This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
TITLE "Perspectives on Semantics, Pragmatics, and Discourse: A festschrift for Ferenc Kiefer"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Perspectives on Semantics, Pragmatics, and Discourse
Pragmatics & Beyond, New Series Editor Andreas H. Jucker, Justus Liebig University, Giessen Editorial Address Justus Liebig University Giessen, English Department Otto-Behaghel-Strasse 10, D-35394 Giessen, Germany e-mail: [email protected] Associate Editors Jacob L. Mey, Odense University Herman Parret, Belgian National Science Foundation, Universities of Louvain and Antwerp Jef Verschueren, Belgian National Science Foundation, University of Antwerp Editorial Board Shoshana Blum-Kulka, Hebrew University of Jerusalem Chris Butler, University College of Ripon and York Jean Caron, Université de Poitiers Robyn Carston, University College London Bruce Fraser, Boston University Thorstein Fretheim, University of Trondheim John Heritage, University of California at Los Angeles Susan Herring, University of Texas at Arlington Masako K. Hiraga, St.Paul’s (Rikkyo) University David Holdcroft, University of Leeds Sachiko Ide, Japan Women’s University Catherine Kerbrat-Orecchioni, University of Lyon 2 Claudia de Lemos, University of Campinas, Brazil Marina Sbisà, University of Trieste Emanuel Schegloff, University of California at Los Angeles Deborah Schiffrin, Georgetown University Paul O. Takahara, Kobe City University of Foreign Studies Sandra Thompson, University of California at Santa Barbara Teun A. Van Dijk, University of Amsterdam Richard Watts, University of Berne
Volume 90 Perspectives on Semantics, Pragmatics, and Discourse: A Festschrift for Ferenc Kiefer Edited by István Kenesei and Robert M. Harnish
Perspectives on Semantics, Pragmatics, and Discourse A Festschrift for Ferenc Kiefer
Edited by
István Kenesei University of Szeged
Robert M. Harnish University of Arizona
Assistant to the Editors Judit Gervain This publication has been supported by a grant from the Hungarian National Research Programs (Grant No. OTKA P33524).
John Benjamins Publishing Company Amsterdam / Philadelphia
8
TM
The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.
Library of Congress Cataloging-in-Publication Data Perspectives on semantics, pragmatics, and discourse : a Festschrift for Ferenc Kiefer / edited by István Kenesei, Robert M. Harnish. p. cm. (Pragmatics & Beyond, New Series, issn 0922-842X ; new ser. 90) Includes bibliographical references and indexes. 1. Semantics. 2. Pragmatics. 3. Discourse analysis. I. Kiefer, Ferenc. II. Kenesei, István. III. Harnish, Robert M. IV. Series. P325.P473 2001 401’.43--dc21 isbn 90 272 5109 6 (Eur.) / 1 58811 053 2 (US) (Hb; alk. paper)
Part III Logical structures and universals in semantics and pragmatics On the lexical typology of modals, quantifiers, and connectives Johan van der Auwera and Bert Bultinck
173
Grelling’s paradox: Its significance for linguistic theory Noel Burton-Roberts
187
Frege on mood and force Robert M. Harnish
203
Leibnizian linguistics Anna Wierzbicka
229
Part IV Dialogue and thematic structure Naming conventions, focalization, and point of view in Balzac’s La Peau de chagrin Kerstin Jonasson
257
Discourse theory and the translation of clefts between English and German Monika Doherty
273
Tectogrammatics in corpus tagging Eva Hajicˇová, Jarmila Panevová and Petr Sgall
293
Capturing differences between social activities in spoken language Jens Allwood
301
An account of innuendo Bruce Fraser
321
Name index
337
Subject index
339
AUTHOR ""
TITLE "Preface"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Preface Ferenc Kiefer’s work in semantics and pragmatics
Ferenc Kiefer started his career as a ‘mathematical linguist’, an occupation largely forgotten today — probably because he had graduated in both mathematics and two foreign languages and literatures (German and French). But he soon found his true vocation in semantics and in an aspect of pragmatics, topicfocus divisions, which later turned out to be of crucial importance in understanding the structural properties of his native Hungarian. His interest in languages and various central problems of semantics determined a course that is marked by a set of articles, introductions and books edited or written in the subsequent years on the semantic classification of adjectives, presupposition and word order, which again include topic and comment. Presuppositions are to be his theme in the following years, leading him to embark on questions (and answers), and then to adjectives and the problems of factivity, and finally to a Doctor of Science degree dissertation for the Hungarian Academy of Sciences on presupposition, subsequently published as a monograph. Modality is another topic between semantics and pragmatics that Kiefer discusses repeatedly. His papers on the semantics of aspect, his treatment of indirect speech acts also show his concern for the borderline between semantics and pragmatics, as is evidenced by articles specifically written on that issue. His work, however, does not stop at the areas of these two rather vast fields; he has written on the syntax, morphology and/or phonology of Hungarian, German, Swedish, French, Danish and Polish, among others, and on the theoretical problems of linguistics at large. His full bibliography runs to 11 single-spaced pages, of which we have selected but the most important ones in the two areas represented in this book and discussed in this Preface. And then we haven’t spoken of Professor Kiefer, the teacher at work from Paris to Moscow and Stockholm to Budapest and Szeged, Director Kiefer of the Research Institute for Linguistics of the Hungarian
viii Preface
Academy of Sciences, and Feri, the man himself, always ready for a good joke, a glass of wine, fine classical music, and a hearty conversation, whether on linguistics or any other subject. May we enjoy his company for another 70 years!
AUTHOR ""
TITLE "Introduction"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Introduction
1.
The structure of the book
1.1 Pragmatics in Grammar Most introductory textbooks take their readers by the hand and lead them through the maze of language by a path that has distinctly marked clearings or stations called Phonology, Morphology, Syntax, Semantics and Pragmatics. This approach gives the impression of levels of grammatical description frozen into separation, with little chance of integrating their achievements. While it may be useful to point at the different natures of various linguistic phenomena, in actual research work such a segregation has long proved untenable. Ever since the early days of Generative Semantics, reference to or even reliance on the pragmatic aspects of language has acquired importance in grammatical analyses. The implementation of the interpretation of, for instance, questions as requests or statements was a direct consequence of the emergence and attraction of speech act theory, and more results of inquiries formerly regarded to be purely pragmatic quickly found their way into formal grammatical theories. One example is the treatment of a subclass of deictic expressions, viz. pronouns. While their extralinguistic reference is still within the domain of pragmatics, their coreferential relations, i.e. the properties that define whether they can have common reference with another expression, are to a great extent accommodated in syntax. But there are more traditional examples that can be quoted for the inclusion of pragmatic information in grammatical descriptions. Mood is after all the expression of the speaker’s relationship to the event or the proposition. Tense represents the relationship between speaker’s time, event time and reference time. A number of adverbial phrases also embody speakers’ attitudes to the events depicted in the rest of the proposition. Theme-rheme divisions grammaticalize the speaker’s view of the information conveyed by the proposition by means of order and/or prosodic devices.
x
Introduction
There is thus no novelty in arguing for the role of pragmatics in the formal subsystems of grammar. But what types of pragmatic information can or must be accounted for in the other modules of grammar is a matter far from being settled. The articles in the section entitled “Pragmatics in Grammar” focus on problems of this nature. In addition to the traditional problems of deixis, new data also highlight the necessity to speculate on how to reconcile pragmatics and syntax. Absentive constructions, that is, grammatical devices serving to express the absence of the subject are found in several languages, and their properties can be accommodated in a unified model (see de Groot’s paper). Speakers’ attitudes are also responsible for more (or less) than properties of full propositions: they arguably must play a role in the meanings of diminutive and augmentative derivational affixation (see Dressler & Merlini Barbaresi’s paper). Hedges are a different device for speakers to weaken their commitment to the truth of the proposition put forward; although their semantic and pragmatic properties are worth studying, their syntactic characteristics are perhaps even more interesting (see van Riemsdijk’s paper). The problems of the grammaticalization of mood and related notions is particularly interesting in such lesser-known or studied languages as Tsez and Bezhta (see Comrie’s paper). 1.2 Semantic compositionality and pragmatics In his pioneering work on the foundations of semantics, Frege proposed that the semantics of a complex expression (sense and reference) is a function of the sense and reference of its constituents, together with their syntactic mode of construction. Much research has been devoted to the exact nature and extent of compositionality in natural language. In some cases the issue is the identity of the constituents involved in the composition, and whether one form is separate from another, or just a shorted version (see Wunderlich’s contribution). One traditional problem area has been the possessive. For instance, John’s book seems to mark any number of different relations between John and his book: he owns it, he wrote it, he is holding it, etc., and this raises the issue of the argument structure of possessive constructions (see Partee & Borschev’s contribution). Another problem area involves such predications as The leaf is green or She’s a good dancer — are these true if the leaf is a Japanese maple painted red or she is a dancer and good? (See Gendler Szabó’s contribution.) The issue of compositionality also has importance for pragmatics, because if compositional mechanisms are defined over syntactic and semantic structures, then pragmatics
Introduction
might be thought to begin where compositional semantics ends. The semantic–pragmatic distinction has been notoriously hard to define. This is partly due to the fact that the word ‘semantics’ comes to us from two traditions: the logical tradition and the linguistic tradition. According to the logical tradition, semantics has to do with reference and truth. According to the linguistic tradition, semantics has to do with meaning encoded in the language. Most researchers agree that pragmatics involves language use and contextual information, and it is easy to see that reference and truth can involve contextual information (e.g. deixis), and that encoded meaning can involve facts of use (e.g. grammatical mood). Again, how to draw the semantics–pragmatics distinction? One strategy has been to expand the conception of semantics to include what is said, while pragmatics is confined to what is implied or implicated. This pushes the problem back to circumscribing what is said (vs. what is implied) and here facts about lexical content, syntactic structure and compositionality can be used (see Bach’s contribution). 1.3 Logical structures and universals in semantics and pragmatics Since at least Aristotle, researchers have tried to relate principles of logic and logical structure (unusually conceived of as universal) to the structure and use of natural language. The use of logic in the study of natural language has taken many guises. One project has been to find logical structure in the syntax of sentences, and work on ‘logical form’ (LF) illustrates this application. Another project has been to apply methods and definitions in logic to linguistic theory. In some cases even paradoxes in logic can be used to get at general issues (see Burton-Roberts’ contribution). Another project is to uncover and systematize relations in the logical vocabulary of languages by finding common patterns of semantic entailment and pragmatic implications (see van der Auwera & Bultinck’s contribution). Still others inquire into the possibility of extending the logical analysis of sentences beyond the traditional fact stating class of assertions, and see what logical relations might obtain between e.g. imperatives and interrogatives (see Harnish’s contribution). Finally, there is the Leibnizian extension of the idea of universal logical structure (going back to Aristotle’s Categories) to an innate universal class of predicates out of which all semantic structures can be constructed, and which can form the basis of translation (see Wierzbicka’s contribution).
xi
xii
Introduction
1.4 Dialogue and thematic structure Since Georg von der Gabelentz in the late 19th century called attention to another dimension in which sentences are articulated, linguists have been captivated by problems of thematic organization, to pick one name for what has been termed as psychological subjects and psychological predicates by Gabelentz and in Germany, functional sentence perspective in the Prague School and in Slavic linguistics at large, theme–rheme relations in Britain among others, and topic–focus division in current grammatical analyses. While there is general agreement regarding the applicability of thematic organization to discourse, it is also universally acknowledged that individual languages differ in the particular mechanisms that implement these distinctions, constituting the grammatical problem of cataloguing and accounting for the means to express them. The question for a pragmatic analysis to address is why and how this apparatus is employed in actual or ideal discourse. Kerstin Jonasson’s article provides one example for the former: it investigates the use of definite and indefinite referential expressions in focusing the reader’s attention in one of Balzac’s novels. Monika Doherty compares the use of clefts, i.e. the most frequent device in English to render focus, in the English translations of German short stories. Hajicˇová et al. provide a marking apparatus for discourse for computer applications. Although the focus of recent linguistics, especially generative grammar, has been on sentences and their constituents, even a quick look at normal language use will reveal that we often speak in nonsentential units, i.e. we use either parts of sentences or we string sentences together to convey our meaning. But things get even more complicated when other people are involved, when we move, as some say, from discourse to conversation and dialogue. Here a speaker’s contribution typically must mesh not only with what has preceded, but also with what is anticipated. The speaker will often formulate their contribution so as to highlight or focus on topics common to the dialogue, and most languages have devices available for this purpose — think of what has gone by the name of ‘stylistic variants’, e.g. active/passive constructions, cleft and pseudo-cleft constructions and so on. The study of discourse, conversations and dialogue covers a wide range of topics from cohesive mechanisms to turn-taking, to openings and closings, to social register. And to make matters even more complicated, not only are these communicative events taking place in a social matrix, but the social matrix is itself embedded in a biological and physical matrix. And because of this it has been very difficult to get a clear picture of
Introduction xiii
how these various dimensions of discourse and dialogue might fit together. Some researchers have chosen to provide a statistical analysis of various types of discourse (see Allwood’s contribution). Others have preferred to select a particular aspect of dialogue, and to dig down deeper and see how it works in detail. For instance, some work on speech acts in general, trying to determine their nature and their categories. Others work on how we implicate things (vs. saying them) — what the mechanisms of implication are, and what their variety can cover. Others are even more specific, choosing particular uses of language for detailed study (see Fraser’s contribution).
2. Overview of the articles Casper de Groot discusses the non-lexical or grammatical expression of absence in a number of languages. It is a special form of deixis, in which the event is displaced from the “deictic center”, i.e. some location specified in the sentence. The crucial properties of the Absentive are as follows. (i) The event has to be remote, (ii) it must be stative, (iii) it cannot be directional, and (iv) there cannot be direct perception of the person involved. These characteristics are then accounted for in a framework of Functional Grammar Wolfgang U. Dressler and Lavinia Merlini Barbaresi contend that pragmatics is a superordinate of semantics. They ground their arguments in the analysis of diminutive and augmentative derivational affixes by claiming that in addition to the semantic features corresponding to ‘small’ and ‘big’, respectively, they contain the pragmatic feature of ‘fictive’, specified as ‘non-serious’. Their evidence is taken from cross-linguistic data, early language-acquisition, and diachrony. Henk van Riemsdijk shows clear cases in which pragmatics ‘intrudes’ into syntax. Examples like (he is) far from innocent or (he bought) what he took to be a guitar must be represented in canonical syntactic tree structures with far constituting the head and the so-called ‘transparent relative clause’ as a headless relative, although the meanings conveyed are thus not captured accurately. To circumvent this difficulty he suggests that in contrast to previous proposals that were based on deletion, duplicate structures be posited on the analogy of botanical grafts. This way the principal meanings (he is) innocent or (he bought) a guitar as well as the hedges they are embedded into can be more accurately analyzed.
xiv Introduction
Bernard Comrie examines apparently identical constructions in two closely related Northeast Caucasian languages and shows that even under such conditions the semantic fine structures may very well differ. Affective verbs in one language can be used in the imperative mood (as in the Bezhta equivalent of ‘Love your enemies!’), but in the other their imperative forms are ungrammatical and alternative strategies must be made use of, such as the causativization of the verb. Dieter Wunderlich pays his tribute to Ferenc Kiefer by confronting the intricacies of comparative constructions in his native language, Hungarian. Since in contrast to other languages, Hungarian has two distinct syntactic devices to express comparison, a case-marked phrasal one and a clausal one, he addresses the question whether there is any semantic difference between them. He concludes that the phrasal comparative evokes no generic reading, whereas the clausal comparative expresses generic quantification with the comparative in its scope. Having more than one option is found to be felicitous because it helps this language escape the constraints imposed by clausal constructions and the syntactic clumsiness of phrasal expressions encountered in other languages. Barbara Partee and Vladimir Borschev’s problem is formulated along similar lines: does the genitive phrase John’s in possessive constructions like John’s team, a team of John’s, and predicate possessives like That team is John’s have unified or distinct analyses? Does the difference of interpretation arise from the way the genitive phrase combines with simple nouns (as team) and relational nouns (as brother)? The case is further exacerbated by the fact that while a number of languages have possessive modifiers along with argument-like possessives, there may be fine semantic differences between their uses in the individual languages. Interesting puzzles of putative elliptical constructions arise in predicate possessives as evidenced by the presence or absence of case and number agreement. Zoltán Gendler Szabó investigates the semantics of certain adjectives and the consequences for the principle of compositionality: the meaning (i.e. the propositional content) of a complex expression is determined only by the meaning of its constituents and the manner in which they are combined. As formulated the principle excludes the possibility that context, both linguistic and nonlinguistic, plays a role in determining meaning (content). In other words, there are no ‘context-dependent’ expressions, expressions that have different contents when uttered in different contexts. After disposing of
Introduction
‘illusory’ examples of content dependence, and noticing that the phenomenon of indexicality is a kind of context dependence, the author formulates the Context Thesis: the content of an expression depends on context only insofar as the contents of its constituents do. Is this correct? Examples such as The leaf is green seem to suggest that in the context of sorting leaves for decorations it could be true of a red leaf painted green, but not in the context of a botany lesson — and perhaps this extends to other kinds of adjectives as well, such as good as in good N, where context seems to fix the function of N that is being evaluated as good. The author finds variables in the logical form of sentences with e.g. green and good which allow the Context Thesis (and hence one form of compositionality) to be preserved. Kent Bach argues that there is a theoretically important notion of ‘what is said’ in uttering a sentence (vs. what is implied or implicated). The idea comes basically from what Bach dubs Grice’s ‘Syntactic Correlation Constraint’, which requires what is said to reflect the elements of the sentence, their order and their syntactic character. This semantic and compositional conception of what is said is contrasted with more inflationary pragmatic notions, such as those popular in, for instance, Relevance Theory. It is then defended against objections based on psychological, epistemological and linguistic evidence for the intrusion of pragmatic factors into what is said. Johan van der Auwera and Bert Bultinck continue the work of Horn and earlier work by the authors on the semantic and pragmatic parallels between modals (necessity, possibility), quantifiers (all, some) and connectives (and, or, not). The starting point of subsequent work on these notions was the original ‘Aristotelian Square of Opposition’ for quantifiers (with its relations of contradictory, contrary, subcontrary and subalternation), which has subsequently been extended by various authors to other notions and in some cases recast as three, five or six sided figures. The present paper explores the similarities between modals, quantifiers and conjunctions at the cognitive level, and differences between them at the lexical — semantic level. These relations are captured in a three-level scalar square, which nevertheless shows that some differences in potential for lexicalization remain. Noel Burton-Roberts hangs, as he puts it, a heavy coat on a small hook. The hook is Grelling’s paradox and the coat is linguistic theory. Grelling’s paradox, you will remember, is often formulated by first defining autological as ‘true of itself ’ and heterological as ‘not true of itself ’, then asking: is heterological autological or heterological? If it is the one, its the other, hence the paradox. After
xv
xvi Introduction
subscribing to Ryle’s dissolution of the paradox, the author pursues the issue deeper into linguistic territory: into issues of use and mention (quotation), and especially of the type — token distinction and its relation to representation in general. Robert M. Harnish proposes seven conditions of adequacy on a theory of sentential (vs. verbal) mood (the conventionalized pairing of sentence structure and speech act potential). He then turns to one neglected aspect of Frege’s philosophy of language, Frege’s remarks on mood and communication for interrogative and imperative sentences. A ‘minimalist’ interpretation of Frege is first offered for these non-indicative sentences. But it is found to have problems and a more systematic ‘extended’ interpretation is proposed. First simple sentences are investigated, then truth functional compounds of various types are examined and compared with embeddings of imperatives and interrogatives. Anna Wierzbicka argues for ‘Leibnizian’ linguistics as distinguished from the more dominant ‘Cartesian’ linguistics, though, as she notes, they are not incompatible. For instance, both Cartesians and Leibnizians see language as “a mirror of the human mind” that will tell us much about human nature. Leibnizian linguistics involves the idea that there is an innate universal alphabet of concepts or “conceptual primes” that can be used to construct the semantic representations of any sentence in any language. It focuses on the study of words (not sentences), and is centered on meaning and translation (not syntax). Leibnizian linguistics involves the trial and error construction of definitions from many languages and the inspection and reduction of the necessary primitives. The author reports the preliminary results of ‘Natural Semantic Metalanguage’ (NSM) project in uncovering about 60 such primitives. Kerstin Jonasson’s interest is centered on focalization, that is, the way in which authors are introduced and referred to in the course of a narrative. Focalization can be external, that is, the reader is given information only of the protagonist’s identity, or internal, in which case the character’s thoughts and feelings are also made known. Jonasson takes the first part of one of Balzac’s novels as her corpus and investigates the French author’s techniques of referring and shifting points of views, a method of changing the narrative perspective or the deictic center. It is by this method, she argues, that Balzac arouses the empathy of the reader with one or another of his characters, or the narrator himself, for that matter.
Introduction xvii
Monika Doherty examines the translations of clefts between English and German. Clefts are a convenient, and sometimes the only, means to express focus, a common device in discourse to highlight or contrast some expression in a sentence. In her view of a hierarchical structure of sentence meaning, elements of one layer can be in the scope of elements of another layer, made up of propositional meanings, illocutionary forces, and information structure. The close analysis of clefts sheds light upon the way in which elements highlighted stand in contrast with the preceding context, which serves the basis of further assumptions concerning the processing mechanisms that such constructions mobilize. Eva Hajicˇová, Jarmila Panevová and Petr Sgall explain how the syntactic tagging of the Czech National Corpus is carried out, making use of a dependency based account of underlying, or tectogrammatical, structures. This is the second stage of the tagging procedure, itself divided into automatic and nonautomatic, or ‘intellectual’, tasks. The latter include recovery of deleted items in the text, the number and gender of various possessive adjectives and pronouns, and, hardly a surprise, topic–focus articulation, which shows again the crucial nature of information structuring in discourse. Jens Allwood offers an overview of recent work on the statistical analysis of the lingusitic properties of 25 different social activities. A corpus of spoken Swedish of one million words served as the basis of the research, tagged for the relevant characteristics. The numbers and ratios relating to lengths of utterances, variation in vocabulary, parts of speech, etc., provide reliable data for further hypotheses and analyses. Bruce Fraser investigates the puzzling phenomena of innuendo, that particular form of negative suggestion, which, unlike hinting, insinuating, suggesting etc., which can be labeled ‘implying’, has no verb (*He innuendoed that S.). Probably the most famous example of innuendo is captured in the following old and often repeated joke: a captain notes in the ship’s log “the first mate was drunk all day”. The next day the first mate notes in the ship’s log “The captain was sober all day”. What exactly is innuendo, how does it work, and how does it fit into linguistics? The author reviews some of the previous work on innuendo, and finds problems with it. Most seriously, that the class of ‘implied’ phenomena is often said to be covert in the sense that a speaker’s intention to imply is not intended to be recognized. The author suggests a different account based on the idea that innuendo is a restricted type of insinuation which itself is a type of implication, a communicative component of a communicative act that is itself intended to be recognized.
xviii Introduction
To conclude this Introduction, the editors wish to express their gratitude to the Hungarian National Research Programs for their grant toward the preparation of the final version of the typescript (OTKA P/33524), to the anonymous reviewers, who were kind and generous enough to devote their time and energy to read and comment on the articles, to Prof. Andreas Jucker, the series editor, for his unfailing trust and support, and to Ms. Judit Gervain for her enthusiastic collaboration to produce a perfect copy.
AUTHOR ""
TITLE "Selected books and articles by Ferenc Kiefer in semantics and pragmatics"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Selected books and articles by Ferenc Kiefer in semantics and pragmatics
1964
1966 1966
“A halmazelmélet egy nyelvészeti alkalmazásáról” [On a linguistic application of set theory]. In Általános Nyelvészeti Tanulmányok I, 187–200. “A jelentéselmélet formalizálásáról” [On the formalization of semantic theory]. In Általános Nyelvészeti Tanulmányok IV, 105–155. “Über Fragen einer formalen semantischen Theorie”. Sprache im technischen Zeitalter 19: 198–210.
——— and Samu Ábrahám 1966 A Theory of Structural Semantics. The Hague: Mouton. 1967 On Emphasis and Word Order in Hungarian [Ural – Altaic Series 76]. Bloomington: Indiana University Press. 1968 Mathematical Linguistics in Eastern Europe. New York: American Elsevier. 1969 “Az emfázis kérdéséhez” [On emphasis]. Magyar Nyelvo˝r 93: 97–155. 1970 “Egy új jelentéselmélet felé: A generatív nyelvelmélet jelentés-komponense” [Toward a new semantic theory: The semantic component of generative grammar]. Nyelvtudományi Közlemények 72: 193–199. 1970 “On the problem of word order”. In Progress in Linguistics, M. Bierwisch and K.E. Heidolph (eds), 127–142. The Hague: Mouton. 1972 “Einleitung: Semantik und generative Grammatik” In Semantik und generative Grammatik, F. Kiefer (ed.), VII–XV. Frankfurt: Athenäum. 1973 “On presuppositions”. In Generative Grammar in Europe, F. Kiefer and N. Ruwet (eds), 218–242. Dordrecht: D.Reidel. 1974 Essais de sémantique générale [Repères linguistiques]. Paris: Mame. 1976 “T. Schiebe’s theory on presuppositions (a review article)”. Acta Linguistica Hungarica 26: 232–240.
xx
Selected books and articles by Ferenc Kiefer in semantics and pragmatics
1976
1976
1977 1978 1978 1978 1978 1978 1980
1981
1981 1983 1983 1983
1984 1984 1986
“Toward a semantic classification of Swedish adjectivals”. In Proceedings of the 3rd International Conference on Nordic and General Linguistics, J. Weinstock (ed.), 462–467. Austin: University of Texas. “Néhány megjegyzés az aktuális mondattagolás és az elo˝feltevések közötti összefüggésro˝l” [Some remarks on the interrelationship between functional sentence perspective and presuppositions]. Nyelvtudományi Közlemények 78 (2): 371–378. “Some semantic and pragmatic properties of wh-questions and the corresponding answers”. SMIL 15: 42–70. “Les présuppositions et la grammaire générative”. Studi di Grammatica Italiana 143–159. “Adjectives and presuppositions”. Theoretical Linguistics 2: 135–173. “Factives revisited”. In Proceedings of the 12th International Congress of Linguists, 194–199. Vienna. “Factivity in Hungarian”. Studies in Language 2 (2): 165–197. “Functional sentence perspective and presuppositions”. Acta Linguistica Hungarica 27: 83–109. “Yes–no questions as wh-questions”. In Speech Act Theory and Pragmatics, J. Searle, F. Kiefer and M. Bierwisch (eds), 197–219. Dordrecht: D. Reidel. “Zur semantischen Klassifizierung ungarischer Adjektive”. In IV. Finnugor Kongresszus, Gy. Ortutay (ed.), 275–286. Budapest: Akadémiai Kiadó. “A kérdo˝ mondatról” [On question sentences]. Nyelvtudományi Közlemények 83: 273–329. Az elo˝feltevések elmélete [A theory of presuppositions]. Budapest: Akadémiai Kiadó. “What is possible in Hungarian?” Acta Linguistica Hungarica 33: 149–187. “A -hat/-het képzo˝ jelentéstanához: Az episztemikus -hat/-het” [On the semantics of the derivational suffix -hat/-het: The epistemic -hat/-het]. In Általános Nyelvészeti Tanulmányok XVI, 131–153. “A magyar aspektusrendszer vázlata” [A sketch of the Hungarian aspectual system]. In Általános Nyelvészeti Tanulmányok XV, 127–149. “Szemantika vagy pragmatika?” [Semantics or pragmatics?] Nyelvtudományi Közlemények 86 (1): 5–22. “Focus and modality”. In Language and Discourse: Test and Protest, J.L. Mey (ed.), 287–312. Amsterdam: John Benjamins.
Selected books and articles by Ferenc Kiefer in semantics and pragmatics xxi
1986
1986 1986 1987 1987 1988 1988 1992
1994 1994
1995 1998 1998 1998
1999 2000
“Epistemic possibility and focus”. In Topic, Focus and Configurationality, W. Abraham et al. (eds), 161–179. Amsterdam: John Benjamins. “A modalitás fogalmáról” [On the notion of modality]. Nyelvtudományi Közlemények 88: 3–38. “Some semantic aspects of indirect speech in Hungarian”. In Direct and Indirect Speech, F. Coulmas (ed.), 201–217. Berlin: Mouton de Gruyter. “On defining modality”. Folia Linguistica XXI (1): 67–94. “Charles Bally modalitás fogalma mai szemmel” [Charles Bally’s notion of modality from today’s perpective]. Tertium non datur 4: 29–45. “The pragmatics of answers”. In Questions and Questioning, M. Meyer (ed.), 255–279. Berlin: Mouton de Gruyter. “Ability and possibility: The Hungarian verb tud ‘to be able to’”. Studies in Language 12 (2): 393–423. “Aspect and conceptual structure: the progressive and the perfective in Hungarian”. In Fügungspotenzen, I. Zimmermann and A. Strigin (eds), 89–110. Berlin: Akademie Verlag. “Modality”. In The Encyclopedia of Languages and Linguistics, 2525– 2530. Oxford: Pergamon Press. “Some peculiarities of the aspectual system of Hungarian”. In Tense, Aspect and Action: Empirical and Theoretical Contributions to Language Typology, C. Bache, H. Basboll and C.E. Lindberg (eds), 185–205. Berlin: Mouton de Gruyter. “Le problème des équivalences pour l’expression de l’aspect et du mode d’action”. Cahiers d’études hongroises 7: 95–101. “Morphology and pragmatics”. In A Handbook of Morphology, A. Zwicky and A. Spencer (eds), 272–279. Oxford: Blackwell. “Modality and pragmatics”. Folia Linguistica XXXI (3–4): 241–253. “Modality”. In The Handbook of Pragmatics. J. Verschueren, J.-O. Östman, J. Blommaert and C. Bulcaen (eds), 1–36. Amsterdam, J. Benjamins “Modality”. In Concise Encyclopedia of Grammatical Categories, K. Brown and J. Miller (eds), 221–235. Amsterdam, Elsevier. Jelentéselmélet [The theory of meaning]. Budapest: Corvina.
DOCINFO
AUTHOR ""
TITLE "Pragmatics in Grammar"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Part I
Pragmatics in grammar
AUTHOR "Casper de Groot"
TITLE "Functional Grammar and the non-lexical expression of absence"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Functional Grammar and the non-lexical expression of absence Casper de Groot University of Amsterdam
1.
Introduction
In a limited number of languages of Europe we find non-lexical expressions which are used to indicate that a person is not present at a particular place. This type of grammatical marking has been given the name of ‘absentive’ (De Groot 1995b, 2000). The data of the Absentive in the languages of Europe offers an interesting case for linguistic theory. The first part of this paper contains data of the Absentive, a relatively unknown category in linguistic literature. I will argue that an appropriate account of the Absentive requires a functional model which must be able to accommodate syntax, semantics and pragmatics. The second part of the paper discusses the treatment of the Absentive in such a linguistic model, viz. Functional Grammar as developed by Simon Dik and others (cf. Dik 1997).
2. The data Languages may have several types of lexical means to express absence, i.e. a way of expressing that somebody is not present at a certain place. A tentative list of possible expressions includes the following: (1) a. b. c. d. e. f.
Peter is absent. Peter is not here. Peter is walking to the railway station. Peter is shopping at the market. Peter is away shopping. Peter went out for lunch.
Examples (1a) and (1b) contain only the information of absence. Examples (1c) and (1d) may, but do not necessarily, denote absence. The last two examples, (1e) and (1f), contain the information of Peter’s absence and also the activity he is, or will be involved in. The examples in (1) give one or more types of information with respect to the absence of Peter. These types of information are restricted to ‘being absent’, ‘being somewhere else’, ‘being on one’s way to some place, or ‘being out (doing something)’. They do not (necessarily) contain information about the period of time that Peter will be absent, or that Peter will come back after some time. There are about eight languages from Europe1 which have non-lexical, i.e. grammatical means to express that a person is not present at a certain place but absent involved in an action somewhere else.2 The following examples from Hungarian, Italian, and Norwegian give an illustration of the form the Absentive can take. Hungarian uses a copula together with the infinitive of a lexical verb; Italian does the same but also employs a preposition; Norwegian uses a different type of expression in which a copula is coordinated with a lexical verb, which takes the finite form. Consider: (2) a.
János levelet volt föladni. John letter.acc was post.inf ‘John was off posting a letter.’ b. Gianni è a mangiare. John is at eat.inf ‘John is off having lunch.’ c. Jan er og handlar. John is and shop.pres ‘John is off shopping.’
(Hungarian)
(Italian)
(Norwegian)
The examples in (2) contain the following types of information: (3) i. John is not present; ii. John is involved in an activity indicated by the lexical verb; iii. Based on pragmatic knowledge, it is predictable how long John will be away, or there is an assumption about the period of time that John will be away; iv. John will return after a period of time.
One context in which the Absentive is most naturally used is the following question-answer pair (on the phone/at the door):
Functional Grammar and the non-lexical expression of absence
(4) A: Péterrel szeretnék beszélni. Peter.com like.cond.1sg speak.inf ‘I would like to speak with Peter.’ B: De hiszen úszni van. well swim.inf is ‘Well, he is off swimming.’
(Hungarian)
Another situation in which the construction is often used is in messages people leave on the door, for instance: (5) We zijn lunchen. we are have.lunch.inf ‘We are off having lunch.’
(Dutch)
Note that the examples in (2), (4B), and (5) do not overtly express absence of the person specified by the subject. Since the expressions in these examples do not denote something else than the types of information specified under (3), I consider these forms to be grammatical expressions or grammaticalized forms of absence, in a similar sense as Progressive aspect is grammaticalized in ‘John is walking.’, namely in the form of copula be together with the stem of the lexical verb with the suffix -ing. A classical, well accepted way of testing the existence of grammatical categories in a language is the use of test frames. For instance, the presence of Perfective versus Imperfective aspect in languages is often straightforwardly motivated on basis of the following test frame: clauses with Perfective expect can be extended with the adverbial phrase “in an hour” and not with the adverbial phrase “for an hour” (“John has read the newspaper in an hour/*for an hour”). For the Imperfective forms hold the opposite (“John was reading the newspaper *in an hour/for an hour”). In the following sections I will give the most important properties of the Absentive by using the method of test frames. This approach serves two purposes: on the one hand, it illustrates the grammatical properties of the Absentive and on the other it presents a comprehensive overview of the linguistic facts. 2.1 Deictic center The Absentive expresses that the person referred to by the subject is not present at what can be called the deictic center. If the deictic center is specified in a clause, the use of the Absentive entails the dislocation of the event from the
5
6
Casper de Groot
deictic center, as is illustrated in example (6a). Collocation of the deictic center and an event is possible by not using the Absentive, but rather using, for instance, the progressive as in (6b). Consider: (6) a.
Toen Peter de kamer binnenkwam was Marie lunchen. (Dutch) when Peter the room came.in was Mary have.lunch.inf ‘When Peter entered the room, Mary was off having lunch.’ b. Toen Peter de kamer binnenkwam zat Marie (daar) te lunchen. when Peter the room came.in sat Mary (there to have.lunch ‘When Peter entered the room, Mary was having lunch (there).’
Example (6b) is appropriate in the context where Mary was having her lunch in the room, which Peter entered. The other example (6a) cannot be used in such a context. Example (6a) can only be understood in the sense that Mary was not present in the room.3 2.2 Remoteness and direct perception Absence also has to do with remoteness, i.e. it is not natural to use the Absentive when the person referred to is in the direct neighborhood of the Speaker. If the Absentive is used in the context that a person is off swimming, the person absent will not be in a swimming pool nearby, for instance his own swimming pool in the garden of his house, but rather somewhere else. The following example from Swedish illustrates that the property of remoteness is relevant, because the use of the Absentive forces a particular interpretation of the utterance. Consider: (7) John är och duschar. John is and have.a.shower.pres ‘John is off having a shower.’
When sentence (7) is used, the Addressee (A) knows or assumes that there is no shower in John’s house or in the hotel room where John is staying, and that John went somewhere else to have a shower. A Speaker will not use the Absentive if he has a direct view on the person absent and (part of) the activity he/she is involved in. The following sequence of utterances are considered ungrammatical in the languages with the Absentive, consider:
Functional Grammar and the non-lexical expression of absence
(8) a.
Ik zie Marie op straat lopen. I see Mary on street walk.inf ‘I see Mary walk down the street.’ b. *Ze is zeker zwemmen. she is certainly swim.inf ‘She must be off swimming.’
(Dutch)
In other words, the Absentive cannot be used with verbs and or expressions with denote direct perception. 2.3 Duration When the Absentive is used, there is an assumption about the duration of the absence of the Subject. Both Speaker and Addressee should hold this assumption. The following dialogue illustrates this observation: (9) A: Péterrel szeretnék beszélni. Peter.com like.cond.1sg speak.inf ‘I would like to speak with Peter.’ B: De hiszen, futballozni van. well play.football.inf is ‘Well, he is off playing football.’ A: Ja persze. Nyolckor visszahívom. indeed at.eight back.call.1sg ‘Indeed. I’ll call back at eight o’ clock.’
(Hungarian)
Participant B could have answered the initial question in a different way had he assumed A did not know about the activities of Peter. In that case he might have used something equivalent to “Peter is not here at the moment. You may call him after eight o’ clock.” One may, of course, also find the following type of dialogue: (10) A: Péterrel szeretnék beszélni. Peter.com like.cond.1sg speak.inf ‘I would like to speak with Peter.’ B: De hiszen, futballozni van. well play.football.inf is ‘Well, he is off playing football.’ A: Nem is tudtam, hogy jár futballozni. not even know.past.1sg that go.3sg play.football.inf ‘I didn’t know that he plays football.’
7
8
Casper de Groot
I do not, however, consider (10) a counter-example to the observation that there must be an assumption about the duration of absence. In the case of (10), participant B made the wrong assumption as to the pragmatic knowledge of A. By using the Absentive, B wishes to inform A about the predictable period of time of the activity Peter is engaged in, but not about the fact that Peter is a football player. The period of time someone is engaged comprises all activities, which are connected to the activity, specified by the lexical verb. The number and type of activities are not determined lexically, but pragmatically. The routine of Peter may for instance be: (11) a. b. c. d. e. f. g. h. i. j. k.
leave the house at 2 p.m. go to the football field change clothes do warm up exercises play a football match take a shower get dressed go to a bar have a beer go home arrive home at 8 p.m.
2.4 The Absentive is a stative expression The basic information expressed by the Absentive is that somebody is not present but somewhere else. That is why the Absentive can be used as an answer to a question of the type Where is X?, and therefore the Absentive must be considered a stative construction. For instance: (12) A: Wo ist der Hans? where is the Hans ‘Where is Hans?’ B: Er ist schwimmen. he is swim.inf ‘He is off swimming.’
(German)
By using the Absentive B leaves open several possibilities. Hans may be on his way to the swimming pool, in the swimming pool, or on his way back home. In many languages, stative constructions cannot be used with deontic modality, with the imperative or as the complement of the verb want, because
Functional Grammar and the non-lexical expression of absence
in all these cases an Agent is required. It is possible to say “Close the door!”, but not “*Be young!” It seems that the Absentive cannot be used with deontic modality (cf. (13a)), with the imperative or as the complement of the verb ‘want’. The Absentive can — like all stative expressions — be used with the optative (cf. (13b)). (13) a. *Je moet vanmiddag zwemmen zijn! you must this afternoon swim.inf be.inf ‘You must be off swimming this afternoon!’ b. Ik wou dat ze zwemmen waren. I would that they swim.inf were ‘I wish they were off swimming.’
(Dutch)
Consider the following example of a stative situation in which the Absentive is used: (14) A: Wann komt der Hans? when comes the Hans ‘When/at what time will Hans arrive?’ B: Er ist schwimmen. he is swim.inf ‘He is off swimming.’
(German)
By using the Absentive B implicitly refers to a point in time, namely the time point Hans normally arrives after swimming. 2.5 Directional verbs Directional verbs such as ‘go to’ or ‘come from’ specify just one direction, either away from the deictic center or towards the deictic center. For that reason, directional verbs cannot be used in the Absentive, because the Absentive encompasses both directions: going away and coming back (cf. (3iv)). Consider: (15) *Feri sétálni van a pályaudvarra. Feri walk.inf is the railway station.subl ‘Feri is off walking to the railway station.’
(Hungarian)
2.6 Discussion and conclusions of this section Above it has been shown that some languages of Europe have grammatical means to express absence. The grammars of those languages should therefore
9
10
Casper de Groot
account for the grammatical aspects of the Absentive. In the same way as grammars account for the order of constituents, agreement, tense distinctions, etceteras, the Absentive should be treated as part of grammar. The question arises which properties of the Absentive should be accounted for in grammar. It is clear that the use of the Absentive serves a communicative goal, i.e. to inform the Addressee about the absence of a person. If one wishes to include this pragmatic aspect of the use of the Absentive in a grammar of a language, such a grammar would then also give rules concerning social interaction. I will not defend this view. I would rather take the position advocated by the model of Functional Grammar, namely the following. In the functional view a language is regarded as an instrument which human beings use to achieve certain goals and purposes. These goals and purposes are taken to lie primarily in the establishment of complex patterns of social interaction. A Speaker uses linguistic expressions to communicate messages to an Addressee so as to change this Addressee in a certain way. From the functional point of view one wishes, whenever this is possible, to understand why languages are organized as they are, in the light of the way they are used. FG therefore seeks for functional explanations to account for structural properties of languages. A functional explanation of a linguistic phenomenon is a statement in which that phenomenon is shown to follow from one or more principles, which crucially refer to any of the functional prerequisites imposed on natural languages. It seems that functional explanations are complex and never simple in the sense of directly accounting for a linguistic phenomenon X in term of ‘THE’ function of X (cf. Dik 1986). It seems to me that grammar should at least account for the following four properties of the Absentive, because they relate to phenomena to be accounted for by grammatical rules and not rules of verbal or social interaction. The phenomena have to do with (1) remoteness, (2) stativity, (3) directionality, and (4) direct perception. Consider the following examples which illustrate that the Absentive has grammatical restrictions in the sense that it is incompatible with a number of types of adverbial phrases and the use of the Absentive as a complement of the verb “see” with direct perception: 1. Remoteness (16) Peter loopt hier naast mij. *Hij is boodschappen doen. Peter walks here next me *he is shop.inf ‘Peter is walking next to me. He is off shopping.’
Functional Grammar and the non-lexical expression of absence
2. Stativity Wat is Suzan aan het doen. ??Ze is zwemmen. what is Suzan to the do ??ze is swim.inf ‘What is Suzan doing? She is off swimming.’ b. *Wees zwemmen! be swim.inf ‘Be off swimming!’
(17) a.
3. Directionality (18) *Mark is naar het station wandelen. Mark is to the station walk.inf ‘Mark is off walking to the station.’
4. Direct perception (19) Ik zie Jan daar lopen. *Hij is zeker zwemmen. I see John there walk *he is certainly swim.inf ‘I see John walking there. He must be off swimming.’
In the next section I will give a brief outline of the Functional Grammar (FG) model. For a detailed outline of the model, I refer to Dik (1997). I will then discuss possibilities for an account of the grammatical aspects of the Absentive.
3. Functional Grammar 3.1 The structure of the clause In FG clauses are described in terms of abstract underlying clause structures, which are mapped onto the actual form of the corresponding linguistic expression by a system of expression rules. Schematically: (20)
UNDERLYING CLAUSE STRUCTURE Ø EXPRESSION RULES Ø LINGUISTIC EXPRESSION
The underlying clause structure is a complex abstract structure in which several levels of formal and semantic organization are distinguished:
11
12
Casper de Groot
(21) CLAUSE PROPOSITION PREDICATION PREDICATE TERM
Æ Æ Æ Æ Æ
speech act possible fact state of affairs property/relation entity
Clause and Proposition form the interpersonal level; the other three form the representational level. All predicates, or contentives, of a language are stored in a Lexicon. Predicates are contained in predicate-frames, structures which specify their fundamental semantic and syntactic properties, such as (1) the syntactic category of the predicate (Verbal, Nominal, Adjectival), (2) the number of arguments, (3) the semantic functions of the arguments (Agent, Goal, Recipient, etc.). Satellites (non-arguments) can extend predicate-frames. The semantic functions of the satellites express the relation between the predication or proposition and the satellites, whereas the semantic functions of the arguments express the relations between the predicates and the arguments. Arguments and satellites can be replaced by terms, i.e. the forms underlying NPs. Each level can be further specified by means of grammatical (morphosyntactic, morphosemantic) operators, such as number (a term operator), tense (a predication operator), and interrogative (a speech act operator). The layer of Predication will be particular relevant to the discussion of the Absentive. Therefore, I will pay some more attention to this level. 3.2 Nuclear, core, and extended predication Nuclear predications consist of predicates and terms. Terms refer to entities in some world, and predicates designate properties of, or relations between such entities. A nuclear predication as a whole designates a set of States of Affairs (SoAs). The term State of Affairs (SoA) is used in the broad sense of ‘conception of something, which can be the case in some world’. SoAs can be divided into different types, according to the values, which they can have for a number of distinguishing parameters. These parameters and their different values together define a semantic cross-classification of SoAs. The most important semantic parameters defining the typology of SoAs are Control, Dynamism, Telicity, and Momentaneousness. Nuclear predications can be further specified by predicate operators and satellites, both operating at Level 1. Operators (or π1-operators): e.g. aspectual distinctions; satellites: e.g. terms with the function of Manner, Speed, and Instrument. The result of these
Functional Grammar and the non-lexical expression of absence
extensions is called core predication. E.g. (22) in which the Perfective operator applies. Also note the satellite with the function of Manner: (22) prog readV (John)Ag (the book)Go (quickly)Manner [John is reading the book quickly.]
Core predications can be further specified by predication operators and satellites, both operating at Level 2. Operators (or π2-operators): e.g. temporal and some modal distinctions; satellites: e.g. terms with the function of Time and Place. The result of these extensions is called extended predication. For instance (23) in which the Tense operator with the value of Past applies. Note the satellite with the function of Location: (23) pres perf readV (John)Ag (the book)Go (superficially)Manner (in the library)Loc [John has read the book superficially in the library.]
3.3 The hierarchical ordering of operators and satellites The operators may apply to different semantic domains and therefore they may have scope differences (Hengeveld 1989). Similarly, satellites may be added to different semantic domains and therefore may have scope differences. Each level has its own set of operators (indicated by π) and set of semantic functions of satellites (satellites are indicated by σ). Schematically: (24) level 1 level 2 level 3 level 4
The four levels together can be schematically presented in the following hierarchical, embedded way: (25) π4 [π3 [π2 [π1 [Predicate + arguments]NUCL (σ1)]CORE (σ2)]EXT (σ3)]PROP (σ4)
Schema (25) clearly indicates that π4 operators have scope over the other operators; that π3 operators have scope over π2 and π1 operators, etc. Similarly, propositional satellites, σ4, have scope over all embedded levels including their satellites. The following example illustrates some of the scope differences (cf. b–e):
13
14
Casper de Groot
π4 int [π3 [π2 past [π1 prog [singV (John)Ag]NUCL (σ1 loud)Manner]CORE (σ2 bathroom)Location]EXT (σ3)]PROP (σ4) [‘Was John singing loudly in the bathroom?’] b. [[John was singing loudly] in the bathroom] c. past [John is singing loudly] d. [[John is singing] loudly] e. question [John was singing loudly in the bathroom]
(26) a.
Within each level there also seems to be an order between the operators. As for the levels 1 and 2 the following hierarchy seems to hold (see De Groot (1995a) for discussion and examples), where the operator right of an “<” has scope over the former one(s): (27) Hierarchy of operators (1) Aspect < Inner phasal aspect < Outer phasal aspect < Quantificational aspect < Participant-oriented modality (MOD1) < State of affairs-oriented modality (MOD2) < Polarity < Tense < Propositional-oriented modality (MOD3) < ………
Level π1 π2
π3
Aspectual operators are: Imperfective and Perfective; Inner phasal aspectual operators are: Ingressive, Progressive, and Egressive; Outer phasal aspectual operators are: Prospective, Immediate prospective, Recent perfect, and Perfect. The hierarchy in (27) posits e.g. that Aspect is within the scope of Inner phasal aspect and not reverse; and that Tense has scope over all aspectual operators, etc. Consider e.g. the following example of English and note the order of operators which reflects the scope differences: (28) π2 past perf π1 prog workV (John)Ag ‘John had been working.’
4. FG and the Absentive In this section I will discuss a treatment of the Absentive within the framework of Functional Grammar. I will first show that the Absentive belongs to the
Functional Grammar and the non-lexical expression of absence
domain of core predication. I will then suggest to account for the Absentive by means of a π2 operator, namely the operator of dislocation. Finally, I will try to find the place of this operator in the hierarchy presented in (27) above. 4.1 The location of the Absentive in the layered model In order to account for the Absentive it is relevant to know its location in the layered model. Since the Absentive relates a state of affairs to a deictic center, i.e. the Subject is involved in an action away from the deictic center, it seems reasonable to assume that an operator on the level of predication accounts for the Absentive. Note that predications can be located in time and space. Because of internal properties of the Absentive, it is, however, difficult to present many tests, which prove the exact location of the Absentive. I will discuss some of the incompatibilities below. First consider the application of tense, a π2 operator, together with the Absentive. The following example illustrates that the Absentive in Hungarian allows for different tense distinctions: (29) János úszni van/volt/lesz. John swim.inf is/was/will ‘John is/was/will be off swimming.’
(Hungarian)
Quantificational aspect, another example of a π2 operator, also applies: (30) Marietje was vandaag twee keer zwemmen. Mary was today two times swim.inf ‘Mary was off swimming twice today.’
(Dutch)
Example (30) also illustrates the application of a temporal satellite (‘today’), which indicates that a predication is involved. These examples suggest that the Absentive is located on the level of core or nuclear predication, because the Absentive can be located in time (cf. (29)), and, moreover, it is possible to quantify over the Absentive (cf. (30)). Tense and quantificational aspect operate on the level of core predication. Now the following difficulty arises. The Absentive does not allow for all operators on the level of core predication, for instance the use of deontic modality is not allowed. The incompatibility, however, has to do with a particular property of the Absentive, i.e. the property of being a non-agentive stative expression. This type of expression does not combine with deontic modality as shown in (31b–c), contrasted with (31a) which shows the opposite features.
15
16
Casper de Groot
(31) a. You must come! (agentive; dynamic) b. *You must be away! (non-agentive; stative) c. *Je moet vissen zijn. (non-agentive; stative) you must fish.inf be ‘You must be off fishing.’
(Dutch)
Note that (31b) and (31c) are ungrammatical under the deontic reading. When we now have a look at the operators and satellites, which relate to the level of the nuclear predication, we see that they all operate — so to speak — within the domain of the Absentive. Here are some examples of nuclear satellites, i.e. (σ1): a satellite with the function of beneficiary in (32a) and comitative in (32b). (32) a.
Peter is voor Marie bloemen kopen. Peter is for Mary flowers buy.inf ‘Peter is off buying flowers for Mary.’ b. Jan is met Marie zwemmen. John is with Mary swim.inf ‘John is off swimming with Mary.’
(Dutch)
Manner satellites do not apply, because they do not combine with stative, nonagentive SoAs. Consider the following examples, where (33b) illustrates the misfit of the Absentive with a manner adverb: (33) a. *Peter is slowly clever. b. *Peter is langzaam zwemmen. Peter is slowly swim.inf ‘Peter is slowly off swimming.’
(Dutch)
As for the operators, which apply on the level of the nuclear predication, Aspect is of particular interest. The Absentive is not compatible with the different forms of Inner phasal aspect. As has been pointed out in Section 2.3 above, the Absentive may encompass a number of activities, which makes reference to a phase in one activity impossible. This then also excludes Progressive aspect. Moreover, the stative nature of the Absentive rules Inner phasal aspect out. As for another type of Aspect, the Imperfective–Perfective opposition, Hungarian shows that only the Imperfective aspect can be used (cf. (34a)), and not the Perfective form (cf. (34b)): (34) a.
Zsuzsa a leckét írni volt. Zsuzsa the lesson.acc ipfv.write.inf was ‘Zsuzsa was off preparing her homework.’
(Hungarian)
Functional Grammar and the non-lexical expression of absence
b. *Zsuzsa a leckét volt meg-írni. Zsuzsa the lesson.acc was pfv.write.inf ‘Zsuzsa was off preparing her homework.’
When Hungarian uses Perfective megír it is not possible to refer to a point of time within the state of affairs, because with the Perfective aspect the state of affairs is presented as an indivisible unit with beginning, middle and end rolled into one (cf. Kiefer 1982). The Absentive allows reference to a point of time. Example (34a) can be combined with expressions such as (35): (35) Amikor János hívott, Zsuzsa a leckét írni volt. when John called Zsuzsa the lesson.acc write.inf was ‘When John called, Zsuzsa was off preparing her homework.’
It thus seems that, like in other stative constructions, the application of aspectual operators in the Absentive is neutralized. The Absentive can be used as the contents of a proposition and can be used under different speech acts. Consider (36a) with epistemic modality (π3-operator) and (36b) with interrogative illocution (π4-operator): (36) a.
Jan zou zwemmen zijn. John would swim.inf be ‘John would be off swimming.’ b. Is Jan zwemmen? is John swim.inf ‘Is John off swimming?’
(Dutch)
The examples in (36) illustrate two things. On the one hand, they illustrate that the location of the Absentive is predication indeed. The π3-operator has scope over the Absentive. On the other hand, they illustrate that the Absentive is fully integrated in the grammatical system of layers in Dutch.4 4.2 A π2-operator of dislocation After establishing the position of the Absentive in the layered model the characteristic property of the Absentive may be associated with an operator of spatial orientation. I suggest the introduction of the following π2-operator: dislocation. (37) The operator dislocation locates the predication distinct from the deictic center.
17
18
Casper de Groot
Example (38b) would then be the (abbreviated) underlying representation of (38a): (38) a.
Jan is vissen. John is fish.inf ‘John is off fishing.’ b. [π2 pres disl [visV (Jan)Ag]NUCL]CORE
(Dutch)
We could postulate that the opposite value of dislocation, namely collocation is an operator by default. The operator of dislocation neatly accounts for a number of grammatical properties of the predication. Among them are the following: 1. The operator excludes the extension of satellites, which specify a collocation with the deictic center, e.g. adverbial phrases of the type ‘next to me’ in instances where the Speaker and the deictic center coincides. 2. The operator avoids the Absentive to become a complement of verbs of perception (‘see’ and ‘hear’), because direct perception relates to collocation and not to dislocation. 3. The predication counts as a stative expression.5 This will prevent extensions of satellites with the function of direction. 4. The operator will serve as an input in the expression rule component, where special expression rules will be triggered. The operator can, of course, not account for extra-linguistic information such as duration of absence and assumptions hold by Speaker and Addressee. A grammar of verbal interaction must provide rules for these aspects of the Absentive. If we want to give the operator Dislocation a place in the hierarchy presented in (27) above, the data suggests that it is a π2-operator with the most narrow scope. We have seen already that it is possible to quantify over absentive states of affairs (cf. (30)). The following example shows that also Outer phasal aspect has scope over the Absentive: (39) John is zwemmen geweest. John is swim.inf been ‘John has been off swimming.’
The location of the Absentive in the hierarchy of operators is then as follows:
Functional Grammar and the non-lexical expression of absence
5. Conclusions In this paper I have brought forward a test case for linguistic theory: how to account for the grammatical expression of absence? I have shown that a functional grammar could account for all grammatical aspects of the constructions involved. Such a theory should accommodate the notion of deictic center and the operator of dislocation. Of course, the way one wishes to treat the Absentive in grammar depends on the scope of the grammar. Grammatical models may in this respect differ widely from each other. I have taken the position, advocated by the Dikian Functional Grammar, that grammar consists of rules relevant to the organisation of natural language. Rules that underlie language use belong to another grammar, namely the grammar of verbal interaction. The Absentive, which at first sight seems to mainly serve a communicative goal, i.e. to express that somebody is not present, has a great number of properties which can be accounted for in terms of grammatical rules.
Notes 1. The languages are: Dutch, German, Finnish, Frisian, Hungarian, Italian, Norwegian and Swedish. 2. I refer to De Groot (2000) for a detailed discussion of the properties of the Absentive and their occurrences in the languages of Europe. 3. The deictic center is not restricted to the place of the Speaker (S). It may be anywhere, e.g. home, office, beach etc. 4. This holds for the other seven languages as well.
19
20
Casper de Groot
5. I assume that the dislocation operator will introduce a π1 operator, which will specify the state of affairs to be stative, i.e. [−con], [−dyn]. This specification will prevent the application of aspectual operators.
