speakers may come to agree on certain pieces of information in the course of a conversation. Assuming with Possible World Semantics that a proposition is associated with the set of worlds in which it is true, the context also contains the Context Set (CS) which represents the set of worlds in which all CG-propositions are true. Stalnaker (1978: 322) writes about the purpose of communication: ‘‘To engage in a conversation is, essentially, to distinguish among alternative possible ways that things may be. The purpose of expressing propositions is to make such distinctions.’’ If an assertion is uttered (cf. e.g. (3)), this aim is pursued directly. Unless the addressee rejects the proposition expressed, it is added to the CG and the CS gets reduced to the worlds in which the CG-propositions are valid (cf. (4)). (3) Anna invites Stephan for co¤ee. (¼p1) (4) input context CG ¼ f g CS ¼ W
output context 1 CG0 ¼ fp1 g CS 0 ¼ CS \ fw 2 W jp1 ðwÞ ¼ 1g
If a constituent question is uttered, it opens up a restricted number of alternatives which the answer (in most cases an assertion) ideally reduces to one possibility. Relying on Partition Semantics for the semantic assumptions on questions (cf. Groenendijk and Stokhof 1984, 1997, Higginbotham and May 1981, Higginbotham 1991, 1996), a wh-question such as (5) induces a space of answers as in (6). Each cell corresponds to a proposition (or, respectively, a set of worlds) which represents a possible complete (within this type of theory strongly exhaustive) answer.2
1. The following abbreviations will be used throughout the article: CG ¼ common ground, CS ¼ context set, p/P ¼ proposition, E ¼ existential implicature of whquestions, Q ¼ presupposition, Di,w ¼ the set of propositions constituting an individual’s beliefs in a world, D i,w ¼ the set of worlds containing the corresponding belief worlds, Ii,w ¼ the set of propositions ignored by an individual in a world, I i,w ¼ the set of worlds containing the corresponding ‘ignorance’-worlds. 2. On alternative approaches to the semantics of questions cf. e.g. Karttunen (1977), Heim (1994), Dayal (1996), Krifka (2001), Reich (2003).
The distribution of knowledge in (un)acceptable questions
151
(5) Who does Anna invite for co¤ee?, D ¼ {Julia, Stephan} (6)
Anna invites Julia as well as Stephan for co¤ee.
(¼ p1)
Anna invites Julia, but does not invite Stephan for co¤ee.
(¼ p2)
Anna invites Stephan, but does not invite Julia for co¤ee.
(¼ p3)
Anna invites neither Julia nor Stephan for co¤ee.
(¼ p4)
As (7) illustrates, the CG gets expanded with one possible answer, the CS can be changed in as many ways as the CG. After an answer has been given, an assertive context update follows. (7) input context CG ¼ f g
CS ¼ W
output context CG0 ¼ CG [ fp1 ga CG [ fp2 ga CG [ fp3 ga CG [ fp4 g CS 0 ¼ CS \ fw 2 W jp1 ðwÞ ¼ 1ga CS \ fw 2 W jp2 ðwÞ ¼ 1ga CS \ fw 2 W jp3 ðwÞ ¼ 1ga CS \ fw 2 W jp4 ðwÞ ¼ 1g
A demand that an adequate proposition has to meet in a context is its informativity, i.e. by adding a proposition the information state of discourse is supposed to change (cf. (8)). (8) CS \ fw 2 W jpðwÞ ¼ 1g 6¼ CS For an informative assertion, (8) means that it should not be intended to add a proposition that is already in the CG. A wh-question should not open up any of the possibilities which have been removed from the CS earlier (cf. Bu¨ring’s condition in (2)). Several attempts have been made to expand Stalnaker’s model with components or functionalities in order to be able to model communication more accurately (cf. e.g. Portner 2005 [imperatives], Farkas and Bruce 2009 [rejections]). One of those aspects that are missing in the original architecture that Stalnaker proposes concerns an individualisation of the contextual parameters in the sense that there should not only exist a set of propositions corresponding to the knowingly joint propositions between
152
Sonja Mu¨ller
speaker and addressee, but also of their individual belief systems. Bartels (1999) e.g. (cf. also Caponigro and Sprouse 2007, Gunlogson 2003) makes a distinction between three systems for each discourse participant: Private beliefs, public commitments, joint public commitments (|| CG) (cf. (9)). With reference to how the cells are filled in di¤erent states of an ongoing discourse, Bartels derives the changes induced by di¤erent moves in a discourse situation. When updating a context with an assertion (cf. (9)), the speaker has a certain belief P in mind that he wants to convey. The addressee can have the same belief, a contradictory belief or no belief at all with respect to the subject. Neither the speaker nor the addressee have publicly committed themselves to the issue. After asserting P, the speaker has publicly committed himself to P. Due to cooperativity, the addressee makes P part of his private beliefs. If he accepts P, P becomes part of his public commitments. As P is by now public commitment of the speaker as well as of the addressee, P can also be considered joint public commitment. (9) (cf. Bartels 1999: 102) illustrates how the speaker’s private belief P finds its way into the other systems and finally – by arriving in the CG – fulfills the true purpose of an assertion. (9) private beliefs
public commitments
joint public commitments
speaker
addressee
speaker
addressee
initial context
P
P, not P, –
–
–
–
after commitmentsS (P)
P
P
P
–
–
after commitmentsA (P)
P
P
P
P
P
In the case of a wh-question such as (5), the cells are filled di¤erently (cf. (10)). Bartels assumes that such questions have a sentential presupposition that there is at least one individual that answers the question positively. She considers this assumption, which sometimes also features as an existential implicature in the literature (represented as E in (10)), as a public commitment by the discourse participants and, therefore, as joint public commitment.3 Furthermore, the speaker assumes that the belief he asks 3. Note that this assumption is not compatible with the assumptions made by the semantics of questions adopted here. However, nothing in the analysis presented at a later stage hinges on this aspect.
The distribution of knowledge in (un)acceptable questions
153
for is present in the addressee’s belief system. If the addressee gives an answer, he publicly commits himself to P and the speaker adds P to his private beliefs. If the speaker accepts the addressee’s answer, P also becomes part of his public commitments. (10) (cf. Bartels 1999: 195) shows how a question-answer-sequence provides an informative contribution to discourse. (10) private beliefs
public commitments
joint public commitments
speaker
addressee
speaker
addressee
initial context
E
P
E
E
E
after WHQS (E)
E
P
E
E
E
after commA (P)
P
P
E
P
E
after commS (P)
P
P
P
P
P
The assumptions about informative assertions and questions as well as the conditions that are found in the three systems assumed by Bartels throughout di¤erent stages in an ongoing discourse can be considered a reflection of the conditions that have to be met in order for the speech acts to be carried out (e¤ectively). As the following analysis is concerned with questions, the illustration of this aspect will be restricted to this speech act. Looking at the illocutionary act question in Searle (1969: 66) (cf. (1)), the first part of the preparatory rule mirrors the fact that the speaker’s private beliefs (if at all) only contain the presupposition and that the CG does not already contain the answer either. In that case, the answer also had to be part of the speaker’s private beliefs and public commitments. However, the speaker knows about the proposition as such and can, therefore, o¤er alternatives. The speaker posing the question spells out the essential rule. The second preparatory rule as well as the essential rule capture Bartels’ assumption that P is part of the addressee’s private beliefs as well as of the implicit assumptions one makes within Stalnaker’s wider picture when adding the context update induced by a question to his model, namely that an assertion that induces an assertive context change will follow the question.
154
Sonja Mu¨ller
3. Extraction constructions If a complex sentence such as (11) is introduced into the context, the context update takes place on two levels: On the level of the main context, the sentence has the e¤ect that it is added to the CG that Julia has the belief of Anna inviting Stephan for co¤ee (¼p2) by simultaneously deleting those worlds from the CS in which her doxastic system does not contain this belief. (11) [p2 Julia thinks [p1 that Anna invites Stephan for co¤ee]]. However, the matrix verb introduces a further embedded context. In contrast to (3) (repeated for convenience in (12)), the truth of p1 is not evaluated with respect to the speaker who, therefore, wants it to become information shared by the discourse participants, but with respect to the individual Julia. (12) Anna invites Stephan for co¤ee. (¼p1) Farkas (1992b) refers to the individual with respect to whose system a proposition’s truth value gets assigned as the individual anchor of a proposition and she further assumes that the individual anchoring of propositions happens indirectly: Propositions are anchored in worlds which are again anchored in individuals. The proposition p1 in (11) is thus true in a world which represents the actual world from Julia’s perspective, i.e. the matrix subject Julia is the individual anchor of p1 because it is connected to the world relative to which p1 is evaluated. In (12), however, p1 is true in a world which represents the actual world from the speaker’s perspective, i.e. the speaker is the individual anchor of p1 because he is connected with the world relative to which p1 is evaluated. For the context update induced by a sentence such as (11), the assumptions by Farkas mean that one has to assume further sets of propositions apart from the CG and the CS, namely, a set of propositions containing the propositions that constitute Julia’s beliefs (Di,w in (13)) and corresponding worlds which make up Julia’s belief worlds (Di,w in (13)). In analogy to a CG/CS-update, p1 is added to Di,w and the worlds in which p1 is not true are deleted from Di,w. (13) CG ¼ f g CS ¼ W
Di;w ¼ f g D i;w ¼ W
CG 0 ¼ fp2 g D 0i;w ¼ fp1 g CS 0 ¼ CS \ fw 2 W jp2 ðwÞ ¼ 1g D 0 i;w ¼ D i;w \ fw 2 W jp1 ðwÞ ¼ 1g
The distribution of knowledge in (un)acceptable questions
155
Assuming for complex sentences such as (11) that the context update takes place on two levels, the context change induced by a complex question such as (14) can be considered to look like (15). (14) Who does Julia think that Anna invites for co¤ee? 4, D ¼ {Stephan, Caro} (15)
main context
Julia’s doxastic system5
CG ¼ f g
Di;w ¼ f g
CS ¼ W
Di;w ¼ W
CG0 ¼ CG [ f p1 ga
D0i;w ¼ Di;w [ f p5 ga
CG [ f p2 ga
Di;w [ f p6 ga
CG [ f p3 ga
Di;w [ f p7 ga
CG [ f p4 g
Di;w [ f p8 g
CS 0 ¼ CS \ fw 2 W jp1 ðwÞ ¼ 1ga D 0 i;w ¼ D i;w \ fw 2 W j p5 ðwÞ ¼ 1ga CS \ fw 2 W j p2 ðwÞ ¼ 1ga
D i;w \ fw 2 W j p6 ðwÞ ¼ 1ga
CS \ fw 2 W j p3 ðwÞ ¼ 1ga
D i;w \ fw 2 W j p7 ðwÞ ¼ 1ga
CS \ fw 2 W j p4 ðwÞ ¼ 1g
D i;w \ fw 2 W j p8 ðwÞ ¼ 1g
4. The anonymous reviewer conjectures that extractions across finite sentence boundaries are ungrammatical in standard German. Since the earliest pieces of work concerned with extraction data, linguists have been confronted with the problem that their acceptability cannot always be clearly determined. Individual judgements vary, dialectal di¤erences between Northern and Southern Germany have been assumed (cf. e.g. Grewendorf 1988: 82, Fanselow 1987: 51f.) and the di¤erences are of a rather gradual than categorial nature. Such issues are not only a matter of extractions from that-complement clauses, but can be extended to further extraction contexts. Apart from ideolectal and dialectal variation, social factors have also been made responsible for di¤erent judgements (cf. Adli 2004). Although the data’s status is surely less clear than the judgements for other linguistic phenomena, it is not the case that only Southern or Austrian speakers have been consulted – as presumed by the anonymous reviewer. The extraction data the following analysis is based on are from speakers from the Western part of Germany (North Rhine-Westphalia: Lower Rhine area, Cologne (area)). However, the speakers have more diverse backgrounds (North RhineWestphalia, Hesse, Baden-Wu¨rttemberg, Schleswig-Holstein). A systematic survey (particularly concerned with those extraction data) is definitely still a desideratum. For a recent experimental piece of work on (di¤erent) extraction constructions cf. Kiziak (2011). 5. Assuming that (14) induces the partitions in (16) means attributing a certain meaning to the question which can be paraphrased as in (i).
156
Sonja Mu¨ller
(16) a. Julia thinks that Anna invites Stephan as well as Caro for co¤ee.
¼ p1
Julia thinks that Anna invites Stephan, but does not invite Caro for co¤ee.
¼ p2
Julia thinks that Anna invites Caro, but does not invite Stephan for co¤ee.
¼ p3
Julia thinks that Anna invites neither Stephan nor Caro for co¤ee.
¼ p4
b. Anna invites Stephan as well as Caro for co¤ee.
¼ p5
Anna invites Stephan, but does not invite Caro for co¤ee.
¼ p6
Anna invites Caro, but does not invite Stephan for co¤ee.
¼ p7
Anna invites neither Stephan nor Caro for co¤ee.
¼ p8
Di¤erentiating between the e¤ect that the extraction construction has on the main and the embedded context, on the one hand, the question asks for the proposition which can be added to the set containing Julia’s beliefs (Di,w) (and thereby deleting Julia’s non-belief worlds from Di,w) by opening up the possibilities in (16b). On the other hand, the question concerns the update of CG and CS with one of the propositions from (16a). 3.1. Factivity A well-known phenomenon in the context of extraction constructions such as (14) is that the acceptability of such sentences largely depends on the matrix verb. From (17) to (18), a degradation of the questions is conceivable. (i) ‘What is the proposition that can be anchored in Julia’s doxastic system, the possibilities all being complete answers to the question: Who does Anna invite for co¤ee?’ This interpretation, however, is not the only one possible. Another reading is also possible which can be paraphrased by (ii). (ii) ‘For which individual is it the case that Julia thinks that Anna invites him/her for co¤ee?’ Although those two interpretations do not necessarily make a di¤erence as far as the meaning of an extraction construction under the occurrence of the matrix verb think is concerned, the two di¤erent interpretations do indeed a¤ect the meaning of such questions under the occurrrence of di¤erent matrix verbs. As this issue is not of central interest in this paper, the interested reader is referred to the full account as elaborated on in Mu¨ller (2010).
The distribution of knowledge in (un)acceptable questions
157
(17) Wen meint/ glaubt/ sagt/erza¨hlt Peter, dass der Chef entla¨sst? Who thinks/ believes/ says/tells Peter that the boss fires ‘Who does Peter think/believe/say/tell that the boss will fire?’ (18) Wen ignoriert/ bedenkt/ verdra¨ngt/ verheimlicht Who ignores/ keeps in mind/ represses/ conceals Peter, dass der Chef entla¨sst? Peter that the boss fires ‘Peter ignores (the fact)/keeps in mind/represses/conceals that the boss fires who?’ (intended meaning) The phenomenon at issue is that of bridge verbs. Certain verb classes (¼bridge verbs) allow extractions from their that-complements while other verb classes (¼non-bridge verbs) do not. Among the verbs whose that-clause is less transparent are factive verbs 6 (cf. (18)).7 The phenomenon being at work here has received a large number of explanations over many years of research on locality phenomena in general. As the research tradition in this area has predominantly been syntactic, it is not surprising that there are lots of approaches trying to account for data such as (17) and (18) in structural terms. As the formation of wh-questions in some models within the generative paradigm is assumed to involve movement of the wh-phrase from its base position to the left periphery, syntacticians have proposed accounts which rely on other well-known syntactic movement constraints (cf. e.g. Kiparsky and Kiparsky 1970, Pu¨tz 1975, Cattell 1978: Ross’ (1967) Complex NP Constraint; Stowell 1981, 1986, Cinque 1990, Fukui 1986: Huang’s 1982 Condition on Extraction Domains; Fanselow 1987, Rizzi 1990, Manzini 1992, De Cuba 2006, Basse 2008: (some version of ) Chomsky’s 1973 Subjacency Condition). In the tradition of less abundantly developed semantic and information structural accounts (cf. e.g. Abrusa´n 2008, Szabolcsi and Zwarts 1993, Erteschik-Shir 1973, 2007, Kluender 6. Many other accounts argue that factive verbs have a negative impact on extractions from that-complement clauses. Cf. e.g. Abrusa´n (2008), De Cuba (2006), Rizzi (1990), Truswell (2007). A recent experimental study which confirms the negative influence of factive verbs in English is described in Ambridge and Goldberg (2008: 374). 7. There are other verb classes which have been attested a negative influence on a that-complement clause’s extraction possibilities, e.g. manner-of-speaking verbs such as flu¨stern (whisper), stammeln (stammer), schreien (shout) (cf. e.g. Erteschik-Shir 1973, 2007, Kluender 1991) or implicative verbs such as verursachen (cause), bewirken (cause), beweisen ( prove) (cf. Mu¨ller 2010). This paper focuses on the class of factive verbs. For a uniform analysis accounting for all those cases cf. Mu¨ller (2010).
158
Sonja Mu¨ller
1991), in the following, an analysis will be suggested that traces the negative influence of factive verbs back to the e¤ect that the resulting questions have on discourse. Their roles within communicative settings as described above will be analyzed as well as the decrease in acceptability from (17) to (18), which is considered evidence for the hypothesis to be tested that violating the pragmatic condition of the speaker not knowing the answer to the question posed by himself does have an impact on the question’s acceptability. Before spelling out an explanation of the degraded status of complex question formation under the occurrence of factive that-complements in Section 3.1.3, Section 3.1.1 and Section 3.1.2 will enlarge on the relevant semantic and, respectively, pragmatic properties of factivity.8 3.1.1. Presuppositions semantically – truth conditions and context change The decisive property of that-complements selected by factive matrix verbs is that the proposition expressed is presupposed, that is (19) cannot be considered true without the subordinate clause being considered true as well – a situation which does not apply to (20) under the occurrence of a non-factive verbum putandi or verbum dicendi. (19) [p1 Peter conceals [p2 that Mary is pregnant.]] ! Mary is pregnant. (20) [p1 Peter thinks/says [p2 that Mary is pregnant.]] – Mary is pregnant. In order to test for the presuppositional nature of an inference one can draw and to delimit it from other logical deductions such as implications, traditionally, a whole battery of contexts (the S-family) (cf. Chierchia and McConnell-Ginet 1996: 281) is given. If an inference ‘survives’ in those contexts (cf. (21)), in all likelihood, it will be a presupposition (but cf. e.g. Chierchia and McConnell-Ginet 1996: 282f., Abbott 2005: 3 on problematic issues). (21) a. b. c. d.
Peter conceals that Mary is pregnant. ! Mary is pregnant. Peter does not conceal that Mary is pregnant. ! Mary is pregnant. Does Peter conceal that Mary is pregnant? If Peter conceals that Mary is pregnant, he will not be very happy to be the father.
8. The matrix verb which is often considered the most classical factive verb (wissen [know]) is factored out on purpose in the following argumentation. The reason for this move is the verb displaying several di¤erent meanings some of which are indeed factive and some of which are not (cf. e.g. Reis 1977: 144f., Farkas 1992a: 49). In order to add this verb to the study, one would have to find a way to control for the respective readings.
The distribution of knowledge in (un)acceptable questions
159
As predicted, the presupposition in the complement of conceal remains in a positive (cf. (21a)) as well as a negative (cf. (21b)) declarative, in a yesno-question (cf. (21c)) and in the antecedent of a conditional (cf. (21d)). From the point of view of the context update induced, it is assumed that presuppositions are propositions which have to be true in the CS that exists before the sentence containing them is uttered and its proposition considered true in CS 0 (cf. Farkas 2003: 4, Kadmon 2001: 14, Heim 1992: 186). In this argumentation, presuppositions are understood as requirements on the context state before the sentence containing the presupposition is introduced into the context. For (19), this means that the CS has to contain worlds in which p2 is true when the worlds in which p1 is not true are intended to be excluded from the CS (cf. (22)). Otherwise, the sentence is not defined relative to this context and a context update cannot apply. (22) CS ¼ fw 2 W jp2 ðwÞ ¼ 1g CS 0 ¼ CS \ fw 2 W jp1 ðwÞ ¼ 1g Looking at the phenomenon from the perspective of truth value assignment, others have expressed the very same intuition about the missing licensing of a sentence whose presupposition is not met by arguing that a sentence whose presupposition is not true cannot be assigned a truth value (cf. e.g. Stalnaker 1974: 472, cf. also Frege 1892, Strawson 1952). However, from both interpretations, it follows that a presupposed proposition has already been assigned a truth value and is not open for truth value assignment any more when it comes to assigning a truth value to the sentence that contains the presupposition. 3.1.2. Presuppositions pragmatically – speaker assumptions Choosing a pragmatic approach to presuppositions means that they are not treated in terms of a relation between a sentence and a proposition as it is practised within the semantic view, but that a presupposition is mirrored in the beliefs, in intentions or expectations of speakers. The approach which is the intuitively most plausible one is reached when understanding presuppositions as CG-information. This assumption can e.g. be derived from the following quotation by Stalnaker (2002: 704). In the simple picture, the common ground is just common or mutual belief, and what a speaker presupposes is what she believes to be common or mutual belief. The common beliefs of the parties to a conversation are the beliefs they share, and that they recognize that they share: a proposition j is common belief of a group of believers if and only if all in the group believe that j, all believe that all believe it, all believe that all believe that all believe it, etc.
160
Sonja Mu¨ller
For a sentence such as (23), the assumptions mean that the speaker assumes that the proposition expressed in the complement clause is already part of the CG, that is the speaker and the hearer know that Mary is pregnant and that they both know that. (23) [p1 Peter verheimlicht, [p2 dass Maria schwanger ist]]. Peter conceals that Mary pregnant is ‘Peter conceals that Mary is pregnant.’ As the context update in the way illustrated above involves both evaluating the truth/falsity of propositions with respect to the worlds in the CS as well as adding the same propositions to the CG, the second part of the update being left out in (22) (cf. (24)) illustrates exactly the pragmatic view on the e¤ect on the context referred to in Section 3.1.1. (24)
CG ¼ fp2 g CG 0 ¼ CG [ fp1 g ¼ fp2 g [ fp1 g ¼ fp1 ; p2 g
Assuming the more di¤erentiated view chosen by Bartels (1999), the update induced by an assertion which expresses the proposition P containing a presupposition Q looks like (25). (25) private beliefs
public commitments
joint public commitments
speaker
addressee
speaker
addressee
initial context
Q P
Q P, not P, –
Q –
Q –
Q –
after commitmentsS (P)
Q P
Q P
Q P
Q –
Q –
after commitmentsA (P)
Q P
Q P
Q P
Q P
Q P
Although the same assumptions concerning the cell’s fillings as made above apply to P again, from the simple and straightforward point of view of presuppositions as assumed at this point, the presupposed proposition Q is part of the CG (¼joint public commitments) and due to the
The distribution of knowledge in (un)acceptable questions
161
relation between this system and the other two, Q is also part of those systems: Being part of the joint public commitments implies being part of the single systems constituting the public commitments which again (presupposing an honest, cooperative communicative setting) implies being part of the respective private beliefs.9 The role of the speaker’s attitude towards the presupposed issue that is assumed when choosing a pragmatic perspective on the topic as well as the precise distribution of knowledge assumed by (25) can be backed up by sequences such as (26) to (29). (26) Peter conceals that the school is on fire, Abut I do not consider it true that the school is on fire. (27) Peter thinks that the school is on fire, but I do not consider it true that the school is on fire. (28) Peter conceals that the school is on fire, Abut I do not know whether the school is on fire. (29) Peter thinks that the school is on fire, but I do not know whether the school is on fire. The speaker can neither commit himself to the (first part of the) assertion in (26) without committing himself to the presupposed issue nor can he present the (first part of the) assertion in (28) by simultaneously denying knowledge of the presupposed content in the complement. Both attitudes, however, can be advanced in (27) and (29) under the occurrence of a nonfactive verb without any problems. 9. There are several factors which complicate the simplified picture presented here. Due to lack of space and the focus of this paper, I cannot elaborate on those at length (cf. Mu¨ller 2010 for the full account capturing the facts I can only mention briefly here). The first aspect concerns the fact that the assumptions made by speaker and hearer do not necessarily present true beliefs, but only have to be made for the purpose of the conversation (cf. Stalnaker 1978: 231). This in fact applies to all operations taking part in up-dating the context and is, therefore, no special issue in the discussion of presuppositions. The second aspect which is of relevance for the discussion of presuppositions more directly is the insight that presupposed information does not necessarily have to be old information by being part of the CG before the utterance containing the presupposition is uttered. This phenomenon known as accommodation (cf. e.g. Stalnaker 1974: 202, Lewis 1979: 340, Kadmon 2001: 17f.) also has to be excluded from the illustration here although it has an influence on the analysis presented in Section 3.1.3.
162
Sonja Mu¨ller
3.1.3. A deficient move in discourse The pragmatic account of presuppositions together with the semantics of questions introduced in Section 2 now allows an analysis of the negative influence that factive verbs have on extractions from that-complement clauses. In Section 2, it was assumed that by asking a question such as (30), a speaker aims at getting that proposition that should be added to the set constituting Peter’s beliefs, the proposition answering the root question Who did Mary invite? (cf. the partition in (31)). (30) Wen meint Peter, dass Maria eingeladen hat?, D ¼ {Hans, Fritz} Who thinks Peter that Maria invited has ‘Who does Peter think that Mary invited?’ (31)
Peter thinks that Mary invited Hans as well as Fritz. Peter thinks that Mary invited Hans, but did not invite Fritz. Peter thinks that Mary invited Fritz, but did not invite Hans. Peter thinks that Mary invited neither Hans nor Fritz.
Interpreting the less acceptable question in (32) in analogy to (30), i.e. regarding the speaker as aiming to know what the proposition that is to be anchored in Peter’s ‘ignorance system’ looks like, this proposition is understood as a complete answer to the question Who did Mary invite? and it opens up the possible complete answers in (33). (32) ??Wen ignoriert Peter, dass Maria eingeladen hat?, D ¼ {Hans, Fritz} Who ignores Peter that Mary invited has ‘Peter ignores the fact that Mary invited who?’ (indended reading) (33)
[p5 Peter ignores [p1 that Mary invited Hans as well as Fritz]]. [p6 Peter ignores [p2 that Mary invited Hans, but did not invite Fritz]]. [p7 Peter ignores [p3 that Mary invited Fritz, but did not invite Hans]]. [p8 Peter ignores [p4 that Mary invited neither Hans nor Fritz]].
The distribution of knowledge in (un)acceptable questions
163
As the proposition expressed by a factive complement clause presents CGcontent, by uttering a question such as (32), the deficient situation arises that the proposition that should be added to Peter’s ignorance system cannot be di¤erent from the one that is already part of the CG. If it was possible that Peter could ignore something di¤erent from what is the case anyway, (34) should present an adequate sequence. (34) Mary invited only Hans, Aand Peter ignores (the fact) that Mary invited Hans as well as Fritz. This means that by uttering (32), the speaker opens up alternatives although no alternatives are available. (35) illustrates this situation in discourse. (35)
main context
Peter’s ignorance system
CG ¼ f p2 g
Ii;w ¼ f g
CS ¼ fw 2 W j p2 ðwÞ ¼ 1g
I i;w ¼ CS
CG 0 ¼ CG [ f p5 ga
0 Ii;w ¼ Ii;w [ fp1 ga
CG [ fp6 ga
Ii;w [ f p2 ga
CG [ f p7 ga
Ii;w [ f p3 ga
CG [ f p8 g
Ii;w [ f p4 g
CS 0 ¼ CS \ fw 2 W j p5 ðwÞ ¼ 1ga I 0i;w ¼ I i;w \ fw 2 CSj p1 ðwÞ ¼ 1ga CS \ fw 2 W j p6 ðwÞ ¼ 1ga
I i;w \ fw 2 CSj p2 ðwÞ ¼ 1ga
CS \ fw 2 W j p7 ðwÞ ¼ 1ga
I i;w \ fw 2 CSj p3 ðwÞ ¼ 1ga
CS \ fw 2 W j p8 ðwÞ ¼ 1g
I i;w \ fw 2 CSj p4 ðwÞ ¼ 1g
Using the factive verb presupposes that the proposition in the embedded clause (here p2) is part of the CG, so that the question whether p1, p2, p3 or p4 is part of Peter’s ignorance system simply does not arise. In parallel, adding p5 to p8 to the CG does not present a serious possibility as the CS’s reduction can only be induced by p6. Looking at which information is available in the di¤erent systems at various stages in the course of discourse within the model assumed by Bartels (1999), the situation arises that P is part of every cell available (cf. (36)).
164
Sonja Mu¨ller
(36) private beliefs
public commitments
joint public commitments
speaker
addressee
speaker
addressee
initial context
P
P
P
P
P
after WHQS (E)
P
P
P
P
P
after commA (P)
P
P
P
P
P
after commS (P)
P
P
P
P
P
As the information asked for is part of the CG, P is already part of the public commitments of the speaker and the addressee as well as their private beliefs. That is, before the actual move in discourse that is supposed to update the context has been carried out, the only state that can be reached by this move already exists. The only answer that the addressee can give is naming P which the speaker adds to his beliefs, accepts, publicly commits himself to and which thereby becomes shared public commitment. However, no progress is furthered throughout such a course of discourse. The same problem, however, does not arise with the acceptable question in (30) as it poses no requirements on the CG and Peter’s belief system can be updated totally independently from the propositions in the CG. In principle, Peter can believe in a completely di¤erent possible answer to the question Who did Mary invite? as it is possibly known that it is the case (cf. (37)). Opening up alternatives is, therefore, legitimate. (37) Mary invited only Hans, but Peter thinks that Mary invited Hans as well as Fritz. Consequently, by asking for a constituent contained in a that-complement clause selected by a factive matrix verb, a question results which does not progress further in discourse, but rather induces a step backwards because it asks for the form of an issue of which the speaker simultaneously expresses that he (as well as the discourse partner) knows about it, and that he knows that they both know about it. This paper therefore suggests that the unacceptability of questions such as (32) (when interpreted as in (33))10 10. Cf. Mu¨ller (2010) and Mu¨ller (to appear) for another interpretation and its consequences concerning the question’s acceptability. Cf. also footnote 5. Therefore, the reviewer is completely right in referring to a scenario in which
The distribution of knowledge in (un)acceptable questions
165
is due to the fact that the questions have to be considered fully uninformative operations in discourse. The questions do not fulfill the function that is usually associated with questions, namely to open up alternatives whose reduction leads to an increase in private as well as consciously shared knowledge. Therefore, factive matrix verbs present the first case of linguistic material in this study whose occurrence in questions yields a violation of the condition of the speaker not knowing the answer to the question he poses. The general claim that a pragmatic condition influences the question’s well-formedness might be considered controversial. The decisive aspect to keep in mind with respect to that claim, however, is that the reasons which lead to the assumption that the speaker displays a certain knowledge state do not result from a particular context being constructed (of course a perfectly acceptable question does not degrade when being asked for the second time in one and the same discourse sequence although it has already been answered), but are of a much more general nature: They are anchored in the properties of the linguistic material employed to yield the two people are invited and the question aims at finding out whether the matrix subject’s attitude is assigned to the invitation of both individuals or only one individual. The di¤erence between those readings is discussed in detail in Mu¨ller (2010) and Mu¨ller (to appear). The correct point the reviewer makes is related to the bigger issue concerning the availability and precise anchoring of negative information in the question’s denotation. Interestingly (because of providing evidence for assumptions made in Mu¨ller (2010)), the reviewer expresses the reading he considers non-deficient by resuming the wh-pronoun in the subordinate clause. He mentions (i) and not the worse (ii). (i) Von wem weiß Peter nicht, dass Maria ihn eingeladen hat? About whom knows Peter not that Mary him invited has ‘About whom doesn’t Peter know that Mary invited him?’ (ii) Wen weiß Peter nicht, dass Maria eingeladen hat? Who knows Peter not that Mary invited has ‘Who doesn’t Peter know that Mary invited?’ Cf. Mu¨ller (2010) for the full approach which also provides an account of the influence of resumptive pronouns. Note at this point that ignorieren is treated here as a member of the class called Beru¨cksichtigungspra¨dikate in Reiss (1977: 202¤.). It is not meant to express lacking knowledge, as the anonymous reviewer seems to assume. At least in my dialect, X ignoriert, dass . . . cannot have this reading at all. However, this aspect does not change the more general point the reviewer makes here (and as which it has been understood above).
166
Sonja Mu¨ller
respective knowledge states of the participants and are, therefore, due to the meaning components that are invariably available for this particular linguistic material (cf. also Gajewski’s (2002, 2008) concept of L-analyticity to cope with similar questions when tracing ungrammaticality back to tautologies and/or contradictions). In the present case, it is the verb’s factivity, in the cases discussed below, it will be modal meaning components. 3.2. Anti-factivity In the case of the presupposition at issue under the occurrence of factive verbs, the speaker has to commit himself to the truth of the proposition expressed by the that-complement clause. It is also conceivable that the commitment has to be a negative one. If one finds the case of a negative presupposition – which for the data under discussion in this paper would have to be a complement clause which has to be considered false by the At a later stage, he further argues that ignorieren is not a factive verb at all under the ‘Beru¨cksichtigungs’-reading. He mentions (iii). (iii) Peter glaubt fa¨lschlicherweise, Peter believes wrongly aber er ignoriert das einfach, but he ignores this just daher gibt es auch keine therefore are there also no
dass Maria ihn betru¨gt, that Mary him betrays
Probleme in ihrer Ehe. problems in their marriage
‘Peter wrongly believes that Mary betrays him, but he simply ignores this, therefore, there are no problems in their marriage.’ Such weaker presuppositions which are only anchored in a matrix subject’s belief system have been assumed to hold for semi-factive verbs (e.g. herausfinden ( find out)) or emotive factives (e.g. bedauern (regret)) (cf. e.g. Seuren 1991). Such verbs can, but do not have to, be interpreted factively (namely in a pragmatic sense as to present common ground content). The reviewer’s example thus raises the interesting and more general question whether there are verbs which can only be interpreted factively at all – although I do have doubts about the true/transparent use of the matrix verb by the speaker when phrasing the same situation by using a dass-complement clause under ignorieren. (iv) sounds quite strange to me (which (iii) does not at all). (iv) Peter glaubt fa¨lschlicherweise, dass Maria ihn betru¨gt, Aaber er ignoriert einfach, dass Maria ihn betru¨gt;daher gibt es auch keine Probleme in ihrer Ehe. ‘Peter wrongly believes that Mary betrays him, Abut he simply ignores (the fact) that Mary betrays him; therefore, there are no problems in their marriage.’
The distribution of knowledge in (un)acceptable questions
167
speaker who commits himself to the truth of the complex clause – the prediction of the analysis in Section 3.1.3 would be that the question resulting from extracting a constituent from this that-clause should be as unacceptable as the parallel question in which factive verbs make up the matrix clause because the same deficient situation would arise in discourse. The negated proposition expressed by the that-clause would then present CGinformation and therefore asking for the precise form of this issue would be an inadequate move in discourse. A (putative) case at issue presents the that-complements selected by volitional predicates. The aim of the following section is to show how predictions made by the hypothesis in Section 3.1.3 can serve the purpose of (indirectly) checking assumptions on the interpretation of the linguistic material made use of in order to test for this hypothesis. 3.2.1. The anti-factivity of complements of volitional predicates in Meinunger (2007) Meinunger (2007) assumes that volitional predicates are anti-factive (or counter-factive), i.e. that they presuppose that the issue expressed in the subordinate clause is not the case in the actual world. His generalisation in (38) and his concluding assumption ‘‘The claim thus is that true volitional predicates presuppose the non-givenness of the proposition contained in their complement clause.’’ (cf. Meinunger 2007: 168f.) are based on data such as (39) to (41). (38) Volitional predicates in a broad sense are anti-factive (or counterfactive). Similar to counterfactual constructions, they refer to eventualities that are not given. (39) a.
b.
(40) a.
Ich wu¨nsche (mir), dass du in den Garten gehst. I wish me that you in the garden go ‘I wish you would go into the garden.’ Ich will, dass er ein Auto kauft. I want that he a car buys ‘I want him to buy a car.’ Ich wollte, dass er ein Auto kauft. I wanted that he a car buys ‘I wanted him to buy a car.’
168
Sonja Mu¨ller
b.
Sie wollte, dass der Vogel stirbt. She wanted that the bird dies ‘She wanted the bird to die.’
c. Hans bat Maria, dass sie ihm eine Goldmu¨nze mitbringt. Hans asked Mary that she him a golden coin brings ‘Hans asked Mary to bring him a golden coin.’ Meinunger (2007: 167f.) (41) a.
b.
Ich wu¨nschte, du ha¨ttest mehr Zeit fu¨r mich. I wished you had more time for me ‘I wish(ed) you had more time for me.’ Seine Eltern wu¨nschten sehr wohl, er ha¨tte sie nie His parents wished very PART he had her never kennen gelernt. got to know ‘His parents DID actually wish he had never gotten to know her.’ Meinunger (2007: 163)
The issue expressed in the complement clause is supposed not to be existent in the actual world regardless of whether the matrix predicate displays the present tense or past tense in the indicative (cf. (39), (40)) or subjunctive (cf. (41)). According to (38), (39) to (41) presuppose that the addressee is not in the garden at the time of the utterance, does/did not buy a car, the bird was alive, Mary did not bring Hans a golden coin, the addressee did not have time and the referent of the pronoun he got to know the referent of the pronoun she. As an exception to this relation between the subordinate proposition and the state of the world Meinunger gives the sentences in (42) in which the wish’s content spans a longer period. (42) a.
b.
Viele Ma¨nner wollen, dass ihre Frauen arbeiten. Many men want that their wives work ‘Many men want their wives to work.’ Hans wu¨nscht sich, dass seine Frau ihr Haar o¤en tra¨gt. Hans wishes himself that his wife her hair open wears ‘Hans wishes that his wife would wear her hair loose.’
The distribution of knowledge in (un)acceptable questions
169
c. Eine Mutter mo¨chte, dass ihr Kind glu¨cklich ist. A mother wants that her child happy is ‘A mother wants her child to be happy.’ Meinunger (2007: 168) From (42), it cannot be derived that the wives do not work, that Hans’ wife does not wear her hair open and that the child is not happy. If Meinunger is right, the account presented in Section 3.1.3 would predict that it should not be possible to extract constituents from thatcomplements under interpretations such as (39) to (41) because of the embedded propositions’ anti-factivity. However, this prediction is not borne out. Volitional predicates in the narrow (cf. (43)) as well as broad sense (cf. (44)) make fairly good bridge verbs. (43) a.
b.
(44) a.
b.
Welchen Kandidaten will/wollte die Studentin, Which candidate wants/wanted the student dass der Pru¨fer aufruft? that the examiner calls ‘Which candidate does/did the student want the examiner to call?’ Wen wu¨nscht/e der Angestellte sich, Who wishes/ed the employee himself dass der Chef entla¨sst/entließe? that the boss fires/would fire ‘Who does/did the employee wish for the boss to fire?’ Wen verlangt/e der Detektiv, Who demands/ed the detective dass der Spion beobachten soll? that the spy observe shall ‘Who does/did the detective demand the spy to observe?’ Welche Fenster bittet/bat die Mutter, Which windows asks/asked the mother dass die Tochter putzen soll? that the daughter clean shall ‘Which windows does/did the mother ask her daughter to clean?’
170
Sonja Mu¨ller
It neither seems to have an influence whether the matrix verb’s tense is present or past nor whether the verbal mood is indicative or subjunctive. Having volitional predicates in the matrix of such extraction constructions thus does not lead to degraded acceptability. The natural consequence of the well-formed data in (43) and (44) seems to be that either the analysis presented in Section 3.1.3 or Meinunger’s generalisation on the anti-factivity of propositions embedded under volitional predicates needs rethinking. In order to decide this issue, the following section will critically analyse Meinunger’s assumptions on the anti-factivity of subordinate clauses selected by volitional predicates and it will be shown why the generalisation in (38) cannot be adhered to, whereas the condition on (un)acceptable extraction constructions suggested in this paper can be kept up. It will be argued that the crucial aspect that Meinunger does not take into account is the question of whose version of the actual world has to be considered when speaking about the embedded proposition’s anti-factivity. 3.2.2. Anti-factivity within speakers’ belief systems Looking at sentences in which wollen constitutes the matrix (cf. (45)), no restrictions apply with respect to the validity of the embedded proposition in the actual world. The issue can be existent or non-existent. (45) Es regnet (nicht), und/(aber) ich/Peter will, dass es regnet. It rains (not) and/(but) I/Peter wants that it rains ‘It is (not) raining, and/(but) I/Peter want(s) it to rain.’ If wu¨nschen constitutes the matrix, the above assumptions shift. As the conditional paraphrase of (46a) in (46b) illustrates (cf. Heim 1992: 205), (46a) indeed presupposes that the issue expressed in the complement clause is not existent. (46) a. b.
Amina wishes that Volker fills in the cupboard. Amina thinks that she would be in a more desirable world than she actually is, if Volker filled in the cupboard.
However, the anti-factivity crucially only has to be valid within the matrix subject referent’s doxastic system and not in the world of the utterance (cf. Heim 1992: 202f.). As (47) illustrates, only Amina’s doxastic system is of relevance here: An individual cannot wish for something that is already existent in its version of reality.
The distribution of knowledge in (un)acceptable questions
(47) a.
b.
c.
d.
171
Volker is filling in the cupboard, but Amina thinks that he is not filling in the cupboard, and wishes that Volker would fill in the cupboard. Volker is not filling in the cupboard. Amina thinks that Volker is not filling in the cupboard and wishes that Volker would fill in the cupboard. Volker is filling in the cupboard, Amina thinks that Volker is filling in the cupboard and Awishes that Volker would fill in the cupboard. Volker is not filling in the cupboard. However, Amina thinks that he is filling in the cupboard, and Awishes that Volker would fill in the cupboard.
Meinunger’s assumption concerning the embedded proposition’s antifactivity is understandable under an analysis of the examples he chooses: The lexical material involves a change of state so that the expressions presuppose their negative counterpart, i.e. buying a certain car (cf. (39b), (40a)) presupposes that the customer has not carried out this particular purchase, the death of a bird (cf. (40b)) supposes that it was alive before and bringing a certain object (cf. (40c)) presumes that the beneficiant does not possess this object yet. On the other hand, Meinunger predominantly refers to examples under the occurrence of first person subjects and second person addressees (cf. (39), (40a), (41a)). This suggests that the world in which the presupposition has to be fulfilled is the actual world that speaker and addressee share. Constructing respective contexts, however, shows that the negative presupposition only has to be valid within the matrix subject’s epistemic system (cf. (48), (49)). (48) At a garden party. A: I wish that Mary would go into the garden. B: Eh? But she is already there. A: Oh, I thought that was Clara who is sitting in the garden. Since they’ve had such similar haircuts, I keep mixing them up. (49) A: I wish that you would go into the garden. B: Eh? Paul, we ARE in the garden. A: What? This little piece of green is meant to be that garden for which Fritz bought an electric lawn-mower?
172
Sonja Mu¨ller
As (48) shows, the requirement has to be met that the matrix subject is convinced of the non-existence of the circumstance. A assumes a di¤erent version of reality. A situation such as (49) suggests that speaker and addressee share their knowledge about the state of the world, but this does not have to be the case. A similar scenario can be made up for past tense and subjunctive matrix verbs (cf. (50), (51)). (50) Peter wished that Mary would break her relationship with Klaus back then, but only because Peter did not know that there was nothing going on between them anymore. (51) A: I wish you broke o¤ your relationship with Klaus. B: My dear, nothing has been going on between us for years. A: Oh, I’m sorry. I thought that was still up to date. The examples show that some volitional predicates require the embedded proposition to be false within the matrix subject’s doxastic system. However, in all of the cases the proposition can be either true or false within the worlds of the CS. This aspect can be transferred to directives such as befehlen (order) and bitten (ask for). It is part of the conditions of a directive speech act that the issue ordered or asked for is not already the case in the actual world (cf. e.g. the propositional content rule for requests in Searle 1969: 66). (52) is not an adequate utterance in a situation in which Mary is just putting on her coat. Similarly, (53) cannot be uttered adequately in a context in which Mary is already leaving by car. (52) Peter to Mary: Put on your coat! (53) Hans to Mary: Please go by car! For complex sentences such as (54) and (55), it has to be assumed that the issue ordered or requested is not already existent in the embedded discourse. (54) Peter befiehlt, dass Maria die Jacke anziehen soll. Peter orders that Mary the coat put.on shall ‘Peter orders Mary to put on the coat.’ (55) Hans bittet Maria, dass sie mit dem Auto fa¨hrt. Hans asks Mary that she with the car goes ‘Hans asks Mary to go by car.’
The distribution of knowledge in (un)acceptable questions
173
The assumption that Mary is not putting on her coat and that Mary is not already going by car is made by Peter and, respectively, Hans and is, therefore, related to their version of reality. (56) seems weird, whereas (57) seems unproblematic. The same assumption applies if one constructs sentences parallel to (56) and (57) for (55). (56) Peter considers it true that Mary is putting on her coat right now, and APeter orders Mary to put on the coat. (57) Mary is putting on her coat right now. Peter does not know it and orders Mary to put on the coat. Consequently, a requirement of an order or request is the person carrying out this speech act to assume the non-existence of the respective issue. In embedded contexts, the matrix subject’s belief worlds are relevant which, however, do not have to match the actual world. In root clauses, the requirement refers to the actual world. If speaker and addressee do not share the same version of reality, uttering an imperative e.g. is an adequate move again (cf. (58)). (58) On the mobile phone: Late at night. B was at a party. A: Go home! B: But I’ve been at home for two hours now. A: Oh, I’m sorry. I didn’t know that. Reflecting on his own speech act, a speaker might even explicate his error (cf. (59)) or admit having made an utterance under wrong assumptions (cf. (60)). (59) I might be wrong and you do work enough for school, but because it doesn’t look like that to me, I beg you to work harder for school. (60) I was indeed wrong and took Mary for Klara and ordered Mary to stop talking although she hadn’t uttered a single word. In the context of the direct utterance of an order or request, those assumptions are usually made tacitly. Consequently, the examples referred to in this section illustrate that Meinunger’s generalisation concerning the embedded proposition’s anti-factivity under the occurrence of volitional matrix predicates cannot be assumed to hold. Anti-factivity does play a role in the interpretation of volitional predicates, however, it only concerns the proposition’s falsity within the doxastic system of the individual originally having the wish or uttering the directive. The condition on well-formed
174
Sonja Mu¨ller
questions asking for constituents contained in that-complement clauses suggested in Section 3.1.3 can thus be kept. Under the occurrence of a volitional predicate, the speaker does not ask for CG-information as he does under the occurrence of factive verbs. As will be shown in the following, factivity is only one aspect of meaning that – when occurring in questions – leads to the deficient pragmatic situation that the speaker asks for a known issue. A further aspect of meaning that can have a similar e¤ect is modality, occurring here in the shape of modal verbs (cf. 4.) and modal particles (cf. 5.).
4. Modal verbs in questions Modal verbs present one further case in the context of the idea pursued in this paper that the distribution of knowledge as coded by particular linguistic material and its meaning has an impact on the possibility of question formation in the following sense: If the pragmatic conditions concerning the distribution of knowledge between speaker and hearer that leave their mark on Searle’s assumptions about the structure of the illocutionary act question (cf. (1)) are not met, the question’s acceptability degrades. The occurrence of modal verbs is interesting because, on the one hand, they can serve the purpose of dissolving the deficient situation ascribed to extractions involving factive matrix verbs (cf. Section 4.1) and, on the other hand, fairly unacceptable questions can be constructed in which a situation arises which is very similar to the one in the bad extraction constructions without modal verbs (cf. Section 4.2 and Section 4.3). 4.1. Modal verbs in the matrix of that-complements – neutralising factivity In Section 3.1.3 it was claimed that the degraded acceptability of complex questions such as (61) is due to the complement clause’s factivity. The decisive property of factive complements was illustrated by sentences such as (62). The embedded proposition needs to be true in the state of the discourse context in which the complex sentence is considered true. If the embedded issue is valid in the CS, then this proposition also has to be part of the CG (or at least part of the speaker’s knowledge)11. A speaker 11. I will come back to the informative use of presupposed material in discourse in Section 6.