References Dik, Simon C. 1986 “On the notion of functional explanation”. Belgian Joural of Linguistics 1: 11–52. 1997 The Theory of Functional Grammar. Part 1 and 2. Berlin: Mouton de Gruyter. Groot, Casper de 1995a “Aspect and tense in Functional Grammar”. In Typological Perspectives [Temporal Reference, Aspect and Actionality. Vol 2.], P.M. Bertinetto et al. (eds), 29–45. Turin: Rosenberg & Sellier. 1995b “The Absentive in Hungarian”. In Levels and Structures [Approaches to Hungarian, Vol 5], I. Kenesei (ed.), 45–61. Szeged: JATEPress. 2000 “The Absentive”. In Tense and Aspect in the Languages of Europe, Ö. Dahl (ed.), 641–667. Berlin: Mouton de Gruyter. Hengeveld, P.C. (Kees) 1989 “Layers and operators”. Journal of Linguistics 25: 127–157. Kiefer, Ferenc 1982 “The aspectual system of Hungarian”. In Hungarian Linguistics, F. Kiefer (ed.), 293–329. Amsterdam: Benjamins.
AUTHOR "Henk van Riemsdijk"
TITLE "A far from simple matter"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
A far from simple matter Syntactic reflexes of syntax–pragmatics misalignments* Henk van Riemsdijk Tilburg University
1.
Introductory remarks
It is a well-known fact that the relationship between syntactic structure and its semantico-pragmatic impact is often less than trivial or straightforward. One of the many cases that show this is found in examples like the following: (1) Would you be kind enough to open the window?
The point here is that this is syntactically speaking a yes/no question but is pragmatically most naturally interpreted as a request. Hence, (2a) is slightly surprising as an answer, while (2b) is the expected answer. (2) a. Yes, I would. b. OK, I will. / No, I won’t.
In the pragmatics literature, such cases are referred to as ‘speech act idioms’ or as ‘short-circuited conversational implicatures’ (cf. Horn 1988 for a concise discussion). Put in yet another, perhaps more syntactic way, one might say that the subordinate clause in (1) is semantico-pragmatically superordinate in that the subordinate clause conveys the content of the request while the superordinate clause translates into ‘please’. There are other constructions in which what appears to be superordinate and subordinate from the syntactic point of view seems to be reversed when interpreted from the perspective of semantics and/or pragmatics. Certain socalled hedges are notorious in this regard. Take an example like (3). (3) This argument is close to convincing.
22
Henk van Riemsdijk
At first sight, close is the predicative adjective while to convincing is a PP dependent on close. The actual meaning, however, is that close to is a hedge modifying the adjective convincing, yielding a meaning roughly like ‘almost fully convincing’. Hedges constitute a rather notorious semantic problem because they cause problems for straightforward accounts in terms of truth conditions. Strictly speaking, anything that is not fully convincing is not convincing. But on the other hand, conventions of understatement might cause you to utter (3) when you can find no fault with the argument. Hedges, in fact, are most often used in cases where the speaker does not want to assume full responsibility for a statement. A variety of terms have emerged in discussions of such matters. In a recent article, Lasersohn (1999) speaks about ‘semantic slack’ and ‘pragmatic slack’, and devises a theory of ‘pragmatic halos’ to deal with phenomena of this general type. I will gladly leave the pragmatic intricacies to the experts, but in the present article I propose to discuss a few cases in which such apparent misalignments of what is subordinate and what is superordinate may well have a structural effect on the syntactic representations that are involved. More specifically, I will discuss three cases in which the superordinate-subordinate conflict reduces to the question of which one of two candidates is the head of the construction.
2. A couple of harder than usual problems 2.1 Direct partitive constructions In an early article on this type of misalignment phenomena, Kajita (1977) points out that in complex nominal expressions of the type Det N1 of Det N2 it is sometimes the second N which is felt to be the semantic head, even though on a standard analysis, the first N is the matrix head, while the second N is embedded within a complement or adjunct to the first noun. Kajita cites examples like the following: (4) a. A couple of weeks passed. b. The report does not contain a fraction of truth. c. We did not find a vestige of evidence.
He points out, among other things, that in certain varieties of English the preposition of can be omitted: (5) A couple weeks passed.
A far from simple matter
As a matter of fact, omitting the preposition is the normal way to express Direct Partitives (cf. Vos 1999) or Pseudo-Partitives (Kubo 1996) in a number of languages including Dutch, German and Swedish. He also mentions the fact that some of the Det-N1 combinations can be used adverbially in certain constructions: (6) Her eyes are a shade too small and a fraction too near together. (Kajita 1977: (21))
A final observation that is worth mentioning is that the quantifier-like nature of the first noun in an example like (4c), as opposed to a regular complex NP like (7), shows up when not is adjoined to (or otherwise a first constituent of) the NP in question. (7) the capital of the nation (8) a. The capital of the nation was not bombed. b. Not the capital of the nation was bombed. (* on the reading corresponding to (8a)) (9) a. Many students did not show up. b. Not many students showed up. (10) a. We did not find a vestige of evidence. b. Not a vestige of evidence was found.
In Kajita’s times, when X-bar theory was still optimally simple: one head per phrase, there was a serious problem in accounting for the relative dominance of the more embedded noun. How could it act as the semantic head of the complex noun phrase as a whole, or, alternatively, how could the upstairs noun, ostensibly the real head, end up semantically and, as in (10), syntactically, demoted? Kajita suggested that this might be achieved by means of a rather complex rule of reanalysis. From a more modern perspective, this particular case is somewhat less daunting. Functional heads have appeared on the scene and the notion of (extended) projection1 makes it possible for a single projection to have one ore more functional heads in addition to the lexical (semantic) head. In a further refinement, it has been suggested that in addition to functional and lexical heads, an intermediate category of semi-lexical heads should be distinguished.2 This refinement seems necessary to account for the fact that even properties relating to semantic headship seem to oscillate between N1 and N2. I will limit myself to a brief illustration here.
23
24
Henk van Riemsdijk
Take the following examples: (11) a. John ate a tray of pastries. b. John carried a tray of pastries. c. John turned over a tray of pastries.
In (11a) it is clearly the pastries which get eaten, hence N2 must be taken to be the semantic head that satisfies the selectional requirements of the verb eat.3 In the second example, however, it is clear that tray is the semantic object of the verb carry. It is only by implication that the pastries are also carried because they happen to be on the tray. Turning now to (11c), we find that this sentence is actually ambiguous between the two readings. Either it means that the whole tray is turned over (presumably causing the pastries to fall to the floor), or it means that all the pastries on the tray are each turned over individually, resulting in a tray full of upside-down pastries, but with the tray itself still in its original position. These semantic indications for a dual structure can be supplemented by a host of indications that syntactically as well N1 and N2 must often be assumed to be members of one and the same (extended) projection, implying that N1 must be taken to be a semi-lexical head. Limiting the discussion again to just one example, consider the fact that a case that is assigned to the whole complex noun phrase can deploy itself on N1 and N2 alike, creating a pattern looking like case agreement. This can be straightforwardly seen in German examples like the following: (12) mit einem vollen Glas rotem Wein with.dat a.dat full.dat glass.dat red.dat wine.dat ‘with a full glass of red wine’
On an analysis in which rotem Wein is an independent noun phrase embedded inside the matrix noun phrase einem vollen Glas this case pattern would be entirely unexpected. For details the reader is referred to the references cited above (cf. note 2). For our purposes in the present article, we may conclude that the problem Kajita brought up in the 70s was a genuine problem of the kind we are interested in here, a misalignment between the syntactic head and the semantic head, but that the revisions of the theory of phrase structure that have developed during the 80s and 90s yield relatively straightforward solutions.
A far from simple matter
2.2 The ‘far from’ construction Whether the second and third problems that Kajita (1977) points to can be dismissed in a similarly straightforward manner is much less clear. I have dealt with the third one, the issue of Transparent Free Relatives (TFRs) elsewhere and will return to it in Section 3 of the present article. The second issue is that of adjective + preposition + X combinations that look as if the adjective is the head with the PP as its complement. One such case was briefly alluded to in the introductory Section 1. And here again, the question arises whether what looks like the matrix head is not really some subordinated, adverb-like element. Take the following two examples, both taken from Kajita (1977: (2/4)): (13) a. The airport is far from the city. b. Those people are far from innocent.
In (13a) there is little doubt that far is the main predicate, the property predicated of airport, and, consequently, that from the city is a PP-complement to the head far. But in (13b) what seems to be expressed is that those people are innocent to a minimal degree. In other words, innocent seems to be the semantic head, while far from is a kind of adverbial modifier. Kajita lists half a dozen more examples that have this same property: close on, close to, next to, other than, nearer to, greater than. Of these I cite two examples: (14) a. There are next to no statistical data available. b. He greeted me with greater than normal politeness.
In addition he also lists a number of cases that are quite similar except that the (pseudo-)head is a non-adjectival element: (15) a.
This did not cause anywhere near the difficulty experienced by the retarded children. b. Your mother as much as told me so.
It is indeed true that such cases can generally be paraphrased with the first part substituted by some kind of adverb: (16) a. Those people are hardly innocent. (cf. (13b)) b. There are virtually no statistical data available. (cf. (14a)) c. Your mother practically told me so. (cf. (15b))
But it would hardly do to claim that far from, next to, as much as, etc. can be simply analyzed as adverbs. For one thing, they do seem to consist of several
25
26
Henk van Riemsdijk
words and cannot easily be identified with compounds of the normal type. Furthermore, it cannot be an accident that they all have an analytical use, that is, a use in which indeed the first part is the head and the second a complement or adjunct embedded under it: (17) a. The airport is far from the city. (= (13a)) b. Her house is next to some general’s mansion. c. Have you been anywhere near the front?
The existence of these analytical uses will account for the choice of the preposition in the adverbial use if the latter is derived in some way from the analytical construction. Such a derivational approach is indeed what Kajita assumes, and he offers some interesting additional evidence. First, he notes that when used attributively there is a contrast between the two cases. Consider the prenominal, attributive equivalents of (13a–b): (18) a. *the far from the city airport b. those far from innocent people
The ungrammaticality of (18a) follows from the principle that says that prenominal adjectives must be adjacent to the noun they modify.4 But on the assumption that the structure of (18b) is identical to that of (18a), the same principle should apply to exclude this one as well. However, if we can say that far from in (18b) is somehow demoted, inserted as an adverb, invisible to the principle in question, then innocent will be the relevant head and there is no violation of the adjacency principle. The Dutch equivalent of far from adds yet another dimension to this argument. In Dutch, prenominal adjectives are generally inflected with a schwa: -e.5 By this test, it is the second adjective, again, which acts as the head of the adjective phrase: (19) a. *een ver/verre van de stad vliegveld a far /far from the city airport cf. een verre stad a far-away city b. twee ver(re)6 van onschuldig-*(e) mensen two far from innocent people
A last fact which we may mention here is that there are cases in which we can detect a clear ambiguity between the two readings. This can be seen in the following examples from Dutch:
A far from simple matter
(20) a.
Zij las alles behalve de bijbel. she read everything except the bible b. Zij reageerde alles behalve enthousiast. she reacted anything but enthusiastically c. Zij was alles behalve een goede partner. she was everything except a good partner
(20a) typically means that she read literally everything, but she did not read the bible. In other words, alles is the object of the verb, not bijbel. In (20b) we have the opposite situation: the adverb modifying the verb must be enthousiast since alles cannot function as an adverb (*zij reageerde alles). But (20c) is ambiguous.7 It can either mean that she was everything to me, except a good partner, or it can mean that she was a far from good partner to me, without anything being said about whether she was anything else to me instead. Hence, there are good reasons to assume two different structures for the two types of meanings. Here, then, we have a case in which a solution along the lines suggested above in connection with partitive constructions does not seem to be viable. I have argued that there is a fundamental theoretical reason why semi-lexical heads must be of the same syntactic category as the lexical head (cf. Van Riemsdijk 1998b). But the above cases show distinct category status for the competing heads. In (14b), for example, greater has got to be an adjective, while normal politeness is evidently a noun phrase.8 Instead, I believe that the far from construction is a genuine case of what I have called a ‘graft’.9 On this type of approach, the noun phrase in (21a) will have a (simplified) structure such as the one given in (21b).
27
28
Henk van Riemsdijk
(21) a. a far from simple matter b. DP
D
N¢
a
AP
N
A
matter
simple from A
far P
AP A
PP AP
In the next section I will turn to Kajita’s third case, that of what Nakau (1971) called pseudo-free relatives and which, following Wilder (1998) I refer to as transparent free relatives (TFRs).
3. Transparent free relatives 3.1 The Japanese legacy: Nakau and Kajita The first linguist, as far as I can determine, who recognized the relevance of Transparent Free Relatives (TFRs) is Nakau (1971). The issue was taken up again and taken further by Kajita (1977). McCawley (1988), referring to Kajita, also devotes a brief discussion to them. And much more recently, it was Wilder (1998, 1999) who took up the issue again with an article to which I have reacted in several papers.10 TFRs look like Free Relatives (FRs). Take the following example which is taken from Kajita (1977: (29a)) and cited in McCawley (1988: 732).
A far from simple matter
(22) The man entered the cockpit carrying a gun, a razor, and a can of what the crew took to be gasoline.
In normal FRs, the invisible head is interpreted either as a definite noun phrase or as a universally quantified one. In other words, an FR like (23a) either means (23b) or (23c). (23) a. I am going to drink what you order. b. I am going to drink the specific beverage that you order. (definite reading) c. I am going to drink whatever beverage you order. (universal reading)
But (22) differs from (23) in this respect because the interpretation is indefinite: (24) The man entered the cockpit carrying a gun, a razor, and a can of something that the crew took to be gasoline.
Another way to put this is to say that there is not really an invisible head here but that the predicate nominal in (23) is, in a sense, the head of the relative clause. The rest of the apparent free relative is indeed, as McCawley points out, most readily interpreted as a ‘hedge’. So we again have a situation in which some element which is, apparently, deeply embedded in the syntactic structure is prominent from the point of view of the semantics and the pragmatics of the sentence, while the surrounding material in which it is embedded is pragmatically demoted to the status of a hedge. There are, indeed, many syntactic and semantic indications that the predicative XP which is apparently inside the TFR has properties that make it look as if this XP, which I call the ‘callus’,11 is part of the matrix rather than deeply embedded inside a relative clause. I will summarize them very briefly here. The first three of these have been established in Wilder (1998, 1999), the fourth, fifth and sixth have been added in Van Riemsdijk (2000a,b). 1.a TFR with a plural callus will take plural agreement in the matrix (as in (25b)) despite the fact that the wh-word what, which seemingly heads the relative clause, can only take singular agreement in normal FRs, cf. (25a): (25) a. What pleases/*please me most adorns/*adorn the living room wall. b. What *seems/seem to be several meteorites *was/were lying on the lawn.
29
30
Henk van Riemsdijk
2.a TFR with a human callus can refer to humans, as in (26b), despite the fact that the wh-word what, which seemingly heads the relative clause, cannot refer to humans in normal FRs, cf. (26a): (26) a. *She invited what was giving a talk that afternoon for lunch. b. She invited what I took to be a policeman. (Wilder’s (21))
3.extraction out of the callus of a TFR is possible without the complex noun phrase constraint effect that one would expect if the callus were deeply embedded inside the FR:12 (27) a. What NP is the verb what Stowell calls adjacent to? b. *What NP does the verb c-command whatever constituent is adjacent to? c. *What NP does the verb c-command the PP that is adjacent to?
4.the callus can be an idiom chunk that is idiomatically licensed in the matrix and not inside the FR, and furthermore, the idiom chunk can contain a pronoun bound by an element in the matrix; in normal FRs and in headed relatives this is not possible: (28) a. Nick lost what seems to be called his marbles. b. #Nick lost whatever round objects are called his marbles. (non-idiomatic reading at best) c. #Nick lost the round objects that are called his marbles. (non-idiomatic reading at best)
5.the callus can contain a bound anaphor that is bound in the matrix and not inside the FR; this is neither possible in normal FRs nor in headed relatives: (29) a. They live in what is often referred to as each other’s backyard. b. *They live in whatever location you used to refer to as each other’s backyard. c. *They live in the place that you used to refer to as each other’s backyard.
6.when the callus is a noun phrase, its case must satisfy not only the requirements inside the TFR but also those of the matrix environment:13
A far from simple matter
(30) a.
Ich habe mir was man als einen.acc I have me what one as a schnellen.acc Wagen bezeichnen könnte gekauft. fast car characterize could bought ‘I have bought what one could characterize as a fast car.’ b. Ich habe mir was von vielen I have me what by many Ï *ein.nom schneller.nom Wagen ¸ als Ì ˝ Ó *einen.acc schnellen.acc Wagen ˛ as a fast car bezeichnet werden würde gekauft. characterized be would bought ‘I have bought what by many would be characterized as a fast car.’ Ï *ein.nom schneller.nom Wagen ¸ c. Was viele als Ì ˝ Ó *einen.acc schnellen.acc Wagen ˛ what many as a fast car bezeichnen würden wird selten gekauft. characterize would is rarely bought d. Was als ein.nom schneller.nom Wagen bezeichnet what as a fast car characterized werden könnte wird selten gekauft. be could is rarely bought
There is actually a seventh consideration, discussed in Kajita (1977), which should be brought up here since both Wilder and I had failed to do so in our recent publications on TFRs. This is an argument derived from the fact that only identical categories can be conjoined in coordinated structures. Consider the following example. (31) a. He served wine and what the Italians call prosciutto. b. *He served wine and what the Italians call al dente. c. He served the wine and what was simmering on the stove.
In these examples, wine is apparently conjoined with a (T)FR. In (31a) prosciutto is a noun. Hence, we would expect the conjunction to be grammatical, regardless of whether it is the (T)FR or the callus (prosciutto) that is conjoined with wine. But in (31b) the callus is al dente, which in Italian is a PP but when used in English is presumably an adjective or an adverb. And indeed on the typical TFR reading (‘something which the Italians call al dente’) this sentence is out. If we forcibly construe it as a true FR (‘the thing that the Italians call al
31
32
Henk van Riemsdijk
dente’), the sentence could be marginally grammatical. In this case (31b) would be on a par with a regular FR case like in (31c), which is perfectly grammatical. These seven considerations demonstrate, convincingly in my mind, that the callus is, in Kajita’s and McCawley’s words, the head of the TFR. The question then arises how TFRs can be analyzed to reflect this insight. Kajita and McCawley suggest that this should be done in terms of a rule of ‘reanalysis’. But how this reanalysis is supposed to work remains entirely vague, except for the fact that a transformational apparatus of considerable power would be required to raise the callus into head position and to adjoin the remnant relative clause to its left.14 Leaving a transformational derivation aside, Wilder (1998, 1999) suggests that there are two viable alternatives, viz. a three-dimensional constituent sharing analysis and a backward deletion or ellipsis approach. Wilder adopts the latter approach, while Van Riemsdijk (1998a) pursues the former option, which traces its antecedents to Lakoff (1974) and can be formalized along the lines proposed in Moltmann (1992). A formalization of three-dimensional structures using minimalist terminology can be found in Citko (1998). Under the deletion approach and under the grafts theory, the structure of an example like (32a) would be roughly as shown in (32b) and (32c) respectively. (32) a. He carried what the crew took to be gasoline. b. He carried [whati the crew took [e]i to be gasoline] gasoline.
A far from simple matter
IP
c.
DP
VP V
he
DP
carried gasoline
whati the crew
took
ei
to be V
DP V DP Spec
DP VP
IP VP
IP CP
In the next and final subsection, I will present an argument against Wilder’s proposal. 3.2 Problems with a deletion account An immediate problem with a simplistic deletion account is that a FR like the bracketed one in (32b) does not really have a source. In such an analysis, the FR is something like an adjunct to a head noun, gasoline. Of course, nouns do take relative clauses, but, at least in English, we do not expect such relative clauses to the left. Rather, relative clauses are normally positioned somewhere to the right of the head noun. Of course, this particular relative clause is a reduced one in that its predicate nominal is deleted, but it is not reduced in the relevant sense. Participles and the like, which are conceivably derived from relative clauses, would occur on the left hand side, but, unlike these, the TFR is finite and still carries its wh-word. That is not all. The callus can also be an AP or a PP, as in (33).
33
34
Henk van Riemsdijk
(33) a. The auk is what biologists term pterorhine. b. They remain what is called of two minds about this issue.
Needless to say, APs and PPs do not take relative clauses. The matter is further complicated by the fact that the callus can also be embedded inside other XPs. For example, the callus can be an attributive adjective or an adverb modifying some other constituent like an adjective. Consider (34). (34) a.
Bill owns three what some people would consider to be extravagant cars. b. In this example, the variable is what most linguists would characterize as improperly bound.
In (34a) we know that the TFR does not have the full noun phrase in its scope since it follows the numeral. And in (34b) the reading in which the TFR has scope only over the adverb is the most natural one given the fact that we are talking about a variable, hence there is no discussion about it being bound. So what the hedge is about is the fact that binding is improper. Again, attributive adjectives and preadjectival adverbial modifiers cannot host relative clauses of any kind. Yet, a deletion analysis would seem to force the assumption on us that they can. Wilder’s proposal, of course, is more sophisticated than this. He assumes that the FR originates as an ‘independent phrase marker’. Subsequently, this independent phrase marker, presumably a CP, is inserted in its surface position, a process which Wilder takes to be part of the phonological side of the grammar. Then deletion applies. This is the way Wilder illustrates his proposal: (35) a.
Syntax: independent phrase markers [he bought [DP a guitar]] [what he took to be [DP a guitar]] b. Phonology: parenthetical placement and deletion John bought ·what he took to be a guitarÒ a guitar
This may be a way to avoid the problems mentioned above, and it is attractive in a number of ways. First, it is true that parentheticals are often used to express hedges in a way very similar to TFRs. The following are some examples. (36) a. He carried, if I am not entirely mistaken, a can of gasoline. b. This is not, as far as I remember, an Italian wine.
In fact, one might say that TFRs and constructions of the far from-type are special cases of parenthetical constructions: basically they are the same except for the fact that for one constituent (the callus) they are parasitic on the host structure.
A far from simple matter
The problem of specifying what is a well-formed parenthetical is not an easy one. And similarly, the question of when and where a constituent may be shared between a parenthetical and the host structure is far from trivial. In the case of far from there is probably no escaping the fact that lexical factors play a role, since these constructions are highly sensitive to grammaticalization. In the case of TFRs, however, this is much less obvious. Observationally, the callus has to be a predicative XP in the TFR. But other than some tentative suggestions in Van Riemsdijk (2000b) I have nothing further to say about this question, pending further investigation. This problem, needless to say, remains neutral between the deletion and the grafting approaches. Second, certain relative clauses, in particular ad-clausal ones, are often inserted inside a sentence in the manner of parentheticals. Consider (37). (37) He was carrying a razor, which is bad enough, and a can of gasoline, which is deadly.
Third, the very TFRs that we are looking at can, when complete, be used as righthand sentential conjuncts: (38) He was carrying a can of gasoline, or (at least) what I thought was a can of gasoline.
There is no denying, then, that there is considerable similarity between parentheticals and grafts. On the three-dimensional theory which I am advocating, parentheticals are treated pretty much like grafts, as pointed out above, except that there is no (visible) callus. So this is not a point of contention. I nevertheless have two problems with Wilder’s proposal. First, the position of an unreduced parenthetical relative clause is different from that of a reduced one: (39) a.
He took his daughter to the movies, or what I thought was his daughter. b. He took his daughter, or what I thought was his daughter, to the movies. c. *He took, or what I thought was his daughter, his daughter to the movies.
(40) a.
He has three obscene pictures on his office walls, or at least what I would consider obscene *(pictures). b. He has three obscene, or at least what I would consider obscene, pictures on his office walls.
35
36
Henk van Riemsdijk
c. *He has three, or at least what I would consider obscene, obscene pictures on his office walls.
Though the theory of grafting has not been developed in any detail yet, it is still rather easy to see that this difficulty will not arise. That is so because it is the shared constituent which determines where the material of the TFR goes. If, as is true in most of the English examples, the shared constituent is at the end of the TFR, then the TFR must be linearized in front of that element.15 The second point, related to this property, is this. As I have argued in Van Riemsdijk (1998a), contra Wilder, the callus can be in the middle of the TFR. This is illustrated in many of the Dutch and German examples discussed above. Consider (30a), repeated here as (41). (41) Ich habe mir [was man als einen.acc I have me [what one as a schnellen.acc Wagen bezeichnen könnte] gekauft. fast car characterize could bought ‘I have bought what one could characterize as a fast car.’
Here the callus (einen schnellen Wagen) is in the predicate nominal position within the relative clause, but because German is verb final it is followed by the verb cluster. From the perspective of an approach in terms of parentheticals, this would mean that parentheticals can sometimes be discontinuous. As far as I am aware, however, discontinuous parentheticals do not exist. Certainly crude attempts at constructing them like the following are hopelessly bad: (42) a. *He was carrying, if I am not, a can of gasoline, entirely mistaken. b. *There was, to the best, no solution, of my knowledge, to this problem. (43) a. #These are, I believe, headed for, you might say, the shredder. b. *if = These are, I believe you might say, headed for the shredder.
(43a) is ungrammatical on the reading where the two chunks of parenthetical are interpreted as constituting a single parenthetical as in (43b), though it is grammatical if the sentence happens to have two independent parentheticals in it. While much remains to be done in working out the details of a theory of grafting, I believe that the positional properties of TFRs that I have just listed argue in favor of it and against an approach along the lines suggested by Wilder.
A far from simple matter
4. Concluding remarks There can be no doubt that the syntactic and pragmatic properties of a sentence are often misaligned. In particular, what is pragmatically the main or dominant element in the message may very often show up as an embedded or even adjoined element in the syntax. There is nothing wrong with this in a theory which has autonomous components for syntax, semantics and pragmatics. What the above examples have shown, however, is that there are cases in which, so to speak, pragmatics has intruded into syntax, resulting in structures that are no longer representable as canonical trees but which are, in a sense, closer to the semantics or pragmatics of what is being conveyed. The idea that I have argued for is that such pragmatic intruders often come in the guise of grafts. In his 1993 article for the Nicolas Ruwet Festschrift, Ferenc Kiefer asks ‘où s’arrête la syntaxe?’, a highly relevant question, as we have just seen. His conclusion in that article was that the idea of an autonomous and formal syntactic component should be weakened to a certain extent. He suggests that cognitive principles that have their roots in the way in which our everyday way of thinking is organized may frequently intrude into the more formal domains of linguistic description. From this perspective, we syntacticians are led to ask ‘où s’arrête la pragmatique?’. What I have tried to do in the present contribution is to suggest that at least certain intrusions into syntax should actually be accommodated within formal syntax proper by means of a theory of grafts which, while sketchy and programmatic for the time being, holds considerable promise when it comes to accommodating precisely this type of interaction between pragmatics and syntax.
Notes * I would like to thank the students in my class at the Thermi/Mytilene (Lesbos) International GLOW Summer School in the summer of 1999 for lively discussion, in particular Elena Anagnostopoulou, Barbara Citko, Angeliki Ralli and Edwin Williams. Thanks are also due to Alex Grosu for stimulating e-mail exchanges and to Masayuki Oishi for sending me some of the harder to find Japanese papers. 1. See Grimshaw (1991), Van Riemsdijk (1998b). 2. Cf. Emonds (1999), Kubo (1996), Van Riemsdijk (1998b) and Vos (1999). 3. I am ignoring here the gastronomical trick of preparing food receptacles that can themselves be eaten such as trays made from deep fried potato dough or what have you.
37
38
Henk van Riemsdijk
4. See Emonds (1976, 1985), Williams (1982), Van Riemsdijk (1993). 5. Roughly speaking this is always true except when the noun phrase is in the indefinite neuter singular, in which case inflection is zero. 6. The preferred form in this adverbial use is verre. This is not to be confused with the inflected form of the adjective. It is a somewhat archaic frozen form, which is incidentally another indication that these constructions have been subject to a substantial degree of grammaticalization. This grammaticalization effect is also evidenced when the ‘adverb’ is used in isolation, elliptically. Here verre van can be used without any object of the preposition, where in English a pronoun is required: (i) (ii)
Hij is niet onschuldig — verre van. He is not innocent — far from *(it).
7. A referee points out, correcly, that according to prevalent Dutch spelling conventions alles behalve can also be spelled as a single word: allesbehalve. Thereby, an example like (20c) is disambiguated through orthographic means. This may quite plausibly be interpreted as a signal for some form of reanalysis such as the ‘graft’ proposal to be detailed below. 8. The categorial status of the complement may not be fully transparent in all cases. Take the examples of the type (19b). If we say that innocent is an AP here, we are allowing the preposition from to take AP-complements. Indeed, from also occurs in expressions like the sea turned from blue to gray, but color terms tend to waver between adjectival and nominal use. On the one hand, color terms are easily usable in nominal positions (blue is a nicer color), but on the other hand, comparatives are not excluded, as in the light changed from darkish to brighter than my eyes could tolerate. Pursuing this matter is beyond the scope of this article, and the main thrust of the present text is unaffected by this issue. 9. Lakoff (1974) had a similar idea and called such structures ‘syntactic amalgams’. I prefer the botanical metaphor since it typically applies to trees: the idea is that a subtree (the graft or the scion) is grafted onto a matrix tree (the stock). The junction of the two substructures, again following botanical terminology, I refer to as the ‘callus’. To the extent that the notion of grafting involves subtrees that are joined together, it might be thought to be identical to the minimalist notion of ‘merge’. The difference is, however, that ‘merge’ inserts the root of one subtree into a preterminal element of another subtree, whereas the grafting process joins two subtrees at the (pre-)terminal level. (In the latter respect, of course, the metaphor is not entirely appropriate.) See Citko (1998) for an extension of the notion ‘merge’ to threedimensional structures. 10. See Van Riemsdijk (1998a, 2000a,b). 11. See note 9 for some comments on the choice of terminology. Under the theory I advocate, the XP is shared between the matrix and the relative clause. Given the botanical metaphor I have adopted, this shared predicative XP is called the ‘callus’. 12. There is an additional argument, related to the one from extraction, which has to do with apparent extraction from noun phrases in Dutch, cf. Van Riemsdijk (1998a). In an example like (i), it appears as if the wh-word has been extracted out of a PP which is embedded within a noun phrase.
A far from simple matter
(i)
Waari heeft Jan [de conceptversie van ei] afgewezen? what has Jan [the draft.version of rejected ‘What did Jan reject the draft version of?’
But it can be shown that this type of extraction is only possible if the PP is somehow detached from its nominal host. For example, matrix adverbs can separate the PP from the NP, as in (ii). See Van Riemsdijk (1997) for details. (ii)
Waari heeft Jan [de conceptversie] gisteren (‘yesterday’) [van ei] afgewezen?
In the absence of an adverb, the detachment in question is string vacuous. Consider now what happens if the noun phrase from which detachment takes place is part of a TFR. (iii) shows that detachment within the TFR, while perhaps possible, does not enable wh-extraction from that position. But if the detachment process is taken to be part of the matrix we expect the linearization of the scion to place the material to the right of the callus between the callus and the detached PP. This is exactly what happens, as (iv) demonstrates. (iii) (iv)
*Waari heeft Jan [wat je als [de conceptversie] [van ei] kunt beschouwen] afgewezen ?Waar heeft Jan [wat je als [de conceptversie] kunt beschouwen] [van e ] afgewezen i i
13. This is modulo case syncretism. If the masculine noun Wagen in the text example is replaced by the neuter noun Auto, the nominative and accusative forms are identical, and thereby all four variants become grammatical. See Van Riemsdijk (2000b) for more discussion. A reviewer suggests that case matching has been shown to be a frequent property of regular FRs, as shown in considerable detail in Groos and Van Riemsdijk (1981) for German and a number of other languages. This, according to the reviewer, suggests that TFRs may be regular FRs after all. Note, however, that case matching in FRs concerns the case on the wh-word, whereas the case matching observed in TFRs concerns the predicate nominal. One could object to that, however, that in TFRs the wh-word and the predicate nominal must agree in case anyway since they stand in a secondary predication relation to each other. This is where case syncretism comes in, though. It was shown in Groos and Van Riemsdijk (1981) that in regular FRs the formal case features need not match precisely in those cases where the surface form is nevertheless identical. The only relevant case in German being the neuter singular was, which is both nominative and accusative. In the case of TFRs, however, it is the predicate nominal itself which is sensitive to case syncretism, as shown in detail in Van Riemsdijk (2000b). The relevance of this distinction is immediately visible in the text example (30b). Here the wh-word is was, which can pass both for a nominative and for an accusative. The corresponding predicate nominals are morphologically distinct, however, and hence a case conflict ensues, regardless of which form is chosen. Case matching effects on the predicate nominal therefore stand out as a unique property of TFRs. 14. I would not be surprised to see an attempt at just such an analysis in a general head raising framework of the type suggested by Kayne (1994). See also Bianchi (1999). 15. Wilder (1998, 1999) argues, in fact, that this is a necessary condition and adduces this fact in support of his backward deletion theory in view of the fact that the right node raising (RNR) construction, if analyzed in terms of deletion, is subject to the same condition. I have
39
40
Henk van Riemsdijk
argued in Van Riemsdijk (1998a) that the argument is circular, since RNR is also a prime candidate for an analysis in terms of grafts, that the argument is empirically incorrect since even in English the shared constituent can be internal, as in there was [what appeared to be a totem pole to most of the bystanders] standing in our front yard, that an internal shared constituent is even the normal case in German and Dutch, see the second point in the text, and that in addition to right peripheral and internal shared constituents left peripheral shared constituents exist as well, one example being regular free relatives of the matching type, cf. also Citko (1998).
References Bianchi, Valentina 1999 Consequences of Antisymmetry: Headed Relative Clauses. Berlin: Mouton de Gruyter. Citko, Barbara 1998 “ATB Analysis of Free Relative Clauses”. In Proceedings from the Main Session of the 34th Annual Meeting of the Chicago Linguistic Society, M.C. Gruber, D. Higgins, K. Olson, and T. Wysocki (eds), 69–81. University of Chicago: The Chicago Linguistic Society. Emonds, Joseph E. 1976 A Transformational Approach to English Syntax. New York: Academic Press. 1985 A Unified Theory of Syntactic Categories. Dordrecht: Foris. 1999 “How clitics license null phrases”. In Clitics in the Languages of Europe, H.C. van Riemsdijk (ed.), 291–367. Berlin: Mouton de Gruyter. Grimshaw, Jane 1991 Extended Projection. Manuscript. Rutgers University. Groos, Anneke and Hendrik C. van Riemsdijk 1981 “Matching Effects in Free Relatives: A Parameter of Core Grammar”. In Theory of Markedness in Generative Grammar, A. Belletti, L. Brandi and L. Rizzi (eds), 171–216. Pisa: Scuola Normale Superiore. Horn, Laurence R. 1988 “Pragmatic Theory”. In Linguistic Theory: Foundations [Linguistics: The Cambridge Survey Volume I], F. Newmeyer (ed.), 113–145. Cambridge: Cambridge University Press. Kayne, Richard S. 1994 The Antisymmetry of Syntax. Cambridge: MIT Press. Kajita, Masaru 1977 “Towards a Dynamic Model of Syntax”. Studies in English Linguistics 5: 44–66. Kiefer, Ferenc 1993 “Où s’arrête la syntaxe?”. In De la musique à la linguistique. Hommages à Nicolas Ruwet, L.Tasmowsky and A. Zribi-Hertz (eds), 313–326. Ghent: Communication & Cognition.
A far from simple matter
Kubo, Miori 1996 “Some Considerations on Noun Classes and Numeral Classifiers: A Study of (Pseudo-)partitives in Japanese and English”. In Keio Studies in Theoretical Linguistics I, Y. Nishiyama and Y.Otsu (eds), 89–124. Tokyo: Institute of Cultural and Linguistic Studies, Keio University. Lakoff, George 1974 “Syntactic amalgams”. In Proceedings of the Chicago Linguistic Society Meeting (CLS 10), M. W. La Galy, R. A. Fox and A. Bruck (eds), 321–344. Lasersohn, Peter 1999 “Pragmatic Halos”. Language 75: 522–551. McCawley, James D. 1988 The Major Syntactic Phenomena of English. Chicago: Chicago University Press. Moltmann, Friederike 1992 Coordination and Comparatives. [Distributed by MIT Working Papers in Linguistics]. Nakau, Minoru 1971 “The Grammar of the Pseudo-Free Relative Pronoun What” English Linguistics 6: 2–47. Riemsdijk, Hendrik C. van 1993 “Complements, Adjuncts and Adjacency”. In De la musique à la linguistique. Hommages à Nicolas Ruwet, L.Tasmowsky and A. Zribi-Hertz (eds), 498–512. Ghent: Communication & Cognition. 1997 “Push Chains and Drag Chains: Complex Predicate Split in Dutch”. In Scrambling, S. Tonoike (ed.), 7–33. Tokyo: Kurosio. 1998a “Trees and Scions, Science and Trees” Chomsky 70th Birthday Celebration Fest-Web-Page. URL: http://mitpress.mit.edu/chomskydisc/riemsdyk.html. 1998b “Categorial Feature Magnetism: The Endocentricity and Distribution of Projections”. The Journal of Comparative Germanic Linguistics 2: 1–48. 2000a “Binnenstebuitenbouw: enkele observaties over de transparante vrije relatiefzin”. In Met taal om de tuin geleid. Huldealbum voor Georges de Schutter, St. Gillis, J. Nuyts and J. Taeldeman (eds). Antwerpen: Universitaire Instellingen. 2000b “Free Relatives Inside Out: Transparent Free Relatives as Grafts”. In Proceedings of the 1999 PASE Conference, B. Rozwadowska (ed.), 223–233. University of Wrocław. Vos, Riet 1999 A Grammar of Partitive Constructions. Ph.D. Dissertation, Tilburg University. Wilder, Chris 1998 “Transparent Free Felatives”. ZAS Papers in Linguistics 10: 191–199. 1999 “Transparent Free Relatives”. In Proceedings of the Seventeenth West Coast Conference on Formal Linguistics, K.N. Shahin, S. Blake and E-S. Kim (eds). CSLI. Cambridge: Cambridge University Press. Williams, Edwin W. 1982 “Another Argument that Passive is Transformational”. Linguistic Inquiry 13: 160–163.
41
AUTHOR "Wolfgang U. Dressler and Lavinia Merlini Barbaresi"
TITLE "Morphopragmatics of diminutives and augmentatives"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Morphopragmatics of diminutives and augmentatives On the priority of pragmatics over semantics Wolfgang U. Dressler and Lavinia Merlini Barbaresi Institut für Sprachwissenschaft der Universität Wien / Dipartimento di Anglistica dell’Università di Pisa
1.
A model of morphopragmatics
Based on previous work by Dressler and Merlini Barbaresi (1987) and on cooperation with the destinee of this festschrift (Dressler and Kiefer 1990), a model of morphologized pragmatics, termed morphopragmatics (Dressler and Merlini Barbaresi 1994, 1999), has been elaborated, which covers the area of the general pragmatic meanings of morphological rules. A morphological rule has a morphopragmatic meaning if it contains a pragmatic variable which is necessary within the description of its meaning. This implies that its basic pragmatic meaning(s) cannot be reduced to a semantic meaning. The main field of application has been the pragmatics of diminutives and augmentatives. This is also the case for this contribution, intended to defend pragmatic priority against recent proposals for a semantic basis of the meaning of diminutives and augmentatives. Concretely contrasting an implicit opponent will make our argument line more perspicuous. In general, we oppose the previous communis opinio whereby all pragmatic meanings of diminutives and augmentatives would derive from ‘small’ and ‘big’ respectively, or from some other morphosemantic meaning (e.g. Ettinger 1974; Volek 1987; Wierzbicka 1991, cf. also the critique in Jurafsky 1996: 537f). Our approach is in line with our more general theoretical claim that pragmatics is a superordinate of semantics. In addition to the semantic features [small] of diminutives and [big] of augmentatives, we have proposed a still more basic, common pragmatic feature [fictive] for both (specified as [non-serious] for
44
Wolfgang U. Dressler and Lavinia Merlini Barbaresi
diminutives). This feature, we have claimed, is responsible for the majority of the pragmatic uses of diminutives (and augmentatives) in Italian and German and in the many other languages that we have investigated. Thus in the following two Italian variants of an apology both the diminutive and the augmentative have the function of downgrading the speaker’s liability (Dressler and Merlini Barbaresi 1994: 441f). (1) Senti, scusami per quelle sciocchezz-ine/sciocchezz-one che listen forgive.me for these silly things-dim/s.t.-aug that ti ho detto, non le pensavo to.you I’ve said not them I.thought ‘Listen, forgive me for that nonsense I said to you. I didn’t mean it.’
Within the derivational class of alteratives or evaluatives, diminutives represent the unmarked category, augmentatives the marked category. As expected, many languages (e.g. English, German) have only diminutives. Due to the priority of the pragmatic meaning [fictive], the use of diminutives in these languages may correspond to some uses of augmentatives in others (e.g. Italian) to such an extent that the best translation of an Italian or Spanish augmentative may be an English or German diminutive, as in (2). (2) a.
biond-one blond-aug ‘blond-ie’ b. angel-ote angel-aug ‘sweet-ie’
(Italian)
(Spanish, Napoli and Reynolds 1995: 170)
Diminutives and augmentatives are evaluative (cf. Corbin et al. 1999), i.e. they express an evaluation ‘as to value’ (not ‘as to fact’), “according to the evaluator’s intentions, perspectives and standards of evaluation” (Dressler and Merlini Barbaresi 1994: 153). Since these are subjective, a purely semantic representation of the evaluative character of diminutives and augmentatives leads to unacceptable reductionism. This particularly holds for emotive and attitudinal aspects of evaluation (cf. also Pupier 1998; Rio-Torto 1999; Moñino and Ortiz 1999; Lorente et al. 1999). For example, the Italian diminutive ripos-ino is responsible for the ironic empathy conveyed by the whole utterance:
Morphopragmatics of diminutives and augmentatives
(3) Non c’è giorno che non si faccia il suo not there.is day that not himself makes the his ripos-ino di due ore nap-dim of two hours ‘There is no day that he would miss his nice two-hour nap.’
Our approach has been extended to the sociopragmatics of Italian diminutives by De Marco (1998), to translation studies by Biscetti (1998) and Costa (1997) and to the investigation of the first-language acquisition of diminutives and augmentatives (cf. Section 4).
2. Jurafsky Specifically we want to oppose the long and important article by Jurafsky (1996). Based on a large cross-linguistic documentation, he has developed a cognitivist model of diminutives. The central prototype among diminutives is identified as the (semantic) meaning ‘child’, from which all other meanings and uses are semantically or pragmatically derived. This radial, polysemous category of diminutives (cf. Lakoff 1987) models one great continuum which stretches from cognition to semantics and pragmatics. Choosing an uninterrupted, borderless continuum instead of drawing (at least for analytic reasons) boundaries between cognition, linguistic semantics and pragmatics presupposes a theoretical axiom and a methodological option of fundamental importance, impossible to discuss in a short contribution. Let us just repeat our view (Dressler and Merlini Barbaresi 1994: 4ff) that (1) pragmatics is cognitively founded, but that this cognitive foundation and the pragmatic superstructure based on it should be separated, and that (2) semantics is included within pragmatics, but can be separated from it by canceling all pragmatic variables. Another very disputable theoretical choice by Jurafsky (1996: 536, 537, 542) is that of neglecting the Saussurean boundary between synchrony and diachrony to such an extent that synchrony reduces to fossilized diachrony, in concordance with the author’s intention to present an ‘archaeology of meaning’. Again, one might object that the author has only ignored an analytical boundary, but this option has negative consequences that we will have to criticize (although we do not follow the Saussurean extreme of negating any relation between synchrony and diachrony, e.g. when using micro-diachronic ‘external’ evidence for synchronic analyses).
45
46
Wolfgang U. Dressler and Lavinia Merlini Barbaresi
Important assets of Jurafsky’s contribution are his universalist approach and his endeavors to make strong claims about the relation between the universal radial category and its language-specific instantiations (p. 543): A.“The diminutive in particular languages can be described as instantiating coherent portions of” the universal category, and this “coherent subset is a connected graph that includes the prototype” (the prototype being ‘child’ in the case of the diminutive category, cf. p. 561).
On p. 561, claim A is watered down to: B.“Every diminutive in our database has either the ‘child’ or ‘small’ sense” (both being ‘central senses’).
Further weakening of claim B is performed on the same page: C.“Falsification of the hypothesis is possible by “morphemes that exhibit many of the senses of the diminutive, but can be shown not to have or have had either the sense ‘child’ or ‘small’”
Taken literally, claim C is unfalsifiable because the prehistory of a diminutive suffix may go back to unattested protolanguages where its meaning cannot be ascertained. In addition to this problem, the vast majority of semanticists and pragmaticists will not accept, for a given language, a central meaning of a category which is only prehistorical, due to the above-mentioned refusal to mix synchrony with diachrony, particularly with macro-diachrony. In trying to falsify these claims, let us start with English examples, first with the female suffix -ette, cited as diminutive by Jurafsky (1996: 545, 545), and with the diminutive suffix -let. Both can mean ‘small’, but neither means ‘child’ semantically or pragmatically, i.e. they are not adequate for child-centered speech situations, not even for hypocoristics. Diachronically, -ette is a loan from French, -let may be due to a reanalysis of the same suffix (Marchand 1969: 326). This falsifies claim A, but not the second parts of the weakened claims B and C. Moreover, smallness has never been the most salient or most frequent use of the suffix -ette, a lack of two commonly assumed properties of a prototype — but Jurafsky does not assume either of these two properties for his own concept of prototype. In marked contrast to these two suffixes, the English suffix -y/-ie has a central meaning “child” in its being typically used for child-centered speech situations or speech situations derived from them (cf. Dressler and Merlini Barbaresi 1994: 193–196), including hypocoristics addressed to children, but it hardly has a meaning of smallness outside these speech situations. An example is:
Morphopragmatics of diminutives and augmentatives
(4) Was it something about not having dark-ies on as conductors?
Despite the lack of a purely semantic meaning of smallness, the suffix -y/-ie may signal contempt, as in the above example and in: (5) I know a rough-ie when I see one. He is just one of those blokes who can’t stay away from trouble.
Since the meaning of contempt is derived by Jurafsky (1996: 542, 547 Table 8) from the meaning of smallness, this meaning distribution falsifies his concept of “coherent portion” in his claim A because in the ordered set child – small – contempt, the meanings ‘child’ and ‘contempt’ do not form a coherent (= continuous) subset, but a discontinuous one. Moreover, since this distribution also appears to have been true for earlier periods (cf. Marchand 1969: 298), not even an earlier meaning bridge ‘small’ would have established coherence (continuity). Similar counter-examples can be found in other languages as well: French -et (masc.), -ette (fem.), according to recent studies by Dal (1997, 1999), Fradin (1999), does not have a semantic or pragmatic meaning ‘child’, and ‘small’, in its literal semantic meaning, is rather marginal. Of course, it appears in hypocoristics, such as Mariette (¨ Marie), but it is (pace Jurafsky 1996: 564) not at all clear that hypocoristics presuppose a meaning ‘child’, unless diminutive formation exhibits other characteristics of a child-centered speech situation, namely extragrammatical morphological operations, such as reduplication (or echo-word formation) and irregular abbreviation (truncation), as in Elizabeth Æ Liz, Bet(t-y), Liz(zy)-Wiz(zy), etc. (cf. Dressler and Merlini Barbaresi 1994:36ff, 105; Dressler 1994). As to its diachronic origin, Fr. -et, It. -etto, Sp. -ito are usually derived from a borrowed Celtic onomastic suffix -ittus (also accepted by Jurafsky 1996: 564), without any trace of an original meaning ‘child’. Or the cognate suffixes, E. -ish and G. -isch, accepted as diminutives by Jurafsky (1996: 565), with relational, pejorative, approximative meanings, have no clear meaning of small and derive diachronically from a relational -isk suffix (cf. Pounder 1999), which, at the time of its formation, had no meaning of ‘child’ or ‘small’ either. Similarly, the diminutive suffixes It. -ino, Sp. -in, Fr. -in go back to a Latin suffix -inus, with relational or appropriation meaning (cf. Magni 1999: 145f; Mutz 1999: 165). Both developments defy Jurafsky’s (1996: 560ff) claims of unidirectionality of diachronic change from central senses (i.e. ‘child’ and ‘small’ for diminutives) to peripheral senses (e.g. ‘related-to’ = relational,
47
48
Wolfgang U. Dressler and Lavinia Merlini Barbaresi
‘approximation’, ‘contempt’ = pejorative). The same argument can be made for the Indo-European diminutive suffix *-ko, for which Jurafsky (1996: 565–569) proposes to reconstruct an original meaning ‘child’. But his proposal, based on rapid inspection of grammars, dictionaries and secondary literature, is not consistent with the solid philological, diachronic and comparative work that supports Magni’s (1999) reconstruction of an original meaning for the same suffix of ‘belonging/membership, possession, resemblance’ (and similar earlier reconstructions). In connection with diachrony, we have finally to mention that Jurafsky does not provide any independent cognitive evidence for the cognitive centrality of ‘child’ and the central derivations from it, particularly from ‘child’ to ‘small’. This is a frequent defect of the cognitivist school Jurafsky adheres to, which reduces to an argumentation of the type: (1) axiom: everything in language is an emanation of the central cognitive system, (2) heuristics: if I find similar phenomena in many languages, then they must have a common cognitive foundation, (3) heuristics: I have to construct a plausible cognitive generalization, (4) empirical evidence: this construction must be applicable to further (if possible, to all other) languages. Such arguments smack of circularity that can only be overcome if independent evidence from cognitive psychology and related psycholinguistics is supplied. But, to Jurafsky’s credit, he does supply one type of independent, albeit only linguistic, evidence, i.e. (5) diachronic evidence for the origin of diminutive affixes in words for ‘child’ (p. 562). Diachronic origin, however, does not equal cognitive origin or base. And he does not supply any example of the origin of diminutive suffixes in adjectives meaning ‘small’. Of course, E. little and Fr. petit have developed some pragmatic meanings shared by diminutives (cf. Dressler and Merlini Barbaresi 1994: 114), but this is a far cry from a change into proper diminutives. Furthermore, Jurafsky views this diachronic evolution as a sort of grammaticalization process. This subsumption has a potential flaw in that diachronic grammaticalization finally results in prototypical inflectional or derivational affixes (possibly before further lexicalization), whereas diminutives are non-prototypical morphological elements which do not further develop into prototypical ones (cf. Dressler and Merlini Barbaresi 1994: 55, Dressler 1994). Finally, Jurafsky’s model faces many unresolved problems, e.g. not accounting for the intimately connected category of augmentatives: what is the node/ prototype of augmentatives? How can Jurafsky’s approach explain the pragmatic similarity between diminutives and augmentatives, which we (Dressler and Merlini Barbaresi 1994: 441f, 447f, 461, 472, etc.) have explained with the
Morphopragmatics of diminutives and augmentatives
feature [fictive] common to both? This is especially important for hedging (dealt with by Jurafsky 1996: 556–561 only for diminutives), as in Dressler and Merlini Barbaresi (1994: 457): (6) Ho da chiederti un piacer-ino/piacer-one I.have to ask.you a favor-dim/favor-aug ‘I’ve got to ask you a favor’
Moreover, if diminutives basically mean ‘child’ and ‘small’, whereas augmentatives basically mean ‘big’ or ‘intense’ (cf. Lorente et al. 1999) and possibly ‘mother’ (cf. Jurafsky 1996: 546f; Creissels 1999), then why are the many actual and legal combinations of diminutives and augmentatives, as in It. pezz-ett-one ‘piece-dim-aug’ (Dressler and Merlini Barbaresi 1994: 447), not contradictory and thus illegal? (Jurafsky 1996: 547, 549, 556 n.3 treats augmentatives as opposed to diminutives). Very many cases of such combinations are recorded in Alberti et al. (1991), e.g. fiasch-ett-one ‘bottle-dim-aug’, trav-icell-one ‘beamdim-aug’, pazz-erell-one ‘crazy-dim-aug’, ghiott-on-cino ‘glutton-aug-dim’. Jurafsky (1996: 571) briefly mentions the relation of diminutives to augmentatives and pejoratives as an unresolved issue. But there is still another issue, related to all three categories: how can his model accommodate evaluation? (cf. Section 1). Note that Sörés (1999) considers the distinction between diminutives and augmentatives to be inapplicable to the category of evaluative verbs. Our critique must be counterbalanced by referring to interesting issues raised by Jurafsky (1996). His notion of ‘coherent subset/portion’ (discussed in Section 3) should exclude derivational gaps. Thus if one assumes Jurafsky’s (1996: 542) derivational chain “small Æ small type of Æ imitation”, then no diminutives of any language which have the meanings ‘small’ and ‘imitation’, should exclude the meaning ‘small type of ’. Such derivational chains should be made explicit and subjected to Jurafsky’s analyses in terms of metaphors, inferences, generalizations and lambda-abstractions on the one hand. On the other hand, Jurafsky’s skeleton of derivational chains (p. 542) should be expanded. Then the criterion of derivational gaps in coherent portions could help to decide among alternative analyses. The synchronic concept of derivational chains can be enriched, as Jurafsky has shown, by the diachronic aspect. But this would need detailed philological work, because mere reconstruction necessarily remains skeletal. Where we also disagree with Jurafsky, are his generous assumptions of unidirectionality in change.
49
50
Wolfgang U. Dressler and Lavinia Merlini Barbaresi
3. In defense of our model Jurafsky (1996: 563) explicitly criticizes our model for three reasons which have to do with our basic diminutive-feature [non-serious]: 1.Our “account has difficulties explaining why so many uses of the diminutive involve children”. However, we have insisted (Dressler and Merlini Barbaresi 1994: 147) that adult humans treat children (particularly small ones) as non-serious partners in interactions, which explains why the feature [nonserious] applies so well and so often to small children. The same holds for very senile people, who are also often addressed with diminutives (cf. p. 190). This ‘diminutivum senile’ is much less obvious than the well-known ‘diminutivum puerile’ because children play a much more central role than senile people. Another reason lies in phenomena of acquisition, see Section 4. Also note that the clearly child-centered Austrian and Hungarian suffix -i does not extend beyond the sphere of child-centered speech, unless in the direct metaphors of pet- or lover-centered speech and in referring to evidently childish behavior of adults (cf. p. 191). In contrast, all the other diminutives (to which Jurafsky assigns the same semantic prototype ‘child’) have no such limitation. Finally, also Italian augmentatives are overused in child-centered speech situations (p. 439) which, according to Jurafsky, are opposed to diminutives and have nothing to do with the prototype ‘child’. 2.“‘Nonserious’ could just as easily refer to non-child recreation; this is not commonly the case”. Actually it is the case: ‘non-childish’ diminutives are commonly used in what we have called homiletic speech (p.87), especially when intimacy, familiarity, sympathy, empathy, jocularity, etc. are involved (cf. p. 197). Note that in such speech situations no native speaker associates a mental image of children metaphorically to the use of diminutives. 3.“It is not clear why ‘affection’, obviously linked with children, should be a necessary extension to ‘nonserious’”. First of all, affection is not at all obviously linked with children, as the psychological and psycholinguistic literature on affection cited in our book shows. Moreover, most of the cases of affective use discussed in our book have nothing to do with children. Second, why should there be any necessary extension of meaning at all? All synchronic derivations (i.e. meaning extensions) assumed both by us and by Jurafsky are optional. Third, we have derived (p. 34) the affective uses of both diminutives and augmentatives from the evaluative character of both categories. From this derives the fourth counterargument: if affective use were derived from children,
Morphopragmatics of diminutives and augmentatives
and if augmentatives are opposed to diminutives and have nothing to do with children, from where should one derive the affective use of augmentatives? Our more basic pragmatic feature [fictive] which applies to both diminutives and augmentatives, can more easily explain the common behavior of these antonymic suffixes.