The distribution of knowledge in (un)acceptable questions
175
posing the question in (61), therefore, asks for the precise form of an issue for which he expresses at the same time that he as well as the addressee know about. (61) ??Wen ignoriert Peter, dass Maria eingeladen hat? Who ignores Peter that Mary invited has (62) Peter ignores (the fact) that Mary invited Hans. a. ABut Mary did not invite Hans. b. ABut I do not know whether Mary invited Hans. The prediction of the account pursued here is that dissolving the deficient situation that is assumed to hold for (61) should lead to an increase in acceptability. Modal verbs can serve the purpose of creating such a situation. While the subordinate clause’s truth cannot be denied in (62a), the continuation in (63) illustrates that the embedded issue, which is the same in (62) and (63), by no means has to be the case and, therefore, does not present information in the CG which the speaker has to know about and which has to exist when the complex sentence is intended to become part of the CG. (63) Peter soll ignorieren, dass Maria Hans eingeladen hat. Peter shall ignore that Mary Hans invited has ‘Peter shall ignore (the fact) that Mary invited Hans.’ However, Mary actually did not invite Hans. Klaus who gave that advice apparently took Hans for Fritz. The latter is in fact invited and Peter cannot stand him. The pairwise comparison of the extraction constructions with and without modalising the matrix clause (cf. (64)) provides evidence for the assumption that the distribution of knowledge does have an impact on the questions’ degree of well-formedness. (64b) certainly displays a higher degree of acceptability than (64a). (64) a. b.
Wen ignoriert Peter, dass Maria eingeladen hat? < Wen soll Peter ignorieren, dass Maria eingeladen hat?
In (64b) neutralising the factive verb is caused by the modal verb sollen in its interpretation as expressing deontic necessity. However, this example does not present an isolated case. As further examples in (65) to (68) show, the same e¤ect can be evoked by modal verbs of di¤erent interpretations.
176
Sonja Mu¨ller
(65) a.
Peter verheimlicht, dass Maria Hans tri¤t. ! Peter conceals that Mary Hans meets ! Maria tri¤t Hans. Mary meets Hans ‘Peter conceals that Mary meets Hans ! Mary meets Hans.’
b.
Peter muss (–nach Fritz–) verheimlichen, Peter must –according Fritz– conceal dass Maria Hans tri¤t. that Mary Hans meets ‘According to Fritz, Peter has to conceal that Mary meets Hans.’ This is the only way for Fritz to understand the successful family life of the Meiers. But maybe Mary in fact does not meet Hans and Fritz only thinks that this is the case. This would also help explaining them peacefully living together.
(66) a.
Wen verheimlicht Peter, dass Maria tri¤t? < Fritz: Well, I’m surprised that there haven’t been quarrels in the family yet. Peter simply has to conceal what is going on there. Who would have expected this from Mary? What’s your opinion on this issue? Paul: I’m not that well-informed what’s going on there. Wen muss Peter verheimlichen, dass Maria tri¤t?
b.
(67) a.
b.
Peter veru¨belt seiner Mutter, dass sie den Mathelehrer liebt. Peter blames his mother that she the maths teacher loves ! Peters Mutter liebt den Mathelehrer. ! Peter’s mother loves the maths teacher ‘Peter blames his mother for loving the maths teacher. ! Peter’s mother loves the maths teacher.’ Peter darf seiner Mutter veru¨beln, dass sie den Peter is allowed his mother blame that she the Mathelehrer liebt. maths teacher loves ‘Peter is allowed to blame his mother for loving the maths teacher.’
The distribution of knowledge in (un)acceptable questions
177
But this is only such a move by the psychologist to oblige Peter. His mother is separated from her husband, but in fact does not have a new partner and also does not love the maths teacher. (68) a. b.
Wen veru¨belt Peter seiner Mutter, dass sie liebt? < Wen darf Peter seiner Mutter veru¨beln, dass sie liebt (in order to come to terms with the situation that his parents are divorced)?
In (65b) and (66b), the modal verb expressing necessity is interpreted epistemically. As (65) shows, it dissolves the complement clause’s factivity and leads to an increase in the extraction construction’s acceptability as the direct comparison between (66a) and (66b) shows. In (67b) and (68b), a modal verb expressing possibility occurs. The extra-subjective-volitive (cf. Zifonun, Ho¤mann and Strecker 1997: 1891) usage of du¨rfen has the impact on the that-complement selected by the factive veru¨beln (blame) to lose its obligatory factivity. A di¤erence in the extraction constructions’ acceptability can be observed again (cf. (68)). 4.2. Performatively vs. descriptively used modal verbs Although the positive influence that modal verbs have on extraction constructions under the occurrence of factive matrix verbs which usually cause the complement’s opacity has been attested for various interpretations and usages, such increases of the structures’ acceptability do not occur under every appearance of modal verbs. E.g. if sollen receives a bouletic interpretation whose source is the person posing the question in (69), the question cannot be considered acceptable. A context which clarifies this reading in an unextracted structure can be found in (70). (69) AWen soll Peter ignorieren, dass Maria eingeladen hat? Who shall Peter ignore that Mary invited has (70) Peter soll ignorieren, dass Maria Hans eingeladen hat. I hope that the party will go peacefully then. Why does there always have to be scu¿e among guests? Although a continuation such as (71) without assuming the subordinate clause’s factivity does not seem very plausible, the relevant factor at issue seems to be another (more general) one. (71) ABut I know that Mary did not invite Hans.
178
Sonja Mu¨ller
The modal verb is used performatively in (69) and (70) (cf. e.g. Portner 2005, Schwager 2005, Ninan 2005) in contrast to a descriptive usage in the examples analysed in (63) to (68). In order to formulate an explanation for the data under consideration, it seems reasonable to have a closer look at the two possible ways of using modal verbs, independently of extraction constructions and factivity. While the sentence in (72) provides a description of the world regarding a certain command, (73) carries out the command itself, that is (73) has the same e¤ect as the imperative in (74), whereas (72) describes what is commanded. (72) You must do the shopping today (as far as I know). (73) You must go now!
Schwager (2005: 47/156f.)
(74) Go now! As (75) (sollen being understood as in (76), that is the speaker who utters a wish by sollen also asks for the argument reference) shows, (wh-)question formation is not possible in general under the occurrence of a performatively used modal verb. (75) AWen soll Peter einladen? Who shall Peter invite ‘Who shall Peter invite?’ (76) Peter soll Maria einladen! Dann wird die Party ein Hit! Peter shall Mary invite then becomes the party a hit ‘Peter shall invite Mary! The party will fly then!’ The reasons for the unacceptability of (69) are thus independent ones as they also take e¤ect when performatively used modal verbs occur in simpler wh-questions such as (75). The degraded status of the question in (69) cannot be argued to be due to the factivity of the complement clause not being dissolved. Scenarios can be construed with performative modal verbs in the matrix in which the embedded proposition’s truth seems to have to be presupposed (cf. (70), (71)) as well as cases in which the truth can be denied (cf. (77)). (77) Karl: Peter soll verheimlichen, dass Maria Hans liebt. Karl: Peter shall conceal that Mary Hans loves ‘Karl: I want Peter to conceal that Mary loves Hans.’
The distribution of knowledge in (un)acceptable questions
179
I want to observe how he behaves in order for the message not to become publicly known. Fritz: But Mary does not love Hans though. What’s the sense of all this? Karl: Yes, I know that, of course. But I want to see how Peter behaves: Whether he gives the information away, stays mum, to whom he maybe confides it. However, as under the reading of (70) and (71) in which the presupposition seems to be maintained, (78b) does not yield an improvement of the rather bad construction in (78a) when being interpreted along the lines of (69) but against the background of (77).12 (78) a. b.
??Wen verheimlicht Peter, dass Maria liebt? AWen soll Peter verheimlichen, dass Maria liebt?
The question on the conditions under which a complement clause’s factivity is (not)/can(not) be dissolved under the occurrence of a performatively used modal verb unfortunately cannot be answered at this point. As factivity or its dissolve cannot be relied on for a parallel treatment of extractions such as e.g. (78a) and (75), another way of arguing has to be chosen. As has been mentioned before, sentences such as (73) in which a modal verb is used performatively are interpreted as imperatives (cf. (74)). It has been oberserved that wh-question formation is not possible if the finite verb displays imperative verbal morphology (cf. e.g. (79), (80)). (79) *Wen ruf an? Who call.IMP to
Schwager (2005: 77)
(80) *Wer geh in die Egon-Bar? Who go.IMP in the Egon-pub
Lohnstein (2000: 85)
While Lohnstein (2000, 2007) suggests a semantic account relying on the interaction between truth value assignment, evaluation domains of propositions and partitioning, Schwager (2005) chooses a more pragmatic perspective. The starting point for her explanation of data such as (79) and 12. Although the scenario construed here suggests the reading that the embedded proposition is known to be false, Karl could also answer Fritz’ question without any incoherences with (i). (iv) Well, we don’t know whether she loves him or not, but actually this isn’t of any relevance for the experiment I’m conducting. I only want to see how Peter behaves [. . .] (continued as in (77)).
180
Sonja Mu¨ller
(80) is the observation that question formation is possible if the question is not interpreted as information-seeking, but rhetorically (cf. (81) taken from Schwager 2005: 77). (81) (Speaking to a child who is carrying around a flower pot it should actually be able to put into the right place.) Na komm, du weißt es doch. Wo stell den Hey come, you know it PART Where put.IMP the Blumentopf hin? flower pot to ‘Hey come on, you know that. Where are you supposed to put the flower pot?’ Schwager explains this observation by assuming that uttering an imperative triggers a presupposition which is understood as an epistemic authority condition. This basically captures the fact that it is plausible to presuppose that a speaker knows about the issue he directs at the addressee and commands, commends or asks for. As the imperative carries out a directive speech act which due to its direction of fit (world-to-word) wants to cause a change of the world, this goal can only be successfully achieved if the speaker provides the information on what the world should be like (cf. Schwager 2005: 158f. for a formal definition of the presupposition). If a speaker poses a question such as (79), he, therefore, violates the authority presupposition as the missing argument reference suggests him having a cognitive deficit with respect to the issue commanded/commended/asked for which he wants to be removed by the addressee. However, if the question is not meant as information-seeking, but e.g. as a rhetorical question, the speaker does not express a cognitive deficit, that is he does not lack the information regarding the place he wants the addressee to put the flower pot. On the contrary, he knows the answer, assumes that the addressee knows it and that the addressee knows that they both know the answer. Schwager’s concept can be used to explain why only very particular readings of performative modal verbs are compatible with question formation (cf. also Schwager 2005: 77f.). Looking again at (75) (repeated in (82)), it was argued that the question does not allow a reading under which the anchor of the performative modal verb is the same person who is posing the question. (82) AWen soll Peter einladen?
The distribution of knowledge in (un)acceptable questions
181
This impression now follows: Assuming that the utterance of a performative modal verb very much like an imperative verb also triggers the authority presupposition, namely the assumption that a speaker expressing e.g. his wish knows what the wish is about, i.e. the lack of knowledge expressed by the constituent question and the existing knowledge as expressed by the performative modal clash (cf. (83a)). However, if the source of the wish is someone di¤erent than the person asking the question, the same problem does not arise (cf. (83b)). The modal is used descriptively and the default interpretation as an information-seeking question is unproblematic. If the modal is used performatively, however, only specific, pragmatically well-defined readings such as the rhetorical one (cf. (83c)) are allowed for (82). (83) a. b. c.
reading 1: AWho do I hereby wish that Peter would invite? reading 2: Who does Karl wish that Peter would invite? reading 3: The answer is obvious, isn’t it? So, come on: Who can I wish that Peter would invite? (I know that I always want Peter to invite his brother, you know that I always want Peter to invite his brother and that we both know that I always want Peter to invite his brother.)
4.3. Evidential modal verbs Apart from the deontic or bouletic reading, sollen can also be interpreted evidentially (cf. (84) taken from Abraham (this volume)). (84) Haider soll betrunken gewesen sein. Haider shall drunk been have ‘Haider is said to have been drunk.’ No problems arise if the modal is interpreted that way when occurring in a wh-question (cf. (85)). (85) Wen soll Peter eingeladen haben? Who shall Peter invited have ‘Who is Peter supposed to have invited?’ The unproblematic nature of (85) can be easily accommodated in the analysis pursued here when assuming with Abraham (this volume) that epistemic modals ‘‘denote a twofold deixis’’, that is enter an anchoring with
182
Sonja Mu¨ller
respect to the source of the proposition expressed and undergo an evaluation by the speaker. The decisive factor for deriving the well-formedness of (85) is Abraham’s assumption that in the case of sollen the source of the information is neither the speaker himself nor the addressee, but a third person. Because of this characteristic of the interpretation, it is excluded from the start that the person posing the question and the source of the information conveyed coincide which would lead to a pragmatic deficiency as in the case of performative modals. In order for the latter to sensibly occur in questions, the source parameter has to shift from the current speaker to someone else (cf. (83b)). Such a shift, however, is not needed in examples such as (85) because the two anchors (the person lacking knowledge vs. the holder of knowledge) are never identical. In case of acceptable extractions such as (86), the question of whether the presupposition that ignore normally triggers gets dissolved can easily be answered positively. (86) Wen soll Peter ignoriert haben, dass Maria eingeladen hat? Who shall Peter ignored have that Mary invited has ‘Peter is supposed to have ignored the fact that Mary invited who?’ (intended meaning) Interpreting sollen evidentially has the same e¤ect as embedding the sentence under a verb of saying. The only di¤erence between (86) and (87) is that (87) names the precise source (¼Hans). (87) Wen sagt Hans, dass Peter ignoriert hat, Who says Hans that Peter ignored has dass Maria eingeladen hat? that Mary invited has ‘Hans says that Peter ignored the fact that Mary invited who?’ (intended meaning) In (86) as well as (87), the information about someone having ignored the embedded issue is shifted into someone else’s system of saying. As long as the current speaker does not commit himself to the fact that an issue indeed gets ignored, the presupposition is not triggered in the current discourse situation. The occurrence of evidential modals in questions in general provides evidence for the assumption that the participating sources of knowledge as expressed by the respective modal verb and the lack of knowledge as
The distribution of knowledge in (un)acceptable questions
183
expressed by the speech act carried out direct the question’s acceptability as they must not be identical for the question to be fully acceptable. Abraham (this volume) argues that di¤erent evidential modals can be ranked on a scale expressing degrees of reliable assessment of the issue’s truth. Comparing sollen and mu¨ssen e.g., he assumes that the latter is ranked higher because the truth assessor is the speaker himself, while in the case of sollen, the responsibility is shifted to a third instance. As in (88), under the use of evidential mu¨ssen, the speaker himself considers something evidence which leads him to e.g. state that Peter must be at home. (88) Peter muss zu Hause sein. Das Licht brennt. Peter must at home be the light burns ‘Peter must be at home. The light is on.’ The ranking assumed by Abraham is in fact mirrored in the readings wellformed questions can get if evidential mu¨ssen occurs. If the evidence for stating that something is necessarily the case is available to the speaker himself, he cannot sensibly ask a question such as (89), that is the modal (and, therefore, the source of the evidence) cannot be assigned to the questioner here. (89) A: Wo muss Peter sein (, wenn man sich hier sein Zimmer so anschaut)? A: Where must Peter be if one self here his room like that looks.at ‘A: Where must Peter be (, just looking around his room)?’ The unavailability of this reading follows from the assumptions made above and by using Schwager’s (2005) authority presupposition: It is not sensibly possible for a speaker to state that he has evidence that is directly accessible to himself and that he alone has chosen to be evidence for him stating something, but does not know exactly about the assumption that he draws from the evidence. If he knows what he considers evidence, he should know what the chosen material is evidence for, otherwise he could not have chosen it from the material available as he only chooses it to base his statement on it. (89) can only be interpreted as asking for the addressee’s opinion on Peter’s whereabouts based on the evidence available in his room. That is, A is the source of the question, B, however, is made the source of the evidence by the questioner. When the addressee gives an answer, he becomes the anchor of the evidentiality. In (90), B
184
Sonja Mu¨ller
arrives at the conclusion that Peter must be in Greenland based on criteria that he considers decisive. (90) B: That’s clear: Er muss in Gro¨nland sein. He must in Greenland be ‘He must be in Greenland.’ See the tourist guide over there, a Greenlandic dictionary over there and all winter clothes are missing in his cupboard. That the assessment of the proposition’s truth is a matter of the speaker’s assessment (which gets shifted to the addressee in (89)) can be derived from the fact that the questioner could easily refute B’s answer by stating that the criteria which lead B to his statement do not convince him (cf. (91)). (91) A: Well, the guide and the dictionary do not convince me. He could just have gone to visit his parents. And he could have taken the books with him to study. Well, and that the winter clothes have gone is not really surprising. Look outside! Germany hasn’t seen such a winter for decades! In order for evidential mu¨ssen to sensibly occur in questions, the knowledge source has to be shifted away from the speaker – a measure which is not necessary with evidential sollen as the evidentiality’s anchoring happens more indirectly through a third person so that this source is automatically detached from the current speaker. Because of those characteristics that apply to mu¨ssen, this modal in (66b) only improves the extraction from (66a) because it is assigned to the individual that is not posing the question, i.e. because mu¨ssen is quoted by Paul who is asking the question and, thereby, assigned to Fritz. It is not Paul who is stating that he has evidence for assuming that Peter necessarily conceals a certain issue; it is Fritz who Paul addresses with his question. Schwager’s (2005) concept of epistemic authority which o¤ers an explanation for the incompatibility of question formation and imperative verbal mood thus also helps to account for the observation made in Section 4.1 that not all uses of modal verbs yield an increase in the acceptability of extractions with factive matrix verbs. It does not only account for the performatively used modals (cf. Section 4.2), which are also discussed by Schwager herself but also for certain evidential uses (cf. Section 4.3). The account sketched above ties in neatly with the explanation suggested for the negative influence that factive verbs have on extraction constructions
The distribution of knowledge in (un)acceptable questions
185
in Section 3.1.3: The interaction of di¤erent types of modal verbs and question formation can be considered a further case where the distribution of knowledge does not match the conditions as assumed for the illocutionary act question to be successfully, that is in Searl’s terms ‘‘nondefectively’’ (cf. Searl 1969: 54), performed. If the use of an imperative, performative modal or certain evidential modals presupposes epistemic authority on the part of the speaker, this situation is again not compatible with the first preparatory rule for the speech act question which states that the speaker does not know the answer (cf. Searle 1969: 66). Before looking at modal particles in Section 5 – yet another linguistic means that is used to express assumptions about the information states of discourse participants – a further issue concerning modal verbs in extraction constructions will be referred to in Section 4.4. As the extraction data presented in this subsection do not fall under the explanation suggested in this paper so far, it is to be seen as a sort of excursus o¤ering German data that (to my knowledge) has not been observed in the literature yet and that illustrates a further set of data on which modality has a positive impact. An explanatory account of the data, however, has to be left open for future research at this point. 4.4. Modal verbs in [þwh]-complement clauses In contrast to the extraction data looked at in this paper so far, namely extractions from that-complements, extractions from complements introduced by ob (whether) or a wh-pronoun always seem to result in a very low degree of acceptability in German (cf. (92), (93)).13 (92) a.
b.
(93) a.
*Wen weiß Peter, ob Maria liebt? Who knows Peter whether Mary loves ‘Peter knows whether Mary loves who.’ (intended reading) *Was fragt Fritz, ob Maria weggera¨umt hat? What asks Fritz whether Mary away.put has ‘Fritz asks whether Mary put away what?’ (intended reading) *Wen weiß Peter, wann Maria gesehen hat? Who knows Peter when Mary seen has ‘Peter knows when Mary saw who?’ (intended reading)
13. Cf. e.g. Rizzi (1990), Kluender (1991), Comorovski (1996), Abrusa´n (2008), Mu¨ller (2010) for accounts of extractions from [þwh]-complement clauses.
186
Sonja Mu¨ller
b.
*Wohin erza¨hlt Paul, warum Fritz gefahren ist? Whereto tells Paul why Fritz gone is ‘Where does Paul tell why Fritz went?’ (intended reading)
Whereas German seems to behave rather restrictively with respect to the possibility of extracting constituents from [þwh]-clauses, it has been claimed e.g. for English that this does present an option in the language. The data cited then, however, do not equal the translations of (92) and (93), but involve non-finite [þwh] þ to-complement clauses (cf. (94), (95)). (94) a. b.
?Which student did he wonder whether to consider intelligent? ?Who does Mary know whether to invite? Abrusa´n (2008: 152)
(95) a. b.
Which problem did John ask how to phrase? Which problem are you wondering how to phrase? Szabolcsi (2002: 14/31)
If parallel extractions take place from finite complements, the judgements are rather bad, similar to the German examples (cf. (96) (although Szabolcsi (2002) writes about (96) that the data do not receive uniform judgements among native speakers of English)). (96) a. b. c.
*About which topic did John ask who was talking? *?Which topic did John ask who was talking about? *How did John ask who behaved? Szabolcsi (2002: 3)
As embedded [þwh] þ to-infinitves do not present an option in German (cf. (97), (98)), extraction constructions which are parallel in all respects cannot be studied. (97) *Maria weiß, ob Paul einzuladen. Mary knows whether Paul to.invite ‘Mary knows whether to invite Paul.’ (98) *Hans hat gefragt, wie das Problem mit dem Wasserhahn Hans has asked how the problem with the water-tap zu formulieren. to phrase ‘Hans asked how to phrase the problem with the water-tap.’
The distribution of knowledge in (un)acceptable questions
187
Such data could look like (99) and (100), the reason for the low degree of acceptability, however, has to be assigned to an earlier level than the level on which a constituent is being asked for. If the underlying structure is not well-formed, the structure resulting from extracting a constituent cannot be well-formed either. (99) *Wen weiß Maria, ob einzuladen? Who knows Mary whether to.invite ‘Who does Mary know whether to invite?’ (100) *Welches Problem hat Hans gefragt, wie zu formulieren? Which problem has Hans asked how to phrase ‘Which problem did Hans ask how to phrase?’ Although the construction which improves the extraction from the [þwh]complement clause in English is not available in German, it can be shown that the extractions also improve in German if the structures receive the respective interpretation. This is where modality enters the picture: Bolinger (1978) and Bhatt (2000) assume that such non-finite complements are interpreted modally (cf. e.g. (101), (102) taken from Bhatt 2000: 1/133f.).14 (101) Tim knows how to solve the problem. QTim knows how one/he could/should solve the problem. 14. Bolinger (1978) argues that there is a minimal di¤erence between the nonfinite and the overtly modalised wh-complement in (i) and (ii), does not assume a di¤erence between (iii) and (iv) and refers to (v) and (vi) to show that there are cases in which the two complement clauses cannot be substituted for another. (i) (ii) (iii) (iv) (v) (vi)
She told me where I should go. She told me where to go. Stop arguing about what to do (about what you should do). She couldn’t think what to try next (what she should try next). You’ll never believe what you should do in a case like that. *You’ll never believe what to do in a case like that. Bolinger (1978: 117f.)
The examples in (i), (ii) and (iv), (v), therefore, illustrate that non-finite and finite modalised [þwh]-complements do not display identical interpretations. How exactly the meaning of a non-finite [þwh]-complement clause has to be determined is analysed in Bolinger (1978) and Bhatt (2000) and cannot be elaborated on at this point. The meanings attributed here are only meant to argue that the semantic factor modality plays a role.
188
Sonja Mu¨ller
(102) Magnus knows whether to invite Siggy to the party. QMagnus knows whether he should invite Siggy to the party. The German structures that correspond to the well-formed English examples from (94) and (95) look like the (b)-versions in (103) to (106) and not like the (a)-versions which in fact receive a very low degree of acceptability. (103) a.
*Welchen Studenten hat er sich gefragt, Which student has he self asked ob er fu¨r intelligent gehalten hat? whether he for intelligent considered has
(13) The man he-came. (he: Agreement marker) (14) The man he-came-he. (he: Agreement marker in Wackernagel clitic position as in the South German substandard; cf. for 2nd sg./pl. pronominals as in (15a,b) below) (15) a. b.
Habt¼s (es/ihr) was zum Essen? have¼you i.cl (you i.pl) something to eat Has¼t (du) was zum Essen? have¼you i.cl (you i.sg) something to eat
(12) represents the diachronic initial stage where the topic the man cooccurs with the subject pronoun he. Subsequent cliticizing weak he yields an agreement a‰x as in (13) and the corresponding clitic Wackernagel representations as in (15a,b). Faarlund (1992), on comparing old and modern states of Scandinavian languages, found that the subject status has continuously grown to gain dominance over earlier topics to finally insert the topic entirely in fronted position. Limiting topics to the clause initial position yields subjects. Leiss (2000a,b) corroborates this by finding the su‰x for definiteness on nominals only in words with positions that were not clause-fronted – i.e., the clause-initial position was the default position for nominal thematic status that needed no extra morphological representation. In other words, when eventually unmarked topic positions coincided with agreement marking, fixed subject status entered the grammatical picture in topicalized position. What do we expect a topic prominent language to do with properties that are characteristic of subject prominent languages? Consider (16), a typical criterion holding for single clause-based valence and its semantic relation. (16) Theta criterion: Each clausal argument must receive one and only one Theta role, and any single Theta role must be ascribed only once to a single-claused argument. The issue crucial for our discussion of Bu¨hler’s legacy in modern linguistics is this: While for a valence-based grammar as in subject and action prominent languages a non-argument cannot be in thematic topic position,
Traces of Bu¨hler’s semiotic legacy in modern linguistics
233
this appears to be the default case in languages where the predications get the semantics of states and properties – as in the East Asian languages among, prominently, Japanese. In the latter languages clause-fronted nominals may appear despite their non-argument status, crucially, however, not obliging the theta criterion. In other words, in statal and property prominent languages neither syntactic nor semantic valence is activated. Topics can be assigned due to contextual appropriateness. This default is suspended only with descriptions of actions. Cf. Fig. 15. Pattern
Explanation
Description of a property
A!B
A is somehow related to B, or A has in part a property of B.
Description of an action
f (x, y, z)
‘‘f ’’ is a function determining how x, y and z are to be replaced
Figure 16. Description of an action vs. that of a property (cf. Tanaka 2011)
The typological dichotomy between action and state descriptions and language types has a strong correspondence with Ikegami’s typological classification. According to Ikegami (1981) Japanese is a werden/becomelanguage where, by preference, a state of event is represented as an agentless occurrence. By contrast, German or English are tun/do-languages in which the perspective of the action is foregrounded. Inspired by Ikegami’s (1981) Hinds (1986) has characterized Japanese as a language focusing on the situation, whereas English is considered as focusing on the person focus. 7.2. Language typologies in terms of do vs. become Ikegami (1981) suggested a new typological classification: i.e., the division between tun-/do- and werden-/become-languages. In a do-language any state of a¤airs is taken to be an action or the result of an action by default. By contrast, in a werden/become-language such an action default is not presupposed necessarily, but, much rather, an event is taken to emerge out of itself. Needless to say that this is not taken to be a discrete division. Quite obviously, German and English would count as do-languages, as opposed to Japanese which is a become-language – one where the actor is not in the foreground of the event. In do-languages the inventory of transitive predicates is expected to be high, whereas in languages of the become-type predicates will be one-
234
Werner Abraham
place, or the actants are adverbs rather than arguments. The fact that the Theta-criterion has only limited e¤ect in Japanese would seem reasonable against this background as a become-language.7 Languages where states of a¤airs are taken to be properties rather than actions appear to suspend the Theta criterion. This is di¤erent in German and in English. German verbs are classified in terms of valence: Syntactic valence according to number and morphological case of their arguments; and semantic valence in terms of theta roles. The combinability of these valences yields a limited number of clause formation patterns. Thematic topics can only be arguments. In other words, any discourse-coherent status is established only by an element which is governed by the predicate and, consequently, has a status determined by the theta criterial limitations. 7.3. The Japanese evaluating sentence, han-bun dan, and Bu¨hler’s origo concept While Western (Indo-) European languages do not attach any particular origo-linked, modalized truth assessibility to propositions in the declarativeindicative mood, Japanese does so in the following sense. Declaratives in the indicative mood may, under certain discourse-linked conditions, bear a degree of subjective evaluation while, under a di¤erent discourse-linked marking, this evaluative subjectivity is missing. See the following minimal pair illustration taken from Tanaka (2008: 310). (17) a.
b.
Kai-*(wa)/ ?ga yakyu-wo shiteiru Kai-top baseball.acc do.progr . . . full predication; evaluative: han-bun dan ‘‘Kai is playing baseball’’ Kai-*(ga)/ ??wa yakyu-wo shiteiru Harappa Kai-nom baseball.acc do.progr field . . . DP with relative modification; descriptive: gensho-bun ‘‘the field on which Kai is playing baseball/baseball-played on by Kai’’
7. In a certain sense, one can think of ergative languages in terms of the werden/ become type. Notice that German werden is inchoative – the most fundamental premise for unaccusativity (Abraham 1995/ 2 2005).
Traces of Bu¨hler’s semiotic legacy in modern linguistics
235
The evaluative version goes with the discourse-topic particle wa. It represents a subjective judgment in contrast to sentences with any of the other possible particles, among which nominative ga creating a non-evaluative, objective description of assertive illocutionary status. The wa-version is the unmarked alternative in the sense that, first, the sentence without the discourse particle would be absolutely unacceptable and, second, the speaker’s absolute factivity judgment is made known to any addressee. It adds a more strongly verifiable asserting quality to the assertion than does its purely descriptive counterpart. It also yields a modal type of assertion stronger than in the Indo-European indicative assertion. Its function yielding this type of truth asserting modality is ‘‘to vouch for one’s assertion’’ or ‘‘to be absolutely trustworthy’’. It thus plays a strong interactive role involving speaker and the credulity at the hands of the addressee. The speaker takes responsibility for what he says, whereas he does not in the alternative case of assertion. This clearly has epistemic status in its own right. It resembles the functions of the modal particles, the epistemic modal verb, and other adverbs of modality in the Germanic languages, most prominently in German and Dutch (Abraham 2010). In terms of Bu¨hler’s origo scenario as well as Peirce’s and Jakobson’s shifter perspective toward deixis, and lastly in the sense of indexicality, epistemicity comes about through the lexical signal passing over into the class of deictic signals. As much as the epistemic reading of a modal verb, vis-a`vis its root meaning, is a transfer from non-deictic to deictic modality, the same semiotic-deictic, or indexical, change takes place between (17a) and (17b). Why is it just the Topic/Thema marker that obtains this deictically pointing function? Notice that Topics, in Japanese, do not create any sort of alignment toward the predicate (as distinct from what the nominativesubject marker does). That is, Topic draws the hearer’s/addressee’s attention in the sense that it is embedded in a running discourse, with the e¤ect that the hearer retains the topic-about information alongside the ensuing report by the speaker. In (18), for example, the raining event has two interpretations: the one with –wa locates the event with the speaker in the middle of the raining scene, while the one without the wa-constituent – i.e., without any topic-about – makes the presupposition that the addressee is aware of the rain as much as the speaker himself. No extra confirmation or highlighting is necessary. (18)–(19) are gleaned from Tanaka 2011.
236
Werner Abraham
(18) a.
kyou-wa ame-da today raining ‘‘Today it is raining’’ b. ame-da raining ‘‘It is raining’’
(19) a. b.
watashi-no-musume-wa otokonoko-desu my daughter is a boy otokonoko-desu
The selection of the topic is entirely at the speaker’s discretion such that an apparently non-sensical sentence as (19a) is acceptable as long as the speaker thinks that the addressee is able to reconstruct the relation intended (Tanaka 2008: 314f.). 8. Pronouns and reflexives It would not be unexpected to find out that pronouns and reflexives have a di¤erent categorial status in the two types of languages: deictic languages vs. anaphoric languages. We expect that deictic languages involve ontologicalsociological criteria to play a far greater role in their pronominal inventories. In what follows this issue is illustrated. 8.1. Reflexives in Japanese: Their non-anaphoric status The reflexive pronoun (‘anaphor’ in Generative Linguistics) in Japanese is not on a par with that in English and other European languages. Consider the case of Long-Distance binding of the Japanese ‘reflexive’ (e.g. Manzini & Wexler 1987, Cole, Hermon & Sung 1990). (20) Johni thinks that Billj hates himself*i/ j /himi/*j . jibun-woi/?j nikunde-iru-to omotte-iru (21) John-wai Bill-gaj John-TOP Bill-NOM REFL-ACC hate-INF-that believe-INF ‘‘John believes that Bill hates himself.’’ A ‘‘John believes that Bill hates him.’’ (20) is a classical example confirming the classical theory of binding according to which an anaphor must be bound within its binding domain. This relates the reflexive to Bill rather than to John. Personal pronouns work the other way around: They can only co-refer outside of the binding
Traces of Bu¨hler’s semiotic legacy in modern linguistics
237
domain. Thus, him instead of himself would only be co-referential with John. Japanese, however, presents a marked exception to Binding theory. Jibun, allegedly a reflexive in (18), refers to both John and Bill, quite in contrast to English as in (20). From among the numerous attempts to explain away this particular deviation from Binding theory, the best is probably to say that jibun is not a PersPro in the first place. Indeed, Sportiche (1986) adopted the position that jibun is not a pronominal, but an R-expression, i.e. a normal common noun in possession of its undelimited class reference. This corroborates the position defended here (Tanaka 2011). 8.2. ‘Personal pronouns’ in Japanese: Pronominal reference vs. full nominal reference Japanese ‘pronouns’ behave di¤erently from their English and German counterparts. Syntactically they are not like English pronouns, but they are nominals. Consider the lexicon of the ‘personal pronouns’. While European pronouns are limited to closed paradigms with but a few entries, Japanese semantic equivalents are open-classed. One and the same person may bear di¤erent denotations according to his/her semantic-ontological specification. (22)–(27) are due to Tanaka (2011). (22) watashi-wa shiri-mase-n (e.g., in a public situation) I-TOP know-HON-not (23) boku-wa shiranai-yo I-TOP know-not-MPart
(e.g., adult to adult friend)
(24) ore, shin-nai I know-NEG
(e.g., toward an aged friend)
(25) Papa shira-nai Papa know-NEG
(e.g., toward one’s own child)
(26) Jibun-wa shiri-mase-n (e.g., common soldier to o‰cer) Self-TOP know-HON-NEG (27) oira shira-nai I know-NEG
(e.g., jokingly toward a friend)
(22)–(27) all have the same propositional value: ‘‘I do not know.’’ However, with the situation changing di¤erent denotations for the speaker and, partially, also for the predicate will be employed. No doubt, the list of options is not unlimited, but it is certainly not exhausted by the alternates above. The syntactic di¤erences of the Japanese correspondences
238
Werner Abraham
and their uses mirror socio-psychological and honorific distinctions which must not be disregarded and which do not permit to flatten them out in terms of European pro-forms. Such distinctions cannot be abstracted away from in favor of the grammatical class of PersPro. They have to be retained in favor of the addressee’s socio-psychological status as well as the speaker’s knowledge that the addressee knows about the speaker’s awareness of this. In Japanese, all three grammatical persons are a¤ected by this strategy (Tanaka 2011). 9. Indexicals 9.1. Indexicals and deictics Just as ‘deictics’ in the sense of Bu¨hler, Tokieda, and Tanaka, also the modern concept of ‘indexicals’ such as I, here, now, today, he, she and that is taken to be linguistic expressions whose reference shifts from context to context (Forbes 1989, 2003; Richards 1992, 2003). Probably, plural pronominals such as we, ours, they, theirs, these, and those will have to be added (for a discussion of plural indexicals see Nunberg 1993). Following the modern language-philosophical tradition, one may say that speakers uttering a single sentence that contains an indexical may say di¤erent things dependent on the concrete situation or context in which the utterance is made. Kaplan (1989), whose work has dealt substantially with indexicals, maintains that indexicals have two sorts of meaning: one that is often called ‘character’ or ‘linguistic meaning’; and the second often called ‘content’. Using this terminology, one can say that the word I ‘‘ego’’ has a single character (or linguistic meaning), but has di¤erent contents in di¤erent contexts. One may see the following main reasons for being interested in indexicals. First, one may wish to describe their meanings and fit them into a more general theory of meaning. And second, reflection on indexicals may give us insight into matters such as the nature of belief, self-knowledge, firstperson perspective, and consciousness – and, derived from the latter, the theory of mind in L1. It becomes already obvious from what has been said above that this concept of ‘‘indexical’’ has a striking similarity with what Bu¨hler and his Japanese followers (of whom at least Tokieda was aware of Bu¨hler) discussed in so much detail and theoretical embedding. The Anglocentric discussion, in its entirety, has not taken notice of Bu¨hler’s work (in all likelihood due to the fact that Bu¨hler’s book of 1934 was in German).
Traces of Bu¨hler’s semiotic legacy in modern linguistics
239
Nor was there awareness of the Japanese authors we mentioned above. The weightiest discussion in Modern English linguistics and the philosophy of language is probably due to the American philosopher-logician David Kaplan – although the semiotic-philosophical ground as well as its terminology had been laid out by Peirce as early as 1857 (Peirce 1857/ 1982), whom, again, no one in the modern discussion appears to refer to. Only Jakobson, in his discussion of the notion shifter, mentions Peirce as well as Bu¨hler (Jakobson 1957/1971: 131f.). The indexicals that philosophers of language have studied most are (here according to Kaplan 1989; see also Giorgi 2010): (28) a. b. c.
Pronouns: I, he, she, this, that Adverbs: here, now, actually, presently, today, yesterday, tomorrow Adjectives: my, his, her, actual, present
An indexical’s referent and content are determined by its linguistic meaning (‘character’) and such contextual factors as the time, location, and intentions of the speaker. Indexicals are also commonly called contextsensitive expressions because their contents vary from context to context. The term ‘indexical’ is commonly restricted to simple expressions such as I and today, whereas the term ‘context-sensitive’ is also applied to complex expressions that contain simple indexicals, such as the man I love and I am female. The essential similarity of the concept of the ‘indexical’ with ‘deictics’ resides in the fact that, according to Kaplan, there are indexical and nonindexical uses of the pronouns – although Bu¨hler, Peirce, and Tokieda did not think of exactly this type of shifting. For example, he, his, she, and her may be used as bound variables in formal languages as in (29). (29) a. b.
Every mani believes that hei is smart. Every girlj loves herj father.
The occurrence of he in (29a) functions like a variable that is bound by the quantifier phrase every man. Similarly, her in (29b) is bound by ‘every girl’. If the same pronouns are not bound by prior quantifier phrases, or other linguistic expressions, they are also called unbound anaphora. This is the simple use of an anaphor which refers back to an antecedent DP, as in (30). It is concept C referred to as ‘‘unbound in its proper domain’’ in Chomsky’s Binding Theory (Chomsky 1982, 1981/1993).
240
Werner Abraham
(30) a. b.
Johnnyi hit a home run. Hei was very happy. [The woman who bought a newspaper] j was in a hurry. Shej forgot her change. Finally, there are the uses of these pronouns in which we shall be interested, the indexical (or demonstrative or deictic) uses, as in (31a,b). (31) a. Hei likes sardines [pointing at Fredi ], but hej does not [pointing at Barneyj ]. b. Hisk car is dirty [pointing at Alfredk], but hism car is clean [pointing at Alonzo m ]. 9.2. Pure indexicals and true demonstratives Kaplan (1989) draws a further distinction between two sorts of indexicals: i.e., pure indexicals, such as he, she, his, her, and that, and true demonstratives, such as I, today, tomorrow, actual, present, and here and now. The two types of indexicals di¤er in how their references and contents are determined. The reference and content of a true demonstrative in a given context/speech situation depends on the speaker’s pointing at something or somebody or his intentions to do so. For example, the reference and content of that in a context is determined (in part) by the speaker’s pointing gestures or by the speaker’s intention to refer to a particular object. The reference and content of a pure indexical in a context is not limited by such gesturing on the part of the speaker. For example, the reference of I in any context/speech situation is always the speaker, whether or not (s)he points at him/herself. The reference of tomorrow in a given context is always the day after the day of the context/speech situation. Thus, summarizing, the reference and context of a pure indexical is automatic, whereas the reference and content of a true demonstrative is not (cf. Perry 1977, 1979, 1997). 9.3. Multiple indexing Kaplan’s theory contains a fundamental distinction between the truth values of sentences as opposed to sentence contents, or propositions. The truth value of a sentence (as opposed to a proposition) depends on two parameters: context, or speech situation, and world. Kaplan’s own illustration is that ‘‘I am female’’ is uttered by Fred (a male) as against Wilma (a female). The content of the sentence ‘I am female’ in context c is the proposition that Fred is female, and this proposition is false with respect to w (the real world in which context c is set). It follows that the sentence is false with respect to both c and w. By contrast, the content of this sentence at context c* is the proposition that Wilma is female, and this
Traces of Bu¨hler’s semiotic legacy in modern linguistics
241
proposition is true in world w. Thus, the sentence ‘I am female’ is true with respect to c* and w. Notice that the reference to the real world is the same both times, but the contexts are di¤erent. Thus, the sentence’s truth value is doubly-related to both context and world. This sort of doublerelativization has been called double-indexing (cf. Vlach 1973, Kamp 1971 for indexicals in the sense of double-indexing). At first sight, this distinction has resemblance with both Jakobson’s concept of shifter (Jakobson 19(77/)85) and Bu¨hler’s (multiple) deixis. However, it is not quite clear how both context and world reference can be linked to speaker, observer, and origo (speaker as oriented in the configuration of here and now) – to which also Jakobson refers. Furthermore, ‘context’ needs to be specified in terms of speech act situation and whether the speaker is part of the speech act situation or whether he acts from outside the speech act situation as an observer. Jakobson’s term shifting refers exactly to this transience between ‘inside and outside of the speech act situation’ (Bu¨hler’s origo). See most recently Wechsler (2010).
10. Comparison 1. Ad ‘‘linguistic expressions whose reference shifts from context to context’’: Kaplan and his followers have not taken the trouble to distinguish such ‘contexts’ in the sense that Bu¨hler, Peirce, Jakobson, Tokieda, Nakamura, and Tanaka have: namely the origo vs. non-origo context distinction. 2. Although the term double indexing seems close to Jakobson’s shifting between ‘inside and outside of the origo’, the fact that it does not include the origo (cognitive anchoring in the here and now) concept makes the comparison weak and results in the advantage of Bu¨hler’s understanding. 3. While the deictic-anaphoric distinction, which is based on the concept of Bu¨hler’s origo, has been pursued for deep typological distinctions (Tokieda, Nakamura, Tanaka; others, too?), the modern theories of indexicals have not had such linguistic consequences. This is no doubt due to the fact that Bu¨hler’s (as well as Peirce’s and Jakobson’s understanding of the) concept of the origo has not found any fruitful reception in modern Anglophile linguistic and language-philosophical discussions. 4. Contrary to occasional assumptions (such as by Schi¤er 1978, 1981; Peacocke 1992) avoids, Bu¨hler’s origo concept does not seem to have direct pertinence in Frege’s distinction of referent and sense (Frege 1892, 1984).
242
Werner Abraham
5. The entire discussion of speech act theory (beginning in 1963 up to Searle 1969) embodies no follow-up to Peirce or Bu¨hler in the sense of the present discussion and in the sense of a solid extension of its fundamental ideas. 6. In particular, no profoundly interesting conclusions have been drawn in the discussion of indexicals with respect to questions of the typological distinction between discourse prominence in, or of, a language vs. grammatical parts-of-speech prominence. This distinction is based on, and implied by, the distinction between anaphoricity as a structural solution to recurrence, on the one hand, and determiner deixis, on the other. See (1)–(3) above. 7. There appears to be a direct correspondence between Kaplan’s automatic (or pure) indexicals set in a context unlimited by gesturing or pointing on the part of the speaker (such as here and now) and Bu¨hler’s origo-immanent deixis, on the one hand, and, on the other hand, that which Kaplan has called true demonstratives, such as I, today, tomorrow, actual, present, and here and now. In the latter case no automatism of the true reference is presupposed, but, much rather, the speaker’s pointing at something or somebody is required for the unambiguous resolution of reference. 8. Kaplan’s link to a world of reference makes us think of the theory of veridicality (or of the reference of possible worlds) – i.e., the worlds of belief and individual reasoning as factors of assessment in the sense of truth assessibility (see, for example, Giannakidou 1998, 2001). That is, it leads us into theory of mind domains which play a role not only in the cognitive and linguistic process of maturation, but also into central domains of modality (Abraham & Leiss 2008, 2009). 11. Conclusion Our preface serves the purpose to show that, even if Bu¨hler’s notions of the Speech act Origo and Organon did not leave a direct legacy in modern linguistic discussion, they have at least been taken up independently. This would seem su‰cient to give due respect to Bu¨hler’s early notions and his lines of argument to peel out these notions. We have witnessed that the grammar of deixis may occupy certain components of a language that distributes its grammar predominantly on the symbolic field (German article and demonstrative deictics distributed on discourse criteria such as thema and rhema). Our main focus, however, lay on the typological division of languages. Japanese as well as Thai were shown to centrally embody deictic
Traces of Bu¨hler’s semiotic legacy in modern linguistics
243
components in their grammars – thus laying di¤erent distributional weight on the relation between lexicon and grammar. More generally, and beyond pronouns and reflexives, we touched upon the implications of the two strategies of reference, Anaphoricity and Deixis, i.e. on the canon of clausal building structure. In the anaphoric mode, which encompasses three fundamental interactors – Speaker, Addressee, and Object of reference – form a three-poled structure. An event or state of a¤airs is prototypically conceived of as employing three indispensable actants, namely Agent, Recipient, and Patient. In this structural scenario, the Agent in Subject function and the predicate occupies the central clausestructuring role. The Subject-Predicate-system, typically valid in many European languages and characterized crucially by verbal agreement and conjugation, is the default choice of the anaphoric language type. By contrast to the three-poled anaphoric system, a two-poled deictic structure mirrors the Observer (simultaneously the Speaker) an event or state of a¤airs in its direct speech act context. The actants of this mode are not subject to a syntactic filter as the Theta-Criterion. The interplay of these predicational actants is licensed by the respective context, in which the Topic-about (Thema) is centrally commented on. The TopicComment structure characterizes both Japanese and German as a bipolar, solidly grammaticalized structure. In Japanese (and, to the best of our typological insight, Thai) as a deictic language, the topic-comment structure is a canonical clause-building principle. German, by contrast, embodies either structural system: anaphoricity/structural parts-of-speech prominence and deixis/structural topic prominence. Any of the distinctions leading to the linguistic and typological diversities and their fundamental components that we have pointed out are due to Bu¨hler’s profound and original insights.8 8. The most recent mention of Bu¨hler’s work I have come across is in Levinson 2003. Bu¨hler’s Origo and Organon concepts are encapsulated in Levinson’s three frames of reference: the intrinsic frame illustrated by the description ‘The ball is in front of me’ (deictic centre); the absolute frame as in ‘The ball is north of me’; and the relative frame as in ‘The ball is in front of the tree (from where I am standing)’. It is noteworthy that the author emphasizes the fact that none of the three frames need have a deictic centre: Thus, in the intrinsic frame one can simply say ‘in front of (Levinson 2003: 50). Clearly, Levinson’s work is inspired by both Gestalt psychology (of the nineteen-twenties) and linguistic typology and the task of describing and characterizing languages (such as Guugu Yimithirr and Wik Mungan) that are very exotic as compared to Indo-European languages. Levinson refers to Hockett’s (1960) design feature of displacement as well as Bu¨hler’s (1982[1934]) concept of transposed deictic centre. He makes no mention of Peirce und Jakobson.
244
Werner Abraham
References (the Japanese literature listed below has been cited secondarily after Tanaka 2011) Abraham, Werner 1997 The base structure of the German clause under discourse functional weight: contentful functional categories vs. derivative functional categories. In: W. Abraham & E. v. Gelderen (eds.) German: syntactic problems – problematic syntax, 11–42. [Linguistische Arbeiten 374]. Tu¨bingen: M. Niemeyer. Abraham Werner 22005 Deutsche Syntax im Sprachenvergleich. Grundlegung einer typologischen Syntax des Deutschen. 2., verbesserte und erweiterte Auflage. [Studien zur deutschen Grammatik 41]. Tu¨bingen: Stauffenburg. Abraham, Werner 2003 Pronomina im Diskurs: deutsche Personal- and Demonstrativpronomina unter ,,Zentrierungsperspektive‘‘. Grammatische ¨ berlegungen zu einer Teiltheorie der Textkoha¨renz. SprachwisU senschaft 27/4: 447–491. Abraham, Werner 2006 The discourse-functional crystallization of DP from the original demonstrative. In: E. Stark, E. Leiss & W. Abraham (eds.) (2006): Nominal determination. Typology, context constraints, and historical emergence, 241–256. [Studies in Language Companion Series 89]. Amsterdam: John Benjamins. Abraham, Werner 2010 ToM or Foreign Consciousness Alignment strategies. Workshop ms. For the SLE Annual Conference Lisbon 2009. Abraham, Werner & Elisabeth Leiss (eds.) 2008 Modality-aspect interfaces. Implications and typological solutions. [Typological Studies in Languages 79]. Amsterdam/Philadelphia: John Benjamins. Abraham, Werner & Elisabeth Leiss (eds.) 2009 Modalita¨t. Epistemik and Evidentialita¨t bei Modalverb, Adverb, Modalpartikel und Modus. [Studien zur deutschen Grammatik 77]. Tu¨bingen: Stau¤enburg Verlag. Almog, Joseph; John Perry & Howard Wettstein (eds.) 1989 Themes from Kaplan. Oxford: OUP. Austin, John L. 1963 How to do things with words. Oxford: Clarendon Press. Baker, N. & Greenfield, P. 1988 The developement of new and old information in young children’s early language. Language Sciences 10/1: 3–34.
Traces of Bu¨hler’s semiotic legacy in modern linguistics
245
Bar-Hille#l, Yehoˆsˇua 1964 Language and information. Selected essays on their theory and application. Reading, Mass.: Addison-Wesley. Bates, E. & B. MacWhinney 1979 A Functionalist approach to the acquisition of grammar. In: E. Ochs & B.B. Schie¤elin (eds.) Developmental pragmatics. New York: Academic Press. Benveniste, Emile 1966/1977 Proble`mes de linguistique ge´ne´rale. Paris: Gallimard. Bu¨hler, Karl 1934/1982 Sprachtheorie. Die Darstellungsfunktion der Sprache. [Ullstein Taschenbuch 1159]. Stuttgart: Gustav Fischer. Bu¨hler, Karl 1990 Theory of language. The representational function of language. Transl. D.F. Goodwin. Amsterdam: John Benjamins. Chafe, Wallace L. 1996 How consciousness shapes language. Pragmatics and Cognition 4: 35–54. Chomsky, Noam 1981/1993 Lectures on government and binding: The Pisa Lectures. Berlin: Mouton de Gruyter. Chomsky, Noam 1982 Some concepts and consequences of the theory of Government and Binding. Linguistic Inquiry Monograph 6. Cambridge, Mass.: MIT Press. Cole, Peter; Gabriela Hermon & Li-May Sung 1990 Principles and parameters of long-distance-reflexives. Linguistic Inquiry 21/1: 1–22. Coseriu, Eugenio 1981 Textlinguistik. Eine Einfu¨hrung. Tu¨bingen: Narr. ¨ sten Dahl, O 2004 The growth and maintenance of linguistic complexitiy. [Studies in Language Companion Series 71]. Amsterdam: John Benjamins. Diewald, Gabriele M. 1991 Deixis and Textsorten im Germanen. [Reihe Germanistische Linguistik 32]. Tu¨bingen: Niemeyer. Ehlich, Konrad 1982 Anaphora and deixis: same, similar, or di¤erent? In: Jarvella, R. J. et al. (eds.) Speech, place, and action. Chichester: John Wiley & Sons. 315–338. Ehlich, Konrad 1983 Deixis und Anapher. In: G. Rauh (ed.), Essays on deixis, 79–97. Tu¨bingen: Narr. Faarlund, Jan Terje 1992 The subject as a thematic category in the history of Scandinavian. Folia Linguistica XXXVI/1–2: 151–159.
246
Werner Abraham
Forbes, Graeme 1989 Indexicals. In D. Gabbay and F. Guenther (Eds.), Handbook of Philosophical Logic, Volume IV, 463–490. Dordrecht: Reidel. Forbes, Graeme 2003 Indexicals. In: D. Gabbay & F. Guenther (eds.) Handbook of Philosophical Logic., 2nd Edition, Volume 10, 101–134. Dordrecht: Kluwer. Revised version. Frege, Gottlob ¨ ber Sinn und Bedeutung. Zeitschrift fu¨r Philosophie und philos1892 U ophische Kritik 100. Translation by Herbert Feigl as ‘‘Sense and Nominatum’’ in Herbert Feigl & Wilfrid Sellars (eds.) 1949. Readings in Philosophical Analysis, 85–102. New York: AppletonCentury-Crofts. Reprinted in Martinich 2001, 199–211. Also translated by Peter Geach and Max Black as ‘‘Sense and Reference’’ in Peter Geach & Max Black (eds.) (1952): Translations from the philosophical writing of Gottlob Frege, 56–78. Oxford: Blackwell. Frege, Gottlob 1984 Thoughts. In: Frege (ed. B. McGuiness, trans. P. Geach and R.H. Stootho¤ ) Collected Papers on Mathematics, Logic, and Philosophy, 351–372. Oxford: Blackwell. Giannakidou, Anastasia 1998 Polarity sensitivity as (non)veridical dependency. [Linguistik Aktuell / Linguistics Today 23]. Amsterdam: John Benjamins. Giannakidou, Anastasia 2001 The meaning of free choice. Linguistics and Philosophy 24: 367– 421. Gil, David 2008 How complex are isolating languages? In: M. Miestamo, K. Sinnema¨ki & F. Karlson (eds.), 109–131. Giorgi, Alessandra 2010 About the speaker. Towards a syntax of indexicality. Oxford: OUP. Givo´n, Talmy 1976 Topic, pronoun, and grammatical agreement. In: Li, Ch. N. (ed.): 25–56. Hauser-Gru¨dl, Nora 2010 Topicality in L1-acquisition. A contrastive analysis of null subject expressions in child French and German. In: C. Breul, & E. Go¨bbel (eds.) Comparative and contrastive studies of information structure. Amsterdam: John Benjamins. Hinds, John 1986 Situation vs. person focus. Tokyo: Kuroshio. Hinds, John 1987 Reader versus writer responsibility: A new typology. In: U. Connor & R.B. Kaplan (eds.) Writing across languages: analysis of L2 texts, 141–152. Reading, Mass: Addison-Wesley.
Traces of Bu¨hler’s semiotic legacy in modern linguistics
247
Hockett, Charles Francis 1959 A course in modern linguistics. New York: Macmillan. Ho¨hle, Tilmann N. 1988 VERUM-Fokus. Sprache und Pragmatik Arbeitsberichte 5. 1–7. Ikegami, Yoshihiko 1981 suru-to-naru-no-Gengogaku. [‘DO-language’ and. ‘BE-COMElanguage’: two contrasting types of linguistic representation]. Tokyo: Taishukan. Iwasaki, Shoichi 1993 Subjectivity in grammar and discourse. Theoretical considerations and a case study of Japanese spoken discourse. Amsterdam: John Benjamins. Jakobson, Roman 1957/1971 Shifters, verbal categories, and the Russian verb. In: Jakobson, Roman: Selected writings 2: 130–147. Jakobson, Roman 1985 Selected writings. Vol. 7: Contributions to comparative mythology. Studies in linguistics and philologty, 1972–1982. Berlin: Mouton. Jakobson, Roman 1971 Selected writings. Vol. 2: Word and language, 254–264. The Hague: Mouton. Jakobson, Roman 1985 Selected writings. Vol. 7: Contributions to comparative mythology. Studies in linguistics and philologty, 1972–1982. Berlin: Mouton. Kamp, Hans 1971 Formal properties of ‘now’. Theoria 37: 227–273. Kaplan, David 1989 Demonstratives. In: Almog, Perry & Wettstein (eds.) 1989: 481– 563. Kuroda, Shigeyuki 1973 Where epistemology, style and grammar meet. In: S. R. Anderson & P. Kiparsky (eds.) Festschrift for Morris Halle, 377–391. New York: Holt, Rinehart and Winston. Langacker, Ronald W. 1985 Observations and speculations on subjectivity. In: J. Haiman (ed.) Iconicity in syntax, 109–150. Amsterdam: John Benjamins. Langacker, Ronald W. 1990 Subjectification. Cognitive Linguistics 1: 5–38. Leiss, Elisabeth 2000a Artikel and Aspekt. Die grammatischen Muster von Definitheit. [Studia Linguistica Germania 55]. Berlin: de Gruyter. Leiss, Elisabeth 2000b Gender in Old High German. In: Unterbeck, B. & Rissanen, M. (eds.), Gender in grammar and cognition, 237–258. Berlin: de Gruyter.
248
Werner Abraham
Leiss, Elisabeth 2000c Verbalaspekt and die Herausbildung epistemischer Modalverben. Germanistische Linguistik 154: 63–83. Leiss, Elisabeth 2009 Sprachphilosophie. [De Gruyter Studienbuch]. Berlin: de Gruyter. Levinson, Stephen 2003 Space in language and cognition. Explorations in cognitive diversity. Cambridge: CUP. Li, Charles N. 1976 Subject and topic. New York: Academic Press. Li, Charles N. & Sandra A. Thompson 1976 Subject and topic: a new typology of language. In: Ch. N. Li (eds.): 457–490. Lyons, John 1982 Deixis and subjectivity: loquor, ergo sum? In: R. J. Jarvella & W. Klein (eds.) Speech place, and action. Studies in deixis and related topics, 101–124. New York: John Wiley & Sons. Manzini, M. Rita & Kenneth Wexler 1987 Parameters, Binding Theory, and learnability. Linguistic Inquiry 18/3: 413–444. Marmaridou, Sophia S. A. 2000 Pragmatic meaning and cognition. Amsterdam: John Benjamins [Pragmatics & Beyond: New series 72]. Martinich, A.P. (ed.) 2001 Philosophy of language. 4th Edition. Oxford: OUP. Miestamo, Matti, Kaius Sinnema¨ki & Fred Karlson (eds.) 2008 Language complexity. Typology, contact, change. [Studies in Language Companion Series 71]. Amsterdam. Benjamins. Nakamura, Yoshihisa 2004 Shukansei-no-gengogaku – Shukansei to bunpoukouzou/-koubun. [Linguistics of subjectivity – subjectivity and grammatical structure and construction]. In: Nakamura, Y. (ed.) Kognitive Grammatik II, 3–51. Tokyo: Taishukan. Narrog, Heiko 2009 Modality in Japanese – the layered structure of the clause and hierarchies of functional categories. [Studies in Language Companion Series 109]. Amsterdam: John Benjamins. Nunberg, Geo¤rey 1993 Indexicality and deixis. Linguistics and Philosophy, 16: 1–43. Ogawa, Akio 2005 Meteo-Pra¨dikate im Sprachvergleich. Neue Beitra¨ge zur Germanistik 4/2: 92–103. Peacocke, Christopher 1992 A study of concepts. Cambridge, MA: MIT Press.
Traces of Bu¨hler’s semiotic legacy in modern linguistics
249
Peirce, Charles S. 1982¤. Writings of Charles S. Peirce. A chronological edition. Vol. 1–6. Bloomington: Indiana University Press. Vol. 1, 1982 Writings 1857–1866. Vol. 2, 1984 Writings 1867–1871. Vol. 3, 1986 Writings 1872–1878. Vol. 4, 1986a Writings 1879–1884. Vol. 5, 1993 Writings 1884–1886. Vol. 6, 2000 Writings 1886–1890. Peirce, Charles S. 1871/1984 The Berkeley review: Fraser’s The work of George Berkeley. In: Peirce 1982¤: Vol. 2 (1984), 461–487. Perry, John 1977 Frege on demonstratives. Philosophical Review 86: 474–97. Reprinted in Perry (1993). Perry, John 1979 The problem of the essential indexical. Nouˆs 13: 3–21. Reprinted in Martinich (2001) and Perry (1993). Perry, John 1997 Indexicals and demonstratives. In: Bob Hale & Crispin Wright (eds.) A companion to philosophy of language, 586–612. Oxford: Blackwell. Richard, Mark 1992 Indexicals. In: William Bright (ed.), International encyclopedia of linguistics, 200–202. Oxford: OUP. Richard, Mark 2003 Indexicals. In: William Bright (ed.) International encyclopedia of linguistics. 2nd edition. Oxford: OUP. Revised version. Rizzi, Luigi (ed.) 2004 The structure of CP and IP. The cartography of syntactic structures. Volume 2. Oxford: OUP. Schi¤er, Stephen 1978 The basis of reference. Erkenntnis 13: 171–206. Schi¤er, Stephen 1981 Indexicals and the theory of reference. Synthese 49: 43–100. Schleicher, August 1863 Die Darwinsche Theorie und die Sprachwissenschaft. O¤enes Sendschreiben an Dr. Ernst Ha¨ckel. Weimar: Bo¨hlau. Schleicher, August ¨ ber die Bedeutung der Sprache fu¨r die Naturgesetze des Men1865 U schen. Weimar: Bo¨hlau. Searle, John R. 1969 Speech acts. An essay in the philosophy of language. Cambridge: CUP.
250
Werner Abraham
Shannon, Claude Elwood & Warren Weaver 1949 The mathematical theory of communication. Urbana, Illinois: Univ. of Illinois Press. Sperber, Dan & Deirdre Wilson 1986 Relevance. Communication and cognition. London: Blackwell. Sportiche, Dominique 1986 Zˇibun. Linguistic Inquiry 17: 369–374. Tanaka, Shin 1992 Fokussierung und Topik im Deutschen. Symposion 7: 53–64. Tanaka, Shin 2003 Topik-Merge und Topikprogression. Papers on Languages and Cultures 12. 137–150. Tanaka, Shin 2008 The aspect-modality link in Japanese: the case of the evaluating sentence. In: W. Abraham & E. Leiss, 309–330. Tanaka, Shin 2011 Deixis and Anaphorik: Referenzstrategien in Text, Satz and Wort. Eine typologisch-kontrastive Untersuchung Deutsch – Japanisch. Berlin: de Gruyter. Tokieda, Motoki 1941 Kokugogaku-Genron. Gengo-Katei-Setsu-no-Seiritsu to so-noTenkai. [Prinzip zur Forschung der Nationalsprache. Entstehung und Entwicklung der Theorie des Sprachprozesses]. Tokyo: Iwanami. Tokieda, Motoki 1950 Nihon-Bunpo. Kougo-hen. [Japanese grammar. Spoken language]. Tokyo: Iwanami. Tokieda, Motoki 1954 Nihon-Bunpo. Bungo-hen. [Japanese grammar. Written language]. Tokyo: Iwanami. Vlach, Frank 1973 ‘Now’ and ‘then’: A formal study in the logic of tense and anaphora. Ph.D. Dissertation, Philosophy Department, UCLA. Wechsler, Stephen 2010 What ,you‘ and ,I‘ mean to each other: person indexicals, selfascription, and theory of mind. Language 86/2: 332–365. Weinrich, Harald 21971 Tempus. Besprochene und erza¨hlte Welt. Stuttgart: Kohlhammer. Wierzbicka, Anna 1988 The semantics of grammar. Amsterdam: John Benjamins. Winter, Werner 1961 Relative Ha¨ufigkeit syntaktischer Erscheinungen als Mittel zur Abgrenzung von Stilarten. Phonetica 7. 193–216.
Part II.
Instances of deixis and origo in sundry languages
Modal particles, speaker-hearer links, and illocutionary force Marco Coniglio Humboldt-Universita¨t Berlin
Abstract In this paper, I will show that the syntactic category of modal particles in German pertains both to the notion of modality and to that of illocution. Since particles are modal words with peculiar properties, such as their speaker deixis and their speaker-hearer link, they do not belong to the propositional part of the utterance, but rather to its illocution. I will prove their link to the illocutionary force by considering their syntactic distribution in subordinate clauses. I will claim that modal particles are licensed only if the utterance in which they occur is an autonomous speech act and, hence, that they are to be considered markers of the illocutionary force of a clause (cf. Doherty 1987, Jacobs 1986, 1991, Thurmair 1989). 1. Introduction Modal particles (MPs, henceforth) are a well studied class of words in German, which mainly occur in spoken language. Examples include denn, eben, ja, wohl, etc. The speaker uses them for di¤erent purposes. In general, by means of a MP, she aims at expressing her mental attitude, opinion, personal belief, interest, etc. with respect to the propositional content of her utterance. Cf. Abraham (1995, 2008a), Borst (1985), Egg (this volume), Meibauer (1994), Ormelius-Sandblom (1997a, 1997b), Thurmair (1989), etc. about the centrality of the speaker’s role in the definition of the class of MPs. In particular, by using a MP, the speaker reports her own knowledge state and in general what she surmises is the knowledge state of the hearer. Exactly the possibility for the speaker to attribute a certain knowledge awareness to others renders MPs relevant from the cognitive point of view. In fact, MPs play an important role in the communicative interaction between speaker and hearer, as was underlined by Abraham (2008a: 261): ‘‘MP tasten im Sinne einer Theory of Mind die ho¨rergebundenen
254
Marco Coniglio
Erfahrungs- und Wissensvoraussetzungen ab und regen damit zu Korrektur bzw. Besta¨tigung durch den Adressaten an.’’ [MPs explore the hearerbound experience and knowledge conditions in the sense of a Theory of Mind and thereby encourage the addressee to correct or to confirm them. MC]. By using a particle, the speaker introduces her representation of the world in the discourse and makes predictions about the hearer’s knowledge, belief or attitude. The hearer can confirm or reject this representation, according to her own knowledge state. In order to illustrate to what extent the role of MPs can be relevant in cognitive processes, let us consider the following simple example: (1) Italiener sind ja sowieso alle doof. Italians are PRT anyway all foolish ‘Italians are anyway all foolish, as you know.’ As observed for instance by Thurmair (1989: 104–109), the function of (unstressed) ja is in general that of signaling that the speaker considers a certain fact to be known and evident to the hearer (and to others). In this example, the speaker is expressing not only the fact that she considers all Italians to be foolish people, but she also wants to convey additional information, i.e. that she considers this fact to be evident and therefore accessible to the hearer, thus negotiating common ground information and shared knowledge with her (also cf. Leiss 2008b). Therefore, the particle has the function both of expressing the speaker’s personal knowledge or conviction about a certain fact and of attributing a certain knowledge state or conviction to the hearer. If we replace the particle ja in example (1) with the particle doch: (2) Italiener sind doch sowieso alle doof. Italians are PRT anyway all foolish ‘Italians are anyway all foolish (although you think/say the contrary).’ We observe an interesting fact. What changes here is not the propositional part of the clause, i.e. the fact that, according to the speaker, Italians are all foolish, but what the speaker assumes is the hearer’s opinion or belief. In this case, the speaker wants to contradict the hearer, since she assumes that the latter entertains another opinion (cf. for example Thurmair 1989: 110–119).
Modal particles, speaker-hearer links, and illocutionary force
255
From a cognitive point of view, MPs (like modal verbs) are therefore highly informative, namely in at least a threefold way: They provide information 1) about the source of the information, which has to do with the propositional part of the utterance; 2) about the speaker’s assessment of this information, which is relevant from the illocutionary point of view; and 3) about the speaker’s estimate of the hearer’s state of knowledge.1 MPs di¤er from modal adverbs, since they have access to the illocutionary force of the clause (cf. Doherty 1987, Jacobs 1986, 1991, Thurmair 1989, Abraham 1995, 2009). They interact with the illocution and must be anchored to a speaker as the ‘‘author’’ of a certain speech act and as the possessor of a certain mental representation. Consequently, from a syntactic perspective, MPs must be considered as crucial criteria for defining which types of clauses are endowed with independent illocutionary force (cf. Thurmair 1989: 73–82). A clause which contains a MP must be necessarily anchored to a speaker, i.e. to the bearer of the illocution. Hence, I will argue that such a clause must have a projection ForceP, where the speaker is encoded. This is certainly true for main clauses, since they have illocutionary force per definition. However, starting from certain assumptions in the literature that some types of subordinate clauses can also be considered root clauses, I will provide evidence for the independent illocutionary status of subordinate clauses that (can) license MPs. In the following sections, I will first describe the main properties of German MPs, how they di¤er from adverbs or other modal words and how they interact with the speaker and the illocutionary force of an utterance. In section 3, I will present Haegeman’s (2002, 2004a, 2004b, 2006) hypotheses about the di¤erent syntactic status of root and embedded clauses. In particular, I will discuss the fundamental role played by illocutionary force in the definition of these two categories of clauses. In 4, we will see how MPs can implement this theory and how they are linked to the speaker in di¤erent contexts. A theoretical analysis of the syntactic interaction between MPs and ForceP will be provided in section 5.
1. I am following Abraham (2008a, 2008b) and Leiss (2008a, 2008b), who – building on Diewald’s (1999) concept of double deixis – consider MPs modal words displaying a triple deixis. See section 2.1.
256
Marco Coniglio
2. Modal particles between modality and illocution 2.1. Including modal particles in the notion of modality MPs are considered adverb-like elements.2 In contrast with adverbs3 (and with other particles), MPs usually have a more abstract meaning and their translation into other languages can often be very di‰cult. If we consider the case of ja in sentence (1) again, we see that the meaning and function of the particle cannot be brought back to ‘yes’, which is its meaning in other contexts. The meanings and functions of MPs can be captured if we consider them to be modal expressions. Conceptually, this can be attained by extending the notion of modality to ‘‘Ausdru¨cke der Sprechereinstellung’’ [expressions of the speaker’s attitude], as proposed by Meibauer (1994: 2. Provided that we assume that the class of adverbs also comprises all the di¤erent classes of polyfunctional particles that have been highlighted by German grammatical tradition, i.e. Gradpartikeln, Steigerungspartikeln and so on. MPs are often homophonous with such adverbial elements (cf. Thurmair 1989). For example, besides its function as a MP (ii), auch can also be used as a Gradpartikel (i). The following examples are taken from Thurmair (1989: 155): (i) Peter: Russisch ist ganz scho¨n schwer. Max: Deutsch ist AUCH nicht einfach. [Peter: Russian is pretty di‰cult. Max: German is not easy either. MC] (ii) Ali: Ich hab von dem Text nicht alles verstanden. Max: Naja, Deutsch ist auch nicht einfach. [Ali: I did not understand the whole text. Max: Well, German is not that easy. MC] 3. MPs have often been described as ‘‘weak’’ adverbs in the sense of the tripartition proposed for pronouns by Cardinaletti and Starke (1999) (see Coniglio 2005, Grosz 2005, Cardinaletti 2007). MPs are adverbial elements that do not project the whole syntactic structure as full-fledged adverbs do. This would explain not only MPs’ typical semantic ‘‘bleaching’’ and their peculiar properties as modifiers of illocutionary force (which I will discuss in the next section), but also the fact that, like other ‘‘weak’’ elements, they can be neither coordinated (ia) nor modified (ib), nor used as a reply to a question (ic), etc.: (i) a. Geh doch (*und ) mal zur Schule! go PRT and PRT to school b. *Geh sehr mal zur Schule! go very PRT to school c. Speaker A: (Wie) kann ich zur Schule gehen? – Speaker B: *Mal ! how can I to school go PRT
Modal particles, speaker-hearer links, and illocutionary force
257
11–15).4 Their function of expressing the speaker’s attitude plays a fundamental role in the definition of the modal nature of MPs. But also their function as means for reporting the speaker’s estimate about the hearer’s knowledge awareness is relevant for capturing the modal dimension of particles (see Abraham 2008a, 2008b). According to Hentschel (1986: 120), MPs ‘‘[. . .] bezeichnen keine Objekte der außersprachlichen Wirklichkeit, sondern dru¨cken Relationen aus, die zwischen den bezeichneten Objekten oder Sachverhalten bzw. zwischen diesen und dem Sprecher und/oder Ho¨rer bestehen.’’ [do not describe objects of the extralinguistic reality, but express relations between the described objects or facts or between these and the speaker and/or the hearer. MC]. The speaker uses a particle to attribute a particular attitude to herself and to others (also see Coniglio 2011: 14–19). Moreover, she encourages the hearer to express her own point of view by correcting or confirming the speaker’s assumption (cf. Abraham 2008b). When producing a certain utterance, the speaker does not limit herself to describing a fact or to expressing her opinion about it. She also makes an estimate about the hearer’s state of knowledge. Based on Sperber and Wilson (1986: 4–5), Abraham (2009: 2) claims in fact that [t]houghts do not simply travel from Speaker to Addressee [. . .]. We don’t send out communications to addressees when we are engaged in linguistic intercourse. What we do, instead and much rather, is that we try to make sure what exists, on the part of the addressee, in terms of shared knowledge relevant to the present linguistic interaction and in terms of what is in need of correction on either part.
The interaction between speaker and addressee is therefore crucial to understand the specificity of MPs in the modal Urmasse ‘primordial mass’ (Abraham 2008a). In the framework of Sperber and Wilson’s (1986) relevance theory, Abraham (2008a, 2008b, 2009) and Leiss (2008b) show that MPs di¤er from other modal words (such as modal verbs and adverbials) in a significant way. Let us first consider epistemic modal verbs. Abraham (2008a, 2008b, 2009) and Leiss (2008a) point out that such verbs display a double deixis (in the sense of Diewald 1999, referring to Jakobson 1971). Abraham (2009: 9) discusses the following case: 4. Meibauer (1994) follows Kiefer (1987), who in turn builds on Bierwisch (1980) and Kratzer (1981).
258
Marco Coniglio
(3) a.
b.
epistemic modal verbs/EMV denote a twofold deixis: according to the source of p, and another one according to the speaker’s assessment of p. Haider soll betrunken gewesen sein. (with p ¼ Haider betrunken gewesen sein) ‘Haider was supposed to have been drunk.’
Epistemic modal verbs therefore unfold a double deixis since they provide both the source of p (the propositional part of the utterance) and the speaker’s assessment of p. Epistemic modal verbs project double shifters (in Jakobson’s terms) referring both to the source of information and to the bearer of a certain opinion.5 In contrast, regarding epistemic and evidential adverbials, Abraham (2009: 9–10) claims that both unfold a simple deixis, as is shown in his representations in (4) and (5): (4) a. b.
epistemic adverbials/epAdv unfold but a simple deixis: Haider ist wahrscheinlich/vermutlich betrunken gewesen. ‘Haider was probably/presumably drunk.’
(5) a.
evidential adverbials/evAdv likewise unfold a simple deixis – speaker deixis missing.
5. Jakobson (1971) uses the term shifter, which was coined by Jespersen (1922), to refer to an element whose meaning can only be established with reference to the message that is being uttered in a specific context. For instance, personal pronouns are shifters. The word I refers to the speaker who utters a certain sentence. Therefore, its meaning varies along with the context.
Modal particles, speaker-hearer links, and illocutionary force
b.
259
Haider ist o¤ensichtlich betrunken gewesen. (with p ¼ Haider betrunken gewesen sein) ‘Haider was obviously drunk.’
In the case of epistemic adverbs, the source branch is missing, since they only report the speaker’s assessment of p. Regarding evidential adverbs, instead, they do provide the source of p, but have nothing to say with respect to the speaker’s assessment of p (cf. Abraham 2008b, 2009; Leiss 2008a, 2008b). Epistemic and evidential adverbs are therefore somehow ‘‘deictically’’ defective, if compared with modal verbs. If we now consider MPs, we can observe that they are shifters (in the sense of Jakobson 1971) in as much as, besides having a basic meaning, they introduce presuppositions with respect to the knowledge shared between speaker and addressee (cf. Abraham 2008a). Therefore, we can argue that they unfold a triple deixis, as illustrated by Abraham (2008b, 2009: 10): (6) a.
b.
modal particles/MPs denote a di¤erent deixis in the following sense: Speaker makes an estimate about the knowledge awareness of Addressee while letting the Addr know about this estimate and giving him a chance to relativize, or correct, this estimate about p. Haider soll ja betrunken gewesen sein (with p ¼ Haider betrunken gewesen sein) ‘Haider was supposed PART to be drunk alright.’
260
Marco Coniglio
In contrast to other modal words, MPs have a surplus. Besides denoting the source of p and the speaker’s assessment of p, they also provide information about the speaker’s estimate of the hearer’s state of knowledge about p, nonetheless giving her the possibility to react to this estimate. In this respect, MPs di¤er not only from modal adverbials, but also from the category of modal verbs. What is crucial is that MPs ‘‘[. . .] induce an intercommunication between Sp and Addr without giving full lexical expression to this exchange.’’ (Abraham 2009: 10). Hence, MPs play a fundamental role in this speaker-hearer link. In this paper, I will show the import of such observations in accounting for the syntactic distribution and behavior of MPs. I will argue that, despite their being modal words and, consequently, syntactically anchored to the higher portion of the IP (in a clausal functional structure a` la Cinque 1999), MPs pertain to the CP layer of a clause, where illocution and speaker are assumed to be encoded. But before putting forward how their interaction with illocution and speaker could be translated into a syntactic model, let us take into account some syntactic evidence for this assumption, which has so far been mainly based on pragmatic considerations. 2.2. Modal particles as modifiers of illocutionary force Many scholars, such as Doherty (1987), Jacobs (1986, 1991) and Thurmair (1989), have argued that MPs do not a¤ect the propositional part of a clause but rather modify its illocutionary force. We would therefore expect to find them in the highest portion of the clausal structure, namely in the CP layer. However, one of the best known syntactic properties of MPs is their confinement to the IP of the clause, i.e. in what topological studies on German would call the Mittelfeld ‘middle field’ of the clause.6 The middle field is the portion of the clause that is comprised between the finite verb in second position and the uninflected verb in final position, if both verbal elements are present. In contrast to adverbs, which – under certain circumstances – can occur more freely in the sentence (8), this is the only position accessible to MPs (7). (7) a.
Hans ist ja schon losgefahren. Hans is PRT already left ‘Hans has already left, as you know.’
6. In the highest IP-internal projections, Cinque (1999) posits mood and modality projections. In recent studies, it was shown that German MPs can only be found interspersed among these projections, namely in specific positions above the (higher) repetitive adverbs (cf. Coniglio 2005, 2006, 2007a, 2007b).
Modal particles, speaker-hearer links, and illocutionary force
b. c.
261
*Ja ist Hans schon losgefahren. *Hans ist schon losgefahren, ja.
(8) a.
Hans ist vermutlich schon losgefahren. Hans is probably already left ‘Hans has probably already left.’ b. Vermutlich ist Hans schon losgefahren. c. (?) Hans ist schon losgefahren, vermutlich.
The middle field restriction is the only generally accepted criterion which allows us to distinguish MPs from other lexical elements, such as adverbs.7 Although MPs only occur in the middle field of the clause, they display peculiar syntactic properties that connect them to the left periphery of the clause, in particular – as will be shown below – to the projection ForceP (also see Abraham 1995, 2008b; Zimmermann 2004a, 2004b). ForceP is the projection where the type of the clause, its illocutionary force and the speaker (to which force is anchored) are syntactically encoded. If proven, this relation between MPs and the CP would further confirm that MPs have a special syntactic status among expressions of modality, as already discussed above from a pragmatic perspective. At least two important facts show the close link between MPs and the CP layer, namely their restrictions to certain types of clauses and their interaction with the illocutionary force of the sentences where they occur. With respect to the first point, it has been generally observed that not all particles can occur in all types of clauses. Each particle is licensed in di¤erent contexts. As Thurmair (1989: 49) pointed out, a particle can generally occur in one or more of the seven illocutionary types that she distinguishes. Thus, for instance, the particle denn can occur in interrogative clauses, both polarity and wh-questions, but in no other types, as is shown in the following examples: (9) a.
Ist er denn schon losgefahren? is he PRT already left ‘Has he already left (I wonder)?’
7. Because of this property, MPs were assumed to be a phenomenon typical of German and of other Germanic languages which display a topological middle field (cf. Abraham 1991). However, recent studies have shown that the presence of MPs is common to many languages and is not always anchored in the presence of a German-like middle field.
262
Marco Coniglio
b.
*Er ist denn schonlosgefahren. he is PRT already left ‘He has already left (I wonder).’
c. *Fahr denn los! leave PRT ‘Leave (I wonder)!’ Each particle displays a di¤erent syntactic distribution. Thus, for instance, the particle wohl could felicitously replace denn both in the interrogative and in the declarative clause in (9a) and (9b), respectively, but not in (9c). MPs therefore interact with the type of the clause. This is encoded in the projection ForceP; hence, the need for postulating a connection between MPs and the CP layer. Let us now take into account the second piece of evidence for the existence of a syntactic link between MPs and the left periphery of a clause, i.e. their interaction with its illocutionary force. This fact has already been observed by many authors. Thurmair (1989: 2, 73), for instance, argues that MPs display their e¤ect in the illocutionary field by modifying or strengthening the illocutionary force of a clause. Jacobs (1986, 1991) formalized this idea by arguing that MPs are modifiers of illocutionary types. In particular, he explains how this modification takes place in the following way (Jacobs 1986: 103): ‘‘Wenn Verbstellung, Verbmodus, Intonation etc. einen bestimmten Illokutionstyp X festlegen, so wird daraus durch Hinzunahme eines Abto¨nungsmittels ein anderer Illokutionstyp X 0 , der in einem zu pra¨zisierenden Sinne eine speziellere, in ihren Anwendungsbedingungen eingeschra¨nktere Version von X ist.’’ [If verb position, verb mood, intonation etc. predefine a certain illocutionary type X, then by adding a modal particle this type turns into another illocutionary type X 0 , which is – in a sense that has not yet been specified – a more special version of X, more restricted in its usage conditions. MC]. This means that a particle takes up a certain illocutionary type and modifies it by restricting and specifying it. Let us illustrate this by means of the following imperative clause: (10) Komm JA nach Hause! come PRT to home ‘Do come home!’ In the absence of the stressed particle JA, the sentence would be a kind of ‘‘default’’ imperative clause, without any further specification: It could
Modal particles, speaker-hearer links, and illocutionary force
263
be an order, a request, a piece of advice, etc. The function of the stressed particle is that of making the illocutionary type more precise. It makes the imperative clause a more ‘‘specific’’ and ‘‘restricted’’ one, namely a cogent order. If we replaced the particle JA with the particle mal, for instance, then the imperative clause should be interpreted as a suggestion or a less peremptory order. Thus, we have just seen that there is a close relation between MPs and the type and illocution of a clause, in particular with the projection ForceP, where the latter is encoded. As I will show in the following sections, this surmise is endorsed by the peculiar distribution of MPs in main and subordinate clauses. Since MPs interact with illocutionary force, we must assume that they necessarily presuppose the presence of force in the sentences where we find them. As was already observed by Thurmair (1989: 73–82), a sentence which licenses MPs is illocutionarily independent. From a semanto-pragmatic perspective, this fact can be easily predicted, since MPs signal the speaker’s attitude to her utterance and, consequently, the presence of a new speech act. From a syntactic perspective, given the necessary presence of illocutionary force, we expect to find MPs in all contexts which display root properties, i.e. not only in main clauses, but – under certain circumstances – also in subordinate clauses. Let us first consider main clauses. It has been generally observed that each MP can only occur in a subset of all illocutionary types available. For example, the (unstressed) particle ja can only occur in declaratives, denn in questions, wohl in declaratives and questions and so on (see example [9] for denn). However, if we consider the whole class of MPs, we observe that they can be licensed in all main illocutionary types, namely at least in the seven types considered by Thurmair (1989: 49). This is easily predicted since main clauses are endowed with illocutionary force per definition. What is interesting, however, is that MPs can occur in subordinate clauses, too. Although some types of subordinate clauses can license no particles (11), there are many others which admit them (12): (11) Wenn es (*schon) Frost gibt, erfrieren die Rosen. if it PRT freeze gives freeze-to-death the roses ‘If it freezes, the roses will be killed by frost.’ (Brauße 1994: 112) (12) Er hat ein schlechtes Gewissen, weil er wohl gelogen hat. he has a bad conscience because he PRT lied has ‘He has a guilty conscience because he probably lied.’ (Asbach-Schnitker 1977: 48)
264
Marco Coniglio
How can we account for such facts? Given that MPs require the presence of illocutionary force, why can they be licensed in subordinate clauses? As will become clear in the next sections, this can be explained by assuming that root properties and illocutionary force are not restricted to main clauses. Recent theories on subordination assume that some types of subordinate clauses, too, unfold illocutionary force (cf. Haegeman 2002 and following works). I will show that these syntactic proposals confirm the validity of earlier observations on the illocutionary potential of MPs (see Doherty 1987, Jacobs 1986, 1991, Thurmair 1989, and Abraham 1995, 2009, for instance). I will claim that MPs are to be considered crucial criteria to recognize the presence of force not only in main clauses, but also in certain types of subordinate clauses, which consequently can be reclassified as illocutionarily independent clauses, irrespective of whether they are traditionally considered independent or not. What is at issue now is which clauses have root properties and, in particular, which are the syntactic conditions that are necessary for MPs to be licensed. Let us consider the following contrast: (13) Er ist nicht durchgekommen, weil er ja schlechte Noten he is NEG past-the-year, because he PRT bad grades bekommen hatte. got had ‘He failed the year because his grades were bad.’ (14) Er ist nicht durchgekommen, weil er (*ja) gute Noten he is NEG past-the-year, because he PRT good grades bekommen hatte got had (. . . sondern deshalb, weil er Mitleid bei den Lehrern erregt hat). (. . . but because the teachers pitied him). ‘He passed the year not because his grades were good, but because the teachers pitied him.’ Both causal clauses contain the MP ja. But, interestingly, while the causal clause in (13) can felicitously license it, the insertion of the particle in (14) would result in an ungrammatical sentence. Thus, an adequate analysis should be able to explain in which respect the causal clauses in (13) and (14) di¤er. I will show below that the semanto-pragmatic independence of a clause necessarily requires that this very clause possess particular syntactic prop-
Modal particles, speaker-hearer links, and illocutionary force
265
erties. It will be shown below that illocutionarily independent clauses must display a more complex syntactic structure and usually must be merged in a higher position than illocutionarily dependent clauses.
3. On the external and internal syntax of root and embedded clauses The fact that main clauses are endowed with independent illocutionary force is self-evident, since they are anchored to a speaker (i.e. to the ‘‘author’’ of the speech act) by definition. But what about subordinate clauses? As mentioned above, the possibility for subordinate clauses to share some properties with main clauses is a well-established fact by now. In the last decades, generative research has been concerned with the necessity of understanding two main aspects of the structure of a clause, i.e. its internal and external syntax. On the one hand, it has been necessary to establish what the internal structure of a clause looks like, which are the syntactic projections building up the skeleton of a clause and so on. On the other hand, it has been crucial to understand in which way clauses merge together. This is what we could call the ‘external’ syntax of a clause, i.e. the way a clause is merged to another in order to build up a complex sentence. Much has been done with respect to the first point. For example, the cartographic approach has mapped the syntactic projections that the clause structure is made up of. Much more remains to be done with respect to the second aspect. The way in which two clauses are merged together still remains unexplained or not fully clear. There are many problems one has to cope with and the comparison between languages makes the analysis even more di‰cult. Many scholars have tried to give an answer to such questions. An important result was achieved, namely the definition of some syntactic and semantic properties which characterize main clauses in contrast to subordinate clauses and which therefore allow us to take a step forward in understanding to what extent they di¤er. These properties are called root phenomena (or else embedded root phenomena, when occurring in subordinate clauses). Cf. Hooper and Thompson (1973), Heycock (2005). Starting from these important studies, Haegeman (2002, 2004a, 2004b, 2006) proposes an interesting analysis of subordinate clauses. In particular, she distinguishes two types of clauses. With respect to adverbial clauses, for instance, she argues for the existence of ‘‘central’’ and ‘‘peripheral’’ adverbials and claims that the latter are endowed with independent illocu-
266
Marco Coniglio
tionary force, thus being very similar to root clauses. Central adverbials, instead, do not display independent force, since they are part of the speech act of the matrix clause. In the next sections, I will briefly take into account Haegeman’s hypotheses. I will discuss her proposals on the external and internal syntax of clauses, in 3.1. and 3.2. respectively. 3.1. On the syntax of subordination The discussion in Haegeman (2002, 2004a, 2004b, 2006) starts from some considerations on the external syntax of subordinate clauses, i.e. on their relation with the main clause to which they are subordinated (also see Reis 1997). Haegeman claims that some clauses, which she calls peripheral adverbials, are characterized by the fact that they are adjoined to the main clause at the discourse level and are therefore syntactically and semantically independent from it, in contrast to central adverbials, which are closely linked to the matrix clause and, consequently, dependent on it. Haegeman (2002) focuses on contrasts such as the one in (15): (15) a. b.
If it rains we will all get terribly wet and miserable. If [as you say] it is going to rain this afternoon, why don’t we just stay at home and watch a video? Haegeman (2002: 117)
The first sentence is characterized by a close relation between the main clause and the subordinate clause: The conditional clause contains a ‘‘cause’’, which has the matrix clause as its e¤ect. We therefore observe the existence of a sequential relation between the event in the subordinate clause and the one in the main clause. That is the reason why Haegeman refers to conditionals such as the one in (15a) as event conditionals. These are adverbial clauses of the central type, since they are closely linked to the proposition in the main clause. The conditional clause in (15b) di¤ers from the one in (15a) in many respects. In (15b), the conditional clause, which Haegeman calls a premise conditional,8 contains the logical premise for the matrix clause. The conditional clause constitutes a speech act in itself, which justifies the speech act 8. In the literature, these types of conditionals are also known as relevance conditionals or biscuit conditionals, after a famous example by Austin (1956). Also see Johnson-Laird (1986), Ko¨nig (1986), Ko¨nig and van der Auwera (1988), and Dancygier and Sweetser (2005).
Modal particles, speaker-hearer links, and illocutionary force
267
(a question act, in this example) of the main clause. The syntactic merge of the matrix clause with the subordinate clause takes place at the discourse level. The conditional is not dependent on the matrix clause and, consequently, it is considered ‘‘peripheral’’ by Haegeman. Notice that we are not dealing with a real subordination in this case, since the syntactic relation between the two clauses is rather a kind of coordination. That is the reason why Haegeman uses the term associated clause when referring to the matrix clause of peripheral clauses. According to Haegeman (2002: 118), the term ‘‘matrix clause’’ would not be appropriate in cases such as (15b), where the subordinate clause is not embedded in the matrix clause, but rather coordinated with it. The conceptual distinction between central and peripheral adverbials is necessary not only from a semantic and pragmatic point of view, but also from a syntactic perspective. As shown by Haegeman (2002 and following works), the two types of adverbial clauses are syntactically quite di¤erent. If we consider their external syntax, i.e. the way they are adjoined to the matrix clause (or associated clause), we can observe that central adverbials are more deeply embedded than premise conditionals. The syntactic configuration for the relation between a central adverbial and its matrix clause is that of a real subordination. Conceptually, central adverbials must be interpreted as depending on the illocution of the matrix clause. Haegeman (2002: 131) therefore assumes that they are merged inside the IP of the matrix clause, as we see in the case of the event conditional in (16): (16) Event conditional (Haegeman 2002: 131)
268
Marco Coniglio
A central adverbial is probably adjoined to the vP or to an aspectual projection of the matrix clause. What is crucial is that such an adverbial has no illocutionary force. It depends on the speech act in the matrix clause, since it is syntactically in its scope. In the case of peripheral adverbials, we have a di¤erent configuration: The adverbial clause is independent from its associated clause, since – as was already said – it is just coordinated with it. Haegeman (2002: 132) provides the following syntactic representation for this relation: (17) Premise conditional (Haegeman 2002: 132)
The peripheral conditional CP2 and the associated clause CP1 are merged in a coordination-like structure. Hence, the premise conditional is adjoined to the CP of the associated clause only after the CP of the latter has been completed. The point where peripheral adverbials are merged with their associated clause is therefore very high and external to the associated clause. That is why Haegeman uses the term ‘‘peripheral’’ when she refers to such adverbials. In contrast to central adverbials, which are illocutionarily dependent on the matrix clause, peripheral adverbials are endowed with their own illocutionary force. In such cases, there are two speech acts, namely the speech act of the associated clause and the one of the adverbial. Despite the apparent subordination structure, the two clauses are coordinated at the discourse level. Given this important structural di¤erence between the two types of adverbial clauses, we expect them to behave di¤erently in many respects. Haegeman (2002 and following works), in fact, provides a number of syntactic tests, which prove the di¤erent external structure of the two types of clauses. By applying these tests, she aims to show that, since central adverbials are embedded in the matrix clause, they are usually in the scope of elements that belong to it. In contrast, peripheral adverbials can never be in the scope of the associated clause, since they are external to it. Here, I will just consider a couple of the tests used by Haegeman. Crucial evidence comes from the di¤erent scope of adverbs that occur in the matrix (or associated) clause. Since, in contrast to a peripheral
Modal particles, speaker-hearer links, and illocutionary force
269
adverbial, a central adverbial is embedded in the matrix clause, we expect that an adverb in the main clause can only quantify over a central adverbial, and not over a peripheral one. Let us illustrate this by means of the following contrast taken from Haegeman (2002: 127): (18) If a 15-year old boy is told to leave, it is often totally impractical for him to go alone, so a whole family, father, mother and two or three younger children may all be forced out. (Guardian, 19.3.2, p. 4, col 2) (19) [. . .] But if the writer and critic in me asked these questions, the reader never did, not for a single moment. (Guardian, 2.2.2., page 14, col 4) [. . .] The event conditional in (18) is in the scope of the adverb often in the matrix clause. The conditional is an adverbial clause of the central type. Despite its placement to the left of the adverb often, the central conditional must be interpreted as if it were embedded in the matrix clause, i.e. below the adverb. This is not the case for the premise conditional in (19). The adverbial clause is not in the scope of the adverb never in the associated clause. The conditional is a peripheral adverbial and cannot therefore be in the scope of elements in the associated clause. A similar contrast can be observed with respect to the scope of the negation and of negative elements in the main clause. As we see in the following examples by Haegeman (2002: 128), the behavior of the two types of adverbials with respect to the scope of the negation is di¤erent: (20) a. b.
John won’t manage if there’s a lot of pressure on him. John won’t manage, if there’s (already) such a lot of pressure on him.
The adverbial clause in (20a) is a central conditional and is therefore in the scope of the negation in the matrix clause. In contrast, the conditional clause in (20b) is not in the scope of won’t, since it is a peripheral adverbial and is consequently adjoined at a higher level than the negation. As Haegeman (2002: 129–130) points out, binding tests, too, are useful to distinguish the two types of clauses. Let us consider her examples: (21) a. b.