4. First-language acquisition Independent evidence for some of our main claims can be found in language acquisition, a traditional field of substantial (or external) evidence for linguistic theory since Jan Baudouin de Courtenay, Paul Passy and Roman Jakobson. The traditional claim that diminutives are the unmarked and augmentatives the marked category in their mutual relation, plus the general view that unmarked categories are acquired before corresponding marked categories, allow the prediction that children should acquire the diminutive category before the augmentative category, if the respective language has both categories. This prediction is born out in all acquisitional studies we know: Italian (cf. Dressler 1994; Ceccherini et al. 1997; De Marco 1998), Greek (cf. Stephany 1997), Polish (cf. Dressler 1994). More important is our claim about the priority of the pragmatic meaning over the semantic meaning of diminutives and augmentatives. From this claim we may derive the expectation that some basic pragmatic meanings should emerge before the semantic meaning [small] (or any other meaning derived from it). The first relevant result of first-language acquisition studies is that first, for a long time, diminutives are used without any meaning of smallness, i.e. without any discernable semantic meaning, very often in linguistic and situational contexts where a meaning of smallness is excluded (e.g. contradicted). This is true for Italian (Ceccherini et al. 1997; De Marco 1998), Greek (Stephany 1997), Russian (Voeykova 1998; Protassova 1999: 23), Polish (Dressler 1994), Czech (Dressler 1994), Lithuanian (Savickiene 1998), Dutch (Gillis 1997), German (Dressler 1994). For example, Italian diminutives appear in Sara’s speech at the age of 1;6 (Ceccherini et al. 1997) and in Matteo’s speech at the age of 1;4, productive with 1;8 (De Marco 1998). But the first appearance of the semantic meaning ‘small’ occurs in Sara’s speech only at 2;0.20:
51
52
Wolfgang U. Dressler and Lavinia Merlini Barbaresi
(7) guarda un micio piccolo, un micio, (corrected:) un mic-ino look a kitten small a kitten a kitten-dim ‘Look a small kitten, a kitten, (no) a small kitten’
Here she realizes that a ‘small kitten’ can be substituted by a ‘kitten-dim’. In languages which have augmentatives besides diminutives, the semantic meaning of smallness is acquired at the latest when augmentatives are acquired. An example is the contrast between ser-ino (¨ seder-ino) and ser-one (¨ sederone) made by a three-year old Italian girl who indicated her own bottom (It. euphemistic sedere) with a diminutive, in contrast to her aunt’s, indicated with an augmentative (cf. Dressler and Merlini Barbaresi 1994: 446). This connection between the acquisition of augmentatives and the acquisition of the basic semantic meaning of diminutives is easily explained by the following assumptions: (1) Children acquire first a basic pragmatic meaning of diminutives, identical or connected with the pragmatic feature of fictiveness. (2) When they start to acquire augmentatives, they acquire them with the same or a very similar pragmatic meaning. (3) If children acquire two categories or items with a shared meaning, they try to differentiate them in meaning, due to the principle of contrast whose importance for children has been well established by Clark (1993). And since the basic meaning difference between diminutives and augmentatives is the semantic distinction between small and big size on whatever semantic dimension, children are induced to acquire this denotative difference. Now what meaning do children assign to diminutives before they acquire the semantic meaning of smallness? Since, most of the time, it is very difficult for the observer to ascertain any precise meaning of a given diminutive that a child uses alongside its simplex base, one might hypothesize that children first use diminutives as synonymous substitutes for their simplicia. However this would violate Clark’s (1993) principle of contrast (cf. Dressler 1997b). Thus we can make the following deduction: (1) adults have both a pragmatic and a semantic meaning of diminutives, (2) children first have no semantic meaning, but already use diminutives alongside their simplex bases, (3) thus they should have a separate meaning for diminutives which is not the semantic meaning [small]. Now we have the choice that this is a meaning either (a) identical with an adult meaning, or (b) different. If it is identical (a), then only a pragmatic meaning is left. If it is different (b), then what should it be? No reasonable proposal has been made as yet. In any case, option (a) is simpler, and the burden of proof is on those who propose a different meaning (b).
Morphopragmatics of diminutives and augmentatives
Moreover, there are positive indications that the pragmatic meaning of diminutives has emerged earlier than the semantic meaning. This is the case with endearment, both in hypocoristics and diminutives (Savickiene 1998) and with emotive meanings (Protassova 1999: 22). Further evidence comes from the fact that early diminutives are especially concentrated in areas of use where pragmatic meanings are most likely, namely body parts, pet animals, food, family members, etc. (Stephany 1997; Voeykova 1998; Savickiene 1998), cf. also the use in appropriation of children (Protassova 1999: 23). Finally, there is no case in the literature on the acquisition of diminutives where the relation between presence and absence of pragmatic meaning would be reversed. In other words, children at these early ages, where they have not yet assigned the semantic meaning of smallness to diminutives, have never been observed using a diminutive with a simplex meaning and simultaneously the corresponding simplex with a pragmatic meaning of a diminutive. Derived pragmatic meanings have been observed only later, e.g. the use of diminutives as a cajoling device in Matteo’s speech at 3;6 (De Marco 1998): (8) Puoi darmi quella pall-ina un attim-ino soltant-ino? can.you give.me that ball-dim a moment-dim only-dim ‘Can you give me that nice ball just for a little moment?’
cf. Protassova (1999: 21f). 5. Diachrony Further substantial (external) evidence for the priority of pragmatics over semantics in the meaning of diminutives and augmentatives comes from diachronic change. The first argument refers to the fact that the same Latin or Proto-Romance suffixes have resulted in diminutives in some Romance languages, in augmentatives in others (cf. Dressler and Merlini Barbaresi 1994: 488). For example, the suffix It. -one, Sp. -ón forms augmentatives (in Portuguese and Romanian as well), but diminutives in French and Catalan, augmentatives and diminutives in complementary distribution in some Spanish regiolects (cf. Rainer 1993: 637). Since the Latin precedent Nom. -o, Gen. -onis signals excessive intensification this implies that augmentative meanings are the earlier ones. Now, the semantic meanings of diminutives and augmentatives are opposed, whereas the pragmatic meanings are partially very similar. This indicates that diachronic change will not have consisted primarily in a jump from [big] to [small], but in a gradual shift of pragmatic meanings.
53
54
Wolfgang U. Dressler and Lavinia Merlini Barbaresi
Similarly, French (Gallo-Romance) -ot forms diminutives, whereas Sp. -ote forms augmentatives. It. -otto is a diminutive, but of rather limited semantic diminution (cf. Dressler and Merlini Barbaresi 1994: 435). Even if this may be interpreted as a gradual semantic change, stability of the main pragmatic meaning can be observed. Also Jurafsky’s (1996) assumption of a first basic meaning of ‘child’ is not compatible with such diachronic changes. For, what corresponding meaning of augmentatives could explain such a change? The choice of Jurafsky (1996: 546f, 549, 556 n.3) seems to point to ‘mother’ (cf. Creissels 1999). But how can the meaning ‘child’ change to ‘mother’ and vice versa? The solution of replacing the augmentative meaning ‘big’ with ‘intense’ (cf. Lorente et al. 1999) cannot be of any help, nor can metaphors, such as ‘small is beautiful/nice’ (Tabakowska 1993) explain such diachronic transitions. Since diminutives are, as is accepted by everybody, important for childcentered speech, much reinterpretation of meanings is liable to take place in early stages of first-language acquisition (cf. Dressler 1997a). If these reinterpretations are not unlearned in later stages of the ontogenetic language development, diachronic change will result. Thus we may assume the following scenario for the origin of diminutive meanings of the Lat. suffix -inus in its development to It. -ino, Fr. -in, etc. The first attested and still existing meaning has been relational. This meaning was extended, in Late Latin, to denote members of the family (including slaves living in the house). The prototypical subset of this new meaning was a patronymic one, e.g. Secund-inus as son of Secundus. When small children were addressed in that way they may have reinterpreted such names as hypocoristics in analogy to already existing hypocoristics used by and towards children, and this is due to the well-known iconic power of the most palatal vowel [i˜] (pace Bauer 1996; cf. Ohala et al. 1994) which is much more important for children than for adults (cf. Dressler 1994). At any rate, the assumption of unidirectional diachronic changes (as discussed in Section 3) cannot be simply established on inductive generalizations or on some high-level cognitive principle, but must be explained within a specific scenario of diachronic change, in the case of diminutives by referring to phenomena of early language acquisition.
Morphopragmatics of diminutives and augmentatives
6. Conclusion In this contribution we have discussed Jurafsky’s (1996) article, because it presents the only systematic recent treatment of the semantics of diminutives which is incompatible with our model. We have tried to give multiple evidence for the priority of pragmatics over semantics within the realm of diminutives and augmentatives: 1. cross-linguistic synchronic evidence for a basic pragmatic meaning, which cannot be reduced to or derived from a semantic meaning (Sections 2, 3); 2. evidence from early language acquisition, where pragmatic meaning precedes semantic meaning (Section 4); 3. diachronic evidence from meaning changes which are much easier to explain via pragmatic than via semantic meanings (Section 5). One may object that our argument is based uniquely on diminutives and augmentatives, a morphological class which does not belong to prototypical derivational morphology, i.e. it represents non-prototypical word formation, and is even marginal within the whole component or grammatical module of morphology. Diminutives (less so augmentatives), that is, often exhibit properties of extragrammatical operations. Such criticism would imply that one should instead seek evidence for the priority of pragmatics over semantics among rather prototypical representatives of morphology. We can counter this line of argument by referring to our assumption (cf. Dressler and Karpf 1995; Karmiloff-Smith 1992) of constructivist modularization in language acquisition. This concept implies that morphology and syntax emerge as submodules, or at least as relatively autonomous components of grammar, due to children’s constructivist self-organization in the process of language acquisition. But — what is important here — this concept also implies that semantics separates from pragmatics due to a constructivist abstraction of more specific semantic meanings from more general pragmatic meanings. Particularly diminutives offer the possibility of highlighting this critical process of differentiation.
55
56
Wolfgang U. Dressler and Lavinia Merlini Barbaresi
References Alberti, Claudia, Ruimy, Nilda, Turrini, Giovanna and Zanchi, Giampero 1991 La donzellétta vien dalla donzèlla: Dizionario delle forme alterate della lingua italiana. Bologna: Zanichelli. Bauer, Laurie 1996 “No phonetic iconicity in evaluative morphology”. Studia Linguistica 50: 189–204. Biscetti, Stefania 1998 Due sistemi morfologici a confronto. Il caso degli alterativi Gaddiani nella traduzione inglese. Thesis, University of Pisa. Ceccherini, Marco, Bonifactio, Serena and Zocconi, Elisabetta 1997 “Acquisition of diminutives in Italian (Sara)”. In Studies in Pre- and Protomorphology, W. U. Dressler (ed.), 165–172. Wien: Verlag der Österreichischen Akademie der Wissenschaften. Clark, Eve 1993 The Lexicon in Acquisition. Cambridge: Cambridge University Press. Corbin, Danielle et al. 1999 La morphologie des dérivés évaluatifs [Special issue Silexicales 2]. Costa, Marcella 1997 “Alterati accrescitivi e peggiorativi”. In Italiano e Tedesco: un confronto, S. Bosco Coletsos (ed.), 173–203. Alessandria: Dell’Orso. Creissels, Denis 1999 “Origine et évolution des diminutifs et augmentatifs dans quelques langues africaines”. Silexicales 2: 29–35. Dal, Georgette 1997 Grammaire du suffixe -et(te). Paris: Didier. 1999 “Suffixation par -et(te) et bases verbales”. Silexicales 2: 37–47. De Marco, Anna 1998 Sociopragmatica dei diminutivi in italiano. Rende: Centro editoriale Università della Calabria. Dressler, Wolfgang U. 1994 “Evidence from the first stages of morphology acquisition for linguistic theory”. Acta Linguistica Hafniensia 27: 91–108. 1997a “‘Scenario’ as a concept for the functional explanation of language change”. In Language Change and Functional Explanations, J. Gvozdanovic´ (ed.), 109–142. Berlin: Mouton de Gruyter. 1997b “Zur Entstehung der Distinktivität in der Kindersprache”. In Polyphonie pour Iván Fónagy, J. Perrot (ed.), 127–136. Paris: L’Harmattan. Dressler, Wolfgang U. and Annemarie Karpf 1995 “The theoretical relevance of pre- and protomorphology in language acquisition”. Yearbook of Morphology 1994: 99–122.
Morphopragmatics of diminutives and augmentatives
Dressler, Wolfgang U. and Kiefer, Ferenc 1990 “Austro-Hungarian morphopragmatics”. In Contemporary Morphology, W. U. Dressler et al. (eds), 69–77. Berlin: Mouton de Gruyter. Dressler, Wolfgang U. and Merlini Barbaresi, Lavinia 1987 Elements of Morphopragmatics. Duisburg: LAUD A 194. 1994 Morphopragmatics. Berlin: Mouton de Gruyter. 1999 “Morphopragmatics”. In Handbook of Pragmatics, J. Verschueren et al. (eds). Amsterdam: Benjamins. Ettinger, Stefan 1974 Diminutiv- und Augmentativbildungen: Regeln und Restriktionen. Tübingen: Narr. Fradin, Bernard 1999 “La suffixation en -et est-elle évaluative?”. Silexicales 2: 69–82. Gillis, Steven 1997 “The acquisition of diminutives in Dutch”. In Studies in Pre- and Protomorphology, W. U. Dressler (ed.), 165–179. Wien: Verlag der Österreichischen Akademie der Wissenschaften. Jurafsky, Daniel 1996 “Universal tendencies in the semantics of the diminutive”. Language 72: 533–578. Karmiloff-Smith, Annette 1992 Beyond Modularity: a developmental perspective on cognitive science. Cambridge, Mass.: MIT Press. Lakoff, George 1987 Women, Fire, and Dangerous Things. Chicago: University of Chicago Press. Lorente, Mercè, Adelstein, Andreina and Kuguel, Inès 1999 “La persistance du caractère évaluatif dans les lexicalisations: le cas du suffixe -azo en espagnol”. Silexicales 2: 127–137. Magni, Elisabetta 1999 “Sémantique du diminutif et relations d’appartenance”. Silexicales 2: 139–148. Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation. München: Beck. Moñino, Yves and Ortiz, Carolina 1999 “Réinterprétation de deux procédés de morphologie évaluative dans un créole de Colombie”. Silexicales 2: 253–261. Mutz, Katrin 1999 “Les dérivés évaluatifs de l’italien: comment la diachronie explique la synchronie”. Silexicales 2: 161–167. Napoli, Donna Jo and Reynolds, Bill 1995 “Evaluative affixes in Italian”. Yearbook of Morphology 1994: 151–178. Ohala, John, Nicols, J. and Hinton, L. 1994 Sound Symbolism. Cambridge: Cambridge University Press. Pounder, Amanda 1999 “Suffixation évaluative et ambiguité”. Silexicales 2: 189–196.
57
58
Wolfgang U. Dressler and Lavinia Merlini Barbaresi
Protassova, Ekaterina 1999 “Rol’ diminutivov v sovremennom russkom jazyke”. Manuscript. Moscow State University. Pupier, Paul 1998 “Une première systématique des évaluatifs en français”. Revue québécoise de linguistique 26: 51–78. Rainer, Franz 1993 Spanische Wortbildungslehre. Tübingen: Niemeyer. Rio-Torto, Graça M. 1999 “Aux limites de la dérivation: évaluatifs et z-évaluatifs”. Silexicales 2: 197–208. Savickiene, Ineta 1998 “The acquisition of diminutives in Lithuanian”. In Studies in the Acquisition of Number and Diminutive Marking [Antwerp Papers in Linguistics 95], S. Gillis (ed.), 115–135. Sörés, Anna 1999 “Comment un procès peut-il être évalué? Etude sémantique de quelques dérivatifs en hongrois”. Silexicales 2: 243–251. Stephany, Ursula 1997 “Diminutives in early child Greek, a preliminary investigation”. In Studies in Pre- and Protomorphology, W. U. Dressler (ed.), 147–156. Wien: Verlag der Österreichischen Akademie der Wissenschaften. Tabakowska, Elzbieta 1993 Cognitive linguistics and poetics of translation. Tübingen: Narr. Voeykova, Maria D. 1998 “Acquisition of diminutives by a Russian child”. In Studies in the Acquisition of Number and Diminutive Marking [Antwerp Papers in Linguistics 95], S. Gillis (ed.), 97–114. Volek, Bronislava 1987 Emotive Signs in Language and Semantic Functioning of Derived Nouns in Russian. Amsterdam: Benjamins. Wierzbicka, Anna 1991 Cross-Cultural Pragmatics: The Semantics of Human Interaction. Berlin: Mouton de Gruyter.
AUTHOR "Bernard Comrie"
TITLE "“Love your enemies”"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
“Love your enemies” Affective constructions in two Daghestanian languages* Bernard Comrie Max Planck Institute for Evolutionary Anthropology, Leipzig
1.
Introduction
Tsez and Bezhta are two closely related languages within the Tsezic branch of the Nakh-Daghestanian (Northeast Caucasian, East Caucasian) family.1 The NakhDaghestanian language family contains the branches shown in (1); further hierarchical structure within the family is possible, but remains controversial. (1) Nakh-Daghestanian Nakh (including Chechen) Avar-Andi (including Avar) Tsezic (Hunzib (Hz), Bezhta (B), Hinuq, Tsez (T), Khwarshi) Lak Dargi Lezgic (including Lezgian)
The list in (1) contains representative languages for some of the branches, but a complete language listing (subject to reconsideration in light of the language/ dialect boundary controversy) of the Tsezic languages. It should be noted that Bezhta and Hunzib form a subbranch, as do Tsez and Hinuq, with Khwarshi standing isolated within the branch. In order to understand the body of this article, some preliminary information relating to Tsez and Bezhta morphology and phonology is necessary. In both languages, (most) vowel-initial verbs agree in gender-number with their absolutive-case argument (for which see Section 2 below). The agreement prefixes are as shown in (2) for Tsez, (3) for Bezhta. The prefixes are clearly
60
Bernard Comrie
cognate between the two languages, with j being the regular Bezhta equivalent of Tsez r; genders II and IV are distinguishable in Bezhta not only because of their different plural prefixes, but also in irregular agreement forms found in pronouns. In the plural, the four-way distinction is reduced to a two-way opposition, with Tsez neutralizing genders II–IV, Bezhta genders I–II and genders III–IV. Each of the first three genders has a clear semantic core, but gender II in Tsez and gender III in Tsez and Bezhta also include a number of inanimate nouns. (2)
sg Ø jbr-
male human female human; some inanimate animals; some inanimate inanimate
pl brrr-
T
I II III IV
sg Ø jbj-
male human female human animals; some inanimate inanimate
pl bbjj-
B
I II III IV
(3)
It is also useful to note two morphophonemic generalizations that will help to follow the morpheme glosses in the examples that follow. These are set out in (4) and (5). (4) Vowel hiatus is avoided by dropping a vowel before another vowel (B,T). (5) Illegitimate consonant clusters (basically, violating CVC syllable structure) are broken up by epenthetic i (B), e (T).
2. Clause constructions in Tsez and Bezhta As generally in Nakh-Daghestanian languages, it is useful to distinguish a number of clause constructions, of which three cover most eventualities and are of direct relevance here. The behavior of the two languages, in terms of morphosyntactic encoding of semantic (“thematic”) roles, is identical (although it is always conceivable that further work will uncover distinctions between the languages, for instance in the details of argument structure). In the intransitive construction, the verb typically has a single argument (though some verbs may also have other oblique arguments). This single
“Love your enemies”
argument appears in the absolutive case, which is identical to the citation form of the noun. If the verb is vowel-initial and capable of agreeing, then it agrees with this absolutive argument, as in (6)–(7). (6) ijo j-ücˇä¯xe-jo. mother.abs II-sleep-pst ‘Mother slept.’
(B)
(7) is b-exu-s. bull.abs III-die-pstwit ‘The bull died.’
(T)
In the transitive construction, the verb typically has two arguments (though it may also have further oblique arguments). The most agent-like argument appears in the ergative case, the most patient-like argument in the absolutive case. If the verb is capable of showing agreement, then it agrees with the patient in the absolutive case, as in (8)–(9).2 (8) kib-ba anq’o b-oxi-jo. girl-erg mouse.abs III-catch-pst ‘The girl caught the mouse.’
(B)
(9) žek’-a¯ bišwa r-ac’-xo. man-erg food.abs IV-eat-prs ‘The man eats the food.’
(T)
The third relevant construction is the affective construction, sometimes also called the inverse construction, which is used primarily with psychological verbs in a broad sense, including verbs of emotion, perception, and cognition. In this construction, the experiencer argument appears in the lative case,3 the argument expressing the source of the experience in the absolutive case. If the verb can show agreement, then it agrees with the source argument in the absolutive case. Examples are shown in (10)–(11). (10) di-l kid j-ac-ca. me-lat girl.abs II-love-prs ‘I love the girl.’
The affective construction will be the main concern in what follows.
61
62
Bernard Comrie
3. Imperative formation in Tsez and Bezhta Since one of the crucial sets of data differentiating Tsez from Bezhta in terms of the affective construction concerns imperatives, it is necessary to say something about imperative formation in the two languages, so that forms can be reliably identified as imperatives. For both languages, a couple of completely irregular imperatives are left out of account. It is convenient to start with Bezhta, with has the simpler morphology in this respect. In Bezhta, a verb stem can end in a consonant or any of the vowels a (or its umlauted variant ä), o (or its umlauted variant ö), e, or i; stems ending in i are however very rare, and there are none ending in u. The stems are given in the lefthand column of (12), and throughout stems that take a gender prefix are marked by means of a leading hyphen; with regard to actually occurring verb forms, stems can be seen in the so-called present-I forms in the middle column. If the verb is transitive, in the imperative the suffix -a is added. If the verb is intransitive (or affective), then there is no suffix in the imperative, which is thus identical to the stem.4 Bezhta and Tsez, unlike the general Nakh-Daghestanian pattern, in general lack labile verbs, i.e. verbs that can be used both transitively and intransitively. I am aware of no labile verb in Tsez. In Bezhta, I am aware of just one, -onq’o- ‘come; bring’, and as indicated in the last line of (12) it has, predictably, two imperative forms depending on transitivity. (12) Stem Present-I Imperative Intransitive (and Affective) verbs -e>-e>-ca -e> -ohda-ohda-s -ohda Qü‘†-ö Qü‘†-ö-s Qü‘†-ö -en†’e-š -en†’e -en†’e-ixQi-ixQi-s -ixQi Transitive verbs: git’git’-ca -ansa-ansa-s xu†-oxu†-o-s n -e he-enhe-š -ixk’i-ixk’i-s Note -onq’o-
-onq’o-s
(B) ‘fight’ ‘work’ ‘wake up’ ‘go’ ‘get warm’
git’-a -ans-a xu†-a -enh-a -ixk’-a
‘pour’ ‘weigh’ ‘drink’ ‘open’ ‘heat’
-onq’o -onq’-a
‘come’ ‘bring’
“Love your enemies”
Tsez has fewer stem types, since verb stems can only end in a consonant or the vowels i and u; the verb stem can be found by deleting the present suffix -x(o).5 Intransitive verbs normally take no ending in the imperative, as illustrated in the examples in (13). However, a handful of intransitive verbs — three are known to me in the speech of Arsen Abdulaev — take the suffix-o, otherwise reserved for transitive verbs; for k’o†i- ‘jump’, this was already noted by Imnajšvili (1963: 191). The exceptional intransitive verbs in question are all “unergative”, i.e. have an agent-like single argument, but it should be noted that most semantically unergative verbs do have a zero ending in the imperative, and that even these irregular verbs behave in other respects like intransitive verbs, in particular with their single argument in the absolutive case. Transitive nonderived verbs take the suffix -o in the imperative; I am aware of one exception, the verb xeci- ‘leave’, which quite anomalously truncates the final vowel to form its imperative. Derived transitive verbs take no ending in the imperative, and thus have the same imperative formation as intransitive verbs. In (13) this is illustrated with -exu-r- ‘kill’, the causative of -exu- ‘die’ and having the imperative -exu-r. Indeed, causative verbs are the only unequivocal instances of derived verbs that have imperatives in Tsez. There are some verbs whose status as derived or underived seems to be dubious, and here there is dialectal, perhaps even idiolectal, variation. Thus, it is not clear whether -egir‘send’ is synchronically to be analyzed as a causative, and there are different imperative forms attested that correspond to different solutions. The verbs te†‘give (to third person)’ and ne†- ‘give (to first or second person)’ could be analyzed as a root -e†- with a deictic prefix, although this would not be a productive pattern, and again imperative variants correlate with the derived versus nonderived analysis. Finally, verbs that in Tsez would otherwise end in a forbidden stem-vowel take a stem-forming increment d, as in the root esa‘wash (transitive)’, stem esad-; for the validity of the root, compare the intransitive equivalent esa-na-, which of course also requires the stem-forming increment to give the intransitive stem esa-nad-. Again, imperative variants for esadcorrespond to its analysis as derived versus nonderived; esanad-, being intransitive, takes a zero ending anyway, so provides no relevant evidence. (13) Stem Present Intransitive verb keckec-xo -ik’i-ik’i-x †exu†exu-x
Imperative kec -ik’i †exu
(T) ‘sleep’ ‘go’ ‘stay’
63
64
Bernard Comrie
but k’o†ik’o†i-x k’o†-o Transitive nonderived verb kurkur-xo kur-o kosikosi-x kos-o Aa†uAa†u-x Aa†-o but xecixeci-x xec Transitive derived verb: Use stem -exu-r-exu-r-xo -exu-r -exu-exu-x -exu but -egi(-)r-egi(-)r-xo -egi-r, -egir-o t(-)e†t(-)e†-xo t-e†, te†-o esa(-)desa(-)j-x esa-d, esad-o
4. Causative formation in Tsez In the previous section the formation of causatives in Tsez was noted, and this is the last piece of information that needs to be added before progressing to the semantics of affective verbs in Tsez and Bezhta. Causative formation is highly productive in Tsez, and involves suffixing of -r. To avoid forbidden consonant clusters, and sometimes even otherwise in order to make the morpheme more salient, the appropriate form is -er. Examples (14)–(15) illustrate causative formation with the nonderived verb -exu- ‘die’ and its causative equivalent -exu-r- ‘kill’. In these simple cases, causativization takes an intransitive verb and makes it transitive; although not relevant to present concerns, it should be noted that it is perfectly normal in Tsez to take a transitive verb and causativize it to give a ditransitive verb, which then enters into the transitive clause construction. (14) kid j-exu-s. girl.abs II-die-pstwit ‘The girl died.’
(T)
(15) už-a¯ kid j-exu-r-si. boy-erg girl.abs II-die-caus-pstwit ‘The boy killed the girl.’
(T)
“Love your enemies”
Of particular interest given the concerns of this article is the fact that it is also possible to causativize affective verbs. This is illustrated in (16)–(18). First, at least with some affective verbs, it is possible to omit the experiencer without the resulting sentence being judged to be elliptical. Thus, (17) simply means that the calf turned up, showed up. This construction can then be causativized as in (18), which might be literally translated as something like ‘the shepherd caused the calf to show up’. Although both (16) and (18) can often be appropriately translated into English as ‘the shepherd found the calf ’, there is a slight difference in meaning between them. The causative version in (18) implies that the shepherd was actively looking for the sheep, whereas (16) is more neutral in meaning, and could apply to situations where the shepherd did not even know that the sheep was lost but came across it in an unexpected location. (16) aAo-r meši b-esu-s. shepherd-lat calf.abs III-find-pstwit ‘The shepherd found the calf (perhaps accidentally).’
(18) aA-a¯ meši b-esu-r-si. shepherd-erg calf.abs III-find-caus-pstwit ‘The shepherd found the calf(by looking for it).’
(T)
The causatives of affective verbs will play an important part in what follows.
5. Constraints on imperatives My attention was drawn to a somewhat surprising difference between Tsez and Bezhta while I was working with a complete Bezhta translation of the Gospel according to Saint Luke by Madzhid Xalilov and a partial Tsez translation of the Gospel according to Saint Luke by Arsen Abdulaev. At three points, this gospel has, in the Russian original from which the translation was made, the imperative of the verb vozljubit´ ‘love, come to love’. In Bezhta, the three translations are given in (19)–(21). It will be noted that in each of these examples, the verb glossed ‘love’ is simply put into the imperative.6 Since it is not a transitive verb, it appears without any imperative suffix. In (19) and (21) the experiencer is not expressed overtly, but in(20), following the original, it is, and is as expected in the lative case.
(20) mizo-l mizo-s-zu tušmal-la b-a¯t’. you.pl-lat you.pl-gen1-refl enemy-pl Ipl-love.pl.ipr ‘Love you your enemies!’(Luke 6.35)
(B)
(21) dibo hicas-na mi-coj-zu Ø-at’. you.gen1 neighbor-even you-like-refl I-love.ipr ‘Love even your neighbor as yourself.’ (Luke 10.27)
(B)
What conclusions can we draw about the semantics of the verb -at’-, and similarly other affective verbs, from these data, which are incidentally judged to be perfectly acceptable by other speakers of Bezhta? In the discussion of affective constructions and their analogs (such as dative subject constructions) in other languages, one question that sometimes arises, but has rarely received a satisfactory answer, is whether the relevant predicates have the same meaning as their English translation equivalents. Thus, does Bezhta -at’- or Tsez -eti‘love’ mean the same as English love? The data in (19)–(21) would support the claim that they do. Just as the English imperative of love is an instruction to someone to enter the state of loving someone else, so too is the Bezhta imperative of -at’-. In Tsez, the situation is quite different. The Tsez translation corresponding to Bezhta (19)–(20) is as given in (22).7,8 Crucially, the causative of the affective verb must be used. To construct (22) with uncausativized b-eti is simply ungrammatical. This is shown with a simpler pair in (23)–(24). (22) b-eti-r mež-a¯ mežu-s tušman-bi. Ipl-love-caus.ipr you.pl-erg you.pl-gen1 enemy-pl ‘Love your enemies!’ (Luke 6.27, 35)
(T)
(23) *meši b-esu. calf.abs III-find.ipr ‘Find the calf!’
(T)
(24) meši b-esu-r. calf.abs III-find-caus.ipr ‘Find the calf!’
(T)
In Tsez, imperatives of affective verbs are simply excluded. This points towards a different semantics for the relevant verbs. In Tsez, if -esu- means not ‘find’ but rather ‘be found, show up’, then it is clear that (23) cannot be used to instruct
“Love your enemies”
someone to find the calf — if it can be assigned any interpretation at all (which seems doubtful, but see below), then it would be an instruction to the calf to show up. The causative, as in (24), is acceptable and readily accounted for under this analysis: the shepherd is being instructed to cause the calf to show up. In fact, I tested a number of affective verbs to see if their imperatives were acceptable, and the general clear judgment was that they are not.9 A list of these verbs is given in (25). (25) -aci- ‘dislike’; -eti- ‘want’; -ij- ‘know’; -ikwa(d- (dialect variant: -ukad-) ‘see’; -iqi- ‘receive’; ko†’i-‘be able’; šu†’i- ‘forget’; teq- ‘hear’ but: ?*-esu- ‘find, be found’
The only affective verb that I tested for which Arsen Abdulaev was prepared to tolerate an imperative form was -esu-, but even this was only under considerable pressure, so that the judgment is not too reliable, especially when taking into account that the meaning of the resulting form -esu was judged to be unclear. In fact, this seems to be part of a more general constraint on imperatives in Tsez. In general, imperatives of intransitive verbs are judged unacceptable when the referent of the single argument would not plausibly have control over the situation in question. A representative list is given in (16). Note that these verbs are all intransitive (even though their English translations are sometimes labile), and that their transitive causative counterparts all freely allow imperative formation. (26) -a†’i- ‘be deceived’; ˇc’a†i- ‘split’; -ecu- ‘break’; -et’u- ‘tear’; geg- ‘break, die’; łas- ‘cool down’; łij- ‘finish’; -o†- ‘be sick’; q’aši- ‘get drunk’ but: -exu- ‘die’, -e>u- ‘lose (be defeated)’
In some instances, fine semantic or cultural judgments can make a difference. Thus, it might seem odd to include q’aši — ‘get drunk’ as not being under control, but in fact one could plausibly argue that while drinkers have control over whether or not they drink and how much, whether or not they get drunk is simply an automatic consequence of this (perhaps in combination with other factors clearly not under their control). The fact that-exu- ‘die’ is judged felicitous in the imperative seems to parallel its use in English expressions like die, dog! beloved of writers of historical fiction describing a fight to the death. The imperative of -e>u- ‘lose’ could be used, for instance, of someone deliberately throwing a game.10 Given the relevance of cultural factors, it would not be surprising if judgments were to vary somewhat from speaker to speaker, although I have not yet been able to test this in detail.
67
68
Bernard Comrie
Tsez has another construction that lacks an imperative, and this is the potential construction, illustrated in (27).The verb takes the potential suffix -ł. The entity that is said to have a certain ability appears in the possessive case (whereas with the underived transitive verb Aa†u- ‘drink’ it would appear in the ergative). There is no imperative corresponding to (27), since one cannot plausibly have control over being able to do something. (27) k’et’u-q >’ay Aa†u-ł-xo. cat-poss milk.abs drink-pot-prs ‘The cat can drink the milk.’
(T)
6. Other expressions of volitionality in Tsez In fact, the behavior of affective verbs and, more generally, verbs denoting uncontrolled situations in the imperative seems to be part of an even broader phenomenon in Tsez, since there are other areas where such verbs refuse to participate in structures that are open to other verbs. For instance, Tsez has a number of ways of expressing the future. Some of these, such as the future indefinite(formed by lengthening the vowel before the last consonant of the stem) or the compound future (formed by combining the infinitive with an appropriate form of the verb ‘be’), seem to be semantically neutral and can be formed from all verbs. But one future form, the so-called future definite, is highly restricted. First, it can only be used with a first-person subject. This is shown for intransitive verbs in (28)–(29), where crucially in (29) the future definite is excluded with the second person subject. (28) di j-äk’i/ j-ik’a joł/ j-ik’-an. me II-go.futind II-go-inf be.prs II-go-futdef ‘I will go (female speaker).’
(T)
(29) mi j-äk’i/ j-ik’a joł/ *j-ik’-an. you II-go.futind II-go-inf be.prs *II-go-futdef ‘You will go (female addressee).’
(T)
The same constraint is shown for transitive verbs in (30), which shows, incidentally, that it is the more agent-like argument whose grammatical person is relevant.11
“Love your enemies”
(30) di/*mi bišwa r-ac’-an. me/you food.abs IV-eat-futdef ‘I/you will eat the food.’
(T)
But even if the grammatical person condition is met, the future definite is only possible if the verb in question is plausibly under the control of the subject, so that the future definite is excluded in (31) even with a first person singular subject. (31) di Ø-a¯†’i/ Ø-a†’-a joł/ me I-be.deceived.futind I-be.deceived-inf be.prs *Ø-a†’-an. *I-be.deceived-futdef ‘I will be deceived (male speaker).’
(T)
Likewise, the future definite is excluded with affective verbs, as shown in (32)–(33), which show, incidentally, that it does not matter whether the first person singular argument is experiencer (as in(32) and (34)) or source of the experience (as in (33)). (32) dä-r mi uka¯j/ ukad-a joł/ *ukad-an. me-lat you see.futind see-inf *be.prs see-futdef ‘I will see you (male addressee.)’
(T)
(33) debe-r di uka¯j/ ukad-a joł/ *ukad-an. you-lat me see.futind see-inf be.prs *see-futdef ‘You will see me (male speaker.)’
(T)
(34) dä-r meši b-äsu/ b-es-a joł/ me-lat calf III-find.futind III-find-inf be.prs *b-es-an. *III-find-futdef ‘I will find the calf.’
(T)
If the affective verb is causativized, to give an ordinary transitive construction, then of course the future definite is perfectly acceptable: (35) di meši b-esu-r-an. me calf.abs III-find-caus-futdef ‘I will find the calf.’
(T)
Finally, the potential construction can only be formed from verbs denoting situations over which the referent of the noun phrase in the possessive case has potential control. Thus, (27), repeated as (36), is acceptable, since the cat has
69
70
Bernard Comrie
control over drinking the milk. Sentence (37) is possible with the causative of an affective verb, since one can have control over whether one causes a situation to come about, but not with the straightforward affective verb, since one does not have control over whether or not something shows up. The generalization here seems to be that it only makes sense to say that an entity is able to do something if that entity has some control over doing it. This thus fits in with a general pattern of Tsez verb semantics whereby it is not possible simply to disregard the non-control semantics of a verb, rather this must be specifically overridden, for instance by use of the causative derivative. (36) k’et’u-q >’ay Aa†u-ł-xo. cat-poss milk.abs drink-pot-prs ‘The cat can drink the milk.’ (37) dä-q meši b-esu-r-eł-xo/ *b-esu-ł-xo. me-poss calf III-find-caus-pot-prs III-find-pot-prs ‘I can find the calf.’
(= (27)) (T)
(T)
7. Conclusions While this article has been fairly wide-ranging in the data considered, I consider the main theoretically relevant point to be that even when constructions are identical in the mapping of semantic roles onto morphosyntactically defined noun phrase positions, it is still possible for them to differ subtly in their semantics, for instance for a translation equivalent to mean ‘see’ in one language but ‘be visible to’ in another. This can even apply to languages as closely related as Bezhta and Tsez. The constraint noted for Tsez preventing the formation of the imperative of affective verbs had already been noted for another language of the Tsezic branch, namely Hunzib, by Van den Berg (1995: 88), who observes further that in order to render the imperative of ‘see’ into Hunzib, it is necessary to causativize the verb, as in (38).12 (38) Ø-ãc’-k’(-o). I-see-caus-ipr ‘See [him]!’
(Hz)
Given that Hunzib stands genetically closer to Bezhta than it does to Tsez, it is perhaps surprising that in this respect Hunzib should so clearly parallel Tsez
“Love your enemies”
rather than Bezhta. A possible explanation is that the Tsez/Hunzib pattern is more archaic, and that Bezhta has innovated, something that would not be unusual given a general tendency for Bezhta, especially the dialect of the village of Bezhta itself, to be more innovative. But only more detailed study of other Tsezic and Nakh-Daghestanian languages may resolve this issue finally.
Notes * I am grateful to Arsen Abdulaev and Madzhid Xalilov for sharing their native-speaker intuitions of Tsez and Bezhta, respectively, with me, and for giving me access to their unpublished translations of the Gospel according to Saint Luke. The dictionaries Xalilov (1995; 1999) and the verb lists in Imnajšvili (1963: 187–190) have been invaluable sources of material. A slightly earlier version of this article was presented to the Department of Comparative Linguistics of the University of Leiden; I am grateful to all those who contributed to the discussion, and especially to Helma van den Berg. The following abbreviations are used: abs – absolutive, caus – causative, erg – ergative, futdef – future definite, futind – future indefinite, gen1 – genitive 1 (used when dependent on noun in absolutive case), inf – infinitive, ipr – imperative, lat – lative, pl – plural, poss – possessive, pot – potential, prs – present, pst – past, pstwit – past witnessed, refl – reflexive, sg – singular. The language of examples is identified as follows: B – Bezhta, Hz – Hunzib, T – Tsez. 1. In Tsez, somewhat unusually for a Nakh-Daghestanian language, there is a specific suffix in the ergative, namely -a¯. In Bezhta, following the more usual Nakh-Daghestanian pattern, the ergative is simply the oblique stem of the noun. Thus the Bezhta word for ‘girl’ is kid; its oblique stem is kibba- (with consonant assimilation), its ergative is kibba, its genitive-1 kibba-s, etc. Glossing kib-ba as ‘girl-erg’ is a convenient, though morphologically not quite accurate, compromise. 2. I use the term lative, rather than dative, because the same case suffix also plays a role in the formation of locational cases. But for present purposes nothing crucial hinges on this terminological choice. The affective construction thus corresponds to so-called dative subject constructions in the grammatical traditions of many other languages. 3. For verbs whose stem ends in a it is of course impossible to distinguish transitive and intransitive subtypes overtly, since with the former the stem-final a drops before the imperative suffix (cf. (4)), while in the latter the stem-final a remains in final position, i.e. both subtypes end in a. But the rule given in the text correctly generates these forms. 4. The variant -xo is used after a consonant, -x after a vowel (and, in some dialects, a semivowel). 5. In Bezhta, most verbs show some kind of internal change sensitive to the number of the absolutive argument. In the case of -at’ ‘love’, the vowel is lengthened if this argument is plural. Tsez has no analog of this.
71
72
Bernard Comrie
6. The Tsez translation does not retain the addressee pronoun in Luke 6.35. At the time of my investigation, Luke 10 had not yet been translated into Tsez. 7. In Tsez, reflexive pronouns are formed by combining the ergative of the personal pronoun with the case form of the same pronoun appropriate to the syntactic/function of the noun phrase in the clause. The initial ergative component is invariant for case and is not to be interpreted as a transitive subject/agent; indeed, it can cooccur with an ergative transitive subject/agent. 8. A possible analysis would be that they are morphologically possible, but semantically anomalous. This seems to underlie the listing of imperative, and also future definite (see below) forms for many of these and other verbs in Xalilov (1999) that are not accepted by Arsen Abdulaev. 9. The fact that the imperative and future definite of this verb are possible suggest that Imnajšvili (1963: 263)is in error in listing it with affective verbs. 10. In Tsez, the first and second person singular pronouns, and only these, use the citation form as transitive subject/agent, i.e. lack an ergative case distinct from the absolutive. 11. In Hunzib, transitive verbs usually take the suffix -o in the imperative, but with derived transitives it is usually omitted.
References Imnajšvili, David S. 1963 Didojskij jazyk v sravnenii s ginuxskim i xvaršijskim jazykami. Tbilisi: Izd-vo AN Gruzinskoj SSR. Van den Berg, Helma 1995 A grammar of Hunzib(with texts and lexicon). Munich: LINCOM Europa. Xalilov, Madžid Š. 1995 Bežtinsko-russkij slovar’. Maxacˇkala: IJaLI DNC RAN. 1999 Cezsko-russkij slovar’. Moscow: Academia.
DOCINFO
AUTHOR ""
TITLE "Semantic Compositionality and Pragmatics"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Part II
Semantic compositionality and pragmatics
AUTHOR "Dieter Wunderlich"
TITLE "Two comparatives"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Two comparatives Dieter Wunderlich Heinrich-Heine-Universität, Düsseldorf
Mi-nél több-et olvasok, an-nál keves-ebb-et tudok. what-adess more-acc read.1sg that-adess little-comp-acc know.1sg ‘The more I read, the less I know’
1.
The comparative debate
The now classical debate on comparatives in the eighties was concerned with two questions. First, is the phrasal comparative, exemplified in (1a), an elliptical subcase of a clausal comparative, exemplified in (1b) (and in what sense), or is it an independent type of construction? The clausal comparative displays a relation between degrees that, in principle, can be determined by any sort of state of affairs. This type of construction is more general, and, evidently, more complex than the phrasal one, and therefore it has attracted the interest of most of the literature on comparatives. Only a minority of researchers (Hoeksema 1983; Heim 1985; Pinkal 1989, 1990) regarded the phrasal comparative as a construction of its own, which is simpler but efficient enough to just compare two objects with respect to a given scale. (1) a. An elephant is bigger than a camel. b. This shelf is longer than the room is wide.
The second question concerned the status of the comparative complement, expressed by than in English: is it an argument introduced by the comparative morpheme when it is applied to an adjective or adverb, or is it a free adjunct? Only Pinkal (1989) argued for the latter; in his adjunct analysis, the comparative morpheme itself only contributes the information that the relevant degree surpasses some contextual value, whereas the sentential complement, being a free relative, adds a specification of this value. For the majority of researchers, all instances in (2) would be elliptical, though it is not so clear what the relevant
76
Dieter Wunderlich
complements have to look like. However, all these examples are complete in the adjunct analysis, although they may be subject to modification. (2) a. The rhinoceros is one of the bigger animals. b. I really would like to take a longer leave. c. Older people usually need more sleep.
Cross-linguistically, there is some evidence that a phrasal and a clausal comparative construction can coexist (Stassen 1984), and also morphological evidence for the adjunct analysis, besides the syntactic evidence given by Karger et al. (1994). In the following, I will consider some such evidence from Hungarian.1
2. Two types of Hungarian comparatives In contrast to English and German, but in company with Latin and probably other languages, Hungarian has two types of comparative constructions: a phrasal type in which the complement is realized with adessive case, and a clausal type in which it is introduced by the particle mint. The following examples show that both constructions are possible when two entities are compared in just one dimension, expressed by an adjective in (3a), an adverb in (3b), or a quantity expression in (3c). (3) a.
Anna érdekes-ebb volt Péter-nél /, mint Péter. Anna interesting-comp was Peter-adess / than Peter ‘Anna was more interesting than Peter.’ b. Anna gyors-abb-an olvas Péter-nél /, mint Péter. Anna fast-comp-adv reads Peter-adess / than Peter ‘Anna reads faster than Peter.’ c. Anna több könyv-et olvas Péter-nél /, mint Péter. Anna more book-acc reads Peter-adess / than Peter ‘Anna reads more books than Peter.’ (KVF: 153)
The adessive contributes the standard of comparison by semantic case (with the literal meaning ‘located at a proximal place’), therefore only the dimension expressed by the comparative morpheme is relevant (‘be more interesting’, ‘read faster’, ‘read more books’). In contrast, mint is a complementizer which selects a certain type of relative clause, and so can contribute a much wider range of degrees as the standard of comparison. The examples in (4) show that mint allows a quite articulated clause structure, even though ellipsis is preferred if the dimension of comparison is held constant.
Two comparatives
(4) a.
Anna érdekes-ebb, mint (a-milyen érdekes) Péter (volt). Anna interesting-comp than (rel-what.kind interesting Peter (was ‘Anna is more interesting than Peter was.’ b. Anna gyors-abb-an olvas, min (a-hogy (gyors-an)) Péter (olvas). Anna fast-comp-adv reads than (rel-how (fast-adv Peter (reads ‘Anna reads faster than Peter does.’ c. Anna több könyv-et olvas, mint (a-mennyi (könyve)-t) Péter Anna more book-acc reads than (rel-how.many (book-acc Peter (olvas). (reads ‘Anna reads more books than Peter does.’ (KVF: 151–154)
Similar observations can be made if the item of comparison has a lower rather than a higher degree on a scale, although with some restrictions. The adessive is only triggered by an explicit comparative form with the suffix /(V)bb/ on the scale expression, and not by the suppletive form kevésbé, which is preferred with adjectives but only allows the clausal construction. (5) a.
Anna kevésbé érdekes volt {*Péter-nél/, mint Péter}. Anna less interesting was {*Peter-adess than Peter ‘Anna was less interesting than Peter.’ b. Anna kevésbé gyors-an olvas {*Péter-nél/, mint Péter}. Anna less fast-adv reads {*Peter-adess than Peter ‘Anna reads less fast than Peter.’ c. Anna keves-ebb könyv-et olvas {Péter-nél/, mint Péter}. Anna less book-acc reads {Peter-adess than Peter ‘Anna reads less books than Peter.’ (KVF: 152–153)
The mint-construction allows a full clause, if both the dimension and the object of comparison are shifted: (6) Anna több könyv-et olvas-ott, mint a-mennyi cikk-et Anna more book-acc read-past than rel-how.many article-acc Péter írt. Peter wrote ‘Anna has read more books than Peter has written articles.’ (KVF: 152)
What is interesting in this construction is the appearance of a relative pronoun, which indicates that the free relative analysis is on the right track. (7) shows that the same clause can indeed be used as a headless relative. Therefore, the complementizer mint itself is semantically empty. This is confirmed by the fact
77
78
Dieter Wunderlich
that mint also introduces equative constructions with a positive form of the adjective, as in (8). (7) A-mennyi cikk-et Péter írt, csodálatos/impozáns volt. rel-how.many article-acc Peter wrote amazing/impressive was. ‘How many articles Peter wrote, was amazing/impressive.’ (8) Anna olyan érdekes, mint (a-milyen) Péter (volt). Anna such interesting as (rel-how Peter (was. ‘Anna was as interesting as Peter.’ (KVF: 155)
3. Restrictions for the two types of comparatives Given these two types of comparative complements, the adessive DP and the relative clause introduced by mint, one may ask why they coexist. The answer is that they underlie different structural constraints: there are some constructions in which only the adessive can appear, but others in which only mint can appear. Recall that the adessive forms a clause-internal DP phrase, but mint a CP adjunct, that is, an independent clause (even if is reduced by ellipsis). This lets us predict a certain distribution of the two comparative constructions. Concerning mint, the following constraints are relevant: (1) A CP adjunct forms a syntactic island; therefore, it does not allow extraction (such as that required by wh-movement). (2) The mint CP cannot precede the comparative morpheme (in particular, when it is elliptical and its content must be reconstructed). (3) No CP can be embedded prenominally in a possessor phrase. None of these constraints is effective with respect to adessive DPs: they do not form an island within the comparative phrase, they may precede the comparative morpheme, and they can be embedded prenominally. That extraction of the adessive complement is possible is shown by the wh-sentences in (9) as well as by the relative clause in (10). The counterparts with mint are totally ungrammatical. (9) a.
Ki-nél volt Péter érdekes-ebb? who-adess was Peter interesting-comp ‘Peter was more interesting than who?’ a1. *Ki volt Péter érdekes-ebb mint? (Who was Peter more interesting than?) a2. *Mint ki volt Péter érdekes-ebb? (Than who was Peter more interesting?)
Two comparatives
b. Ki-nél olvas-ott Péter több könyv-et? who-adess read-past Peter more book-acc ‘Peter read more books than who? (10) Ismer-t-em egy ember-t, a-ki-nél senkisem olvas-ott know-past-1sg a man-acc, rel-who-adess nobody read-past több könyv-et. more book-acc ‘I knew a man, nobody read more books than he.’
Moreover, that the adessive phrase, but not the mint-complement, can be embedded into a prenominal possessor phrase is shown in (11). (11) a.
Péter [Anná-nál érdekes-ebb emberek Peter [Anna-adess interesting-comp people’s könyv-e-i-t] olvasta. books-3sg.poss-pl-acc read.def ‘Peter has read books by people more interesting than Anna.’ b. *Péter [érdekes-ebb, mint Anna emberek könyvei-t] olvasta. (KVF: 154)
For the purpose of expressivity, it is thus plausible that the adessive exists. On the other hand, the mint clause should also exist, if only for the purpose of expressing equatives. Moreover, the intended meaning of (6) above cannot readily be expressed by an adessive comparative, unless Hungarian allowed as much nominalization as Turkish. Turkish lacks a clausal comparative formed with a particle but permits the expression of rather complex comparisons by means of nominalization (the Turkish comparative case is the ablative). (12) a.
Hasan san-a ben-den fazla kitap ver-di. (Turkish) Hasan 2sg-dat 1sg-abl more book give-past ‘Hasan gave you more books than I (gave you).’ b. Hasan san-a [ben-im ver-digˇ-im-den] fazla kitap Hasan 2sg-dat [1sg-gen give-fnoml-1sg-abl more book ver-di. give-past ‘Hasan gave you more books than I gave (you).’ c. Hasan san-a [ben-im Ali-ye ver-digˇ-im-den] fazla kitap Hasan 2sg-dat [1sg-gen Ali-dat give-fnoml-1sg-abl more book ver-di. give-past ‘Hasan gave you more books than I gave Ali.’ (Kornfilt 1997: 179)
79
80
Dieter Wunderlich
Note that the Turkish ablative, when it takes a simple DP rather than a nominalized verb, is restricted to the contrast with the nominative. As I will show later, the Hungarian adessive is less restricted, although for the price that some ambiguities arise.
4. Semantic considerations As expected, the mint-clause, but not the adessive, shows semantic effects that are unique for a clausal comparative, where the comparative is in the scope of the standard of comparison. (13a) shows that an ‘or’ coordination in the mintclause preferably gets the same interpretation as the ‘and’ coordination (13b); such an interpretation is impossible for the ‘or’ coordinated adessive in (13c).2 (13) a.
Péter okos-abb volt, mint Hans vagy Fritz. Peter clever-comp was than Hans or Fritz ‘Peter was more clever than both Hans and Fritz.’ (preferred) ‘Peter was more clever than either Hans or Fritz.’ (dispreferred) b. Péter okos-abb volt, mint Hans és Fritz. Peter clever-comp was than Hans and Fritz ‘Peter was more clever than both Hans and Fritz’ c. Péter okos-abb volt Hans-nál és Fritz-nél/ Peter clever-comp was Hans-adess and Fritz-adess/ *Hans-nál vagy Fritz-nél. *Hans-adess or Fritz-adess ‘Peter was more clever than both Hans and Fritz’
This result is derived in Pinkal’s (1989) approach, in which the mint-clause expresses generic quantification with the comparative in its scope. (14a) represents the mint-clause under the assumption that its content is (elliptically) reconstructed in the context of okos, while (14b) represents the main clause, with ‘ι’ as the Russellian operator for definite descriptions. This yields the representation (14c) for (13a); this formula is equivalent to the conjunction (14d).3 mint Hans vagy Fritz (okos volt): λD "d [clever(hans,d) ⁄ clever(fritz,d)ÆD(d)] b. Péter okosabb volt: λd ιd¢[clever(peter,d¢) > d], abbreviated as λd more.clever(peter,d))
(14) a.
Two comparatives
"d [clever(hans,d) ⁄ clever(fritz,d) Æ more.clever(peter,d)] “For all degrees d such that Hans is d-clever or Fritz is d-clever, Peter is more clever than d.” d. "d (clever(hans,d) Æ more.clever(peter,d)) & "d (clever(fritz,d) Æ more.clever(peter,d)) c.
For the adessive construction I assume that the adessive phrase is in the scope of the comparative. In (15a), the comparative is extended by means of the operation ARG (described in Wunderlich 1997), which adds a variable Q as the place-holder for the relation expressed by the adessive phrase. The adessive itself expresses an identity relation between degrees on some scale P, as shown in (15b). Functional application then yields the result (15c) for the VP in (13c) above. Finally, in (15d), the subject term is integrated, and the value of P is contextually specified by the predicate clever. One can easily see that ‘or’ and ‘and’ coordination would yield different results. This analysis thus clearly shows that the phrasal comparative leads to different results from the clausal one. (15) a. ARG(okosabb): λQ λx {more.clever(x,d) & Q(d)} b. Hans-nál: λd [d = ιd¢ P(hans,d¢)] c. λx λd {more.clever(x,d) & [d = ιd¢ P(hans,d¢) & d = ιd¢ P(fritz,d¢)]} d. $d {more.clever(peter,d) & [d = ιd¢ clever(hans,d¢) & d = ιd¢ clever(fritz,d¢)]} “The degree to which Peter is clever exceeds the degree to which Hans is clever and it exceeds the degree to which Fritz is clever.”
If an ‘or’ coordinated adessive phrase is focused, having scope over the comparative (16a) (see fn.2), one only gets the ‘or’ reading given in (16b). This is because the adessive does not evoke a generic reading. (16) a.
Péter Hans-nál vagy Fritz-nél volt okos-abb. Peter Hans-adess or Fritz-adess was clever-comp ‘It was Hans or Fritz that Peter was more clever than.’ b. $d {[d = ιd¢ P(hans,d¢) ⁄ d = ιd¢ P(fritz,d¢)] & more.clever(peter,d)]} “The degree to which Hans is clever or the degree to which Fritz is clever is exceeded by the degree to which Peter is clever.”
That the adessive is interpreted with respect to the same scale as the adjective is a default assumption. Theoretically, it should be possible that the adessive DP delivers a value on another scale, given that the adessive case is applied to a
81
82
Dieter Wunderlich
deadjectival noun.4 Consider (17b), which is the adessive counterpart to the mint-comparative in (17a); this sentence is judged grammatical by some of my informants, although Kenesei, Vago & Fenyvesi (1998: 153) regard it as ungrammatical. For comparison, the parallel Turkish example is shown in (18). (17) a.
A polc hossz-abb, mint amilyen széles a szoba. the shelf long-comp than what wide the room ‘The shelf is longer than the room is wide.’ b. A polc hossz-abb a szoba széles-ség-é-nél. the shelf long-comp the room wide-noml-3sg.poss-adess ‘The shelf is longer than the room’s width.’
(18) Masa kapG-nGn genis¸-ligˇ-in-den daha enli-dir. table door-gen wide-noml-3sg.poss-abl more broad-3sg ‘The table is broader than the door’s width.’
Since it is possible that the adessive phrase introduces its own scale, the analysis in (15), which assumes identification with the scale of the comparative element as the default rather than as the only option, is correct.
5. Adessive comparatives can suppress a structural case If the comparative complement is realized by an adessive, any other case must be suppressed, regardless of whether the complement contrasts with a subject or object. Investigating which cases can be suppressed, I observed two varieties of reaction among my informants: a more liberal one, which (under certain conditions) allows all structural arguments to be contrasted phrasally, and a more restrictive one, which forbids most of these phrasal constructions in favoring the mint-comparative. I suggest that the liberal variety represents the more conservative stage which disfavors the unifying mint-comparative. The following observations are based on the liberal judgments, whose default interpretations correspond to the only ones of the more restrictive judgments. In the context of a polyadic verb with an adverbial comparative, the adessive DP can often contrast with more than one of the verb’s arguments. Such a case is illustrated in (19a), where ‘you’ can be contrasted with either the nominative subject or the accusative object, the latter being the preferred reading. (19b) shows that the mint-clause can resolve this ambiguity by overt case-marking. That the adessive can also relate to the nominative is shown in (19c), where the preferred reading of (19a) is rather improbable.
Two comparatives
(19) a.
(Én) Péter-t jo-bb-an szeret-em nál-ad. (I Peter-acc good-comp-adv love-1sg adess-2sg ‘I love Peter more than (I love) you.’ (preferred) [mint tégedACC] ‘I love Peter more than you (love Peter).’ [mint teNOM] b. Péter-t jo-bb-an szeret-em, mint téged/ mint te. Peter-acc good-comp-adv love-1sg than 2sg.acc/ than 2sg.nom ‘I love Peter more than (I love) you/than you (love Peter).’ c. A matek-ot jo-bb-an szeret-em nál-ad. the math-acc good-comp-adv love-1sg adess-2sg ‘I love math more than you (love math).’
A similar observation can be made for ditransitive verbs, which allow the adessive DP to contrast with either the dative recipient or the nominative subject; however, the latter reading needs a different word order. (20) a.
Anna több könyv-et ad-ott Ferenc-nek Péter-nél. Anna more book-acc give-past Ferenc-dat Peter-adess ‘Anna gave more books to Ferenc than to Peter.’ b. Anna Péter-nél több könyv-et ad-ott Ferenc-nek. Anna Peter-adess more book-acc give-past Ferenc-dat ‘Anna gave more books to Ferenc than Peter.’
The following examples show that the adessive DP never contrasts with a semantic case such as inessive, ablative, or instrumental (21), nor with a possessor of the noun (22). (21) a.
Inkább a kert-ben ült, mint a ház-ban/ more the garden-iness sat.3sg than the house-iness/ *ház-nál. *house-adess ‘She sat more in the garden than in the house.’ b. Bélá-tól több level-et kap-t-am, mint Péter-to˝l/ Bela-abl more letter-acc receive-past-1sg than Peter-abl/ *Péter-nél. *Peter-adess ‘I received more letters from Béla than from Peter.’ c. Jo-bb-an szeret-nék Karcsi-val táncol-ni, mint good-comp-adv like-1sg.cond Charlie-instr dance-inf than Péter-rel/ *Péter-nél. Peter-instr/ *Peter-adess ‘I more like to dance with Charlie than with Peter.’
83
84
Dieter Wunderlich
(22) Ez inkább Péter feladat-a, mint az enyém/ *nál-am. dem more Peter task-3sg.poss than the mine/ *adess-1sg ‘This is more Peter’s task than mine.’
The adessive in the comparative construction suppresses any other case; it turns out that semantic case cannot be ignored, whereas both accusative and dative can. This fact constitutes evidence for assuming the Hungarian dative to be a structural case.5 Let us see whether the various instances of lexically marked dative in Hungarian also allow an adessive comparative: tetszik ‘like’ assigns dative to the highest argument (the experiencer) (23a), both nehéz ‘difficult’ and kell ‘need/must’ assign dative to the subject of the dependent verb (23b,c), and fontos ‘important’ assigns dative to the lower argument (23d). All of these constructions are judged grammatical by the more liberal informants, only with a slight question mark for (23a). The lexically marked dative thus turns out to be as structural as the canonical dative of ditransitive verbs. (23) a.
Nek-i jobban tetszik a könyv nál-am. dat-3sg better like the book adess-1sg. ‘He/she likes the book more than I.’ b. Anná-nak Péter-nél nehez-ebb franciául olvas-ni. Anna-dat Peter-adess difficult-comp French read-inf ‘For Anna it is more difficult to read French than for Peter.’ c. Anná-nak több level-et kell ír-ni-a Péter-nél. Anna-dat more letter-acc need write-inf-3sg.poss Peter-adess ‘Anna has to write more letters than Peter.’ d. A GB fontos-abb volt Anná-nak Péter-nél. the GB important-comp was Anna-dat Peter-adess ‘GB was more important for Anna than for Peter.’ ‘GB was more important for Anna than Peter.’
Note that (23d) is ambiguous between object reading (preferred) and subject reading for the adessive DP. Another candidate is the dative possessor, triggered by the copula (24a); here the adessive DP can be interpreted as the standard of comparison or as the place where the possessed things are located. This attributive construction also allows the adessive DP to contrast with the head noun if the sortal requirements are fulfilled (24b). (24) a.
Dániel-nek érdekes-ebb könyv-e-i vannak Péter-nél. Daniel.dat interesting-comp book-3sg.poss-pl are Peter-adess ‘Daniel has more interesting books than Peter.’ ‘Daniel has more interesting books at Peter.’
Two comparatives
b. Dániel-nek érdekes-ebb könyv-e-i vannak a Daniel.dat interesting-comp book-3sg.poss-pl are the GB-nél. GB-adess ‘Daniel has more interesting books than GB.’
Finally, the different behavior of structural vs. semantic case can be seen nicely from the following minimal pair: telefonál ‘call up’ can be used either with dative or with instrumental, see (25a) vs. (25b). The respective comparative constructions allow besides the irrelevant subject reading (‘I called him more often up than Peter did’) an object reading (even preferred) with dative (26a), but only a locative case reading with instrumental (26b). (25) a.
Anna Péter-nek telefonál. Anna Peter-dat telephone ‘Anna calls Peter up.’ b. Anna Péter-rel telefonál. Anna Peter-instr telephone ‘Anna calls Peter up.’
(26) a.
Gyakra-bb-an telefonál-t-am nek-i Péter-nél. often-comp-adv telephone-past-1sg dat-3sg Peter-adess ‘I called him up more often than (I called up) Peter.’ ??‘I called him up more often at Peter.’ b. Gyakra-bb-an telefonál-t-am vel-e Péter-nél. often-comp-adv telephon-past-1sg instr-3sg Peter-adess ‘I called him up more often at Peter.’ *‘I called him up more often than (I called up) Peter.’
6. Attributive adessive comparatives can suppress all kinds of case In attributive comparatives, the adessive DP preferably contrasts with the head noun, independent of its case, but with a different order it may also contrast with another argument of the verb (see also (24) above). The contrast with the head noun is the more natural option because the attribute is part of the NP; in such a case, the denotation of the adessive DP must be contained in that of the head noun, for instance, GB Ã book.
85
86
Dieter Wunderlich
(27) a.
Anna érdekes-ebb könyv-ek-et olvas-ott a GB-nél. Anna interesting-comp book-pl-acc read-past the GB-adess ‘Anna read more interesting books than GB.’ b. Anna Péter-nél érdekes-ebb könyv-ek-et olvas-ott. Anna Peter-adess interesting-comp book-pl-acc read-past ‘Anna read more interesting books than Peter.’
Other instances of attributive comparatives are shown in (28). (28) a.
Nál-a unalmas-abb ember-t még soha sem látt-am. adess-3sg boring-comp person-acc yet never not saw-1sg ‘I’ve never seen a more boring person than him.’ b. Anna a könyv-et egy Péter-nél érdekes-ebb ember-nek Anna the book-acc a Peter-adess interesting-comp person-dat küld-te el. send-past.3sg prev ‘Anna sent the book to a more interesting person than Peter.’ c. Ez a könyv még egy Péter-nél butá-bb ember-nek is this the book even a Peter-adess stupid-comp person-dat also tetszett. please.past ‘This book has pleased an even more stupid person than Peter.’
The attributive construction even allows an adessive complement to contrast with a head noun that bears semantic case, as shown in (29). This is expected, because the adessive DP, being part of the attribute, contrasts with the nominal head of the DP to which it belongs; there is no external contrast to another DP with semantic case, as in (21) above. (29) a.
Még Péter-nél is érdekes-ebb ember-to˝l even Peter-adess also interesting-comp person-abl kap-t-am level-et. receive-past-1sg letter-acc ‘I received letters from an even more interesting person than Peter.’ b. Szeret-n-ék egy Péter-nél érdekes-ebb ember-rel like-cond-1sg a Peter-adess interesting-comp person-instr táncol-ni. dance-inf ‘I would like to dance with a more interesting man than Peter.’
Two comparatives
7. Summary This brief study has presented some interesting facts about Hungarian and I hope it has established the view that two options are better than one. The adessive comparative can escape the striking constraints the clausal construction is confronted with, whereas the mint comparative can escape the clumsy nominalizations of Turkish. There is good evidence that the comparative semantics should distinguish between a clausal and a phrasal type with different scope relations. Along the way, it has been shown that Pinkal’s free relative analysis of the comparative clause complement is transparently realized in Hungarian. Furthermore, it has turned out that the Hungarian adessive comparative can override a structural case (including dative), but not a semantic case. Hence, the prominent distinction of structural vs. semantic case is reflected in the distribution of these two comparatives. A final question, not yet addressed, is why the Hungarian correlative construction is best expressed with two adessives (‘The more I read, the less I know’).
Notes 1. When I wish that our addressee is a happier and wiser man, I do not have in mind any person that I want to compare him with. All that I would like to emphasize is that Hungarian, being home of our addressee, is happy and wise in providing us with two standards of comparison. Feri Kiefer once edited a paper of mine on comparative clauses (1973); this now is the second one about this fascinating topic (which justifies the title ‘two comparatives’). Between then and now lies what I call the ‘classical period’ of comparative research, with a list of famous names (Cresswell 1976; Klein 1980, 1991; Hellan 1981; Hoeksema 1983; von Stechow 1984; Heim 1985; Bierwisch 1987; Pinkal 1989, 1990), from all of them I learned much, although only little has made its way into this paper. The Hungarian examples indicated with KVF are taken from Kenesei, Vago & Fenyvesi (1998). I am especially grateful to Tünde Vallyon and Renata Meyer for providing me with further data, to Chris Piñón as well as two reviewers for their constructive comments, and to Barbara Stiebels, who introduced me to Hungarian morphology. 2. It is possible to focus the phrase Hansnál vagy Fritznél, so that it gets scope over the comparative. Then only the ‘or’ reading is possible, see below. 3. For the conjunction expressed in (13b), the interpretation parallel to (14c) would make sense only if Hans and Fritz are equally clever; therefore, the ellipsis is reconstructed differently, and the sentence directly expresses (14d). 4. If there exists a base noun, the deadjectival noun is blocked:
87
88
Dieter Wunderlich
(i)
A csomag nehez-ebb a mérleg súly-á-nál/ the package heavy-comp the scale weight-3sg.poss-adess/ *súlyos-ság-á-nál. *weighty-nom-3sg.poss-adess ‘The package is heavier than the scale’s weight/*weightiness.’
5. There is more evidence for this conclusion: the Hungarian dative agrees with the possessor, as well as with the inflected infinitive (triggered by the modals such as kell ‘must’, kellene ‘ought to’, and szabad ‘permitted’), and it is lexically assigned to either the highest argument (tetszik ‘like’) or the lowest argument (telefonál ‘call up’), thereby forcing the remaining argument to be realized in the nominative.
References Bierwisch, Manfred 1987 “Semantik der Graduierung”. In Grammatische und konzeptuelle Aspekte von Dimensionsadjektiven, M. Bierwisch & E. Lang (eds), 91–286. Berlin: Akademie Verlag. Cresswell, Max 1976 “The semantics of degree”. In Montague Grammar, Barbara Partee (ed.), 261–292. New York: Academic Press. Heim, Irene 1985 Notes on comparatives and related matters. Manuscript. Austin, Texas. Hellan, Lars 1981 Towards an integrated analysis of comparatives. Tübingen: Narr. Hoeksema, Jack 1983 “Negative polarity and the comparative”. Natural Language and Linguistic Theory 1: 403–434. Karger, Reinhard, Lerner, Yves-Jean and Pinkal, Manfred 1994 “Zur syntaktisch-semantischen Analyse attributiver Komparative”. In Kognitive Linguistik: Repräsentation und Prozesse, S. Felix, C. Habel, and G. Rickheit (eds), 107–128. Opladen: Westdeutscher Verlag. Kenesei, István, Vago, Robert M. and Fenyvesi, Anna 1998 Hungarian. London: Routledge. Klein, Ewan 1980 “A semantics for positive and comparative adjectives”. Linguistics and Philosophy 4: 1–45. 1991 “Comparatives”. In Semantics. An international handbook on contemporary research, A. v. Stechow and D. Wunderlich (eds), 673–691. Berlin: de Gruyter. Kornfilt, Jaklin 1997 Turkish. London: Routledge.
Two comparatives
Pinkal, Manfred 1989 “Die Semantik von Satzkomparativen”. Zeitschrift für Sprachwissenschaft 8: 206–256. 1990 “On the logical structure of comparatives”. In Natural language and logic: Lecture notes in Artificial Intelligence, R. Studer (ed.), 146–167. Berlin: Springer. Stassen, Leo 1984 “The comparative compared”. Journal of Semantics 3: 143–182. von Stechow, Arnim 1984 “Comparing semantic theories of comparison”. Journal of Semantics 3: 1–77. Wunderlich, Dieter 1973 “Vergleichssätze”. In Generative Grammar in Europe, F. Kiefer and N. Ruwet (eds), 629–672. Dordrecht: Reidel. 1997 “Argument extension by lexical adjunction”. Journal of Semantics 14: 95–142.
89
AUTHOR "Barbara H. Partee and Vladimir Borschev"
TITLE "Some puzzles of predicate possessives"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Some puzzles of predicate possessives* Barbara H. Partee and Vladimir Borschev University of Massachusetts, Amherst / VINITI, Moscow
1.
Background: Possessives and the argument-modifier distinction in NPs
Possessive constructions like John’s teacher, John’s team, John’s cat, friend of John’s offer an interesting test-bed for the argument–modifier distinction in NPs, both in English and cross-linguistically. Many, perhaps all, possessives seem to have some properties of arguments and some of modifiers, but some seem more argument-like and some more modifier-like. Recent proposals by Jensen and Vikner (1994), Vikner and Jensen (1999), Partee and Borschev (1998), Borschev and Partee (1999a,b) analyze all possessives as argument-like, a conclusion we are no longer sure of. It is not easy to settle the question of whether there is a substantive difference between these two “roles” of possessives, and it may well be the case that all or many possessives play both roles at once. One central question about possessive constructions, then, is the following: Are all, some, or no possessives arguments of nouns, and if so, which ones (and how can we tell?), and of what kind, and at what ‘level’ of analysis? Within this larger question, we discuss here a relevant narrower question: Do predicate possessives provide strong evidence against a unified treatment of all possessives as arguments? 1.1 Possessives/genitives and related constructions The terminology surrounding “possessives” and “genitives” is confusing, since the correspondences among morphological forms, syntactic positions, grammatical relations, and semantic interpretations are complex and debated, and vary considerably across languages. For clarification, let us distinguish at least the following:1
92
Barbara H. Partee and Vladimir Borschev
1. Possessive pronouns: E. my, his; R. moj ‘my’, ego ‘his’; E. predicative forms mine, his and postnominal forms of mine, of his. 2. English “Saxon genitives”: John’s, and the postnominal Saxon genitive of John’s. 3. English PP with of + Acc. 4. Russian postnominal genitive NP: Mendeleeva ‘of Mendeleev’, tigra ‘of a/the tiger’ 5. Russian prenominal possessive: Mašin dom ‘Masha’s house’. Some of the problems of the semantics of possessives affect all of these, while some of the problems require making distinctions. Very similar problems arise in corresponding constructions in many other languages, and related problems arise with the English verb have and its lexical and constructional counterparts in other languages (Freeze 1992; Landman and Partee 1984; Szabolcsi 1994; Jensen and Vikner 1996; Partee 1999b). The present work concerns the possible need for a distinction between possessives as modifiers and possessives as arguments, and the role that predicate possessives may play in resolving that issue. One starting point is the following data from Partee (1983/97: 464): (1) a. John’s team b. A team of John’s c. That team is John’s (2) a. John’s brother b. A brother of John’s c. (#) That brother is John’s (3) a. John’s favorite movie b. A favorite movie of John’s c. (#) That favorite movie is John’s
Informally, a unified interpretation of genitive phrase “John’s” that applies to all of these cases is that the genitive phrase always expresses one argument of a relation, for which we will use the descriptive term “genitive relation”, following Jensen & Vikner (1994). But the relation can come from any of three sources: (1) the context, as in (1) (“plays for”, “owns”, “is a fan of”, etc.); this happens when the noun is a plain 1-place predicate; (2) an inherently relational noun like “brother”; (3) an inherently relational adjective like favorite. The puzzles include these: can (and should) examples (1a) and (2a) be given a uniform analysis, and if so, how? Or does the genitive construction combine
Some puzzles of predicate possessives
differently with plain and relational nouns, and if so, are these differences predictable from some general principles? Should the first case be split into two distinct cases, one being a default preference of the “genitive” construction itself for a genitive relation in the family of “owns”, “possesses”, “controls”, possibly with a distinct syntactic source? The examples in (3) show that argument-like genitives cannot always simply be analyzed as complements of a lexical noun, since it is the whole N-bar favorite movie that provides the relation of which John is an argument.2 The Russian “genitive modifier” (GM) construction exemplified in (4) presents similar challenges, showing a similarly diverse range of “genitive relations”, with a similar range of relational and non-relational nouns, although there are interesting differences between English and Russian to account for as well. (4) a.
b.
c.
d.
e.
f.
g.
h.
i.
ljubitel’ košek lover.nom.sg cat.gen.pl ‘lover of cats, cat-lover’ rost ˇceloveka height.nom.sg man.gen.sg ‘height of the/a man’ nožka stola leg.nom.sg table.gen.sg ‘leg of the table, table leg’ krug syra circle.nom.sg cheese.gen.sg ‘circle (wheel) of cheese’ stakan moloka glass.nom.sg milk.gen.sg ‘glass of milk’ portret Peti portrait.nom.sg Petja.gen ‘picture of Petja’ sled tigra track.nom.sg tiger.gen.sg ‘track of the/a tiger’ sobaka docˇeri dog.nom.sg daughter.gen.sg ‘the daughter’s dog’ nebo Andreja Bolkonskogo sky.nom.sg Andrej.gen Bolkonsky.gen ‘Andrej Bolkonsky’s sky’
93
94
Barbara H. Partee and Vladimir Borschev
In the case of Russian, the question of whether the examples in (4) are all instances of a single construction is even more difficult than in the case of English, since the uses of the Russian genitive NP cover uses analogous to both the English Saxon genitive in (1)–(3) and English PPs with of + Acc. At a descriptive level, virtually all authors who have grappled with the semantics of genitive constructions are agreed that in some cases the genitive NP seems argument-like and in other cases it seems modifier-like. The “argument” nature of at least some genitives is clearest in the case of some deverbal nouns, those called “Complex Event Nominals” by Schoorlemmer (1995), “Derived Nominals” by Babby (1997), and “process nominals” by Rappaport (1998). To be slightly more precise about our relatively neutral assumed syntax for the first of these constructions, and for the Russian postnominal genitive construction, we represent the syntactic structure as in (5) below, a linearized form of the schematic phrase structure tree of Borschev and Partee (1999b): (5) [N N NPGEN], where N is a cover term for N0 and non-maximal N-bar (Montague 1973’s CN and CNP), and NP is a cover term for both NMAX and DP.
The semantic question is: do the possessive constructions [N N NPGEN] have a uniform compositional interpretation? There are in principle three possibilities. (1) Assimilate all cases to the “free R” reading. That option was proposed by Hellan (1980). Partee (1983/97) argued against it on the basis of the contrast among the (c) examples in (1)–(3). (2) Posit two different possessive constructions, treating “inherent R” possessives as type-raised arguments and “free R” possessives as (intersective) modifiers (Partee 1983/97). (3) Assimilate all cases to the “inherent R” reading. This option was introduced by Jensen and Vikner (1994), and further explored in Partee and Borschev (1998), Borschev and Partee (1999a,b), and Vikner and Jensen (1999). 1.2 Two theories of possessives Partee (1983/97) proposed two distinct genitive constructions with relational and non-relational nouns, the latter incorporating a “free relation variable R” whose value must be supplied by context. On the other hand, (a modified version of) Jensen and Vikner (1994) offers a uniform interpretation of the genitive, with coerced type-shifting of the N-bar to a relational reading when necessary. The investigation of the differences between these two approaches, in part through an ongoing dialogue which Borschev and Partee have been
Some puzzles of predicate possessives
carrying on with Jensen and Vikner over the past two years, has led us to an appreciation that the problem of the semantics of the genitive construction(s) is a much richer domain of inquiry than we had originally imagined, and to convergence on some issues and new questions on others. A note about notation: In what follows we use CN for a (“plain”) N-bar of type ·e,tÒ (one-place predicate, with only a “referential” θ-role (Williams 1981, the R role of Babby 1997), and TCN for a (“transitive” or “relational”) N-bar of type ·e,·e,tÒÒ like father, favorite movie. The analysis of Partee (1983/97) posits an ambiguity in the construction, with the N-bar supplying the relation if it is relational, and with the construction supplying a “free relation variable” if the N-bar is not relational. We illustrate the postnominal genitive, as in (1b), (2b), (3b), which Partee (1983/97) analyzed as a modifier, treating the prenominal genitive in (1a), (2a), (3a) as a composition of the postnominal genitive with an implicit definite determiner. Postnominal genitive (of John’s): combines with CN or TCN to make a CN When a genitive NP combines with a plain CN, type ·e,tÒ: the construction provides a “free R”, a variable of type ·e,·e,tÒÒ which we write as Ri.3 (6) of John’s: team of John’s:
When a genitive NP combines with a TCN, type ·e,·e,tÒÒ, the TCN provides its “inherent R”. (7) of John’s: λR[λx[R(John)(x)]] or equivalently, λR[R(John)] teacher of John’s: λx[teacher(John)(x)]]
Jensen and Vikner (1994) propose that an analysis which incorporates coerced type-shifting in the sense of Partee (1987) should be able to do without two separate rules for the genitive. They present an alternative analysis, building on the framework of Pustejovsky (1993, 1995): the genitive must always combine with a relational common noun (phrase), coercing a one-place predicate noun to a two-place relational meaning (“team” to an appropriate sense of “teamof”). Their analysis corresponds to the “inherent R” case of Partee (1983/97), and with a relational noun like teacher the two analyses agree. The difference arises with a plain one-place CN like chair or team, which on their analysis is coerced to a TCN interpretation. Jensen and Vikner follow Pustejovsky in appealing to the qualia structure of the lexical entry to guide the coercion, so
95
96
Barbara H. Partee and Vladimir Borschev
that for instance the telic role of chair (“chairs are to sit in”) licenses the shift of CN chair to TCN chair illustrated below. (8) CN chair: TCN chair:
λx[chair(x)] λyλx[chair(x) & sits-in(x)(y)]
Initially we had some important differences with Jensen and Vikner concerning the degree to which lexical meaning drives coercion. In their current work and our most recent published work, we are agreed that on the most general version of their approach, the genitive construction should always demand a TCN to combine with, and if it finds instead a CN it will coerce it by whatever means are available and “natural”, sometimes lexical, sometimes pragmatic. (We make a less sharp distinction between lexically and contextually supplied shifted meanings than Jensen and Vikner do, because of the outlook on the integration of information from lexical and other sources described in Partee and Borschev 1998; Borschev and Partee 1998.) A “pragmatic” coercion is seen as shifting the noun to a relational reading that incorporates the “free relation variable” of Partee (1983/97) into the shifted noun meaning. (9) TCN team:
λyλx[team(x) & Ri(x)(y)]
As in Partee’s analysis, a felicitous use of an expression with a free variable requires that the context make a particular choice of value for the variable salient. Partee and Borschev (1998) propose extensions to Jensen and Vikner’s coercion approach to cover also the “contextual” cases, and point to a need for more fine-grained coercion principles to cover phenomena involving the relational adjective favorite and the difference in “most likely relation” in the interpretation of examples like John’s movie and John’s favorite movie. One main difference between the two approaches is then in where a “free relation variable” is added in a case where context is driving a pragmatically based coercion. Let’s suppose that team of Mary’s is such a case. (10) Jensen and Vikner: of Mary’s: (shifted) team: team of Mary’s:
(11) Partee (1983): of Mary’s: (non-shifted) team: team of Mary’s:
λPλx[P(x) & Ri(Mary)(x)] team λx[team(x) & Ri(Mary)(x)]]
Some puzzles of predicate possessives
The final result is the same; but for Jensen and Vikner the free relation variable comes in as part of the meaning of the shifted noun, while for Partee (1983/97) it comes in as part of the meaning of the genitive construction itself. Does this difference in “where” the free relation variable is situated ever make a detectable difference? Yes. Partee and Borschev (1998) give an empirical argument in favor of Jensen and Vikner’s approach, based on an analysis of the example Mary’s former mansion, suggested to us by Norvin Richards (p.c.). The argument rests on four assumptions, as spelled out in (12) below. (12) Assumptions: 1. mansion is lexically a 1-place noun. 2. former is an endocentric modifier, lexically a CN/CN, shiftable to a TCN/TCN. former as CN/CN: former monastery, former dancer. former as TCN/TCN: former owner, former friend. 3. The “free relation” variable in this case has as one of its most salient values something like “owns” or “lives in”. 4. Mary’s former mansion has two readings: “Reading A”: a former mansion (perhaps now just a ruin) that is (now) Mary’s. I.e., now Mary’s, formerly a mansion; and “Reading B”: something that was formerly Mary’s mansion; it may still be a mansion, but it’s no longer Mary’s.
On the Partee (1983/97) account, there is no motivation for any type-shifting to occur, and the “free relation” “owns” will be introduced with the possessive Mary’s, after former has combined with mansion. This means that the free relation (“owns”) in the interpretation of the possessive Mary’s will never be under the scope of former. As a result, Partee (1983/97) can derive Reading A above, but not Reading B. Tree (13) shows the compositional structure of Mary’s former mansion on the account of Partee (1983/97).
97
98
Barbara H. Partee and Vladimir Borschev
(13)
NP NP’s Mary’s
CNP CN/CN
CN
former
mansion
free R introduced here
But Jensen and Vikner’s account, with coercion of CN to TCN, does provide derivations for both readings, which Partee’s account cannot. For Jensen and Vikner, Mary’s coerces former mansion to a relational TCN. Given our assumptions, there are two ways that former mansion could shift to a TCN. 1.Initially leave mansion as a CN, treat former as CN/CN, combine them to form a CN, as on Partee account; then shift that CN to a TCN, bringing in the free variable at that stage to get the shifted meaning of former mansion shown below in (14): (14) λy[λx[former(mansion)(x) & Ri(y)(x)]]
[Ri: “is owned by”]
This corresponds to Reading A above, with the free R introduced at the point where the CNP shifts to become a TCNP. The compositional structure would be almost identical to that in tree (13), differing only in “where” the free R is introduced. 2.Or shift mansion to a TCN, and former to a TCN/TCN, combine them to form a TCN as shown below in (15): (15) λy[λx[former(mansion-of)(x)(y)]], where mansion-of is an abbreviation for λy[λx[mansion(x) & Ri(y)(x)]]
This corresponds to Reading B above, with compositional structure as in (16) below.
Some puzzles of predicate possessives
(16)
NP NP’s Mary’s
TCNP TCN/TCN
TCN
CN/CN
CN
former
mansion free R introduced here
We assume that both of these ways of coercing the phrase former mansion are structurally available; different choices of lexical items or different contexts may favor one over the other, but since both are consistent with all the principles that we are aware of, the Jensen and Vikner approach successfully predicts the ambiguity and therefore has a clear empirical advantage over the Partee (1983/97) approach.4
2. Predicate possessives: A problem for the “one genitive” approach? In spite of the theoretical appeal of the “one genitive” approach and its ability to solve the problem of Mary’s former mansion, we are still not convinced that it is correct. One of our main worries concerns predicate possessives. We earlier observed that predicate possessives seem to favor “free R” interpretations; and predicate possessives are not in a structural argument position unless one posits an empty head noun accompanying all predicate possessives. Both of these factors support the treatment of predicate possessives in Partee (1983/97) as type ·e,tÒ, “free R” only. There the predicative John’s was interpreted as λx[Ri(John)(x)]. The modifier-type postnominal genitive of John’s of Partee (1983/97), given in (6) above, which combines with a plain CN, amounts simply to intersective modification of the noun by this predicate possessive. And if some possessives can occur as basic ·e,tÒ predicates, that would suggest that when those same possessives occur inside the NP, they are basically modifiers, and not arguments, returning us to the distinction posited in the earlier Partee (1983/1997) approach. If there are no possessives that demand a treatment as basic type ·e,tÒ predicates, that would be an argument in favor of treating all ‘modifier possessives’
99
100 Barbara H. Partee and Vladimir Borschev
occurring inside an NP within the “one genitive” approach of Jensen and Vikner. That can be done on their approach by type-shifting the head CN or CNP into a relational reading incorporating the free R, so that it then takes the possessive term as an argument. But if we find in some languages that there are systematic differences in form and/or interpretation between certain genitives/ possessives that occur only NP-internally and others that occur both predicatively and NP-internally, that would present a serious challenge to the “one genitive” approach, at least for those languages. But the issue is empirically complex for at least two reasons: (1) there may be independent reasons (syntactic or morphological) why some kinds of possessives (e.g. Russian genitives) cannot occur as predicates; (2) and some predicate possessives may be elliptical full NPs; it is not always easy to tell. Much of what follows is concerned with the latter problem. In the following sections, we look at evidence about predicate possessives in English, Russian, German, and Polish. The evidence supports the idea of two semantically different kinds of possessives, with some forms, such as English Saxon genitives, used for both. One kind are argument possessives, which fit the Jensen and Vikner analysis; these occur in construction with a relationally interpreted noun (or with an adjective like favorite plus a noun). Argument possessives do not occur in type ·e,tÒ, so when they occur alone, they are interpreted as elliptical NPs5 with a relational noun implicitly understood. The Russian genitive appears to be of this type, and we consider the Jensen and Vikner analysis correct for the Russian genitive construction. The other kind are true predicative possessives, basically of type ·e,tÒ, interpreted approximately as in the corresponding analysis of Partee (1983/97), but with the “free R” preferentially interpreted as some kind of “possession” or “control”. To represent the way this distinction differs from the original distinction of Partee (1983/97), we will stop referring to the “free R” and refer instead to RPOSS. When this kind of possessive occurs inside an NP, it is a modifier rather than an argument. We believe that the Russian prenominal possessive forms discussed in Section 2.2 are of this type. Since the English Saxon genitives, as well as possessive pronouns in all four of the languages looked at here, have both uses, we conclude that the “one genitive” approach cannot be correct for those constructions. But if we do conclude that in English, for instance, there is a distinction to be made between argumental and modifier constructions both expressed by Saxon genitives like prenominal John’s, we are left with a puzzle concerning the large proportion of cases which could seemingly be analyzed either way: are
Some puzzles of predicate possessives
they all “ambiguous”? We will return to this puzzle, which remains open, in Section 3. 2.1 Predicate possessives in English The nature of predicate possessives is less clear in English than in some other languages. It is difficult to be sure whether an apparent predicate possessive like John’s in (1c), repeated below, is a simple one-place predicate with an RPOSS or “possession” reading, or is an ‘argument genitive’ occurring as part of an elliptical NP, i.e. with John’s implicitly in construction with another occurrence of team. (1) c.
That team is John’s.
In Stockwell, Schachter and Partee (1973), it was asserted that predicate possessives as in (1c) allow only a “free R” reading, but that conclusion, maintained in Partee (1983/1997) and in Partee and Borschev (1998), was probably based on too small a sample of data and not enough careful examination of possibilities. So while (17a,b) are indeed bad, (17c) (example from Ash Asudeh, p.c.) seems to be able to get a relational reading all right. And (17d,e) (from Ekaterina Rakhilina, p.c.) together with the badness of (17f) strongly support the hypothesis that there is something wrong with the subject in (17a,b), not with the unavailability of a suitable interpretation for the predicate possessive as hypothesized by Stockwell et al (1973). (17) a. b. c. d. e. f.
*That father is John’s. *That favorite movie is John’s. That teacher is John’s. His [pointing] father is also John’s. Dad’s favorite movie is also mine. *That father is John’s father.
The good examples in (17), namely (17c–e), all have predicate possessives that may be interpreted as elliptical NPs:6 John’s teacher, John’s father, my favorite movie. The bad examples (17a,b,f ) all have intrinsically relational head nouns (or common noun phrase in the case of 17b) that have to be interpreted nonrelationally in the subject but relationally in the predicate, assuming that (17a,b) have elliptical predicate possessives.7 The head noun in the subject in examples (17a,b,f ) must shift to a non-relational reading in order to be compatible with the demonstrative determiner that; with such strongly relational expressions as father and favorite movie, it is not easy to construe them non-relationally, and
101
102 Barbara H. Partee and Vladimir Borschev
a strong context is required to interpret the subject phrases at all. It may be that there is a restriction (perhaps a processing restriction) on shifting an expression “away from” its basic meaning and then “back again” (The “bad” sentences are probably indeed not ‘ungrammatical’, but are nearly impossible with respect to the intended readings ‘John’s father’, etc.) In the good examples (17d,e) we have the relational readings of the head noun (phrases) in both the subject and the (elliptical) predicate. And the relevant difference between the good (17c) and the bad (17a) may be that unlike father, teacher is lexically supplied with equally salient and closely related relational and non-relational readings, so that one wouldn’t have to ‘suppress’ the relational reading by shifting in order to interpret teacher in the subject NP non-relationally. The data above, reinforced by the Dutch data mentioned in footnote 6, strongly suggest that predicate possessives may sometimes be “elliptical” NPs or “Determiner-only” NPs. And if all bare possessives in all languages could be interpreted as elliptical NPs, then predicate possessives would not pose a problem for the “one-genitive” analysis; the difference between possessive or genitive forms that can and that cannot occur “bare” as predicates would simply reflect constraints on NP ellipsis. But we believe that not all predicate possessives are elliptical. We do not have conclusive arguments for English; there are several complicating factors, including problems in the analysis of copular sentences. We briefly mention here some data that we consider relevant, but rather than trying to build any arguments based on English, we will then turn to other languages where the structures in question can be identified more clearly. There are some uses of “bare” mine, yours, John’s, etc., that occur in argument positions that are otherwise occupied by full NPs of type e, as in (18). We take this as clear evidence that bare possessive pronouns or Saxon genitives do occur as elliptical NPs.8 (In Russian, for instance, the same is possible with possessive pronouns but not normally with bare genitive NPs.) (18) Mine is over there. John’s is over here. Sue already took yours and Mary’s.
There are postcopular uses as in (19) where the possessives seem very much like inverted elliptical subjects, arguably of type e although perhaps of type ·e,tÒ. The debate relates to the problem of “inversion around be” and the types of NPs in apparent “inversion” sentences discussed in Williams (1983), Partee (1986), Moro (1997), Heycock and Kroch (1998, 1999), Partee (1999a).
Some puzzles of predicate possessives 103
(19) a. Some of the most interesting results were John’s. b. One of the most interesting suggestions was John’s. c. Another proposal we will have to consider is Sara’s.
Examples (20)–(22) below seem to us to be candidates for genuinely predicative possessives that are not elliptical full NPs, but we have no really convincing structural arguments for making the distinction in English, so our hypothesis that these are predicative rather than elliptical is based so far on semantic intuitions and translation possibilities into languages where the distinction seems clearer. (20) The house, the barn, and the land are finally ours and ready to move into. (21) It’s already/now/finally/almost ours. (22) Anything we find on this land is John’s.
Rather than try to support our intuitions about the English examples, we turn to some languages where we have found some syntactic and/or morphological distinctions that provide evidence for a distinction between modifier possessives and argument possessives. 2.2 Russian prenominal possessives vs. genitives In Russian, possessive pronouns and the normally prenominal quasi-adjectival possessive forms can occur in predicate position but genitive NPs cannot.9 This suggests that Russian genitive NPs may always be argument-like, and that the Jensen & Vikner uniform analysis with coercion of CNs to TCNs (extended to Russian in Borschev and Partee 1999a,b) is correct for the Russian genitive construction. It also suggests that the Russian prenominal possessive forms are at least sometimes modifier-like, and the same for the possessive pronouns. (Evidence of ambiguity of the roles of the possessive pronouns is in Section 2.4.) The Russian prenominal possessive construction studied by KoptjevskajaTamm and Šmelev (1994) and by Babyonyshev (1997) is illustrated in (23) and the genitive construction in (24). (23) a.
Petin stul Petja.poss.masc.sg chair.masc.sg ‘Petja’s chair.’ b. Mamin portret Mama.poss.masc.sg portrait.masc.sg ‘Mama’s portrait.’
104 Barbara H. Partee and Vladimir Borschev
(24) a.
stul Peti chair.masc.sg Petja.gen.sg ‘Petja’s chair.’ b. portret mamy portrait.masc.sg Mama.gen.sg ‘Mama’s portrait.’
In these examples, both constructions can be used in describing the same range of cases; the possible relations of Petja to the chair or of Mama to the portrait are as various as with the English prenominal genitive. But the meanings do not “feel” identical. In the possessive construction in (23), we would like to claim (as did Schoorlemmer 1995) that the possessive Petin, mamin acts as a modifier of the head noun. We believe that the prototypical interpretation of the possessive modifier is indeed ‘possession’ (of the object denoted by the head noun, by the (animate) entity denoted by the noun in the possessive form.) To maintain such a claim, it seems that ‘possession’ must be understood in a broadly extended sense to apply to a diverse range of relations; see Heine (1997). Thus in example (23b), possession may be possession proper, ‘authorship’, or the relation of ‘being portrayed’. But the possibility of expanding the sense of ‘possession’ is evidently not unlimited. Thus ‘murderer of Petja’ can be expressed in Russian by (25a) but not by (25b). (25) a.
ubijca Peti murderer.masc.sg Petja.gen.sg ‘Petja’s murderer’ (murderer of Petja) b. Petin ubijca Petja.poss.masc.sg murderer.masc.sg # ‘Petja’s murderer’ [OK only as e.g. ‘a murderer Petja has hired’]
In the genitive construction in (24a), we analyze Peti as an argument of the relation which connects it to stul. In the given case, the most salient relation could alternatively be seen as some kind of possession as well; but ‘possession proper’ is not the prototypical interpretation for the genitive construction. The range of possible relations expressed with a genitive is extremely broad (cf. Knorina 1985, 1988, 1990, 1996; Borschev and Knorina 1990; Partee and Borschev 1998; Borschev and Partee 1999a,b). While this data is not completely conclusive, it supports the hypothesis that the Russian genitive construction is correctly analyzed as uniformly argumental, i.e. that Jensen and Vikner’s approach to English genitives is correct instead for Russian genitives. And we believe that the Russian prenominal “adjectival”
Some puzzles of predicate possessives 105
possessives are basically modifiers, with the “free” RPOSS as the core of their meanings (see the analysis in (31) below for German). But the high overlap in possible interpretation of the two constructions, as illustrated in (23) and (24), is an example of the puzzle mentioned at the end of the introduction to Section 2, to which we return in Section 3. 2.3 German possessive pronouns Tony Kroch (p.c.) suggested looking for languages that would given evidence from agreement behavior as to whether predicate possessives are more like simple (adjectival) predicates or more like full NPs. Sten Vikner (p.c) observed that German is a language that gives some evidence: Predicate adjectives in German do not agree with subjects, but predicate possessives do, suggesting that predicate possessives are indeed more like elliptical NPs than like simple ·e,tÒ predicates10. (26) Diese Bücher sind alt/ *alte. these.neut.pl books.neut.pl are old/ *old.pl ‘These books are old.’ (27) Diese Bücher sind meine/*?mein these.neut.pl books.neut.pl are mine.pl/*mine ‘These books are mine.’
This would suggest that the “one genitive” approach may be correct for German, if all apparent predicate possessives give morphological evidence of being elliptical NPs. But it was further observed by Hans Kamp (p.c.) and others that actually, the non-agreeing form can sometimes be used. It is used only in “standard” German, not in colloquial German, and it has an “archaic” flavor. Most interestingly, it seems that there are semantic differences between the agreeing and the non-agreeing predicate possessive, and if this data stands up, it is extremely interesting. (28) a.
Diese Bücher sind meine: can be any relation. these.neut.pl books.neut.pl are mine.pl b. Diese Bücher sind mein: (archaic) “Possession” only. these-neut.pl books-neut.pl are mine (no agreement)
Further examples are given in (29) and (30). A newly naturalized citizen might say (29a), but (29b) suggests a conqueror is speaking. Any relation is possible in
106 Barbara H. Partee and Vladimir Borschev
(30a), with the most likely possibility being the parent-child, but (30b) suggests a custody fight, i.e. a dispute about who is to be in ‘possession’ of the children. (29) a.
Das Land ist (jetzt) meins. the.neut.sg land.neut.sg is (now mine.neut.sg b. Das Land ist jetzt mein. the.neut.sg land.neut.sg is now mine
(30) a.
Die Kinder sind meine. the children are mine.pl b. Die Kinder sind mein. the children are mine.
In all of (28b), (29b), (30b), the form which shows absence of agreement in the way a predicate adjective would is limited in its interpretation to “possession”. In other words, the form in which the possessive pronoun appears to be a simple predicate of type ·e,tÒ is interpreted in terms of a relation that appears to be associated with the possessive construction itself rather than with the semantics of any governing noun. In contrast, the forms which appear to be elliptical NPs have a range of interpretations including possession but also including relations typical of ‘argument’ genitives, where the relevant relation is determined11 principally by the noun to which the genitive supplies an argument. Typical choices for the ‘genitive relation’ for the ‘argument’ genitive interpretations in (28a), (29a), (30a) might be authorship, citizenship, and the parent-child relation, respectively. Of course “possession” itself can have metaphorical extensions, so the “possession” cases do not always have to be about ownership in a literal sense. But if these distinctions are correct, this is important evidence for the idea of two distinct genitives. So we are now inclined to believe that some predicate possessives really are plain ·e,tÒ predicates, and that those have just a possession/control reading, which we take to be the semantics of the ·e,tÒ possessive, as shown in (31) below. And other predicate possessives may be elliptical NPs, and their interpretation may have the full range of possibilities that would be displayed by a full NP with a prenominal genitive occurring in such a position. (Note that a full NP may itself have meanings of types e, ·e,tÒ, or ··e,tÒ,tÒ, depending on both its internal makeup and the position in which it occurs, so the study of the full range of meanings of bare possessives as elliptical NPs needs more study.) (31) [John’s]PRED:
λx[RPOSS(John)(x)]
type: ·e,tÒ
Some puzzles of predicate possessives 107
2.4 Russian and Polish possessive pronouns In Russian, in the past tense, predicate nominals may be in the Instrumental case, particularly when indicating temporary relations. Babby (1973), Siegel (1976) and others have used case and other agreement behavior to argue that some predicative adjectives are elliptical NPs and others are simple APs. The following data may provide a basis for distinguishing among predicate possessive pronouns that are and are not elliptical NPs. (32) a.
Èta strana byla kogda-to that.fem.nom.sg country.fem.nom.sg was.fem.sg once moej. my.fem.instr.sg ‘That country was once mine.’ [‘possession’ or citizenship] b. Èta strana byla kogda-to that.fem.nom.sg country.fem.nom.sg was.fem.sg once moej stranoj. my.fem.instr.sg country.fem.instr.sg ‘That country was once my country.’ [‘possession’ or citizenship]
(33) a.
Èta strana byla kogda-to that.fem.nom.sg country.fem.nom.sg was.fem.sg once moja. my.fem.nom.sg ‘That country was once mine.’ [‘possession’ only] b. *Èta strana byla kogda-to that.fem.nom.sg country.fem.nom.sg was.fem.sg once moja strana. my.fem.nom.sg country.fem.nom.sg ‘That country was once my country.’
A full predicate nominal is impossible in the nominative in the context of (33b), and in the same context, a nominative predicate possessive pronoun can be interpreted only as a possessive, not as an ‘argument’ genitive (even with a seemingly ‘free’ relation.) Thus the predicate possessive in (33a) cannot reasonably be analyzed as an elliptical NP, but must be a simple ·e,tÒ predicate, and it is this occurrence of the predicate possessive that unambiguously denotes “possession”. This data is similar to the German data, supporting the idea that there is a ‘possessive’ predicate of type ·e,tÒ instantiated at least by some possessive pronouns in German and Russian and possibly also by some
108 Barbara H. Partee and Vladimir Borschev
predicative “NP’s” forms in English, distinct from other cases of predicate possessives which are elliptical full NPs and in which the possessive may be an argument of an implicit relational noun. Wayles Browne (p.c.) suggested that we should get data on Polish, because in Polish NP – be – NP requires Instrumental on the predicate NP, whereas in Russian the predicate NP may or may not be Instrumental. And in Polish NP – be – Adj requires Nominative on the Adjective, whereas in Russian the predicate AP may be (1) short-form Adjective, (2) long-form Nominative Adjective, or (3) long-form Instrumental Adjective. The corresponding Polish data are as follows.12 (34) a.
Ten kraj był kiedys’ that.masc.nom.sg country.masc.nom.sg was.masc.sg once moim. my.masc.instr.sg ‘That country was once mine.’ [‘possession’ or citizenship] b. Ten kraj był kiedys’ that.masc.nom.sg country.masc.nom.sg was.masc.sg once moim krajem. my.masc.instr.sg country.masc.instr.sg ‘That country was once my country.’ [‘possession’ or citizenship; citizenship preferred.]
(35) a.
Ten kraj był kiedys’ that.masc.nom.sg country.masc.nom.sg was.masc.sg once mo’j. my.masc.nom.sg ‘That country was once mine.’ [‘possession’ only] b. *Ten kraj był kiedys’ that.masc.nom.sg country.masc.nom.sg was.masc.sg once mo’j kraj. my.masc.nom.sg country.masc.sg ‘That country was once my country.’ [ungrammatical] c. Ten kraj to był kiedys’ that.masc.nom.sg country.masc.nom.sg part was.masc.sg once mo’j kraj. my.masc.nom.sg country.masc.sg ‘That country was once my country.’13 [‘possession’ or citizenship]
The Polish data confirm the hypothesis that when a predicate possessive pronoun allows an “argumental” reading, it is the remnant of an elliptical NP,
Some puzzles of predicate possessives 109
and when it doesn’t, it isn’t. The “possession” reading, which seems to be emerging as the clearest case of a non-argumental, or modifier, reading, can show up either in a remnant of an NP or as a bare ·e,tÒ predicate. This reinforces the idea that a possessive inside an NP can be either an argument or a modifier. But a possessive which is an ·e,tÒ predicate in a predicational construction cannot be an argument, presumably because it is not in construction with a head of which it could be the argument.14
3. Conclusion and remaining puzzles The examination of predicate possessives15 in Section 2 leads us to the conclusion that not all predicate possessives are remnants of elliptical NPs, and that there do exist ·e,tÒ predicate possessives distinct from argument possessives. When possessives which are identified as ·e,tÒ possessives on the basis of their behavior in predicate position are also found inside an NP, it is therefore natural to assume that they are acting as modifiers, like other ·e,tÒ predicates (intersective adjectives, relative clauses, locative and other prepositional phrases). Some possessives seem to occur only in NPs and these are therefore reasonably regarded as arguments, or arguments that have been type-lifted to become modifier-like functors (which we also call ‘argument genitives’ or ‘argument possessives’ insofar as they directly or indirectly saturate an argument position in the relational CN or CNP). Argument genitives may be in construction with the head N or with a modified N such as favorite movie or former mansion (but see footnote 2). In some languages, like Russian, it may be possible to argue for a correlation of the semantic distinction with a distinction in form: we hypothesize that the prenominal possessives are modifier possessives, while true genitives are always (possibly type-lifted) arguments16. The clearest examples of predicative possessives seem to have meanings that relate to the notional concept of ‘possession’, not to the “free R” readings posited as predicative in the early work of Partee. This suggests that we should work with a different basic split: between (i) predicative or modifier readings that have a (possibly extended) meaning of “possession”, a relational notion RPOSS that comes from the possessive form rather than from a noun, and (ii) ‘argument’ readings that involve either an inherently relational noun or a plain noun coerced to a relational reading. This is thus not exactly a return to the early Partee (1983/97) analysis, but an analysis that agrees with Partee and
110
Barbara H. Partee and Vladimir Borschev
Borschev’s extensions of Jensen and Vikner’s analysis except for the recognition of a separate predicative “possessive” reading. What Partee (1983/97) analyzed as “free R” cases may be a mixture of predicative possessives of type ·e,tÒ, acting as intersective modifiers when they occur inside an NP, and argument possessives in construction with a noun which has been coerced to a relational reading by the incorporation of some lexically accessible relation (such as ‘part-whole’ or ‘created by’) or of a salient free R. There are at least two major interrelated puzzles remaining open at this point. One is how to think about the notion of “possession” that has been appealed to here, since it is evidently extremely broad. The second, mentioned at the end of Section 1, is how to think about English Saxon genitives in the light of the hypothesized split: are English genitives inside NPs ambiguous between modifier and argument? Given that Russian very often allows the same relation to be expressed either way, it would be difficult to rule out one analysis or the other of an unclear case on any semantic grounds. And for the same reason, the great semantic elasticity of interpretation shown by both constructions, this seems a very puzzling kind of “ambiguity” and very hard to get clear evidence for. In a language like English where the same form is used for both modifier and argument possessives, is one construction the dominant or default one that is used wherever there is not clear evidence for the other? Or is this just a benign instance of the fluidity of the argument–modifier shifting possibilities discussed by Dowty (1997)? Does one simply use in a given case whichever analysis requires the least type-shifting? These issues and many others remain to be explored.
Notes * The authors wish to thank many colleagues for suggestions and discussion, especially Sigrid Beck, Manfred Bierwisch, Wayles Browne, Johannes Dölling, Irene Heim, Per Anker Jensen, Hans Kamp, Tony Kroch, Godehard Link, Elena Paducheva, Ekaterina Rakhilina, Carl Vikner, Sten Vikner, Ilse Zimmermann, and an anonymous referee. The first author presented this material in October 1999 at the second annual MIT-UMass-UConn Semantics Workshop at MIT, and we thank the participants for helpful discussion. Parts of it were presented by the first author at a Graduiertenkolleg Klausurtagung held by the Linguistics Department of the University of Stuttgart in Kleinwalsertal, Austria, and by both authors in Berlin and in Munich in June 1999. This material is based upon work supported by the National Science Foundation under Grant No. BCS-9905748. Any opinions, findings, and conclusions or recommendations
Some puzzles of predicate possessives
expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. This paper is dedicated with affection to our oldest joint friend Ferenc Kiefer. 1. We use English and Russian for illustrative purposes, abbreviated below as E and R. 2. An anonymous referee suggests that one should explore a possible approach on which the genitive in (3) is a complement of the lexical adjective favorite, so that genitives, when complements, would always be complements of some lexical item. That could certainly be made to work semantically, as long as the adjective favorite is always a function applying to the noun’s meaning. As the referee notes, “it does complicate the syntax at first blush”; a fuller investigation might best be carried out in connection with a study of the interaction of genitives with superlative and superlative-like constructions as in John’s best picture, John’s first picture. 3. As with the use of free variables like xi to represent pronouns used without linguistic antecedent, we assume as a felicity condition on the use of free Ri that the context should make it sufficiently clear to the hearer what particular relation the speaker has in mind. 4. An alternative analysis of the ambiguity, based on different assumptions which we do not share, has since been offered by Larson and Cho (1999). 5. We use the term “elliptical NP” for occurrences of bare John’s or mine, etc., when they occur in NP positions and are interpreted as if a noun were present, without intending to take sides between a deletion approach and an interpretive approach to such ellipsis. 6. We thank our anonymous referee for pointing out that in Dutch, the predicate possessive in example (17c) is even more clearly an elliptical NP than in English, and that Dutch furthermore is a language which clearly distinguishes elliptical from non-elliptical predicate possessives. In Dutch, in the rendition of (17c), the d-word die, signaling the presence of nominal structure, is obligatory, as shown in (i). (i)
Die docent is *(die) van Jan. that teacher is (*that of Jan ‘That teacher is Jan’s.’
By contrast, in (ii) both options are possible. (ii)
Die auto is (die) van Jan. that car is (that of Jan ‘That car is Jan’s.’
7. In fact, if the predicate possessive in (17a) is not understood as elliptical for John’s father, the sentence is well-formed but requires a strong context that can both support a one-place predicate reading of father and some kind of ‘free’ or ‘possessive’ relation in the interpretation of John’s. A possible scenario could be a gathering of fathers and daughters and a team of reporters (including John) who among them are supposed to interview all the fathers and all the daughters. Similar remarks apply to (17b). 8. See Stockwell, Schachter and Partee (1973) for an analysis of such ellipsis and its relation to the occurrence of the pro-CNP one(s) in transformational terms. In this paper we are implicitly assuming that possessive pronouns like my/mine have the same kind of semantic
111
112
Barbara H. Partee and Vladimir Borschev
analysis as the Saxon genitive John’s, with the morphological alternation between my and mine, etc., attributable to surface syntactic factors which we can safely ignore (see Stockwell, Schachter and Partee 1973). 9. There are exceptions to the statement that genitives can never occur predicatively; these require further study. But both traditional grammars and native speakers seem agreed that the statement is basically correct. 10. Further evidence that these predicate possessives are elliptical NPs was provided by Sigrid Beck and Irene Heim (p.c.): the possessive pronoun in (27) can be followed by adjectives (i.e. there can be ellipsis of just the head noun), while the adjective in (26) and the adjective-like possessive pronoun in (28b) cannot be. (i)
(ii)
Dieser Bücher sind meine alte. these.neut.pl book.neut.pl are my.neut.pl old.neut.pl ‘These books are my old ones.’ *Dieser Bücher sind teuer neu(e). these.neut.pl book.neut.pl are expensive new ‘These books are expensive new ones.’
11. Recall from the discussion in Section 1 that the “genitive relation” in the case of an “argument genitive” may either be intrinsic to the lexical semantics of the noun, as with father, teacher, or may result from shifting the noun from a simple CN to a relational TCN by incorporating a lexically or contextually salient relation. In order to find examples that will support both ‘possessive’ and ‘argument’ genitives, we have looked for nouns for which both possibilities are reasonable, i.e. nouns which can easily be construed either relationally or as plain nouns denoting ‘possessable’ entities. Not all nouns can be expected to support both readings, and not all speakers will find both possibilities equally reasonable in the examples we have given in this and the following section. 12. Thanks to Ania Łubowicz and Anita Nowak for judgments. For (34a), Anita reports no preference for one reading or the other, while for (34b) she reports a preference for the ‘citizenship’ reading. Both rejected (35b) as ungrammatical; Ania suggested that it should be corrected to (35c), which she finds possibly ambiguous. Both agreed that (35a) is unambiguously “possession” only, whereas (34a) allows either reading. The basic judgments given above in the text for (34a,b) and (35a,b) were further confirmed by Janusz Bien, Boz˙ena Cetnarowska (and by a substantial majority of a group of 12 students of hers), Boz˙ena Rozwadowska, Piotr Banski, and Joanna Blaszczak, to all of whom we are grateful. Janusz Bien noted that similar contrasts can be gotten, possibly even more clearly, with the Polish dom, which can mean either ‘house’ (a plain CN) and ‘home’ (a TCN). Boz˙ena Rozwadowska reports that when you use the noun portret ‘portrait’, the nominative adjectival predicate moj (as in (35a)) gives a normal sentence with a ‘possession’ sense: ‘That portrait was once mine’. But when you use the instrumental moim as in (34a), the result seems nonsensical because it seems to have to mean ‘that the person in the portrait was me some time ago but that now it is somebody else.’ That judgment suggests that the instrumental form requires the argument reading; we have no explanation for why the possession reading is not an alternative possibility here as well.
Some puzzles of predicate possessives
13. There was a problem about whether or not to include this sentence in the paper. The argument is clear and simple without it, and for non-Polish speakers, including it is simply a puzzling distraction, as our anonymous referee pointed out. However, when we present our work without it to audiences that include Polish speakers, someone always brings it up, so our compromise is to include it just to show that we are aware of its existence but to leave for further study the question of its grammatical and topic–focus structure. 14. Although it might seem self-evident that a predicate separated from a subject by a copular verb cannot function as an argument to the head noun of the subject, English sentences like This photo is of John show that the situation is more complicated and requires further investigation of the sorts of constructions Grimshaw (1990) calls ‘argument adjuncts’. 15. One related bit of evidence which has come to light recently and which we have not had time to explore fully enough to discuss in this paper concerns the Russian negative possessive pronoun nicˇej ‘no one’s’. Normally a negative polarity item like other Russian ni- words (words corresponding to English n-words nothing, no one, nowhere, etc., except that Russian is a ‘multiple negation’ language), it occurs with and without accompanying sentential negation in the following minimal pair, for discovery and discussion of which we are grateful to Elena Paducheva and Ekaterina Rakhilina. Sentence (ib) is not considered fully acceptable by the second author. (i)
a.
b.
Èto nicˇej ne portret. that [∆-copula] no-one’s.masc.sg.nom neg portrait.masc.sg.nom Èto kartina. that [∆-copula] picture ‘That’s not anyone’s portrait. That’s (just) a picture.’ Èto nicˇej portret. that [∆-copula] no-one’s.masc.sg.nom portrait.masc.sg.nom Ego nikto ne kupil. it.masc.sg.acc no-one neg bought ‘That’s not anyone’s portrait. No one bought it.’
The word nicˇej here acts as a normal negative polarity item on its argument use in (ia); but when it occurs as a possessive modifier meaning ‘belonging to no one’ in (ib), it does not take the usual accompanying sentential negation ne. While a full analysis of these examples awaits further research, we hypothesize that the argument nicˇej in (ia) in the semantically relevant structure of the NP is governed by the (relational) noun, while the modifier nicˇej in (ib) expresses a predicate that has negation internal to it, and is semantically conjoined (intersected) with the noun (here a one-place predicate). Even without full analysis, such a pair would seem to further support the existence of a distinction between argument possessives and modifier possessives. 16. We have somewhat more confidence in the conclusion that Russian genitives are always arguments than in the conclusion that Russian prenominal possessives are always modifiers; the latter claim needs further research. The history of these forms merits further investigation; Richards (1976) reports that in Old Russian, the ‘possessive adjective’ forms were much more common than the use of the genitive, the opposite of the situation in modern Russian, where the possessive adjectives are much more limited in their use. Her work suggests that
113
114
Barbara H. Partee and Vladimir Borschev
many factors other than the semantics of the construction have been relevant through its history, and she cites examples from late Old Russian (XV-XVII c.) from Makarova (1954) of two-word proper names in which one of the words is in the genitive and the other in the possessive adjective form. Thus we recognize that we need to have caution in possibly trying to read too much semantics into the distinction between the two forms in contemporary Russian.