No one will answer the phone if he thinks it’s his supervisor. Why does no one answer the phone, if he probably thinks it’s his supervisor?
270
Marco Coniglio
In (21a), the quantifier no one in the matrix clause can bind the variables he and his, which occur in the conditional clause, and be coreferential with them. This means that the conditional is in the scope of the quantifier and is therefore to be considered a central adverbial. In the sentence in (21b), in contrast, the quantifier cannot bind he and his since it does not take scope over them. As a consequence, he and his cannot corefer with the quantifier, but must obligatorily refer to another person. In this case, we are dealing with a peripheral conditional, which – given its very high merge point – cannot be in the scope of quantifiers in the associated clause. Here, I mentioned just a couple of tests used by Haegeman in order to show the di¤erent syntactic behavior of the two types of clauses. What is crucial is that lexical material in the matrix (or associated) clause can take scope over central adverbials, since the latter are truly embedded in the matrix clause. In contrast, the same elements cannot take scope over peripheral adverbials. Peripheral adverbials are merged with the associated clause only once the CP of the latter has been completed and are therefore out of the scope of material belonging to it. Notice that the distinction between central and peripheral clauses holds not only for conditional clauses, to which the examples above refer, but also for other types of subordinate clauses. Thus, for instance, we distinguish between central and peripheral causal clauses. Let us consider the following examples: (22) a. b.
I did not read that book because I liked it, but because I had to. I did not read that book because I did not like it.
In (22a), the causal clause is a central adverbial, since it is in the scope of the negation belonging to the matrix clause, while in (22b) the causal clause is of the peripheral type: it is out of the scope of the matrix negation. In conclusion, regarding the external syntax of subordinate clauses, we must distinguish between central and peripheral adverbials depending on their relation to the main clause. Central adverbials are characterized by their syntactic dependence on the matrix clause, whilst peripheral adverbials are illocutionarily and syntactically independent. In the next section, I will show that the internal syntax of the two types of clauses di¤ers as well. 3.2. On the structure of the left periphery of subordinate clauses Since the important studies by Rizzi (1997, 2001, 2004), generative linguists have assumed that the clausal left periphery, which was once con-
Modal particles, speaker-hearer links, and illocutionary force
271
sidered to be a unique syntactic projection, consists of several projections. The existence of a complex CP layer is based on the observation of important phenomena in Italian and in other languages, which show that a more fine-grained structure is needed. Here, I will not take Rizzi’s arguments into consideration. I refer to his works for further explanations. For her purposes, Haegeman starts from the following structure a` la Rizzi (1997 and following works): (23) Force Top* Focus Mod* Fin The CP domain is comprised between the two projections ForceP and FinP, which encode the illocutionary force of the clause and rudimentary information about the embedded IP, respectively. Between these two projections, one can find fronted and focused elements. According to Haegeman (2002), who follows Rizzi (1997 and following works), TopP is responsible for licensing fronted arguments, FocusP is the projection where focused elements and wh-phrases are hosted, while in ModP one finds fronted adjuncts. Since Rizzi’s proposals were adopted, it has become clear that not all phenomena that are encoded in the projections of the CP layer are available in all types of clauses. Haegeman’s great merit was to provide an important generalization about the conditions under which these phenomena can be licensed in subordinate clauses. In the following sections, I will briefly present her proposals with respect to each subtype of subordinate clause. 3.2.1. Adverbial clauses When Haegeman (2002 and following works) distinguishes between central and peripheral adverbials, she claims that their internal syntax, too, di¤ers, in as much as their CPs display a di¤erent number of projections. I.e., central adverbials have a reduced CP, while peripheral adverbials display the complete structure of the left periphery.9 Let us briefly see what that means. It was shown above that the CP comprises about five projections. However, Haegeman argues that not all the phenomena encoded in these projections are available in all types of clauses and that consequently certain projections are not always activated. 9. Notice that, in more recent works, Haegeman (2008a, 2008b) assumes that the highest part of the IP is impaired, too, since also Cinque’s (1999) higher adverbs and certain modal verbs cannot be licensed in central clauses (cf. fn. 12).
272
Marco Coniglio
In particular, ForceP, TopP and FocusP seem to be only available in peripheral clauses, in contrast to ModP and FinP, which are activated in both central and peripheral adverbials. Let us just consider the availability of TopP, FocusP and ModP, since they have a clear lexical realization in English. If we first take subordinate clauses of the peripheral type into account, we observe no contrast in the licensing of fronted arguments (hosted in TopP) and adjuncts (in ModP). The following examples for English are taken from Haegeman (2004a: 149, 160): (24) a. b.
If these problems we cannot solve, there are many others that we can tackle immediately. If with all these precautions you have not succeeded, why do you want to try again?
In peripheral adverbials, both the licensing of fronted arguments (24a) and of fronted adjuncts (24b) is possible. If we now take into account two central adverbials such as the following event conditionals, we observe a di¤erent pattern. Both examples are taken from Haegeman (2002: 148): (25) a. b.
*If these final exams you don’t pass, you won’t get the degree. If with these precautions you don’t succeed, you should try again next week.
While fronted adjuncts are still available, as we see in (25b), fronted arguments are excluded from such contexts (25a). As for focused phrases, which are licensed in the projection FocusP, they pattern with fronted arguments, since they cannot occur in central adverbials. See the following example for Italian, which is taken from Haegeman (2002: 152): (26) *Se IL MIO LIBRO riesci a leggere, supererai if the my book you-manage to read you-will-pass 10 l’esame. the exam ‘If MY BOOK you manage to read, you will pass the exam.’
10. However, this sentence is grammatical for some speakers, as pointed out by Haegeman (2002: 152) herself.
Modal particles, speaker-hearer links, and illocutionary force
273
Therefore, these phenomena, which are generally referred to as ‘‘root phenomena’’ (cf. Hooper and Thompson 1973; Heycock 2005; etc.), are possible only in what Haegeman calls ‘‘peripheral’’ adverbials. Notice that such phenomena are assumed to be licensed in CP-internal projections. Thus, their absence from central adverbials leads Haegeman (2002 and subsequent works) to consider it a piece of evidence of the structural reduction of these clauses.11 From a cartographic perspective, this idea is translated by Haegeman (2002) into the structures in (27). She assumes that, in contrast to root clauses and peripheral adverbials, central adverbials are structurally reduced:12 (27) a. b. c.
Central adverbials: Sub Mod* Fin Peripheral adverbials: Sub Force Top* Focus Mod* Fin Root clauses: Force Top* Focus Mod* Fin Haegeman (2002: 159)
While peripheral adverbials unfold the complete structure of the left periphery, central adverbials are truncated above the projection ModP.13 This would explain the fact that the latter do not have illocutionary force and cannot license root phenomena, as was shown above. According to Haegeman (2002 and following works), the absence of root phenomena in central clauses depends on the absence of the projection ForceP in these same contexts. Hence, ForceP is responsible not only for 11. The observation that, in (truly) embedded clauses, a limited number of syntactic phenomena is available reminds us of an important principle formulated by Ross (1973: 397), namely the Penthouse Principle: (i) The Penthouse Principle More goes on upstairs than downstairs. The idea that Ross pursues is that a fewer number of phenomena are available in subordinate clauses (downstairs) than in main clauses (upstairs), since embedded phenomena are only a subset of the options that one has in unembedded clauses. This principle could be updated by including peripheral adverbials under the label ‘‘upstairs’’, given that they are syntactically adjoined at a very high structural level (cf. 3.1.). 12. In more recent works, Haegeman (2008a, 2008b) no longer resorts to the hypothesis of a ‘‘truncation’’ of certain clauses, but to an alternative explanation, based on operator movement, which takes place only in central clauses: The presence of root phenomena would yield an intervention e¤ect, which blocks this movement. 13. Here, I will not consider the problem posited by the projection of the subordinating conjunction (SubP).
274
Marco Coniglio
encoding the illocutionary force of a clause, but also for the activation of the projections that license root phenomena. 3.2.2. Complement and relative clauses Haegeman (2006) puts forward a similar analysis for complement clauses. Moreover, she tentatively suggests that this proposal could be extended to relative clauses as well (Haegeman 2002: 166). Let us first consider complement clauses. Haegeman (2004a, 2006) distinguishes two types of complement clauses, i.e. those depending on factive verbs, which presuppose the truth of the embedded utterance, and those depending on non-factive verbs, which are all other types of complement clauses. Based on tests similar to those shown above for adverbial clauses, Haegeman argues that factive complements are structurally impoverished if compared with non-factive complements, since they lack the projection ForceP and the other related projections: (28) a. Non-factive complements: that (Top) (Focus) Force Mod* Fin b. Factive complements: that Mod* Fin Haegeman (2004a: 171) I will not discuss the reasons why Haegeman (2004a) inserts the projection ForceP in a lower position in this structure (also see Haegeman 2006: 1662–1663 on this issue). What matters here is that, in this case, too, precisely this projection is crucial for licensing fronted arguments and focused elements. Root phenomena are only available if the complement clause has illocutionary force. Complement clauses are more problematic than adverbials, since they are syntactically embedded in the matrix clause, even in the case of nonfactive complements (see the discussion on example [31] below). But it was shown that the presence of force usually correlates with a more peripheral merger position. According to Haegeman, non-factive complements are nonetheless endowed with illocutionary force. Their force would derive from the predicate of their matrix clauses. In section 4.1., I will discuss the distribution of MPs in complement clauses and I will show that it is in fact the matrix predicate that makes a speaker available for the embedded clause. When discussing the speaker of complement clauses, Haegeman (2002) uses the term ‘‘potential’’ speaker, since this does not necessarily coincide with the real speaker but, in general, it is the subject of the predicate in the matrix clause. This means that, in the case of com-
Modal particles, speaker-hearer links, and illocutionary force
275
plement clauses, the projection ForceP would provide a link to the speaker encoded in the superordinate clause (see also Sections 4.1 and 5). If we now consider relative clauses, we could argue that a truncation hypothesis also holds in this case. As suggested by Haegeman (2002: 166), we should distinguish between restrictive and appositive relative clauses, since the former seem to involve a reduced structure and to be adjoined at a lower level when compared with the latter. However, Haegeman (2002 and following works) does not take these types of clauses directly into consideration. She only refers to the vast literature on this topic, which shows that appositive relative clauses can license certain root phenomena and unfold illocutionary force. In contrast, restrictive relatives are much more similar to central adverbials.
4. Modal particles as markers of illocutionary force in subordinate clauses Summarizing Haegeman’s (2002 and following works) proposals, it was shown that we should distinguish each type of subordinate clause on the basis of two criteria: 1) according to its external syntax, i.e. whether it is embedded in the main clause or unembedded and 2) according to its internal syntax, i.e. whether it displays a projection ForceP. We observe a correlation between these two aspects: Peripheral clauses have illocutionary force, central clauses do not. There is only one exception, i.e. non-factive complements, which are embedded, but nonetheless have the projection ForceP, due to the properties of the selecting verb. Hence, we can conclude that the presence of Force is crucial for the licensing of root phenomena in a clause – independently of its exact degree of embedding. In the following part, I will present some occurrences of MPs in subordinate clauses. In particular, I will consider their distribution in complement clauses, adverbials and relative clauses. I will argue that MPs are important tools for detecting the illocutionary force of a clause. Although on independent grounds, the results of my investigation will confirm the proposals in Doherty (1987), Jacobs (1986, 1991), and Thurmair (1989) among others. The distribution of MPs will be examined on the basis of concrete examples. By means of simple syntactic tests, it will be shown that MPs are only compatible with clauses with root properties. Following Thurmair (1989: 73–82), I will hence claim that the presence of particles disambiguates a clause and assigns illocutionary force to it, since MPs are markers of this force.
276
Marco Coniglio
I will conclude that Haegeman’s theories about the illocutionary independence of certain types of subordinate clauses match very well with the distribution of MPs, i.e. MPs only occur in those subordinate clauses that display the complete structure of the CP and are illocutionarily independent from their associated clause. A central aspect will be the definition of the way in which the anchoring of a clause in the speaker takes place. In fact, given the di¤erent external syntax of complement and subordinate clauses and the di¤erent mechanism in which they are assigned illocutionary force, we expect that the speaker, too, is made available in a di¤erent way. 4.1. Modal particles in complement clauses On the basis of the class of predicates that select a complement clause, we distinguish factive and non-factive complements. According to Haegeman (2006), the first do not display illocutionary force, while the latter do, since they encode a speaker. Consequently, we expect that, in contrast with factive complements, non-factive complements can license MPs, which require the presence of a speaker to whom they can be anchored. This prediction is borne out by the following contrast: (29) Katharina bedauert es, dass Bayern Mu¨nchen (*wohl ) Katharina regrets it that Bayern Munich PRT gewonnen hat. won has ‘Katharina regrets that Bayern Munich won.’ (30) Katharina hat gesagt, dass Bayern Mu¨nchen wohl gewonnen hat. Katharina has said that Bayern Munich PRT won has ‘Katharina said that Bayern Munich won.’ The MP wohl cannot occur in the complement clause in (29), because the selecting predicate (es) bedauern is a factive verb. An important property of factive complements is that they contain a statement which is presupposed by the selecting predicate in the matrix clause. Since this statement is a fact, the speaker has no possibility of expressing her assessment of p (see 2.1.) by means of a MP. In contrast, the non-factive complement selected by the predicate hat gesagt in (30) does encode a speaker and therefore can felicitously license the particle wohl.
Modal particles, speaker-hearer links, and illocutionary force
277
The occurrence of particles in complement clauses signals the presence of illocutionary force in the same clauses. We see a problem with this analysis. It was shown above that the presence of illocutionary force in a clause generally correlates with an important syntactic property, i.e. the more peripheral position occupied by it. Although this holds for adverbial clauses, it does not seem to be the case for complement clauses. The latter are always embedded in the matrix clause, i.e. not only in the case of factive complements (which is expected, since they do not have illocutionary force, nor can they license MPs), but also in the case of non-factive clauses, as hinted at above. This can be proven by using the binding tests that were shown in 3.1. Consider the following example: (31) Jederi der Studenten hat behauptet, dass eri wohl lieber each of-the students has a‰rmed that he PRT preferably einen anderen Film sehen mo¨chte. an other movie see would-like ‘Each of the students a‰rmed that he preferred to see another movie.’ Since, in this example, the quantifier jeder ‘each’ and the variable er ‘he’ can be coreferential, we can conclude that the non-factive complement clause is embedded in the matrix clause. Nonetheless, the complement clause can unexpectedly license the MP wohl. Above, I assumed that MPs are to be considered indicators of the presence of illocutionary force in a clause (cf. Thurmair 1989: 73–82). This means that, given the presence of a MP, the complement clause in (31) must be illocutionarily independent and have the projection ForceP. How can we reconcile these facts with the embedded position occupied by non-factive complements? The answer to this question was provided by Haegeman herself. With respect to adverbial clauses, Haegeman (2002 and following works) postulates that the presence of force correlates with the more peripheral syntactic position occupied by these subordinate clauses. In contrast with adverbial clauses, non-factive complements – according to Haegeman (2002 and following works) – derive their illocutionary force from the predicate in the matrix clause. The main clause unfolds special properties, which can endow the embedded clause with force. This entails that the illocutionary force of complement clauses is not ‘‘anchored’’ to the speaker, but rather to what Haegeman calls the ‘‘potential’’ speaker (Haegeman 2002: 159–160). This usually coincides with the subject of the matrix predicate. We can prove this by means of MPs, which are claimed to be shifters in the sense of Jakobson (1971). Consider
278
Marco Coniglio
the sentence in (30) once again. In contrast to particles occurring in subordinate clauses, the particle in the example above does not refer to the actual speaker. The reference point of the particle wohl is Katharina, i.e. the subject of the matrix clause, which is the potential speaker in the reported speech.14 In this case, too, the accessibility to a speaker, to whom the illocution can be referred, is crucial for the licensing of MPs. But what makes illocutionary force available in such contexts are not the intrinsic properties of the subordinate clause, but rather the selecting properties of the matrix predicate. 4.2. Modal particles in adverbial clauses In this section, I will present some occurrences of MPs in adverbial clauses in order to demonstrate that these signal the presence of illocutionary force in these cases as well. In fact, the use of a MP indicates the speaker’s commitment to her utterance. I.e., she does not mean to accept the content of the adverbial clause as a mere fact, but she rather intends to express her assessment of the propositional content of her utterance. Therefore, we expect the existence of a correlation between the peripheral merging point of a clause and the possible licensing of MPs. In contrast, central adverbials should be incompatible with MPs, since they are devoid of illocutionary force. Let us consider the sentences in (13) and (14) again, which are repeated here as (32) and (33), respectively: (32) Er ist nicht durchgekommen, weil er ja schlechte Noten he is NEG past-the-year, because he PRT bad grades bekommen hatte. got had ‘He failed the year because his grades were bad.’ (33) Er ist nicht durchgekommen, weil er (*ja) gute Noten he is NEG past-the-year, because he PRT good grades bekommen hatte got had (. . . sondern deshalb, weil er Mitleid bei den Lehrern erregt hat). (. . . but because the teachers pitied him). ‘He passed the year not because his grades were good, but because the teachers pitied him.’ 14. Also see Do¨ring (2007), who discusses similar shifts in the reference of MPs.
Modal particles, speaker-hearer links, and illocutionary force
279
As I have already pointed out above, at first glance, this contrast is striking. Why can the same particle be licensed in the causal clause in (32), but not in (33)? With Haegeman’s proposals in mind, we have a straightforward explanation for it. In (32), the causal clause can optionally license the MP ja because the adverbial clause is not in the scope of the negation nicht in the associated clause. That means that the causal clause is illocutionarily independent and, thus, a peripheral adverbial. Consequently, the presence of the projection Force allows the licensing of the particle. In the second sentence, instead, the causal clause is in the scope of the negation nicht in the matrix clause and, as expected, cannot felicitously license the particle ja. The causal clause in this example is a central adverbial and has no illocutionary force, which would be necessary for the particle to be licensed. From the analysis of this contrast, we can conclude that Haegeman’s distinction between central and peripheral adverbials is corroborated by the distribution of MPs. Hence, the optional presence of MPs can help us distinguish between clauses which are illocutionarily dependent on the matrix clause and clauses with independent force. Notice that the anchoring of a particle to the speaker plays a crucial role for its licensing. If we consider the grammatical example in (32), we can prove that the particle ja must be directly referred to the speaker, and not, for instance, to the subject of the associated clause (er), which could be a potential speaker (cf. 4.1.). The presence of the projection ForceP in the subordinate clause makes the actual speaker directly accessible to the adverbial clause (and consequently to the particle too). Similar observations can be made with respect to conditional clauses, too. In section 3, it was shown that Haegeman (2002 and following works) distinguishes event and premise conditionals on the basis of their adjunction point and of the optional presence of illocutionary force. If we consider the following contrast, taken from Brauße (1994), we can argue that Haegeman’s theories are supported by the distribution of particles in this case as well: (34) Wenn es (*schon) Frost gibt, erfrieren die Rosen. if it PRT freeze gives freeze-to-death the roses ‘If it freezes, the roses will be killed by frost.’ (Brauße 1994: 112)
280
Marco Coniglio
(35) Wenn es schon Frost gibt, ko¨nnte es wenigstens auch schneien. if it PRT freeze gives could it at least also snow ‘If it really has to freeze, it should at least snow as well.’ (Brauße 1994: 112) In the case of (34), the insertion of the MP schon is ungrammatical.15 Given the cause-e¤ect relation between the conditional and the matrix clause, the adverbial clause must be interpreted as a central conditional. Thus, it lacks independent illocutionary force. The case in (35) is di¤erent. The conditional is a peripheral adverbial. It expresses an event, which has as its consequence not another event, but the speech act of the associated clause. Thus, the conditional clause is illocutionarily independent, as is signaled by the presence of the particle. In the grammatical example in (35), too, the particle schon directly refers to the speaker of the whole utterance, i.e. to the actual speaker. In any case, no other potential speakers in the associated clause would be accessible in this case. Similar conclusions could be drawn for all other types of adverbials, which I did not mention here (cf. Coniglio 2011: 130–206 for a deeper investigation). Some types of subordinate clauses can be both central and peripheral, as in the case of causal and conditional clauses, which we have just seen, but some subordinate clauses only have one option: They can be either central adverbials (such as temporal clauses) or peripheral adverbials (as in the case of concessive clauses). What matters is that the possible presence of MPs in a subordinate clause is crucial for signaling the speaker’s commitment and, consequently, the presence of illocutionary force in the same clause.16 4.3. Modal particles in relative clauses With respect to relative clauses, we observe that MPs can occur in appositive relatives (37), but not in restrictive ones (36): (36) Sie hat den einzigen Roman von Eco gekauft, she has the only novel of Eco bought den ich (*ja) nicht zu Ende lesen konnte. which I PRT not to end read could ‘She bought the only novel by Eco that I could not finish reading.’ 15. Notice that schon as a temporal adverbial would be possible in this sentence. 16. For a complete list of the embedded contexts where MPs can (or cannot) occur, see the table in (43).
Modal particles, speaker-hearer links, and illocutionary force
281
(37) Sie hat den Roman ,,Der Name der Rose‘‘ gekauft, she has the novel The Name of the Rose bought den ich ( ja) nicht zu Ende lesen konnte. which I PRT not to end read could ‘She bought the novel ‘‘The Name of the Rose’’, which I could not finish reading.’ The hypothesis of a structural di¤erence between the two types of clauses, which Haegeman (2002: 166) hinted at, and which is confirmed by the vast amount of literature on this topic (see, for instance, de Vries 2002) is therefore supported by the distribution of MPs in these contexts. Restrictive relatives are devoid of illocutionary force, since they are dependent on an antecedent in the main clause for their interpretation. Thus, they cannot license MPs. Appositive relatives, instead, constitute autonomous speech acts and, thus, often contain expressions of the speaker’s assessment of p (cf. 2.1). In such cases, particles function as clear markers of illocutionary force. As was claimed for instance by Thurmair (1989), the use of a particle generally turns a relative clause which is ambiguous between a restrictive and a non-restrictive reading (38) into a non-restrictive clause (39), as in the following examples taken from Thurmair (1989: 80): (38) Autos, die laut sind, sollten mit einer geschlossenen cars which loud are should with a closed Motorkapsel versehen werden. engine encloser supplied be ‘Cars(,) which make a lot of noise(,) should have an engine encloser.’ (39) Autos, die ja laut sind, sollten mit einer geschlossenen cars which PRT loud are should with a closed Motorkapsel versehen werden. engine encloser supplied be ‘Cars, which as you know make a lot of noise, should have an engine enclosed.’ The particle17 has the function of underlining the presence of an independent speech act. By using it, the speaker intends to signal her commitment to the utterance, hence the appositive reading of the relative clause. 17. Notice that similar observations are valid for certain high adverbs, too.
282
Marco Coniglio
The incompatibility of a particle with a truly ‘‘embedded’’ relative clause can be proven by means of di¤erent syntactic tests, for instance the binding test seen in section 3.1. Consider the following sentence: (40) Jede Frau, die (*wohl) ein Kleid von Louis Vuitton each woman who PRT a dress of Louis Vuitton gekauft hat, hat einen Rabatt bekommen. bought has has a discount got ‘Each woman that bought a Louis Vuitton dress got a discount.’ In (40), the quantifier jede binds the relative pronoun die in the subordinate clause. Only if the particle is absent can the relative clause be embedded and receive a restrictive interpretation. The presence of a particle, instead, entails a shift in the interpretation of the relative clause, which consequently becomes appositive and, hence, an independent speech act. However, this shift is incompatible with quantifiers such as jede in (40), which require an embedded restrictive clause.18 Therefore, in presence of the particle, the sentence is ungrammatical. We can conclude that a particle signals the illocutionary independence and the appositive interpretation of a relative clause. Only appositive relatives display the projection ForceP, to which the epistemic reference of a particle must be anchored. As was already shown above for adverbial clauses, we observe that, in this case, too, the speaker to whom the illocution of the appositive relative is referred is the actual speaker. In (37), for instance, the particle refers to the speaker, and not to the potential speaker sie ‘she’. Notice however that it is possible to find MPs in certain restrictive relative clauses, too, namely only in the presence of a potential speaker and of certain predicates which allow the anchoring to this speaker. Consider the following example:
18. In the absence of a quantifier, the relative clause would be grammatical and receive an appositive interpretation: (i) Die Frau, die wohl ein Kleid von Louis Vuitton gekauft hat, the woman who PRT a dress of Louis Vuitton bought has hat einen Rabatt bekommen. has a discount got ‘The woman, who bought a Louis Vuitton dress, got a discount.’
Modal particles, speaker-hearer links, and illocutionary force
283
(41) Marco sucht ein Auto, das es wohl gar nicht gibt. Marco looks-for a car which it PRT not at all gives ‘Marco is looking for a car that does not exist.’ Although this is a restrictive relative clause, it tolerates the presence of the MP wohl. Interestingly, the particle is not anchored to the actual speaker, but to the ‘‘potential’’ one in the matrix clause, i.e. Marco. As in the case of complement clauses, besides the presence of a potential bearer of the illocution, a special predicate is needed, too. If we replace the verb suchen ‘to look for’ in (41) with haben ‘to have’, for instance, the sentence becomes ungrammatical: (42) *Marco hat ein Auto, das es wohl gar nicht gibt. Marco has a car which it PRT not at all gives ‘Marco has a car that does not exist.’ In relative clauses, too, it is therefore crucial that a speaker who commits herself to the utterance be available. This could also be a potential speaker, but it must be compatible with the predicate in the matrix clause. In any case, their special status as shifters (cf. Jakobson 1971) allows particles to refer either to a potential or to the actual speaker, according to the situation.
5. Looking for a syntactic link between modal particles and the speaker Summarizing the results so far, we have seen that if a clause licenses a MP then it must be considered an autonomous speech act (also cf. Thurmair 1989, for instance). The presence of MPs signals the presence of an active speaker, who wants to stress her own mental representation of a certain fact and to attribute a certain attitude and state of knowledge to the hearer. The speaker, who is the actor of this cognitive process, finds her syntactic representation in ForceP, i.e. in the projection where the illocutionary force is encoded. Since many types of subordinate clauses do not admit the syntactic representation of the speaker, we can assume that these clauses do not have such a projection, nor can they license MPs. In the table below, I report the results of my survey on the distribution of MPs in subordinate clauses. In particular, this table improves the results provided by Thurmair (1989: 81) and Hentschel (1986: 201).
284
Marco Coniglio
(43) The distribution of MPs in subordinate clauses:
Complement/content clause
non-factive complement clauses
þ
factive complement clauses
indirect questions adversative/contrastive clauses causal/reason clauses
Adverbial clauses
þ /þ
concessive clauses
þ
conditional clauses
/þ
consecutive/result clauses
/þ
final/purpose clauses
/þ
locative/place clauses
modal clauses
Relative clauses
/þ
/þ
temporal/time clauses
restrictive relative clauses
non-restrictive relative clauses
þ
other relative clauses
/þ
From the observation of the distribution of MPs, we can conclude that they can only occur in clauses which are endowed with independent illocutionary force, i.e. in those clauses that – according to Haegeman’s predictions – display root properties. The table must be interpreted in the following way. On the one hand, there are subordinate clauses with complete CP structure, which are illocutionarily independent and can consequently license MPs (marked with þ). These are non-factive complements, peripheral adverbials (such as contrastive clauses) and appositive relatives. On the other hand, some subordinate clauses display a reduced CP and, since they are devoid of illocutionary force, they do not allow for MPs (marked with ). These are complements of factive verbs, central adverbials (such as locative clauses) and restrictive relatives. Notice that there are some clauses, which can be either of one type or of the other (/þ), depending on the context (this is the reason why my analysis improves the results in Thurmair 1989: 81, among others). In 4.2., for instance, we discussed the cases of causal and conditional clauses.
Modal particles, speaker-hearer links, and illocutionary force
285
These ambivalent types of clauses can or cannot license MPs according to their interpretation. If we now include main clauses in the group of clauses with illocutionary force, which can license MPs, we arrive at the important generalization that MPs allow us to distinguish between independent speech acts and mere dependent clauses. I have shown that MPs are markers of the presence of illocutionary force. This is particularly clear in cases such as relative clauses, where particles – often in combination with other strategies – are used to disambiguate a potentially ambiguous relative clause and to make it explicitly non-restrictive, by stressing the role of the speaker, of her assessment of p and of her estimate of the hearer’s state of knowledge. A clause that can license a MP is a speech act. I argued that MPs are crucial evidence for this. In particular, the presence of force correlates with important syntactic properties, such as the more peripheral position where such clauses are generated, at least in the case of peripheral adverbials or of appositive relative clauses. In contrast, in the case of factive complements, the presence of force goes hand in hand with the presence of qualified predicates, which can embed a speech act. I argued that a particle must be anchored to a speaker. Given their special nature as shifters (cf. Jakobson 1971), MPs will be anchored to the ‘‘contextual’’ speaker, i.e. the speaker in a specific context, who happens to vary during a normal conversation. Since the contextual speaker is syntactically encoded in the projection ForceP, a clause can license MPs only if it contains this projection. ForceP makes either the actual speaker or – in the case of complement clauses – a ‘‘potential’’ speaker available. For these reasons, we could postulate that, syntactically, the particle must search for the projection Force. If it cannot find it, it will not be licensed, because it fails to receive its epistemic reference point. Notice that the mechanisms for licensing MPs in adverbial and complement clauses are quite di¤erent. In the case of adverbial clauses, the anchoring to the speaker takes place in the following way:19
19. However, there are rare cases of illocutionarily independent adverbials (such as some types of consecutive and purpose clauses) which behave as complement clauses, thus being embedded in the matrix clause.
286
Marco Coniglio
(44) Anchoring to the speaker of an adverbial clause:
The relation between the main clause and the adverbial clause is that of a conjunction. Both clauses have direct access to the speaker. Consequently, once a particle in the adverbial clause is linked to ForceP, it will be automatically anchored to the actual speaker. As for complement clauses, the anchoring to the speaker takes place in a di¤erent way. As represented below, only the illocutionary force of the matrix clause will be anchored to the real speaker. The complement clause, in turn, will only have access to the potential speaker, who is made available by the predicate in the matrix clause: (45) Anchoring to the speaker of a complement clause:
Modal particles, speaker-hearer links, and illocutionary force
287
Therefore, a MP in a complement clause cannot be anchored to the actual speaker.20 It must be referred to the potential speaker inside the matrix clause. Finally, regarding relative clauses, the anchoring of MPs usually takes place in the same way as was shown for adverbial clauses, i.e. they are linked to the actual speaker of the utterance. However, as was shown in example (41) above, they can sometimes be anchored to a potential speaker – similarly to complement clauses – provided that they find an ‘‘illocution-bearing’’ element in the matrix clause, which can legitimate the illocution of the relative clause.21 Such a potential speaker makes the necessary Force available and selects an embedded clause with complete CP structure. ForceP is the projection where not only the illocutionary type and the force of a clause are encoded, but also the speaker with all her linguistically unexpressed features. ForceP marks the presence of a speech act and implies the presence of an ‘‘actor’’, namely the speaker, with all her cognitive representations. In this speech-act-related projection, there is also place for the syntactic encoding of the pragmatic interaction between speaker and hearer (see section 2), more precisely for the speaker’s estimate of the knowledge awareness of the hearer (cf. Abraham 2008b, 2009; Leiss 2008a). Only in the presence of Force does this speaker-hearer bridging become possible. From a theoretical point of view, it is therefore necessary to postulate a syntactic link between MPs and the projection ForceP, to which MPs must have access in order to be licensed. So far, two alternative proposals have been put forward in order to account for this link: Either that MPs move to ForceP at Logical Form or that they enter into an Agree-relation 20. Except when the main clause encodes no concrete potential speaker at all or a generic one, as in (i), or when actual speaker and potential speaker coincide (ii): (i) (?)Es ist denkbar, dass Marco es wohl scha¤en wird. it is thinkable that Marco it PRT manage will ‘One could think that Marco will probably manage it.’ (ii) Ich glaube, dass Marco es wohl scha¤en wird. I think that Marco it PRT manage will ‘I think that Marco will probably manage it.’ In both cases, it is as if the particle could bypass the matrix clause and be directly linked to the actual speaker. 21. Notice that this holds for some adverbial clauses as well (cf. fn. 19).
288
Marco Coniglio
with the head Force . Since the first proposal is problematic for recent proposals within a minimalist framework,22 I will only discuss the second approach, which was provided by Bayer (forthcoming) with respect to the particle denn. I believe that such an explanation could be extended to all particles. Bayer claims that a MP such as denn enters into an Agree-relation with a C-head, as shown in (46), where I have slightly changed his proposal in order to make it suitable for all particles. In the representation below, only the salient step of the syntactic derivation is provided, where, after the merger of Force , the Agree operation between the MP and this head takes place: (46) Agree relation between MPs and Force : a. Force . . . [ IP . . . MP . . . [ VP . . . ] ] agree ! b. Force . . . [ IP . . . MP . . . [ VP . . . ] ] Each particle is endowed with two relevant features. On the one hand, it has an interpretable feature (illocutionary feature), which regards the semanto-pragmatic function of the particle. For example, as proposed by Bayer (forthcoming), we could postulate a feature conc(ern) for the particle denn, since this particle expresses concern or interest on the speaker’s side. This feature therefore expresses the way the particle modifies the illocutionary force of a clause. On the other hand, each particle has an uninterpretable feature uSpeaker. The particle must be linked to a speaker, but this can only be made accessible by the projection ForceP. Since the head Force encodes information about the ‘‘actor’’ of the speech act, it has an interpretable feature Speaker.
22. According to this cartographic proposal, which was provided by Coniglio (2007c, 2009, 2011), who in turn builds on Abraham (1995) and Zimmermann (2004a, 2004b), a particle covertly moves from its IP-internal position to the specifier of ForceP in order to be licensed, as represented below: (i) Covert movement of MPs (Coniglio 2007c, 2009, 2011): [ ForceP MP [ Force [ . . . [IP . . . MP . . . ] ] ] ] In Force, the particle has access to information about the speaker and the illocutionary force of the clause. At the same time, it transfers its own semantic features to ForceP, thus being able to take scope over the whole clause.
Modal particles, speaker-hearer links, and illocutionary force
289
Once the head Force has been merged, the Agree-operation between the particle and this head takes place. The uninterpretable feature uSpeaker on the particle gets valued. Thus, the particle is anchored to the speaker. At the same time, the particle transfers its own illocutionary feature to the head Force . As a result of this operation, even though the particle did not move from its IP-internal position, it will refer to the speaker and modify the illocutionary force of the clause. This operation does not only take place in main clauses, where force is necessarily anchored to the speaker, but also in illocutionarily independent subordinate clauses, where force is anchored either to the actual speaker or to the potential one, according to the type of clause. Linking MPs to Force is hence crucial in order to anchor them to the speaker’s mental representations. I want to claim that precisely this connection with the projection ForceP syntactically justifies the peculiar properties of MPs, which – as argued in section 2 – di¤er from those of other modal words in at least one important way. It was claimed that, in the modal Urmasse (cf. Abraham 2008a), MPs not only make available the source of information and the speaker’s assessment of the proposition, but also provide information about the speaker’s estimate of the knowledge awareness of the addressee (cf. Abraham 2008a,b). Since such information finds its syntactic encoding in ForceP, the availability of a link to this projection is necessary for MPs to be licensed.
6. Conclusions In conclusion, we have seen that not only main clauses, but also certain subordinate clauses have illocutionary force. It was shown that MPs are crucial criteria that allow us to distinguish root and embedded clauses. MPs are to be considered indicators of the speaker’s commitment to the utterance (and of her assessment of the hearer’s beliefs, cf. Abraham 2008b) and, consequently, of the illocutionarily independent status of the clause. This explains the peculiar properties of MPs within the class of the modal words. Many issues have not been discussed here. For example, similar observations can be put forward for MPs in other languages – both Germanic and non-Germanic ones. See, for instance, Cardinaletti (2010) and Coniglio (2008)
290
Marco Coniglio
for Italian, Coniglio and Zegrean (2010) for Romanian, and Meisnitzer (2008) for Romance languages in general. This analysis can therefore be extended also to languages that do not display a German-like middle field.23 Moreover, there are some languages which display CP-particles.24 It would be interesting to see whether the distribution of these elements is the same as in the case of German IP-particles and which are the common properties and specific di¤erences between the two types of particles. The underlying question should be whether there is a class of MPs (or of equivalent lexical elements) in all languages or not. From a cognitive perspective, it is in fact a fundamental issue to understand whether these syntactic categories really are language-specific and, if so, which alternative linguistic tools MP-less languages resort to.
Acknowledgements I am very grateful to Werner Abraham and Elisabeth Leiss for their invaluable support. Also, I am deeply indebted to Anna Cardinaletti, Guglielmo Cinque, Manfred Krifka and Roland Hinterho¨lzl for their constant help with my research. Many special thanks go to Bryan Leferman and Iulia Zegrean for kindly volunteering to read and comment a draft of my paper, to Claudia Katharina Mertens for her German grammaticality judgments, and to all the audience members at the SLE Meeting 2008 in Lisbon ‘‘Workshop on Theory of Mind or Relevance – linguistic description and communicative explanation’’.
23. This was assumed to be a special property, which allows for the presence of MPs in a language (cf. fn. 7). Notice, however, that Italian MPs usually occur in a middle-field-like structure, as pointed out in Coniglio (2008). 24. Cantonese, for instance (cf. Law 2002). Its sentence-final particles are assumed to be generated in the CP. Another example is Romanian (cf. Coniglio and Zegrean 2010).
Modal particles, speaker-hearer links, and illocutionary force
291
References Abraham, Werner 1991 The grammaticization of the German modal particles. In: Elizabeth Closs Traugott and Bernd Heine (eds.), Approaches to Grammaticalization, Volume 2, 331–380. Amsterdam/Philadelphia: John Benjamins. Abraham, Werner 1995 Wieso stehen nicht alle Modalpartikel in allen Satzformen? Die Nullhypothese. Deutsche Sprache 23: 124–146. Abraham, Werner 2008a Die Urmasse von Modalita¨t und ihre Ausgliederung. Modalita¨t anhand von Modalverben, Modalpartikel und Modus. Was ist das Gemeinsame, was das Trennende, und was steckt dahinter? In: Werner Abraham and Elisabeth Leiss (eds.), Modalita¨t. Epistemik und Evidentialita¨t bei Modalverb, Adverb, Modalpartikel und Modus, 251–302. (Studien zur deutschen Grammatik 77.) Tu¨bingen: Stau¤enburg. Abraham, Werner 2008b Fremdbewusstseinsabgleich in Syntax und Semantik. Paper read at the Annual Societas Linguistica Europaea Meeting 2008, Lisbon, Workshop on Theory of Mind or Relevance – linguistic description and communicative explanation. Abraham, Werner 2009 Illocutive force is speaker and information source concern. What type of syntax does the representation of speaker deixis require? Templates vs. derivational structure? Ms., University of Vienna. Asbach-Schnitker, Brigitte 1977 Die Satzpartikel wohl. Eine Untersuchung ihrer Verwendungsbedingungen im Deutschen und ihrer Wiedergabemo¨glichkeiten im Englischen. In: Harald Weydt (eds.), Aspekte der Modalpartikeln: Studien zur deutschen Abto¨nung, 38–61. Tu¨bingen: Max Niemeyer. Austin, John Langshaw 1956 Ifs and cans. Proceedings of the British Academy 42: 109–132. Bayer, Josef forthcoming From Modal Particle to Interrogative Marker: A Study of German denn. Ms., University of Constance. Bierwisch, Manfred 1980 Semantic structure and illocutionary force. In: John Rogers Searle, Ferenc Kiefer and Manfred Bierwisch (eds.), Speech Act Theory and Pragmatics, 1–35. Dordrecht: Reidel. Borst, Dieter 1985 Die a‰rmative Modalpartikeln ,,doch‘‘, ,,ja‘‘ und ,,schon‘‘. Ihre Bedeutung, Funktion, Stellung und ihr Vorkommen. Tu¨bingen: Max Niemeyer.
292
Marco Coniglio
Brauße, Ursula 1994 Lexikalische Funktionen der Synsemantika. Tu¨bingen: Gunter Narr. Cardinaletti, Anna 2007 Fu¨r eine syntaktische Analyse von Modalpartikeln. In: EvaMaria Thu¨ne and Franca Ortu (eds.), Gesprochene Sprache – Partikeln, 89–101. Frankfurt: Peter Lang. Cardinaletti, Anna and Michael Starke 1999 The typology of structural deficiency: A case study of the three classes of pronouns. In: Henk van Riemsdijk (eds.), Clitics in the Languages of Europe. 145–233, Berlin/New York: Mouton de Gruyter. Cardinaletti, Anna 2010 German and Italian Modal Particles and Clause Structure. To appear in: Theresa Biberauer and Michelle Sheehan (eds.), The Linguistic Review (special edition on particle syntax). Cinque, Guglielmo 1999 Adverbs and Functional Heads: A Cross-Linguistic Perspective. Oxford: Oxford University Press. Coniglio, Marco 2005 Deutsche Modalpartikeln: eine syntaktische Analyse. MA thesis, University of Venice. Coniglio, Marco 2006 German Modal Particles in the Functional Structure of IP. University of Venice Working Papers in Linguistics 16: 57–95. Coniglio, Marco 2007a Deutsche Modalpartikeln: ein Vorschlag zu ihrer syntaktischen Analyse. In: Eva-Maria Thu¨ne and Franca Ortu (eds.), Gesprochene Sprache – Partikeln, 103–113. Frankfurt: Peter Lang. Coniglio, Marco 2007b German modal particles in the IP-domain. Rivista di grammatica generativa 32: 3–37. Coniglio, Marco 2007c German Modal Particles in Root and Embedded Clauses. University of Venice Working Papers in Linguistics 17: 109–141. Coniglio, Marco 2008 Modal particles in Italian. University of Venice Working Papers in Linguistics 18: 91–129. Coniglio, Marco 2009 Deutsche Modalpartikeln in Haupt- und Nebensa¨tzen. In: Werner Abraham and Elisabeth Leiss (eds.), Modalita¨t. Epistemik und Evidentialita¨t bei Modalverb, Adverb, Modalpartikel und Modus, 191–221. (Studien zur deutschen Grammatik 77.) Tu¨bingen: Stau¤enburg.
Modal particles, speaker-hearer links, and illocutionary force
293
Coniglio, Marco 2011 Die Syntax der deutschen Modalpartikeln: ihre Distribution und Lizenzierung in Haupt- und Nebensa¨tzen. (Studia Grammatica 73.) Berlin: Akademie Verlag. Coniglio, Marco and Iulia Zegrean 2010 Splitting up Force, evidence from discourse particles. University of Venice Working Papers in Linguistics 20: 7–34. Dancygier, Barbara and Eve Sweetser 2005 Mental Spaces in Grammar: Conditional Constructions. (Cambridge Studies in Linguistics 108.) Cambridge: Cambridge University Press. Diewald, Gabriele 1999 Die Modalverben im Deutschen. Grammatikalisierung und Polyfunktionalita¨t. Tu¨bingen: Niemeyer. Doherty, Monika 1987 Epistemic Meaning. Berlin: Springer. Do¨ring, Sophia 2007 Zur Kontextverschiebung bei deutschen Diskurspartikeln. Bachelorarbeit, Humboldt University of Berlin. Egg, Markus this volume Discourse particles at the semantics-pragmatics interface. Grosz, Patrick 2005 ‘‘Dn’’ in Viennese German. The Syntax of a Clitic Version of the Discourse Particle ‘‘denn’’. MA thesis, University of Vienna. Haegeman, Liliane 2002 Anchoring to speaker, adverbial clauses and the structure of CP. Georgetown University Working Papers in Theoretical Linguistics 2: 117–180. Haegeman, Liliane 2004a Topicalization, CLLD and the left periphery, ZAS Papers in Linguistics 35. Proceedings of the Dislocated Elements Workshop, Zentrum fu¨r Allgemeine Sprachwissenschaft Berlin, November 2003, Volume 1, 157–192. Haegeman, Liliane 2004b The syntax of adverbial clauses and its consequences for topicalisation. Antwerp Papers in Linguistics 107: 61–90. Haegeman, Liliane 2006 Conditionals, factives and the left periphery. Lingua 116: 1651– 1669. Haegeman, Liliane 2008a The syntax of adverbial clauses. Ms., University of Ghent. Haegeman, Liliane 2008b The internal syntax of adverbial clauses. Ms., University of Ghent.
294
Marco Coniglio
Hentschel, Elke 1986 Funktion und Geschichte deutscher Partikeln. Ja, doch, halt und eben. (Reihe Germanistische Linguistik 63.) Tu¨bingen: Niemeyer. Heycock, Caroline 2005 Embedded root phenomena. In: Martin Everaert and Henk van Riemsdijk (eds.), The Blackwell Companion to Syntax, Volume 2, Chapter 23, 174–209. Oxford: Blackwell. Hooper, Joan and Sandra A. Thompson 1973 On the applicability of Root Transformations. Linguistic Inquiry 4: 465–97. Jacobs, Joachim 1986 Abto¨nungsmittel als Illokutionstypmodifikatoren. Groninger Arbeiten zur Germanistischen Linguistik 27: 100–111. Jacobs, Joachim 1991 On the semantics of modal particles. In: Werner Abraham (ed.), Discourse particles. Descriptive and theoretical investigations on the logical, syntactical, and pragmatic properties of discourse particles in German, 141–162. Amsterdam/Philadelphia: John Benjamins. Jakobson, Roman 1971 Shifters, verbal categories, and the Russian verb. In: Selected Writings II, 130–147. The Hague: Mouton. Jespersen, Otto 1922 Language; Its Nature, Development And Origin. London: Allen and Unwin. Johnson-Laird, Philip Nicholas 1986 Conditionals and mental models. In: Elizabeth Closs Traugott, Alice ter Meulen, Judy Snitzer Reilly and Charles A. Ferguson (eds.), On Conditionals, 55–75. Cambridge: Cambridge University Press. Kiefer, Ferenc 1987 On defining modality. Folia Linguistica 21: 67–94. Ko¨nig, Ekkehard 1986 Conditionals, concessive conditionals and concessives: Areas of contrast, overlap and neutralization. In: Elizabeth Closs Traugott, Alice ter Meulen, Judy Snitzer Reilly and Charles A. Ferguson (eds.), On Conditionals, 229–246. Cambridge: Cambridge University Press. Ko¨nig, Ekkehard and Johan van der Auwera 1988 Clause integration in German and Dutch conditionals, concessive conditionals, and concessives. In: John Haiman and Sandra A. Thompson (eds.), Clause Combining in Grammar and Discourse, 101–133. Amsterdam/Philadelphia: John Benjamins.
Modal particles, speaker-hearer links, and illocutionary force
295
Kratzer, Angelika 1981 The notional category of modality. In: Hans-Ju¨rgen Eikmeyer and Hannes Rieser (eds.), Words, Worlds, and Contexts. New approaches in Word Semantics, 38–74. Berlin/New York: Mouton de Gruyter. Law, Ann 2002 Cantonese sentence-final particles and the CP domain. UCL Working Papers in Linguistics 14: 375–398. Leiss, Elisabeth 2008a Drei Spielarten der Epistemizita¨t, drei Spielarten der Evidentialita¨t und drei Spielarten des Wissens. In: Werner Abraham and Elisabeth Leiss (eds.), Modalita¨t. Epistemik und Evidentialita¨t bei Modalverb, Adverb, Modalpartikel und Modus, 3–24. (Studien zur deutschen Grammatik 77.) Tu¨bingen: Stau¤enburg. Leiss, Elisabeth 2008b Epistemicity, evidentiality, and Theory of Mind. Paper read at the Annual Societas Linguistica Europaea Meeting 2008, Lisbon, Workshop on Theory of Mind or Relevance – linguistic description and communicative explanation. Meibauer, Jo¨rg 1994 Modaler Kontrast und konzeptuelle Verschiebung: Studien zur Syntax und Semantik deutscher Modalpartikeln. (Linguistische Arbeiten 314.) Tu¨bingen: Max Niemeyer. Meisnitzer, Benjamin 2008 Modality in communication in Romance languages: modal verbs and particles. Paper read at the Annual Societas Linguistica Europaea Meeting 2008, Lisbon, Workshop on Theory of Mind or Relevance – linguistic description and communicative explanation. Ormelius-Sandblom, Elisabet 1997a The modal particle schon: Its syntax, semantics and pragmatics. In: Toril Swan and Olaf Jansen Westwik (eds.), Modality in Germanic Languages. Historical and Comparative Perspectives, 75–131. (Trends in Linguistics, Studies and Monographs 99.) Berlin/New York: Mouton de Gruyter. Ormelius-Sandblom, Elisabet 1997b Die Modalpartikeln ,,ja‘‘, ,,doch‘‘ und ,,schon‘‘. Zu ihrer Syntax, Semantik und Pragmatik. Lund: Almqvist and Wiksell International. Reis, Marga 1997 Zum syntaktischen Status unselbsta¨ndiger Verbzweit-Sa¨tze. In Christa Du¨rscheid, Karl-Heinz Ramers, and Monika Schwarz (eds.), Sprache im Fokus. Festschrift fu¨r Heinz Vater zum 65. Geburtstag, 121–144. Tu¨bingen: Niemeyer.
296
Marco Coniglio
Rizzi, Luigi 1997 Rizzi, Luigi 2001
Rizzi, Luigi 2004
Ross, John R. 1973
The Fine Structure of the Left Periphery. In: Liliane Haegeman (ed.), Elements of Grammar, 281–373. Amsterdam: Kluwer. On the Position Int(errogative) in the Left Periphery of the Clause. In: Guglielmo Cinque and Giampaolo Salvi (eds.), Current Studies in Italian Syntax. Essays o¤ered to Lorenzo Renzi, 287–296. Amsterdam/New York: Elsevier. Locality and Left Periphery. In: Adriana Belletti (ed.), Structures and Beyond. The Cartography of Syntactic Structures, Volume 3, 223–251. Oxford: Oxford Studies in Comparative Syntax.
The penthouse principle and the order of constituents. In: Claudia Corum, Thomas Cedric Smith-Stark and Ann Weiser (eds.), You Take the High Node and I’ll Take the Low Node. Papers from the Comparative Syntax Festival. 397–422. Chicago: Chicago Linguistic Society. Sperber, Dan and Deirdre Wilson 1986 Relevance. Communication and cognition. Oxford: Blackwell. Thurmair, Maria 1989 Modalpartikeln und ihre Kombinationen. (Linguistische Arbeiten 223.) Tu¨bingen: Max Niemeyer. Vries, Mark de 2002 The Syntax of Relativization. Amsterdam: LOT. Zimmermann, Malte 2004a Zum Wohl: Diskurspartikeln als Satztypmodifikatoren. Linguistische Berichte 199: 1–35. Zimmermann, Malte 2004b Discourse particles in the left periphery. ZAS Papers in Linguistics 35, Proceedings of the Dislocated Elements Workshop, Zentrum fu¨r Allgemeine Sprachwissenschaft Berlin, November 2003, Volume 2, 543–566.
Discourse particles at the semantics-pragmatics interface1 Markus Egg Humboldt-Universita¨t Berlin Abstract This paper presents an approach to the semantics of discourse particles that aims at providing very specific analyses which are nevertheless applicable to a wide range of uses of these particles. It is shown that these competing goals can be pursued simultaneously by taking into account that discourse particles can refer semantically not only to the utterances they form part of or to utterances previously made, but also to felicity conditions of such utterances. This approach is first illustrated by a detailed analysis of the unstressed uses of the discourse particle schon, and is then shown to be fruitfully applicable to stressed uses of the particle as well. 1. Introduction The topic of this paper is the semantics of discourse particles, in particular, of the particle schon. The paper proposes an approach to the semantics of discourse particles that yields very specific analyses which are nevertheless applicable to a wide range of uses of the particles in question. Before entering the discussion of schon it seems advisable to define the group of particles that are relevant for this paper and that are dubbed ‘discourse particles’, because the terminology is far from transparent. These particles are morphologically invariant, they do not contribute to truthconditional semantics, they are (in German) restricted to the Mittelfeld in declarative sentences2, and they have no status as an independent constituent (they cannot be moved or be the target of wh-questions). This is in 1. The author wishes to thank Werner Abraham, Volker Gast, and Elisabeth Leiss for extensive feedback on earlier versions of this paper. 2. This is the position in the sentence structure between the finite element in the 2nd position and any nonfinite verb forms.
298
Markus Egg
line with Thurmair (1989) and Abraham (2000); unlike these authors, however, I do not consider stressed uses of the relevant elements to fall outside the class of discourse particles; see Section 4.3. To avoid terminological confusion, let me point out that these particles are sometimes referred to as ‘modal particles’, whereas ‘discourse particle’ has been used to denote another class of particles that is excluded from consideration here, viz., the group of particles like a¨h ‘eh’ or also ‘well’, which are used to handle the flow of conversation, turn taking and problems of language production. 1.1. Basic intuitions This section outlines the basic intuitions that shape the proposed analysis and relates it to the Theory of Mind/Foreign Consciousness Alignment (FCA) outlined by Abraham and Leiss (this volume). There are two ways in which discourse particles contribute to semantic construction at the discourse level. First, these particles characteristically refer to the common ground (CG), the background assumptions that are taken for granted by the interlocutors for the purpose of the communication. Hence, a comprehensive description of their semantic contribution must account for the ways in which they are used to manage the common ground. Discourse particles can e.g. be used to negotiate potential updates of the CG, or to remind an interlocutor of the fact that a specific proposition is already part of the CG. Both functions can simultaneously be illustrated by the schon-utterance in (1):3 (1) A: Ich habe nicht genug fu¨r die Pru¨fung gelernt. I have not enough for the exam studied. scha¤en. B: Du wirst es schon you will it SCHON succeed ‘A: I haven’t studied hard enough for the exam. B: You will pass nevertheless.’ My reconstruction of the CG-relevance of schon in (1) is the following. B’s reply acknowledges A’s statement, which consequently updates the common ground. At the same time, this reply states that even though one could deduce from A’s confession that he is going to fail the exam, B still 3. The English glosses are intended as rough paraphrases of the German example sentences, not as their functional equivalents. This holds in particular of my attempts to mimick the e¤ect of the discourse particles in the glosses.
Discourse particles at the semantics-pragmatics interface
299
believes that this consequence will not come to pass. This deduction would be very straightforward, because the inference pattern from insu‰cient preparation to failing is itself part of the CG. In more general terms, the schon-utterance reacts to a statement by accepting it yet pointing out that a very natural inference from this statement (with CG status) is not applicable, thus restricting the range of potential consequences of A’s statement. This analysis spells out the role of the common ground reference for relational particles such as schon and doch in the alignment of the individual beliefs of the interlocutors (FCA; Abraham and Leiss 2011). The crucial point is that interlocutors use inference patterns from the common ground as the basis of synchronising individual beliefs. If a speaker B reacts to a statement q of another speaker A in terms of an utterance schon p, he refers to an inference pattern in the common ground to first communicate his assumptions about A’s own background: B would not point out that the relevant inference pattern does not hold unless he assumed that A considers this pattern applicable to his statement q, and thus that A believes at least tentatively that Bp, the contrary of B’s schon-utterance, holds. In a second step, this amounts to suggesting that A should revise this tentative belief, because there are factors that block the application of the inference pattern (otherwise p, the opposite of the outcome of applying the pattern to A’s utterance could not hold). This revision would synchronise the individual beliefs of A and B and successfully complete the negotiation about the way in which the CG should be developed further. The line of research in this paper continues previous research on the semantics of discourse particles (Abraham, 1991; Ko¨nig, 1997; Karagjosova, 2004; Zimmermann, 2011). Such analyses of discourse particles will yield further insights into the exact nature of the common ground, a hotly debated topic; see e.g. Stalnaker (2002) or Abbott (2008). The role of discourse particles as markers for the way in which specific pieces of information relate to the CG means that they can be used as a window to the CG, because analysing their semantics helps deepen our knowledge about the exact nature of the CG. For example, if the fact that a specific piece of information is part of the CG (as signalled e.g. by the discourse particle ja) can be informative in itself, this has repercussions for an analysis of the common ground: It would entail that CG information may be di¤erently activated for di¤erent interlocutors in a communicative situation, and that ja-utterances can be explained as attempts to activate or foreground CG relevant informa-
300
Markus Egg
tion. Analogous conclusions can be drawn from the described intuitions about schon: The CG cannot be closed under deduction if schon is indeed used to accept utterances as updates for the CG while at the same time blocking potential inferences for the CG, even though these inferences are suggested by the CG itself. Many discourse particles contribute to discourse-level semantics, too, viz., by relating discourse segments.4 Investigating their semantics thus contributes to the analysis of the connection between sentence semantics and discourse semantics, a topic that deserves further attention in spite of the pioneering work of e.g. Asher and Lascarides (2003) and van Leusen (2007). For the discourse particle schon, the relation between the two segments is one of denial of expectation, which comes about by simultaneously accepting a proposition and denying a reasonably expectable consequence of this proposition. This insight highlights the close connection between discourse particles and discourse structure:5 The notion of denial of expectation figures prominently in both fields. In particular, it is used by Lagerwerf (1998) to characterise the discourse relation of concession. Therefore, investigations of discourse particles are highly relevant for a comprehensive account of discourse structure. They complement other work on explicit discourse structure marking, which so far has mainly focussed on conjunctions (Sanders et al. 1992, Knott and Sanders 1992, Knott and Sanders 1998, Miltsakaki et al. 2004) and discourse-relevant adverbials such as then (Webber et al. 2003). It will be argued in this paper in Section 3 that discourse particles have the same discourse-structuring potential as these adverbials. 1.2. Semantic and pragmatic aspects of discourse particles The next step in the analysis is to identify the semantics of schon on the basis of the interpretation of schon-utterances. This interpretation emerges from the interaction of the semantic contribution of the discourse particle with general pragmatic principles. 4. This applies to discourse particles like schon, noch, or halt, as opposed e.g. to ja. In the remainder of the paper, I will concentrate on the first group of particles. 5. Such a discourse structure emerges because so-called discourse relations (e.g., cause, elaboration, or contrast) link minimal discourse units (roughly, clauses), which combines them into larger discourse units; see e.g. Mann and Thompson (1988) or Webber (2004).
Discourse particles at the semantics-pragmatics interface
301
In particular, the way in which (1) is used to make predictions about the hearer’s assumptions and to negotiate these assumptions and the common ground can be put down to pragmatics, it is due to the dialogue situation. To see this, consider a monologic version of this example: (2) Ich habe nicht genug fu¨r die Pru¨fung gelernt, werde es aber I have not enough for the exam studied will it but schon scha¤en. SCHON succeed ‘I haven’t studied hard enough for the exam, but will pass nevertheless.’ In this example, the speaker does not use the doch-utterance to suggest that the speaker of the preceding statement (which he happens to be himself in (2)) holds a belief deducable from this statement in terms of an inference pattern from the common ground (viz., that the speaker of the first utterance will fail). In other words, this part of the interpretation of (1) is not part of the semantics of schon, while the reference to inference patterns from the common ground and the subsequent blocking of a potential inference from this pattern (by stating the opposite of the result of this potential inference) is. These preliminary remarks on the question of where to draw the boundary between semantics and pragmatics are intended to motivate a ‘parsimonious’ approach to the semantics of discourse particles that formulates semantic descriptions in such a way that they do not comprise aspects that could be explained in terms of general pragmatic processes. However, the danger of pursuing a ‘parsimonious’ approach to semantics is ending up with a semantic description for discourse particles that fails to distinguish the meanings of di¤erent (not synonymous) lexical items. For schon, a case in point is the discourse marker trotzdem ‘nevertheless’, which also introduces a denial of expectation. Since both schon and trotzdem share the constellation in which an inference pattern is not applicable (because the contrary of the plausibly expectable consequence is asserted), their di¤erence must lie elsewhere. There are two major di¤erences between discourse particles like schon and discourse markers like trotzdem. First, only particles can link an utterance to a non-adjacent element of the preceding discourse, i.e., they are not subject to the Right Frontier Constraint (RFC) in the definition of Webber (1988). Thus, B can react to the first clause of A’s utterance (3a)
302
Markus Egg
in terms of (3b) to rule out Peter’s failing as a potential consequence of A’s claim that Peter did not study hard enough, but not in terms of (3c), in which schon is replaced by trotzdem: (3) a.
A: Obwohl Peter nicht genug fu¨r die Pru¨fung gelernt hat, although Peter not enough for the exam studies has geht es ihm gut, denn er wird im na¨chsten Monat goes it him well because he will in.the next month heiraten. marry ‘A: Although Peter has not studied hard enough for the exam, he is fine, because he will get married next month.’
b.
bestehen. B: Er wird die Pru¨fung schon he will the exam SCHON pass ‘B: He will pass the exam nevertheless.’
c.
B: *Er wird die Pru¨fung trotzdem bestehen. he will the exam nevertheless pass ‘B: He will pass the exam nevertheless.’
(3c) is inacceptable as a reaction to (3a), because it suggests that Peter’s failing the exam is ruled out even though it were an expected consequence of his well-being or his upcoming marriage. The second di¤erence between this schon und trotzdem lies in the fact that only schon-utterances can react to non-declarative utterances or be non-declarative themselves (see the discussion of these examples in the Sections 3.2 and 3.3): (4)
A: Was soll ich meinen Eltern zu Weihnachten schenken? what shall I my parents for Christmas give B: Dir wird schon/*trotzdem etwas einfallen. you will SCHON/nevertheless something think of ‘A: What should I give to my parents for Christmas? B: You will be able to think of something, don’t worry.’
gerne sterben? (5) Wer wu¨rde schon/*trotzdem who would SCHON/nevertheless happily die ‘Who would like to die?’
Discourse particles at the semantics-pragmatics interface
303
1.3. Previous accounts Research on all kinds of explicit discourse markers is confronted with a very wide and diverse range of concrete usages of these markers in context, which raises the question of whether it is sensible at all to presume that they contribute uniformly to the semantics of larger discourses. For analyses of discourse particles this means that in order to be a serious competitor to analyses that regard them as polysemous (e.g., GornikGerhardt 1981 for schon), uniform semantic interpretations of the particles must comply with two seemingly conflicting requirements: They must be su‰ciently specific to allow the derivation of the interpretation of concrete examples, and at the same time su‰ciently general to cover a wide range of concrete usages. Many previous approaches to discourse particles in general and schon in particular focus on the second of these requirements. For example, Borst (1985), Thurmair (1989), Ko¨nig (1997), Karagjosova (2004), or OrmeliusSandblom (1997) capture a wide range of usages of discourse particles in their analyses by assuming rather general meanings. For the case of schon, such a general meaning can be described as a two-place relation between the utterance of which schon is a part and information from the context. Often this information is expressed by a previous utterance to which the schon-utterance is a reaction, as e.g. in (1). The majority of the attempts to describe the semantic contribution of the discourse particle schon in terms of a uniform basic meaning characterise it as a simultaneous confirmation and restriction of a previous statement: Borst (1985) states that the schon-utterance a‰rms and at the same time restricts a previous statement. The intuition of Thurmair (1989) that ‘parts of the preceding utterance are confirmed, but the overall validity of the utterance is restricted’ is represented in terms of a feature validity restriction. Ko¨nig (1991) o¤ers a less specific interpretation, viz., that schon in an utterance expresses that the information of the utterance is not a matter of course in the given context. These analyses fit in with the semantic representation of schon assumed in this paper: A proposition p is restricted if reasonably expectable consequences are ruled out, and a proposition q is unexpected, if it is the negation of a potential consequence of a true statement p. Ormelius-Sandblom (1997) models the meaning of schon in a rather special way, as a function from statements p to the statement that ‘it is not a fact that Bp’. I will show that this analysis runs into problems with some uses of schon, even though it is quite underspecified.
304
Markus Egg
But for simple cases like (6) [¼(1)], all these descriptions are applicable. Here, B’s reply confirms A’s confession while at the same time restricting it by pointing out that a very natural consequence (viz., failing) will not obtain: (6)
A: Ich habe nicht genug fu¨r die Pru¨fung gelernt. I have not su‰ciently for the exam studied B: Du wirst es schon scha¤en. you will it SCHON succeed ‘A: I haven’t studied hard enough for the exam. B: You will pass nevertheless.’
These descriptions of schon in the literature capture intuitions about its semantic contribution, but are not intended to be specific enough to spell out in detail the way in which the meaning of schon interacts with the meaning of utterances in the semantic construction of larger discourses. In particular, a semantic ingredient that is neglected even in the more specific accounts of the meaning of schon is that a declarative main clause schon-utterance must refer to something positive. A variant of (6) in which the polarities are reversed sounds strange because it would only be acceptable if A’s failure is regarded as positive: (7)
A: Ich habe genug fu¨r die Pru¨fung gelernt. I have su‰ciently for the exam studied B: Du wirst es schon nicht scha¤en. you will it SCHON not succeed ‘A: I have studied hard enough for the exam. B: You will not pass nevertheless.’
This explains why schon-utterances often have a soothing flavour, because they point out an unexpected positive fact in the light of information that suggests the opposite (in our example (6), A’s confession). This e¤ect can be seen very clearly in the jocular (8), whose literal interpretation is contradictory (because failing is not positive), and in whose eventual interpretation schon enforces a non-literal interpretation of schiefgehen ‘fail’ as its antonym, i.e., the eventual interpretation of (8) is ‘it will go fine, don’t worry’: (8) Wird schon schiefgehen. will SCHON go wrong ‘It will go wrong, don’t worry.’
Discourse particles at the semantics-pragmatics interface
305
While examples such as (6) show that even the most specific semantic analyses of schon in the literature capture only part of its semantic contribution, even the less specific analyses are still too restrictive in that they do not account for the whole range of uses of schon. As an example, consider (9), in which there is neither a restriction of the validity of a preceding utterance nor information that is not a matter of course: (9)
A: Karen hat sich sehr u¨ber ihre Niederlage gea¨rgert. Karen has herself very about her defeat be angry B: Wer verliert schon gern? who loses SCHON gladly ‘A: Karen was very angry at her defeat. B: Well, who likes losing in the first place?’
In this example, the schon-reply (a rhetorical question, semantically, the statement that no one likes losing) does not restrict the validity of A’s statement in any way. On the contrary, the schon-expression even extends the validity of A’s statement by pointing out that it is just a special case of a general phenomenon (Karen is just like anyone else in that she hates losing). What is more, the particle in (9) is used in a rhetorical question, which indicates that the involved statement (that no one likes losing) is already part of the common ground (see Egg 2007 for details). Therefore, schon does not express that the information of its utterance is not a matter of course, i.e., (9) is not captured by the semantic description of the particle as proposed by Ko¨nig (1991). Similarly, representing the semantic contribution of schon as a function from statements p to the statement that ‘it is not a fact that Bp’ (Ormelius-Sandblom, 1997) would not appropriately describe the e¤ect of the particle in (9): The statement it is part of definitely is a fact. Worse still, schon can also show up as a discourse particle in imperatives, e.g., in (10): (10) Nun mach schon! now hurry SCHON ‘Come on!’ But imperatives refer to states of a¤airs that do not yet obtain or have not yet happened, otherwise they would be infelicitous. These states of a¤airs are described in terms of a proposition p and, contrary to the prediction of Ormelius-Sandblom (1997), it is a fact that Bp.
306
Markus Egg
1.4. Plan of attack In this paper, I will present an analysis of discourse particles focussing on schon that o¤ers the necessary generality to account for a wide range of uses while at the same time being specific enough for a semantic construction for discourses that comprise these particles. My attempt to pursue these two competing goals simultaneously (developed in Egg 2010 for the discourse particle doch) is based on the following observations: Most schon-utterances, e.g., in (6), are used as a reaction to a previous utterance (of a di¤erent or of the same speaker). At the same time, schon is inherently relational, i.e., semantically, it relates two propositions. Sometimes, e.g., in (6), the semantic arguments of the particle can be identified with the meanings of these two utterances, which I will call ‘involved utterances’. This is the simplest case, which I will use to develop an account of the semantics of the discourse particle schon. Based on this account it is then possible to tackle more di‰cult cases, starting with those in which the involved utterances can no longer directly provide the semantic arguments of schon, e.g., in (11) [¼(4)]: (11)
A: Was soll ich meinen Eltern zu Weihnachten schenken? what shall I my parents for Christmas give B: Dir wird schon etwas einfallen. you will SCHON something think of ‘A: What should I give to my parents for Christmas? B: You will be able to think of something, don’t worry.’
B’s reply is a reaction to the statement implicit in A’s question that A does not know what to give to his parents for Christmas (otherwise, A would not have asked). B does not refute this statement but rules out one of its potential consequences, viz., that A’s state of ignorance will persist. Thus, the implicit statement is one of the arguments of doch in (11), although it di¤ers from the meaning of A’s utterance (a question, whose meaning is not a proposition at all). Examples such as (11) thus show that the semantic arguments of a discourse particle need not coincide with the meanings of the involved utterances. This observation is reflected in the nomenclature used in this paper: Utterances that comprise discourse particles are called ‘p(article)-utterances’ (or ‘part-utterances’, for a given particle part.) If a p-utterance is used to react to a preceding utterance, the other utterance is called ‘a(ntecedent)utterance’.
Discourse particles at the semantics-pragmatics interface
307
The meanings of a- and p-utterance are to be distinguished from the semantic arguments of the discourse particle, which are called ‘a-proposition’ and ‘p-proposition’, respectively. A failure to distinguish these would enforce one to assume extremely weak meanings for discourse particles and require extra assumptions to explain the use of discourse particles with nondeclarative utterances. For example, in the case of schon, the meaning of the p-utterance can potentially contradict or generalise the meaning of the a-utterance, as illustrated by (6) and (9), respectively. Hence, a semantic analysis of schon based on the literal meanings of these utterances only would have to postulate an extremely weak meaning for the particle. I presume that this observation explains some of the problems of pinning down the semantics of these particles in the literature. Using pairs of utterances to derive the semantics of discourse particles is fine in principle (I will proceed in this manner, too), but must be complemented by the distinction of the meanings of the involved utterances and the semantic arguments of discourse particles. The intuition that the semantic arguments of discourse particles need not be identical to the meanings of the involved utterances is related to suggestions to let discourse relations relate either to the content of the discourse segments that they link or to the corresponding intensions of the speaker or the intended e¤ects on the hearer (see Sweetser 1990 and Knott 2001). But distinguishing the semantic arguments of a discourse particle from the meanings of the two involved utterances is not more than a first step. I will base my analysis on simple examples in which the involved utterances provide the semantic arguments of schon, like in (6). As soon as the meaning of schon is identified, I will then tackle the more di‰cult cases. These cases will show that the choice of semantic arguments of a discourse particle is restricted to the meanings of the involved utterances and utterancerelated statements, in particular, those that express felicity conditions of these utterances. This analysis thus integrates insights of Jacobs (1991) and Karagjosova (2004) that discourse particles can make reference to, and hence be sensitive to, di¤erent types of illocutionary acts. The chosen approach to the semantics of discourse particles therefore assigns a di¤erent role to these particles in discourse processing. They do not simply relate two independently given utterances, they indicate a semantic relation between a first proposition (which is very often but not always the interpretation of the p-utterance) to another proposition (the a-proposition), which must be selected from the context of the p-utterance.
308
Markus Egg
In this respect, discourse particles function like adverbials like then, which point anaphorically at preceding discourse segments (Webber et al. 2003). One could also relate this intuition to the claim of Ko¨nig and Requardt (1991) that such particles are ‘metapragmatic instructions’ that relate the p-proposition to a suitable proposition in the context. The plan of the paper is the following. I will first o¤er some background ideas and assumptions concerning discourse particles and their relation to the common ground and delimit the scope of the present analysis in Section 2. Then I will analyse the semantics of (unstressed) schon in detail in Section 3. Finally, Section 4 will o¤er an outlook to further research.
2. Background and scope of the analyis This section lays the ground of the proposed analysis by first introducing the background assumptions on which it is built and then delimiting its scope to unstressed uses of the particle. Like in much previous work, the meaning of discourse particles is crucially characterised by the way in which they refer to the common ground (CG) (Ko¨nig, 1997; Karagjosova, 2004; Weinrich, 2005; Zimmermann, 2011). Common ground and the interlocutors’ individual backgrounds are modelled in terms of common or individual belief (Stalnaker, 2002). Individual belief is defined as the set of propositions that are true in all possible worlds compatible with the individual’s beliefs; common belief, with the set of propositions believed by all members of the respective group of believers.6 Even preliminary accounts of utterances with discourse particles suggest that CG material might have di¤erent activation status and that the CG is not closed under deduction. Otherwise, it would not make sense to point out that information is already part of the CG (the e¤ect of the discourse particle ja), or that potential inferences do not apply to a proposition in spite of the CG status of the inference (the e¤ect of schon, as it will be analysed in this paper). I model reasoning on the CG and these backgrounds in terms of defeasible deduction (Asher and Lascarides, 2003): Defeasible deduction uses 6. Following Stalnaker, this is an idealisation in that the CG might comprise propositions not shared by the background of every member of the group. This possibility will not be relevant for the interpretation of discourse particles as proposed in this paper, however.
Discourse particles at the semantics-pragmatics interface
309
statements of the form ‘p > q’ ( p defeasibly entails q), which together with p allow one to defeasibly deduce q. This defeasible Modus Ponens applies only if Bq does not already hold and if it is not possible to deduce Bq at the same time (the latter constellation is known as ‘Nixon diamond’; Asher and Lascarides 2003). It is exactly this reference to the common ground that makes the semantics of discourse particles dependent on the context, because the CG is relative to the interlocutor(s) of the respective a- and p-utterances. The shifting e¤ects observed in Do¨ring (2010) illustrate this point; consider e.g. embedding (6) in a report like in (12): (12) A sagte, er habe nicht genug fu¨r die Pru¨fung gelernt. A said he had not su‰ciently for the exam studied B entgegnete, er werde es schon scha¤en. B replied he would it SCHON succeed ‘A said he had not studied hard enough for the exam. B replied that he would pass nevertheless.’ The shifting e¤ect shows up in (12) in that the CG relevant for the interpretation of schon is calculated with respect to A and B, not with respect to speaker and hearer of (12). This CG crucially comprises the inference pattern that insu‰cient preparation for an exam leads to failing the exam (which is taken up by schon). But the speaker of (12) need not commit himself to this inference pattern. Similarly, the positive evaluation of the p-proposition is due to B in (12), i.e., the speaker of (12) need not have a positive attitude towards this proposition. The use of defeasible deduction to model the tension between two propositions as indicated by schon is closely related to accounts of the discourse relation of concession in Grote et al. (1997), Oversteegen (1997), Lagerwerf (1998), and Knott (1996). Since schon has a notoriously wide range of meanings and uses, it is necessary to delimit the range of uses to be covered in the proposed analysis before setting out on the analysis. First, since the topic of this paper are discourse particles, I will not cover uses of schon as a temporal adverb, like in (13), and related uses in which the temporal scale is replaced e.g. by a spatial one, as in (14). See e.g. Abraham (1980), van der Auwera (1993), Lo¨bner (1999), Krifka (2000), Greenberg (2009), and Klein (2010) for extended investigations of these uses of schon:
310
Markus Egg
(13) Fritz ist schon weg. Fritz is SCHON gone ‘Fritz is already gone.’ (14) Basel ist schon in der Schweiz. Basel is SCHON in the Switzerland ‘Basel is already in Switzerland.’ Another distinction among the uses of schon (orthogonal to the one between temporal and other scalar uses and discourse-relevant ones) distinguishes a group of focus-sensitive uses that associate with a focussed part of the utterance: (15) Ariadne ist schon GEstern angekommen. (scalar) Ariadne is already yesterday arrived ‘Ariadne arrived already yesterday.’ (16)
A: Ich gehe nicht hin. B: ICH SCHON. (discourse-relevant) I go not there I SCHON ‘A: I will not attend. B: But I will.’
It has been argued in the literature (Krifka, 2000; Fe´ry, 2010) that all uses of schon can be analysed as focus-sensitive. In the analysis o¤ered in the present paper, I will concentrate on those that do not relate to an associated focussed proper part of the schon-utterance, without discussing their potential focus-sensitivity (but see Section 4.3 for the discussion of some focus-sensitive cases). The third relevant distinction is the one between di¤erently stressed uses of schon. Among its discourse-relevant uses we find compulsorily stressed as well as unstressed uses, some allow both patterns. Stressed uses, which are also identified as an important use of the particle in Ko¨nig, Stark, and Requardt (1990), are used for focus-sensitive contrasts, as in (16), for denials of denials, as in (17), for sentence-equivalent confirmations as illustrated in (18) (typically accompanied by a clause stating a restriction of the confirmation): (17)
A: Ame´lie ißt kein Fleisch. B: Sie ißt SCHON Fleisch. Ame´lie eats no meat she eats SCHON meat ‘A: Ame´lie does not eat meat. B: Actually, she does.’
Discourse particles at the semantics-pragmatics interface
(18)
311
A: Ame´lie ißt kein Fleisch. B: SCHON, aber . . . Ame´lie eats no meat SCHON but ‘A: Ame´lie does not eat meat. B: Indeed she does not, but . . .’
Obligatorily unstressed uses occur e.g. in rhetorical questions or imperatives: (19) A: Wen wird Gu¨nther zu seiner Feier einladen? whom will Gu¨nther to his party invite B: Wen wird er schon einladen? (Seine langweiligen whom will he SCHON invite his boring Kollegen natu¨rlich.) colleagues of course ‘A: Who will Gu¨nther invite to his party? B: Well, who will he invite? (His boring colleagues, of course.)’ (20) Nun mach schon! now hurry SCHON ‘Come on!’ However, in some cases, both stressed and unstressed uses are possible, but with a di¤erent meaning: (21)
A: Wie war die Feier? B: Es war schon/SCHON toll. how was the party it was SCHON great ‘A: How was the party? B: It was great’
In (21), B’s answer with unstressed schon indicates that the party just barely classifies as great, whereas the variant with stressed schon introduces no such hedge but explicitly refutes the statement (which must be su‰ciently activated in the utterance situation) that the party was not great. In the remainder of this section, I will investigate unstressed discourserelevant uses of the particle that have no associated focussed element. But the semantic analysis developed for this group is not irrelevant for the other uses of schon. In Section 4 I will show that the analysis proposed for unstressed uses of schon naturally extends to some stressed uses as well. 3. The analysis of schon 3.1. Starting with a simple example This section introduces and motivates the proposed analysis of the semantics of schon, using the simplest case (6), in which the semantic arguments of the
312
Markus Egg
particle are identical to the meaning of the declarative a- and p-utterance, respectively. This example clearly illustrates the intuition on schon sketched in the first section. The schon-utterance confirms the a-utterance, but at the same time, rules out one of its potential consequences, i.e., B accepts A’s claim that A did not study hard enough for the exam, but rules out the tacit common-sense inference (which is part of the CG) that this will make him fail the exam. This denial of expectation characterises the p-proposition (here, that A will pass) as unexpected and somewhat surprising. The fact that A’s passing the exam is evaluated positively can be seen by comparing (6) to (7). In (7), there is likewise a denial of expectation in that, according to the CG, su‰cient preparation for the exam should lead to passing the exam, which is denied by B. The only di¤erence between (6) and (7) lies in the fact that B refers to something pleasant in the former but not in the latter, which accounts for the di¤erence in acceptability between the two. Formally, the meaning of schon is reconstructed as follows. Here p stands for the semantic argument of schon that is related (though not necessarily identical) to the meaning of the p-utterance (analogously for q and the a-utterance): (22) [[schon]] ( p)(q) i¤ both p and q hold, p is evaluated positively, and, according to the common ground, q defeasibly entails Bp (q > Bp) Both the positive evaluation of p as well as the CG status of the defeasible entailment have the status of presuppositions, i.e., violating them renders a schon-utterance infelicitous. This reconstruction of the semantics of schon can also explain why schon-utterances often sound less confrontative than the corresponding utterance without the particle: Either variety introduces a proposition which the hearer might not subscribe to, but using schon admits that assuming the opposite of this proposition is reasonable in principle, since this assumption is suggested by a previous utterance of the hearer (which the spaker accepts) and the common ground. An example of Thurmair’s that illustrates this pattern (adapted) is (23); here p is the (positive) claim that it is not possible that A will be shot, even though the fact that the situation is dangerous (¼q) defeasibly implies the opposite (that he might be shot):7 7. This example shows that the CG must take into account the context of the dialogue, e.g., B’s reaction would not be appropriate in a context in which there are no other people.
Discourse particles at the semantics-pragmatics interface
(23)
313
A: Es ist gefa¨hrlich. it is dangerous B: Es wird schon keiner auf dich schießen. it will SCHON no one at you shoot ‘A: It is dangerous. B: No one will shoot you, don’t worry.’
The formalisation of the semantics of schon is close to the definition of one of the three kinds of contrastive discourse relations in Lagerwerf (1998), which he calls denial of expectation. It furthermore resembles the analyses o¤ered for doch in Karagjosova (2001) and Abraham (1991). While I do agree that schon and doch are similar in that they both introduce the notion of denial of expectation, the crucial di¤erence between the two is the ordering of the arguments. See Section 4 for details. I will conclude this section with an example of an embedded schonutterance, a use which is discussed extensively in Gornik-Gerhardt (1981) and Ormelius-Sandblom (1997). Here the discourse particle suggests that the speaker distances himself personally from the idea expressed in the schon-utterance, here, from the idea of buying upholstery. (24) Wenn wir schon Polster kaufen, dann wenigstens ordentliche. if we SCHON upholstery buy then at least decent ‘If we really buy upholstery, we should at least buy decent stu¤.’ This example can be analysed in terms of the proposed analysis, too. The idea is that the presupposed elements of the semantic contribution of schon are una¤ected by the fact that the schon-utterance is embedded under if. There is thus no factivity for wir schon Polster kaufen ‘we buy upholstery’, but this proposition (our p) is in contrast to another (a-)proposition q from the context, which, according to the CG, would allow one to derive Bp defeasibly. Introducing this thwarted expectation by using schon expresses the notion of distance of the speaker as noted above. For instance, such a q could be a statement that buying furniture is not a good idea. The striking di¤erence to declarative main clause schon-utterances like (6), however, is that the positive evaluation disappears. In (24), the schonutterance is definitely not regarded as something positive, which is shown e.g. by the fact that one can use a clearly negatively evaluated utterance in this position. Consequently, (24) or (25) do not introduce any kind of soothing e¤ect into the conversation: (25) Wenn wir schon so einen Blo¨dsinn machen . . . if we SCHON such a nonsense make ‘If we really do such stupid things. . . .’
314
Markus Egg
But even if we ignore this absence of the positive evaluation for the time being, these examples are challenging for analyses that try to render the e¤ect of schon in terms of a tree-like discourse structure:8 The autterance for schon is not in the scope of if, and if both schon and if were to introduce discourse relations, the resulting discourse structure would not be a tree. (26) illustrates this structure, here the a-utterance q of schon is rendered as C0 , and the two clauses of (24) as C1 and C2, respectively: (26)
We can retain the treenees of discourse structure for these cases, if we assume that discourse particles behave in the same way as discourse adverbials like then: They introduce additional relations between discourse segments that would go beyond tree structures if they were integrated into the discourse structure (Webber et al. 2003). In particular, they would introduce multiple parenthood, e.g., in (26), C1 would be an argument of both a concession and a condition relation. In other cases, there would also be crossing dependencies. The following example from Webber et al. (2003) exhibits both of these phenomena: (27) High heels are fine for going to the theater. (C1) But wear comfortable shoes (C2) if instead you plan to go to the zoo. (C3) While the discourse structure of this example is (28a), the discourse adverbial instead relates C1 and C3 as alternatives, which would introduce another parent for these nodes and cut across the conditional relation between C2 and C3 as introduced by if, as illustrated in (28b):
8. The exact nature of discourse structure is a heavily debated topic; while analyses in the spirit of Rhetorical Structure Theory (Mann and Thompson, 1988; Taboada and Mann, 2006), e.g., the work of Marcu (2000), presume that discourse structure can be modelled by trees, other researchers point out problems for this view (e.g., Wolf and Gibson 2006 or Lee et al. 2009).
Discourse particles at the semantics-pragmatics interface
(28) a.
315
b.
The consequence of this line of reasoning is that discourse particles are to be classified as cohesive devices in the sense of Halliday and Hasan (1976). These devices complement discourse structure in the organisation of discourse but are not themselves part of the discourse structure (Egg and Redeker, 2008). This characterisation of discourse particles can now explain the di¤erences noted above for (3b) and (3c) as answers to (3a): As a cohesive device, schon can link discourse segments freely, whereas the discourse marker trotzdem indicates textual coherence and is more restricted in its combinatory potential.9 3.2. Non-declarative a-utterances Schon-utterances can also react to non-declarative a-utterances, e.g., in the following two examples. (29)
nichts passieren. A: Paß auf Dich auf! B: Mir wird schon look after yourself up me will SCHON nothing happen ‘A: Take care. B: Nothing will happen to me, don’t worry’
(30)
A: Was soll ich meinen Eltern zu Weihnachten schenken? what shall I my parents for Christmas give etwas einfallen. B: Dir wird schon you will SCHON something think of ‘A: What should I give to my parents for Christmas? B: You will be able to think of something, don’t worry.’
Since the a-utterances are non-declarative, their meanings cannot serve as a semantic argument of schon. Instead, the schon-utterances target felicity conditions of the a-utterances. In (29), B’s reaction targets the sincerity condition of A’s piece of advice, viz., that A believes that B will benefit from his advice to take care (¼q). This defeasibly entails that there is some impending danger for B, which is 9. Here I deviate from Halliday and Hasan (1976), who regard discourse markers as cohesive devices.
316
Markus Egg
ruled out by B’s reply (¼p). Note that even though the schon-utterances in (29) and (30) react to non-declarative a-utterances, they introduce a positively evaluated proposition, as they are declarative main clauses. Discourse (30) starts with a question, whose first preparatory condition states that A does not know the answer yet. B refers to this condition in his reply, he does not refute it, but points out that a proposition defeasibly deducable from this condition (that A’s state of ignorance will persist) does not hold. These examples show that the semantic arguments of a discourse particle and the meanings of the involved utterances need not coincide, in particular, this holds for non-declarative a-utterances whose interpretation is not a proposition. The proposed analysis receives further support from the observation (dating back to Ormelius-Sandblom 1997) that schon-utterances sometimes behave di¤erently in the context of biassed and unbiassed questions. This is illustrated in (31) and (32): (31)
A: Werden unsere Jungs es denn scha¤en? will our boys it then succeed B: Sie werden es schon scha¤en. they will it SCHON succeed ‘A: Will our boys really make it? B: Yes, they will.’
(32) A: Wird der Kanzler vom Parlament gewa¨hlt? gets the chancellor by parliament elected B: *Er wird schon vom Parlament gewa¨hlt. he gets SCHON by parliament elected ‘A: Is the chancellor elected by the parliament? B: Yes, he is.’ Here a schon-utterance can react to a biassed question like A’s utterance in (31). Its bias is marked by the particle denn, which suggests that A believes that the true answer is the negative one. In contrast, a schonutterance is not appropriate as a reaction to A’s unbiassed question in (32). The first step of the explanation is to note that this contrast does not show up for schon-utterances reacting to questions in general, as illustrated by e.g. (30), it only shows up for schon-utterances that are an answer to the question. Second, uttering a biassed question shows that the speaker believes one answer to be the right one, which is not the case for unbiassed questions. This belief of the speaker of a biassed question, for (31), that our boys are
Discourse particles at the semantics-pragmatics interface
317
not going to make it, can then be targeted by a schon-utterance on the basis that one can defeasibly deduce from A’s belief that our boys will not make it that they will indeed not make it, a potential consequence that is ruled out by the schon-utterance. In contrast, A does not exhibit a preference for one of the answers in the case of the unbiassed question in (32). Hence, reacting to this question in terms of a schon-question is not appropriate. At this point, one might argue that asking a question presents all possible answers as potentially true (a clausal implicature in the sense of Gazdar 1979), which would allow one to deduce every one of them defeasibly. But then the ‘Nixon diamond’ constellation outlined in Section 2 applies: As long as A is undecided, either of the mutually exclusive answers to A’s question (a‰rmative and negative) could be derived, hence, the derivation is blocked right from the start, and there is no a-utterance for schon to refer to. This line of argumentation can be extended to wh-questions in order to explain the phenomenon noted by Ormelius-Sandblom (1997) that schonutterances are not appropriate reactions to them:10 (33) A: Wo arbeitet Otto? where works Otto B: *Otto arbeitet schon in der Hauptstraße. Otto works SCHON in the Main Street ‘A: Where does Otto work? B: Ok, Otto works on Main Street.’ If we assume that answers to wh-questions are exhaustive (Groenendijk and Stokhof, 1997), they are mutually exclusive as well, hence incompatible, which blocks their simultaneous derivation (another Nixon diamond constellation). But if one is not willing to consider such answers as exhaustive, it is not clear which a-proposition could be targeted by B’s answer: A presents all possible answers to his question as potentially true, but neither of them would defeasibly entail the opposite of any other. An interesting consequence of this use of schon is that it can be used in a reaction to an interrogative a-utterance in order to insinuate that the speaker of this utterance had a specific answer in mind but did not make this explicit. Consider e.g. a variant of (31) without the particle denn:
10. Her examples show that she wants to discuss schon-utterances that provide an answer to the wh-questions, not those that react them in a di¤erent way (as, e.g., example (30) above).
318
Markus Egg
(34) A: Werden unsere Jungs es scha¤en? will our boys it succeed B: Sie werden es schon scha¤en. they will it SCHON succeed ‘A: Will our boys really make it? B: Yes, they will.’ Here A’s question is not necessarily biassed, yet B’s use of schon indicates that he assumes that A leans towards a negative answer to his question. 3.3. Non-declarative schon-utterances Finally, schon-utterances can not only react to non-declarative a-utterances, they can be non-declarative themselves. The first group of these utterances are questions. All of them are rhetorical, thus, schon functions as a marker of rhetoricity in questions. While all rhetorical questions present a statement as part of the common ground, the exact nature of this statement may vary. Consider first rhetorical schon-questions whose interpretation is negative (no entity fits the question), as in (35): (35) A: Du ha¨ttest Hans helfen mu¨ssen. you had Hans help have to B: Was ha¨tte ich schon tun ko¨nnen? what had I SCHON do be able to ‘A: You should have helped Hans. B: What could I have done?’ B’s reaction in (35), a rhetorical question, can be paraphrased as ‘it is CG information that B could not have done anything’ (¼p). This reaction targets a felicity question of A’s (a-)utterance in (35): The utterance is a reproach, A is telling B o¤ for not o¤ering any help. The relevant felicity condition for reproaches (which functions as the a-proposition q) is that the speaker believes that the hearer would have been able to act in the way that the speaker considers appropriate (here, that the hearer could have helped Hans). This defeasibly entails the opposite of B’s statement: If the speaker holds a specific belief, here, that the hearer could have helped, this defeasibly entails that the opposite of this belief is not CG information. This result of the defeasible reasoning is then ruled out by B’s reaction. This example shows that the notion of positive evaluation, which is extremely strong in declarative main clause schon-utterances (which showed
Discourse particles at the semantics-pragmatics interface
319
up especially in the discussion of (8)), is tied to these syntactic properties of the schon-utterance. Even though rhetorical questions function as statements, the proposition of B’s utterance (that he could not have done anything) is not evaluated positively. Another example that works exactly in the same way is (36) [¼(9)]. (36) A: Karen hat sich sehr u¨ber ihre Niederlage gea¨rgert. Karen has herself very about her defeat be angry B: Wer verliert schon gern? who loses SCHON gladly ‘A: Karen was very angry at her defeat. B: Well, who likes losing in the first place?’ Here B’s reaction (that it is CG information that no one likes losing, our p) targets the second felicity condition for statements (that it is not obvious to A that B knows already, our q). But if it is not obvious to A that B does not know a specific piece of information (that Karen hated losing), this suggests the defeasible inference that a logically stronger statement (not just Karen, but everyone hates losing, which entails that Karen hates losing) cannot be a piece of CG information. This inference is then blocked by B’s reaction. This account solves the puzzle presented in Section 1.3: The denial of expectation interpretation applies to (36), too, but it does not link the meanings of the involved utterances directly. Therefore there is no need to water down the semantic interpretation of schon in order to capture the semantic relation between the meanings of the involved utterances themselves in this example (which would be a generalisation). In addition, example (36) shows once again that not all parts of the CG need be activated in the mind of the interlocutors all the time, otherwise, there would be a contradiction between A’s belief and B’s statement, which intuitively is not there. And, like in example (35) above, the positive evaluation component is missing, even though the schon-utterance functions as a statement. The observation that discourse particles can be used to ascribe belief states to hearers is further corroborated by these schon-utterances: The speaker not only refers to a potentially applicable inference pattern shared by both interlocutors, by referring to felicity conditions of an utterance previously made by the hearer he expresses further assumptions about the hearer’s state of mind, which motivated the hearer to produce the preceding utterance.
320
Markus Egg
Other kinds of rhetorical questions have a non-negative interpretation, they indicate that there is a specific individual (or group of individuals) that fit(s) the question. Such questions can comprise schon, too, e.g., B’s reaction (37b) in response to (37a): (37) a.
A: Wen wird Gu¨nther zu seiner Feier einladen? whom will Gu¨nther to his party invite ‘A: Who will Gu¨nther invite to his party?’
b.
B: Wen wird er schon einladen? whom will he SCHON invite (Seine langweiligen Kollegen natu¨rlich.) his boring colleagues of.course ‘B: Well, who will he invite? (His boring colleagues, of course.)’
c.
B: *Er wird schon seine langweiligen Kollegen einladen. he will SCHON his boring colleagues invite ‘B: He will invite his boring colleagues.’