References Babby, Leonard H. 1973 “The deep structure of adjectives and participles in Russian”. Language 49: 349–60. 1997 “Nominalization in Russian”. In Formal Approaches to Slavic Linguistics: The Cornell Meeting 1995, W. Browne, E. Dornisch, N. Kondrashova, and D. Zec (eds), 54–83. Ann Arbor, MI: Michigan Slavic Publications. Babyonyshev, M. 1997 “The possessive construction in Russian: a crosslinguistic perspective”. Journal of Slavic Linguistics 5: 193–233. Borschev, Vladimir and Knorina, L.V. 1990 “Tipy realij i ix jazykovoe vosprijatie [Types of entities and their perception in language]”. In Language of Logic and Logic of Language, 106–134. Moscow. Borschev, Vladimir and Partee, Barbara H. 1998 “Formal and lexical semantics and the genitive in negated existential sentences in Russian”. In Formal Approaches to Slavic Linguistics 6 [The Connecticut Meeting 1997], Ž. Boškovic´, S. Franks and W. Snyder (eds), 75–96. Ann Arbor, MI: Michigan Slavic Publications. 1999a “Semantika genitivnoj konstrukcii: raznye podxody k formalizacii [Semantics of genitive construction: different approaches to formalization]”. In Tipologija i teorija jazyka: Ot opisanija k ob”jasneniju. K 60-letiju Aleksandra Evgen’evicˇa Kibrika [Typology and Linguistic Theory: from Description to Explanation. For the 60th birthday of Aleksandr E. Kibrik], E.V. Rakhilina and Y.G. Testelets (eds), 159–172. Moscow: Jazyki Russkoj Kul’tury. 1999b “Semantic Types and the Russian Genitive Modifier Construction”. In Formal Approaches to Slavic Linguistics 7 [The Seattle Meeting 1998], K. Dziwirek et al. (eds), 39–57. Ann Arbor, MI: Michigan Slavic Publications. Brown, Sue 1999 The Syntax of Negation in Russian: A Minimalist Approach. Stanford: CSLI Publications. Dowty, David 1997 Adjunct-to-argument reanalysis in a dynamic theory of grammar: The problem of prepositions. Paper presented at the Blaubeuren Semantics Conference, University of Tübingen, March 1997. Freeze, Ray 1992 “Existentials and other locatives”. Language 68: 553–595.
Some puzzles of predicate possessives
Heine, Bernd 1997 Possession: Cognitive Sources, Forces, and Grammaticization. Cambridge: Cambridge University Press. Hellan, Lars 1980 Toward an Integrated Theory of Noun Phrases. Doctoral dissertation, Trondheim University. Heycock, Caroline and Kroch, Anthony 1998 “Inversion and equation in copular sentences”. In Papers in Linguistics 10, A. Alexiadou, N. Fuhrhop, U. Kleinhenz, and P. Law (eds), 71–87. Berlin: Zentrum für allgemeine Sprachwissenschaft. 1999 “Pseudocleft connectivity: Implications for the LF interface level”. Linguistic Inquiry 30: 327–364. Jensen, Per Anker and Vikner, Carl 1994 “Lexical knowledge and the semantic analysis of Danish genitive constructions”. In Topics in Knowledge-based NLP Systems, S.L. Hansen and H. Wegener (eds), 37–55. Copenhagen: Samfundslitteratur. 1996 “The double nature of the verb have”. In LAMBDA 21: 25–37. Institut for Datalingvistik, Handelshøjskolen i København. Knorina, L.V. 1985 “Ob interpretacii genitivnyx konstrukcij [On the interpretation of genitive constructions]”. [Theses of the workshop “Semiotic foundations of intellectual activity”]. Moscow, VINITI. 1988 “Klassifikacija leksiki i slovarnye definicii [Lexical classification and dictionary definitions]”. In Nacional’naja specifika jazyka i ee otraženie v normativnom slovare, J. N. Karaulov (ed.), 60–63. Moscow: Nauka. 1990 “Narušenija socˇetaemosti i raznovidnosti tropov v genitivnoj konstrukcii [Violations of cooccurence and varieties of tropes in genitive construction]”. In Contradictions and Anomalies of Text, 115–124. Moscow. 1996 “The range of biblical metaphors in smikhut”. Moscow Linguistic Journal 3: 80–94. Koptjevskaja-Tamm, Maria and Šmelev, Aleksej 1994 “Alešina s Mašej stat’ja (o nekotoryx svojstvax russkix ‘pritjažatel’nyx prilagatel’nyx’) [Alješa and Maša’s article (on some properties of the Russian ‘adjectival possessives’)]”. Scando-Slavica 40: 209–228. Landman, Fred and Partee, Barbara H. 1984 Weak NPs in HAVE sentences. Draft abstract. Amherst, MA: University of Massachusetts. Larson, Richard and Cho, Sungeon 1999 “Temporal adjectives and the structure of possessive DPs”. In Proceedings of WCCFL 18, S. Bird, A. Carnie, J. D. Haugen and P. Morquest (eds), 299–311. Cambridge: Cascadilla Press. Makarova, S. Ja. 1954 “Roditel’nyj padež prinadležnosti v russkom jazyke XI-XVII vv.” [Genitive of possession in the Russian language of the 11th-17th centuries”.] In Trudy Instituta Jazykoznanija AN SSSR III: 7–31.
115
116
Barbara H. Partee and Vladimir Borschev
Montague, Richard 1973 “The proper treatment of quantification in ordinary English”. In Formal Philosophy, R. H. Thomason (ed.), 247–270. New Haven: Yale University Press, 1974. Moro, Andrea 1997 The Raising of Predicates. New York: Cambridge University Press. Partee, Barbara H. 1986 “Ambiguous pseudoclefts with unambiguous be”. In Proceedings of NELS l6, S. Berman, J.-W. Choe, J. McDonough (eds), 354–366. 1987 “Noun phrase interpretation and type-shifting principles”. In Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers, J. Groenendijk, D. de Jongh, and M. Stokhof (eds), 115–143. Dordrecht: Foris. 1983/1997 “Uniformity vs. versatility: the genitive, a case study”. In The Handbook of Logic and Language, J. van Benthem and A. ter Meulen (eds), 464–470. Amsterdam: Elsevier. 1999a “Copula inversion puzzles in English and Russian”. In Formal Approaches to Slavic Linguistics 7 [The Seattle Meeting 1998], K. Dziwirek et al. (eds), 361–395. Ann Arbor, MI: Michigan Slavic Publications. Also in UMOP 23: Issues in Semantics and its Interface, K. Kusumoto and E. Villalta (eds), 183–208. Amherst: GLSA, 2000. 1999b “Weak NP’s in HAVE sentences.” In JFAK [a Liber Amicorum for Johan van Benthem on the occasion of his 50th Birthday; CD-ROM], J. Gerbrandy, M. Marx, M. de Rijke, and Y. Venema (eds). Amsterdam: University of Amsterdam. Available: http://turing.wins.uva.nl/~j50/cdrom/ Partee, Barbara H. and Borschev, Vladimir 1998 “Integrating lexical and formal semantics: Genitives, relational nouns, and type-shifting”. In Proceedings of the Second Tbilisi Symposium on Language, Logic, and Computation, R. Cooper and T. Gamkrelidze (eds), 229–241. Tbilisi: Center on Language, Logic, Speech, Tbilisi State University. Pustejovsky, James 1993 “Type coercion and lexical selection”. In Semantics and the Lexicon, J. Pustejovsky (ed.), 73–94. Dordrecht: Kluwer. 1995 The Generative Lexicon. Cambridge, MA: MIT Press. Rappaport, Gilbert 1998 “The Slavic noun phrase”. [Position paper for Comparative Slavic Morphosyntax]. Available: http://www.indiana.edu/~slavconf/linguistics/ download.html [1999, Jan. 4] Richards, Karen Rondestvedt 1976 “Toward a history of the possessive in literary Russian: The demise of the possessive adjective”. In Papers from the Parasession on Diachronic Syntax, Chicago Linguistic Society, S.B. Steever et al. (eds), 260–73. Chicago: Chicago Linguistic Society. Schoorlemmer, Maaike 1995 Participial Passive and Aspect in Russian. Doctoral dissertation, OTS [Utrecht Institute of Linguistics]: Utrecht.
Some puzzles of predicate possessives
Siegel, Muffy 1976 “Capturing the Russian adjective”. In Montague Grammar, B. Partee (ed.), 293–309. New York: Academic Press. Stockwell, Robert P., Schachter, Paul and Partee, Barbara H. 1973 The Major Syntactic Structures of English. New York: Holt, Rinehart and Winston. Szabolcsi, Anna 1994 “The Noun Phrase”. In Syntax and Semantics, Vol. 27 [The Syntactic Structure of Hungarian], F. Kiefer and K. É. Kiss (eds), 179–275. New York: Academic Press. Vikner, Carl and Jensen, Per Anker 1999 A semantic analysis of the English genitive: Interaction of lexical and formal semantics. Manuscript. Copenhagen and Kolding. Williams, Edwin 1981 “Argument structure and morphology”. Linguistic Review 1: 81–114. 1983 “Semantic vs. syntactic categories”. Linguistics and Philosophy 6: 423–446.
117
AUTHOR "Zoltán Gendler Szabó"
TITLE "Adjectives in context"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Adjectives in context* Zoltán Gendler Szabó The Sage School of Philosophy, Cornell University
1.
The Context Thesis
One of the fundamental assumptions of contemporary semantic theory is the principle of compositionality, the thesis that the meaning of a complex expression is determined by the meanings of its constituents and by the manner in which these constituents are combined. When challenged, defenders of the principle tend to appeal to the undeniable fact that competent speakers routinely understand complex expressions they have never heard before. This fact, they contend, can only be adequately explained if we assume that linguistic competence requires implicit grasp of a recursive system of semantic rules which assign meanings to complex expressions on the basis of the meanings of their constituents and the way those constituents are combined. But this is much too quick. Nobody denies that the meaning of a complex expression depends on the meanings of its constituents and on its structure; the bite of compositionality is that it depends on nothing else. The appeal of the principle stems from the conviction that those who can understand a new expression must work out its meaning from facts they knew about the expression antecedently. And of course, the only candidate facts are lexical or syntactic. But this piece of reasoning neglects the possibility that the context in which someone encounters a new expression can provide additional clues relevant for interpretation. And if it does, there is no reason to assume that the lexical and syntactic facts by themselves fix the meaning of the new expression. So at the very least, attention to context undermines the unreflective assumption that compositionality is a trivial matter.1 Let us make our terminology more precise. The word ‘meaning’ is used in semantics to refer to all sorts of semantic values; in what follows, I want to focus on a particular semantic value often called ‘content’. The content of a declarative
120 Zoltán Gendler Szabó
sentence is the proposition it expresses and the contents of sub-sentential expressions are the semantic contributions of such expressions to propositions expressed by declarative sentences within which they occur as constituents. There is no general agreement concerning the nature of propositions; the only thing I will assume about them is that they determine truth-conditions, that is, that if two sentences express the same proposition they cannot differ in truthvalue. Semantic theories typically aim at ascribing content to complex expressions in a compositional fashion, hence they are committed to the principle of compositionality of content.2 I take the context of utterance to be a wide and heterogeneous collection of facts concerning the linguistic and non-linguistic environment of a particular use of an expression. It includes facts about the time and the location of the utterance, facts about the speaker, the hearer, and the salient objects around them, facts about their shared background knowledge, about the form and content of the conversation they had before the utterance in question was made, and perhaps much more. There are several important foundational questions about contexts, but for the purposes of this paper I wish to remain more or less neutral with regard to these.3 I will consider the context of an utterance simply as the sum total of all the non-lexical and non-syntactic facts relevant for determining what a speaker conveys by the utterance. Let us say that an expression is context-dependent if and only if it can have different contents when uttered in different contexts. A context-dependent complex expression none of whose constituents are context-dependent would then be a counterexample to the principle of compositionality of content. Is there such a counterexample? There are many cases when we seem to be presented with context-dependent expressions, but in fact we are not. Such cases come in two varieties, which I will call phonetic and pragmatic illusions of context-dependence. Phonetic illusions of context-dependence occur when we have different expressions articulated in the same way. Consider the following two sequences of sentences: (1) a.
I should stop digging this hole in the bottom of this ship, but I cannot resist the temptation. Sinking boats can be entertaining. b. I should stop watching sea battles on video, but I cannot resist the temptation. Sinking boats can be entertaining.
According to their most plausible interpretations, (1a) is about my activity of sinking boats, while (1b) is about boats that are sinking. So it seems that the content of ‘Sinking boats can be entertaining’ varies depending on the larger linguistic context in which it is used. But this is not so. The illusion that we have
Adjectives in context
expressions with multiple contents arises because our ordinary notation is ambiguous: ‘ ‘Sinking boats can be entertaining’ ’ refers to two different sentences.4 The justification for the claim that we are dealing with two different expressions here is syntactic: the linear order of ‘sinking’ and ‘boats’ disguises glaring structural differences. Pragmatic illusions of context-dependence may arise when utterances of an expression convey different things in different contexts. Form this, one might be tempted to conclude immediately that the expression says different things in different contexts. But this does not follow. Consider these sequences: (2) a.
I could not solve the crossword puzzle, even after I wasted the whole weekend on it. The puzzle was extremely difficult. b. I solved the crossword puzzle in five minutes while cooking dinner and listening to the radio. The puzzle was extremely difficult.
It is quite plausible to take ‘The puzzle was extremely difficult’ as ironical in (2b), but not in (2a). Still, there is no good reason to postulate that this sentence expresses different propositions in these two cases; its content is invariably the proposition that the puzzle was extremely difficult. But since this proposition is in conflict with the content of the first sentence of (2b), a sensible hearer will conclude that in uttering the second sentence of (2b) the speaker did not mean to convey what he literally said. We have strong reasons to think that irony is a pragmatic matter: there is no foolproof conventional mark indicating that an utterance is ironical. Moreover, we couldn’t even introduce such a conventional sign. For suppose we agreed that, say, a winking of the left eye signals that the utterance must be interpreted ironically; those who wanted to be genuinely ironical would certainly refrain from winking in making their remarks. Despite the prevalence of phonetic and pragmatic illusions, cases of genuine context-dependence are not hard to come by. Consider (3) and (4), for example: (3) Some basketball players are giants. (4) I am now at the North Pole.
What (3) and (4) say is fixed in part by the context of utterance. If a giant is supposed to be a “legendary being of great stature and strength and of more than mortal but less than godlike power” then (3) is false, but if it is merely “a living being of great size”, it is true.5 Since the fact that ‘giant’ can have different contents is clearly marked in the lexicon, (3) is not a likely candidate for a pragmatic illusion of context-dependence. Saying that there are two words in English — ‘giant1’ and ‘giant2’ — and that consequently we are dealing with
121
122 Zoltán Gendler Szabó
a phonetic illusion of context-dependence in (3) would also be implausible.6 Neither semantic intuition, nor etymology would support such a claim. A speaker who never heard ‘giant’ be used in referring to people of extraordinary size would probably still be able to figure out how to understand (3); by contrast a speaker who doesn’t know that ‘bank’ can be used to refer to the edges of rivers will not know what to make of ‘I closed my account in the bank.’ ‘Giant’ is a polysemous expression, i.e. ‘‘giant’’ refers unambiguously to ‘giant’, but this lexical item has multiple closely associated contents.7 The case of (4) is even more straightforward: different people can say different things at different times uttering this sentence. I have never been in a position to use it to express a truth, but I might be in the future and others have been in the past. ‘I’ and ‘now’ are indexical expressions; their contents are fixed in accordance with a linguistic rule relative to the context of use. And the context-dependence of indexicals spreads to the sentences that contain them as constituents. Neither polysemy nor indexicality is a threat to compositionality: (3) and (4) have multiple contents because they contain constituents whose contents depend on the context. To find a compelling counterexample to compositionality we would need a complex expression whose content changes with the context in which it is used even though the contents of all of its constituent expressions remain fixed. Whether or not there is such a counterexample, coming up with one is not a trivial matter. I will call the claim that there is no such counterexample the context thesis: Context Thesis: The content of an expression depends on context only insofar as the contents of its constituents do. According to the context thesis the context-dependency of a complex expression must be indirect. Once one has fixed what content the simple constituents have in a given context, the way in which the content of the complex expression depends on the context is also fixed. In the next section, I will consider an argument against the context thesis, based on the semantic behavior of certain complex nominal expressions containing adjectives.
2. The color of a painted leaf Charles Travis has argued that many different things might be said with the same words, even if the subject matter of the utterances remains the same:
Adjectives in context
As an arbitrary example, consider the words ‘The leaf is green’, speaking of a given leaf, and its condition at a given time, used so as to mean what they do mean in English. How many distinct things might be said in words with all that true of them? Many. That emerges when we note that one might speak either truth or falsity in such words, if the leaf is the right way. Suppose a Japanese maple leaf, turned brown, was painted green for a decoration. In sorting leaves by colour, one might truly call this one green. In describing leaves to help identify their species, it might, for all the paint, be false to call it that. So words may have all the stipulated features while saying something true, but also while saying something false. Nothing about what it is to be green decides whether the colour of a thing is the way it is with, or the way it is without the paint. What being green is is compatible with speaking either truth or falsity in calling the leaf green. For all that, the painted leaf is as it is sometimes, but not other times, said to be in calling it green.8
What the example shows is subject to debate. Let me start with what is relatively uncontroversial. First, the truth-value of an utterance of ‘The leaf is green’ made about the painted leaf depends on the context. Roughly, the utterance is true if it is made when one is sorting leaves for decoration, and false if it is made when one is trying to identify the species of the leaf. Second, the sequence of phonemes uttered in the two scenarios stand for the same unique sentence, i.e. ‘‘The leaf is green’’ refers unambiguously to the sentence ‘The leaf is green’. Third, the words ‘leaf ’ and ‘green’ are used in both cases “so as to mean what they do mean in English”, which is presumably something fixed. Consequently, we have a single sentence which says different things in different contexts, even though the sentence is neither syntactically nor lexically ambiguous. Is this a challenge to the context thesis? It depends. ‘Meaning’ is an elusive phrase and it is not entirely clear how Travis intends to use it here. If by ‘meaning’ we mean the conventions of use for an expression knowledge of which guarantees understanding then the fact that sentences with a fixed meaning can say different things is no surprise. Any sentence containing an indexical has a single meaning in this sense even though it can expresses a plethora of different propositions. But suppose ‘meaning’ is taken to mean the contribution an expression makes to the proposition expressed by sentences containing the expression. Then the claim that ‘The leaf is green’ can express different propositions even though the words this sentence is constructed from have a fixed meaning is in direct conflict with the context thesis. The obvious reply to this challenge is to reject the claim that the constituent expressions have identical contents in the two utterances. One idea would be to say that this is a case of referential indeterminacy: ‘the leaf ’ can denote the leaf
123
124 Zoltán Gendler Szabó
with or without the paint. There is some plausibility to this idea, given that we would be strongly inclined to call both of those entities leaves. No doubt, this line gives rise to some metaphysical puzzles (How come we have two leafs roughly at the same spatio-temporal location? Do we really create a new leaf in painting one? If we remove the paint is the leaf with paint destroyed?), but the solution is implausible independently of these puzzles. For if in some contexts ‘the leaf ’ denotes what is below the paint, then presumably in those contexts ‘The leaf is unpainted’ is true. But this is unacceptable: there is no context in which one can hold up a painted leaf and say truthfully ‘The leaf is unpainted’. Clearly, it is better to locate the context-dependence in the adjective. One might say that there is no such thing as the content of ‘green’, just as there is no such thing as the content of ‘giant’ or the content of ‘I’. What it is to be green varies with the context. But there are two problems with this reply. The first is that according to Travis the example was merely illustrative, and the phenomenon has nothing to do with specificities of ‘green’. The second is that ‘green’ does not seem to be a polysemous or indexical word. Let us look at these problems in more detail. According to Travis, the example of painted leaf can be easily generalized. In fact, he thinks the moral applies to “whatever else words speak of”: The words ‘is green’, while speaking of being green, may make any of indefinitely many distinct contributions to what is said in words of which they are part. The above variation is illustrative. The same holds of any English predicate. The fact that ‘is green’ speaks of being green does not alone decide what is required for a thing to be as it, on a speaking, says a thing to be. Similarly for whatever else words speak of.9
I think this is an overstatement. If one uses the sentence ‘The number is even’, talking about the number four, the sentence expresses a truth. One cannot construct some special scenario where due to some camouflage similar to the painting of the leaf ‘The number is even’ says something false. There is a trivial way that the example can be generalized to practically all non-mathematical and non-logical cases. Take the sentence ‘The book is a novel’. There are all sorts of borderline instances of novelhood, like longish short-stories, modern epics written in free verse, etc. Let the phrase ‘the book’ be used to denote some such entity. Then there are scenarios where the sentence is true according to the contextual norm, and others where it is false. This is not especially interesting, since it does not threaten the context thesis: intuitively, whenever the content of ‘… is a novel’ is vague in this way, it is because the content of ‘novel’ is vague. What is challenging about the example of the
Adjectives in context
painted leaf is that it is independent of problems of vagueness. Even if a brown maple leaf is painted the most paradigmatic green one can imagine, the problem whether the leaf is green remains. If we neglect vagueness, however, Travis’s example is not widely generalizable. There are indeed different scenarios where utterances of the sentence ‘The ring is gold’ are about the same object, but have different truth-values. If the ring contains silver besides pure gold the utterance can be true in ordinary circumstances and false in a laboratory. But speaking of a ring made of pure gold, ‘The ring is gold’ expresses a truth, no matter what the speaker thinks, or what the purpose of her statement is. So, we can conclude that the kind of context-dependency that we see in ‘The leaf is green’ is not universal: ‘The number is even’ or ‘The ring is gold’ are rather similar sentences that do not have it. Of course, it is equally obvious that the example is generalizable to many other adjectives, besides ‘green’. It is quite easy to come up with parallel cases for ‘The apple is red’, ‘The move is smart’, ‘The problem is interesting’, or ‘The soup is good’. The question is then, what is responsible for this limited phenomenon? This brings us to the second problem concerning the suggestion that ‘green’ is a context-dependent adjective. One can articulate the intuition behind Travis’ example thus: “Of course, the challenge can be countered by saying that ‘green’ expresses one content in one scenario and another content in the other. But intuitively, this is not the case. What it contributes to a proposition expressed by ‘The leaf is green’ in either scenario is the same: the property of being green. ‘Green’ is not a polysemous expression and it is not an indexical either. To stipulate that it is nevertheless context-dependent requires that we introduce context-dependence of a new, hitherto unknown sort. And this seems like an ad hoc maneuver.” If we look up ‘green’ the Webster Dictionary, what we find — alongside the non-literal meaning specifications, which are irrelevant here — is the barely exciting information: something is green if it is “of the color green”. This fact is indeed in conflict with both the assumptions that ‘green’ is polysemous and that it is an indexical. If ‘green’ were a polysemous word like ‘giant’, we would expect the dictionary to contain different numbered sub-entries specifying the different contents the word can have on different occasions of its use. If ‘green’ were an indexical like ‘I’, we would expect the dictionary to contain an informative clause which tells us how to select its content in a given occasion of its use. If ‘green’ is context-dependent, its context-dependence is of a different kind.
125
126 Zoltán Gendler Szabó
The question is whether a convincing story can be told about how the content of ‘green’ and other adjectives is supposed to depend on the context. If there is no such story, the sentence ‘The leaf is green’ refutes the context thesis and provides a genuine counterexample to the principle of compositionality of content. But I don’t think the situation is as bad as all that. In the rest of this paper, I will present the outlines of an account of the semantics of a large class of context-dependent adjectives. And I will argue that this class includes ‘green’.
3. Problems with ‘Good’ It is a curious fact that contemporary views on the semantics of adjectives evolved from a debate that started in moral philosophy. In their effort to articulate theories about what goodness consists in, philosophers turned to questions about the semantics of ‘good’. ‘Good’ is an adjective with some peculiar characteristics, so any analysis of its content had to say something about the interpretation of adjectives in general as well as about the semantic features that distinguish ‘good’ from less problematic adjectives, like ‘round’ or ‘tall’. Like most adjectives, ‘good’ occurs in predicative as well as attributive positions. The former cases suggest that ‘good’ is a predicate, while the latter that it is a predicate-modifier. Which of these indications is to be taken seriously is a question that might affect the way in which we make sense of judgments to the effect that an act, a person, or a state of affairs is good. There is a certain difficulty about pronouncements that ‘good’ is a predicate, or that it is a predicate-modifier. Categories like ‘noun’, ‘verb’, ‘determiner’, ‘pronoun’, ‘adjective’, or ‘sentence’ are syntactic, and categories like ‘predicate’, ‘connective’, ‘quantifier’, ‘predicate-modifier’, ‘variable’, or ‘formula’ are logical. These logical categories, though they are syntactic categories of certain formalized languages, are certainly not syntactic categories of English.10 But what is it supposed to mean that an expression of English belongs to a syntactic category of some other language? One natural way to understand such claims is to regard them as being about logical forms. I take the logical form of an expression e in a natural language L to be an expression e¢ in a suitable formalized language L¢ which has the same logical properties as e. If L¢ has a clear, well-understood semantics e¢ can be used to illuminate and explain the inferential behavior of e in L.11 This means that an expression could have more than one logical form. If we can choose among the competing logical forms one that explains the inferential
Adjectives in context 127
behavior of an expression in the most perspicuous way, we can call this the logical form of the expression. To call ‘good’ a predicate (or to say that the semantic category for ‘good’ is that of predicate) is then to say that in the logical forms of English sentences containing the word ‘good’ we always find the same corresponding predicate. Similarly, to call it a predicate-modifier is to claim that in the logical forms of English sentences containing ‘good’ we always find the same corresponding predicate-modifier. These claims can be formulated as (5) and (6):12 (5) The logical form of ‘good’ is the predicate ‘good(x)’ (6) The logical form of ‘good’ is the predicate-modifier ‘(good(F))(x)’ (Here ‘x’ is an individual variable and ‘F’ is a predicate variable.)
G. E. Moore (1903) claimed that ‘good’ is a predicate that expresses a simple property. Certainly, there is a prima facie plausibility in the idea that ‘good’ expresses goodness, like ‘hard’ hardness or ‘green’ greenness. If there is no reason to think otherwise, we can suppose that expressions in the same grammatical category have the same kind of contents. A semantic analysis based on Moore’s suggestion regards the predicative occurrences of ‘good’ as paradigm cases and treats sentences containing attributive occurrences as complexes built up from simpler sentences in which there are only predicative occurrences. According to Moore, something is a good book if and only if it is good and it is a book.13 One way to implement these ideas is to say that the logical form of (7) is (7a), where the language of logical forms is a formalized extensional first-order language: (7) Sue is a good dancer. (7a) dancer(Sue) Ÿ good(Sue)
There is an argument due to Peter Geach (1956), which shows that this simple Moorean analysis cannot be the whole story about ‘good’. The argument is based on the observation that inferences of a certain type are incorrectly predicted to be valid by the Moorean analysis. Let us call an inference a transferinference, just in case it fits the following pattern: its first premise is of the form ‘a is AN’, its second premise has an entailment of the form ‘a is M’ and its conclusion is of the form ‘a is AM’. Moore’s analysis is in trouble with the following simple transfer-inference:14 (7) Sue is a good dancer. (7a) dancer(Sue) Ÿ good(Sue) (8) Sue is a pianist.
(8a) pianist(Sue)
(9) Sue is a good pianist. (9a) pianist(Sue) Ÿ good(Sue)
128 Zoltán Gendler Szabó
The inference on the right is trivially valid, but according to its most natural interpretation, the inference on the left is invalid; Sue might be a good dancer and a pianist without being a good pianist. The fact that ‘good’ fails to validate certain transfer inferences is regularly referred to in the literature as the nontransparency of this adjective.15 The non-transparency of ‘good’ shows that at least one of the Moorean translations from the English sentences to the logical formalism must be incorrect. Geach’s suggestion was to give up the Moorean assumption that ‘good’ is a predicate and regard it instead as a predicate-modifier. On this approach, the attributive occurrences are the paradigm cases and the predicative occurrences must be explained away. To give an account of them, one has to assume that they are elliptical attributive occurrences. Thus the sentence ‘These gloves are good’ is usually to be understood as ‘These gloves are good gloves’. Although this is a natural interpretation of the sentence, in certain situations it is incorrect. For example, if the sentence is uttered by someone who wants to clean the windows and is looking for some piece of cloth appropriate for that purpose, the suggested analysis would fail. In this situation the sentence is to be understood as saying ‘These gloves are good pieces of cloth for cleaning the windows’. According to Geach something is good if and only if there is an appropriate noun N such that that thing is a good N.16 One way to build Geach’s idea into a semantic analysis would be to assign (7b) to (7) as its logical form, where the language of logical forms remains extensional. Here the semantic value of a predicate-modifier is a function that maps predicate extensions to predicate extensions. (7) Sue is a good dancer. (7b) (good(dancer))(Sue)
This translation is not subject to the previous criticism. From the assumption that the complex predicate ‘(good(dancer))(x)’ applies to the bearer of the name ‘Sue’ it does not follow that the complex predicate ‘(good(pianist)) (x)’ applies to her. So Geach’s move blocks the transfer-inference from (7) and (8) to (9). In assigning (7b) as the logical form to (7), this analysis provides an implicit explanation why the inference is invalid. It goes like this: The inference is invalid, because ‘(good(dancer))(x)’ and ‘(good(pianist))(x)’ have different extensions, which, in turn, is true because ‘dancer(x)’ and ‘pianist(x)’ have different extensions. That is, Sue can be a good dancer and a pianist, and still fail to be a good pianist, because there are people who are dancers and not pianists or vice versa. Now, this explanation cannot be right. For, the inference would be
Adjectives in context 129
invalid even if we added the assumption that all and only dancers are pianists. (Of course, if we add this, we can remove the premise ‘Sue is a pianist’.) (7) Sue is a good dancer.
(7b) (good(dancer))(Sue)
(10) All and only dancers are pianists. (10b) "x(dancer(x) ´ pianist(x)) (9) Sue is a good pianist.
(9b) good(pianist))(Sue)
Geach’s analysis blocks one transfer-inference but it falls prey to another. Hence, it is not an adequate account of the non-transparency of ‘good’. Richard Larson has suggested a more promising theory about ‘good’ and a number of other problematic adjectives.17 His idea is to return to Moore’s idea that ‘good’ is a predicate in (7), but abandon the assumption that it is a predicate that applies to Sue.18 What is being said to be good is Sue’s dancing, not Sue herself. The logical form of the relevant reading of (7) could be captured as (7c):19 (7) Sue is a good dancer.
(7c) Qe[dancing(e, Sue)](good(e))
(Here, ‘e’ is an event variable and ‘Q’ is a generic quantifier. The different brackets in the notation are intended to express that ‘dancing(e, Sue)’ is in the restrictor of the generic quantifier and ‘good(e)’ is in its nuclear scope.) A good paraphrase of (7c) would be ‘Generally dancing events performed by Sue are good.’ Given a Davidsonean analysis of adverbs, we get the same logical form for the sentence ‘Sue dances well’. Prima facie, this is good news: the relevant reading of ‘Sue is a good dancer’ seems indeed synonymous with ‘Sue dances well’. How does the introduction of event variables block the inference from ‘Sue is a good dancer’ to ‘Sue is a good pianist’? Even if dancers and pianists were the exact same people, their dancing and piano-playing performances would still be distinct events. Consequently, it can easily be true that Sue’s dancing performances are excellent, while her piano performances are terrible. That is, ‘Qe[dancing(e, Sue)](good(e))’ might be true and ‘Qe[piano-playing (e, Sue)](good(e))’ false. But the problem of non-transparency is still with us. Imagine a world where dances are performed with a single aim: to persuade the gods to give rain for the crops. Furthermore, dancing is the only method people ever use to try to persuade the gods to give rain for the crops. In this world, not only are the dancers and the rainmakers the same people, dancing and rainmaking are the very same events.20 Now, suppose Sue is a shaman-dancer-rainmaker in this world. She had perfected her dancing and she is much admired for her graceful
130 Zoltán Gendler Szabó
moves. Unfortunately, she is not particularly successful in making rain: year after year, despite her best efforts there is a terrible drought. It seems to me that under such circumstances we could say that Sue is a good dancer but not a good rainmaker. But now we have a problem: (7) Sue is a good dancer.
(7c)
Qe[dancing(e, Sue)](good(e))
(11) All and only dances are rain-makings.
(11c)
"e(dancing(e) ´
(12) Sue is a good rainmaker.
(12c)
rainmaking(e)) Qe[rainmaking(e, Sue)](good(e))
The inference on the left is invalid,21 but the corresponding inference run on the putative logical forms on the right is valid. So, Larson’s logical forms cannot be correct either. There is a standard way to get around the problem of non-transparency. The idea is that the logical forms should be drawn from an intensional language.22 The interpretation of an intensional language distinguishes between (at least) two different semantic values of expressions: the extension and the intension. The extension of the expression is its ordinary semantic value (a truth-value for sentences, an object for names, a set of objects for one-place predicates, a function from sets to sets for predicate-modifiers, etc.); the intension of the expression is a function from possible worlds to its extension. One can have an intensional language where — although there are certain operators that are sensitive to the intensions of the expressions to which they are applied — quantification over possible worlds remains covert. Languages of modal logics are intensional languages of this type. Alternatively, one can have an intensional language, where reference to possible worlds is explicit. In these languages there are variables ranging over possible worlds.23 I think intensional languages of the second type are more perspicuous, so I will use this technique to present the intensional solution to the problem posed by the transfer-inferences.24 The intensionalized version of Geach’s proposal is (7d): (7) Sue is a good dancer. (7d) (good(dancer))(Sue)(w)
In the logical form, ‘w’ is a variable that will be assigned the possible world with respect to which the sentence is evaluated. In most cases this is the actual world, so (7d) is true iff Sue is among the individuals that are good dancers in the actual sworld. But (7) can also occur as a constituent in a complex sentence with modal operators, which are translated as quantifiers binding the possible world variable.
Adjectives in context
Such a move blocks the transfer-inference from (7) and (10) to (9). Even if the predicates ‘dancer(x)’ and ‘pianist(x)’ have the same extensions in the actual world, they have different extensions in other possible worlds. Then the intensions of the complex predicates ‘good(dancer)(x)’ and ‘good(pianist)(x)’ are different functions, and therefore the expressions ‘(good(dancer))(Sue)’ and ‘(good(pianist))(Sue)’ have different intensions too. But then there is no reason why the sentences ‘Sue is a good dancer’ and ‘Sue is a good pianist’ should have the same truth-value in the actual world. One might wonder whether the intensional move solves the problem. What if pianists and dancers are necessarily the same individuals? Doesn’t intuition tell us that even under such circumstances Sue could be a good dancer and a bad pianist? It is hard to tell. Our intuitions might be confused here, for it is not clear whether the assumption that necessarily all and only pianists are dancers is conceptually coherent.25 I will not pursue this line of objection for there is a more convincing one. By assigning (7d) as logical form to (7), we implicitly commit ourselves to an explanation why the inference fails. The reason why Sue may be a good dancer and a pianist without being a good pianist even if all dancers are pianists and vice versa is that there is another possible world where someone (she, or someone else) is a dancer, but not a pianist, or a pianist, but not a dancer. This does not sound very convincing. How could the possibility that someone is a dancer, but not a pianist have anything to do with the question whether Sue has to be a good pianist, given that she is a good dancer and a pianist? It is interesting to compare this with the following case. The inference from ‘Sue is a prospective dancer’ and ‘All and only dancers are pianists’ to ‘Sue is a prospective pianist’ is also invalid. Here the intensional explanation says that the reason why Sue may be a prospective dancer and a pianist without being a prospective pianist even if all dancers are pianists and vice versa is that there is another possible world where someone (she, or someone else) is a dancer, but not a pianist, or a pianist, but not a dancer. In this case the explanation is much more plausible. The failure of this inference can be explained by means of possibilities, because if someone is a prospective N, then she may become, or is likely to become an N. However, similar modal consequences cannot be drawn from the claim that someone is a good N.26 We are in a quandary. Extensional approaches to the adjective ‘good’ seem inadequate because they cannot account for its deep non-transparency. They all incorrectly predict that if the extensions of ‘N’ and ‘M’ are sufficiently closely tied, the inference from ‘a is a good N’ to ‘a is a good M’ goes through. The
131
132
Zoltán Gendler Szabó
failure of the extensional accounts naturally suggests that we should regard ‘good’ as an intensional adjective. However, the explanation of non-transparency gained from the introduction of intensions appears to be the wrong one. The failure of the transfer inferences has nothing to do with modal considerations and so, by going intensional we avoid wrong predictions only through sacrificing explanatory value. A different approach is called for.
4. Ways of being good On the surface the answer to the question why someone may be a good dancer and a bad pianist is quite obvious. Dancing and playing the piano are very different skills, so there is no reason why excellence in one should have anything to do with excellence in the other. In other words, if Sue is a good dancer then she is good at dancing, which is perfectly compatible with her being quite bad at playing the piano. So, goodness — at least in these cases — does not directly attach to Sue; it attaches to her only through one or another description that is true of her. Some of these descriptions, as that she is a dancer or that she is a pianist, may specify certain roles, and she may be skillful, interesting, enthusiastic or well-paid in one of her roles, but unskillful, boring, indifferent or badly paid in another. Being good at dancing and being good at playing the piano are different ways of being good.27 This seems to be the correct analysis of ‘good’ and many other adjectives. As it stands, this explanation is in accordance with the analysis that regards ‘good’ as a predicate-modifier. However, theories that regard ‘good’ as a predicate-modifier tend to make a further step: they typically assume that the relevant roles are fully specified by the noun to which the adjective ‘good’ is attached. As Geach himself noted, there are cases in which this move is problematic. The difficulties are exemplified by sentences in which one speaks about a good event, or a good thing to happen. Geach says that such sentences often do not have a fixed content, since “ ‘event’, like ‘thing’, is too empty a word to convey either a criterion of identity or a standard of goodness.”28 This means that according to Geach, the reason why ‘good event’ does not have clear truthconditions is that ‘event’ does not have a clear content. But, surely, the content of ‘pebble’ or ‘differential-equation’ are clear enough, and still, one does not know what to make of ‘good pebble’ or ‘good differential-equation’.29 The content of the sentence ‘This is a good pebble’ depends on the context of its use. In a certain situation, an utterance of this sentence might say that the pebble is
Adjectives in context
good for playing marbles, in another that it is good for breaking a window. So the noun N in a phrase ‘a is a good N’ is often not sufficient in determining the role in which a is supposed to be good. Is it at least always necessary? Geach certainly thinks so, since he believes that in order to interpret ‘a is good’, one always has to provide some suitable N that will yield the standards of goodness. I think this is implausible. Consider the scenario where students are performing an experiment in the laboratory. They are trying to produce some substance that can be used in a later experiment. The teacher points to a certain blue liquid and says to the students: ‘This is good’. Is it really true that in order to interpret this, the students have to be able to come up with some noun N such that the blue liquid is a good N? It seems that even if they know close to nothing about what the blue liquid is, and hence the nouns they could come up with do not provide standards of goodness, they can know what the teacher’s utterance said. The blue liquid was said to be good for the purposes of the experiment. Of course, Geach might reply that the noun (or, more precisely, the nominal expression) in question is something like ‘stuff that can be used in a later experiment’. But why should we believe that the standards of goodness are provided by the content of this complex noun, rather than simply by the context in which the utterance was made? Here is another problem with the predicate-modifier approach. Imagine a case when students are arguing about what to read next in their reading group. There are two candidates: a short book and a lengthy paper. One of the students points to the book and says: ‘This is a good book’. Another, as a reply, points to the paper and says: ‘This is good too’. What the second student said can be paraphrased roughly as ‘This is good to read too’; it cannot be paraphrased as ‘This is a good book too’. But standard accounts of ellipsis assume that the second sentence predicates something of the paper that is predicated of the book in the first. Given the predicate-modifier analysis of ‘good’, a suitable predicate is simply unavailable.30 All of this suggests strongly that the content of the noun ‘N’ does not play a direct semantic role in fixing the way in which a good N is said to be good. I suggest that we should return to the Moorean idea that the logical form of ‘good’ is a one-place predicate of individuals. Of course, there is a price for this: one certainly cannot hold a theory according to which the extension of ‘good’ is the class of good things. There may be nothing in common to all of the good things. However, one can say that ‘good’ is an incomplete one-place predicate, one that is associated with a set of individuals only if additional information is provided. ‘Good’ can be completed in many different ways; for example under
133
134 Zoltán Gendler Szabó
one completion it is associated with a set of those individuals who are good at dancing, under another with the set of those individuals who are good at playing the piano. The analysis I am suggesting follows Moore in the sense that it regards the predicative occurrences of ‘good’ as the paradigm cases: ‘good’ is an incomplete expression, its semantic category is one-place predicate. I also accept Moore’s analysis of the attributive occurrences: Sue is a good dancer if and only if Sue is good (as a dancer) and Sue is a dancer; Sue is a good pianist if and only if Sue is good (as a pianist) and Sue is a pianist. So the translation I suggest for (7) is (7e): (7) Sue is a good dancer. (7e) dancer(Sue) Ÿ (good(R))(Sue)
‘R’ is a variable standing for a certain role31 in which something can be good. The value of this variable is fixed by the context in which (7) is used.32 Building variables into the lexical representation of ‘good’ may prompt the reaction that we are no longer discussing the semantics if ‘good’ and sentences containing ‘good’, we switched to the pragmatics of such expressions. The terms ‘semantics’ and ‘pragmatics’ are used in a variety of ways and in some of these senses this is undoubtedly true. But if we think that the minimal aim of semantics is to provide a systematic theory of the truth-conditions of assertive uses of well-formed declarative sentences then truth-conditional effects of context are properly regarded as lying within the domain of semantic theory. This, however, does not preclude the possibility that the mechanism used to determine the value of ‘R’ within a context is similar (or perhaps identical) to the mechanism applied in pragmatic implications.33 An advantage of this sort of semantics for ‘good’ is that it can account easily for the intuition that ‘Sue is a good dancer’ entails ‘There is a way in which Sue is good’ or ‘Sue is good in some respect’. On the incomplete predicate analysis, the latter sentences can be regarded as existentially quantifying over the variable within the logical form of ‘good’ whose value is normally fixed by the context. It is not clear to me how one could account for these inferences on the alternative accounts. Another advantage is that it accounts well for cases when something is said to be a good N, but the way it is good is not determined by ‘N’. For example, consider a case when someone is looking for an appropriate object to prop up a television set. Upon finding a thick book, he might say ‘This is a good book.’ What his utterance says is, of course, that the book is good for propping up the television set, not that it is good as a book. Again, the predicate modifier approach has difficulties with capturing such intuitions.
Adjectives in context
One of the consequences of this analysis is the claim that in some contexts (7) could express something different from the proposition that Sue is good as a dancer and Sue is a dancer. And indeed, one can imagine a situation in which what is at question is her moral character, not her skill at dancing. Suppose some dancers are threatening to burn down the Met, because they are unhappy about their salary. Other dancers are trying to convince the members of the first group that burning down the Met would be a bad idea. Meanwhile the singers are having a drink in a bar, and they are discussing which of their colleagues are in which group. Most of them believe that Sue is the leader of those who are threatening to set the building on fire. One of the singers tries to defend Sue, and argues that unlike most of the dancers, she is actually in the second group, trying to save the building. The singer gives several reasons for this claim, and at the end of her argument she utters the sentence ‘Sue is a good dancer’, perhaps putting a strong emphasis on ‘good’.34 But even if there are intricate examples where the proposition that Sue is a good dancer does not say that she is good at dancing, in normal contexts it does: ‘good’ within ‘good N’ normally has the content of ‘good as an N’. This means that in order to accept (7e) as the logical form of (7), one has to slightly expand the notion of context informally introduced at the beginning of this paper. There I said that the content of previous utterances might play a role in determining the content of an expression. But this is not enough. Now I suggest that the content of certain expressions within the utterance may be part of the context that contributes to determining the content of other expressions within the same utterance. The content of ‘good’ in an utterance of (7) may depend, in part, on the content of ‘dancer’.35
5. Varieties of incompleteness The story about the semantics of ‘good’ told in the previous two sections has obvious implications for Travis’s example concerning the color of the painted leaf. If it can be made plausible that the analysis given for ‘good’ — according to which its logical form is a contextually incomplete predicate — applies to other adjectives, including ‘green’, then we may have a semantic account of these expressions that does not violate the context thesis. If ‘green’ is an incomplete predicate, the context-dependence of ‘The leaf is green’ is attributable to the context-dependence of ‘green’. The resistance to the predicative analysis in the semantic literature is largely
135
136 Zoltán Gendler Szabó
based on the non-transparency of most adjectives. Once it is acknowledged that the invalidity of these inferences can be attributed to contextual incompleteness, the resistance looses much of its force. The advantages of a uniform predicative approach to all adjectives are considerable. First of all it is uniform: it assigns the same kind of semantic value to all adjectives.36 This is a theoretical gain for it keeps the syntactic and semantic structures close to one another. Second, assuming a broadly Davidsonean semantics for adverbs, we can account for similarities between adjectival and adverbial modification. Adjectives and adverbs can both treated as predicates: the former as predicates of individuals, the latter as predicates of eventualities.37 Proponents of a unified predicate-modifier approach have no similarly straightforward account to offer about modification in general.38 Finally, there is some psycholinguistic evidence that the predicative use of adjectives is acquired (and, in the case of aphasiaand dysphasia- patients, re-acquired) first.39 This suggests that the predicative use of adjectives is more basic. Nevertheless, the predicative analysis cannot be universally applied to all adjectives. Some (e.g. ‘utter’, ‘former’, ‘chief ’, ‘final’, etc.) have no predicative occurrences, so the assumption that they stand for incomplete predicates is extremely implausible. And ‘Sue is a retired pianist’ does not have the same content as ‘Sue is retired (as a pianist) and Sue is a pianist’.40 I suggest that we set the adjectives which do not support the inference pattern from ‘a is AN’ to ‘a is N’ aside, and accept for all the others a incomplete predicate analysis. In this way we maintain a uniform semantic analysis for the vast majority of adjectives.41 Contextual incompleteness in adjectives has different dimensions. The interpretation of scalar adjectives, like ‘tall’, ‘heavy’, ‘fast’, ‘expensive’ or ‘old’, requires a comparison class. Short basketball players are usually tall people, light whales are heavy animals and — for many of us — cheap airplane tickets to Europe are expensive gifts. The interpretation of evaluative adjectives, like ‘lucky’, ‘delicious’, ‘time-consuming’, ‘simple’ or ‘fitting’, requires that context provide an individual or group of individuals from whose perspective the evaluation is made. A particular event can be fortuitous for me, but not for others; an exercise can be simple for the expert and taxing for the novice, the outcome of an election may be lucky for one party, but not for the country. And of course, there are adjectives that belong to both of these categories; ‘expensive’ and ‘time-consuming’ would be good examples. I suggest that different dimensions of incompleteness correspond to different sorts of variables in the logical form. The variables can receive their
Adjectives in context
values from the context but also from linguistic material. In English, prepositional phrases tend to perform the latter job. For example, ‘clever at doing cross-word puzzles’ and ‘time-consuming as a short-paper topic for graduate students’ are wholly specified one-place predicates. Accordingly, I suggest that the logical form of ‘clever’ is ‘(clever(R))(x)’, the logical form of ‘timeconsuming’ is ‘(time-consuming(C1, C2))(x)’, where ‘R’ is a role-variable standing for a way of being clever, ‘C1’ is a class-variable, standing for a class of with respect to which something counts as time-consuming, and ‘C2’ is another class-variable, standing for a class from whose perspective something counts a time-consuming.42 For predicates that are fully specified, transfer-inferences are valid. For example, from ‘Sue is a clever dancer’ and ‘Sue is a pianist’ it does follow that ‘Sue is a clever pianist’, provided that e.g. the context had already specified that ‘clever’ is to be understood as ‘clever at solving crossword-puzzles’ in both the first premise and the conclusion.43 Of course, if only cleverness in solving crossword puzzles is concerned, it would be quite misleading to say that Sue is a clever dancer or a clever pianist. One should put it rather as ‘Sue, who is a dancer (pianist) is clever’. The reason for this is that the sentence ‘Sue is a clever dancer’ strongly suggests that the kind of cleverness that is ascribed to Sue has something to do with the fact that she is a dancer. So what should we say about ‘green’? It is certainly not an evaluative adjective: no matter how much one believes in the possibility of an inverted spectrum, it is incorrect to say that a leaf is green for me. We use color terms as if color were an objective, observer-independent feature of the world. We have another construction, namely ‘seems green’ which is correctly used in cases when the speaker wants to express uncertainty with respect to veridicality of her subjective experience. ‘Green’ is probably a scalar adjective.44 But setting degrees of greenness aside, one might be tempted to think that ‘green’ is a paradigm example of a contextually complete adjective. To be green is one of the most obvious examples of a feature that is simple. It is often difficult to decide whether something is green because of the condition of the object, the environment, or the nature of our perceptual mechanism; there are also borderline cases, when there might be no fact of the matter whether something is green. But all things that are green are green in the very same way. I think this intuition is misleading. There are at least two ways in which an apple can be green: from the outside, or from the inside. In the former case, it can be ripe, in the latter it cannot.45 There are at least three ways in which a
137
138
Zoltán Gendler Szabó
book can be green: due to having a green dust jacket, a green cover, or green pages. And a corridor can be green in many ways: having green walls, or green ceiling, or green carpet, or green doors, etc.46 An object is green if some contextually specifiable (and presumably sufficiently large) part of it is green. The logical form of ‘green’ is ‘(green(C, P))(x)’, where ‘C’ is a class standing for a comparison class and ‘P’ is a variable standing for a certain part of the object.47 It seems to me that the case of the painted maple leaf fits this pattern. If one is sorting leaves for decoration, what matters is the color of the outside, if one is trying to identify the species, what matters is what we find under the camouflage. This suggests that the context-dependency that appears in Travis’s example is a relatively easily characterizable kind: it is a matter of different contextually specified values for the variable ‘P’.
6. Conclusion My strategy of defending the compositionality of content against Travis’ challenge was to locate the source of the context-dependence of ‘The leaf is green’ in the adjective ‘green’. I argued that the best semantics for ‘green’ postulates a variable within the logical form of this word. Since I also argued that — despite Travis’ contention — this challenge couldn’t be broadly generalized beyond sentences containing adjectives, the answer is to the point. One might wonder whether in responding to similar challenges we have to postulate variables in other lexical items as well. I think we do. Consider, for example, the sentence ‘Every leaf is green’. Arguably, many different things can be said with this sentence, depending on what the contextually relevant domain of quantification is supposed to be. And of course, one may have the prima facie intuition that none of the lexical items has a context-dependent content. Here again, I think the first intuition is misleading; a general account of domain restriction may demand that we postulate the presence of a domain variable within the logical form of ‘leaf ’.48 So, the strategy of defense of compositionality pursued here is likely to lead to a lexicon where many, perhaps most entries contain contextual variables. Whether the defense of compositionality I offered is ultimately convincing depends largely on the plausibility of such a lexicon. And that, in turn, depends on whether we can provide an account of how the values of these variables are to be determined in context. At the beginning of the paper I set foundational
Adjectives in context 139
questions regarding the nature of context and our knowledge of its features aside. At the end, I concede that as long as these issues remain unresolved, the semantic proposal I gave remains questionable. But even if that is so, I hope that I succeeded in pointing out where the real issues lie regarding the compositional interpretation of adjectives.
Notes * This paper is a substantially revised version of Chapter 4 of my dissertation. I want to thank the late George Boolos, Richard Cartwright, Robert Stalnaker and Judith Thomson for discussion and advice regarding drafts of that chapter. The final version of the paper benefited from discussions with Tamar Szabó Gendler, Harold Hodes, Jeff King, and Jason Stanley. I also wish to thank two anonymous referees for this volume whose comments led to further revisions. 1. As I understand the principle of compositionality, it says that the meaning of complex expressions — phrases, clauses and sentences — depends on two kinds of facts. Syntax tells us what the lexical constituents — words and smaller morphemes — occurring in a complex expression are and how they are combined; and the lexicon tells us what those constituents mean. According to the principle, such syntactic and semantic features of a complex expression are jointly sufficient to determine the meaning of that expression. But what does it take for certain features of an expression to determine what that expression means? It is customary in semantics to assume that the relevant determination relation is simply functional dependence. I believe this interpretation of the principle of compositionality is unreasonably weak. In Szabó (2000) I argue that the principle is best understood as the claim that there is a single function across all possible human languages that assigns the meanings of complex expressions to the meanings of their constituents and their syntactic manner of composition. This, however, will not matter for the purposes of the present discussion. 2. One might object to calling the content of an expression its “meaning”. If we think that the meaning of an expression is a feature knowledge of which guarantees understanding of the expression then the proposition expressed by a sentence is not a good candidate for being the meaning of that sentence. (Intuitively, we all know what ‘I am tired’ means but we don’t know what proposition it expresses on a given occasion unless we know who made the utterance.) Perhaps, we should follow David Kaplan and think of meanings as characters, i.e. functions from contexts to contents; cf. Kaplan (1977). Be that as it may, compositionality of content is at least as important in semantics as compositionality of meaning proper. Just as we can speak about understanding expressions, we can also speak about understanding utterances and the latter is presumably tied to grasping the proposition expressed by the utterance. So, insofar that it is plausible that we can understand not only sentences we never hear before but also particular utterances of such sentences, we have some intuitive support for the principle of compositionality of content.
140 Zoltán Gendler Szabó
3. There are two major traditions of thinking about context within semantics. There are those who follow Montague (1968) in representing contexts as indices, i.e. as an n-tuples of speaker, hearer, time, place, a series of salient objects, etc. And there are those who follow Stalnaker (1970) in thinking of contexts as information states which can be represented as sets of possible worlds (or sets of situations) compatible with the shared knowledge of the participants in the conversation. 4. I use the following conventions of quotation. An expression e between single quotes is a (possibly ambiguous) name for e. I also use the iterated single quote to name quotation names of expressions. Double quotes differ from iterated single quotes in appearance: in the case of iterated single quotes I leave a space between the inverted commas, in the case of double quotes I do not. Double quotes are strictly speaking not quotation devices: the expression between the double quotes is used, not mentioned. (Or perhaps, partly used, partly mentioned.) The following examples are illustrative: (i) ‘Cambridge’ refers (ambiguously) to the cities Cambridge, England and Cambridge, Massachusetts. (ii) ‘ ‘Cambridge’ ’ refers to an ambiguous quotation name which is used in (i). (iii) According to a friend of mine, “Cambridge is a reasonably lively place.” He probably has Cambridge, Massachusetts in mind. These conventions are not supposed to be taken as an analysis of how quotation is ordinarily used in English. I have grave doubts about the coherence of our ordinary use quotation marks. 5. The definitions are from the Webster Dictionary. 6. Here and throughout the paper I use the Courier font with or without indices to indicate disambiguation. 7. Unfortunately, there is no terminological unity in linguistics concerning types of ambiguity. I prefer the following definitions. A linguistic type (a phoneme or a sequence of phonemes) is ambiguous if it has more than one content. If an ambiguous linguistic type stands for a single expression, it is polysemous; otherwise it is homonymous. 8. Travis (1994: 171–2). 9. Travis (1994: 172). 10. What makes a category syntactic is that we can determine which expressions belong to that category through morpho-syntactic means. Syntactic categories of formal languages can be ‘read off’ the rules which specify the class of well-formed formulas in the language. 11. One might wish to steer between an extremely weak and an extremely strong conception of logical form. According to the weak conception, the only criterion that we might use to evaluate the adequacy of a certain logical form is whether it gets all the inferences right. This would mean that we can assign the same logical form to any two sentences of English that are logically equivalent. But surely, we should not ascribe ‘Snow is falling’ and ‘Snow is falling and snow is self-identical’ the same logical form. The strong conception of logical form requires that an expression and its logical forms share all their semantically relevant properties. This would demand that semantic features, like [inanimate] be represented in logical form. But it is not clear whether we really want to ascribe different logical forms to ‘It is here’ and ‘She is here’.