The interpretation of B’s question, our p, can be paraphrased as ‘it is known information that Gu¨nther will invite specific persons to his party (viz., his boring colleagues)’. This statement targets the first preparatory condition of questions (that A does not know the answer to his question, our q): If A does not know the answer then it makes sense to derive defeasibly that the answer is not already in the common ground. B points out, however, that this assumption does not hold. Note that the rhetoricity of B’s question is relevant here: A’s ignorance does not imply defeasibly that Gu¨nther will not invite his boring colleagues; but it does imply that Gu¨nther’s intention is not CG information. Therefore B could not react to (37a) by the non-rhetorical (37c), which introduces the fact that Gu¨nther will invite his boring colleagues, but does not characterise it as part of the CG. Finally, there are cases of imperative schon-utterances like (38) [¼(20)]. They raise the question of how to determine the p-proposition, if the schonutterance itself is not declarative: (38) Nun mach schon! now hurry SCHON ‘Come on!’ Following the approach sketched for doch in Egg (2010), I assume that p in such cases is the fact that the utterance was made. Schon is used to
Discourse particles at the semantics-pragmatics interface
321
signal the fact that this utterance is unexpected: In the given context, there is information (our q), which according to the CG would imply that no such utterance should be made. This q must be salient in the context, it might have been introduced in terms of a previous utterance of the hearer, which introduced a potential obstacle for complying with the speaker’s request, or merely been conveyed by the hearer’s hesitation to comply. This reconstruction fits in with the intuition of Thurmair (1989) that these imperatives relate to obstacles/hesitations in the context. Just like for the interrogative schon-utterances, there is no positive evaluative e¤ect of schon in schon-imperatives. In sum, it is possible to assume a semantic representation for schon which is very specific yet covers a wide range of uses of the particle: Schon relates two propositions p and q, which are derived from (though not necessarily identical to) the meaning of the schon-utterance and a preceding utterance to which the schon-utterance reacts. The relation between p and q holds i¤ both are true, and, according to the common ground, q defeasibly entails Bp. Thus, schon-utterances do not deny the proposition q, but point out that a reasonably expectable consequence of q does not hold. In this respect, schon introduces a denial of expectation. For declarative main clause schon-utterances, the particle characterises the proposition p of the utterance as positive. 4. Further work In the paper so far, I presented a research programme for the semantics of discourse particles, which was illustrated with an analysis of the discourse particle schon. What was described in detail up to now is the core of the approach, which consists in identifying the semantic contribution of discourse particles on the basis of small discourses, which typically consist of two utterances. In this section, I want to sketch three further research topics that build on and extend the line of approach proposed in this paper. This is not meant as an exhaustive list, rather, as an illustration of the scientific potential of the proposed approach. 4.1. Interaction with sentence mood A major challenge for a uniform analysis of the di¤erent uses of schon as sketched in the present paper is to explain the interdependence between positive evaluation and sentence mood in schon-utterances. This inter-
322
Markus Egg
dependence shows up e.g. in minimal pairs of rhetorical questions and their corresponding negative statements as in (39): (39) a.
b.
Wer liest das schon? who reads this SCHON ‘But who will read this?’ Das liest schon keiner. this reads SCHON nobody ‘No one will read this, don’t worry’
Even though both (39a) and (39b) convey the same message (modulo the common ground status of the former), there is only a positive evaluation (and with it, a soothing flavour) in the latter. Further work is also required to explain why schon-utterances cannot be true (i.e., non-rhetorical) interrogatives. 4.2. Comparing discourse particles On the basis of detailed analyses of discourse particles, it is possible to compare these particles in detail, and to account for slight di¤erences in meaning for minimal pairs of utterances in which either of two discourse particles can be used. As announced in Section 3.1, I will perform such a comparison for the discourse particles schon and doch, which will bring out the common ground and the di¤erences between these particles. Consider first the definition of the semantics of doch as proposed in Egg (2010). Like in the definition of the semantics of schon, p represents the semantic argument of doch that is related to the meaning of the p-utterance (analogously for q and the a-utterance): (40) [[doch]]( p)(q) i¤ both p and p > Bq are part of the common ground If the common ground comprises p and the fact that p defeasibly entails Bq, then defeasible Modus Ponens would allow one to infer Bq, if q did not hold. The direction of the implication for doch is still debated in the literature, e.g., Abraham (1991), Gast (2008), and Grosz (2010) assume the implication q > Bp, i.e., the one which I assumed for schon. However, in order to model the fact that doch-utterances introduce potential impediments for their a-utterances, whereas schon-utterances rule out potential consequences of their a-utterances, I reserve q > Bp for schon.
Discourse particles at the semantics-pragmatics interface
323
Introducing p in spite of its CG status presupposes assuming that it is not su‰ciently activated for the hearer, which integrates the semantic description of doch in Zimmermann (2011) into the proposed analysis. As a simple illustration for this account, consider e.g. (41), adapted from Karagjosova (2004): Doch indicates that the speaker would not expect his frequent illness because of his healthy lifestyle. (41) Ich bin oft krank. Dabei lebe ich doch gesund. I am often ill but live I DOCH healthily ‘I am often ill. But I have a healthy lifestyle.’ The e¤ect of doch p as a reaction to an a-proposition q against the common ground C is to remind the hearer that C comprises a potential impediment p for q, which expresses either surprise at the fact that nevertheless q should hold or puts doubt on q. Still, q is not explicitly denied, since it is compatible with p, therefore one person can utter both involved utterances without contradicting himself, as in (41). This example shows that the suggestion of Ko¨nig (1997) (that doch introduces a contradiction) is too strong. In other words, this particle introduces a denial of expectation that is based on a defeasible inference that is part of the CG. This is just like in the case of schon, but this time the arguments are reversed: The a-utterance introduces the opposite of the expected state of a¤airs, the p-utterance characterises the state of a¤airs that triggers this expectation. The second di¤erence is the CG status of the first semantic argument of doch, neither argument of schon is marked as CG information. This contrastive description of schon and doch can now be used to analyse ‘minimal pairs’ of utterances that are identical except for doch and schon. As an example, consider the dialogues (42a)/(42b) and (42a)/ (42c): (42) a.
A: Ich habe ein Loch in meiner Hose. I have a hole in my trousers ‘A: There’s a hole in my trousers.’
b.
B: Das merkt doch keiner. this notices DOCH no one ‘B: But no one will notice.’
c.
B: Das merkt schon keiner. this notices SCHON no one ‘B: No one will notice, don’t worry.’
324
Markus Egg
Intuitively, both (42b) and (42c) as a reaction to (42a) express the idea that the hole is not a problem, but (42b) sounds slightly more annoyed and less soothing than (42c). This contrast can now be explained, it follows from a di¤erent reaction to a speech act: B’s reaction in (42a)/(42b) casts doubt on the speech act (complaint) of A by pointing out information accessible to A as part of the CG (that no one will notice the hole), which suggests that one of the felicity conditions of the complaint (that the situation referred to is a problem for A) does not obtain. In other words, if A had taken into account this information (which he should have), he would not have uttered the complaint at all. But in (42a)/(42c), B accepts A’s speech act, which is less confrontative than refusing it altogether. B’s reaction merely tries to soothe A by pointing out that a plausibly expectable negative consequence of the fact about which A is complaining is not going to happen (i.e., people will not notice). In the case of schon, the information conveyed by B is is not already part of the CG, i.e., it is not A’s fault if he did not take it into account when uttering (42a). This also makes (42c) less annoyed as an answer to (42a) than (42b). This is just a first example for a very natural extension of the line of approach that is presented in this paper. I believe that it is necessary to sort out the semantics of the individual particles first before it makes sense to compare them, be it in isolation or in the context of minimal pairs of utterances. As another example, consider the discourse particle auch. As a very preliminary description of its semantics (which does not attempt to cover the whole range of its uses), I follow the intuition of Ko¨nig (1997) and Karagjosova (2004) that auch indicates a relation of inference (as part of the CG) between its arguments. Once again, p represents the semantic argument of auch related to the meaning of the p-utterance (analogously for q and the a-utterance): (43) [[auch]]( p)(q) i¤ both p and p > q are part of the common ground The e¤ect of auch is to indicate that an a-utterance is of less informative value, because its semantics q could have been defeasibly deduced on the basis of CG information by defeasible Modus Ponens (from p and p > q). For example, in (44), B points out that A’s utterance is not particularly surprising, since it is CG information that Peter is ill, and a defeasible inference from being ill to not coming along is also part of the CG:
Discourse particles at the semantics-pragmatics interface
(44)
325
A: Peter kommt nicht mit. B: Er ist auch krank. Peter comes not along he is AUCH ill ‘A: Peter will not come along. B: He is ill.’
This semantic account of auch is very similar to the one of doch: both refer to information from the CG, but auch introduces it as a potential reason, doch, as a potential impediment. Therefore, one would expect that (44) has a counterpart that is derived by (a) replacing auch by doch and (b) inserting a negation into the p-utterance. This prediction is borne out for (44), its reformulation (45) is wellformed. Here, B’s reply points out a potential impediment for Peter’s coming along (as announced by A), which expresses surprise and a reason for not accepting A’s statement without further discussion: (45)
A: Peter kommt mit. B: Er ist doch krank. Peter comes along he is DOCH ill ‘A: Peter will not come along. B: But he is ill.’
However, this kind of reasoning must be embedded in a sophisticated theory of discourse particles and their semantics in order to account for the whole range of attested examples. Consider e.g. the statement (46a) and two possible reactions to it, viz., (46b) and (46c), which di¤er only in auch and doch, respectively: (46) a.
A: Der Pullover ist aber weich. the sweater is but soft ‘A: The sweater is amazingly soft.’ A: Der Pullover ist aber weich.
b.
B: Das ist auch Schurwolle. this is AUCH new wool ‘B: Well, this is new wool.’
c.
B: Das ist doch Schurwolle. this is DOCH new wool ‘B: But this is new wool.’
Both (46b) and (46c) are appropriate reactions to (46a), there is no need to include a negation into A’s statement to make it amenable as an a-utterance for (46c). Thus, the reasoning sketched for (44) and (45) does not cover the minimal pairs (46a)/(46b) and (46a)/(46c).
326
Markus Egg
In the proposed analysis, the minimal pair can be explained in a similar fashion to the dialogues (42a)/(42b) and (42a)/(42c) sketched above: (46b) introduces a potential reason for (42c), i.e., for sweaters, being made of new wool defeasibly implies being soft. In contrast, (46c) targets a felicity condition of (46a), which is used to express surprise. The relevant felicity condition is that it is not obvious to the speaker that the state of a¤airs about which he expresses surprise will obtain in the normal course of events. But, if it is CG information that the sweater is made of new wool, this is a potential impediment to this felicity condition. I expect that the lines of research sketched in this subsection can be fruitfully applied to other discourse particles as well. 4.3. Stressed uses of schon Another natural continuation of the work presented in this paper is the extension of the analysis of unstressed discourse particle uses of schon to other uses of the particle, in particular, to stressed uses. In this subsection, I will not attempt to o¤er an exhaustive account of all stressed uses of schon, I will just focus on two stressed uses to show that the account proposed for unstressed uses of the particle can be extended naturally to these stressed uses, which provides additional evidence for the intuition that both stressed and unstressed schon are related (an intuition which I share with e.g. Krifka 2000 and Fe´ry 2010). The first use is ‘propositional’ in that schon by itself confirms a preceding utterance and typically is accompanied by a restricting clause introduced by aber ‘but’: (47) A: Das Fest war sehr scho¨n. the party was very nice B: SCHON, aber die Musik war zu laut. SCHON but the music was too loud ‘A: The party was very nice. B. It was, but the music was too loud.’ This use is captured by the analysis so far, if we assume that q is the meaning of the confirmed utterance, and p is the meaning of the aberclause. B’s utterance of stressed schon confirms q, but points out that there are reasonably expectable consequences of q that do not hold. These consequences can be denied explicitly in the aber-clause. The e¤ect is a rather weak confirmation of q, which is exactly what the analysis of Fe´ry (2010) (in terms of a free focus) predicts.
Discourse particles at the semantics-pragmatics interface
327
In this use, q cannot be a felicity condition of the a-utterance, as illustrated by a variant of (11): (48) A: Was soll ich meinen Eltern zu Weihnachten schenken? what shall I my parents to Christmas give B: *SCHON, aber dir wird etwas einfallen. SCHON but you will something think of ‘A: What should I give to my parents for Christmas? B: Yes, but you will be able to think of something, don’t worry.’ This observation also holds good for cases in which schon-utterances target felicity conditions of declarative utterances, as in the following variant of (36): (49) A: Karen hat sich sehr u¨ber ihre Niederlage gea¨rgert. Karen has herself very about her defeat be angry B: *SCHON, aber wer verliert gern? SCHON but who loses gladly ‘A: Karen was very angry at her defeat. B: Well, who likes losing in the first place?’ A second very interesting use of stressed schon is a focus-sensitive use of schon in elliptical utterances that introduces a contrast. There is a contrast between mammals and duckbilled platypi in that the former usually do not lay eggs: (50) A: Ein Sa¨ugetier legt keine Eier. a mammal lays no eggs B: Das SCHNAbeltier SCHON. the duckbilled platypus SCHON ‘A: Mammals do not lay eggs. B: Duckbilled platypi do.’ (50) can be covered by the proposed analysis immediately, the statement that mammals do not lay eggs (our q) allows one to deduce defeasibly that duckbilled platypi do not either, because they are mammals, too. This expectation is denied in B’s schon-utterance.11
11. This account does not attempt to describe these examples exhaustively. For example, in (i), there is a contrast between Algeria and England in that only the latter got through to the 2nd round:
328
Markus Egg
Usually, these elliptical schon-utterances are a reaction to negated sentences, but there are other cases: (51) A: Italienische Autos sind elegant. B: LamborGHInis SCHON. Italian cars are elegant Lamborghinis SCHON ‘A: Italian cars are elegant. B: Lamborghinis are.’
t
B’s reaction does not completely refute A’s claim that Italian cars are elegant, but restricts it to Lamborghinis. We can explain this example by Gricean reasoning: Asserting P(Y ) in a context in which P(X ) has been uttered (a potential update for the CG) and where Y refers to a subgroup of X triggers first a scalar implicature that P does not hold for any Y 0 such that Y Y 0 , and then the implicature that P does not hold for the complement of Y w.r.t. X. In (51), the relevant complement are all Italian cars that are not Lamborghinis, and B implies that they are not elegant. At this point, the proposed analysis of (unstressed) schon can once again be applied in the same way as to (50): B’s implied statement (that Italian non-Lamborghinis are not elegant, our p) refutes a consequence of A’s claim that Italian cars are elegant (¼q). This analysis predicts that B must refer to a subgroup of Italian cars, which is borne out by (51), in which Lamborghinis have been replaced by Volkswagens in B’s inappropriate reaction: (52) A: Italienische Autos sind elegant. B: *VOLKSwagen SCHON. Italian cars are elegant Volkswagens SCHON ‘A: Italian cars are elegant. B: Volkswagens are.’ To sum up, this paper presented a research programme for discourse particles which aims at deriving a very specific uniform semantic analysis of these particles that nevertheless manages to capture a wide range of uses of the particles. These two competing goals could be pursued in one (i) Algerien hat sich nicht fu¨rs Achtelfinale qualifiziert. Algeria has itself not for the second round qualified ENGland SCHON. England SCHON ‘Algeria didn’t get through to the second round. England did.’ Nevertheless, there is no straightforward account of (i) along the lines sketched for (50), because there is no immediate causal link between Algeria’s elimination and England’s qualification.
Discourse particles at the semantics-pragmatics interface
329
analysis because it was assumed that there is more flexibility in combining the meaning of a discourse particle with the meanings of surrounding utterances. Discourse particles like schon and doch relate two propositions semantically, but the meaning of the utterance of which the particle is a part, and the meaning of the utterance to which this first utterance reacts are not the only feasible semantic arguments of the particles. In particular, felicity conditions of these two utterances can serve as semantic arguments, too. This research programme was illustrated in detail by investigating the unstressed uses of the particle schon. The proposed analysis was shown to provide a suitable basis for the comparison of the semantics of di¤erent discourse particles and to be applicable to some stressed uses of the particle as well. Further work is needed to cover the whole range of uses of schon at this point, as well as for a comparison of schon and other discourse particles.
References Abbott, Barbara 2008 Presuppositions and common ground. Linguistics and Philosophy, 31:523–538. Abraham, Werner 1980 The synchronic and diachronic semantics of German temporal noch and schon, with aspects of English still, yet and already. Studies in Language, 4:3–24. Abraham, Werner 1991 Discourse particles in German: How does their illocutive force come about? In Werner Abraham, editor, Discourse particles in German. John Benjamins, Amsterdam, pages 203–252. Abraham, Werner 2000 Modal particles in German: word classification and legacy beyond grammaticalization. In Petra Vogel und Bernard Comrie, editors, Approaches to the typology of word classes. Mouton, chapter Berlin, pages 321–350. Abraham, Werner and Elisabeth Leiss 2012 Introduction. This volume. Asher, Nicholas and Alex Lascarides 2003 Logics of conversation. Cambridge University Press, Cambridge. van der Auwera, Johan 1993 ‘‘Already’’ and ‘‘still’’: Beyond duality. Linguistics and Philosophy, 16:613–653.
330
Markus Egg
Borst, Dieter 1985
Die a‰rmativen Modalpartikeln doch, ja und schon. Ihre Bedeutung, Funktion, Stellung und ihr Vorkommen. Niemeyer, Tu¨bingen.
Do¨ring, Sophia 2010 On context shift in German discourse particles. BA thesis, Humboldt University, Berlin. Egg, Markus 2007 Meaning and use of rhetorical questions. In Maria Aloni, Paul Dekker, and Floris Roelofsen, editors, Proceedings of the 16th Amsterdam Colloquium, pages 73–78. Universiteit van Amsterdam, ILLC. Egg, Markus 2010 A unified account of the semantics of discourse particles. In Raquel Ferna´ndez, Yasuhiro Katagiri, Kazunori Komatani, Oliver Lemon, and Mikio Nakano, editors, Proceedings of the SIGDIAL 2010 Conference, University of Tokyo, 24–25. Sept. 2010, pages 132–138. University of Tokyo. Egg, Markus and Gisela Redeker 2008 Underspecified discourse representation. In Anton Benz and Peter Ku¨hnlein, editors, Constraints in Discourse. Benjamins, Amsterdam, pages 117–138. Fe´ry, Caroline 2010 Information structure of schon. In Thomas Hanneforth and Gisbert Fanselow, editors, Language and Logos. A Festschrift for Peter Staudacher. Akademie-Verlag, Berlin. To appear. Gast, Volker 2008 Modal particles and context updating – the functions of German ja, doch, wohl, and etwa. In Heinz Vater and Ole Letnes, editors, Modalverben und Grammatikalisierung. Wissenschaftlicher Verlag, Trier, pages 153–177. Gazdar, Gerald 1979 Pragmatics: implicature, presupposition, and logical form. Academic Press, New York. Gornik-Gerhardt, Hildegard 1981 Zu den Funktionen der Modalpartikel ,,schon‘‘ und einiger ihrer Substituentia. Narr, Tu¨bingen. Greenberg, Yael 2009 Presupposition accommodation and informativity considerations with aspectual still. Journal of Semantics, 26:49–86. Groenendijk, Jeroen and Martin Stokhof 1997 Questions. In J. van Benthem and A. ter Meulen, editors, Handbook of logic and language. Elsevier, Amsterdam, pages 1055– 1124. Grosz, Patrick 2010 German doch: An element that triggers a contrast presupposition. In CLS 46, Tilburg, The Netherlands.
Discourse particles at the semantics-pragmatics interface
331
Grote, Brigitte, Nils Lenke, and Manfred Stede 1997 Ma(r)king concessions in English and German. Discourse Processes, 24:87–118. Halliday, M. and Ruqaiya Hasan 1976 Cohesion in English. Longman, London. Jacobs, Joachim 1991 On the semantics of modal particles. In Werner Abraham, editor, Discourse Particles. Benjamins, Amsterdam, pages 141–162. Karagjosova, Elena 2001 Towards a comprehensive meaning of German doch. In Proceedings of the ESSLLI 2001 Student Session, Helsinki. Karagjosova, Elena 2004 The meaning and function of German modal particles. Ph.D. thesis, Universita¨t des Saarlandes. Klein, Wolfgang 2010 About the German particles schon and noch. Manuscript, MPI Nimwegen. Knott, Alistair 1996 A data-driven methodology for motivating a set of coherence relations. Ph.D. thesis, University of Edinburgh, Department of Artificial Intelligence. Knott, Alistair 2001 Semantic and pragmatic relations and their intended e¤ects. In Ted Sanders, Joost Schilperoord, and Wilbert Spooren, editors, Text representation: linguistic and psycholinguistic aspects. Benjamins, Amsterdam, pages 127–151. Knott, Alistair and Ted Sanders 1998 The classification of coherence relations and their linguistic markers: An exploration of two languages. Journal of Pragmatics, 30:135–175. Ko¨nig, Ekkehard 1991 The meaning of focus particles. A comparative perspective. Routledge, London. Ko¨nig, Ekkehard 1997 Zur Bedeutung von Modalpartikeln im Deutschen: Ein Neuansatz im Rahmen der Relevanztheorie. Germanistische Linguistik, 136:57–75. Ko¨nig, Ekkehard and Susanne Requardt 1991 A relevance-theoretic approach to the analysis of modal particles. Multilingua, 10:63–77. Ko¨nig, Ekkehard, Detlef Stark, and Susanne Requardt 1990 Adverbien und Partikeln. Ein deutsch-englisches Wo¨rterbuch. Julius-Groos-Verlag, Heidelberg.
332
Markus Egg
Krifka, Manfred 2000 Alternatives for aspectual particles: the semantics of still and already. In Proceedings of the Annual Meeting of the Berkeley Linguistics Society 26, pages 401–412. Lagerwerf, Luuk 1998 Causal connectives have presuppositions. E¤ects on coherence and discourse structure. Ph.D. thesis, Universiteit van Tilburg. Lee, Alan, Rashmi Prasad, Aravind Joshi, and Bonnie Webber 2008 Departures from tree structures in discourse: Shared arguments in the Penn Discourse Treebank. In Proceedings of the workshop ‘Constraints in Discourse III ’, Potsdam. van Leusen, Noor 2007 Description Grammar for Discourse. Ph.D. thesis, Radboud Universiteit Nijmegen. Lo¨bner, Sebastian 1999 Why German schon and noch are still duals: a reply to van der Auwera. Linguistics and Philosophy, 22:45–107. Mann, William and Sandra Thompson 1988 Rhetorical Structure Theory: Towards a functional theory of text organization. Text, 8:243–281. Marcu, Daniel 2000 The theory and practice of discourse parsing and summarization. MIT Press, Cambridge. Miltsakaki, Eleni, Rashmi Prasad, Aravind Joshi, and Bonnie Webber 2004 Annotating discourse connectives and their arguments. In Proceedings of the HLT/NAACL Workshop on Frontiers in Corpus Annotation, Boston. Ormelius-Sandblom, Elisabet 1997 The modal particle schon: Its syntax, semantics, and pragmatics. In Toril Swan and Olaf Westvik, editors, Modality in Germanic languages: historical and comparative perspectives. de Gruyter, Berlin, pages 75–133. Oversteegen, Leonoor 1997 On the pragmatic nature of causal and contrastive connectives. Discourse Processes, 24:51–85. Sanders, Ted, Wilbert Spooren, and Leo Noordman 1992 Towards a taxonomy of coherence relations. Discourse Processes, 15:1–35. Stalnaker, Robert 2002 Common ground. Linguistics & Philosophy, 25:701–721. Sweetser, Eve 1990 From etymology to pragmatics. Metaphorical and cultural aspects of semantic structure. Cambridge University Press, Cambridge. Taboada, Maite and William Mann 2006 Rhetorical Structure Theory: looking back and moving ahead. Discourse Studies, 8:423–459.
Discourse particles at the semantics-pragmatics interface
333
Thurmair, Maria 1989 Modalpartikeln und ihre Kombinationen. Niemeyer, Tu¨bingen. Webber, Bonnie 2004 D-LTAG: extending lexicalized TAG to discourse. Cognitive Science, 28:751–779. Webber, Bonnie 1988 Discourse deixis and discourse processing. Technical report, University of Pennsylvania, Department of Computer and Information Science. Available from http://repository.upenn.edu/cisreports/610. Webber, Bonnie, Matthew Stone, Aravind Joshi, and Alistair Knott 2003 Anaphora and discourse structure. Computational Linguistics, 29:545–587. Weinrich, Harald 2005 Textgrammatik der deutschen Sprache. Georg Olms, Hildesheim, 3rd edition. Wolf, F. and E. Gibson 2006 Coherence in natural language: data stuctures and applications. MIT Press, Cambridge. Zimmermann, Malte 2011 Discourse particles. In Claudia Maienborn, Klaus von Heusinger, and Paul Portner, editors, Semantics: an international handbook of natural language meaning. Mouton de Gruyter, Berlin, 2011– 2038.
Modality in the Romance languages: Modal verbs and modal particles Benjamin Meisnitzer Ludwig-Maximilians-Universita¨t Mu¨nchen
Abstract1 The aim of this paper is to approach modality in the Romance languages, introducing the Theory of Mind (henceforth ToM) into the discussion. This has not been done so far, although it is a very useful tool to decide whether a verb or a particle is modal or not. Therefore, modal verbs as well as modal particles will be focused on, due to their crucial role in the speaker’s deixis and their higher rank in the speaker’s/addressee’s deixis, when compared to seemingly equivalent adverbials. Bearing in mind the defining criteria for modals proposed by modern linguistics, it will be shown that in contrast to what is widely assumed in the literature on the subject there is no paradigm of modal verbs in the Romance languages. From a cross-linguistic perspective, it will be argued that the only Romance modals that are polyfunctional, having both epistemic and deontic readings, are Span./Port./Cat. poder, Fr. pouvoir, It. potere, Rom. a putea. Regarding the emergence of modal particles it will be made clear that the modal particle-like lexemes in the Romance languages do not represent a specific paradigmatically ordered class of words as they do in German.
1. I would like to thank E. Leiss (Munich) and W. Abraham (Munich/Vienna) for the various pieces of advice and correction as well as for the inspiration provided by a very challenging course on modality at LMU and insightful corrections on earlier versions of this paper, as well as U. Detges (Munich) for all the interesting ideas generated by his lectures on the subject in regard to the Romance languages. Moreover, I would like to thank P. M. Bertinetto (Pisa), J. Grimshaw (Rutgers), M. Coniglio (Venice), M. Squartini (Torino) and P. Ramat (Padova) for the interesting discussions and feedbacks during the SLE-workshop. For additional comments and discussion I would like to thank C. Knels (Hamburg), M. Arden (Eichsta¨tt), T. Scharinger (Erlangen) and especially W. Oesterreicher (Munich).
336
Benjamin Meisnitzer
1. Modal verbs in the Romance languages 1.1. Current research and findings Despite the overwhelming number of investigations on modality, the notion of modality is still very controversial. Recent studies assume a link between the child’s development of the ability to assess other people’s mental states about a state of a¤airs (ToM) and the use of epistemic modality (cf. Papafragou 2002: 185). According to a ToM based approach, modal verbs and modal particles are di¤erent techniques to negotiate shared knowledge. Few Romance scholars have devoted their research to modal verbs2 and their categorization, mainly due to the fact that there is still no generally accepted definition of the term modal verb. In correspondence to the rather clear-cut paradigm of genuine modal verbs in German, one usually encounters the following list of modal verbs for the Romance languages: It. potere ‘can’, dovere ‘must’/‘have to’, volere ‘want to’, Fr. devoir ‘must’/‘have to’, pouvoir ‘can’, vouloir ‘want to’ and sometimes savoir 3 ‘to know how to’, Span. deber ‘must’/‘have to’, poder ‘can’, tener (que) ‘must’/ ‘have to’ and desear ‘wish to’; Port. dever ‘must’/‘have to’, poder ‘can’, ter (de/que) ‘must’/‘have to’, haver de ‘will’4 and querer ‘want to’ and Rom. a putea ‘can’, a s¸ti ‘to know how to’, a trebui ‘must’/‘have to’ and a vrea ‘want to’ (cf. Beldarraı´n Jime´nez 1979: 272; Engel/Sta˘nescu et al. 1993: 412; Milan 2001: 7¤.). Still some scholars criticize such a list as an inaccurate reduction of modals, arguing that there are many more verbs used to express modality in the di¤erent Romance languages (cf. Beldarraı´n Jime´nez 1979: 272; Sctrick 1971: 118). Some scholars extend the concept of modal verbs to what they call a semantic subsystem of modals in the 2. For a detailed overview of papers on modals in Romance languages, see Milan (2001: 4) and Johnen (2003: 11; 33–52). 3. Although savoir can be combined with verbs expressing physical abilities, it is mainly combined with verbs denoting intellectual capacities. The remaining semantic component suggests an uncompleted semantic bleaching and a low degree of grammaticalization compared with Fr. pouvoir and devoir. savoir expresses mere evidentiality, and not speaker’s appraisals or judgments. Unlike modals, savoir cannot be combined with other modals and it has no epistemic reading. Hence, it is not a prototypical modal and will not be discussed in this paper. 4. Port. haver de is essentially a future marker. Still, we sometimes find it classified as a modal verb, because it can express the speaker’s will, his conviction or his appraisal of how probable he considers the verbal event.
Modality in the Romance languages
337
Romance languages, leading to expansive inventories (of modal forms) like the one suggested by Johnen (2003: 471) with 84 core modals and 24 peripheral modals. Nevertheless, there seems to be broad consensus that the Romance languages have fewer genuine modal verbs than the Germanic languages (Milan 2001: 10; Leiss 2008: 16). Although Johnen lists more than 230 modals for Portuguese (cf. Johnen 2003: 11), most linguists agree that only two or three of them are genuine modal verbs, according to the defining criteria for modals in modern linguistics. The largest consensus in the literature on modal verbs in Romance languages concerns the forms that derive from Latin posse and debe#re. According to Abraham (2002), the criteria used to determine genuine modal verbs (in German[ic]) are: 1. Modal verbs/MVs have to be polyfunctional, that is supporting both epistemic and deontic readings5 (Abraham 2002: 27¤.), 2. MVs, as they may have the syntactic status of a full verb, may stand in (albeit highly constrained) combination with each other (Abraham 2002: 28f.), 3. They must derive from an independent lexical verb (Abraham 2002: 25f.), 4. MVs have no imperative form (Abraham 2002: 25) and 5. Non-finite forms do not allow epistemic readings (Abraham 2008a: 333).6 These criteria are met by all derivatives of Latin posse in the Romance languages and by most derivatives of Lat. debe#re. 1.2. The Romance MV-derivatives of Latin posse and debe#re When we look at modal verbs in the Romance languages in a geographical perspective from west to east, in Portuguese we find poder ‘can’/‘to be able to’ and dever ‘must’/‘have to’, both of which allow deontic and epistemic readings. Whether a deontic (DM) or an epistemic reading (EM) can be inferred, depends mainly on the specific semantic features of the accompanying full verb.
5. In German and Italian, perfective aspect converges strongly with root modality and imperfective aspect with epistemic modality (Abraham 2008a: 344; Leiss 2009: 7; Pietrandrea 2005, chapters 7–8). In Spanish and French there is no connection between modal readings and aspect (cf. Abraham 2008b: 77). The same holds for Catalan and Portuguese. 6. These criteria are deduced from German modal verbs, because they are considered prototypical modals (cf. Hammerich 1960: 66; Abraham 2002: 22 and Abraham 2009: 297), illustrating the di¤erence between grammatical and lexical modality.
338
Benjamin Meisnitzer
(1) Ele pode estar em casa.7 p pr 3rd pers. m s MV 3rd pers s pres ‘to be’ inf prep N ‘home’ DM – ‘He can be at home.’ (¼‘He is allowed to be at home.’) (t s < t e )8 EM – ‘He may be at home.’ (¼According to the speaker’s assessment) (ts ¼ t e ) (example from Oliveira 2001: 175) In contrast to German, the simple present in Portuguese triggers an epistemic reading. In the epistemic reading mood, tense and aspect are specified in the epistemic modal, this being in sharp contrast to the Germanic languages where it is the main verb that establishes a reference to the event (cf. Abraham 2008b). To express deontic modality the speaker would opt for ficar ‘to stay’, which does not permit an epistemic reading. The modal verb dever also allows both readings, but with a stronger a‰nity to the epistemic reading than poder. (2) Ele deve ganhar (cerca de)9 p pr 3rd pers m s MV ‘has to/must’ ‘to earn’ inf ‘about/around’ 3rd pers. s. pres. 5000 Euros por meˆs. 5000 N ‘euro’ prep N ‘month’ DM – ‘He has to/must earn 5000 euros a month.’ (t s < t e ) EM – ‘He must earn (about) 5000 euros a month.’ (t s ¼ t e )
7. Abbreviations used in this paper: acc – accusative, adj – adjective, adv – adverb, adv conj – adversative conjunction, adv neg – adverb of negation, aux – auxiliary, CP – compound perfect, def art – definite article, dem – demonstrative, dir obj – direct object, f – feminine, fut – future, imperat – imperative, imp – imperfect, ind obj – indirect object, inf – infinitive, int – interrogative, m – masculine, MP – modal particle, MV – modal verb, N – noun, nom – nominative, p part – past participle, p pr – personal pronoun, pers – person, poss – possessive, pr. – pronoun, pred adj – predicative adjective, prep – preposition, pres – simple present, subj – subjunctive, subj m – subjunctive marker, unstr. – unstressed, V – verb. 8. In the following, the notation of event tensing according to Reichenbach (1960: 288) is used, with the symbols ts and te referring to the point of speech (ts) and the point of the event (te). 9. The speaker would use this phrase to express an epistemic reading, unless the context clearly favoured a deontic reading.
Modality in the Romance languages
339
According to the ToM approach, in (1) and (2) the speaker assumes the roles of assessor and source of information (double deixis), while the addressee does not participate in the assessment of the truth-value. The same examples with both readings can be found in Spanish ( poder, deber), French ( pouvoir, devoir) and Italian ( potere, dovere). Therefore these verbs qualify as modals in these languages. As mentioned above, in Portuguese (cf. (1)) the speaker would use different main verbs to express either a deontic ( ficar) or an epistemic reading (estar). The same holds true for speakers of other Romance languages: For expressing deontic modality, in Spanish quedarse ‘to stay’ and in French rester ‘to stay’ would be preferred (instead of Span. estar and Fr. eˆtre). Apart from the fact that the main verb can bias towards a deontic or an epistemic reading, the readings are strongly contextual. In Italian, for instance, due to the semantic di¤erence between essere and stare, the first triggers an epistemic reading (3) and the second a deontic reading (4). (3) Puo` essere a casa. MV ‘can’ 3rd pers s pres ‘to be’ inf. prep N ‘home’ EM – ‘He can be at home.’ (¼‘It is possible according to the speaker’s assessment’) (ts ¼ te) (4) Puo` stare a casa. MV ‘can’ 3rd pers s pres ‘to be’ inf prep N ‘home’ DM – ‘He can be at home.’ (¼‘He is allowed to be at home.’) (ts < te) When looking at epistemic modality in the various Romance languages, attention can be drawn to the fact that many utterances which in Italian and Romanian express both an EM and a DM almost exclusively allow a deontic reading (5) in the other Romance languages.10 (5) Gli puo` scrivere. p pr 3rd pers s dative MV 3rd pers s pres inf ‘to write’ DM – ‘He can write him.’ (ts < te) EM – ‘He may write him.’ (ts ¼ te)
10. Cf. Span. Puede escribirle ‘He/She can write him/her’ (DM/*EM), Fr. Il lui peut e´crire ‘He can write him.’ (DM/*EM), but Rom. Iˆi poate scrie with the readings: DM – ‘He can write him’ (ts < te) and EM – ‘He may write him’ (ts ¼ te).
340
Benjamin Meisnitzer
In Romanian and Catalan, only Rom. a putea and Cat. poder can be considered modal verbs due to the distinctive features of Rom. a trebui ‘have to’ and Cat. deure ‘have to’ (cf. 1.3). Like in the other Romance languages, to express deontic modality the speaker would prefer a fi ‘to stay’ in Romanian (6) and quedar-se ‘to stay’ in Catalan in the utterances equivalent to (1).11 A general trend in the Romance languages seems to be that the deontic meaning is expressed by the full verb. It has to be pointed out that in Romanian a modal verb can be combined with an infinitive complement (6) or a subjunctive complement (7) (cf. Motapanyane/Avram 2001: 159f.). The use of a subjunctive complement, which is more common in spoken language, is typical of Balkan languages, while the infinitive complement is typical of the Romance languages. There is no di¤erence in meaning. (6) Maria-l poate vizita. N-p pr 3rd pers s m ac MV 3rd pers s pres inf ‘to visit’ (7) Maria poate sa˘-l N MV 3rd pers s pres subj m-p pr 3rd pers s m acc viziteze. V 3rd pers s subj DM/EM – ‘Maria can visit him.’ (examples from Motapanyane/Avram 2001: 149) 1.3. Distinctive features in the use of Latin debe#re in Romanian and Catalan In Romanian, due to Slavic influence, a trebui ‘have to’/‘must’ is only used in the 3rd person singular, and therefore has no full inflectional paradigm. Furthermore, it allows only a deontic reading: (8) Ea trebuie sa˘ cas¸tige p pr 3rd pers f s MV 3rd pers s pres sub m V ‘to earn’ 3rd pers s subj 5000 de euro. 5000 prep N ‘euro’ DM – ‘She has to/must earn 5000 euros a month.’ (ts < te) EM? – ‘For sure she earns 5000 euros a month.’? (ts ¼ te) 11. In a diachronic perspective this shows an increasing a‰nity of the modal verb to an epistemic reading, as is also the case for may, must and can in American English (cf. Abraham 2002: 22).
Modality in the Romance languages
341
By contrast, in Catalan deure þ infinitive can only express epistemic modality (10) and has no deontic meaning (9). (9) *El nou director ha dit que def art m s poss adj m s N m s ‘director’ V ‘to say’ 3rd pers s CP conj tothom deu arribar a l’hora. ‘everybody’ MV 3rd pers s pres inf ‘arrive’ ‘on time’ *‘Our director told us that everybody should be punctual.’ *(ts < te) (10) Ja deu haver entrat ‘Already’ MV 3rd pers s pres. inf. ‘to have’ P.Part ‘get in’ tothom. ‘everybody’ ‘Everybody should already be inside.’ (ts ¼ te) EM To express deontic modality the modal has to be replaced by a verbal periphrasis: El nou director ha dit que tothom ha d’arribar a l’hora. (ha d’ – ‘has to’) (ts < te) (‘Our director told us that everybody should be punctual.’). 1.4. Conclusion: Are there modal verbs in the Romance languages? In the Romance languages only the reflexes of Latin posse ‘can’ fulfill the criteria to be classified as prototypical modals. debe#re meets the criteria of a modal in Portuguese (dever), Spanish (deber), French (devoir), and Italian (dovere), whereas Rom. a trebui and Cat. deure lack the aspect of polyfunctionality. It is important to note that in the Romance languages the readings often seem to be restricted by semantic features of the embedded full verb. There are no aspectual constraints for deontic and epistemic modality in the Iberian Romance languages and in French. Modals in the Romance languages do not constitute a proper closed paradigm, but rather resemble open word-classes. Nevertheless, the Romance derivatives of Latin posse and – at least in some Romance languages – the derivatives of debe#re are comparable to modal verbs for the following reasons: They (1) allow deontic as well as epistemic readings, (2) They can stand in combination with each other, (3) They have no imperative forms, and (4) Their infinitival forms do not allow an epistemic reading. On the syntagmatic level they have both lost syntagmatic variability, since they occur at fixed syntactic positions closely bound to the main verb.
342
Benjamin Meisnitzer
According to a cognitive ToM approach, both modal verbs and modal particles imply the displacement factor12 as well as the reference to the source of the asserted information (cf. Abraham 2002: 295). Therefore they can be subsumed as techniques for negotiating shared knowledge. In the following chapter I will discuss the existence and the status of modal particles in the Romance languages.
2. Modal particles in the Romance languages Over the last decades, linguistic research in the study of German has led to the identification of a small class of words which are generally referred to as modal particles, whereas grammarians and lexicographers of Romance languages usually have ignored the existence of such particles13 (cf. Franco 1989: 240). However, a number of lexemes in the various Romance languages might qualify as modal particles in view of their syntactic behavior and semantic-pragmatic function, as will be shown in this chapter. The identification and categorization of modal particles is often di‰cult due to the existence of homonymous lexemes, generally adverbs. This, however, also holds for German modal particles. Functionally, modal particles and epistemic modal verbs can be put on one level, since they are used to ‘‘express the speaker’s mental attitude toward, or belief about, what he or she is saying’’ (Coniglio 2006: 57). Both position the speaker’s origo in relation to the statement (Diewald 1991: 250). However, modal particles, in contrast to epistemic modals, include the position of the addressee (threefold deixis) (Abraham 2009: 284). If the complex ToM process is taken into consideration modal particles allow the speaker to negotiate the truth-appreciation of his/her utterance. Truth-appreciation is located in two di¤erent persons: the speaker and the hearer (cf. Leiss, this volume). While they are similar to speakeroriented adverbials, they display a higher degree of grammaticalization and have peculiar characteristics that distinguish them from this class of adverbials. We can resort to the following criteria for defining modal particles/MPs: 1. MPs have a homonymous full semantic correspondent; 12. Partition of the speaker into two di¤erent persons assuming two viewpoints (double origo). 13. It has only been sporadically observed that some lexemes (such as It. mai and poi) present peculiar phonetic, semantic and syntactic characteristics distinguishing them from the traditional class of adverbs (cf. Burkhardt 1985: 265).
Modality in the Romance languages
343
in other words, they are polyfunctional (cf. Abraham 2009: 296), 2. MPs occur in specific positions in the clause (contrary to the source lexicals they are derived from; they have lost syntactic variability14) (Franco 1989: 248f.), 3. MPs may occur in restricted illocutionary contexts (cf. Abraham 2009: 296), 4. MPs are essentially restricted to root clauses (cf. Thurmair 1989: 44f.).15 We conclude that modal particles are full lexemes and that they are the result of a grammaticalization process in which they have lost syntactic mobility in comparison to the source lexemes (Wegener 1998: 37). On the pragmatic level they can strengthen or modify the illocutionary type of clauses and speech act operators (cf. Coniglio 2007: 138; Thurmair 1989: 49). They can primarily be found in spoken language and in what Koch & Oesterreicher call ‘‘communicative immediacy’’ (Na¨hebereich) (cf. Weydt 1969: 95f.; Koch/Oesterreicher 2007: 96f.). On the pragmatic level, modal particles have a communicative function, which is considered to be more important than their semantic value. As we have already noted, semantic bleaching goes hand in hand with any grammaticalization process. The process of semantic bleaching makes modal particles lose their denotative and referential meanwhile increasing their meta-communicative, pragmatic and illocutionary force (Wegener 1998: 43). Verbalization strategies determine the use and the pragmatics of modal particles (cf. Lo´pez Serena/ Borreguero Zuloaga 2010: 431). The aim of this chapter is to test potential candidates according to the criteria identified by modern research on modality. It will be tested whether they are restricted to root sentences. Furthermore, the range of illocutionary contexts will be specified. 2.1. Modal particles in the Iberian Romance languages In the Iberian Romance languages there seems to be a significant di¤erence between Spanish, Galician and Catalan, which exhibit very few modal 14. With regard to their syntactic behavior, the only generally accepted criterion to distinguish modal particles from adverbs in German is that they are confined to the middle field of the clause (Abraham 1988: 457). Since there is no middle field in Romance languages, this criterion cannot be adopted to detect potential modal particles. 15. Modal particles are root phenomena that are licensed in a matrix clause. There are a few exceptions where modal particles can occur in full-embedded clauses (cf. Coniglio 2007: 110). In this case modal particles are limited to clauses displaying root properties and constituting independent speech acts (Coniglio 2008: 106).
344
Benjamin Meisnitzer
particles, and Portuguese, which features several highly grammaticalized modal particles (cf. I). In Portuguese we can find several candidates for modal particles (cf. I), which have homonymous lexemes, and are used by the speaker to negotiate the truth-value with the addressee. In view of the classificatory criteria, all these particles occur in restricted positions, in root sentences and have illocutionary restrictions. (I) Portuguese modal particles Modal particle
Homonymous lexeme
afinal
‘after all’ ‘although it was not expected’
Adverb ‘at last’; ‘all in all’
bem
‘so’; ‘indeed’
Adverb ‘well’
la´
‘all along’; ‘after all’
Adverb ‘there’
sempre
‘all along’
Adverb ‘always’
(11) Ele sempre/afinal vem a Lisboa. P.pr 3rd pers s MP V pres. 3rd pers s prep N ‘Lisbon’ ‘He is going to come to Lisbon.’ (12) Ele vem sempre a Lisboa. P.pr 3rd pers s V pres. 3rd pers s adv ‘always’ prep N ‘Lisbon’ ‘He always comes to Lisbon.’ (examples (24) and (25) from Franco 1989: 248) (13) Ele la´ foi para casa. P.pr. 3rd pers s MP V SP 3rd pers s prep N f s ‘home’ ‘He ended up going home.’ The Portuguese modal particles are restricted to declarative clauses and interrogative clauses maintaining the same word order (e.g. – Ele sempre vem para Lisboa?). They occur preceding the negative adverb or the finite verb. A change in the syntactic position implies a change of the lexeme’s function and of its word class (cf. (11) vs. (12)). sempre, afinal, bem and la´ express a contradiction between the speaker’s and the addressee’s appraisal/expectation (cf. ToM), and the truth-value of the proposition is located in both, as in (11). In both examples (11 and 13), the speaker is unsure about what he is saying and gives the addressee a chance to negotiate the truth-value of the proposition. The assessment of truth is
Modality in the Romance languages
345
therefore transitive. Example (13) can also express the idea that the speaker was not expecting the subject of the sentence to go home, thus reacting extremely surprised or having doubts about him really going home. The speaker wants to discuss the apparent evidentiality with the addressee, and gives him the chance to contradict him, or to negotiate the speaker’s astonishment about the fact that he is going home. In Spanish we find modal particles that are root phenomena and which always precede the main verb. They can be coordinated with each other. (14) Yo bien te lo p pr 1st pers s MP p pr ind obj p pr dir obj habı´a dicho. V past perfect 1st pers s ‘to say’ ‘I told you.’ By using the modal particle (14), the speaker wants to emphasize that he presumes the addressee’s awareness about the content of his proposition. The modal particle bien can only occur in declarative emphatic clauses and syntactically, it precedes the verb. The particle also functions as an adverb (‘well’). The same criteria can be established for the corresponding forms in Portuguese (bem) and Catalan (be´). Contrary to Beerbom’s classification (cf. Beerbom 1992: 460), Spanish pues is a highly frequent modal particle in imperative clauses (15), with a homonymous causal conjunction, used in the same syntactic and illocutionary context as French donc (cf. 2.2). (15) Sie´ntate pues! V imp 2nd pers s ‘to sit’ MP ‘Do sit down.’ The modal particle highlights the speaker’s surprise or impatience with regard to the fact that the addressee has not yet carried out the propositional action, since the speaker’s assessment is that the addressee knows what the speaker expects from him. In other words, pues can enforce the order, but it can also be used as a politeness strategy to weaken the imperative aspect implied in the clause, depending on the intonation. This use of pues is peculiar to Spanish. A similar use of the equivalents in Catalan, Galician or Portuguese is not possible. 2.2. Modal particles in French In French there seem to be fewer elements that might be regarded as modal particles than in Portuguese or Italian. Only bien, donc, and quand
346
Benjamin Meisnitzer
meˆme are used to negotiate shared knowledge or assumptions. bien as a modal particle occurs mainly in interrogative clauses: (16) Vous prendrez bien un P pr 2 pers pl V fut 2 pers pl ‘to take’ MP ind art m s petit ape´ritif ? adj m s N m s ‘Would you like an aperitif?’ (example from Dalmas 1989: 232) With the use of bien the speaker expresses his assumption of the addressee wanting an aperitif, while giving him the opportunity to negate the speaker’s assessment. The modal particle quand meˆme is used to increase the expectation of possible objections from the addressee, as the truth-value of the shared knowledge has to be negotiated with the hearer and is located in both hearer and speaker. (17) J’ ai quand meˆme dormi. p pr 1st pers s aux V 1st pers s MP V p part pres (¼passe´ compose´ [CP]) ‘I did sleep.’ (18) J’ ai dormi quand meˆme. p pr 1st pers s aux V 1st pers s pres V p part (¼CP) concessive adv ‘I slept anyway.’ (examples from Waltereit 2006: 81) Whereas quand meˆme is a concessive adverb in (18), it functions as a modal particle in (17), being syntactically restricted to the position between auxiliary and the main verb, preceding the comment or rheme of the clause. The use of the modal particle expresses the speaker’s conviction that the addressee did not expect him to sleep under the given circumstances, thus giving the addressee the chance to contradict the speaker’s assessment. In example (17) there is an explicit reference to the knowledge of the addressee (ToM), who, according to the speaker’s appraisal, has a contrary expectation. The contradiction here lies in the illocutionary act. While in (18) the source of the truth-value is the speaker only, in (17) he leaves a margin for the addressee to disagree and negotiate whether the circumstances would lead the addressee to think that the speaker did sleep or not. Unlike in (17), in (18) there is reference to a causal sequence that is canceled.
Modality in the Romance languages
347
Examples (17) and (18) demonstrate the subtle di¤erences between adverb and modal particle and therefore make a definite classification a di‰cult matter. However, the illocutionary and the syntactic restrictions in the use of quand meˆme as well as the assessment of the truth-value, located in two di¤erent people (ToM), allow us to define it as a modal particle. In imperative and in emphatic declarative clauses we can find the modal particle donc, which has a homonymous form functioning as a conjunction. donc is syntactically restricted to the position after the finite verb and is used to emphasize the speaker’s expectation that the addressee should take action: (19) Regardez donc! V imperat 2nd pers pl ‘to watch’ MP ‘(Well,) Look (out)!’ (example from Koch/Oesterreicher 2007: 99) At the same time the speaker gives the addressee a chance to contradict the speaker’s proposition, i.e. in (19) the addressee could explain why he is not looking. The addressee is given the chance to negotiate whether he is going to take action or not. 2.3. Modal particles in Italian In Italian, particles that could qualify as modals, such as poi, mai, pur and ben, can only occur in clauses that display the typical features of matrix clauses, i.e. in root contexts as in German or Dutch (see also Coniglio 2011). They occur at fixed syntactic positions in the clause. The only exception is mai, being situated between the auxiliary and the main verb like the other modal particles or following the inflected (main) verb. It can also occupy the position adjacent to the wh-element without changing the meaning of the utterance. Italian modal particles have homonymous lexemes (mai – temporal adverb ‘never’; poi – temporal adverb ‘then’; ben(e)16 – adverb ‘good’; pur(e) – adverb ‘also’). The particle mai only occurs in interrogative contexts, mainly in wh-questions and its function is to signal the speaker’s incapacity to answer the question as well as his assumption that the addressee will not be able to answer either:
16. In contrast to pur(e), ben only has a phonologically reduced form (cf. Coniglio 2008: 123).
348
Benjamin Meisnitzer
(20) Cosa significheranno mai Int pr ‘what’ V ‘to signify’ 3rd pers p MP Future quelle parole? dem.adj f pl N f pl ‘word’ ‘distance to the speaker’ and the addressee’ ‘What will those words mean?’ (example from Coniglio 2008: 108) poi is mainly found in wh-questions, but also in declaratives. Similar to mai it signals the speaker’s as well as the addressee’s inability to find an answer to the posed question, according to the speaker’s assessment: (21) Chi avra` poi telefonato? Int pr ‘who’ aux. V 3rd pers s fut MP V p part ‘to call’ ‘Who might have called?’ (example from Coniglio 2008: 112) It can also make the speaker’s concern or interest with respect to the inquired information explicit. In both cases, there is a transitive assessment of the truth-value (cf. ToM). The particle pur(e) occurs in declarative clauses to signal that the speaker, although convinced of its truthfulness, has no evidence to prove that the assertion is in fact true (22). In concessive and imperative clauses we find the full form (23). (22) Deve aver pur letto il mod v 3rd pers s pres inf ‘to have’ MP V p part ‘to read’ libro. def art s m N ‘book’ ‘He/ She must have read the book.’ (23) Lascialo pure sul V imperat 2nd pers s þ p pr dir obj m (lo) MP prepþdef art m s tavolo! N s m ‘table’ ‘Leave it on the table!’ (examples from Coniglio 2008: 115–116) In the imperative sentence (23) the modal particle attenuates the order, providing the addressee with space for negotiation. In emphatic declarative clauses the modal particles ben and pur can be combined:
Modality in the Romance languages
349
(24) Lui aveva pur ben detto P pr 3rd pers s V imp 3rd pers s MP MP p part ‘to say’ Nom ‘to have’ che non voleva andare al cine. conj adv neg V imp 3rd pers s V inf ‘to go’ prep N m s ‘cinema’ ‘want to’ [aprep þil def.art m s] ‘He insisted that he did not want to come to the cinema.’ (example from Coniglio 2011: 93) According to Coniglio (2011: 93), in such a case of coordinated syntactic structure, a change in the linear order does not result in a change of meaning ( pur ben ¼ ben pur in example (24)). Nevertheless, if both particles are unstressed, the order pur ben is favored. 2.4. Modal particles in Romanian In Romanian the majority of the candidates for modal particles found in grammars (cf. Engel/Sta˘nescu et al. 1993: 947–950) and in linguistic literature on modality (cf. Sta˘nescu 1989: 272 ¤.) are modal adverbs, since they unfold a simple deixis and do not indicate the source of evidence on which the inferential process in the speaker is based on. Moreover, there are interjections which are used at the beginning of sentences only. They influence the intonation and subsequently the reading of the sentence. The only candidates for modal particles in Romanian are pa˘i, ma˘i and possibly doar. They all have homonymous adverbs: doar ‘only’, ma˘i ‘to a greater extent’ and the temporal pa˘i ‘later’ (from apa˘i < apoi). The modal particle ma˘i is used in imperative, interrogative and emphatic declarative clauses preceding the finite verb. By using the modal particle ma˘i, the speaker in (25) expresses that there is a contradiction between the speaker’s and the addressee’s will, according to the speaker’s assessment. According to the ToM approach, by using the modal particle the speaker gives the addressee a chance to negotiate the order: ˆın pace! (25) Dar ma˘i la˘sa-ma˘ Adv Conj MP V imperat ‘to leave’-P.pr. unstr ac Prep N m s ‘peace’ ‘Please, leave me alone!’ The modal particle pa˘i is restricted to emphatic declarative and interrogative clauses. Like ma˘i it occurs clause-initially. It is used to express the rhetorical character of the question and emphasizes the appraisal of the
350
Benjamin Meisnitzer
speaker as opposed to a di¤erent opinion of the addressee (cf. Thun 1984: 127), although it opens the possibility of negotiating the truth-value, which is based on the viewpoint of the speaker and the hearer. (26) A: Iˆti place P.pr 2nd pers unstr dat V pres. 2nd pers s ‘to like’ carnea? N s f ‘meat’ [carne noun þ a def. art. fem. sing.] B: Pa˘i cum nu? MP adv ‘like’ adv ‘not’ ‘A: Do you like meat? // B: Guess!?’ Finally, the modal particle doar is restricted to emphatic declarative clauses and occurs clause-initially or after verbs of knowledge. The modal particle is used to express the speaker’s conviction that the addressee knows what he is asking and it can also be used to express a discrepancy between what the speaker is saying and the expected or previously expressed opinion of the hearer (cf. Thun 1984: 117). (27) A: De ce spui asta? int. adv. V pres 2nd pers s ‘to say’ dem.pron B: Doar s¸tii prea bine! S¸tii doar! MP V pres 2nd pers s ‘to know’ adv ‘much’ adv ‘well’ ‘A: Why do you say that? // B: You know it/why! You know it/why!’ The speaker signals this di¤erence of opinion, giving the addressee the chance to change his position or to contradict the speaker’s conviction about the addressee’s assessment. 2.5. Di¤erent grammaticalization pathways and di¤erent levels of grammaticalization of bien in Spanish and French To conclude, in this section we will take a more detailed look at Fr. and Span. bien in order to show that the same source lexeme can follow di¤erent grammaticalization pathways (even within one language). Furthermore it will be demonstrated that modal particles in di¤erent languages can be at di¤erent levels of grammaticalization although they originate from the same root lexeme.
Modality in the Romance languages
351
Fr. bien can be an adverb meaning ‘a lot, to a large extent’ as well as a homonymous modal particle. Span. bien apart from being an adverb (source lexeme) and a modal particle (as shown previously, cf. 14), can also function as a discourse marker (cf. 28). It is often used to structure and arrange the discourse, for example, to close a topic addressed in the preceding contribution (28) or to introduce a new topic, rather than as a signal of approval of the speaker’s previous argumentation (cf. Detges/ Waltereit 2009: 44). In this function, it is a discourse marker and has to be distinguished from modal particles, because according to the ToM approach there is no shifting or displacement of the origo involved. At the same time, it is not an adverb, because it does not modify the sentence or the verb on a semantic level, like a sentence adverb, and it does not modify the VP like a modal adverb. (28) Bien discourse marker . . . creo que hubo un tema que no quedo´ demasiado claro . . . ‘Well, I guess there was a subject that did not remain very clear . . .’ (example from Waltereit/Detges 2007: 62) As a discourse marker bien has the function to ‘‘overtly indicate the relationship of a given chunk of discourse/text to a wider stretch of the same discourse/text’’ (Detges/Waltereit 2009: 44). While modal particles are used to assess and negotiate shared knowledge (cf. (14) and (29)), discourse markers are techniques used by speakers to negotiate their further verbal interaction (28) (cf. Detges/Waltereit 2009: 59). In French, bien is situated on the boundary between modal particle and focus particle. In (29), bien rec¸u builds a syntactic unit – a behavior typical of modal particles – with bien accentuating the lexeme which it precedes, like a focus particle. Bearing in mind the ToM-complex, bien opens the possibility for the addressee to negotiate the truth-value. (29) Vous avez bien rec¸u mon p pr 2nd pers p aux V 2nd pers pl MP p part ‘to receive’ poss adj m s pres ‘to have’ (¼passe´ compose´ [CP]) message? N m s ‘message’ ‘Have you received my message?’ (example from Waltereit/Detges 2007: 63)
352
Benjamin Meisnitzer
By using the modal particle bien the speaker expresses his expectation of a positive answer and his conviction of the factuality of his utterance. It occurs in assertive contexts as well as in yes/no-questions (cf. Detges/ Waltereit 2009: 45). From the examples discussed above we can conclude that Span. bien has followed two grammaticalization pathways: One leading to a discourse marker, and the other to a modal particle. Fr. bien became a modal particle forming a tight syntactic unit with the main verb rec¸u and showing less semantic integrity than its Spanish counterpart. Comparing examples (29) and (14) we notice that in French the semantic bleaching which the source lexeme has undergone is more advanced than in Spanish. With regard to Italian, we see that the modal particle ben has lost phonological substance (cf. footnote 17), indicating an even more advanced grammaticalization process. 2.6. Conclusion: Are there modal particles in the Romance languages? The aim of this section was to clarify which linguistic items in the Romance languages fulfill the criteria in order to qualify as modal particles. According to ToM, modal particles are used to negotiate the truth-appreciation of shared knowledge with the assessments of the truth located in two di¤erent persons: speaker and hearer. The defining criteria were deduced from studies on German modal particles, since German modal particles are well developed and thus considered to be prototypical. As shown in this paper, there are particle-like lexemes in the Romance languages whose features are very similar to those of German modal particles. They can therefore be classified as modal particles. The following characteristics are shared by all modal particles discussed in this paper: On the pragmatic level they are used to negotiate truth-value in cases of transitive assessments, they are root phenomena which hardly occur in full embedded clauses, they occur in a restricted set of illocutions under syntactic restrictions (cf. II below) and they all have a homonymous lexeme, generally an adverb. Other features shared by the modal particles discussed in this paper are the loss of scope and predicativity and the fact that they have a very low degree of syntagmatic variability compared to the source lexeme.
Modality in the Romance languages
353
(II) Overview of modal particles in the Romance languages17 Language
Portuguese
Modal particle
Illocutionary restriction
afinal, sempre
Declarative and interrogative clauses maintaining the word order of the corresponding declarative clause
Used preceding the negative adverb or the finite verb
Emphatic declarative clauses
Used preceding the verb
pues
Imperative clauses
Used sentence-finally
bien
Emphatic declarative and interrogative clauses
quand meˆme
Emphatic declarative clauses
Used after the finite verb or between auxiliary and main verb preceding the comment or rheme
donc
Imperative clauses
Used sentence-finally
ben, pur
Emphatic declarative clauses
mai
Interrogative clauses
poi
Mainly interrogative clauses with some occurrences in declarative clauses
ma˘i
Imperative, interrogative and emphatic declarative clauses
pa˘i
Emphatic declarative and interrogative clauses
doar
Emphatic declarative clauses
la´
Syntactic restriction
bem Catalan
be´ bien
Spanish
French
Italian
Romanian
Used preceding the main verb (with compound tenses after the finite verb)
Mainly used sentenceinitially and always preceding the finite verb
Used sentence-initially or following verbs of knowledge
17. This survey does not claim to give a complete overview of modal particles in the Romance languages, because there are still other potential candidates. The table contains only the modal particle-like lexemes discussed in this paper, qualifying as modal particles according to the criteria outlined above.
354
Benjamin Meisnitzer
The number of modal particles varies considerably from language to language, and they do not derive from the same source lexemes. Moreover, the same lexeme can follow di¤erent grammaticalization pathways, and the same modal particle can be located at di¤erent stages of the grammaticalization process, as the cross-linguistic perspective has shown in the analysis of Fr. and Span. bien. As the modal particle-like lexemes are rather rare in each of the Romance languages discussed in this paper, they do not constitute a paradigmatically ordered word class like in German. Nevertheless, there seems to be evidence of ‘undiscovered’ modal particles (e.g. Port. ca´, sim, pois; Span. sı´, ya; Fr. mais and It. sı`, mica, proprio, vai, ma, magari, appunto, certo among others), in the Romance languages, making further research necessary to explain their semanticpragmatic function.
3. Conclusion and perspectives: Modality in the Romance languages In this paper the ToM complex was introduced into the discussion of modality, since it revealed to be a very useful tool to decide whether a verb is modal or not and whether a lexeme qualifies as a modal particle or not. The first question we intended to answer was whether there are ‘real’ modal verbs in the Romance languages and whether they form a paradigm like in the Germanic languages. We have shown that only the verbs that derive from Latin posse (Port., Span. and Cat. poder, Fr. pouvoir, It. potere and Rom. a putea) qualify as polyfunctional modal verbs that fulfill the semantic, pragmatic and morphosyntactic criteria for defining modal verbs. Nevertheless, the restrictions concerning the selection of DM and EM readings show that even these modals are not as grammaticalized as German modals. The modal derivative of Latin debe#re is polyfunctional in Portuguese, Spanish, French and Italian, but not in Romanian and Catalan. Thus Rom. a trebui allows a deontic reading only (8), whereas Cat. deure triggers an epistemic reading (cf. (9) and (10)). Compared to the modals derived from Latin posse, there are fewer restrictions in the selection of full verbs to express DM or EM. The occurrence at fixed syntactic positions closely bound to the main verb, shows that both modals have lost their syntagmatic variability and have undergone a reduction in scope. According to Lehmann (1985: 306¤.), this results in a loss of predicativity, which corresponds to a loss of integrity on the paradigmatic level. However, one or two modal verbs are not enough to form a proper paradigm. Thus we can confirm the assumption made by Leiss (2008: 16) that,
Modality in the Romance languages
355
in contrast to the Germanic languages, the Romance languages do not have a real paradigm, but rather some isolated modals (cf. ‘‘muss-kannUrmodalita¨t’’ which is probably a cross-linguistic phenomenon (Abraham 2009: 296)). In the second part of the paper we have discussed the existence of modal particles in the Romance languages. In conclusion, it can be stated that there are a few particle-like lexemes that can be considered as being similar to German modal particles. However, they do not represent a specific paradigmatically ordered class of words in the Romance languages as they do in the Germanic languages (cf. also Koch/Oesterreicher 2007: 97; Sta˘nescu 1989: 270). Speakers often use alternative strategies and techniques to negotiate shared knowledge when the addressee’s opinion counts for the truth-assessability (see Waltereit 2006). Nevertheless, it is desirable to pursue the investigation of modal particle-like lexemes in the Romance languages, since there are still several potential candidates (e.g. Port. ca´, sim, pois; Span. sı´, ya; Fr. mais and It. sı`, mica, proprio, vai, ma, magari, appunto, certo, among others), many of which are rarely, if ever, discussed in literature on modality. References Abraham, Werner 1988 ‘‘Vorbemerkungen zur Modalpartikelsyntax im Deutschen’’. In: Linguistische Berichte 118. 443–465. Abraham, Werner 2002 ‘‘Modal verbs: Epistemics in German and English’’. In: Barbiers, Sjef; Beukema, Frits & Wur¤, Wim van der (eds.): Modality and its Interaction with the Verbal System. Amsterdam: John Benjamins (Linguistik Aktuell/Linguistics Today; 47). 19–50. Abraham, Werner 2008a ‘‘Aspektuelle und sprecher- bzw. persongebundene Bestimmungskomponenten deutscher Modalverben’’. In: Dekker, Kees/MacDonald, Alasdair/Niebaum, Hermann (eds.): Northern Voices: Essays on Old Germanic and Related Topics, O¤ered to Professor Tette Hofstra. Leuven, Paris, Dudley: Peeters. 327–347. Abraham, Werner 2008b ‘‘Modale, Tempus und Aspekt: Markiertheitsbefunde im Romanischen, Englischen und Deutschen’’. Romanistisches Jahrbuch 2008. 71–100. Abraham, Werner 2009 ‘‘Die Urmasse von Modalita¨t und ihre Ausgliederung. Modalita¨t anhand von Modalverben, Modalpartikeln und Modus. Was ist das Gemeinsame, was das Trennende, und was steckt dahinter?’’.
356
Benjamin Meisnitzer
In: Abraham, Werner & Leiss, Elisabeth (eds.): Modalita¨t – Epistemik und Evidentialita¨t bei Modalverben, Adverb, Modalpartikel und Modus. Stau¤enburg: Tu¨bingen (Studien zur deutschen Grammatik; 77). 251–302. Beerbom, Christiane ¨ bersetzungsprobleme. Eine kontrastive Studie 1992 Modalpartikeln als U zum Sprachenpaar Deutsch – Spanisch. Frankfurt am Main, Berlin, New York, Paris: Lang (Heidelberger Beitra¨ge zur Romanistik; 26). Beldarraı´n Jime´nez, Roquelina 1979 ‘‘Intento de descripcio´n confrontativa de los verbos modales en alema´n y en espan˜ol’’. Fremdsprachen 23. 271–277. Burkhardt, Armin 1985 ‘‘Der Gebrauch der Partikeln im gesprochenen Deutsch und im gesprochenen Italienisch’’. In: Holtus, Gu¨nter & Radtke, Edgar (eds.): Gesprochenes Italienisch in Geschichte und Gegenwart. Tu¨bingen: Narr. 236–275. Coniglio, Marco 2006 ‘‘German Modal Particles in the Functional Structure of IP’’. University of Venice Working Papers in Linguistics 16. 57–95. Coniglio, Marco 2007 ‘‘German Modal Particles in Root and Embedded Clauses’’. In: University of Venice Working Papers in Linguistics 17. 109–141. Coniglio, Marco 2008 ‘‘Modal Particles in Italian’’. University of Venice Working Papers in Linguistics 18. 91–129. Coniglio, Marco 2011 Die Syntax der deutschen Modalpartikeln. Ihre Distribution und Lizensierung in Haupt- und Nebensa¨tzen. Berlin: Akademieverlag (Studia Grammatica; 73). Dalmas, Martine 1989 ‘‘Sprechakte vergleichen: ein Beitrag zur deutsch-franzo¨sischen Partikelforschung’’. In: Weydt, Harald (ed.): Sprechen mit Partikeln. Berlin, New York: de Gruyter. 228–239. Detges, Ulrich/Waltereit, Richard 2009 ‘‘Diachronic Pathways and Pragmatic Strategies: Di¤erent Types of Pragmatic Particles from a Diachronic Point of View’’. In: Hansen, Bjo¨rn; Maj-Britt Mosegaard/Visconti, Jacqueline (eds.) Current Trends in Diachronic Semantics and Pragmatics. Bingley: Emerald. 43–61. Diewald, Gabriele 1991 Deixis und Textsorten im Deutschen. Tu¨bingen: Niemeyer (Reihe germanistische Linguistik; 32).
Modality in the Romance languages
357
Engel, Ulrich & Sta˘nescu, Sperant¸a et al. 1993 Kontrastive Grammatik deutsch-ruma¨nisch. Volume 1 and 2. Heidelberg: Groos. Franco, Anto´nio 1989 ‘‘Modalpartikeln im Portugiesischen – Kontrastive Syntax, Semantik und Pragmatik der portugiesischen Modalpartikeln’’. In: Weydt, Harald (ed.): Sprechen mit Partikeln. Berlin, New York: de Gruyter. 240–255. Hammerich, Louis ¨ ber die Modalverba der neugermanischen Sprachen. (Mit 1960 ,,U besonderer Beru¨cksichtigung des Da¨nischen)‘‘. Zeitschrift fu¨r deutsche Wortforschung 16. 46–70. Johnen, Thomas 2003 Die Modalverben des Portugiesischen (PB und PE). Semantik und Pragmatik in der Verortung einer kommunikativen Grammatik. Hamburg: Dr. Kovacˇ (Philologia – Sprachwissenschaftliche Forschungsergebnisse; 60). Koch, Peter & Oesterreicher, Wulf 2007 Lengua hablada en la Romania: Espan˜ol, France´s, Italiano. Madrid: Gredos (Biblioteca Roma´nica Hispa´nica – Estudios y Ensayos; 448). Lehmann, Christian 1985 ‘‘Grammaticalization: Synchronic variation and diachronic change’’. In: Lingua e Stile 20. 303–318. Leiss, Elisabeth 2008 ‘‘The silent and aspect-driven patterns of deonticity and epistemicity: A chapter in diachronic typology’’. In: Abraham, Werner/Leiss, Elisabeth (eds.): Modality-Aspect Interfaces. Implications and typological solutions. Amsterdam: Benjamins (Typological Studies in Language; 79). 15–41. Leiss, Elisabeth 2009 ‘‘Drei Spielarten der Epistemizita¨t, drei Spielarten der Evidentialita¨t und drei Spielarten des Wissens’’. In: Abraham, Werner/ Leiss, Elisabeth (eds.): Modalita¨t – Epistemik und Evidentialita¨t bei Modalverben, Adverb, Modalpartikel und Modus. Stau¤enburg: Tu¨bingen (Studien zur deutschen Grammatik; 77). 3–24. Lo´pez Serena, Araceli & Borreguero Zuloaga, Margarita 2010 ‘‘Los marcadores del discurso y la variacio´n lengua hablada vs. ´ scar/Acı´n Villa, Esperanza lengua escrita’’. In: Loureda Lamas, O (eds.): Los estudios sobre marcadores del discurso en espan˜ol, hoy. Madrid: Arco/Libros. 415–495. Milan, Carlo 2001 Modalverben und Modalita¨t. Eine kontrastive Untersuchung Deutsch-Italienisch. Tu¨bingen: Niemeyer (Linguistische Arbeiten; 444).
358
Benjamin Meisnitzer
Motapanyane, Virginia & Avram, Larisa 2001 ‘‘The Syntax of putea and its Mixed Typology’’. In: van der Auwera, Johan/Dendale, Patrick (eds.): Modal Verbs in Germanic and Romance Languages. Amsterdam: John Benjamins (Belgian Journal of Linguistics; 14). 149–165. Oliveira, Fa´tima 2001 ‘‘Some Issues about the Portuguese Modals dever and poder’’. In: van der Auwera, Johan/Dendale, Patrick (eds.): Modal Verbs in Germanic and Romance Languages. Amsterdam: John Benjamins (Belgian Journal of Linguistics; 14). 167–184. Papafragou, Anna 2002 ‘‘Modality and Theory of Mind. Perspectives from language development and autism’’. In: Barbiers, Sjef/Beukema, Frits/ Wur¤, Wim van der (eds.): Modality and its Interaction with the Verbal System. Amsterdam: John Benjamins (Linguistik Aktuell/Linguistics Today; 47). 185–204. Pietrandrea, Paola 2005 Epistemic modality. Functional properties and the Italian system. Amsterdam: John Benjamins (Studies in Language Companion Series; 74). Reichenbach, Hans 1960 Elements of Symbolic Logic. New York: Macmillan. Sctrick, Robert 1971 ,,Quelques proble`mes pose´s par la description de surface du syste`me des modalite´s en franc¸ais‘‘. In: Langue franc¸aise 12. 112–125. Staˇnescu, Sperant¸a 1989 ‘‘Zum Status der Partikeln im Deutschen und im Ruma¨nischen’’. In: Weydt, Harald (ed.): Sprechen mit Partikeln. Berlin: de Gruyter. 267–275. Thun, Harald 1984 Dialoggestaltung im Deutschen und im Ruma¨nischen: eine strukturell-kontrastive Studie zu den Existimatoren. Tu¨bingen: Narr (Tu¨binger Beitra¨ge zur Linguistik; 239). Thurmair, Maria 1989 Modalpartikeln und ihre Kombinationen. Tu¨bingen: Niemeyer (Linguistische Arbeiten; 223). Waltereit, Richard 2006 Abto¨nung. Zur Pragmatik und historischen Semantik von Modalpartikeln und ihren funktionalen A¨quivalenten in romanischen Sprachen. Tu¨bingen: Niemeyer (Beihefte zur Zeitschrift fu¨r Romanische Philologie; 338). Waltereit, Richard/Detges, Ulrich 2007 ‘‘Di¤erent functions, di¤erent histories. Modal particles and discourse markers from a diachronic point of view’’. In: Catalan Journal of Linguistics, 6. 61–80.
Modality in the Romance languages
359
Wegener, Heide 1998 ‘‘Zur Grammatikalisierung von Modalpartikeln’’. In: Barz, ¨ hlschla¨ger, Gu¨nther (eds.): Zwischen Grammatik Irmhild & O und Lexikon. Tu¨bingen: Niemeyer (Linguistische Arbeiten; 390). 37–55. Weydt, Harald 1969 Abto¨nungspartikel. Die deutschen Modalwo¨rter und ihre franzo¨sischen Entsprechungen. Bad Homburg: Gehlen (Linguistica et litteraria; 4).
The epistemological treatment of information and the interpersonal distribution of belief in language: German Modal Particles and the typological challenge Pierre-Yves Modicom Universite´ Paris-Sorbonne, E´cole Normale Abstract The present chapter contains a discussion of the links between the di¤erent domains of the epistemological treatment of information in speech and discourse. It is based on a confrontation of concepts coined by Guentche´va & Landaburu (2007) with a set of German modal particles. According to Guentche´va & Landaburu, Theory of Mind is part of a wider domain of ‘‘epistemological treatment of information’’ together with mirativity and evidentiality. The concepts created by typologists are used to describe the distribution of modal functions between a few German modal particles examined in pairs. Both components of each pair have the same value in terms of Theory of Mind. The analysis makes the case for an internalist, monosemantic description of the meaning of those particles. Their di¤erent values in context are postulated to be inferable from the combination of their core meanings with other syntactic and illocutionary parameters. It is finally shown that the complex meaning of German modal particles is best accounted for in a triangular framework including Theory of Mind, mirativity, and evidentiality as the permanent parameters followed by the speaker to track the distribution of knowledge between herself and the hearer. 1. Introduction German Modal Particles (henceforth GMPs) are short, uninflected words with specific distributional properties and a modal meaning. Among other specificities, they are morphologically identical with lexemes belonging to other morphosyntactic categories like, for instance, ‘‘yes’’ ( ja), ‘‘well’’ (wohl ), ‘‘plain’’ (eben). They have now been intensively investigated for four decades, following Weydt’s first monograph on this topic in 1969. GMPs are now widely considered to encode (at least parts of ) the felicity
362
Pierre-Yves Modicom
conditions of the utterance they form part of. They do not contribute to the propositional content of the utterance; rather, they seem to intervene in the interaction between speech act participants (see Abraham 1991a). Two central interpretation factors have been proposed so far to characterize their pragmatic contribution. On the one hand, they encode what the speaker knows to be the state of her own knowledge and often what she assumes the hearer’s propositional attitude to be. On the other hand, they also seem to contribute to the overall politeness and vivification of the exchange. The latter interpretation can easily be shown to follow from the first: The fact that the speaker takes into consideration what she assumes to be her partner’s state of knowledge is a signal of willingness to cooperate. According to this interpretation, GMPs are fundamentally related to Theory of Mind. Nevertheless, research on this field (see Zimmermann 2008) has also shown how those particles interfere with the speaker’s truth-functional evaluation of the propositional content of her utterance. Apparently, GMPs do not mark the sole intersubjective distribution of knowledge. They are not restricted to the first or second persons, but also encode the third-person status assigned to the propositional content: What matters is not only whether the speaker or the hearer had knowledge of the proposition before or whether they could have expected it, but also with which value it was or will be part of the common knowledge ground. So far, GMPs seem to be related to both Theory of Mind and the truth functional evaluation of the propositional content. Both dimensions constitute the topics of the present study. Independently, the question of the borders and bridges between the subdomains of knowledge evaluation has often been addressed by typologists in the last years, (see Evans 2007, 2009 and Landaburu & Guentche´va 2007). The aim of the present paper is to show how some of these typological attempts of definition and delimitation are worth comparing with data provided by GMPs. Those observations might match up with the domain of Foreign Conscience Alignment (FCA, see Abraham 2010 and this volume), defined as a global theory of three-person epistemic alignment in the scope of Theory of Mind and of Donald Davidson’s theory on the triangulation of knowledge, where language also plays a key role (see Davidson 2001c). 2. A typological account for the epistemological treatment of information 2.1. The realm of information treatment In the foreword to their typological overview, Landaburu & Guentche´va (2007: 1–3) introduce a set of definitions based on their works as well as
The epistemological treatment of information
363
on previous literature. They first discuss the concept of me´diation (or e´nonciation me´diatise´e, litt. ‘‘mediated utterance’’), which corresponds to what is usually called evidentiality. According to them, it regroups all phenomena marking reported speech, hearsay, inferential knowledge and indirect knowledge of any kind. The paradigm of knowledge sources plays a key role in the context of evidential characterization, but in many languages, this domain seems to have relations to other questions such as assertive commitment or modality. Following Chung & Timberlake (1985), Landaburu & Guentche´va (2007: 2) subsequently propose to define a broad category of ‘‘epistemological treatment of information’’ covering the whole realm of ‘‘the grammatical expression of the qualities which the speaker assigns to her own knowledge of the information she is transmitting’’. This domain has to be divided into three parts: The first of them deals with the speaker’s expression of her commitment to the propositional content of the utterance. As such, it is closely related to epistemic modality as well as to what the authors consider to be the ‘‘intersubjective relation between the speaker and the co-speaker’’. There are thus two branches in this first domain: The scale of commitment on the one hand, and Theory of Mind on the other. The second domain is that of evidentiality stricto sensu: It highlights the conditions under which the speaker has acquired the piece of information she is transmitting, and especially the source of that knowledge. Both dimensions have been treated separately by typologists for many years now. A third dimension has been addressed more recently and is still being discussed: that of expectability or surprise, whose autonomy as ‘‘the mirative’’ has been advocated by De Lancey (2001), mainly on the basis of Tibetan data. 2.2. Illustration from Andoke The division between the first and the second type is illustrated by Andoke, the Columbian language studied by Landaburu. In this language, the subject of the verbal predicate is always encoded by a personal marker to the left of the verb. TAM categories are distributed between this subject marker and the verbal lexeme. The verb is morphologically marked for aspect and its inflection varies according to the mood (Realis vs. Irrealis). Tense, modality and optionally prospective modality are marked by suffixes at the very end of the subject lexeme. At this stage, the propositional content is defined in the real world and can already be granted a truth value. There is still no FCA marker and at the same time no illocutive force has been assigned to what is not an utterance yet, but can already
364
Pierre-Yves Modicom
be used as a subordinate clause. Following Dik’s model (see Dik 1989), Landaburu explains that what is now at stake is the closure of the proposition and its transformation into a felicitous utterance. This is coded in two distinct ways corresponding to the first two epistemological domains: A commitment prefix has to be added to the subject for the sentence to be grammatically acceptable as a main or a free clause, and an optional evidential su‰x might come immediately after the core morpheme of the subject lexeme. (1) dui t bß da¨ß ‘‘White people are coming’’ (Landaburu 2007: 24) The structure of this sentence can be analyzed as follows: (10 ) dui’t
b-
White:Pl (class.3-3)
b-
(Noun) NP
ß
-Ø
-(Ø)
Subj:class.3-3 dir.exp nPot
com subject.trace
evid
Subject
pot
-Ø
da¨-
ß-
Pres
Ingr
move
Agr:class3
tense
asp
(Verb)
agreement
VP
There are five possible commitment markers. The standard one, standing for strong commitment, is b-. When the speaker assumes that the hearer has no possibility to know whether she is saying the truth, she might use ke˜-, which marks that the speaker has the monopoly of knowledge; the hearer is reported to have no access to the described matter of facts, whereas the speaker is strongly committing herself to her utterance. The marker ba˜- is used when the speaker is questioning herself on the validity of what she is saying, and k- and d- mark the fact that the speaker (i) does not know of the truth of what she is saying and (ii) assumes that the hearer has more to say on this topic. Building sentences with one of those markers as commitment prefixes is the normal way to form a question in Andoke, without further syntactic changes. The marker k- stands for global interrogation, whereas d- restricts the focus of the question to one argument of the predicate. It is therefore possible to arrange those markers along a scale representing how the speaker evaluates her own knowledge, with ke˜- at the top and k- at the bottom. Yet, as Landaburu puts it, another classification is more relevant, based on Theory of Mind and reproduced in the following chart:
The epistemological treatment of information
365
If one wants to take into account the complexity of the encountered phenomena, one should integrate the intersubjective relationship within the basic oppositions and arrange the values of the four prefixes alongside two polar axes dealing with the speaker’s knowledge as well as with the I/Youconnection. [. . .] Together with the speaker’s knowledge, what is at stake here is thus the relationship of epistemic authority between the speaker and the hearer (Landaburu 2007: 30). [þKnow; I]
[–Know; I]
[þKnow; You]
b-
k- / d-
[–Know; You]
ke˜-
ba˜-
This account makes the case for the decisive role of FCA in language, in as much as the utterance is grammatically incorrect if there is no overt coding of the speaker’s FCA-stance. Illocutive force is defined as the speaker’s concern with the integration of the propositional content of her utterance in the common knowledge ground of the speech act participants. Interestingly enough, evidentiality appears as a secondary category in Andoke: It is overtly marked only if the direct source of knowledge is not the speaker’s own experience, but hearsay (-ha´-) or rational inference (-dı˜-). It might also be derived from morphemes originally coding FCA: Ba˜ can appear before the speaker-a‰rmative FCA-markers b- or ke˜- and turn the proposition into a postulate, the combination FCA-FCA can be understood as EV-FCA. According to Landaburu (2007: 43), there are as many intersubjective commitment markers as illocutionary independent clauses, and no more. For instance, in a relative clause, the subject lexeme is only deprived of its commitment marker, whereas the subject of the main clause presents all the usual morphemes. This syntactic data leads Landaburu to make the case for a hierarchy of morphemes with intersubjective commitment markers at the top, corresponding to Dik’s fifth and last level of constitution. This empirical data corroborates Landaburu & Guentche´va’s theory on the subdivision of epistemological patterns, which was then commented upon by Evans (2009). Evans uses the concept of engagement, re-using the French word used by Landaburu, which has been translated in the present summary (as is usual) by ‘‘commitment’’. Generalizing data from several languages and authors, Evans (2009:2f.) defines engagement as ‘‘the mental directedness of speech act participants towards a denoted state of
366
Pierre-Yves Modicom
a¤airs’’, involving four parameters: the ‘‘type of cognitive modality’’; the ‘‘cognitive locus’’ (defined as the speaker, the hearer or a third party and ‘‘including speaker’s model of the hearer, presumed mutual-knowledge based representations, etc.’’); the ‘‘domain’’, roughly corresponding to the scope of the engagement marker; and the ‘‘main vs. complement status of the marked proposition.’’ Even though Theory of Mind is but one of four criteria in this representation, it still plays a decisive role which is all the more crucial for the purpose of this paper since Evans explicitly quotes the GMP doch as an example of such an ‘‘engagement marker’’ in Western European languages. Indeed, GMPs are often supposed to mark assumed intersubjective states of knowledge and to have a wide scope over the whole propositional content. It might therefore be relevant to take the data from Landaburu & Guentche´va as a challenge, and to try to apply typologically obtained categories to GMPs. A restriction which we shall observe in the comparison of GMPs with these concepts is the fact that they primarily insist on the sharing of positive knowledge, i.e. on whether the information is assumed to be known or not to the speech act participants. In the present paper, what will be focused on is rather the question of whether the propositional content is part of the participants’ set of beliefs, positive knowledge being a more determined kind of propositional attitude, yet always presupposing belief.1 The present paper will illustrate how GMPs fulfill this epistemological function with partially covert and diverse patterns and how FCA appears to be the unifying criterion to bring this diversity to a categorial solidarity. The following section(s) will focus on a few pairs of particles that seem to be equivalent as regards intersubjective commitment or Theory of Mind and yet di¤er when the whole scope of epistemological marking is concerned. The results will then be summarized and interpreted in a conclusive part.
1. As Davidson showed in his essays Three varieties of knowledge (Davidson 2001d) and First-Person Authority (Davidson 2001b), if there are good reasons to postulate that you know what you believe, estimating what is going on in someone else’s head is primarily a question of what she believes (and not what she knows), i.e. to which propositions she is committing herself (given her concept of objective truth, which Davidson presents as the third-person status in the triangulation of knowledge, grounding – and at the same time being grounded on – first- and second-person stances).
The epistemological treatment of information
367
3. Doch vs. schon 3.1. Schon: re-a‰rming potentially disputed propositions According to Thurmair (1989: 50) schon (originally meaning ‘‘already’’) introduces a ‘‘restriction over possible counterarguments’’,2 which OrmeliusSandblom (1997: 106) symbolizes as follows: (2) lp [PFakt Pp] Q‘‘it is not true that p is not true’’, Here, p is the proposition over which schon takes scope. This may explain why schon is very often used with epistemic verbs, as in the expression ich denke schon (‘‘I schon think’’, meaning roughly ‘‘I believe it altogether’’), or in the following example: (3) In den gro¨ßeren Sta¨dten wie Kapstadt oder Johannesburg darf man schon davon ausgehen, ‘‘In the bigger cities such as Cape Town or Johannesburg, one may schon suppose dass die Stadien auch weiterhin benutzt werden. . . . that the stadiums will be used further on.’’ (DR Kultur 12th July, 2010) Ormelius-Sandblom tends to regard schon as ‘‘speaker-oriented’’ since it encodes the attitude of the speaker towards the content of her utterance. Nevertheless, the dialectical dimension of schon, which addresses possible opposition to the speaker’s assertion, could also justify an intersubjective or interpersonal interpretation. In a study on schon, Pe´rennec (2002: 191) defended the idea that the core meaning common to all uses of schon as a GMP, but also its use as a temporal adverb, represents the trespassing of a limit beyond which what was not true suddenly becomes real. He consequently interprets the uses of schon as GMP as ‘‘hearer-oriented’’ and proposes the following interpretation for the use as a GMP: ‘‘Here, too, schon points beyond a limit, even though the limit is not designated here, be it as the utterance time thanks to deixis, be it thanks to an non-deictic adjunct. [. . .] The speaker uses as a fictive limit an element that might be called ‘‘the evaluation moment of the utterance’’, following Vuillaume’s ‘‘moment of verification of the utterance’’. I prefer ‘‘evaluation’’ rather than ‘‘verification’’ so as to make it clear that not only the mere truth value of the propo2. ,,Einschra¨nkung mo¨glicher Gegenargumente‘‘
368
Pierre-Yves Modicom
sition is at stake, but also the relevance of an utterance or of a speech act. This moment of evaluation, which is actually hearer-oriented, is hinted at by the speaker, who is putting herself in the role of the hearer. [. . .] With schon, the speaker makes it clear that this moment has already taken place, and that the evaluation has already given a positive result.’’
Commenting upon the case of assertive sentences, Pe´rennec writes: ‘‘From what has just been said, it is obvious that the speaker, when she marks the anticipation of the evaluation moment with schon, acts as the one who has knowledge as opposed to the ignoring hearer.’’ This overall interpretation of schon as the anticipation of possible doubts or reluctance of the hearer is advocated for by the DWDS dictionary, according to which schon, in assertive sentences, ‘‘reinforces an assertion against which a counterargument or a doubt has been expressed’’. 3.2. Does schon restrict the validity of p? Yet, according to some sources such as Grimm’s Wo¨rterbuch der Deutschen Sprache, schon can also entail a restriction of the validity of p: Even though its first modal meaning consists in re-a‰rming a proposition against contrary expectations or assumed doubts,3 according to the authors of the dictionary, schon can also ‘‘prepare an objection or a restriction’’ while admitting that the proposition is basically true.4 In this respect, schon rather seems to undermine the validity of the proposition. Yet, there is a strong suspicion that this supposed other meaning actually has to be regarded as a secondary, usage-based reinterpretation in context. First of all, this apparent contradictory meaning can result from the fact that when she uses this lexeme, the speaker makes explicit that she is aware of the hearer being possibly reluctant to agree with her. This revelation pragmatically implicates that the proposition is not beyond doubt and could therefore seem to undermine it in some cases – or in Pe´rennec’s words, it might indicate that the validity limit has just been exceeded. But a compositional interpretation can also be proposed to solve more problematic cases, for instance those where schon is used in a clause before the same speaker utters a second clause apparently restricting the validity of the first one. In such 3. ,,Es wird hier eine thatsache schlechthin der gegentheiligen, anderer erwartung, ausgesprochenem oder angenommenem zweifel gegenu¨ber hervorgehoben‘‘ (Vol. 15, 1459–1464). 4. Coniglio (2007: 1), following Thurmair (1989: 200), also assumes that there might be such a value of ‘‘validity restriction’’ for schon.
The epistemological treatment of information
369
cases, the authors of Grimm’s dictionary suppose that schon occurs in coconstructions with an adversative marker and therefore ‘‘prepares’’ the undermining of the proposition it is embedded in. This theory is illustrated by the following example from Lessing’s Nathan der Weise (1779): (4) Nathan: Dann, Daja, ha¨tten wir ein neues [Haus] uns / Gebaut; und ein bequemeres. N.: ‘‘Then, Daja, we would have bought us a new one [¼a house]; and a more comfortable one.’’ Daja: Schon wahr! Doch Recha wa¨r bei einem Haare mit / verbrannt. D.: ‘‘Schon true! But Recha could have burnt within a hair’s breadth.’’ (Quotation from the WDS, Vol. 15: 1463) The second speaker (Daja) might well be anticipating Nathan’s rea‰rmation of p (the fact that it would have been possible to have a new house built) being valid in spite of all counterarguments. This could suggest that the meaning of schon has to be reconsidered, or that it primarily corresponds to an external strategy of communication in context. Yet, another, more systematic explanation can be proposed. Schon p, aber q may be glossed as follows: ‘‘Indeed, in spite of all I could say to undermine it, it is not true that p is not true, and yet I a‰rm q in order to restrict this consensual validity to a precise domain.’’ Actually, p has already been asserted and must therefore be accepted as part of the common ground unless the second speaker strongly denies it, which she does not do here. She concedes the common ground status of p and then tries to restrict it with q. The most overt syntactic strategy corresponding to this pattern is the use of a concessive subordinate clause – inside which schon may occur, as explained by Coniglio (2007: 12). (5) Hans hat sich mit Depressionen in seinem Zimmer eingeschlossen, ‘‘Hans locked himself up in his room feeling depressed obgleich er das Examen schon bestehen wird. even though he will schon pass the exam.’’ (Borst 1985: 120, quoted in Coniglio 2007) In a concessive reasoning, the acknowledgement of the validity of p has dramatic consequences for the following developments: If p is part of the common knowledge ground, its positive truth value has been endorsed by the speaker, even though she wants to present counterarguments or
370
Pierre-Yves Modicom
restrict the validity of p. This corresponds to the value of schon. The only specificity of this kind of usage is the fact that both roles are played by one and the same speaker. Nevertheless, in spite of their being uttered by the same person, the status of both asserted propositions is not identical. The schon-marked one has already been introduced into the common knowledge ground and has therefore the status of a third-person truth, whereas the other clause is newly asserted under first-person commitment.5 The apparently special value of schon in such contexts can be compositionally explained if we assume that the case in (4) is just an informal equivalent of an overt concessive structure such as in (5), which should be regarded as the main case for such configurations. Concessive clauses represent a case of polyphony in the sense that the conceded proposition is endorsed by the third person. This third-person status proceeds from the concessive subordination which scopes over the rest, including schon. In this perspective, the third person strongly commits herself to the propositional content and rejects in advance any claim that forthcoming restrictions could undermine the validity of the proposition per se. Then the main illocutive act of the first person takes scope over the third-person proposition. This firstpersonal act contains the assertion of those restrictions, whose domain is now self-limited, since the third person is nothing but a convergence of the first with the second one. This means it is not schon that weakens the commitment to p, but the overall logical relation between both propositions, which might be overtly marked by a concessive clause. The interplay between the voice of the common ground and the voice of the first person can account for such usages without externalizing the motivations for the use of schon.6 There are not two values of schon, but only one, which is contextually interpreted with respect to whether the proposition over which schon scopes conveys old or new information, or rather, 5. The sentence whose acceptance as part of the common ground is only re-asserted corresponds to the subordinate clause in the concessive sentences involving overt subordination. The newly asserted proposition gets the status of a main clause, which can be seen as the surface equivalent of its pragmatic status as main assertive goal of the utterance. Yet, both are being asserted and can therefore include a GMP. That such uses of schon correspond to covert subordination can be argued for on the basis of other lexical phenomena such as the grammaticalization of obschon as a concessive complementizer equivalent to obwohl and obgleich. 6. Far from making the case for external strategies, this value of schon rather suggests that the personal voices at stake here are cognitive instances (analogous to the points of Davidson’s triangle?) and not (or at least not necessarily) concrete persons of the external world.
The epistemological treatment of information
371
whether the propositional content is already part of the assumed common knowledge ground. Thus, we can assume that schon always marks: (i) the rea‰rmation of the speaker’s commitment to p (ii) her assumption that the status of p could still be disputed by other speech act participants. In other words, the assumed interpersonal belief coordinates in the case of schon are relatively similar to what we see in Andoke with ke˜-.7 This also makes the case for the idea that the attributed third-person value of a proposition actually proceeds from the alleged distribution of belief and authority between both speech act participants, thus making Theory of Mind a basic category for modality in general. 3.3. Introducing mirativity: the case of doch Another GMP is fundamentally linked to the assumption that the hearer disagrees with the speaker, namely doch. In its basic, non-GMP use, doch is an adversative adverb meaning ‘‘nevertheless’’. The GMP can be identified when there is no explicit proposition to be countered, and when the lexeme appears in the characteristic GMP position on the left part of the middle field (whereas it is generally sentence-initial as an adverb) without being accentuated (see Abraham 1991b or Coniglio 2007 for syntactic considerations). It has a clear adversative dimension, as in the following example. (6) Ja, das ganze war ja bewusst als vertrauensbildende Maßnahme konzipiert, ‘‘Yes, the whole thing was conceived as a measure to bring back some confidence, man wollte zeigen, dass die Banken doch besser da stehen, one wanted to show that the banks are doch finer als das was gedacht wurde, und deswegen hat man ja auch die Tests so konzipiert, than what people thought, and therefore, one has also conceived the tests so, dass manchen Risiken gar nicht aufschneiden. that some risks do absolutely not come into part.’’ (DLF, July 24th, 2010.) 7. Under the restriction named above regarding knowledge vs. belief. In a beliefbased framework, we can say that ke˜- is felicitous in configurations where the speaker strongly believes what she is saying and assumes that the hearer does not necessarily believe it. Thus, the opposition of standpoints is probably more emphasized in German.