Adjectives in context
12. I use the Courier font for logical forms, which is the notation introduced in the previous section to indicate disambiguation; cf. fn. 6. Given that we use logical forms to clarify and explain the logical behavior of natural language expressions, the demand that the languages of logical forms be free of ambiguity seems reasonable. If I had the goal of absolute precision, I would use corner quotes whenever I mention logical forms containing free variables. Since I don’t, I won’t. 13. One of the many passages where Moore endorses this view is the following: “For ‘good conduct’ is a complex notion: all conduct is not good; for some is certainly bad and some may be indifferent. And on the other hand, other things, beside conduct, may be good; and if they are so, then, ‘good’ denotes some property, that is common to them and conduct …” Moore (1903: 2). 14. Cf. Geach (1956). 15. Hamann (1991: 666). 16. Geach endorses this view in the following passage: “Even when ‘good’ or ‘bad’ stands by itself as a predicate, and is thus grammatically predicative, some substantive has to be understood; there is no such thing as being just good or bad, there is only a being a good or bad so-and-so.” Geach (1956: 65). 17. Larson (1998). 18. Larson believes that ‘Sue is a good dancer’ is ambiguous and one of its readings is more or less adequately captured by the original Moorean analysis: Sue is good and Sue is a dancer. I am skeptical about the existence of such a reading, for reasons that will become clear later. In any case, even if a Moorean reading exists, the interesting question is how to capture the other one. 19. The logical form I present is different from Larson’s official logical forms in one respect. Larson assumes that the logical form of ‘good’ contains another argument place filled by a contextual variable ‘C’ whose values are comparison classes. This is necessary because a dancing event can be good relative a narrow class of dancing events but not so good relative to a broader comparison class. I have neglected this complication here. 20. One might deny that there could be a world where dancing events and rainmaking events coincide: such events could occur always simultaneously, but they would nevertheless be distinct events. I find this suggestion ad hoc. An anthropologist may observe dances, describe them in detail, and only later discover that their function is rainmaking. He would naturally think that he thereby found an alternative way to describe the very same events he had already observed. 21. I am assuming that whatever the semantics of the generic quantifier, it must validate such an inference. 22. Cf. Montague (1970a, esp. 201, 211–3), and Montague (1970b, esp. 242–3). 23. Cf. Lewis (1970, esp. 193–200). 24. In doing this, I don’t want to prejudge the complicated issue of ontological commitment. Philosophers tend to like non-explicit quantification over possible worlds because they believe that it fails to carry ontological commitment to those entities. I am doubtful whether pushing explicit quantification into the meta-language can help the ontologically scrupulous,
141
142 Zoltán Gendler Szabó
but I don’t want to press the issue here. There is another typical motivation for using modal operators, rather than explicit quantification. It has been argued that natural language lacks the expressive power of full-blooded quantification over possible worlds. This claim, however, had been challenged on empirical grounds by a number of philosophers and linguists. 25. But note the following objection: Arguably, ‘sell’ and ‘buy’ are expressions such that they are necessarily true of the very same events. Still, a particular transaction can be a good sell and a bad buy. So ‘The transaction between Bill and Sam was a good buy’ does not entail ‘The transaction between Bill and Sam was a good sell.’ But the argument is not fully convincing. A defender of the intensional approach may reply that selling and buying events must be distinguished. Alternatively, one might argue that although the events are the same, the actions of buying and selling are different, and that the logical forms of action verbs quantify over actions, rather than events. 26. This criticism of the use of the intensional machinery in blocking certain inferences has a predecessor in the literature. Sally McConnell-Ginet writes: “But the intensional machinery does not provide a good model of how we think about why those walking quickly might be different from those talking quickly, even though walkers and talkers are the same. The explanation lies not in the existence of an alternative situation (where individuals have different properties), but simply in the possibility of a different sorting of the individuals, given a refinement of the sorting principles.” McConnell-Ginet (1982: 163). 27. The phrase ‘way of being good’ is borrowed from Thomson (1992). 28. Geach (1956: 68–9). 29. I thank Judith Thomson for bringing this problem to my attention. Her discussion of it can be found in Thomson (1994). 30. A defender of the predicate-modifier approach will have to say that this is not a case of genuine ellipsis. One could suggest, for example, that ‘This is good too’ is proxy for ‘This is good read too’. This means that to interpret ‘This is good too’, one has to realize that if something is a good book then it is a good read. 31. ‘Role’ is used here as a more or less technical term. An actor can be good in a given role. Stretching the meaning of ‘role’ a little, one can say that a good dancer, or pianist is good in that role. Perhaps one can say that a good pencil is good in a role, but it certainly makes no sense to say that a good nap, a good sunset, or a good painting is good in some role. The variable ‘R’ stands for some contextual information that specifies the incomplete predicate ‘good’. 32. How is the value of such a variable determined in context? This is a question that cannot be reasonably addressed without taking a firm stand on foundational questions regarding the nature of contexts and the way the conversational participants know about it. I have no such comprehensive view to offer, so I will leave this issue open. My attitude here is similar to the attitude of those who argue that the demonstrative pronoun ‘this’ picks out some contextually determined object, but take no stand on whether the referent of the demonstrative is fixed by intentions, by demonstrations, by salience or some other means. 33. “…pragmatic factors may contribute to semantic interpretation without leading to the consequence that the idea of an independent semantic level should be given up. Pragmatic factors may trigger syntactic rules as well and in spite of this we would not like to say that
Adjectives in context 143
there is no such thing as syntax and everything should be incorporated into pragmatics.” Kiefer (1978: 161). 34. One might object that in this example I exploited an ambiguity. It seems plausible that ‘good’ has a special sense in which it means something like ‘morally good’, and it is this ambiguity, rather than some sort of context-dependence that properly accounts for the case. But I don’t think it is right to say that in my example the defender of Sue declared her to be morally good. This is revealed by the fact that uttering ‘Although morally reprehensible, Sue is a good dancer’ (again, giving a strong emphasis to ‘good’) would not have been a contradictory assertion in the given context. 35. Fortunately, there do not seem to be cases when the content of a noun ‘N’ within a ‘good N’ depends on the content of ‘good’ in that phrase. This accounts for the strong intuition that within a noun phrase of the form [NP AN], the semantic status of the adjective and the noun are different. The content of the adjective is influenced by the content of the noun, but not the other way around. 36. There are many who defend a double classification for adjectives; cf. Parsons (1972) and Siegel (1976). 37. Eventualities are events, states and processes. 38. Classic proponents of the uniform predicate-modifier approach include Montague (1970a) and Cresswell (1976). 39. Cf. Osgood (1971). 40. The fact that ‘retired’ has predicative uses is something of an anomaly. Note that ‘Sue is a retired pianist’ is hardly distinguishable from ‘Sue is a former pianist’, even though the former entails ‘Sue is retired (as a pianist)’ but the latter does not entail *‘Sue is former (as a pianist)’. The difference between ‘retired’ and ‘former’ is probably due to the fact that ‘retired’ is derived from a participle, whereas ‘former’ isn’t. 41. Is it legitimate to set these cases aside? The trouble, of course, is that it might turn out that a semantic account which covers these cases can account for those dealt with the predicative analysis as well. Still, I think it would be bad methodology to assimilate all adjectives to the relatively few which fail to support the inference from ‘a is AN’ to ‘a is N’. For then the fact that these inferences are valid for the vast majority of adjectives would have to be explained via ad hoc meaning-postulates. Those who insist on treating all adjectives as they would treat fairly exceptional cases are “generalizing to the worst case”, just as Montague did when he assigned intensional types to all expressions. 42. It is important to notice that the prepositional phrase that can complete adjectives can modify different constituents of a sentence and that the results are not all equivalent. ‘This mouse, for a mouse, is big’ is equivalent to ‘This mouse is big for a mouse’, and ‘This, for a mouse, is big’ is equivalent to ‘This is big for a mouse’, but the first two sentences are not equivalent to the last two. The first two sentences entail that the object demonstrated is a mouse, the last two do not. Cf. Bartsch (1987: 16). 43. Bartsch (1987) notices that ‘John’s paper is stylistically good’ entails ‘John’s paper is good’ if we evaluate the premise and the conclusion within a context in which the same
144 Zoltán Gendler Szabó
respects of goodness are relevant. (p. 5) Nevertheless, instead of the approach advocated here, she endorses a context-sensitive predicate-modifier account of the semantics of adjectives. 44. Languages may vary in how they categorize color adjectives, there might be even variation within the same language. For example, in Hungarian comparative forms of basic color adjectives are all acceptable, but comparative forms of complex color adjectives are not. Cf. Kiefer (1978: 154). 45. Could it be that in one of these cases the apple is not literally green? This is unlikely, for there seems to be non non-arbitrary way to decide which of the two cases are cases on nonliteral greenness. 46. Lahav (1989) rejects the idea that the context sensitivity of ‘red house’ can be accounted for in this manner. He points out that we might be interested in the color of the inside of a house as much as we are interested in the color of its external surface, the color of the rooms does not normally count towards the color of the house. Although this observation is correct, it does not undermine the analysis. In some contexts it may well be correct to call a house red on account of its interior. The fact that such contexts are unusual is of no semantic concern. 47. In this regard, ‘bald’ behaves much like color adjectives. One can be bald on the top of one’s head, in the front, in the back, none of which means that one is bald altogether. 48. We argue for this view in Stanley and Szabó (2000).
References Bartsch, Renate 1987 “Context-dependent Interpretations of Lexical Items”. In Foundations of Pragmatics and Lexical Semantics, J. Groenendijk, D. de Jongh and M. Stokhof (eds), 1–26. Dordrecht: Foris. Cresswell, Max 1976 “The Semantics of Degree”. In Montague Grammar, B. Partee (ed.), 261–292. New York: Academic Press. Geach, Peter 1956 “Good and Evil”. In Theories of Ethics, P. Foot (ed.), 64–74. Oxford: Oxford University Press. Hamann, C 1991 “Adjectival Semantics”. In Handbuch der Semantik, A. von Stechow (ed.). Berlin: Walter de Gruyter. Kaplan, David 1977 “Demonstratives: An Essay on the Logic, Metaphysics, and Epistemology of Demonstratives and Other Indexicals”. Reprinted in Themes from Kaplan, J. Perry and H. Wettstein (eds), 1989, 481–563. Oxford: Oxford University Press. Kiefer, Ferenc 1978 “Adjectives and Presuppositions”. Theoretical Linguistics 2: 135–73.
Adjectives in context
Lahav, Ron 1989
“Against Compositionality: The Case of Adjectives”. Philosophical Studies 57: 261–79. Larson, Richard 1998 “Events and Modification in Nominals”. In Proceedings From Semantics and Linguistic Theory (SALT) VIII, D. Strolovitch and A. Lawson (eds), 145–168. Ithaca, NY: Cornell University Press. Lewis, David 1970 “General Semantics”. Reprinted in Philosophical Papers, vol. 1., 189–229. Oxford: Oxford University Press. McConnell-Ginet, Sally 1982 “Adverbs and Logical Form: A Linguistically Realistic Theory”. Language 58: 144–184. Montague, Richard 1968 “Pragmatics”. Reprinted in Formal Philosophy, R. Thomason (ed.), 1974, 95–118. New Haven: Yale University Press. 1970a “English as a Formal Language”. Reprinted in Formal Philosophy, R. Thomason (ed.), 1974, 188–221. New Haven: Yale University Press. 1970b “Universal Grammar”. Reprinted in Formal Philosophy, R. Thomason (ed.), 1974, 222–246. New Haven: Yale University Press. Moore, G. E. 1903 Principia Ethica. Cambridge: Cambridge University Press. Osgood, Charles E. 1971 “Where do Sentences Come From”. In Semantics: An Interdisciplinary Reader in Philosophy, Linguistics and Psychology, D. Steinberg and L. Jacobovits (eds), 497–529. Cambridge: Cambridge University Press. Parsons, Terence 1972 “Some Problems Concerning the Logic of Grammatical Modifiers”. In Semantics of Natural Language, D. Davidson and G. Harman (eds), 127–141. Dordrecht: Reidel. Siegel, Muffy E. A. 1976 Capturing the Adjective. Ph.D. Dissertation, Amherst: University of Massachusetts. Stanley, Jason and Szabó, Zoltán Gendler 2000 “On Quantifier Domain Restriction”. Mind and Language 15: 219–61. Stalnaker, Robert 1970 “Pragmatics”. Reprinted in Context and Content: Essays on Intentionality in Speech and Thought, 31–46. Oxford: Oxford University Press. Szabó Zoltán Gendler 2000 “Compositionality as Supervenience”. Linguistcs and Philosophy 23: 475–505. Thomson, Judith Jarvis 1992 “On Some Ways in Which a Thing Can Be Good”. Social Philosophy and Policy 9: 96–117. 1994 “Goodness and Utilitarianism”. In Proceedings of the American Philosophical Association 67: 7–22.
145
146 Zoltán Gendler Szabó
Travis, Charles 1994 “On Constraints of Generality”. In Proceedings of the Aristotelian Society. New Series 44: 165–88. London: Aristotelian Society Publications.
AUTHOR "Kent Bach"
TITLE "Semantically speaking"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Semantically speaking Kent Bach San Francisco State University
In the course of introducing his theory of conversational implicature, Grice adopted a notion of what is said with which to contrast what a speaker implicates in uttering a sentence. He imposed the Syntactic Correlation constraint, as I call it, according to which what is said must correspond to “the elements of [the sentence], their order, and their syntactic character” (1989: 87). So if any element of the content of an utterance, i.e., of what the speaker intends to convey, does not correspond to any element of the sentence being uttered, it is not part of what is said. How could anyone object to the Syntactic Correlation constraint? It might seem that this constraint limits what is said to the context-invariant content of an utterance, but Grice explicitly allowed for indexicality and ambiguity. Even though the referent of an indexical element varies from context to context, it still corresponds (relative to the context) to that element. And even if the uttered sentence is ambiguous, either because it contains ambiguous expressions or is structurally ambiguous, the meaning that is operative in a given context is still a projection of the syntax of the sentence as used in that context. Nevertheless, recent critics have contended that a syntactically constrained notion of what is said is theoretically useless and empirically unmotivated. One common reason for this contention is that many sentences are semantically incomplete or underdeterminate — they fail to express a complete and determinate proposition even after any indexical references are assigned and any ambiguities are resolved — in which case nothing is said in the utterance of such a sentence. Another common reason is that many semantically complete sentences intuitively do not express, and typically are not used to convey, the proposition which, by Syntactic Correlation, comprises what is said. A third reason is that what is said (according to Syntactic Correlation) generally plays no role in the pragmatic processes involved in understanding utterances.
148 Kent Bach
I will argue that a syntactically constrained notion of what is said does have a legitimate theoretical role to play and is empirically significant. In order to show this, I will first (in Section 1) suggest several simple refinements to Grice’s original notion. Then (in Section 2) I will explain what I take to be right and what I take to be wrong with Grice’s distinction between what is said and what is implicated. This distinction is not exhaustive, for it neglects the phenomenon of conversational impliciture (that’s implic-i-ture, as opposed to implic-a-ture), but taking that into account requires merely refining his distinction. It has been objected that this distinction, however refined, is ultimately misguided, in that the notion of what is said is inherently pragmatic, just like implicature. This objection has been based on arguments that take roughly the following form: 1. Any sentence can be used to communicate something without implicating it. 2. If an utterance of a given sentence communicates a proposition that it doesn’t implicate, then that proposition is what is said in the utterance of that sentence. 3. Even so, the determination of that proposition is bound up in pragmatics. 4. Therefore, what is said is not a purely semantic matter. Bad argument. Aside from mistakenly assuming that the distinction between what is said and what is implicated must be exhaustive, this form of argument relies on an overly expansive conception of what is said. The rationale for this conception derives mainly from certain psychological and epistemological considerations involving “pragmatic” processes and “semantic” intuitions. I will argue (in Section 3) that such considerations are misdirected and do not undermine a syntactically constrained, purely semantic notion of what is said. As I will explain in Section 4, this overly expansive conception of what is said can seem attractive only insofar as one ignores certain distinctions, each between something semantic and something pragmatic. Once these distinctions are recognized, there is no reason to reject a purely semantic notion of what is said.
1.
Refining the notion of what is said
In contrasting saying with implicating, Grice took our “intuitive understanding of the meaning of say” (1989: 24–25) to comport with Syntactic Correlation, which requires that what is said correspond to “the elements of [the sentence], their order, and their syntactic character”. His use of ‘order’ and
Semantically speaking 149
‘syntactic character’ suggest some sort of compositional principle, such that what is said is determined by the contents of the constituents (‘elements’) of the sentence as a function of their syntactic relationships. And considering that a sentence can include unpronounced elements (Grice did not consider this), such as empty categories, implicit arguments, and syntactic ellipses, Syntactic Correlation should not be construed as requiring that every element of what is said corresponds to an uttered element of the uttered sentence. But it does require that every element of what is said correspond to some element of the uttered sentence.1 There was one respect in which Grice’s favored sense of ‘say’ was a bit stipulative. For him saying something entails meaning it. This is why he used the locution ‘making as if to say’ to describe irony, metaphor, etc., since in these cases one does not mean what one appears to be saying. Here he seems to have conflated saying with stating. It is more natural to describe these as cases of saying one thing and meaning something else instead. That’s what it is to speak nonliterally (at least if one does so intentionally). For example, if I utter “You’ll be my friend for life” to someone who has just betrayed me, I am saying that this person will be my friend for life — that’s just not what I mean. Similarly, if I utter “You are pond slime” to this person, I am saying that this person is pond slime, although I mean something rather different (though equally unsavory). Besides nonliterality, there are two other reasons for denying that saying something entails meaning it. A speaker one can mean one thing but unintentionally say something else, owing to a slip of the tongue, a misuse of a word, or otherwise misspeaking. Also, one can say something without meaning anything at all, as in cases of translating, reciting, or rehearsing, where one utters a sentence with full understanding (one isn’t just practicing one’s pronunciation) but is not using it to communicate anything.2 So we can replace Grice’s idiosyncratic distinction between saying and merely making as if to say with the distinction (in indicative cases) between explicitly stating and saying (in Austin’s locutionary sense). In so doing, we can distinguish nonliterality from implicature.3 In this way we have a notion of what is said that applies uniformly to four situations: (1) where the speaker means what he says and something else as well (cases of implicature and of indirect speech acts in general), (2) where the speaker (intentionally) says one thing and means something else instead (nonliteral utterances), (3) where the speaker does not say what he intends to say, as in the misuse of a word or a slip of the tongue, and (4) where the speaker says something but doesn’t mean anything at all. On this conception, what is said, being closely tied to the (or a)
150 Kent Bach
meaning of the uttered sentence provides (allowing for indexicality and ambiguity) the hearer with the linguistic (semantic) component of his basis for inferring what, if anything, the speaker is communicating. Whether a speaker means exactly what he says, something else as well, or something else instead, clearly it is never part of the meaning of a sentence that on a particular occasion of use it is being used to communicate anything at all. That is something the hearer presumes from the fact that the speaker is uttering a sentence at all. This communicative presumption, as Bach and Harnish call it (1979: 7), comes into play even if what the speaker means does not extend beyond or depart in any way from the meaning of the sentence he utters. Also, it is never part of what a sentence encodes that it has to be used literally — the hearer must infer (if only by default) that it is being used literally. The utterance does not carry its literalness on its sleeve. It might even contain the word ‘literally’, but even that word can be used nonliterally. So in no case does what a speaker says determine what he is communicating in saying it. To that extent it is true, though trivially so, that every utterance is context-sensitive. In sum, I am suggesting that we reject Grice’s arbitrary stipulation that saying something entails meaning it but retain his appropriately restrictive Syntactic Correlation constraint on what is said. This has important consequences. First, Syntactic Correlation leaves open the possibility that what is said in uttering a sentence need not be a complete proposition. The mere fact that a sentence is syntactically complete (well-formed) does not entail that it is semantically complete. Second, since saying something does not entail meaning it, what one says does not depend on whether or not one has any communicative intention in saying it. So the speaker’s communicative intention cannot contribute to the determination of what is said.4 Communicative intentions enter in only at the level of what the speaker means (is communicating). Even if the speaker means exactly what he says (nothing else and nothing more), that he means exactly what he says is a matter of his communicative intention. Third, any proposition that the speaker is communicating that is distinct from what he says is explicitly cancellable (Grice 1989: 44). That is, the speaker can continue his utterance and, without contradicting himself, go on to explicitly indicate that this proposition was not was not part of what he meant. This third feature is really a consequence of the first two. All three of these consequences should be kept in mind in the course of what follows.
Semantically speaking
2. Refining Grice’s distinction Now we can ask whether the distinction between what is said and what is implicated is exhaustive. Here I assume familiarity with Grice’s account (1989: Chs. 2 & 3) and will not go into the details or trot out the stock examples. Suffice it to say that he proposes a general Cooperative Principle and specific “maxims” — of quality, quantity, relation (relevance), and manner — to account for the rationale and success of conversational implicatures. It is arguable, however, that being cooperative (at least in making constative utterances) consists simply in being truthful and relevant, hence that there is no need for four separate maxims. For example, observing the maxim of quantity is just being relevantly informative. Also, the maxims might better be thought of as presumptions (Bach and Harnish 1979:62–65), for it is on the presumption that the speaker is being truthful and relevant (and intends to be taken to be) that the hearer figures out what the speaker means in saying what he says. The question of whether Grice’s distinction is exhaustive is the question of whether a speaker, in saying something, can mean (communicate) something other than by saying it or by implicating it. As I have argued previously (Bach 1994a), his distinction is not exhaustive, because it leaves out the phenomenon of conversational impliciture (as opposed to implic-a-ture). Implicitures go beyond what is said, but unlike implicatures, which are additional propositions external to what is said, implicitures are built out of what is said.5 Here are some simple examples of sentences likely to be used to convey implicitures: (1) Jack and Jill went up the hill. (2) Jack and Jill are engaged. (3) Jill got married and became pregnant.
Although these sentences express complete propositions, in uttering them a speaker is likely to have meant something more specific, a qualified version of what he said: (1e) Jack and Jill went up the hill [together]. (2e) Jack and Jill are engaged [to each other]. (3e) Jill got married and [then] became pregnant.
where the words in brackets, which are not part of the sentence, indicate part of what the speaker meant in saying what he did. He would have to utter those words (or roughly equivalent words — the exact words don’t matter) to make
151
152
Kent Bach
what he meant fully explicit.6 Let’s call the fuller version an expansion of the original and what is left out its implicit qualification. Utterances of sentences like (1)–(3) illustrate what I call sentence nonliterality (Bach 1994b: 69–72), as opposed to constituent nonliterality, since no expression in the sentence is being used nonliterally. The proposition the speaker means is not the exact proposition, as compositionally determined, that is expressed by the sentence. So, for example, someone who utters (2) is not saying that Jack and Jill are engaged to each other, any more than he would be saying this with “Jack and his sister Jill are engaged”. Rather, this is implicit in what he is saying or, more precisely, in his saying of it. That it is not part of what is said is clear from the fact that it passes Grice’s test of cancellability. That is, it may be taken back without contradiction. For example, there is no contradiction in saying, “Jack and Jill are engaged, but not to each other” (or “Jack and Jill went up the hill (1) but not together” or “Jill got married and became pregnant (3) but not in that order”). There is another kind of case, in which the uttered sentence, though syntactically complete, seems to be semantically incomplete, given Syntactic Correlation (and assuming semantic compositionality). Sentences like (4)–(6) do not seem to express complete propositions: (4) Jack and Jill climbed far enough. (5) Jack and Jill are ready. (6) Jill first got married.
On the plausible assumption that what a speaker means must be a complete proposition, it follows that in these cases the speaker cannot mean merely what is expressed by the sentence. What he means must include a completion of what he says, as illustrated in (4c)–(6c) by the contents of the bracketed material: (4c) Jack and Jill climbed far enough [to get a pail of water]. (5c) Jack and Jill are ready [to get married]. (6c) Jill first got married [and later became pregnant].
A speaker might intend such completions to be understood, but this does not mean that (4)–(6) are syntactically elliptical versions of (4c)–(6c). There is no reason to suppose that they contain suppressed, unpronounced versions of the bracketed material.7 Now it might seem that if sentences like (4)–(6) do not express complete propositions, nothing is said in utterances of them. It is interesting to observe, however, that utterances of these sentences, and of semantically incomplete
Semantically speaking
sentences in general, pass what I call the IQ Test.8 That is, there is no problem in indirectly quoting them, even though they do not express complete propositions (let’s imagine they’re uttered by Mother Goose): (4IQ) Ms. Goose said that Jack and Jill climbed far enough. (5IQ) Ms. Goose said that Jack and Jill are ready. (6IQ) Ms. Goose said that Jill first got married first.
So although (4)–(6) do not express complete propositions, Ms. Goose did say something in uttering them. It’s just that in each case what she said, as opposed to what she meant and asserted, was not a complete proposition. The semantic incompleteness of some sentences has no special significance in respect to the distinction between what is said and what is communicated, although it does refute the grammar-school adage that a complete sentence expresses a complete thought. That a sentence, despite being well-formed syntactically, can be semantically deficient is an underappreciated phenomenon of natural language. Examples like (1)–(3), since they are perfectly ordinary, illustrate the pervasiveness of the gap between linguistic meaning and speaker’s meaning, and those like (4)–(6), though somewhat less common, also illustrate the phenomenon of semantic incompleteness. Such examples have been thought to show that context sensitivity and linguistic meaning are so intermeshed that the notion what is said is a bogus category. As I will argue in the next section, I think this extreme conclusion is unwarranted. It cannot be established by appealing to intuition, introspection, or armchair psychology.
3. Spurious objections to the notion of what is said I have endorsed a literalistic notion of what is said, or what François Recanati (2001) calls “a purely semantic notion of what is said”. Before defending it against various objections, I should indicate what I take to be the relevant difference between semantic and pragmatic. This distinction has been drawn in various ways, but no matter how it is drawn the term ‘semantic’ has something to do with meaning and ‘pragmatic’ has something to do with use.9 Having previously discussed different formulations of the distinction (Bach 1999a), here I will simply explain the version of it that I prefer and why I prefer it. A semantic-pragmatic distinction can be drawn with respect to various things, such as ambiguities, implications, presuppositions, interpretations,
153
154
Kent Bach
knowledge, processes, rules, and principles. For me, it applies fundamentally to types of information. Semantic information is information encoded in what is uttered — these are stable linguistic features of the sentence — together with any extralinguistic information that provides (semantic) values to contextsensitive expressions in what is uttered. Pragmatic information is (extralinguistic) information that arises from an actual act of utterance. Whereas semantic information is encoded in what is uttered, pragmatic information is generated by, or at least made relevant by, the act of uttering it. This way of characterizing pragmatic information generalizes Grice’s point that what a speaker implicates in saying what he says is carried not by what he says but by his saying it (1989: 39). The act of producing the utterance exploits the information encoded but by its very performance creates new or otherwise invokes extralinguistic information. That information, combined with the information encoded, provides the basis for the hearer’s identification of the speaker’s communicative intention. Notice that this pragmatic information is relevant to the hearer’s inference only on the supposition that the speaker is producing the utterance with the intention that the information in question be taken into account. There is no such constraint on contextual information of the semantic kind, which has its effect independently of the speaker’s communicative intention and the hearer’s recognition of that intention. This sort of contextual information is limited to a short list of parameters associated with indexicals and tense, such as the identity of the speaker and the hearer and the time of an utterance.10 So, for example, “even if I am fully convinced that I am Napoleon, my use of ‘I’ designates me, not him. Similarly, I may be fully convinced that it is 1789, but it does not make my use of ‘now’ about a time in 1789” (Barwise and Perry 1983: 148). Now we can consider objections to the literalistic or purely semantic notion of what is said. On such a notion, what is said by a speaker in uttering a sentence excludes anything that is determined by his communicative intention (if it included that, then what is said would be partly a pragmatic matter). I am aware of several objections, which I will call the Propositional, the Not-anImplicature, the Psychological, and the Intuitive objections. Since the first two have in effect been answered in the previous section, here they will be quickly stated and disposed of. The last two deserve more serious attention. According to the Propositional objection, what is said when somebody utters a sentence must be a proposition. Since, as we have seen, some sentences do not express complete propositions (even after ambiguities and vaguenesses are resolved and indexical references are fixed), nothing is said when such a
Semantically speaking
sentence is uttered. Reply: as illustrated by utterances of sentences (4)–(6), indicative sentences may be indirectly quoted even if they are semantically incomplete — they pass the IQ Test. So there is no basis for the assumption that what is said must be a complete proposition. Besides, why should the syntactic requirements on well-formed sentences preclude the possibility that the semantic interpretation of some well-formed sentences might be an incomplete proposition? It is plausible to suppose that a speaker, when he means something, must mean a complete proposition, but it doesn’t follow that any complete sentence used to communicate a complete proposition must itself express one. The Not-an-Implicature objection is based on resisting what Robyn Carston has called “the compulsion to treat all pragmatically derived meaning as implicature” (1988: 176). Once it is recognized that the contribution of pragmatic processes is not limited to the determination of implicatures, “there is no reason”, according to Carston, “why pragmatics cannot contribute to the explicature, the truth-conditional content of the utterance”, which she equates with what is said. However, insofar as the explicature can be the truth-conditional content of the result of an expansion or completion of the utterance, it cannot be identified with what is said, the proposition, expressed relative to the context, by the uttered sentence. Since, as we have seen, the distinction between what is said and what is implicated is not exhaustive, it is a mistake to suppose that if a conveyed proposition is not an implicature, it must be identified with what is said. The two main objections to the semantic notion of what is said apply to sentences like (1)–(3), which do express a complete proposition but are typically used to convey not that proposition but an expanded or enriched version of it. Recanati (2001) suggests that although the semantic interpretation of such a sentence yields a complete proposition, this “minimal” proposition is an “abstraction with no psychological reality”.11 He claims that “semantic interpretation by itself is powerless to determinate what is said” in cases like (1)–(3), because “we cannot sever the link between what is said and the speaker’s publicly recognizable intentions”. He reaches these conclusions, as we will see, by adverting to the “pragmatic processes” involved in understanding an utterance and to intuitions about what is said.12 3.1 The psychological objection This objection assumes that what is said is in the strict, literalistic sense can have no psychological reality unless it is something that a hearer must identify before
155
156 Kent Bach
inferring the speaker’s communicative intention. In developing this objection, Recanati draws several distinctions among what he calls pragmatic processes (the cognitive processes involved in figuring out what a speaker is conveying in uttering a sentence). A primary process yields the first proposition identified by the hearer, whereas any further proposition is arrived at by way of a secondary process. Recanati regards the difference here to be psychologically real, not merely terminological, because only secondary processes are inferential. Primary processes cannot be inferential, according to Recanati, because they do not take a proposition as input. He also distinguishes mandatory from optional processes. A process is mandatory if needed to arrive at any proposition at all, and optional otherwise. A process is mandatory if the uttered sentence is semantically incomplete. It is a further question whether or not the sentence actually contains a covert argument slot (needing to be assigned a semantic value for the sentence to express a complete proposition, relative to a given context). So a process is mandatory not because interpreting the sentence requires it but because the semantic interpretation of the (semantically incomplete) sentence does not yield a complete proposition. Finally, notice (this is key to Recanati’s objection) that a process can be primary without being mandatory, because the first proposition that a hearer actually arrives at might not be the one expressed by the sentence in the context. These are interesting distinctions, although it is not obvious that either of these distinctions marks a psychologically real difference. Perhaps a primary process can be inferential if, when what is said is a complete proposition, it bypasses that proposition, or even if it begins with an incomplete proposition. After all, there is a distinction to be drawn between the possibly non-propositional semantic interpretation of a sentence ‘S’ and the hearer’s thought that the speaker said that S. Since, as we have seen, utterances of semantically incomplete sentences pass the IQ test, the hearer can think that the speaker said that S and proceed to make an inference about why he said it, i.e., about his communicative intention in saying it, even though what is said is not a complete proposition. The process of utterance comprehension is obviously a very interesting topic for psychology, but it’s hard to see why facts about hearers’ cognitive processes should be relevant to semantics. The relevant question is whether the nature of these pragmatic processes, whatever the details, has any bearing on the legitimacy of a purely semantic notion of what is said. It is a mystery to me why facts about what the hearer does in order to understand what the speaker says should be relevant to what the speaker says in the first place. How could the fact
Semantically speaking
(if it is a fact) that what is said sometimes has no psychological reality for the hearer show that it is a mere abstraction? All this shows is that hearers can infer what a speaker is communicating without first identifying what the speaker is saying. Employing the semantic notion of what is said does not commit one to an account of the temporal order or other details of the process of understanding. This notion pertains to the character of the information available to the hearer in the process of identifying what the speaker is communicating, not to how that information is exploited (Bach and Harnish 1979: 91–93). Besides, suppose that it is true that what is said, in the strict, semantic sense, is sometimes not consciously accessed. It is still consciously accessible. This is evident from the fact that people recognize (as we saw with (1)–(3) above) that expansions of it are cancellable. Furthermore, even if in some cases the “minimal” proposition “is not actually computed and plays no role in the interpretation process as it actually occurs” (Recanati 1993: 318), e.g. because of what Recanati calls “local processing” on constituents of the sentence, it can still play a role. Even if a hearer doesn’t explicitly represent what is said by the utterance of a sentence, hence does not explicitly exclude it from being what is meant, still he makes the implicit assumption that it is not what is meant. Implicit assumptions are an essential ingredient in default reasoning in general (Bach 1984) and in the process of understanding utterances in particular. Communicative reasoning, like default reasoning in general, is a case of jumping to conclusions without explicitly taking into account all alternatives or all relevant considerations. Even so, to be warranted such reasoning must be sensitive to such considerations. This means that such considerations can play a dispositional role even when they do not play an explicit role. They lurk in the background, so to speak, waiting to be taken into account when there is special reason to do so. 3.2 The intuitive objection Here the complaint is that the literalistic notion of what is said disregards ordinary intuitions about the truth or falsity, hence about the content, of what is said. For example, it may seem that sentence (7), as uttered on a certain day, (7) I haven’t taken a bath.
is true if the speaker hasn’t taken a bath that day, although on the literalistic view what he said is false if he has ever taken a bath. Of course what the speaker means is true, but that includes the implicit qualification ‘today’, which is not present in the sentence (it is doubtful that a time frame variable is present either).
157
158
Kent Bach
Since the implicit qualification does not affect what is said but is understood right along with what is said, it can seem counterintuitive that what is said is really false. In fact, this is often given as a reason for adopting a more inclusive conception of what is said, one that is not limited to the “minimal proposition” that the literalistic conception of what is said would insist on. Recanati even formulates a principle, the Availability Principle, which prescribes that intuitions about what is said be “preserved” in our theorizing. But he needs to explain why the intuitions he takes to concern what is said actually do concern that, as well as why we should be confident in their accuracy. Of course it is true that we have a perfectly good handle on what our words mean, on how they are put together into sentences, hence on what our sentences mean. Who could deny that competent speakers have such knowledge? Grice himself regarded it “as a sort of paradox [that] if we, as speakers, have the requisite knowledge of the conventional meaning of sentences we employ to implicate, when uttering them, something the implication of which depends on the conventional meaning in question, how can we, as theorists, have difficulty with respect to just those cases in deciding where conventional meaning ends and implicature begins?” (1989: 49). Grice’s paradox may seem especially troubling if it is supposed, as it often is, that accounting for our ordinary judgments about the truth-conditions of sentences is the central aim of semantics. This worry is unfounded. It is the central aim of semantics is to account for semantic facts, not intuitions. People’s spontaneous judgments or “intuitions” provide data for semantics, but it is an open question to what extent they reveal semantic facts and should therefore be explained rather than explained away. Since they are often responsive to non-semantic information, to what is implicit in what is said, they should not be given too much weight. Besides, these intuitions don’t seem to play a role in ordinary communication. In the course of speaking and listening to one another, we generally do not consciously reflect on the semantic content of the sentences we hear or on what is said in their utterance. We are focused on what we are communicating and on what is being communicated to us, not on what is said. Moreover, we don’t have to be able to make accurate judgments about what information is semantic and what is not in order to be sensitive to semantic information. To “preserve intuitions” in our theorizing about what is said would be like relying on the intuitions of unsophisticated moviegoers about the effects of editing on a film. Although people’s cinematic experience is dramatically affected by cuts and camera angles, there is no reason to suppose that their intuitions are reliable about these effects or about how they are produced. Intuitions about what is said may be similarly
Semantically speaking 159
insensitive to the difference between the contribution that is made by the semantic content of a sentence and that made by extralinguistic factors to what an utterance communicates. What worried Grice was not a real paradox.13 Moreover, it is easy to sensitize people’s intuitions about what is said to Grice’s cancellability test for what is not said. Just present them with sentences like (1)–(3) followed by cancellations of what is not explicit in the utterance, as in (1xx)–(3xx), (1xx) Jack and Jill went up the hill but not together. (2xx) Jack and Jill are engaged but not to each other. (3xx) Jill got married and became pregnant but not in that order.
and see if they sense a contradiction or just a clarification. Or ask them whether what is said by an utterance of explicitly expanded versions of (1), (2), or (3) is the same as what is said by utterances of (1), (2), and (3) themselves: (1ee) Jack and Jill went up the hill together. (1) Jack and Jill went up the hill. (2ee) Jack and Jill are engaged to each other. (2) Jack and Jill are engaged. (3ee) Jill got married and then became pregnant. (3) Jill got married and became pregnant.
They are likely to say that the explicitly expanded utterances say something not said by the original ones. Similarly, ask them whether what a speaker says in uttering (8ee) is the same as in uttering (8), (8ee) Howard hasn’t taken a bath today. (8) Howard hasn’t taken a bath.
and they are likely to discern the difference. This contradicts the intuition that in uttering (8) a speaker says that Howard has taken a bath that day. At the very least, the verdict of intuition is less clear when we take into account people’s cancellability judgments and their comparative judgments about what is said by explicitly expanded vs. unexpanded ones.14 So it seems that cognitive and intuitive considerations do not undermine a purely semantic notion of what is said. Furthermore, there is a theoretical need for such a notion, namely, to account for the semantic content of an utterance that is independent of the fact that the utterance is actually made. Its semantic content, relative to a context, does not depend on the sentence actually being uttered.
160 Kent Bach
The context can be an actual or possible context of utterance. Appealing to pragmatic processes and to intuitions is irrelevant to an account of what is said, or would be said in a hypothetical context, because they do not enter in at the locutionary level. It is only at the illocutionary level that such processes are primary or secondary, mandatory or optional. And it is only at that level that the “intuitive truth-conditions” of an utterance come into play. It is irrelevant that when someone utters (9), (9) I have had breakfast.
and means that he has already had breakfast that morning that his utterance would be deemed false if he hadn’t had breakfast that day. What is false is what the speaker is claiming, not what he is saying. What he is saying, the content of the sentence relative to the identity of the speaker and time of the utterance, is true even if he hasn’t had breakfast that day, but this is not what people’s intuitive judgments are focused on. The basic trouble with the various arguments we have considered is their implicit assumption that what is said in uttering a sentence in a context (the semantic content of that sentence relative to that context) can be affected by the speaker’s communicative intention. In fact, what is said is determined linguistically. When a speaker utters a given sentence in a given context, the only intention that is relevant what he is saying is his semantic intention, i.e., his intention concerning the resolution of any ambiguities and the fixing of any indexical references. These may be insufficient to determine a proposition, as when the sentence is semantically incomplete, but that only goes to show that what is said need not be a complete and determinate proposition. This has no special impact on the hearer’s inference to the speaker’s communicative intention (i.e., to what the speaker could plausibly be taken to intend the hearer to take him to mean in saying what he says). Just as the hearer must look for something that is relevantly specific and informative, so, when the sentence is semantically incomplete, he must look for a complete proposition.
4. Diagnosing the objections: Overlooked distinctions How could it even seem that psychological or epistemological considerations about the hearer are relevant to the question of what the speaker says? I believe this is due to overlooking or disregarding certain fundamental distinctions. These distinctions, when heeded, help support a purely semantic notion of
Semantically speaking
what is said. As enumerated below, they all contrast something pragmatic with something semantic. In fact, they derive in one way or another from the basic distinction drawn at the beginning of previous section, between semantic and pragmatic information: D1.Semantic information is encoded in what is uttered, whereas pragmatic information is generated by, or at least made relevant by, the act of uttering it. The distinction between semantic and pragmatic information needs to be heeded whenever it is said, as it often is, that the content of an utterance “depends on context” or is “a matter of context”. It is easy to slide from this platitude to the very strong claim that utterance content is “determined by context” (together with linguistic meaning). That brings in our next distinction: D2.It is one thing for content to be determined in context and quite another for it to be determined by context. To say that content is determined in context is just to say that it can vary from context to context, whereas to say that it is determined by context is to say that it is a function of context. The latter occurs only when the sentence contains context-sensitive elements whose semantic values are functions of objective contextual parameters. This is context, i.e., contextual information, in the narrow, semantic sense. The difference between that and contextual information in the broad, pragmatic sense prompts our next distinction: D3.It is one thing for content to be determined by context in the sense of being ascertained on the basis of contextual information and quite another for it to be determined by context in the sense of being constituted by contextual factors. Determination of the first sort is epistemic determination. It does not make it the case that an utterance has a certain content or, in particular, that the speaker has said a certain thing. Rather, it is what is involved in the hearer’s figuring these things out (as well as what the speaker is communicating). Constitutive determination makes it the case that the speaker has said a certain thing and, more generally, that an utterance has a certain content. Only constitutive determination of what is said is relevant to semantics. What that is can be confusing if we do not heed a further distinction:
161
162 Kent Bach
D4.What a sentence is most typically used to communicate is one thing; what it means may well be something else. As illustrated by our earlier examples (1)–(3) and (8)–(9), e.g., (8) Howard hasn’t taken a bath.
the most natural use of many sentences is not their literal use. The most natural interpretation of an utterance need not be what is strictly and literally said. It might be suggested that to the extent that the meaning of the uttered sentence does not determine the content of an utterance, the speaker’s communicative intention does, and that this intention is part of the context. However, D5.The speaker’s communicative intention is distinct from and not part of the context of utterance. Now if context were defined so broadly as to include anything other than linguistic meaning that is relevant to determining what a speaker means, then of course the speaker’s intention would part of the context. However, if the context is to play the explanatory role claimed of it, it must be something that is the same for the speaker as it is for his audience, and obviously the role of the speaker’s intention is not the same for both. Context can constrain what the speaker can succeed in communicating given what he says, but it cannot constrain what he intends to communicate in choosing what to say. In implementing his intention in order to achieve communicative success (recognition of his communicative intention), the speaker needs to select words whose utterance in the context will enable the hearer to figure out what he is trying to communicate. But this concerns only the epistemic, not the constitutive, determination of the content of his utterance. Finally, we need to apply the elementary distinction between sentences and utterances of sentences to draw a distinction between their contents: D6.The content of a sentence relative to a context is the proposition(s) (possibly incomplete) it expresses relative to that context, whereas the content of its utterance is whatever the speaker is using it to communicate. In my view, the notion of the content of an utterance of a sentence has no independent theoretical significance. There is just the content of the sentence the speaker is uttering, which, being semantic, is independent of the speaker’s communicative intention, and the content of the speaker’s communicative intention, i.e., what the speaker is using the sentence to communicate. When
Semantically speaking 163
one hears an utterance, one needs to understand the sentence the speaker is uttering and to recognize the communicative intention with which he is uttering it. Understanding the sentence is independent of context except insofar as there are elements in the sentence whose semantic value are context-relative. Recognizing the speaker’s communicative intention is a matter of figuring out the content of that intention. Utterances themselves (or sentence tokens) have no content apart from the semantic content of the uttered sentence and the content of the speaker’s intention in uttering it.
5. Conclusions and questions In urging a semantic notion of what is said, I am not implying that the hearer must determine what is said before identifying what is being communicated. So, for example, in hearing (10), (10) The table is covered with dust.
the hearer does not have to entertain the minimal descriptional proposition expressed by (10), that the only table there is is covered with dust, before determining that the speaker is stating of a certain salient table (say the one in front of him) that it is covered with dust. The hearer can figure out that the speaker is talking about that table before he even hears what the speaker is asserting of it. He can figure out that the speaker is conveying an expansion of (10), (10e) The table [in front of me] is covered with dust.
without first having to entertain the proposition expressed by (10). Similarly, if a teacher utters (11) to a class, (11) Everyone should study hard.
he may be presumed to be speaking of the students in the class he is speaking to. They don’t have to entertain the proposition expressed by (11), that all people should study hard, before they can figure out that he is conveying is an expansion of what he said:15 (11e) Everyone [in this class] should study hard.
Even if the hearer does not have to identify what is said before figuring out what the speaker means, the category of what is said still plays important explanatory roles. For one thing, it is needed to exploit the distinction between the process of
164 Kent Bach
identifying what the speaker is communicating and the information available to that process. Then it makes sense to suppose that the process can be sensitive to certain information without the hearer actually entertaining it. This is so because if the sentence didn’t express what it does, the speaker wouldn’t have uttered the sentence that he did and the hearer wouldn’t reason as he does. Moreover, anything that goes beyond what is strictly said is cancellable. Cancellability is a psychologically real phenomenon, and the semantic notion of what is said is needed to explain it. Finally, the semantic notion of what is said is, as we saw in Section 1, needed for formulating the distinction between direct and strictly literal speech and speaking nonliterally or indirectly, and for characterizing misspeaking as well as speech acts that aren’t communicative at all. In defending a purely semantic conception of what is said, I have stressed the following points: –
–
– – – –
The distinction between the semantic interpretation of a sentence and the pragmatic interpretation of an utterance needs to be recognized and respected. What counts as what is said in an utterance of a sentence consists of nothing more than its semantic interpretation, that is, the (possibly incomplete) proposition semantically expressed by the sentence (with respect to the context of utterance). The nature and role of context that is relevant to semantics is limited to the semantic interpretation of indexical elements. The speaker’s communicative intention is not part of the context. Intuitions about what is said should be respected only insofar as they are semantic. Questions about the nature of the cognitive processes involved in understanding an utterance have little if anything to do with the proper formulation of the semantics-pragmatics distinction or with drawing the line between what is said and what isn’t.
Certain questions remain, and answering them requires looking at particular linguistic phenomena. Here is one such question: where there is semantic incompleteness, must it be explained by the presence of free variables? Obviously a free variable can be inserted into a formal representation of a semantically incomplete sentence, but this doesn’t show that the alleged covert variable is really there in the sentence.16 That is a linguistic question, and the answer depends on the case. For example, it is plausible to suppose that relational terms, such as ‘enemy’ and ‘mentor’,
Semantically speaking 165
and perhaps also relative terms like ‘small’ and ‘fast’, have variables associated with them, because, e.g., being an enemy is not a property but a relation and being fast is relative to a category. It is less plausible to suppose this of, say, ‘is noon’ or ‘weighs 80 kg’, even though time of day is relative to a time zone and weight is relative to a gravitational field. In general, we should not assume that metaphysical facts automatically intrude into semantics. Another question concerns how to distinguish cases of expansion (of semantically complete sentences) from cases of completion (of semantically incomplete sentences). Contrary to my assumption that sentences like (10) and (11) above are semantically complete but typically used to mean an expanded version of what they say, it might be argued (see note 15) that they are semantically incomplete and that their completion requires contextually provided domains for the quantificational phrases they contain. Again, it seems that distinguishing cases of expansion from cases of completion is a linguistic question, whose answer depends on specific features of the case. For example, consider the phenomenon of polysemy, as in the following adjectival modifications: (12) sad person / sad face / sad song / sad episode
A sad face is not sad in the way that a sad person is. Rather, it expresses the person’s sadness. A sad song isn’t sad in the way that a sad person is, and need not express any particular person’s sadness, such as the writer’s or the performer’s (nor need it make the listener feel sad). And a sad episode isn’t itself sad — it causes sadness. Now it might be supposed that ‘sad person’ is the semantically basic kind of case and that the others derive from it by pragmatic supplementations or expansions of various sorts. Alternatively, it might be thought that the relation between ‘sad’ and the noun it modifies semantically underdetermines how the two interact (notice that one could use ‘sad’ to say, however absurdly, that a face, song, or episode, is sad in just the way that a person can be sad). I don’t know how to decide between these choices. Finally, here is a pair of cases where the contrast seems clear. (13) I was in Hungary. (14) I have been in Hungary.
The past-tensed (13) strikes me as semantically incomplete, requiring for its completion a temporal qualifier, such as a time-specifying adverb or a ‘when’clause. (14), in the present perfect, seems semantically complete, meaning that I was in Hungary at some time before the time of the utterance. However, I could use it to communicate an expansion of (14), such as (14e),
166 Kent Bach
(14e) I have been to Hungary [on this trip].
say if it were understood that I was on a European business trip during which I was supposed to go to Hungary. Here I will leave open the questions of how to distinguish cases of expansion from cases of completion and of whether semantic incompleteness can only be explained by the presence of free variables. To answer them, perhaps I should return to Hungary.
Notes 1. If logical form is construed as a level of linguistic representation (see May 1991), as opposed to a means of formal regimentation, then Syntactic Correlation in effect requires that what is said by uttering a sentence in a context is its interpreted logical form in that context. This assumes that if what is said is a proposition, it is a proposition whose structure mirrors the syntactic structure of the sentence that expresses it. For a detailed discussion of this conception of structured propositions (see King 1995). 2. To reckon with these various ways of saying something without meaning it, Grice should have invoked Austin’s distinction between locutionary and illocutionary acts. Austin, it may be recalled, defined the locutionary act (specifically the “rhetic” act) as using certain “vocables with a certain more-or-less definite sense and reference” (1962: 95). That sounds a lot like Grice’s notion of saying, except that for Grice saying something entails meaning it: the verb ‘say’, as Grice uses it, does not mark a level distinct from that marked by such illocutionary verbs as ‘state’ and ‘tell’, but rather functions as a generic illocutionary verb that describes any constative act whose content is made explicit. 3. Considering that he describes nonliteral utterances like irony and metaphor not as saying but as making as if to say, it is puzzling that Grice classified them as implicatures. Besides, intuitively one thinks of implicating as stating or meaning one thing and meaning something else in addition, not as meaning something else instead. Implicature is a kind of indirect speech act, whereas irony and metaphor are species of nonliteral but direct speech act (see Bach and Harnish 1979: Ch. 4). 4. Where there is ambiguity or polysemy, how the speaker’s words are to be disambiguated is a matter of the speaker’s intention but not of his communicative intention. Obviously he is not communicating that he is using his words in a certain way. Rather, the supposition that he is using them in a certain way is necessary for inferring his communicative intention, i.e., for inferring what he means. 5. What I call “impliciture” is roughly what Sperber and Wilson (1986) call explicature. But their neologism is misleading, because while it appears to mean the explication or spelling out of what is meant by an utterance, they use it to mean the explicit content of an utterance and, moreover, what they count as explicit includes the “development of the logical form encoded by [the uttered sentence]” (1986: 182), where development involves adding
Semantically speaking 167
elements that are not syntactically represented. Supplementing the sentence with lexical material that is not there would, of course, explicate what the speaker is communicating and thereby make it explicit, but obviously this does not mean that the material was explicit in the original utterance. Rather, to spell out what the speaker could have said to make what he meant fully explicit is to indicate what was merely implicit in what he actually said. 6. Or at least ‘more explicit’. In (1), for example, the words in brackets do not make fully explicit what is meant because the relevant hill is not specified. I suspect that most real-life cases require more elaborate qualifications, often in more than one place in the sentence, to make fully explicit what the speaker means. 7. It might be claimed that they contain covert variables in the position occupied by the bracketed phrases. To support this claim one would have to provide a lexical basis (derived from lexical information about ‘enough’, ‘ready’, or ‘first’) for positing such a variable and an account of the semantic type that a value of this variable must exemplify. Presumably these types would be of the sort designated by the bracketed phrases in (4c)–(6c). Given all this, it could be argued that sentences like (4)–(6) are indexical rather than semantically incomplete. See Stanley (2000) for one defense of this view. 8. In Bach (1999b), I use the IQ Test to show that terms like ‘but’ and ‘nevertheless’ do not generate what Grice called conventional implicatures but make contributions which are included in what is said. 9. One common misconception of the distinction needs to be dispelled. It is often said that semantics is concerned with truth-conditional and pragmatics with non-truth-conditional aspects of meaning. This conception is misguided on several grounds. First, implicatures, which are not semantically determined, have truth conditions. Second, it confuses truth conditions with items that have truth conditions, such as propositions and, as a result of that, leads to the false supposition that what is said must be truth-conditional. It makes sense to speak of open or incomplete propositions but not of open or incomplete truth conditions. 10. The story must be complicated a bit to handle, e.g., recorded utterances of ‘I am not now here’ on a telephone answering machine. In these cases, as Stefano Predelli (1998) has pointed out, it is the time of hearing, not speaking, that is relevant to fixing semantic content. 11. Here I will be discussing Recanati’s most recent formulations of his objections. Earlier formulations and other objections appeared in Recanati (1989, 1993: Chs. 13 and 14, and 1995). I addressed other objections in Bach (1994a: 137–139 and 157–161). 12. The term ‘interpretation’ needs clarification here. Like many verbal nominals, ‘interpretation’ exhibits a process-product ambiguity. With respect to the object sense, the semantic interpretation of an utterance of a sentence is simply the semantic interpretation of the sentence relative to the context of utterance, i.e., its operative linguistic meaning filled in with the semantic values of any of its context-sensitive constituents. The semantic interpretation of a sentence relative to a context constitutes what is said by an utterance of it in that context. The pragmatic interpretation of an utterance is whatever the speaker is attempting to communicate in making the utterance (sentences don’t have pragmatic interpretations). As for the corresponding processes, semantic interpreting is the abstract “process” of assigning meanings to the constituents of a sentences, determining the meaning of the sentence as a function of the meanings of its constituents and their syntactic relationships,
168 Kent Bach
and assigning semantic values to any indexical elements. This process is independent of what particular hearers do when they hear a sentence, although in virtue of their semantic knowledge, together with their knowledge of the values of the relevant contextual parameters, presumably this abstract process is instantiated by the psychological process involved in their understanding of a sentence as uttered in a context. There is no analogously abstract pragmatic process of interpreting. That is something hearers do in the course of identifying speakers’ communicative intentions. 13. Our seemingly semantic intuitions are especially unreliable when there is a recurrent pattern of nonliterality associated with particular locutions or forms of sentence, as with most of the examples above. These are cases not of conventionalization but of standardization (Bach 1995, 1998), in particular, of standardized nonliterality (Bach 1994b: 77–85). Like the more commonly recognized phenomenon of standardized indirection (Bach and Harnish 1979: 192–219), including what Grice called generalized conversational implicature (1989: 37–39), in these cases too it seems that the hearer’s inference to what the speaker means is short-circuited, compressed by precedent (though capable of being worked out if necessary), so that the literal content of the utterance is apparently bypassed. For a monumental study of generalized conversational implicature, with a huge range of data, many of which illustrate what I regard as implicitures (see Levinson 2000). 14. The only empirical (as opposed to armchair) research I’m aware of on intuitions about what is said, by Gibbs and Moise (1997), does not test for this. They claim to show that people’s intuitions accord with the Availability Principle, but their experimental design imposed a false dichotomy on their subjects by forcing them to choose between what is said and what is implicated. Subjects weren’t offered the in-between the category of implicit qualification. Also, Gibbs and Moise did not ask subjects to make cancellability judgments or comparative judgments about what is said by explicitly expanded utterances as opposed to unexpanded ones. 15. In my view, intended restrictions on quantified noun phrases fall within the realm of pragmatics. Stanley and Szabó (2000), however, have given an extensive argument for the claim that such phrases contain covert domain restriction variables, hence that domain restriction is a semantic matter. Bach (2000) is my reply to this argument. 16. Jason Stanley (2000) has argued that, in general, semantically incomplete sentences must contain free variables whose values are somehow supplied by the context, hence that sentences are semantically incomplete only in the way that sentences containing indexicals are incomplete. In my view, it is an empirical question whether a given form of semantically incomplete sentence contains any covert variables, and there is no pat answer to this question — it depends on the type of case.