372
Pierre-Yves Modicom
It can also be used in questions or in imperative sentences (with deontic illocution). This last case exhibits the di¤erence between doch and schon. In imperative sentences, doch usually manifests impatience or anger at the addressee not doing what she is supposed to do. In questions, according to the Wo¨rterbuch der Deutschen Sprache, it often contains a latent reproach (see Vol. 2: 1204). Those two connotations tend to suggest that doch is used when the speaker expects the hearer to agree with her. Zimmermann (2011: 2015) summarizes this when he describes the triggering of doch as ‘‘an adversative attitude to certain background assumptions’’. An informal description would thus be that doch is felicitous i¤: (i) the speaker expected the proposition to be immediately accepted by the hearer on the basis of the assumed common knowledge ground and (ii) the latest developments tend to suggest this expectation of hers was false: The hearer is not aware of the validity of what should have been a matter of consensus. As a consequence, the speaker has to switch her second-person alignment from ‘‘from our common knowledge ground, my evaluation for the truth value of p must be shared by the hearer’’, a distribution that would classically be marked with ja, to the schon distribution: ‘‘I believe p is true but I assume the hearer has doubts about it’’. Doch encodes this switch in Foreign Conscience Alignment: Schon depends exclusively on the speaker’s FCA at the moment of utterance and remains underspecified as regards the past epistemic background, whereas doch is specified in time and can be felicitous if and only if the belief discrepancy between both speech act participants was unexpected by the speaker.8 This FCA-switch is not so di¤erent from what is now generally called ‘‘mirativity’’ (after De Lancey 2001), which designates all forms where the speaker manifests that the matter of facts the proposition accounts for was unexpected for him: Doch marks a brutal and unexpected revision of second-person epistemic marking. This confirms Evans’ suggestion that expectation is a dimension of ‘‘engagement’’ and constitutes an argument for the integration of mirativity into the first pool of epistemological categories sketched by Landaburu & Guentche´va (2007).
8. This time is not the propositional time (comparable to what narratologists call ‘‘narrated time’’) but the speaker’s time (so-called ‘‘narration time’’).
The epistemological treatment of information
373
4. The structure beyond apparent polyfunctionality of ja 4.1. A compositional account for ja’s apparent polyfunctionality The meaning of the GMP ja (‘‘yes’’) has been explored quite intensively in the previous literature. Two contexts of usage are generally distinguished, depending on whether ja occurs in exclamative or in declarative sentences. Ja encodes both strong commitment and the expectation that the addressee will not contradict the speaker. Given its syntax and felicity conditions, ja seems to occupy a very high place in the structural hierarchy, probably scoping over the illocutionary type as an assertion modifier (see Jacobs 1991: 141 ¤. and Zimmermann 2008). Ja highlights supposed agreement between the speakers. It not only expresses that p is consistent with or entailed by the common knowledge ground, thus being already virtually part of it, but it generally encodes the fact that the hearer is assumed to already believe p. Ja therefore has to be defined as a marker of assumed shared commitment, or bi-a‰rmativity. Nevertheless, ja should not be regarded as a strict cross-linguistic equivalent to Andoke b-, which is defined by Landaburu as the default commitment prefix and which encodes a sort of bi-a‰rmativity only insofar as it is opposed on the one hand to particles standing for the speaker’s ignorance, and on the other hand to ke˜-, which focuses upon the knowledge discrepancy in favour of the speaker. Much rather, b- seems to be the unmarked prefix, thus probably encoding the standard configuration of common ground compatibility and conversational cooperation, which is but a weakened form of assumed joint acceptance, the strong form being bi-a‰rmativity. One should not forget that in Andoke, the use of FCA markers is required for every illocutionarily autonomous sentence, whereas it is optional in German, where this underspecified type of commitment is rather expressed by the absence of a particle. In this configuration, ja is rather restricted to those cases where the speaker wants to focus on the first and the second person being on the same wavelength. Yet, those cases are typically what Grice’s maxim of quantity should rule out (see Grice 1975), since they just deal with redundant information. If we consider that overt Theory of Mind is not primarily an external, almost rhetorical strategy, but a systematic cognitive alignment of the epistemological coordinates of the exchange, and if we consider that Grice’s conversational rules then bias the performance of this (internal) alignment, we can reduce the several values of the particle ja to this core bi-a‰rmative meaning. More explicitly, given the rule of quantity, sentences whose propositional content is not only known to both speech act participants, but also believed or known to be such are
374
Pierre-Yves Modicom
more likely to be uttered in contexts where what matters is precisely this statutory recognition as common knowledge ground. This distortion accounts for context-bound bias. For instance, corpus-based studies show a very high proportion of usages in clauses whose textual function is thematic and which are often initial concessions preceding a contradiction, or reminders introducing questions: Ja is used when the existence of a common knowledge ground is actually what has to be focused on rather than the propositional content in itself. (7) Es gibt ja auch den Vorschlag, Medikamente erst mal vorla¨ufig zuzulassen. Wa¨re das eine Mo¨glichkeit? ‘‘There is ja also the possibility to authorize medicaments temporarily at first. Would that be a possibility?’’ (Bayerischer Rundfunk, Samstagsforum 12. 06. 2010) 4.2. Is there really a mirative ja? Supposing pragmatic bias to interfere with an internal cognitive process can account for the supposed ‘‘other’’ meaning of ja, namely its mirative meaning in exclamations such as example (8). (8) Das ist ja eine Unverscha¨mtheit! ‘‘This is ja an impertinence!’’ Mirativity can be induced by the very illocutionary type of the exclamation. As was already sketched before, mirativity is not to be considered as a category in itself, but as an over-specification of FCA in first-person time. For this reason, exclamative sentences are far from exhausting the field of mirativity in German. The example of doch has shown that it was present in declarative sentences, and the GMP denn probably has to be interpreted as a mirative operator over the illocutionary type of questions, given that it indicates that the speaker did not expect to find herself ignorant (see Abraham 1991b: 210¤ ).9 The mirative sentences including ja manifest a switch in the speaker’s overall Theory of Mind. In this case, the speaker suddenly switched to a propositional attitude which she 9. Denn is the basic German interrogative particle. Therefore, in terms of Theory of Mind, it might be assumed to encode a supposed distribution of beliefs where (i) the speaker has no commitment to a precise answer to the question she asks; (ii) she assumes that the hearer could possibly know more about it. If we regard questions as sets of propositions, this means that the speaker does not know which proposition of the set would be a true belief, but she assumes the hearer could have reduced this set of propositions to one token corre-
The epistemological treatment of information
375
assumes must be shared by the hearer. The peculiarity of the ‘‘mirative ja’’ is due to the fact that there seems to be a second switch to explain, which regards the personal orientation of the GMP. In assertive sentences, ja is rather second-person oriented; the speaker tells the hearer what she assumes to be common knowledge ground (thus, ja is often being glossed as ‘‘the speaker expects the hearer not to contradict her’’). In exclamations on the other hand, ja is first-person oriented: What is revised in the speaker’s FCA is her own state of knowledge, whereas the second-person status of the proposition is not (or at least not obligatorily) a¤ected, as in the following example from Hentschel/Weydt (1994: 283), supposed to be uttered after the speaker has just eaten for the first time something cooked by the hearer, and the hearer has not expressed special concerns about her capacities as a cook. (9) Du bist ja ein guter Koch! ‘‘You’re ja a good cook!’’ This shift of personal focus can be accounted for thanks to the proprieties of the respective illocutionary types of the sentences ja is embedded in. According to Grice’s rule of quality, an assertion is defined by the speaker’s commitment to a proposition, which in its turn is supposed not to be evident for the hearer because of the rule of quantity. The information is therefore new for the hearer. On the contrary, exclamations are defined by the novelty of their propositional content for the speaker. This means that the switch of the first-person status of a proposition10 in exclamative sensponding to her belief, which the speaker could subsequently also commit herself to. Nevertheless, the use of GMPs being always optional, the question of what denn brings to such an FCA-configuration can be raised, since it could actually fit for any interrogative clause. Commenting on the sentence Wo ist denn meine Brille? Abraham (1991: 4) suggests that ‘‘in using the original conjunction denn in the middle field, the speaker indicates that she expected to find her glasses in a particular place, but didn’t.’’ Denn can be compared to doch as a marker encoding not only an online evaluation of belief distribution, but also a switch of this evaluation by the speaker in the recent time. This also means that denn is primarily first-person oriented, since it encodes more epistemological details on the speaker’s side than on the hearer’s one. Meanwhile, questions are principally perlocutive speech acts, and in this sense they are second-person oriented. Denn therefore seems to modify the balance between the speech act participants and to re-evaluate the weight of first-person states of mind. 10. This switch is also a switch of the third-person status, because the latter is just a composition of the first- and the second-person ones.
376
Pierre-Yves Modicom
tences is actually not encoded by ja, but is already present in the illocutive type. As for ja, it seems that it merely indicates (i) the fact that there is hic et nunc a convergence of all three personal stances and (ii) a peculiar attention to the recipient of third-person knowledge (which is often the addressee, but in an exclamative sentence, it is the speaker). As we have seen, given the rule of quantity, ja mainly occurs when the propositional attitude is what is being focused upon. It is therefore highly compatible with an exclamative sentence, whose purpose is precisely to encode a switch in the speaker’s state of knowledge. Such an explanation would also imply that the only contribution of ja is to anchor more explicitly the switch to the relationship between the first and the second person.11 This might explain why this alleged mirative ja is the most general and unspecified particle occurring in exclamative sentences (see Hentschel & Weydt 1994: 283 for a more precise account). The mirative ja does not exist per se, but it is a syncretism of several FCA-determinations: The illocutive type encodes first-person epistemic focus and FCA-switch, and ja just carries what can be identified as its core value, namely joint a‰rmativity. It is felicitous i¤ the third-person value of the propositional content is assumed to be recognized as positive by both speech act participants. 4.3. On the evidential specialization of wohl Wohl, originally the adverb for ‘‘well’’, seems to be a Janus-faced particle as well, since it has basically two meanings according to whether it is accentuated or not. If it is accentuated, it highlights the speaker’s strong commitment to the propositional content, which is added to the common ground. If it is not, it has an almost opposite meaning of weak commitment, making the felicity of the utterance compatible with states of the real world not corresponding to what the proposition describes. The first value is easily compatible with the core meaning of the lexeme, whereas the second is not (see Abraham 1991b for further analysis). At first sight, the accentuated wohl seems to be redundant with ja, yet, interestingly enough, wohl is not oriented to the information recipient as ja, but to the ‘‘epistemic reference point’’, that is to the knowledge holder (see Zimmermann 2008). This peculiarity is shared by both the accentuated and the unaccentuated wohl. The interpretation to be defended here is that the 11. Intersubjectivity being the ground of third-person evaluation, this implication of the second person feeds the third-person commitment, so that ja is often glossed as reinforcing the assertion (e.g. in the DWDS dictionary).
The epistemological treatment of information
377
latter, unaccentuated wohl is structurally a bi-a‰rmative particle being pragmatically biased. This occurs along other lines than in the case of ja because of the di¤erent point of focus we already mentioned. Evans (2007) has developed a conceptual distinction between double perspective and joint perspective. Joint perspective is the speaker’s perspective insofar as it is supposed to converge with the hearer’s, whereas double perspective is the presence of two assumed epistemic standpoints. In the framework of this paper, FCA resorts to double perspective, and joint perspective is but a special case of it, that of congruence of propositional attitude, as illustrated by ja. How did wohl receive another specialization, and how can this be related to the question of its orientation? If we consider that GMPs are functional operators which take the propositional content and the speech act participants as arguments (in the sense of Dik 1989) and if we take into account the fact that functions in Dik’s sense might select their arguments according to the kind of predicate they represent, then the following hypothesis can be raised: Wohl operates as a bi-a‰rmative assertion marker over propositions whose truth-functional evaluation has been suspended. This interpretation follows from a reconstruction of wohl’s epistemological function and felicity conditions from the core meaning ‘‘well’’. A modal reinterpretation of the adverb as scoping over the whole proposition would logically be expected to be a validation of the propositional content by the knowledge holder or epistemic point of reference (prototypically, by the speaker in an assertion). Being assigned an FCA function, wohl becomes the origo-oriented equivalent of ja, thus being exposed to the bias of Grice’s maxim of quantity, as well. Given wohl’s weaker reference to the second person, the usage conditions would nevertheless collide with a second rule, namely the rule of quality. Since the knowledge holder is per default supposed to strongly commit herself to whatever she says, origo-oriented strong commitment markers are irrelevant in this second respect, too. This problematic orientation to the speaker plays a crucial role in the functional value of wohl. Given Grice’s maxims, this core meaning of origo-oriented strong commitment with expected acceptance from the recipient is relevant in only one case: When it is focused upon, e.g. prosodically. This is the origin of the specific meaning of the accentuated wohl. When the GMP is not accentuated, its meaning tends to be reinterpreted so as to be conform to the conversational rules: If wohl is used by the speaker to validate the propositional content, this should mean that for her, this commitment was not evident so far, suggesting that the speaker has no direct evidence for the correspondence of the proposition with reality and that the acceptance follows from reflection: Wohl becomes
378
Pierre-Yves Modicom
an inferential marker. This can be observed cross-linguistically: Strong commitment markers can often be reinterpreted as inferentials. Thus, in French, sans doute (‘‘without doubt’’) actually means ‘‘probably’’ or suˆrement (‘‘certainly’’), ‘‘I suppose.’’ In German, gewiss (‘‘for sure’’/‘‘certainly’’) has also the meaning of an inferential standing for propositions accepted after reflection (see WDS, Vol. 6, 6206). Wohl’s orientation to the origo seems to lead one to reinterpret its modal use as the subsequent acceptance of a proposition for the validity of which one has no direct evidence. Thus, wohl would add propositions to the common ground cum grano salis, and under the restriction that their positive truth value was not granted at first sight, but has been set as a postulate after reflection. This turns the unaccentuated wohl into an origo-oriented commitment operator upon unverified propositions, thus doomed to encode weak commitment. (10) Das mag wohl sein. ‘‘It may be so.’’ (literally: ‘‘That might wohl be’’) First empirical evidence for this interpretation would be the fact that wohl scopes over question formation (see Zimmermann 2008: 205¤.). In Dik’s words, question formation occurs one level lower than the determination of the illocutive type, where GMPs are located.12 Questions being sets of propositions among which the speaker does not designate the true one, they are formed at the level where propositions receive their truth value. In this case, the evaluation ends with an e powh´. Further arguments would be found in the list of wohl’s usages in the Wo¨rterbuch der Deutschen Sprache. Wohl appears to occur preferably in a future, potential, or conditional syntactic context. Other contexts of appearance are sentences containing ethical datives, whose modal function in German is often to help the speaker distance herself from the propositional content of the utterance, or together with epistemic modal verbs with a strong evidential meaning, which, as Guentche´va & Landaburu report, often correspond to a covert pattern of non-commitment. There is therefore a strong suspicion that wohl operates over propositions whose truth value has already been suspended so that it does not directly contribute to weakening them. It could rather be used to legitimate what is a mere hypothesis, and to present the assumption as a default common knowledge ground for lack of more reliable information. 12. Respectively the fourth level for truth-functional evaluation and the fifth for the speech act. The latter roughly corresponds to the ‘‘Left Periphery’’ of the Logical Form in recent Generative Grammar.
The epistemological treatment of information
379
This means that wohl synthesizes all three personal stances of FCA. The knowledge holder and the knowledge recipient (corresponding to the first and the second person irrespectively of which of both epistemic roles they play) are di¤erently treated. The knowledge holder is at the center of the described epistemological treatment of information. The proposition is marked as something she assumes to be true without empirical evidence but as an inference and for the purpose of the conversation. The knowledge recipient is clearly in the background, being just asked to accept it for lack of better evidence. The third-person status of the proposition is the other crucial axis, together with the first-person one. Wohl selects its propositional argument as an element of unverified knowledge and sets it as a postulate: This is typical of the kind of epistemological trackers defined by Guentche´va & Landaburu as the second category, that of evidentiality or source-marking. More precisely, it is covertly cognate to inferential markers, designating what is neither empirically proved nor mere hearsay but presented conjectures of the mind. This category is shown by Landaburu to be subordinate to commitment marking, and as we have seen, GMPs operate over propositions whose status in this respect is already determined. But unlike most of them, which are drastically under-determined and indi¤erent as regards source-tracking, wohl seems to select only propositions having this inferential status.
5. Conclusion The study of this sample of German Modal Particles following Guentche´va & Landaburu’s typological data has delivered two results: At first sight, the distribution of features in German might be so di¤erent and pervasive that no isomorphism could be found. But actually, the categories coined to account for an epistemological treatment of information in the languages of the world are relevant for German as well. Schon, doch, denn, ja and wohl can all be integrated into a system of knowledge distribution between the first, the second, and the third person. Yet, the pairs constituted by schon and doch, on the one hand, and ja and wohl, on the other hand, were composed of two particles filling the same blank in the classification along the lines of Theory of Mind. The components of each pair were distinguished by criteria such as [emirative] for doch and schon or [eevidentially restricted] for wohl and ja. The mirative was defined as a specification of FCA in time and as a marker of realignment. Evidentiality cannot be subsumed under the description of belief distribution as
380
Pierre-Yves Modicom
mirativity can, but the case of wohl showed that this second branch of evidential tracking was directly linked to the first one insofar as a systematic pragmatic bias interacting with the core semantic value of the GMP caused it to operate exclusively over a precise evidential class. Thus, evidentiality appears to belong to the epistemological tracking of the first-person stance. It is exhibited together with the encoding of the assumed propositional attitude of the speech act participants regarding the utterance. It has also been shown that the di¤erent uses of each GMP can be reduced to one core meaning directly derived from the original lexeme and related to FCA when analyzed at a su‰cient level of abstraction. Their apparent diversity can be explained compositionally by their coconstruction with other FCA strategies and by a systematic implementation of the conversational rules at the point of the utterance. Acknowledgement The volume editors and one anonymous reviewer have helped immensely to improve earlier versions of this paper. Glosses Com Dir.exp Evid Ingr
Commitment Direct experience evidential Ingressive aspect
Pres Pot nPot Subject.trace
Present tense Potentiality Non-potential Anaphorical class marker for the subject
Sources and literature Primary sources Deutschlandfunk (DLF) Deutschlandradio Kultur (DRK) Grimm, J. & W. et al. 1852–1971 Wo¨rterbuch der Deutschen Sprache. Leipzig/Berlin/Go¨ttingen.
Secondary sources Abraham, Werner 1991a ‘‘Introduction to Discourse particles. Descriptive and theoretical implications on logical and pragmatic particles in German.’’ In: Abraham, W. (Ed.), Discourse particles. Amsterdam: J. Benjamins. 1–10.
The epistemological treatment of information
381
Abraham, Werner 1991b ‘‘Discourse particles in German: How does their illocutive force come about?’’ In: Abraham, W. (Ed.): Discourse particles. Amsterdam: John Benjamins Publishing. 203–252. Abraham, Werner 2010 ‘‘Diskurspartikeln zwischen Modalita¨t, Modus und Fremdbewusstseinsabgleich (Theory of Mind)’’, in Harden, Th. & Hentschel, E. (Ed.) 2010. 40 Jahre Partikelforschung. Tu¨bingen: Stau¤enburg. 33–70. Chung, Sandra & Timberlake, Alan 1985 ‘‘Tense, aspect and mood’’, In: Grammatical categories and the lexicon, Shopen, T. (ed.), Cambridge: Cambridge University Press, p. 202–258. Coniglio, Marco 2007 German modal particles in root and embedded clauses. Venice: University of Venice Working Papers in Linguistics Vol. 17. Davidson, David 2001a Subjective, intersubjective, objective. Oxford: Clarendon. Davidson, David 2001b ‘‘First-person authority’’ (first publ. 1984), in Davidson, David. 2001a: 3–14. Davidson, David 2001c ‘‘The second person’’ (first publ. 1992), in Davidson, David. 2001a: 107–122. Davidson, David 2001d ‘‘Three varieties of knowledge’’ (first publ. 1991), in Davidson, David. 2001a: 205–220. De Lancey, Scott 2001 ‘‘The mirative and evidentiality’’, in Journal of Pragmatics 33/3. 369–382. Dik, Simon C. 1989 The theory of functional grammar. Part I: The structure of the clause. Dordrecht: Foris. Evans, Nicholas 2007 ‘‘View with a view: towards a typology of multiple perspective’’. Berkeley Linguistics Society, 93–120. Evans, Nicholas 2009 The grammar of expectation, talk at the Ludwig-MaximiliansUniversita¨t Mu¨nchen, Dec. 2009. Grice, H. Paul 1975 ‘‘Logic and conversation’’, in Cole, P., and Morgan, J. L. (Eds.) Syntax and semantics: speech acts. Vol. 3. New York: Academic. 41–58. Guentche´va, Zlatka & Landaburu, Jon 1997 Introduction, in Guentche´va, Z. & Landaburu, J. (Eds.): l’e´nonciation me´diatise´e II. Le traitement e´piste´mologique de l’informa-
382
Pierre-Yves Modicom
tion. Illustrations ame´rindiennes et caucasiennes. Louvain, Paris: Peeters. 1–19. Hentschel, Elke & Weydt, Harald 1994 Handbuch der deutschen Grammatik. Berlin/New York: de Gruyter. Jacobs, Joachim 1991 ‘‘On the semantics of modal particles. Illocutionary semantics.’’ in Abraham, W. (Ed.): Discourse particles. Amsterdam: John Benjamins Publishing. 141–162. Landaburu, Jon 2007 ‘‘La modalisation du savoir en langue andoke (Amazonie Colombienne)’’, in Guentche´va, Z. & Landaburu, J. (Eds.). 23–48. Ormelius-Sandblom, Elisabet 1997 ‘‘The modal particle schon; its syntax, semantics and pragmatics’’, in Swan, T. and Westvik, O. J. (Eds.): Modality in germanic languages, Berlin-The Hague: Mouton de Gruyter (Trends in Linguistics – studies and monographs 99). Pe´rennec, Marcel ¨ berschreitung eines Grenzwertes: U ¨ berlegungen 2002 ‘‘Unter- und U zu schon und noch’’, in: Sur le texte. E´nonciation et mots du discours en allemand. Lyon: Presses Universitaires de Lyon. 185–204. Thurmair, Maria 1989 Modalpartikeln und ihre Kombinationen. Tu¨bingen: Niemeyer. Weydt, Harald 1969 Abto¨nungspartikel. Die deutschen Modalwo¨rter und ihre franzo¨sischen Entsprechungen. Bad Homburg v.d.H., Berlin, Zu¨rich: Gehlen (Linguistica et litteraria; 4). Zimmermann, Malte 2008. ‘‘Discourse particles in the left periphery’’. In: Shaer, B., Cook, Ph., Frey, W. & Maienborn, Cl. (Eds.) Dislocated elements in discourse. Syntactic, semantic, and pragmatic perspectives. London: Routledge 200–231. Zimmermann, Malte 2011 ‘‘Discourse particles’’. In: Portner, P., Maienborn, C. & von Heusinger, K. (Eds.), Handbook of semantics. Berlin, 2011–2038. Mouton de Gruyter: Handbu¨cher zur Sprach- und Kommunikationswissenschaft.
Construction-dependent person hierarchies1 Kees de Schepper & Helen de Hoop Radboud Universiteit Nijmegen
Abstract In sentences, the two entities that enjoy a special status are the speaker of the utterance containing the sentence, and the hearer of this utterance. One can refer overtly to the speaker and the hearer in a sentence – pronouns like I and you can serve as the subject of a sentence, for example. But not only subjects can make reference to the speaker and the hearer; both of them also play an important role in the semantics of sentence mood (declaratives, interrogatives, imperatives) and evidentiality (direct evidence, hearsay, inference). In this paper we investigate the e¤ects of having first person and second person subjects in combination with di¤erent types of sentence mood and evidentiality.
1. First examples An analysis with a prominent place for the speaker has a lot of explanatory power for the study of deixis (see Abraham’s Preface, this volume). But also for the study of sentence mood (cf. Tanaka 2008) and evidentiality does such an analysis look promising. The reason is that the two phenomena are rather closely related to the utterance level in speech. At the level of the utterance it is important who speaker and hearer are, because with each new utterance the identity of speaker and hearer shifts. This is why the grammatical category of person plays an important role in both sentence mood and evidentiality. 1. We would like to thank the audience of the SLE conference in Lisbon, September 2009, the audience of the UBC colloquium in Vancouver, January 2010, an anonymous reviewer, and the editors of this volume for their valuable comments. Also, we gratefully acknowledge the financial support of the Netherlands Organisation for Scientific Research (NWO) for our Local Pronouns project.
384
Kees de Schepper & Helen de Hoop
As an example, consider the following sentence from Shipibo-Konibo (Valenzuela 2003: 50)2. (1) jakon baken-ti waste r-iki ainbo bi-ti good give.birth-inf herb dir.ev-cop completely true ‘The herb for easy births is really e¤ective’ In this example there is an evidentiality marker r- coined the direct evidence marker, which means that the speaker has direct evidence for the proposition uttered. Therefore, the information source of the sentence is first person. The subject in (1), jakon baken-te waste ‘herb for easy births’ is neither first nor second person. Next, consider the following example from Shipibo-Konibo in (2). (2) e-a r-iki Bawanixo-nko-ni-a 1sg-abs dir.ev-cop Bawanixo-loc-lig-abl ‘I am from Bawanixo’ Here the information source is the speaker and the subject is first person. In such cases interaction e¤ects may arise – as speakers refer to themselves with the first person – and in this paper we will look into these person e¤ects. We will visualize these e¤ects in the form of person hierarchies. The concept of hierarchies will be explained in section 2. In sections 3 to 5 several new person hierarchies will be posited. This leads to the question 2. Abbreviations used in this paper: ½ 1, 2, 3 abl abs acc caus conj cop cnjv csm decl dir.ev disj firsth imp infer inf
inclusive first, second, third person ablative absolutive accusative causative conjunct marker copula conjunctive change of state marker declarative direct evidential disjunct marker firsthand evidential imperative inferential evidential infinitive
ipfv lig loc n neg nonfirsth or part perf poss.refl prs pst q sg top vis
imperfective ligature locative neuter negative nonfirsthand evidential orientation marker particle perfective possessive reflexive prospective aspect past question singular topic visual evidential
Construction-dependent person hierarchies
385
how many person hierarchies there really are, which will be addressed in section 6. We will end this paper with a conclusion in section 7. 2. Person Hierarchies One way to represent the e¤ects of di¤erent parameters in a certain construction is by using a hierarchy. The most well-known person hierarchy is the Silverstein Hierarchy (Silverstein 1976), as seen in (3). The symbol ½ will represent the inclusive category, as described in section 2.1 below, in this paper. (3) Silverstein Hierarchy ½ < 1,2 < 3 The hierarchy consists of four grammatical categories, inclusive, first person, second person and third person (third person can be split up into several categories – human, animal, inanimate, for example – but these are of no importance in this paper). First, in section 2.1 we will address the question why the inclusive serves as a grammatical category in its own right in the hierarchy in (3), and in all other hierarchies in this paper. Then, in section 2.2 we will provide a general introduction on hierarchies. Finally, in section 2.3 we will look at the Silverstein hierarchy in greater detail. 2.1. The inclusive Silverstein included the inclusive category in his hierarchy. An example of an inclusive pronoun, which is referred to as a first person plural inclusive pronoun, is given in (4). In this Evenki sentence the subject comprises both the speaker and the hearer (Nedjalkov 1997). In (5) another pronoun – referred to as the first person plural exclusive pronoun – is used, and here the subject comprises the speaker but not the hearer. (4) Esi mit oron-mi e-get sokor-ro now ½ reindeer-poss.refl.sg neg-½.imp lose-part ‘Let us (inclusive) not lose our (inclusive) reindeer’ (5) Bu oro-r-vor etejet-chere-v 1pl reindeer-pl-poss.refl.pl guard-prs-1pl ‘We (exclusive) guarded our (exclusive) reindeer’ Daniel (2005) argues against the traditional view that the inclusive is a special instance of the first person plural. His argument is that there are
386
Kees de Schepper & Helen de Hoop
very few instances in which the inclusive pronoun is derived solely from a first-person singular pronoun. Therefore, Daniel argues for the inclusive as an additional, fourth person category. In contrast, the pronouns in (5), which are traditionally referred to as first person plural exclusive pronouns, can be seen as true representatives of the first person plural. As a consequence, the term ‘first person plural exclusive’ could be abolished in favor of the term ‘first person plural’. This leaves us with a grammatical category of person with four values: inclusive, first person, second person, and third person. No further values are needed to describe the person systems in the languages of the world (cf. Cysouw 2003). 2.2. How hierarchies manifest themselves The hierarchy in (3) is an example of what has been called a markedness hierarchy. Haspelmath (2006), building on Greenberg’s (1966) work, argues that it is frequency that drives such hierarchies. To the left in a hierarchy is the most frequent category, and to the right the least frequent category. He states four ways in which a hierarchy can manifest itself: structural coding, facultative expression, inflectional di¤erentiation, and text frequency. Structural coding means that the marking of a less frequent category cannot be shorter than the marking of a more frequent category. Ergative case marking in combination with the Silverstein Hierarchy in (3) is a good example of structural coding (to be more precise, ergative marking for reasons of distinguishability; see de Hoop & Malchukov (2008) and the papers in de Hoop & de Swart (2008) for an in-depth discussion). It is very common for the inclusive to function as a topic, and because topics are often agents, the inclusive is often an agent. Because of this, the inclusive does not require any additional case marking to signal the agent function. For first and second person it is less common to function as a topic/ agent, and for third person even less so. The prediction then is that if a language has an overt ergativity marker for one of these categories, it should also have an overt ergativity marker for all categories to the right in the hierarchy. Crucially, it cannot have zero ergativity marking in the categories to the right, because zero marking is shorter than overt marking. A classical example is Dyirbal (Dixon 1979); in this language third person nouns and pronouns carry an ergativity marker, but the other pronouns have zero marking. Thus, no category to the right of third person has zero marking, since there are no such categories. Therefore Dyirbal obeys the above prediction, as can be verified in (6).
Construction-dependent person hierarchies
387
(6) Ergativity marking in Dyirbal Person category
½
1
2
3
Ergativity marking
-Ø
-Ø
-Ø
-Ðgu
The second way in which a hierarchy manifests itself, facultative expression, means that the marker of a less frequent category may only be dropped if the marker of a more frequent category may also be dropped. The relation with frequency is only indirect: More frequent categories tend to be more predictable, and predictable categories do not need to be expressed. A good example is pro-drop: Pronouns are very frequent and highly predictable, which is why they can be dropped in some languages, but not noun phrases with a common noun. Inflectional di¤erentiation stands for the phenomenon that a less frequent category may not have a richer paradigm of forms than a more frequent category. The idea behind this is that only for frequent categories individual forms may be stored separately. Therefore, only frequent categories have a rich irregular paradigm. An example is the verb to be, which has a rich, irregular paradigm in many languages. Text frequency, finally, is a very obvious manifestation of frequency. Categories that are uttered more often in speech will also appear more often in texts. 2.3. Beyond Silverstein As mentioned earlier, the Silverstein Hierarchy is the most well-known person hierarchy. Silverstein himself noted that the inclusive is more unmarked than first or second person. This is important as we will predict in section 6 that ½, 1, 2 < 3 does not exist as a hierarchy. Thus the inclusive should always behave di¤erently from first and second person in the Silverstein Hierarchy. But does this hierachy su‰ce for all attested person e¤ects? Silverstein did not only look at ergative case-marking, but also at accusative casemarking. Interestingly, accusative case-marking follows the reversed pattern of ergative case-marking: A third person is the least likely to receive accusative case-marking, then second person and first person. The inclusive is the most likely person to be accusative case-marked, since it is very uncommon for the speaker and the hearer to act as the patient in some predicate together. This leads to a person hierarchy one could call the Reversed Silverstein Hierarchy.
388
Kees de Schepper & Helen de Hoop
(7) Reversed Silverstein Hierarchy 3 < 1,2 < ½ Still, not all person e¤ects follow the Silverstein Hierarchy or the Reversed Silverstein Hierarchy. In the following sections we will look at other person hierarchies present in language. 3. Sentence mood The three main sentence moods in language are the imperative, declarative, and interrogative mood. In this section we will look at the person e¤ects for each of them. 3.1. Imperatives A cross-linguistic investigation on imperatives and hortatives has been done by Van der Auwera, Dobrushina and Goussev (2004). An example of a prototypical imperative is given in (8a) and an example of a hortative is provided in (8b) (8) a. b.
Stop hurting me! Let’s stop hurting each other!
Some researchers see hortatives as non-second person imperatives, and indeed, imperatives and hortatives are very similar semantically. The meaning of both may be formulated as: speaker expresses to hearer that person x should do y, whereby x is the subject of the imperative/hortative, and y is the relevant predicate. Furthermore, in a number of languages the syntactic marking of hortatives and imperatives constitutes one paradigm, see Van der Auwera, Dobrushina and Goussev (2004) for references. One example of the connection between hortatives and imperatives is Dutch. In Dutch the hortative construction with the verb laten ‘to let’ is possible for third persons (9a), the inclusive (9b) and first persons (9c), but not for second persons (9d) where there is a special imperative construction (10). (9) a.
Laten ze het zelf maar op-lossen! let.pl 3pl 3sg.n self part part-release ‘They should solve it themselves!’
b.
Laten we het zelf maar op-lossen! let.pl 1pl 3sg.n self part part-release ‘Let’s solve it ourselves!’
Construction-dependent person hierarchies
389
c. Laat ik het zelf maar op-lossen! let.sg 1sg 3sg.n self part part-release ‘I should solve it myself !’ d.
*Laat je het zelf maar op-lossen! let.sg 2sg 3sg.n self part part-release
(10) Los het zelf maar op! release 3sg.n self part part ‘Solve it yourself!’ Assuming that imperatives and hortatives are part of one and the same concept, Van der Auwera, Dobrushina and Goussev constructed a person hierarchy for the structural coding of the imperative. Ignoring plurality, duality and the like, the hierarchy has the form shown in (11). (11) Imperative Hierarchy 2 < ½,3 < 1 As second person is the most common person to be the (semantic) subject of a commanded predicate, second person features as the left-most category in the hierarchy. Indeed, second person is in many languages the only category that receives an unmarked imperative-hortative marking; in those an imperative-hortative meaning for the other persons is only allowed when using a marked, periphrastic construction. For the inclusive and third person sometimes a special imperative-hortative marking is available, but usually this marking is longer than the marking for second person, compare the Dutch marking in (9a–c) and (10). A dedicated imperative-hortative marking for first person is very uncommon. 3.2. Declaratives and interrogatives Now that we have an Imperative Hierarchy, we may expect to find person hierarchies for other sentence moods. Western Apache (also known as Athabaskan) shows how person matters for the declarative mood (De Reuse 2003: 93). In this language it is inappropriate to have declaratives with a second person subject (12a). Instead, an inference evidentiality marker – see the following section – should be added to the sentence, as in (12b). Presumably, the same goes for an inclusive subject. (12) a.
*?Nił gozk’az with.2 3sg.n.perf.be.cold
390
Kees de Schepper & Helen de Hoop
b.
Nił gozk’az la˛a˛ with.2 3sg.n.perf.be.cold infer ‘You are cold’
The reason for the inappropriateness of (12a) is that it is impolite for a speaker to assert something about the hearer. Therefore it is unsure whether (12a) is fully ungrammatical, or simply impolite. Nevertheless, we see the same person e¤ects in English: It is very uncommon to have a declarative with you as subject, or with we as subject in its inclusive meaning. This suggests a tentative Declarative hierarchy as in (13). (13) Declarative Hierarchy 1,3 < ½,2 Parallel to the Declarative Hierarchy, an Interrogative Hierarchy may be observed. In English questions with a first person or the inclusive as subject are rather uncommon. This suggests that there is a hierarchy as in (14), with the second and third person as the left-most categories. (14) Interrogative Hierarchy 2,3 < ½,1 3.4. Suggestions Besides imperatives, declaratives and interrogatives there are other sentence moods to be discerned that have less obvious markings; commissives (I promise you I’ll come), expressives (What a day!), and declarations (I hereby resign as president) are examples. Another instance of sentence mood is suggestions. In Georgian (Harris 1984) there is a special construction for expressing suggestions, in which the subject of the embedded clause may only be second person, as in (15), or inclusive. (15) rogora xar imaze rom 'vino momit ano? ˙ how 2.are 3sg.n.on that wine 2.bring.1sg.3sg.n.cnjv ‘How about bringing me some wine!?’ Similar e¤ects may be observed in English, where the auxiliary might can be used to express a suggestion to a second-person subject or an inclusive subject, but not to a first-person or a third-person subject, see (16) from Foolen and de Hoop (2009). (16) a. b. c. d.
You might try to put the key into this slot We might try to put the key into this slot I might try to put the key into this slot He might try to put the key into this slot
Construction-dependent person hierarchies
391
From this we can conclude that second person and the inclusive are more common as subjects of a suggestion than first and third person. The subsequent Suggestion Hierarchy can be found in (17). (17) Suggestion Hierarchy ½,2 < 1,3 4. Evidentiality Evidentiality is the grammatical category that encodes the information source of the sentence. Aikhenvald (2004: 65) distinguishes six semantic types of evidentials, as seen in (18). (18) Type
Source of information
I.
Visual
sight
II.
Sensory
hearing (often extended to smelling, tasting and feeling)
III.
Inference
inference based on visible evidence
IV.
Assumption
assumption based on general knowledge or logic
V.
Hearsay
third-party, non-specific (John went away, they say)
VI.
Quotative
third-party, specific (John went away, he said )
Person is an important category for evidentiality. The information on which any sentence is based always originates from some individual, and because evidentiality makes reference to this information, attention is drawn to the individual behind it. For evidentiality types I through IV the source of information is typically the speaker herself. For evidentiality types V and VI the source is typically not the speaker. Besides person preferences for source, there are also person preferences for subject in evidentiality marking (evidentials for short). There is an important distinction between internal state predicates (e.g. I’m hungry) and external state predicates (e.g. John went to the market). External state predicates will be discussed first, as most predicates are of this type. 4.1. External state predicates External state predicates are predicates that can be verified in an objective way. When such a predicate has an evidential, the subject’s predicate is
392
Kees de Schepper & Helen de Hoop
preferably not a first person. In other words, it is uncommon to have a first person subject in a sentence with an evidential. The reason for this is simple: It is unnecessary for a speaker to dwell on the evidence for something she did herself. These observations also apply to the inclusive – Aikhenvald (2004: 217–218) found ‘no significant di¤erences’ between the first person and the inclusive. The resulting hierarchy is the same as the Interrogative Hierarchy in (14). (19) External State Hierarchy (¼Interrogative Hierarchy) 2,3 < ½,1 The hierarchy correctly predicts a facultative expression e¤ect: There are languages where null subjects may only occur in evidentials with a second or third person subject (Aikhenvald 2004: 236). Yet, there is a potential problem: People very frequently talk about their own actions, but it is not desirable to mark a sentence containing a first person subject with an evidential. There are languages where evidentials do not occur with the first person at all, but in many languages that have evidentials, it is obligatory to have them in a sentence. Which evidentiality marking is used in such cases? It turns out that languages consider Visual evidentiality the least marked option. In Jarawara if a person got drunk the other night, he uses a marking called the firsthand evidential when he tells about it, as in (20). The firsthand evidential is used to mark Visual and Sensory evidentiality (Dixon 2003: 170). (20) o-hano-hara oke 1sg-be.drunk-pst.firsth 1sg ‘I got drunk’ If a di¤erent evidential is used the sentence gains an additional meaning. In Jarawara, using the non-firsthand evidential in a similar sentence may signal that the speaker does not remember getting drunk, see (21). This non-firsthand evidential is used for marking Inference, Assumption, Hearsay and Quotative evidentiality in Jarawara. (21) o-hano-hani oke 1sg-be.drunk-pst.nonfirsth 1sg ‘Apparently, I got drunk’ Across languages it can be seen that the evidential used for Visual evidentiality is used for neutral cases like (20), and that other evidentials may
Construction-dependent person hierarchies
393
induce e¤ects like the one in (21). Aikhenvald calls these meaning extensions first person e¤ects. Other meaning extensions observed as first person e¤ects are: new information, surprise, denial and the marking of unintentional, uncontrolled, non-volitional actions. As predicted, in almost all of the cases these first person e¤ects pertain to evidentials that do not mark Visual evidentiality. Only when evidentiality marking is optional, first person e¤ects may arise from the use of Visual evidentiality. In Qiang, for example, under normal circumstances a predicate with a first person subject will be expressed without an evidential (LaPolla 2003: 63), see (22). (22) qa ts tu-wsu-z˛-ja 1sg water or-boil-caus-csm.1sg ‘I brought the water to a boil’ If a speaker uses the visual evidential, an additional meaning is added to the sentence. The sentence in (23), for example, has the additional meaning that the action was unintended. (23) qa the: ta de-we-z˛-u-a 1sg hit him or-exist-caus-vis-1sg ‘I hit him (by accident)’ The above data shows which type of evidentiality is preferred in the context of a first person subject. For normal, intentional actions a first person subject prefers to have no evidentiality marking at all, and if that is not possible in a language, Visual evidentiality marking is preferred over any other type of evidentiality marking. The two preferences can be seen in (24). (24) Preference for evidentiality marking for first-person subjects a. No evidential over overt evidential b. Evidential for Visual evidentiality over other evidentials Alternatively, one could say that the External State Hierarchy in (19) applies to both Visual and other types of evidentiality, but that it is less strong for Visual evidentiality. 4.2. Internal state predicates Internal state predicates describe processes internal to a person, for example emotions, desires and pain. For such predicates the speaker is the preferred subject of an evidential. The marker for Sensory evidentiality is the
394
Kees de Schepper & Helen de Hoop
appropriate evidential for internal state predicates, as Sensory evidentiality covers feelings. The person e¤ects for the subject of an internal state predicate are reflected in the hierarchy in (25). (25) Internal State Hierarchy ½,1 < 2,3 The hierarchy may not come as a surprise, as a speaker can only feel her own feelings. It is virtually impossible to feel someone else’s feelings and therefore the combination of internal state predicates and Sensory evidentiality does not occur. A language where the workings of the Internal State Hierarchy in (25) can be observed is Tariana (Aikhenvald 2003: 149). In Tariana the feelings of second and third person subjects may not be described with the Sensory evidentiality marking. Visual or Inferential evidentiality marking should be used instead. 5. Mood plus evidentiality: conjunct-disjunct systems In some languages there is an interesting interaction between person and sentence mood called a conjunct-disjunct system. In such a system a first person subject in a declarative sentence is marked with a special conjunct marker, which is also used for interrogative sentences with a second person subject. A textbook example is the system in Awa Pit (Curnow 1997). In this language a declarative sentence with a first person subject has the same marker as interrogative sentences with a second person subject, as can be seen in (26a–c) and (27a–c). (26) a.
b.
na¼na pala ku-mtu-s 1sg¼top plantain eat-ipfv-conj ‘I am eating plantains’ nu¼na pala ku-mtu-y 2sg¼top plantain eat-ipfv-disj ‘You are eating plantains’
c. us¼na atal ayna-mtu-y 3sg¼top chicken cook-ipfv-disj ‘He/she is cooking chicken’
Construction-dependent person hierarchies
(27) a.
b.
395
min¼ta¼ma ashap-tu-y? wh¼acc ¼q annoy-ipfv-disj ‘Whom am I annoying?’ shi¼ma ki-mtu-s? what¼q do-ipfv-conj ‘What are you doing?’
c. min¼ta-s a-mtu-y? where¼loc-abl come-ipfv-disj ‘Where is he coming from?’ An overview of the verbal paradigm – adapted from Aikhenvald (2004: 124) – in languages with a conjunct-disjunct system is given in (28). (28) Verbal paradigm of a conjunct-disjunct system 1
2
3
Declarative
Conjunct
Disjunct
Disjunct
Interrogative
Disjunct
Conjunct
Disjunct
One analysis of conjunct-disjunct systems is that the conjunct marker marks whether the subject of a sentence can also verify the content of the sentence. In declaratives, it is the speaker who knows whether the information is true, and in interrogatives it is up to the hearer to comment on the truthfulness of the proposition. Aikhenvald (2004: 127) states that conjunct-disjunct systems ‘are not evidential in nature’, but evidentiality marking and conjunct-disjunct marking are closely related. Creissels (2008), for example, shows the relatedness between conjunct-disjunct marking and Quotative evidentiality. Another example is that conjunct-disjunct systems can lead to the same first person e¤ects that we saw for evidentiality marking in the previous section: meaning extensions involving new information, surprise, denial or the marking of unintentional, uncontrolled, non-volitional actions. An example of such a first person e¤ect is given in the sentences in (29a–b), which are from the Tsafiki language (Dickinson 2000). (29) a.
kala ta-yo-e money have-conj-decl ‘I have money’
396
Kees de Schepper & Helen de Hoop
b.
kala ta-i-e money have-disj-decl ‘I have money (– what a surprise)!’
The sentence in (29a) is uttered when a speaker wants to state that she has money. Here the sentence is marked with a conjunct marker in accordance with the information in (28). In (29b) the fact that the speaker has money comes as a surprise to her. This meaning extension involving surprise leads to a disjunct marker in this sentence. Assuming that inclusives behave like first persons in a conjunct-disjunct declarative (since inclusives include the speaker – the content verifier), disjunct marking of subjects in a declarative sentence follows the Disjunct Declarative Hierarchy in (30). Second persons and third persons are commonly marked with a disjunct marker. First persons and inclusives are commonly not marked with a disjunct marker, and if they are they express additional meanings like new information, surprise, denial and the marking of unintentional, uncontrolled, non-volitional actions, as exemplified in (29b). This is what Aikhenvald (2004) has called first person e¤ects. (30) Disjunct Declarative Hierarchy (¼Interrogative Hierarchy) 2,3 < ½,1 The Disjunct Declarative Hierarchy is the same as the Interrogative Hierarchy in (14). It may seem counterintuitive that disjunct marking on declaratives follows a hierarchy for interrogatives. However, what the Interrogative Hierarchy reflects is the inherent information the speaker has about a person. The speaker has no inherent knowledge of the actions of second and third persons, and therefore they feature more often in questions than first person. Similarly, second and third person are often marked with a disjunct marker, since a disjunct marker marks that the speaker has no inherent knowledge of their actions. For disjunct marking in interrogatives the assumption for inclusives will be that they behave as second persons. As a consequence, disjunct marking for subjects in interrogatives follows the Disjunct Interrogative Hierarchy in (31), which is the same as the Declarative Hierarchy in (13). As a rule, the hearer knows more about herself than the speaker does, which is why the speaker refrains from making declarative statements
Construction-dependent person hierarchies
397
about the hearer, and why the speaker refrains from disjunct-marking the hearer in questions. (31) Disjunct Interrogative Hierarchy (¼Declarative Hierarchy) 1,3 < ½,2 6. Restricting the hierarchies By now we have seen a number of person hierarchies. One hierarchy that has not been mentioned so far is the Gender Hierarchy, as seen in (32). Cysouw (2003), building on Corbett (1991), notes that gender is marked most commonly on third person forms and sometimes also on second person forms. Gender marking on first person forms is ‘rather uncommon’, and gender marking on inclusives is ‘almost unattested’. (32) Gender Hierarchy 3