Semantically speaking 169
References Austin, John L. 1962 How To Do Things With Words. Oxford: Oxford University Press. Bach, Kent 1984 “Default reasoning: Jumping to conclusions and knowing when to think twice”. Pacific Philosophical Quarterly 65: 37–58. 1994a “Conversational impliciture”. Mind & Language 9: 124–162. 1994b Thought and Reference, paperback ed., revised with postscript. Oxford: Oxford University Press. 1995 “Standardization and conventionalization”. Linguistics and Philosophy 18: 677–686. 1998 “Standardization revisited”. In Pragmatics: Critical Concepts, vol. IV, Asa Kasher (ed.), 712–722. London: Routledge. 1999a “The semantics-pragmatics distinction: what it is and why it matters”. In The Semantics-Pragmatics Interface from Different Points of View, Ken Turner (ed.), 65–84. Oxford: Elsevier. 1999b “The myth of conventional implicature”. Linguistics and Philosophy 22: 327–366. 2000 “Quantification, qualification, and context”. Mind & Language 15: 262–283. Bach, Kent and Harnish, Robert M. 1979 Linguistic Communication and Speech Acts. Cambridge, MA: MIT Press. Barwise, Jon and Perry, John 1983 Situations and Attitudes. Cambridge, MA: MIT Press. Carston, Robyn 1988 “Implicature, explicature, and truth-theoretic semantics”. In Mental Representations: The Interface Between Language and Reality, Ruth Kempson (ed.), 155–181. Cambridge, Eng.: Cambridge University Press. Gibbs, Raymond and Moise, Jessica 1997 “Pragmatics in understanding what is said”. Cognition 62: 51–74. Grice, Paul 1989 Studies In the Way of Words. Cambridge, MA: Harvard University Press. King, Jeffrey 1995 “Structured propositions and complex predicates”. Noûs 29: 516–535. Levinson, Stephen 2000 Presumptive Meanings: The Theory of Generalized Conversational Implicatures. Cambridge, MA: MIT Press. May, Robert 1991 “Syntax, semantics, logical form”. In The Chomskyan Turn, Asa Kasher (ed.), 334–359. Oxford: Blackwell. Predelli, Stefano 1998 “I am not here now”. Analysis 58: 107–115. Recanati, François 1989 “The pragmatics of what is said”. Mind and Language 4: 294–328.
170 Kent Bach
1993 Direct Reference: From Language to Thought. Oxford: Blackwell. 1995 “The alleged priority of literal interpretation”, Cognitive Science 19: 207–232. 2001 “What is said”, Synthese 127: 62–79. Sperber, Dan and Wilson, Deirdre 1986 Relevance. Cambridge, MA: Harvard University Press. Stanley, Jason 2000 “Context and logical form”, Linguistics & Philosophy 23: 391–434. Stanley, Jason and Szabó, Zoltán Gendler 2000 “On quantifier domain restriction”, Mind & Language 15: 219–261.
DOCINFO
AUTHOR ""
TITLE "Logical Structures and Universals in Semantics and Pragmatics"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Part III
Logical structures and universals in semantics and pragmatics
AUTHOR "Johan van der Auwera and Bert Bultinck"
TITLE "On the lexical typology of modals, quantifiers, and connectives"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
On the lexical typology of modals, quantifiers, and connectives* Johan van der Auwera and Bert Bultinck University of Antwerp (UIA)
1.
Introduction
Ferenc Kiefer is an authority both on modality and on the interface between semantics and pragmatics (e.g., Kiefer 1983, 1987, 1998, to appear). He makes use of notions from logic and adopts insights from the Gricean framework. As such, he joins linguists and philosophers of language who handle a semantics– pragmatics distinction and test the border lines through investigations of such classic topics as modality, quantifiers and connectives. One of the linguists that he joins is Larry Horn, who has been claiming, since 1972 and most elaborately in 1989 that systems of modals, quantifiers, and conjunctions, and of their combinations with negation, show substantial semantic and pragmatic parallels. The point of the present paper is (1) to take the discussion of the parallels a bit further — a bit further also than van der Auwera (1998b) — and (2) to discuss some of the differences. This will be done from a cross-linguistic point of view, to which extent this paper constitutes an exercise in lexical typology.
2. Similarities The basic relations between the modals for necessity and possibility, all and some, the connectives and and or, and their negations can be described with a ‘semantic map’. One most commonly finds a square, as in the Aristotelian ‘Square of Oppositions’ (see Horn 1989: 6–21) or in the ‘Duality Square’ (Löbner 1985, 1987, 1990). The Aristotelian square is shown in (1).
174 Johan van der Auwera and Bert Bultinck
(1)
A
$
contrariness
E
∀¬, ¬∃ ∧¬, ¬∨
∀ ∧ subalteration I
$ ¬, ¬◊
◊ ∃ ∨
contradictoriness
subcontrariness
subalternation ◊¬, ¬$ ∃¬, ¬∀ ∨¬, ¬∧
O
On the left, we find the positive notions of necessity and possibility. As in the work of Horn, the left-hand corners of the square can be identified with “A” and “I” — from Latin AffIrmo ‘I affirm’. On the right, we find the negative versions: ‘it is necessary that not p’, which is taken to be equivalent to ‘it is not possible that p’ at the top, and ‘it is possible that not p’ and the equivalent ‘it is not necessary that p’ at the bottom. The right-hand corners carry the abbreviations “E” and “O” — from Latin nEgO ‘I negate’. Some of the terminology which developed throughout the ages is explained in (2), as applied to modality. ‘ p’ and ‘ ¬p’ / ‘¬ p’ are contrary; they cannot both be true but they can both be false; b. ‘ p’ and ‘ ¬p’ / ‘¬ p’ are contradictory; they cannot both be true and they cannot both be false; c. ‘ p’ and ‘ ¬p’ / ‘¬ p’ are subcontrary: they cannot both be false but they can both be true
(2) a.
We will not go into the properties of the so-called ‘Duality square’ (Löbner 1985, 1987, 1990). Suffice it to say that it rearranges the four notions, thus brings out different relations, but is nevertheless fully compatible with the Aristotelian square. One does not only find quadripartite schemes; there are also proposals for representations with three, five and six corners or slots (see Horn 1990; van der Auwera 1996a). Why should that be so? Applied to modality, the basic problem, one that Aristotle already struggles with, is that possibility allows two readings. The possibility that is captured in (1) — but also in the Duality square — is interdefinable in terms of necessity in the way shown in (3) (cf. also Kiefer 1987, 1998, to appear). (3) a. b.
p = ¬ ¬p p = ¬ ¬p
On the lexical typology of modals, quantifiers, and connectives
(4)
p Æ p
But there is another sense, one which allows people to say things like (5) and which suggests the equivalence in (6). (5) John may be there and he may not be there. (6)
p = ¬p
The possibility sense of (5) and (6) must truly be different from the one of the Squares, as the demonstration ad absurdum in (7) shows. (7) \
p = ¬ p = ¬
¬p ¬p =
p p
(= (3a)) (= (6))
(8) is a classification of the main “solutions” to the problem described: (8) A. The geometry contains only one possibility type, i.e. a. the possibility characterized by (3), the geometry yields a square, and α. one has no theory about the possibility characterized by (6) (e.g., Aristotle) β. one has a theory about the possibility characterized by (6) (e.g., Horn 1989) b. the possibility characterized by (6), and the geometry yields a triangle (implicitly Jespersen 1917) B. The geometry contains both possibility types, and a. one represents the contradictory of the possibility characterized by (6), and the geometry yields a hexagon (e.g., Blanché 1969)1 b. one does not represent the contradictory of the possibility characterized by (6), and the geometry yields a three-layered square (e.g., implicitly Hintikka 1960 and explicitly van der Auwera 1996a)
The B.b.-type solution that is proposed in van der Auwera (1996a) involves the three-layered square in (9). To avoid confusion, the two possibility readings are given different symbols: ‘ ’ stands for ‘possible’ in the sense of (3), and ‘ ’ for ‘possible’ in the sense of (6).
175
176 Johan van der Auwera and Bert Bultinck
(9) Modality’s three-layered scalar square A 1 $ p, ¬◊ ¬ p ¬$p, ◊¬p
2
I ◊p, ¬$ ¬p
♦p, ♦¬p
$¬p, ¬◊p
3 O
E
‘ ’ = ‘necessary’, ‘¬’ = ‘not’, ‘ ’ = ‘possible’ in the sense of ‘at least possible’, ‘ ’ = ‘possible’ in the sense of ‘only possible’, sometimes called ‘contingent’, the sense in which ‘ p’ = ‘ ¬p’, ‘Æ’ = ‘entailment’
The representation in (9) is a square. Like the Aristotelian square it represents the interdefinability of necessity and possibility (3). But unlike the Aristotelian square, (9) has three layers, and thus conforms with Jespersen’s (1917) tripartition of the modal field as well as with Hintikka’s (1960) reconstruction of Aristotle’s views on modality. But (9) is also scalar, and as such it allows us to incorporate some of Horn’s insights on the scalar nature of modal concepts. In van der Auwera (1996a) this incorporation happened along orthodox Grice– Horn lines; in particular, ‘ p’ was argued to have the negation of the higher value ‘ p’ as a scalar Quantity-1 implicature, i.e. ‘¬ p’, which, in combination with ‘ p’, would yield ‘ p’. But a non-Gricean account is equally feasible, cf. Bultinck (1999), who accounts for the usual “exact”-interpretation of the linguistic realizations of the I-corner semantically rather than pragmatically. In this perspective, the linguistic expressions that have traditionally been associated with the I-corner, are considered as the linguistic realizations of the very center of the square (representing the ‘only’-readings). Also, the scalarity that is built into the square takes care of the (rather few) linguistic phenomena that Griceans explain by means of an ‘at least’-semantics. The problems that exist for ‘possible’ also exist for the corresponding quantifier and conjunction. The quantifier is ‘some’. Its two readings are illustrated in (10). (10) I read some books. a. ‘some if not all’ b. ‘only some, not all’
The connective is ‘or’ and its two readings, called ‘inclusive’ and ‘exclusive’, are illustrated in (11).
On the lexical typology of modals, quantifiers, and connectives 177
(11) He spoke with journalists or academics. a. ‘with people that were journalists or academics or both’ [inclusive] b. ‘with people that were either journalists or academics but not both’ [exclusive]
The solution for the quantifiers and connectives is not different from that for modality. Again, one can construct a three-layered scalar square, and the two readings of ‘some’ and ‘or’ can be captured, whether pragmatically (Grice, Horn, van der Auwera) or semantically (Bultinck). The three-layered scalar square that generalizes over modal, quantifiers and connectives is shown in (12). (12) The three-layered scalar square for modals, quantifiers, and connectives A
I
$
1
∀ ∧
2
¬$ , ◊ ¬ ¬∀, ∃ ¬ ¬∧, ∨ ¬
3
◊ ∃ ∨
$ ¬, ¬ ◊ ∀ ¬, ¬ ∃ ∧ ¬, ¬ ∨
O
E
$ ‘some’ in the sense of ‘only some’ ‘or’ in the sense of ‘only or’, exclusive disjunction
3. “Cognitive” and “lexical” semantics So far we have worked towards the first goal of the paper, viz. to clarify the similarity between modals, quantifiers and connectives. This similarity, we now claim, is to be situated mainly at what could be called a level of “cognitive” semantics. In cognitive semantics one studies concepts, i.e. meanings that are in some way related to expressions, but the exact relation is of no concern. One can study, e.g., what it means to say that it is not necessary that not p. In what could be called “lexical semantics”, on the other hand, one studies meanings in as far as they are expressed directly in lexical items or conventionalized phrases.
178 Johan van der Auwera and Bert Bultinck
Relevant are in this context the existence of the English items must and need and the idiomatic phrases must not and need not, but one does not look at the phrase it is not necessary that not. The latter is English, but it has not been conventionalized into a lexeme or an idiomatic phrase. Let us now illustrate the difference between cognitive and lexical semantics a bit more, with three examples from the realm of modality. First, the two basic modals, possibility and necessity, are cognitively interdefinable (see (3)). However, as noted independently by Wunderlich (1981: 41) and Horn (1991: 97), with a foreboding in Döhmann (1974: 66, 72), languages can be found to exploit the Aristotelian equivalence in the construction of conventionalized expressions for ‘ p’, but not for ‘ p’. That is to say that languages regularly express ‘necessary’ as ‘not possible not’ but perhaps they never express ‘possible’ as ‘not necessary not’. An example of a ‘not possible not’ strategy conveying necessity comes from Bangla. (13) Bangla (Dasgupta 1980: 114) o na hese parlo na. he not laughing could not “He couldn’t not laugh.” ‘He couldn’t help laughing / He had to laugh.’
Second, nobody in his/her right mind would doubt that necessity is different from possibility. (14)
p π p
Let us call this a cognitive law. As to the lexicon, however, one does not have to look hard for languages that have modal words that do not distinguish between possibility and necessity. One example is Danish må, which is shifting from a possibility to a necessity sense. (15) Danish (cf. Davidsen-Nielsen 1990: 187) Nu må du fortælle. now may/must you tell ‘Now you may/must tell a story.’
This process is not at all exotic. It happened in English, Dutch and German too, with the cognates of må, but there the process ended half a millennium ago. Another type of vagueness between possibility and necessity is illustrated with the German “modal passive”.
On the lexical typology of modals, quantifiers, and connectives 179
(16) German (cf. Drosdowski 1984: 106) Wesensprobleme sind mit den Mitteln der Ontologie zu lösen. problems.of.being are with the means of.the ontology to solve “Problems of being are to be solved with the means of ontology.” ‘Problems of being can/must be solved with the means of ontology.’
And yet another type is found in the Swedish “acquisitive” modal få ‘get’. (17) Swedish (Wagner 1976: 56) Lasse få köra bil. Lasse may/must drive car “Lasse gets to drive the car.” ‘Lasse may/must drive the car’.
Our third example concerns the cognitive fact that possibility is not the same as impossibility. (18)
p π ¬ p
Still, we have come across expressions that are vague between possibility and impossibility. One is the case of varieties of spoken American English not distinguishing between stressed can and can’t. The other case concerns varieties of Romani that reinterpreted the original impossibility particle našti into a possibility marker — through associating negativity with a free standing originally optional negative element and reinterpreting the -ti part of našti as a subjunctive element (Boretzky 1996).
4. Lexical differences Modals, quantifiers and connectives may be very similar on a cognitive level, on the lexical level, however, it seems that they are quite different. This point can be made with a review of the three examples of the preceding section. First, we have seen that languages regularly express ‘necessary’ as ‘not possible not’. We have not come across any language, however, that regularly expresses ‘all’ as ‘not some not’ or ‘p and q’ as ‘not not-p or not-q’. Second, modal expressions commonly allow a vagueness between necessity and possibility. How about quantifiers that are vague between ‘all’ and ‘some’ readings, and connectives that are vague between ‘and’ and ‘or’ readings? With respect to quantifiers, we are aware of only one case.2 In Hungarian (of all languages!) the element valamennyi allows both ‘all’ and ‘some’ readings.
180 Johan van der Auwera and Bert Bultinck
(19) (Hungarian) Itt van valamennyi. here is some/all ‘Here is some/all.’
The two readings can often be kept apart by stress and, if valamennyi is the object, by the definite vs. indefinite object agreement on the verb (see (20)), but at least in (19), valamennyi is truly vague. (20) (Hungarian) a. Olvas-ok valamennyi-t. read-1sg.indef some-acc ‘I am reading some.’ b. Olvas-om valamennyi-t. read-1sg.def all-acc ‘I am reading all’.
For a non-Gricean, diachronic account of this vagueness, one can consult Haspelmath (1997: 127–128). With respect to conjunctions, the situation is a little tricky. First, it is clear that there are contexts in which the difference between ‘and’ and ‘or’ gets neutralized, in the sense that the overall semantics can be arrived at with the meaning of ‘and’ as well as with the meaning of ‘or’ (see Oetke 1981). Still, the occurrences of the ‘and’ and ‘or’ words retain separate meanings. Second, Gil (1991) describes Maricopa as a language that lacks an expression for ‘and’ and in which the mere juxtaposition of, e.g., two noun phrases could correspond to both ‘and’ and ‘or’. Still, we are not dealing with one conjunction that is vague between ‘and’ and ‘or’. Third, modal expressions have been documented that allow a vagueness between possibility and impossibility. We have not, however, seen any mentioning of a quantifier that would be vague between ‘some’ and ‘no’ or a connective that would be vague between ‘or’ and ‘neither … nor’. We conclude that on the lexical level, modals, quantifiers and conjunctions may well be more different than on the cognitive level. Still, lexical similarities also exist. The cases known to us concern quantifiers and conjunctions with identical cognitive coordinates that use identical morphemes. In Tamil the clitic -u¯ means ‘and, all’ and -o¯ means ‘or, some’ (Annamalai and Steever 1998: 104) — see also Steever (1998: 135, 146–147) on Kannada, and Gil (1994) on the ‘and’ — ‘all’ similarity in a wider variety of languages (but interestingly, he claims that the relevant quantifier is not really ‘all’ but ‘each’).
On the lexical typology of modals, quantifiers, and connectives
5. A cognitive difference? The one difference between modals, quantifiers and conjunctions that has been hinted at in the literature (Horn 1989: 260) is the fact that languages may have special expressions for ‘¬ p’. In English this expression is the auxiliarized need, without -s in the third person. (21) can only mean ‘¬ p’, not ‘ ¬p’, and (22) shows that this use of need is impossible in a positive context. We thus see that need is specialized for the ‘¬ p’ meaning.3 (21) He needn’t do that. (22) *He need do that.
Other languages exhibiting this kind of specialization include Old English (mister, thurfen), Dutch (hoeven), German (brauchen), Old German (dürfen), Danish (behøve), Finnish (pitää), Greek (chriazete), Mandarin (yóng) and Kashmiri (lag) — see De Haan (1994), Van der Wouden (1996b) and van der Auwera (1996b, 1998a). A specialized lexicalization for the ‘¬ p’ meaning may well be relatively rare — this is the perspective from which need type verbs have been approached by Horn (1989) and Van der Wouden (1996a,b). The point that is relevant now is that this lexicalization is at least attested, whereas the corresponding lexicalization for quantifiers and connectives is unattested. That is, we know of no quantifier that is specialized for the ‘not all’ meaning or of a connective specialized for the ‘not and’ meaning. This difference between modals, on the one hand, and quantifiers and connectives, on the other hand, we will claim below, may well be situated at the cognitive level. The ‘middle’ modal, quantifier, and connective, ‘possible’, ‘some’, and ‘or’, so we argued in Section 2, is vague between two readings: ‘at least possible/ some/or if not necessary/all/and’ and ‘only possible/some/or and not necessary/ all/and’. The two readings are represented in (23) by shading.
181
182 Johan van der Auwera and Bert Bultinck
(23) ‘possible’, ‘some’, and ‘or’ as vague between ‘at least’ and ‘only’ A
I
$
1
∀ ∧
◊ ∃ ∨
2 ¬$, ◊ ¬ ¬∀, ∃ ¬ ¬∧, ∨ ¬
3
$ ¬, ¬ ◊ ∀ ¬, ¬ ∃ ∧ ¬, ¬ ∨
O
E
For ‘possible’, however, but not for ‘some’ and ‘or’, there is a third reading, viz. ‘at most possible’. When something is at most possible, it is possibly possible, and when something is possibly possible, this entails that it is possible. That is to say that ‘at most possible’ is a subtype of ‘possible’. This argument does not go through for quantifiers and connectives. When there are at most some, there are possibly some, but the latter does not entail that there are some. Similarly, ‘at most or’ means ‘possibly or’, but ‘possibly or’ does not entail ‘or’. The threeway vagueness of ‘possible’ is represented by the shading in (24). (24) ‘possible’ as vague between ‘at least’, ‘only’, and ‘at most’ A
I
$
1
◊
2
¬$, ◊ ¬
$ ¬, ¬ ◊
3 O
E
We thus see that the cognitive map for the modals is a bit different from that for the quantifiers and the connectives, and that the difference concerns the “O” value. Perhaps it is fair to say that the “O” value is more prominent for modals than it is for quantifiers and connectives. Only in the case of the modals does the “O” value take part in the vagueness of the mid-value. The vagueness of ‘possible’ extending to ‘at most possible’ is not of the type that Griceans or anti-
On the lexical typology of modals, quantifiers, and connectives
Griceans are concerned with, hence it does not fall under the ban of the separate lexicalizability of ‘only’ and ‘at least’ meanings. Hence the modal “O” value can be lexicalized separately, and if this happens, it makes sense that a language does not do this in terms of a possibility word — for the issue that we are dealing with here is precisely the vagueness of possibility words — but rather in terms of a necessity word.
6. Conclusion This paper argues that it is worthwhile making a distinction between the “cognitive” and “lexical” semantics of modality, quantifiers, and conjunctions. The semantic similarity between modals, quantifiers, and conjunctions is to be situated at the cognitive level and can be captured with a three-layered scalar square. But even at that level, the similarity is not complete. In particular, there is a distinction with respect to the “O” corner of the said square, a distinction that may be related to the possibility of separately lexicalizing the “O” corner in the case of the modals, but not in the case of quantifiers or conjunctions. There are many other lexical differences between modals, quantifiers, and conjunctions. These have to be dealt with at the level of “lexical semantics”.
Notes * Thanks are due to an anonymous referee and to Edith Moravcsik for drawing our attention to the vagueness of Hungarian valamannyi ‘some/all’, and to Juan de Dios Luque Durán for allowing us to use some of the material in van der Auwera (1998b). 1. In terms of the layered model in (9), this extra notion would combine layers 1 and 3 but exclude 2. 2. We almost have a second example in Warlpiri. Bittner and Hale (1995: 90–94) indicate that the Warlpiri cardinality nominal panu can receive two interpretations: in a weak reading it means ‘many’, while in a strong interpretation it means ‘all’. ‘Many’, however, is a midrange quantifier, unlike ‘some’. 3. ‘Need’ is allowed in other negative polarity contexts too. This issue is addressed by Van der Wouden (1996a,b).
183
184 Johan van der Auwera and Bert Bultinck
References Annamalai, E. and Steever, Sanford B. 1998 “Modern Tamil”. In The Dravidian Languages, S. B. Steever (ed.), 100–128. London and New York: Routledge. Bittner, Maria and Hale, Ken 1995 “Remarks on definiteness in Warlpiri”. In Quantification in Natural Languages, E. Bach, E. Jelinek, A. Kratzer and B. Partee (eds), 81–105. Amsterdam: Kluwer. Blanché, Robert 1969 Structures intellectuelles. Essai sur l’organisation systématique des concepts. Paris: Vrin. Boretzky, Norbert 1996 “Zu den Modalia in den Romani-Dialekten”. Zeitschrift für Balkanologie 32: 1–27. Bultinck, Bert 1999 “The Cardinal Debate”. Paper presented at the Annual Meeting of the Societas Linguistica Europaea, Ljubljana. Dasgupta, Probal 1980 Questions and relative and complement clauses in a Bangla grammar. Dissertation, New York University. Davidsen-Nielsen, Niels 1990 Tense and mood in English. A comparison with Danish. [Topics in English Linguistics 1]. Berlin: Mouton de Gruyter. De Haan, Ferdinand 1994 The interaction of negation and modality. A typological study. Dissertation, USC. Döhmann, Karl 1974 “Die sprachliche Darstellung der Modalfunktoren”. In Logik und Sprache [Exempla Logica 1], A. Menne and G. Frey (eds), 57–91. Bern and München: Francke. Drosdowski, Günther 1984 Duden. Grammatik der deutschen Gegenwartssprache. Mannheim: Dudenverlag. Gil, David 1991 “Aristotle goes to Arizona, and finds a language without and”. In Semantic Universals and Universal Semantics, D. Zaefferer (ed.), 96–130. Berlin: Foris Press. 1994 “Conjunctive operators in South-Asian languages”. In Papers from the fifteenth South Asian Language Analysis Roundtable Conference 1993. A. Davison and F. M. Smith (eds), 82–105. Iowa City: University of Iowa. Haspelmath, Martin 1997 Indefinite pronouns. Oxford: Oxford University Press.
On the lexical typology of modals, quantifiers, and connectives
Hintikka, K. and Jaakko, J. 1960 “Aristotle’s different possibilities”. Inquiry 3: 18–28. Horn, Laurence 1972 On the semantic properties of logical operators in English. Dissertation, UCLA. 1989 A natural history of negation. Chicago: The University of Chicago Press. 1990 “Hamburgers and truth: why Gricean explanation is Gricean”. In Proceedings of the sixteenth Annual Meeting of the Berkeley Linguistics Society, J.-P. Koenig, M. Meacham, S. Reinman and L. A. Sutton (eds), 454–471. Berkeley: Berkeley Linguistics Society. 1991 “Duplex negatio affirmat …: The economy of double negation”. In Papers from the 27th Regional Meeting of the Chicago Linguistics Society. Part two: The parasession on negation. L. M. Dobrin, L. Nichols and R. M. Rodriguez (eds), 80–106. Chicago: Chicago Linguistics Society. Jespersen, Otto 1917 Negation in English and other languages. Copenhagen: Høst. Kiefer, Ferenc 1983 “What is possible in Hungarian?”. Acta Linguistica Academiae Scientiarium Hungaricae 31: 147–185. 1987 “On defining modality”. Folia Linguistica 21: 67–94. 1998 “Presidential address. Modality and pragmatics”. Folia Linguistica 31: 241–253. (to appear)“The pragmatics of modality”. In Handbook of Pragmatics 1998, J. Verschueren, J.-O. Östman, J. Blommaert and C. Bulcaen (eds). Amsterdam and Philadelphia: Benjamins. Löbner, Sebastian 1985 “Natürsprachliche Quantoren. Zur Verallgemeinerung des Begriffs der Quantifikation”. Studium Linguistik 17/18: 79–113. 1987 “Quantification as a major module of natural language semantics”. In Studies in discourse representation theory and the theory of generalized quantifiers, J. Groenendijk, D. de Jongen and M. Stokhoff (eds), 53–85. Dordrecht: Foris. 1990 Wahr neben Falsch. Duale Operatoren als die Quantoren natürlicher Sprache. Tübingen: Max Niemeyer. Oetke, Claus 1981 Paraphrasenbeziehungen zwischen disjunktiven und konjunktiven Sätzen. Tübingen: Max Niemeyer. Steever, Sanford B. 1998 “Kannada”. In The Dravidian languages, S. B. Steever (ed.), 129–157. London and New York: Routledge. van der Auwera, Johan 1996a “Modality: the three-layered scalar square”. Journal of Semantics 13: 181–195. 1996b “On the typology of negative modals”. Paper Colloquium “Perspectives on Negation”, Groningen (to appear in a Benjamins volume, edited by J. Hoeksema et al.) 1998a “On combining negation and modality”. In Proceedings XVIième congrès international des linguistes, B. Caron (ed.), Paper No. 0071. Oxford: Pergamon.
185
186 Johan van der Auwera and Bert Bultinck
1998b
“Sobre la lexicalización de los modales vs. cuantificadores y conjunciones”. In Estudios de tipología lingüistica, J. de Dios Luque Durán and A. Pamies Bertrán (eds), 179–193. Granada: Método Ediciones. Van der Wouden, Ton 1996a Need we say more about polarised auxiliaries? Manuscript. Groningen. 1996b “Three modal verbs”. Paper Colloquium “The Germanic Verb”, Dublin (to appear in a Niemeyer volume, edited by Sheila Watts) Wagner, Johannes 1976 “Eine kontrastive Analyse von Modalverben des Deutschen und Schwedischen”. International Review of Applied Linguistics in Language Teaching 14: 49–66. Wunderlich, Dieter 1981 “Modalverben im Diskurs und im System”. In Sprache und Pragmatik. Lunder Symposium 1980, Inger Rosengren (ed.), 11–53. Lund: Gleerup.
AUTHOR "Noel Burton-Roberts"
TITLE "Grelling’s paradox"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Grelling’s paradox Its significance for linguistic theory* Noel Burton-Roberts University of Newcastle upon Tyne
In this paper, I am going to hang a heavy coat on what might seem a small hook. I hope the attempt will at least amuse Ferenc on his seventieth birthday. The hook is Grelling’s paradox, which involves the (invented) predicates ‘autological’ and ‘heterological’. The paradox is generally agreed in philosophical circles to have been resolved. Nevertheless, I argue that Grelling’s predicates (as usually construed) pose a much more general problem of central relevance to linguistic theory. I will suggest this problem is resolved by an idea I’ll call “the representational conjecture”. This idea has wide-ranging implications but here I concentrate just on its implications for the semantic analysis of Grelling’s predicates.
1.
Grelling’s paradox
‘Autological’ and ‘heterological’ are generally taken to apply to linguistic expressions (LEs) and to mean respectively “true of self” and “not true of self”. On these terms, ‘short’ and ‘polysyllabic’ are autological — because ‘short’ is short and ‘polysyllabic’ polysyllabic. By contrast, ‘long’ is not long, nor ‘monosyllabic’ monosyllabic. So they are heterological. Philosophical interest in these predicates stems from the paradox that seems to arise when we ask whether ‘heterological’ is itself heterological or autological. On the one hand, if ‘heterological’ is heterological, that makes it autological; on the other, if ‘heterological’ is autological, that makes it heterological. Which makes it autological. Which makes it heterological. And so on. This is Grelling’s paradox.
188 Noel Burton-Roberts
It is generally agreed that Ryle (1950) succeeded in dispelling the paradox. He in essence pointed out that the meaning of those predicates is parasitic on the meaning of the predicates they are applied to. For example, ‘heterological’ means “not long” when applied to ‘long’. Ryle would express this by the use of what he calls a “namely rider”: (1) ‘Long’ lacks the property for which it stands, namely the property of being long.
But when we come to unpack what it might mean to say that ‘heterological’ is heterological (or not), we encounter an infinite semantic regress. In Ryle’s words, (2) “‘Heterological’ lacks the property for which it stands, namely that of lacking the property for which it stands, namely that of lacking the property for which it stands, namely…”.
There is no final paraphrase of ‘heterological’ when applied to itself. Without independent identification of “the property for it stands”, it is meaningless. In short, ‘heterological’ applied to itself issues a semantic check that can’t be cashed. The same incidentally goes for ‘autological’. Hence the questions whether ‘autological’ and ‘heterological’ are autological or heterological, and any purported answer to those questions, are simply meaningless rather than paradoxical. I am satisfied that Ryle succeeds in dismantling the paradox. But I want to suggest the problems with ‘autological’ and ‘heterological’ go deeper — and in a way (not generally noticed) that is of fundamental significance for linguistics and the architecture of grammar. In the first place, there is a respectable tradition in which use and mention of LEs are held to be mutually exclusive. In (3) the name ‘John’ is used (to refer to an individual so named), not mentioned, while in (4) the name is mentioned, not used. (3) John is a popular guest. (4) ‘John’ is a popular name.
However, if we accept the usual account of the meaning of ‘auto(hetero)logical’, there is a clear sense in which their use demands that the predicates they apply to be simultaneously used and mentioned. A predicate is true when used in making a statement and it is true/false-of what is mentioned by that statement. So, when a predicate is asserted to be (not) true of itself, it must be regarded as thereby both used and mentioned.
Grelling’s paradox 189
There are theories of Q(uotational)-mention — the mention of LEs that involves the use of quote marks — which allow for simultaneous use and mention. (5) ‘Short’ is an adjective.
These theories have it that in (5), for example, ‘short’ is both used and mentioned — it is used to mention/refer to itself. Lyons (1977: 5–10) explicitly defends this “reflexive use” account of Q-mention and I believe Searle’s account (1969: 73–76) of Q-mention implies it. However, this account analyses Q-mention as involving the use of an expression to mention the expression used to mention it — that is, the use of (an expression)i to mention (the expression used to mention (it)i)i. This feels tricky. It flouts Peirce’s (and others’) stricture that ‘a sign must be other than its object’ (Peirce 1933: II, para 230) and a linguistic ban on i-within-i. It also suggests that ‘John’, say, is both the name of an individual and the name of that name. Other theories of Q-mention are designed to avoid all this, for example Tarski’s (1956) naming theory, and Quine’s (1940). Although these other theories have their own problems (see e.g. Davidson 1965, Saka 1998), they reflect a strong instinct (which I share) that reflexive use as understood above is impossible. However, if reflexive use so understood is impossible, then ‘autological’ and ‘heterological’ should not be coherently usable, not just in application to themselves or each other, but under any circumstances. Against this, we have seen, they are generally thought to be coherent and true of e.g. ‘short’ (‘long’). This might suggest we need to allow for reflexive use after all. However, I shall consider an alternative possibility. In due course I will be proposing that ‘auto(hetero)logical’ do not in fact mean what they are generally taken to mean. But first I need to motivate the proposal. There is a second problem with how ‘auto(hetero)logical’ are generally understood. I suggest there is a significant independent reason to believe ‘auto(hetero)logical’ cannot in fact mean what they are generally taken to mean. Consider (6)–(7): (6) ‘Short’ is autological. (7) ‘Short’ is true of itself.
As usually construed, these are synonymous. Take (7) first. ‘Short’ describes a spatial (or temporal) property of physical objects (or events). Objects/events correctly described by ‘short’, i.e. physical objects/events, are not such as to be true or false. It is only things with logical properties that can be true/false. But
190 Noel Burton-Roberts
(7) demands that something be both short and true (of itself). In other words, I am suggesting that, on the usual construal, (7) is sortally incorrect. And the same must go for (6), if synonymous with (7).
2. Type and token Given the way physical inscription/utterance phenomena produced by speakers are usually thought to be related to linguistic expressions, this sortal incorrectness is barely noticeable. The usual (‘classical’) account of this relation is that LEs on the one hand and physical inscription/utterance phenomena on the other stand in a type–token relation. Classical type/token thought insists that, while types are abstract, tokens are spatio-temporally unique physical phenomena (e.g. Peirce 1933; Lyons 1977; Bromberger 1989; Bach 1995). More specifically, as applied in linguistics and philosophy of language, it insists that linguistic tokens are perceptual phenomena external to mind/brain. In what follows I will refer to such phenomena as “E-physical”. It is classically assumed that speakers “token” — that is, E-physically produce examples/instances/occurrences of — LEs, which are then to be thought of as (abstract) types. On these terms, with LEs thought of as abstract types, it is appropriate to attribute to them abstract (e.g. semantico-syntactic, logical) properties. As regards relevant E-physical phenomena, since these are classically regarded as the tokens (that is, as the examples, instances or occurrences) of those linguistic types, they can be thought of as instantiating the properties of those types. For that matter, they must be thought of as instantiating those properties. Phenomena are assigned to a type (are regarded as its tokens) precisely on the basis that they exhibit the properties that define the type, surely. In this way, classical type/token thought seems to legitimize the assignment of abstract (semanticosyntactic, logical) properties of LEs (the supposed types) to certain E-physical phenomena (their supposed tokens). If this really is an appropriate way to think of the relation between linguistic expressions and relevant E-physical phenomena, then the sortal incorrectness of (6)–(7) is not just difficult to notice — it is doubtful whether they could even be regarded as incorrect. Yet, even on these classical type–token assumptions, it would generally be thought incorrect to ask, for example, what color an LE has. LEs are not generally thought of as colored. But if (6)–(7) are correct, coherent and true, the same must surely go for:
Grelling’s paradox
(8) ‘Black’ is true of itself / autological / black. (9) ‘Red’ is not true of itself / heterological / not red (but black).
and we could not damn the question about the color of LEs as sortally incorrect. It is also worrying that I could have arranged with the printer to make (8)–(9), if coherent, false. We surely want what we say of LEs (and the linguistic in general) not to vary by occasion but to hold good, in Grice’s 1989 word, ‘timelessly’. In this connection, consider (10)–(11): (10) ‘Italic’ ‘Underlined’ ‘In example (10)’
¸ ˝ are autological. ˛
(11) ‘Italic’ ‘Underlined’ ‘In example (10)’
¸ ˝ are heterological. ˛
In respect of this last point, it might be thought we should appeal to the distinction between type and token here and say that (8)–(9) (and (10)–(11)) have to do, not with linguistic types, but just their tokens. But then what of (6)–(7)? (6)–(7) are generally thought to be true irrespective of tokens — and thus as true of the types as such (but see note 6 below). And if linguistic types can be spatially short (or long), we surely cannot deny that they can be colored. Everything else with spatial extent is colored. Classical type/token does not help here, then. On the contrary, I suggest it is classical type/token that is responsible for the sortal incorrectness. As applied in linguistics, I believe, it introduces a sortal error into the foundations of linguistic theory. This sortal error shows up (if you’re looking for it) as a problem for a standard concept of syntax. Consider ‘I introduced him to your friend’. In asking this, I am asking you to consider, not an utterance, but an LE. Translating into classical type/token, the request concerns a linguistic type. Now that type (the LE) includes an instance/occurrence/token of a certain lexical item, the LE ‘friend’. ‘I introduced your friend to my friend’ includes two tokens of that LE. The problem is this. We need to allow, as I just have, that LEs as such (the types) can be syntactically composed of LEs — that LEs are instantiated (tokened) in the structure of other LEs. Classical type–token, with its insistence that types are abstract but tokens E-physical, cannot permit this. Since, classically, tokens are E-physical phenomena, we cannot say of any LE as such (any type)
191
192 Noel Burton-Roberts
that it includes a token of any LE. Types, being abstract, are not composed of physical phenomena; hence they can’t be composed of (classical) tokens of LEs. On these terms, only the E-physically occurrent token of an LE could be composed of tokens of LEs. Lyons (1977: 14), though only in passing, insists on this. But (classically) it is utterance phenomena that are linguistic tokens. Classical type/token thus obliges us to insist that an LE can occur/be instantiated only in the utterance (U) of an LE, never in the LE itself (the type). Thereby the LE itself (as type) is emptied of all structural/lexical properties, now only to be found in U, its supposed token. This makes one wonder, after all, why any physical utterance phenomenon should (or could) be regarded as the token of an LE in the first place. If anything has syntactic composition, surely it is LEs themselves. On the one hand, this consequence of classical type/token thought is radically anti-realist with respect to the usual concept of syntax; on the other, it attributes to E-physical (e.g. phonetic, graphic) phenomena something they arguably don’t have — namely, syntactic structure (otherwise syntax would reduce to phonetics or graphics). In short, classical type/token buys the idea that the linguistic is tokened in perceptual phenomena by forfeiting a standard concept of syntactic structure. I suggest there is a simple solution to this problem. Apart from the classical insistence that linguistic tokens are E-physical (i.e. perceptual) phenomena, it seems entirely reasonable to allow that complex LEs (as types) are indeed composed of one of more tokens of other LEs. So let’s abandon the classical insistence that tokens by definition are E-physical phenomena and allow that there are two distinct concepts of ‘occurrence’ (or ‘token’) in play here. Let’s allow that there are different sorts of tokens and, correspondingly, different sorts of types. Certainly, types are “abstract” in virtue of being types but — independently of that — we need to distinguish between E-physical (perceptual) types and Abstract (formal) types. A type is an E-physical type if and only if it is a type having physical phenomena external to mind/brain as tokens (e.g. mercury, forked lightning, alphabet letters, phonetic phenomena). E-Physical types, then, are the classical types usually assumed in linguistics. By contrast, a type is a Abstract (formal) type (independently of its abstractness qua type) if and only if its tokens are Abstract, formal. Armed with this sortal distinction between types (in terms of the sort of their tokens), we can now insist that linguistic types are Abstract types. As tokens of Abstract types, tokens of LEs are no less abstract or formal than the types in which they occur as syntactic constituents. On these terms, the usually assumed idea of syntactic composition poses no problem.
Grelling’s paradox 193
This “sorted” approach to type–token reflects an assumption quite standard in linguistics: that occurrence-within-the-formal-structure-of-an-LE is a sort of occurrence — namely, linguistic occurrence — absolutely distinct from E-physical occurrence. I believe “sorted” type–token was anticipated by Bar-Hillel’s (1970: 367) distinction “between some abstract linguistic entity… and its equally abstract occurrences” (my emphasis). But all this contradicts the (classical) idea that the linguistic is tokened in physical utterance/inscription phenomena. On these (sorted) terms, what we produce in utterance are tokens of E-physical — not linguistic — types. In short, the foundations of linguistic theory seem to rest on two distinct and mutually inconsistent concepts of type/ token (“classical” vs. “sorted”).1 The conflict between them reflects a more general tension in linguistic theory. The attraction of classical type/token is that it appears to offer itself as a model of the relation between (what Chomsky calls) the internal and the external, between the I-linguistic and the E-physical, LE and U. On inspection, this has proved problematic. Classical type/token also implies that certain E-physical phenomena — those regarded as linguistic tokens — are themselves linguistic. What emerges from this is a view of language as having both an internal facet (in terms of types) and an external facet (in terms of tokens). However problematic (on the above inspection) this is, it is a popular view. Sorted type/token is in marked contrast to classical type/token in both of those respects. Firstly, it has nothing to say about the relation between the linguistic and the E-physical, between LEs generated by an internally constituted grammar and the E-physical products of speakers’ utterance/inscription behavior. Secondly, in denying that any E-physical phenomenon is a linguistic token, sorted type/token denies that the E-physical is linguistic. If the linguistic is “abstract” because mentally constituted, it implies that linguistic tokens are no less mentally constituted (internal) than their types. On these terms, the linguistic has no “external facet”. It is worth noting that this is consistent with Chomsky’s internalist contention that “language has no objective existence apart from its mental representation” (1972: 61). Nevertheless — even if they are not themselves linguistic — it is undeniable that certain E-physical phenomena (viz. those we regard as utterance/inscription phenomena) clearly are related to the linguistic. What then is the nature of this relation? Having dismissed classical type/token as an account of the relation, I urgently need to offer an alternative.
194 Noel Burton-Roberts
3. Representation The alternative that I propose is that speakers produce external phenomena in aid of E-physically representing linguistic expressions (LEs). Call this the “representational conjecture”. Here I can only sketch the main idea and some of its more general implications (see my 1994, 1998, 1999, 2000; BurtonRoberts & Carr 1999; Chng 1999). To avoid misunderstanding, I shall be referring to “M-representation”, ‘M’ for Magritte. This is intended as a reminder of his famous painting La Trahison des Images, which includes some writing, “Ceci n’est pas une pipe”, itself a reminder that in looking at the painting we are not looking at a pipe, but only at a representation of a pipe. In Chomskian linguistics we are used to another, specialized sense of ‘representation’, which I shall call “C-representation”, ‘C’ for ‘Chomsky’ or ‘Constitutive’. Here ‘representation’ is “not to be understood relationally, as ‘representation of’” (Chomsky 1995b: 53, my italics). A C-representation is constitutive of what it (C-)represents. For example, in “language has no objective existence apart from its mental representation” (quoted above), we clearly have C-representation: language as not distinct from, but constituted in, that mental “(C-) representation”. “Linguistic levels of representation”, as usually understood, are C-representations. “Phonological representation”, for example, isn’t a representation of anything, it just is (constitutively) the phonology. The rationale of my appeal to ‘representation’, by contrast, is that an M-representation (M-rep) is not constitutive of what it is a representation of. M-rep emphatically is a relation — a relation between x and y, where x π y. The distinction between representation (M-representans) and what-is-represented (M-representatum) is essential to the conjecture. The (Magrittian) point, in this context, is that what is M-representational of the linguistic is not thereby linguistic. On the contrary, it is thereby not linguistic. In terms of this conception of the relation, Chomsky’s internal/ external distinction is to be thought of as that between the linguistic and what is non-linguistically, E-physically representational of the linguistic. It is in this (M-rep) sense that E-physical phenomena, without being linguistic, can be “relevant” to the linguistic. The relation of M-representation is absolutely distinct from any type–token relation. In producing an M-rep x of y, we produce x, not a token (instance, occurrence) of y.2 M-rep(x,y) implies that y is instantiated independently of the fact of representation. On these terms, speakers do not produce or utter LEs;
Grelling’s paradox 195
they produce or utter only E-physical M-representations of LEs. As for the LEs themselves, these are generated, not produced (or produceable). The conjecture is that, differences between painting and speech notwithstanding, those activities and their products stand in the same logical relation to their objects. It is then no more appropriate to think of utterance behavior as linguistic performance than it is to think that a representational painter, painting a wooden table, is performing (/producing a piece of) carpentry. M-representation is an asymmetric relation, with (E-physical) representans unidirectionally oriented on its object, the (linguistic, mentally constituted) representatum. Furthermore, the fact that it is the object of external M-rep is not a fact about the nature of the linguistic. The linguistic is not, as object of external M-rep, oriented toward the external or “externalizable”, any more than any object is by its nature oriented towards what is M-representational of it. The representational conjecture departs quite radically from tradition in this respect. Rather than the linguistic being thought of as “accommodated to the human sensory and motor apparatus” (Chomsky 1995a: 221), exactly the reverse holds in the conjecture. Here the E-physical (and the sensorimotor) is to be thought of as “accommodated” to the linguistic, in the sense of being produced so as to be M-representational of it. This distinguishes M-representation from the more usual realisation. Underlying “realization” is the idea that the linguistic is in part derivationally targeted on sensory output. By the conjecture, in contrast, sensory output is M-representationally targeted on the linguistic. M-representation is a non-natural relation. As noted, given M-rep(x,y), it is not a natural fact about y that it is the object of some M-rep. No more is it a natural or intristic fact about x that it is an M-rep. It is a matter of the intentions of x’s producer that M-rep(x,y). As regards the perceiver of x, it is a matter of his interpretation of those intentions whether he responds to x as a representation and, if so, what he takes it to be a representation of. Nor is it necessary for any physical M-rep to have any property in common with its representatum. This is crucial. M-rep can be iconic but, being nonnatural, it doesn’t have to be. It can be non-natural in the further sense of being conventional. An iconic M-rep (like Magritte’s of a pipe) needs to share perceptual properties with (i.e. resemble) its representatum. Conventional M-reps, by definition, do not. What is needed instead is a convention. When x and y, M-rep(x,y), lie in the same perceptual domain, we can choose iconic or conventional, or a combination. Otherwise, it has to be wholly conventional. In graphic M-rep of the phonetic and musical, for example, x and y lie in different
196 Noel Burton-Roberts
perceptual domains. And with ‘π’, ‘"’, ‘%’, y isn’t even perceptual, but cognitive (conceptual). The same goes for the linguistic, I assume. In short, E-physical M-rep of the linguistic is wholly conventional. The properties related by conventional M-representation are disjoint. That is, the E-physical M-representations are not (and, given conventions, do not need to be) possessed of the (linguistic) properties they are used to M-represent. This representational approach to the relation allows us to dispense with the idea — and in fact denies — that any single object can possess both the properties we generally think of as core linguistic (lexical, syntactico-semantic, logical properties) and perceptual (phonetic, graphic) properties. Both the M-representational relation and its conventionality are extralinguistic. As one of the terms thereby related, the linguistic itself is not constituted in the relation. On these terms, the thesis that the linguistic itself is a matter of (and that LEs are) sound-meaning pairs, and thus the locus of Saussurian arbitrariness, is to be seen as resulting from conflating one of the terms related (the M-representatum) with the (M-representational) relation itself. In other words, it conflates what is M-represented with how it is M-represented. This idea has a wide range of implications, explored in a preliminary way in Burton-Roberts (2000), which mainly concerns the nature and status of phonology. The usually assumed motive for including phonology within the language faculty is to allow for generated linguistic expressions to be made “visible”, to be “manifested” or “realized”, in phonetic phenomena. The inclusion of phonology is thus closely bound up with classical type/token thought and, I maintain, is heavily implicated in the sortal incorrectness noted earlier.3 The whole point of assuming that the language faculty has a phonological component in addition to semantic and syntactic components (or, as in Minimalism, an “overt” syntactic derivation that splits into (a) a “covert” syntactic derivation serving LF and (b) a “realizational” derivation serving PF) is precisely to conjoin — within the notion of “linguistic expression” — sortally disjoint properties, effectively in aid of allowing that speakers utter/produce what the grammar generates. The representational conjecture (RC) offers a radical (and radically “minimalist”) alternative to this picture of things. In its light, the human sensory and motor apparatus is seen to be necessary, not for the linguistic, but only — and independently — for its external M-representation (in communication among non-telepathic humans). On these terms, there is no reason to assume that the linguistic has to be accommodated to sensorimotor systems and thus no motive for including a phonology component within the human language faculty. The
Grelling’s paradox 197
RC entails that a phonological system is in fact the locus of the (dialect-specific) conventions governing the phonetic M-representation of the properties in terms of which the linguistic faculty operates. Phonology, then, is external to the language faculty. Phonology is the M-representational interface between (wholly non-linguistic) phonetic phenomena and the (wholly non-phonetic) properties defined by a phonology-free language faculty. Burton-Roberts 2000 pursues the implications of the RC for the distinction and relation between Language (L — the unique, invariant and wholly natural object of inquiry of UG theory) and particular languages (regarded as conventional systems for the physical representation of L — CSPR(L)s). Rather than delve further here into the pockets of this coat, I now return to the semantic hook I hung it on.
4. Grelling’s predicates re-analyzed In the light of the representational conjecture, consider again (8)–(9), repeated here. (8) ‘Black’ is true of itself / autological / black. (9) ‘Red’ is not true of itself / heterological / not red (but black).
I suggest that, if classical (E-physical) type/token thought applies to anything in this context, it applies, not to LEs, but only to their E-physical M-representations. Producing an M-rep of an LE entails producing tokens of E-physical (acoustic or graphic) types, not linguistic types. It is the M-reps that physically occur. It is particular occurrences, instances, tokens, not of the LE ‘black’, but of its E-physical representation {black}4 that are black (or not). There being no convention governing what color graphic M-reps must have, they can vary in this respect (though of necessity they will have color). I have suggested that, independently of any paradox, ‘auto(hetero)logical’ — as usually construed — call for (i) reflexive use and/or (ii) sortal incorrectness. Either way (or both ways), I want to insist that they are not — so construed — possible predicates. But then, as mentioned, we need to explain what is going on when, with all appearance of coherence, they are in fact used. The representational idea provides a semantic analysis of ‘auto(hetero)logical’ that captures this, while avoiding both sortal incorrectness and reflexive use of any LE. In its light, those predicates do not apply to LEs. They apply to physical M-representations of LEs. Nor do those predicates mean “true
198 Noel Burton-Roberts
of self”: the M-representations they apply to have only physical, not logical, properties. It cannot then be the physical M-reps that are true/false; it is the linguistic predicates they are M-representational of that are true/false. The revised semantics of those predicates is given in (12). (12) A (physical) representation R of a linguistic predicate P is autological iff P is true of R — and R is heterological iff P is not true of R.
In these terms, the subject of a predication involving ‘auto(hetero)logical’ is not a linguistic predicate but its conventional physical M-representation. It is the M-reps that have color and/or length, not the LEs (the predicates) they are M-reps of. (6)–(7) above must therefore be replaced by (6*)–(7*) respectively. (6*) {short} is autological. (7*) {short} is the M-rep of a predicate (viz. ‘short’) which is true of that M-rep.
Although ‘autological’ and ‘heterological’ receive a very different analysis under the conjecture, intuitions regarding their truth or falsity in particular cases are preserved intact under the re-analysis. Where (6) was thought to be coherent and true, (6*) is coherent — and also true. Where (13) was thought coherent and true, (13*) is coherent and true. (13) !‘long’ is heterological. (13*) {long} is heterological.5
In view of this, I feel entitled to say that the representational analysis accurately captures what is in fact meant by their use. Construed representationally, there is nothing in Grelling’s predicates that is in itself objectionable. This does not re-instate the paradox, however. Independently of any paradox, it is still meaningless to apply ‘autological’ to its own representation as in (14a) (= (14b)). (14) a. {autological} is autological. b. {autological} is the M-rep of a predicate (viz. ‘autological’) which is true of that M-rep.
We can see why, as soon as we try to identify the LE we are mentioning, not (as in (14b)) by means of its usual M-rep, but more directly, as in (14c). (14) c.
{autological} is the M-rep of a predicate (viz. ‘being the M-rep of a predicate (viz., ‘being the M-rep of a predicate (viz… (and so on) …’)… … which is true of that M-rep.
Grelling’s paradox 199
The result is a semantic regress, comparable but not identical to that identified by Ryle. The question whether ‘heterological’ is true of its own conventional M-rep can be no more meaningful than the same question in respect of ‘autological’. Neither has a meaningful answer, hence neither has a paradoxical answer. It is characteristic of the (classical) type/token thought that underlies so much of linguistic theory and philosophy of language — and revealing of the difference between it and the representational idea — that linguistic predicates which happen quite innocuously to be true of the acoustic or graphic phenomena employed to M-represent them should be thought of as true of themselves.
Notes * Thanks to Robyn Carston, Soke Chng, Kasia Jaszczolt and a referee for comments. I am particularly endebted to Sir Peter Strawson for very helpful discussion of Grelling’s paradox in the light of the Representational Conjecture. I gratefully acknowledge that the initial work for this paper was supported by a grant from the British Academy (Research Leave Scheme). 1. Adjusting terminology might appear to make the problem go away: e.g. by restricting ‘occurrence’ to occurrence-within-an-LE and stipulating that ‘tokens’ are not ‘occurrences’. But the adjustment is surely artificial: tokens are occurrences (what is a token of forked lightning but an occurrence of forked lightning?); furthermore, the adjustment simply reaffirms the distinction made here between ‘linguistic occurrence’ and ‘E-physical occurrence’. Legislating against the use of ‘token’ in respect of linguistic occurrence is beside the point provided we don’t confuse the tokening of linguistic types and the tokening of E-physical types. In connection with this proposed adjustment, see Bach 1995; Bromberger & Halle 1997: n.6. Oddly enough, neither of these mentions any problem that the adjustment might be thought to solve. Incidentally I believe that, stripped of the classical insistence that tokens have to be E-physical phenomena, the type/token distinction/relation loses much of its special interest for linguistic theory. 2. I believe Peirce (1933, Vol. 4: 423) actually confuses his type/token (instantiation) relation with that of (M-)representation (the signing relation). An M-representational sign for y is not an instance of y. 3. The sortal incorrectness is what underlies the anomaly of (i) and the invalidity of (ii): (i) (ii) \
John’s book is fascinating and weighs half a kilo. a. That phrase is in John’s book. b. John’s book is in the kitchen. c. That phrase is in the kitchen.
Notice that, if “that phrase” is actually intended as a reference to an M-rep of the phrase, (ii) might well be regarded as valid (if still not entirely felicitous).
200 Noel Burton-Roberts
4. Since linguistic expressions are to be distinguished from their M-representations, I need a means of indicating which I am mentioning. I use the usual Q-marks for the mention of LEs themselves and {braced italics} for the mention of their conventional M-reps. It won’t have escaped the reader’s notice, however, that the M-representations themselves figure (between Q-marks) in the citation of the expressions. This needs explaining. In mentioning an LE, we need some means of identifying which LE is being mentioned. This can be done (i) linguistically — that is, by means of the use of some LE (e.g. ‘the copula’), or (ii) M-representationally — that is, by means of the use, not of any LE, but of its conventional representation (e.g. {be}). Putting Q-marks around the M-representation is the conventional means of indicating that the expression M-represented is being mentioned, not used. Q-mention thus involves the use of an M-representation without the use of the expression it is an M-representation of. I pursue this theory of Q-mention in Burton-Roberts (1999), showing that it offers a genuine alternative to the reflexive use theory, naming theories, and Davidson’s (1979) ‘demonstrative’ theory. 5. (6*)–(7*) and (13*) are judged true in principle and by comparison with other M-representations in English. However, M-representations can be produced in a variety of ways. In particular, as a referee comments, {long} and {short}, or rather their phonetic co-M-reps, could be pronounced with such an exaggerated drawl as to make (6*)–(7*) and (13*) false.
References Bach, Kent 1995
“Type/token distinction”. In Cambridge Dictionary of Philosophy, R. Audi (ed.). Cambridge: Cambridge University Press. Bar-Hillel, Yehoshua 1970 Aspects of Language. Jerusalem: Magnes. Bromberger, Sylvain 1989 “Types and tokens in linguistics” In Reflections on Chomsky, A. George (ed.), 58–89. Oxford: Blackwell. Bromberger, Sylvain and Halle, Morris 1997 “The contents of phonological signs”. In Derivations and Constraints in Phonology, I. Roca (ed.), 93–122. Oxford: Oxford University Press. Burton-Roberts, Noel 1994 “Ambiguity, sentence and utterance: a representational approach”. Transactions of the Philological Society 92 (2): 179–212. 1998 “Language, linear precedence, and parentheticals” In The Clause in English: in honour of Rodney Huddleston, P. Collins (ed.), 33–52. Amsterdam: John Benjamins. 1999 “Ambiguity, quotation, Grelling’s paradox: a representational approach”. Newcastle & Durham Working Papers in Linguistics 5: 75–98. 2000 “Where and what is phonology? A representational perspective” In Phonological Knowledge: Conceptual and Empirical Issues, N. Burton-Roberts, P. Carr and G. Docherty (eds), 39–66. Oxford: Oxford University Press.
Grelling’s paradox 201
Burton-Roberts, Noel and Carr, Philip 1999 “On speech and natural language”. Language Sciences 21: 371–406. Chng, Soke. 1999 Language, Thought and Literal Meaning. Ph.D. Dissertation, University of Newcastle. Chomsky, Noam 1972 Language and Mind. Harcourt, Brace, Jovanovich. 1995a The Minimalist Program. Cambridge, Mass.: MIT Press. 1995b “Language and nature”. Mind 104: 1–61. Davidson, Donald 1965 “Theories of meaning and learnable languages”. In Logic, Methodology and Philosophy of Science, Y. Bar-Hillel (ed.), 383–394. Amsterdam: North Holland. 1979 “Quotation”. Theory and Decision 11: 27–40. Reprinted in his Inquiries into Truth and Interpretation (1984). Oxford: Oxford University Press. Grice, Paul 1989 “Utterer’s meaning and intentions” In Studies in the Way of Words, P. Grice, 86–116. Cambridge, Mass.: Harvard University Press. Lyons, John 1977 Semantics, Volume 1. Cambridge: Cambridge University Press. Peirce, Charles 1933 Collected Papers. Cambridge, Mass.: Harvard University Press. Quine, Willard V.O. 1940 Mathematical Logic. Cambridge, Mass.: Harvard University Press. Ryle, Gilbert 1950 “Heterologicality”. Analysis 11 (1): 61–69. Saka, Paul 1998 “Quotation and the use–mention distinction”. Mind 107: 113–35. Searle, John 1969 Speech Acts. Cambridge: Cambridge University Press. Tarski, Alfred 1956 Logic, Semantics and Metamathematics. Oxford: Oxford University Press.
AUTHOR "Robert M. Harnish"
TITLE "Frege on mood and force"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Frege on mood and force Robert M. Harnish University of Arizona, Tucson
1.
Introduction
Mood is what differentiates such sentences as:1 (1) a. b. c. d. e.
John left the room. [Indicative] Did John leave the room? [Yes/No interrogative] Who left the room? John. [Wh-interrogative] John, leave the room! [Imperative] What, John leave the room?! [Exclamative]
The differences here are not due to words, or at least words alone, for the same words can occur in sentences of different moods. Rather structure plays the major role in determining the identity of mood. Structure includes word order, syntactic categorization, and intonation contour. A theory of mood in part accounts for the intuitive similarities and differences among these sentences, especially with regard to their semantic and pragmatic aspects.2 There are a number of frameworks within which to formulate a theory of mood, but it is widely acknowledged that mood poses a serious obstacle to semantic theories such as Frege’s and Davidson’s which make truth the central concept. Despite attempts (such as Lewis’) to read all moods as indicative by treating them as implicit performatives, consensus is solidly behind the common man in supposing nonindicatives to lack truth values.
2. Conditions of adequacy on a theory of mood What do we want from a theory of mood and what constraints should we put on such a theory? Various authors have implicitly or explicitly suggested constraints on such a theory. Here is a summary.3
204 Robert M. Harnish
2.1 Frege’s Conditions 1.The theory should account for semantic force and content compositionally, insofar as the expression is compositional. 2.Nonindicatives are not true or false. 3.An unambiguous indicative sentence can be uttered nonassertorically as in a disjunction, a conditional, or in a joke or a play (mutatis mutandis for nonindicatives). 2.2 Davidson’s Conditions4 4.The information must be general enough that all sentences with the same mood share relevant aspects of meaning and use, and sentences in different moods can share aspects of meaning and use. 5.The theory must assign an element of meaning to utterances in a given mood that is not present in utterances in other moods. And this element should connect with the difference in force between assertions, questions and commands in such a way as to explain our intuition of a conventional relation between mood and use. 2.3 McDowell’s Condition 6.The theory should allow us to infer what act(s) is being directly performed in the (serious and) literal utterance of sentences in each mood, given contextual information and definitions of the various illocutionary acts. That is, the theory should allow us to infer an illocutionary act (IA) from an utterance act (UA) given the appropriate contextual information (McDowell 1976): (UA) S (seriously and) literally uttered e, (IA) In (seriously and) literally uttering e, S directly F-ed that p.
2.4 Grice’s Condition 7.The theory must not multiply ambiguity, nonliterality or indirection beyond necessity (Grice 1975).
Frege on mood and force 205
3. Frege Frege’s semantics was based on a number of general principles concerning the structure of sentences. Most centrally there are the following: 3.1 General Principles Syntactic Compositionality Principle: All complex expressions are syntactically composed by fitting function names to argument names (singular terms, or names of the appropriate level functions). Syntactic Decompositionality Principle: Complex expressions can be decomposed (not necessarily uniquely) into function names and argument names. Semantic Compositionality Principle: Let expression E be composed of e1 + e2 + … + en, then, the sense, and reference, of E is determined by (is a function of) the sense, and reference respectively, of e1 + e2 … + en, plus their syntactic relations. Consequences of these principles: 1.Substitution: The sense, and reference, of a complex expression E remains unchanged by substitution of co-sensical, and co-referential constituent expressions respectively. 2.Failure: If a constituent expression of E fails to have a sense, or to refer, then E fails to have a sense, or to refer respectively. Indirect sense and reference: The indirect reference of an expression is its customary sense. 3.2 Mood Data Here are some comments of Frege’s on the various moods. We begin with indicatives as a kind of reference point, since it was the focus of most of Frege’s research. 3.2.1 Indicatives 1.An indicative sentence ‘contains’ an assertion and a thought:5 Two things must be distinguished in an indicative sentence: the content … and the assertion. The former is the thought, or at least contains the thought. (Frege, “The Thought”, 21)
206 Robert M. Harnish
2.The sense of an indicative sentence is a thought and its reference is a truth value: The thought … cannot be the reference of a sentence, but must rather be considered as the sense. (Frege, “On Sense and Reference”, 62) We are therefor driven into accepting the truth value of a sentence as constituting its reference. (Frege, “On Sense and Reference”, 63)
3.The thought expressed is that the conditions associated with the sentence for naming the True are fulfilled: The sense of this name [indicative sentence] — the thought — is the thought that these conditions are fulfilled. (Frege: The Basic Laws of Arithmetic, #32)
4.A judgment is the acknowledgment of the truth of a thought: judging, we may say, is acknowledging the truth of something; what is acknowledged to be true can only be a thought” (Frege, “Negation”, 126)
5.An assertion is the manifestation of a judgment: Judgment is made manifest by a sentence uttered with assertive force. (Frege, “Compound Thoughts”, ftnt. 2) The manifestation of this judgment … [is] assertion. (Frege, “The Thought”, 22)
6.The truth claim of an assertion arises from the form of the declarative sentence: We declare the recognition of truth in the form of an indicative sentence (Frege, “The Thought”, 22) We have no particular clause in the indicative sentence which corresponds to the assertion, that something is being asserted lies rather in the form of the indicative. (Frege, “The Thought”, ftnt 1)
7. Assertive force is absent if the requisite seriousness is lacking: Therefore it must still always be asked, about what is presented in the form of an indicative sentence whether it really contains an assertion. And this question must be answered in the negative if the requisite seriousness is lacking. (Frege, “The Thought”, 22)
Frege on mood and force 207
3.2.2 Interrogatives 1.Questions in general have two components: “In a question we can distinguish between the demand for a decision and the special content of the question, the point we are to decide. (Frege, “Negation”, 117)
2.A Y/N interrogative can be formed from every indicative: A sentence question can be formed from every indicative sentence. (Frege, “The Thought”, 21)
3.A Y/N interrogative ‘contains’ a request and a thought: An interrogative sentence [−question] and an indicative one contain the same thought: … but the interrogative sentence contains something more too, namely a request. (Frege, “The Thought”, 21) A sentence question contains a demand that we should either acknowledge the truth of a thought, or reject it as false. (Frege, “Negation”, 117)
4.A Wh-interrogative ‘contains’ a request, and obtains a thought through completion of what is asked: In a word-question we utter an incomplete sentence which only obtains a true sense through the completion for which we ask. (Frege, “The Thought”, 21)
3.2.3 Imperatives 1.An imperative ‘contains’ a request and a sense, but not a thought: One does not want to deny sense to an imperative sentence, but this sense is not such that the question of truth could arise for it. Therefore I shall not call the sense of an imperative sentence a thought. (Frege, “The Thought”, 21)
2.The sense of an imperative lies on the same level as a thought: A command, a request, are indeed not thoughts, yet they stand on the same level as thoughts. (Frege, “On Sense and Reference”, 68)
3.2.4 Subordinate interrogatives and imperatives These have questions and directives as their indirect reference: In the cases so far considered the words of the subordinate clauses had their indirect reference, and this made it clear that the reference of the subordinate clause itself was indirect, i.e. not a truth value but a thought, a command, a
208 Robert M. Harnish
request, a question. The subordinate clause could be regarded as a noun, indeed one could say: as a proper name of that thought, that command, etc., which it represented in the context of sentence structure. (Frege, “On Sense and Reference”, 68)
3.2.5 Exclamations These do not express thoughts: Sentences expressing desires or requests [optatives?] are ruled out in the same way [from expressing thoughts]. (Frege, “The Thought”, 21) An exclamation cannot be regarded as a communication on this account, since no corresponding sentence-question can be formed. (Frege, “The Thought”, 21)
3.2.6 The challenge To construct a theory which meets the conditions of adequacy and conforms to Frege’s general principles and specific observations.
4. Simple sentences: A minimalist Fregean semantics Following Frege we will say that a sentence ‘contains’ its force, ‘expresses’ its sense, which ‘determines’ its reference, and we will say the sentence as a whole ‘refers’ to its truth value. 4.1 Declarative (/-)6 1.‘/-S’ contains an assertion: the manifestation of a judgement which is the recognition of the truth of the thought expressed. 2.‘/-S’ expresses a sense: the thought that the conditions for the sentence naming the True are fulfilled. 3.‘/-S’ refers to a truth value: the True or the False. 4.2 Y/N Interrogative (Y/N?)7 1.‘Y/N?S’ contains a request: the manifestation of a demand that H assert whether or not the thought expressed names the True. 2.‘Y/N?S’ expresses a sense: a thought. 3.‘Y/N?S’ refers to: ?
Frege on mood and force 209
4.3 Wh-Interrogative (Wh-?)8 1.‘Wh-?S’ contains a request: the manifestation of a demand that H specify an argument that falls under the concept. 2.‘Wh-?S’ expresses a sense: a thought function.9 3.‘Wh-?S’ refers to: ? 4.4 Imperative (!)10 1.‘!S’ contains a request: the manifestation of a directive (“a command, request, etc.”) that H is to fall under the concept. 2.‘!S’ expresses a sense: a sense-concept. 3.‘!S’ refers to: ?
5. Simple sentences: An extended Fregean semantics The sketch given so far is fairly faithful to Frege’s fragmentary comments, but there are lacunae in the theory. In particular: (1) what are the referents, if any, of interrogative and imperative sentences, and (2) how are these related to the reference of their constituents? It might be questioned whether such lacunae should be filled. For instance, after reviewing Frege’s comments on imperatives from “On Sense and Reference” just quoted, Carl (1994: 93) concludes: “Thus, the distinction between sense and reference cannot be applied to all sentences.” However, all that really follows from Frege’s remarks is that the sense of imperatives is not a thought, and the reference is not a truth value. Is there something else that could play these roles? Note that the considerations that induced Frege to find a reference for indicative sentences also apply to nonindicative sentences.11 1.For instance, Frege argues in “On Sense and Reference” that: The fact that we concern ourselves about the reference of a part of the sentence indicates that we generally recognize and expect a reference for the sentence itself. (Frege, “On Sense and Reference”, 63)
By parity of reasoning, interrogatives and imperatives should have references too, since surely their constituents also have references just as indicatives do. 2.Frege also says in “On Sense and Reference”:
210 Robert M. Harnish
What else but the truth value: (i) belong generally to every sentence if reference of its components is relevant, (ii) remains unchanged by substitution of co-referring expressions. (Frege, “On Sense and Reference”, 64)
Analogous points holds for interrogatives and imperatives. For instance, if an interrogative such as ‘Does John smoke?’ is truly answered by ‘Yes’, then if John is Mary’s brother, the question ‘Does Mary’s brother smoke?’ is also answered by ‘Yes’. The same for complying with an imperative. 3.Finally, Frege comments in “On Sense and Reference”: The striving after truth drives us always to advance from sense to reference. So we are driven to accepting the truth value of a sentence as its reference. (Frege, “On Sense and Reference”, 63)
Likewise, one could argue, with imperatives the striving for compliance, and for interrogatives the striving for true answers, drives us from sense to reference, and so we are driven to accepting references for these sentences as well. And this suggests what the analog of truth values for nonindicatives will be. We call these ‘satisfaction values’: Satisfaction Values /-: True False ?: TA (True answer) FA (False answer)12 !: C (complied with) NC (not complied with) Suppose we generalize the Fregean schema derived from Section 3.2.1, point 3 above: Generalized Fregean Sense Schema (GFSS): The sense of a sentence is the condition13 that must obtain if the sentence is to refer to the: T, TA, C. In this way, indicatives will have truth-conditions as their sense, interrogatives will have true answerhood conditions as their sense, and imperatives will have compliance conditions as their sense. We can say that when these conditions are fulfilled the sentence is ‘satisfied’, and we can call these conditions ‘satisfaction conditions’.14
Frege on mood and force
5.1 A potential problem In his discussions of assertion Frege had mood take scope over the thought expressed. Given that predicates name concepts and concepts are functions from objects to truth values, a sentence of the form /-[Fa] will manifest a judgment that is true if and only if a is F. Here name, predicate and mood all fit together. Now suppose the sentence is a simple nonindicative. The basic problem lies in the fact that the satisfaction value of indicative sentences (i.e. its ‘reference’ in Frege’s terms) is inherited ultimately from the values of the references of its predicates which are functions from objects to truth values. So how do we get a non truth value as the satisfaction value for the sentence without changing the semantics of predicates? This is a serious, nasty problem, and it may indicate that Frege’s approach is at bottom unalterably assertoric. But let’s see if anything can be done. There seem to be two broad options, either: 1. the range of values for concepts must be extended to include other satisfaction values (changing their semantics), or 2. the mood must operate on truth values to yield these other values. The former has serious problems and is so unFregean that we should try to pursue option 2. Recall that Frege decomposes what has come to be called the assertion sign ‘/-’ into the vertical stroke ‘/’ which expresses assertion proper15 and the horizontal ‘-( )’ which maps everything onto the truth values True and False. Suppose we decompose the signs of the interrogative ‘?’ and the imperative ‘!’ moods in the same way, for instance ‘?*’ for the interrogative and ‘!#’ for imperatives. We stipulate that, just as for assertion, the ‘*( )’ portion of the ‘?*’ sign maps everything that follows it onto TA or FA, and the ‘#( )’ portion of the ‘!#’ sign maps everything onto C or NC. Two plausible principles for this might be: *( ): T Æ TA, F Æ FA #( ): T Æ C, F Æ NC The other portion of the mood sign signals the usual force. In this way the predicate can still yield truth values, while the sentence has answerhood and compliance values. It might be objected that we have yet to make out an intuitive connection between sense and reference for nonindicative sentences — how are questions and commands ‘modes of presentation’ of their satisfaction values? But this oddity is no different for Frege’s account of indicatives. Think about it. Frege’s
211
212 Robert M. Harnish
conception of the relation of sense to reference is that the sense is the mode of presentation of the reference and a model of this relation is the way being the Morning Star is a mode of presentation for Venus. For Frege truth values and thoughts are objects among other objects, and it is just as hard to view thoughts as ‘modes of presentation’ for truth values (on the above model of singular terms) as it is to view a question as a mode of presentation of an answer, and a command as a mode of presentation of a compliance. 5.2 Simple sentences: Extended Fregean semantics for mood We can now extend our previous semantics of indicatives to nonindicatives including interrogatives and imperatives. 5.2.1
Indicative (as before)
5.2.2 Y/N Interrogative (Y/N)16 1.‘Y/N?S’ contains a request: the manifestation of a demand that H assert whether or not the thought expressed names the True. 2.‘Y/N?S’ expresses a sense: a thought. 3.‘Y/N?S’ refers to TA, FA.17 5.2.3 Wh-Interrogative18 1.‘Wh-?S’ contains a request: the manifestation of a demand that H complete the expressed sense. 2.‘Wh-?S’ expresses a sense: a thought function. 3.‘Wh-?S’ refers to TA, FA. 5.2.4 Imperative19 1.‘!S’ contains a request: the manifestation of a directive (“a command, request, etc.”). 2.‘!S’ expresses a sense: a directive. 3.‘!S’ refers to C, NC.
6. Compound sentences So far we have restricted our attention to simple sentences. There are two main types of compounding to consider: in the first mood takes scope over the
Frege on mood and force
connective (mood has ‘wide scope’), and in the second the connective takes scope over mood (mood has ‘narrow scope’). In what follows we will only investigate the standard logical connectives (and: &, or: V, if-then: Æ, Not), but for a complete theory all connectives that can participate in compound moods would have to be considered.20 6.1 Moods of truth functions: Wide scope Consider the case where mood takes scope over the connective. In that case our previous account of simple sentences extends to complex sentences, since the connective is embedded in the structure of the content. This case is no more, and no less, problematic than the simple case. 6.1.1 Truth functions of moods: Narrow scope The problem: Frege viewed the logical connectives as naming functions from truth values to truth values. So if these connectives can connect sentences in nonindicative moods, with nonindicative semantic values, we have the problem of how to interpret the connectives in these cases: 1. Generalize the connectives to include all satisfaction values 2. Define different connectives for different moods and combinations of moods We will not try to defend the first choice here, lets just explore it and see where it leads. If it works, that will be something to be said for it, if it doesn’t, that will count against it. There are at least two kinds of cases to consider: (1) first, there are the uniform compound moods — where the connective connects sentences in the same moods. (2) Second, there are mixed cases, where the connective connects sentences in different moods. Caveats: (1) we should not assume that each of these possibilities is logically distinct, (2) we should not assume that each of these makes sense semantically, and (3) we should not assume that a given natural language such as English realizes all these possibilities. 6.1.2 Uniform moods There are choices among nine possibilities here: four logical connectives, two interrogative moods, imperative mood, and two scopes for mood. We can organize these possibilities as shown in Table 1 below (brackets ‘[ ]’ indicate scope; read ‘?x’ as specify the x which …).
213
214 Robert M. Harnish
Table 1.Uniform mood possibilities and actualities in English Wide/Narrow
Let’s look at these possibilities to see which are realized in English, for example, and which might be ‘logically’ equivalent to others. 6.1.2.1 Yes/No Interrogatives. Conjunction (2) a.
Wide scope: there seems to be no way to form these in English which makes them distinct from narrow scope b. Narrow scope: ‘Does John smoke and does Mary smoke?’
This is truly answered positively (‘Yes’) if and only if each conjunct is truly answered positively: [TAP & TAP] ´ TAP. This is truly answered negatively (‘No’) if and only if at least one conjunct is truly answered negatively: [TAN ⁄ TAN] ´ TAN. Disjunction (3) a.
Wide scope: there seems to be no way to form these in English which makes them distinct from narrow scope b. Narrow scope: ‘Does John smoke or does Mary smoke?’
This is truly answered positively (‘Yes’) if and only if one of the disjuncts is truly answered positively: [TAP ⁄ TAP] ´ TAP. It is truly answered negatively (‘No’) if and only if both of the disjuncts are truly answered negatively: [TAN & TAN] ´ TAN. Conditional These are not well formed in English:21
Frege on mood and force
(4) *If did you leave the room, then did you drink some milk?
The problem: How to account for this fact? Negation (5) a. Wide scope: ‘Is it the case that John does not smoke?’22 b. Narrow scope: *‘Not John smokes?’ (*‘It is not the case that does John smoke?’)23
(5a) is truly answered positively (‘Yes’) if and only if the questioned clause (‘John does not smoke?’) is truly answered positively. It is truly answered negatively (‘No’) if and only if the questioned clause (‘John does not smoke?’) is truly answered negatively. (5b) is not well formed in English. The problem: How to account for this fact? 6.1.2.2 Wh Interrogatives. Conjunction (6) a. Wide scope: ‘Who ate and drank?’ b. Narrow scope: ‘Who ate and who drank?’
(6a) is truly answered by specifying some individual(s) if and only if those individuals truly answer each conjunct i.e. it asks for those individuals who both ate and drank: [TA & TA] ´ TA. (6b) is truly answered by specifying those individuals who (at least) ate, and those individuals who (at least) drank. Since the list of individuals who truly answer (6a) can be different from those who truly answer (6b), these interrogatives are logically distinct. Disjunction (7) a. Wide scope: ‘Who ate or drank?’ b. Narrow scope: ‘Who ate or who drank?’
These seem to be equivalent to each other, and also equivalent to narrow scope conjunction (6b) above. Conditional These are not well formed in English: (8) *If who ate, then who drank?24
The problem: How to account for this fact?
215
216 Robert M. Harnish
Negation (9) a. Wide scope: ‘Who (did) not eat?’ b. Narrow scope: ‘*Not who ate?’ (‘*It is not the case that who ate?’)25
(9a) is truly answered by a specification of individual(s) who do not truly answer the question ‘Who ate?’. A[?x(Not Fx)] ´ Not[A?x(Fx)]. (9b) is not well formed in English. The problem: How to account for this fact? 6.1.2.3 Imperatives. Conjunction (10) a.
Wide scope: there seems to be no way to form these in English which makes them distinct from narrow scope b. Narrow scope: ‘Go shopping and go to the movies!’
A hearer H complies with the conjunction if and only if H complies with each conjunct [C & C] ´ C. A hearer H fails to comply with the conjunction if and only if H fails to comply with at least one conjunct [NC ⁄ NC] ´ NC. Disjunction (11) a.
Wide scope: there seems to be no way to form these in English which makes them distinct from narrow scope b. Narrow scope: ‘Go shopping or go to the movies!’
This is complied with if and only if one of the disjuncts is complied with: [C ⁄ C] ´ C. Conditional There do not seem to be any good examples of unmixed conditional imperative moods. However, ‘If A then B’ is logically equivalent to ‘NotA or B’ and to ‘Not(A and not B)’, the disjunctive and conjunctive forms are acceptable: (12) a. *If leave the room, then drink some milk! [!p] Æ [!q] b. Either don’t leave the room or drink some milk! [! Notp] ⁄ [!q] c. Don’t both leave the room and not drink any milk! ![Not(p & Notq)]
The problem: Explain this fact.
Frege on mood and force 217
Negation (13) a. Wide scope: Do not leave! b. Narrow scope: ‘*Not do leave!’ (‘*It is not the case that leave!’)26
A hearer H complies with (13a) if and only if H does not comply with ‘Leave!’. C[Notp] ´ Not[Cp]. (13b) is not well formed in English. The problem: How to account for the absence of narrow scope? 6.1.3 Mixed moods Logically, there are of course many more possible mixed moods than the 24 uniform moods we just briefly reviewed. We will not investigate them in detail here, but we will note that some mixed moods are actualized in English, for instance the following conditional seems syntactically well formed: (14) If Mary goes shopping then tell John I am going to the movies [/-p] Æ [!q]
This seems not to be satisfied if Mary goes shopping and the hearer H does not tell John I am going to the movies, but is it satisfied otherwise (on analogy with the indicative material conditional)? And what is its satisfaction value — truth, compliance? It is not clear with mixed moods what the answer should be. But in general, mixed mood compounding is much more restrictive than for uniform moods: (15) a. b. c. d.
*It’s a nice day and/or do you have the time. *It’s a nice day and/or pick me some flowers. *Do you have the time and/or pick me some flowers. *Pick me some flowers and/or do you have the time.
There are ways of improving these. For instance, replacing ‘and’ with ‘and by the way’ makes them almost acceptable: (16) a. b. c. d.
It’s a nice day and by the way do you have the time. It’s a nice day and by the way pick me some flowers. Do you have the time and by the way pick me some flowers. Pick me some flowers and by the way do you have the time.
Also, splitting them into two sentences makes them acceptable:27 e.
It’s a nice day. Do you have the time?
218 Robert M. Harnish
If the second clause comments on the first, the sentence is almost ok: f.
It’s a nice day, or did I say that already?
And there is some question whether connectives can take scope over mixed moods i.e. whether mixed moods can have narrow scope. We leave these matters for another time. All in all, then, mixed moods are highly restricted, and there is no obvious explanation for this in the extended Fregean theory — maybe it is just a fact about English syntax, but maybe it is common to other languages as well. 6.2 Embedded imperatives and interrogatives Earlier we outlined a somewhat controversial version of Frege’s theory of imperatives and interrogatives viewed as sentence types correlated with speech act types. Here we briefly investigate complex sentences with simple imperative and interrogative shaped sentences as a subordinated constituent. Frege considers constructions such as: (17) Embedded imperatives a. for someone to command (that S). b. for someone to ask (that S). c. for someone to forbid (that S). (18) Embedded interrogatives a. for someone to doubt (whether S). b. for someone not to know (what S). (19) a. b. c. d. e. f.
Frege claims that the dependent clauses in both (17) and (18) occur with indirect reference, but his argument is obscure: [1] A subordinate clause with ‘that’ after ‘command’, ‘ask’, ‘forbid’, would appear in direct speech as an imperative. [2] Such a clause has no reference but only a sense. [3] A command, a request, are indeed not thoughts, yet they stand on the same level as thoughts. [4] Hence in subordinate clauses depending
Frege on mood and force 219
upon ‘command’, ‘ask’, etc. words have their indirect reference. [5] The reference of such a clause is therefore not a truth value but a command, a request and so forth. (Frege, “On Sense and Reference”, 68)
Claim (1) seems true (with qualifications) but of marginal interest because of the unnaturalness of this form, and the consequently small class of cases that are well formed, e.g.: (20) a. b. c.
?I command that (you take out the garbage). ?I ask that (you take out the garbage). ?I forbid that (you take out the garbage).
In English it is much more natural to use a ‘for-to’ complementizer28 rather than ‘that’, e.g.: (21) a. I command (you to take out the garbage). b. I ask (you to take out the garbage). c. I forbid (you to take out the garbage).
But in this form the subordinate clause cannot be uttered independently at all, let alone as an imperative. Claim (2) is not clear as to whether Frege means the clause in its occurrence in the complex sentence or whether he means even when uttered by itself as e.g. an imperative. Suppose he means the latter — why should we think it true? Why could we not follow our earlier course and say that there are for imperatives ‘obedience-values’ and for interrogatives ‘true answerhood values’ defined in the obvious ways analogous to truth values for indicatives? As before, the sense of an imperative or interrogative sentence would be the ‘mode of presentation’ of one of these values. And, as the sense of an indicative expresses the conditions under which it is true, the sense of an imperative would express the conditions under which it is obeyed, and the sense of an interrogative the conditions under which it can be said to have been answered etc. If Frege means the former interpretation of (2) then he has simply begged the question since this is what is to be established. Claim (3) seems true and it says simply that the sense of these sentences involves a command, request etc. In claim (4) Frege has moved to his conclusion already, but it is not obvious how (1)–(3) entail (4). The following is a possible reconstruction of Frege’s train of thought here: We can change the truth-value of sentences like those in (17) and (18) by substituting different sentences in the dependent clause position. Truth depends on reference, but since these kinds of sentences do not normally have a reference, but they do have a sense, their sense must be their reference which means that
220 Robert M. Harnish
they must have an indirect occurrence. So viewed the argument is an inference to the best explanation. 6.2.1 Embedded interrogatives29 1.‘/-S(?S)’ contains an assertion: the manifestation of a judgment which is the recognition of the truth of the thought expressed.30 2.‘/-S(?S)’ expresses a sense: the thought that the conditions for the sentence naming the True are fulfilled. 3.‘/-S(?S)’ refers to a truth value: the True or the False. 6.2.2 Embedded Imperatives31 1.‘/-S(!S)’ contains an assertion: the manifestation of a judgment which is the recognition of the truth of the thought expressed. 2.‘/-S(!S)’ expresses a sense: the thought that the conditions for the sentence naming the True are fulfilled. 3.‘/-S(!S)’ refers to a truth value: the True or the False. As can be seen, embedded interrogatives and imperatives are just assertions relating speaker/cognizers to questions and commands.
7. Conclusion In this paper I proposed six conditions of adequacy on a theory of mood. I then reviewed Frege’s scattered remarks on the various moods, especially his remarks on nonindicative sentences, and I systematized these remarks in accordance with his general semantic theory. I then noted an asymmetry in his treatment of the reference of indicative versus nonindicative sentences. Finally, I explored the consequences of adding references to nonindicative sentences, first for simple sentences, then for coordinate and subordinate compound sentences, thereby making Frege’s semantics symmetrical and homogeneous. This theory, when combined with Frege’s general semantic theory, can meet the original conditions of adequacy for a theory of mood — though arguing for this in detail here would take too long. For now just notice that from what we have said so far: (1) nonindicatives, like indicatives, are compositional, nonindicatives are not true or false, and nonindicatives can occur unquestioned and uncommanded (Frege). (2) And all sentences with question or command force in virtue of mood share an element that accounts for that force, and sentences
Frege on mood and force 221
which differ in force due to mood have different elements (Davidson). (3) Furthermore, the theory, when conjoined with a theory of speech acts can allow us to infer the illocutionary force of its utterance from its mood.32 (McDowell) And finally (4), the theory does not require any special ambiguity, nonliterality or indirection (Grice). On the basis of this I suggest that we might have a potentially interesting and ‘Fregean’ account of the moods.33
Notes 1. We must distinguish these types of mood from the speech acts that they are typically used to perform: stating, questioning, directing (requesting, ordering, commanding), and exclaiming. 2. For an attempt to work out a theory of mood for English see Harnish (1994). 3. See Harnish (1994) for further discussion of these conditions as well as source citations. 4. Davidson adds the condition “(3) Finally, the theory should be semantically tractable.” (1979: 14–15). We assimilate this to compositionality. 5. Or something that becomes a thought upon contextual completion of indexical elements. We will understand thoughts in this way throughout. 6. See Figure 1. 7. See Figure 2. 8. See Figure 3. 9. A sense that becomes a thought upon completion by an appropriate sense. 10. See Figure 4. 11. We are not endorsing these arguments, only repeating them. 12. We must distinguish ‘not answered’ from ‘false answer’. 13. We cannot say ‘thought’, since thoughts are true or false, but interrogatives and imperatives are not true or false. 14. The idea of generalizing truth conditions and systematically collecting indicatives, interrogatives, and imperatives under a common analysis can be found in Katz (1977), Searle (1978), Bach and Harnish (1979). For a different approach see Davidson (1979). Dummett (1973/81: Ch. 10) discusses indicatives and imperatives collected under conditions of ‘correctness’. For a survey see Harnish (1994). 15. More exactly, it manifests the judgment that the thought expressed determines the True. 16. See Figure 5. 17. Y/N interrogatives can have true positive (‘Yes’) answers or true negative (‘No’) answers, which will be important later. 18. See Figure 6.
222 Robert M. Harnish
19. See Figure 7. 20. Frege (1923) called sentences/thoughts constructed out of these connectives “mathematical compound” sentences/thoughts. 21. Mood must be distinguished from performatives. According to Bach and Harnish (1979), Harnish (1988) and others, performatives are indicatives (and so true or false). And because they are indicatives, they are more free, as we will see, to enter into compounds than are the nonindicative moods (another argument that moods do not come from performatives). But notice that even the narrow scope performative here is awkward: ‘If I ask you did he leave the room, then I ask you did you drink some milk.’ 22. To be distinguished from ‘biased’ Y/N questions: ‘Doesn’t John Smoke?’ asks for confirmation that John smokes — it implies that the speaker already believes he does. 23. Again, mood must be distinguished from narrow scope performatives. The sentence ‘It is not the case that I ask you does John smoke (whether John smokes)?’ seems, on the face of it, to be clumsy but true or false. 24. The narrow scope performative is better: ‘If I ask you who ate, then I ask you who drank’. 25. The narrow scope performative is fine: ‘It is not the case that I ask you who ate.’ But, as I have been saying, performatives are not mood. 26. The narrow scope performative is fine: ‘It is not the case that I tell you to leave.’ But, as I have been saying, performatives are not mood. 27. So discourse concatenation is not always equal to sentential conjunction. 28. See P. Rosenbaum (1967) for an exposition of these notions. 29. See Figure 8. 30. ‘/-S(?S)’ is a sentence with an imbedded interrogative in it of either the Y/N or Wh sort. When we need to distinguish these two kinds of embedded interrogatives we will use ‘S(Y/N?)’ for Yes/No, and ‘S(Wh?S) for wh. 31. See Figure 9. 32. See Harnish (1994) for further discussion of this. 33. The present article is an edited and slightly revised version of a talk given to the Department of Logic, Czech Academy of Sciences, Prague, October 19, 1998. I want to thank the audience for provocative comments and especially to Petr Kolars for much good philosophy and managing a delightful stay. These topics are developed further in Harnish (1999) Chapters 1, 5, and 6.
Frege on mood and force 223
Figures We will represent each mood in a Frege diagram, where C = contains, E = expresses, R = refers to, D = determines. These diagrams are modelled on the one by Frege in his (1891) letter to Husserl.
Sentence C
R
E Sense: the thought that the conditions for the sentence naming the True are fulfilled D
Referent: True, False Force: /- Assertion: the manifestation of a judgment which is the recognition of the truth of the thought expressed
Figure 1.Indicatives Sentence C
R
E Sense: a ‘sense-concept’
Referent: ?
D Referent: True, False
Force: Y/N? Request: the manifestation of a demand that H assert whether or not the thought expressed names the True
Figure 2.Y/N-Interrogatives
224 Robert M. Harnish
Sentence C
R
E Sense: a ‘sense-concept’
Referent: ?
D Referent: a concept
Force: WH/? Request: the manifestation of a demand that H specify an argument that falls under the concept
Figure 3.Wh-Interrogatives Sentence C
R
E Sense: a ‘sense-concept’
Referent: ?
D Referent: a concept
Force: ! Request: the manifestation of a directive (“a command, request, etc.”) that H is to fall under the concept
Figure 4.Imperatives Sentence C
R
E Sense: a thought
Referent: TA, FA
D Referent: True, False
Force: Y/N? Request: the manifestation of a directive that H assert whether or not the thought expressed names the True
Figure 5.Extended Y/N-Interrogatives
Frege on mood and force 225
Sentence C
R
E Sense: a ‘sense-concept’
Referent: TA, FA
D Referent: a concept
Force: WH/? Request: the manifestation of a demand that H specify an argument that falls under the concept
Figure 6.Extended Wh-Interrogatives Sentence C
R
E Sense: a ‘sense-concept’ D
Referent: C, NC
Referent: a concept
Force: ! Request: manifestation of a directive that H fall under the concept
References Bach, Kent and Harnish, Robert M. 1979 Linguistic Communication and Speech Acts. Cambridge, Mass.: MIT Press. Carl, Wolfgang 1994 Frege’s Theory of Sense and Reference. Cambridge: Cambridge University Press. Davidson, Donald 1979 “Moods and Performances”. In Meaning and Use, A. Margalit (ed.), 9–20. Dordrecht: Reidel. Dummett, Michael 1973/81 Frege: Philosophy of Language. Cambridge, Mass.: Harvard University Press. 1st/2nd edition. Frege, Gottlob 1892 “On Sense and Reference”. In Translations From the Philosophical Writings of Gottlob Frege, P. Geach and M. Black (eds)(1960), 56–78. Oxford: Blackwell. Reprinted in Basic Topics in the Philosophy of Language, R. Harnish (ed.) (1994), Englewood Cliffs, N.J.: Prentice Hall. 1893 Grundgesetze der Arithmetik, translated by M. Furth (1964) as Basic Laws of Arithmetic, Berkeley: University of California Press. 1918a “The Thought”. Mind (1956). Reprinted in P. Strawson (ed)(1967), Philosophical Logic, Oxford. Reprinted with corrections in Basic Topics in the Philosophy of Language, Harnish, R. (ed)(1994), Englewood Cliffs, N.J.: Prentice Hall. 1918b “Negation”. In Translations From the Philosophical Writings of Gottlob Frege, P. Geach and M. Black (eds) (1960), 117–135. Oxford: Blackwell. 1923 “Compound Thoughts”, Mind 1963. Reprinted in E. Klemke (ed)(1968), Essays on Frege, Urbana, Ill.: University of Illinois Press. Grice, Paul H. 1975 “Logic and Conversation”. In Syntax and Semantics, R. Cole and J. Morgan (eds), 41–58. New York: Academic Press. Harnish, Robert M. 1988 “Performatives are Default-Reflexive Standardized-Indirect Acts”. Acta Linguistica Hungarica 38 (1–4): 83–106. 1994 “Mood, Meaning and Speech Acts”. In Foundations of Speech Act Theory. S. Tsohatzidis (ed.), 407–459. London: Routledge. 1999 Frege’s Philosophy of Language. Manuscript. University of Arizona. Katz, Jerrold 1977 Propositional Structure and Illocutionary Force. Cambridge, Mass.: Harvard University Press. McDowell, John 1976 “Truth Conditions, Bivalence, and Verificationism”. In Meaning and Truth. G. Evans and J. McDowell (eds), 42–66. Oxford: Oxford University Press. Rosenbaum, Peter 1967 The Grammar of English Predicate Complement Constructions. Cambridge, Mass.: MIT Press.
"har-r23">
228 Robert M. Harnish
Searle, John R. 1978 “Literal Meaning”. Erkenntnis 13 (1): 207–224, reprinted in Expression and Meaning, Searle, J. R. (1979), Cambridge: Cambridge University Press.
AUTHOR "Anna Wierzbicka"
TITLE "Leibnizian linguistics"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Leibnizian linguistics Anna Wierzbicka Australian National University, Canberra
1.
Introduction
According to Leibniz (1981 [1765]: 330), “languages are the best mirror of the human mind, and (…) a precise analysis of the significations of words would tell us more than anything else about the operations of the understanding”. Over the centuries, and especially in this century, many scholars have repeated Leibniz’s observation, often, however, changing his plural form (“languages”) to the singular (“language”). But in fact the profundity of Leibniz’s insight lay precisely in that plural. “Language” in the singular is not a good “mirror of the human mind”. Languages are diverse and each language reflects features of the culture associated with it. Hence attention to “language” (in the singular) as a guide to “human cognition” has often led scholars astray, resulting either in the absolutization of the conceptual distinctions and patterns suggested to them by their native language or in the abandonment of any attention to real languages in favor or sterile formalisms and quasi-mathematical abstractions. “Language” in the singular — English, French or Japanese — can be, as Sapir (1949) said, a guide to social reality, a guide to culture. It can also be a guide to the human psyche as culturally constituted, that is, as shaped not only by the innate and universal features of “human nature”, but also by the particular features of historically transmitted “local” cultures (cf. Shweder 1990). It is only “languages” in the plural which allow us to see and appreciate both the diversity of cultures and what Franz Boas and others have called “the psychic unity of mankind”. To study the universal aspects of human cognition we need to pay attention to linguistic universals: it is the shared rather than the idiosyncratic features of languages which provide a guide to the workings of the generic human mind.
230 Anna Wierzbicka
The fundamental importance of Leibniz’s thought to contemporary linguistics lies in the way he was able to combine a profound interest in different languages with a search for the universals of language and thought. As Umberto Eco (1995: 270) comments: Leibniz had always been fascinated by the richness and plurality of natural languages, devoting his time to the study of their lineages and the connections between them. He had concluded that it was not possible to identify (much less revive) an alleged Adamic language, and came to celebrate that very confusio linguarum that others were striving to eliminate (see Gensini 1990, 1991).
This celebration of the diversity of languages was linked with an appreciation of their value in providing different perspectives on the world. As George Steiner (1992: 78) notes in his classic work After Babel: In 1697, in his tract on the amelioration and correction of German, Leibniz put forward the all-important suggestion that language is not the vehicle of thought but its determining medium. Thought is language internalized, and we think and feel as our particular language impels us to do. But tongues differ as profoundly as do nations. They too are monads, ‘perpetual living mirrors of the universe’, each of which reflects or, as we would now put it, structures experience according to its own particular sight-lines and habits of cognition.
At the same time, as was also noted by Steiner (1992: 78), Leibniz was profoundly interested in the possibilities of a universal semantic system, immediately accessible to all people, and based on a set of pan-human conceptual primitives (cf. also Eco 1995: 270).
It is this combination of an interest in the diversity of languages with a belief in the underlying “psychic unity of humankind”, grounded in a system of panhuman conceptual primitives, which makes Leibniz’s thought so fascinating.
2. “Leibnizian linguistics” vs. “Cartesian linguistics” “Leibnizian linguistics” can be seen as an alternative to Chomsky’s (1966) “Cartesian linguistics” — to the generative grammar advanced by Chomsky as the rightful heir to the theories of language and thought put forward by the great seventeenth century French thinkers, notably Descartes and Arnauld. After decades of domination, the “generative grammar” paradigm is in decline: the Chomskyan revolution, so eagerly embraced by many linguists and
Leibnizian linguistics
hailed by many scholars in adjacent disciplines in the sixties and seventies as a break-through in the study of language and mind, has clearly not fulfilled its promise. Arguing for an approach which would enable us “to account for the nature and possibility of relations between languages as they actually exist and differ”, Steiner (1992: 113) declares Chomskyan linguistics to be irrelevant to those genuinely interested in languages, in translation, in problems of cross-cultural understanding, or in human cognition, and notes: Noam Chomsky’s Theory of Government and Binding, his The Science of Language, show a sharpening distance from the universalist innocence of sovereign claims of earlier theories. The promises initially offered to non-technical readers of literature, of philosophy, of translation, have been relinquished.
Steiner (1992: 112) advocates an approach “whose bias of interest focuses on languages rather than language; whose evidence will derive from semantics (with all the implicit stress on meaning) rather than from ‘pure syntax’; and which will begin with words, difficult as these are to define, rather than with imaginary strings of ‘pro-verbs’ of which there can never be any direct presentation.” Sharpening his attack on Chomskyan linguistics in the preface to the second edition of his classic, Steiner (1992: xv) states: The dissent from transformational generative claims throughout After Babel bear on the failure of such grammars to produce substantive examples of ‘universals’ in natural languages, and on the cardinal irrelevance of the Chomskyan project to poetics and hermeneutics. Today, generative grammars have retreated into almost total formality, into a degree of analytic and meta-mathematical algorithmic abstraction so great as to have hardly any bearing on the matter of actual speech worlds and of the creative differences between them.
In his influential recent book The rediscovery of the mind John Searle (1994: 242) also challenges the basic assumptions of Chomskyan linguistics, pointing out that “the alleged rules of universal grammar” (as conceived by Chomsky and his followers) are not accessible to consciousness and consequently are not verifiable. Similarly, the psychologist and philosopher Rom Harré and the psychologist and neuro-surgeon Grant Gillett (1994:77–78) argue (against Chomsky and Chomskyans) that in the exploration of human cognition priority must be given to intuitively intelligible natural languages which are accessible to the consciousness: Another important consequence of the second cognitive revolution is the priority that must be given to ordinary languages in defining what are the
231
232 Anna Wierzbicka
phenomena for a scientific psychology. We will endeavour as far as possible to present and understand cognition in terms of the ordinary languages through which we think, rather than looking for abstract representations of them. That is radical because it resists the idea that a new, formal calculus must be devised to represent thought. Such calculi lie at the heart of the artificial intelligence project, the methodological principles of Chomsky and the transformational grammarians, and the assumption of formalists of all kinds.
Similar views have also been expressed by one of the main authors of the “first cognitive revolution”, Jerome Bruner (1990). I also have long argued that an alternative to Chomskyan (and what Chomsky calls “Cartesian” linguistics) can be found in “Leibnizian linguistics”. This alternative “Leibnizian” approach has produced a substantial body of research (see the references listed at the end of this paper).1 In what follows, I will discuss those of Leibniz’s ideas on language and thought which have proved particularly fruitful in this work and which can be seen as a basis for the kind of linguistics that Steiner, Bruner and others have been calling for: a linguistics grounded in language universals, and centered on meaning and translation, a linguistics which begins with the study of words, difficult as these may be to define, and which is prepared to serve, and guide, lexicography; a linguistics which combines an interest in universal grammar with in-depth empirical study of lexical and grammatical systems of real languages, in all their specificity and richness.
3. The “alphabet of human thoughts” From a linguistic point of view, the most important and fruitful of Leibniz’s ideas is his view that there must be a set of innate and universal human concepts — conceptual primes — which can provide a basis for the analysis of all human thoughts, however complex. These innate conceptual primes are “conceived by themselves” and are so clear to us that no explanation can make them any clearer. Leibniz called this innate set of conceptual primes “the alphabet of human thoughts”: The alphabet of human thoughts is the catalogue of those concepts which can be understood by themselves, and from whose combinations our other ideas arise.2
The reason why this idea is so crucial for linguistics is that a set of shared concepts provides a common measure, without which languages cannot be meaningfully described and compared. The need for such a common measure
Leibnizian linguistics 233
has often been stressed, in particular, with respect to linguistic typology, but in fact it applies to all linguistic theory. To quote one eminent typologist, William Croft (1990: 11): The characteristic feature of linguistic typology … is cross-linguistic comparison. The fundamental prerequisite for cross-linguistic comparison is comparability, that is the ability to identify the ‘same’ grammatical phenomena across languages … This is a fundamental issue in all linguistic theory, in fact. Nevertheless this problem has commanded remarkably little attention relative to its importance.
Croft’s comment that in the past “this problem has commanded remarkably little attention relative to its importance” echoes Leibniz’s comments on the importance of fixed and firm basic concepts in linguistic analysis and foreign language teaching. This consideration [how some concepts arise from others] allows us to avoid vicious circles, which are constantly taken recourse to, and the mind can fasten to some firm and fixed concepts which can be decided upon. How very important this is, few people understand, because few people reflect how very important it might be to determine what the first elements in all things are. (Couturat (ed) 1903: 160)3
The task of discovering the ultimate simples (the “building blocks” of human thoughts) was seen by Leibniz as difficult and time-consuming, but by no means impossible. It had to be pursued by trial and error, that is, by sustained, systematic attempts to define as many words as possible, so that one could identify on an empirical basis those concepts which serve as the building blocks from which all others are constructed. The primary terms, the indefinables, cannot be easily recognized by us except in the way that the prime numbers are: we can only recognize them as such if we try to divide them [by the smaller ones]. (Couturat (ed.) 1903: 187)4
The basic guideline in this search, then, was the requirement that the set of simple concepts should contain only those which are truly necessary for defining all the others. Whatever can be defined is conceptually complex and should be defined; whatever cannot be defined (without circularity and without going from simple to complex and from clear to obscure) should not be defined. “Reducenda omnia alia ad ea qua sunt absolute necessaria ad sententias animi exprimendas” (Couturat 1903: 281), ‘All other [expressions] should be reduced to those which are absolutely necessary for expressing the thoughts in our minds.’ If we do not discover this alphabet of necessary concepts which cannot
234 Anna Wierzbicka
be made clearer by any definitions (“quae nullis definitionibus clariores reddere possumus”, Couturat 1903: 435), we can never successfully elucidate meanings conveyed in language, because without this basic tool we will only be able to translate unknowns into other unknowns. Leibniz illustrates the need to reduce all complex meanings to self-explanatory components with the following comparison. “Suppose I make you a gift of a large sum of money saying that you can collect it from Titius; Titius sends you to Caius; and Caius, to Maevius; if you continue to be sent like this from one person to another you will never receive anything.” (Couturat 1903: 430) Only by decomposing complex meanings into self-explanatory components can any true understanding ever be achieved. If nothing could be understood in itself nothing at all could ever be understood. Because what can only be understood via something else can be understood only to the extent to which that other thing can be understood, and so on; accordingly, we can say that we have understood something only when we have broken it down into parts which can be understood in themselves. (Couturat (ed.) 1903: 430)5
It was, above all, this emphasis on the need for the ultimate primes to be anchored in intuitively intelligible natural language which distinguished Leibniz’s program from other seventeenth century efforts to construct a “philosophical language”, such as that of John Wilkins.
4. Leibniz and Wilkins: Two contending traditions Wilkins, like Leibniz, sought to link the search for a universal philosophical language with a semantic analysis of natural language. His book, An essay towards a real character and a philosophical language, was published in London in 1668, and Leibniz was certainly aware of it. Wilkins’ philosophical language, however, differed fundamentally from Leibniz’s, in that it was artificial. Instead of indefinable words from natural language, which formed the basis of Leibniz’s work, Wilkins used arbitrary symbols, which he called “transcendental particles”. According to Wilkins, the transcendental particles should correspond to real semantic relations — and in this sense they were not arbitrary. But the choice of transcendental particles was partly a matter of mere convenience. Thus, for instance, the number of transcendental particles was, to some extent, the decision of the investigator rather than an absolute and objective fact:
Leibnizian linguistics 235
“What may be the most convenient number of such Transcendental Particles is not easy to determine” (Wilkins 1668: 320). Leibniz could not accept the idea of “the most convenient number” of elementary units, since he rejected all arbitrariness in his work on the “alphabet of human thoughts”. (Indeed, he criticized Raymond Lulle, a thirteenth-century logician whom he considered one of his forerunners, for classifying concepts into schemata with equal numbers of elements (Gerhardt (ed.) 1960–61 Vol. 4: 63).) Leibniz had a great deal of respect for Wilkins’ work, but he saw it as quite distinct from his own: I considered this matter before the appearance of Mr Wilkins’ book, when I was a young man of nineteen, in my little book de Arte combinatoria, and it is my opinion that the genuinely real and philosophical characters should correspond to the analysis of thoughts. (Couturat (ed.) 1901 [1697]: 54)6
Unlike Wilkins, Leibniz conceived of his projected philosophical language as being, above all, a tool for reasoning: For their [i.e. Wilkins and his followers’] language or script only allows a convenient form of communication to be set up between people divided by language; but the true Characteristica Realis, as I conceive of it, ought to be considered one of the most apt tools of the human mind, as it would have unrivalled power with respect to the development of ideas, storing those ideas, and distinguishing between them. (Gerhardt (ed.) 1960–1 Vol. 7: 7)
Leibniz felt that the strength of his philosophical language would be derived from the fact that at the same time as it modeled the semantic relationships among words, it would also model the logical relationships. He believed that even elementary reasoning, such as that required to demonstrate that ‘Solomon is David’s son’ given that ‘David is Solomon’s father’, is not logically justified until a semantic analysis reveals how one proposition follows from the other: The aim of our Characteristica is to use such words that all inferences which can be drawn could be seen from the words or symbols themselves … . Inferences of this kind cannot be shown on the basis of Latin words alone, unless they are analyzed into other equivalent ones. (Courat (ed.) 1901: 75n)7
In Leibniz’s view, such a relationship cannot be revealed equally well by an analysis using “transcendental particles”, since artificial symbols are not understandable by themselves, and hence the relationships shown by combinations of such symbols are not immediately comprehensible. Only the indefinables can be understood by themselves.
236 Anna Wierzbicka
It seems that Leibniz kept devising semantic explications almost all his life. As he writes in an unfinished letter, “… et je m’ay fait quantité de definitions [‘I have made myself lots of definitions’]”. Leibniz’s nineteenth-century editor, Gerhardt wrote: There are a great many manuscripts among Leibniz’s papers which deal with elementary concepts and definitions … . Convinced that he would never be able to complete the great task alone, how often he yearned for help! Finally, in the second last decade of his life, he had his secretary Johann Friedrich Hodann draw together a collection of his definitions, which is still in existence … . (Gerhardt (ed.) 1960–61, Vol. 7: 30)8
Hodann’s collection of definitions forms the text of the Table of definitions (Couturat (ed.) 1903) — a pioneering work in systematic lexicography.9 Leibniz’s dream of finding collaborators never came true, and his explications remained an isolated phenomenon for almost three centuries. Initially with the rise of semantics as a branch of linguistics in the twentieth century the Leibnizian approach was overshadowed by the Wilkinsian. The Danish linguist Louis Hjelmslev, one of the pioneers in this century of Wilkinsian approach to linguistic semantics, wrote: There is a gulf separating [structural semantics] from the old attempts to establish a universal semantics or ars magna, which culminated in the scientia generalis or characteristica generalis of G. W. Leibniz, but which relied on the ars generalis of Raymond Lulle both for its methodological principles and for the essence of its ideas … . (Hjelmslev 1957: 98)10
Although Wilkins is not widely identified as the source, the idea of modeling real semantic relations by means of artificial elementary units has gained wide acceptance in twentieth century linguistics. Hjelmslev’s (1953) “figurae of the content plane”, Katz and Fodor’s (1963) “semantic markers”, or the “atomic predicates” of the generative semanticists (e.g., Lakoff 1970; McCawley 1973) were all very much in the spirit of Wilkins’ work. Although the linguists who used such symbols usually adopted for the purpose words taken from natural language, they always stressed emphatically that the meaning of these symbols must not be identified with their meaning in ordinary usage. Thus in spirit they were Wilkinsians rather than Leibnizians. In the twentieth century, Leibniz’s ideas were revived (though not under that name) by another Danish linguist, Holger Steen Sørensen, who argued (opposing Hjelmslev) that the elementary semantic units should not be sought in artificial “figurae of the content plane” unrelated to the “expression plane”,
Leibnizian linguistics 237
but in the indefinable words of natural language (Sørensen 1958). The indefinables sought by Sørensen, however, were not identical with Leibniz’s alphabet of human thought, because they were not necessarily universal. Sørensen did not claim that the sets of indefinables in different natural languages were in a one-to-one relationship. Proposing an inquiry into the “indefinables of the English language” (Sørensen 1958:42–48), he did not postulate investigating the language of human thought. To that extent he was not a true Leibnizian. A fully-fledged Leibnizism first surfaced in linguistics in the writings of the Polish linguist Andrzej Bogusławski, who in the mid-sixties initiated semantic investigations based on the idea of a non-arbitrary, universal set of elementary units of thought, identified with the meaning of the indefinables of natural language (Bogusławski 1966, 1970). This became the starting point for the socalled “NSM” movement in contemporary linguistics (see Wierzbicka 1996; Goddard 1997a). The acronym “NSM” stands for “Natural Semantic Metalanguage”, but since the idea comes from Leibniz, this movement could equally well be call “Leibnizian linguistics”. The “Natural Semantic Metalanguage” (to be discussed more fully in the next section) is a name given to the language-like irreducible core of all languages, which has been established on the basis of extensive cross-linguistic investigations, and which is based on empirically established lexico-grammatical universals. In the past, Leibniz’s semantic claims have tended to be brushed aside, not because anybody was able to present any coherent arguments or counterevidence against them, but because Leibniz had failed to advance even a tentative list of those fundamental conceptual building blocks whose existence he was postulating. The following comment by a modern commentator on Leibniz is characteristic in this respect (Martin 1964: 26): When he [Leibniz] says that there are only a few fundamental concepts, he is expressing himself in the most cautious terms. What are we to understand by this statement? … The obvious question as to the number of fundamental concepts must presumably be left unanswered.
Having pointed out the difficulties which the program of searching for the ultimate concepts involves, the same commentator adds (p.25): In these circumstances it is understandable that Leibniz should consistently avoid the obvious question as to the number and type of fundamental concepts. The approach would be more convincing if one could at least gain some clue as to what the table of fundamental concepts might look like.
238 Anna Wierzbicka
These criticisms are, if not justified, at least understandable. The whole program of paring down the redundancy contained in every natural language till the indispensable kernel is reached may indeed look implausible, perhaps even utopian, until some tentative table of fundamental concepts is actually drawn up. But the fact that Leibniz did not draw up any such tables himself does not mean that it cannot be done. As Leibniz himself foresaw, empirical crosslinguistic investigations can achieve what philosophical speculations alone could not.
5. Cross-linguistic investigations as a key to universals of language and thought This was another of Leibniz’s brilliant insights, unique at the time and not sufficiently appreciated before the last quarter of the twentieth century: the vital role that empirical cross-linguistic investigations have to play in the discovery of the shared core of all languages, both with respect to universal concepts (“embodied” in universal words) and to universal grammar (“embodied” in universal grammatical constructions). In his New Essays on Human Understanding Leibniz (1981 [1765]: 302) wrote: It is true that someone who wanted to write a universal grammar would be well advised to move on from the essence of languages to their existence and to compare the grammars of various languages; just as an author seeking to write a universal jurisprudence, derived from reason, would do well to bring in parallels from the laws and customs of the nations. This would be useful not only in a practical way, but also theoretically, prompting the author himself to think of various considerations which would otherwise have escaped his notice.
And he elaborated, predicting the birth of descriptive and typological linguistics and its role in the study of human cognition: Eventually, every language in the universe will be recorded, and contained in dictionaries and grammars; and comparisons will be made amongst them. This will be extremely useful for the knowledge of things, since their properties are often reflected in their names (…), as well as for the knowledge of our mind and of the marvelous variety of its operations. (Leibniz 1981 [1765]: 336–37)
The Leibnizian program of “recording every language in the universe” has not yet been fully carried out, but very substantial efforts in this direction have been undertaken in recent decades, resulting in hundreds of professionally
Leibnizian linguistics 239
made grammars and dictionaries, and in wide-ranging typological investigations. One of the fruits of these investigations is the tentative identification of “universal words” — that is, words which apparently can be found in all languages of the world. Since these “universal words” (in some cases, morphemes or phrasemes) provide semantic primes in terms of which all complex meanings can be expressed, the empirically established set of such words corresponds to what Leibniz postulated as the “alphabet of human thoughts”. The full set of these universal words and concepts, which has emerged from empirical crosslinguistic work in the last three decades can be presented in the form of the following table. (For detailed discussion and justification see Wierzbicka 1996; Goddard 1997a., 1998):11 Table of Universal Concepts Substantives:
I, you, someone/person, something/thing, people, body
Determiners:
this, the same, other
Quantifiers:
one, two, some, all, many/much
Attributes:
good, bad, big, small
Mental predicates:
think, know, want, feel, see, hear
Speech:
say, word, true
Actions, events, movement:
do. happen, move
Existence, and possession:
there is, have
Life and death:
live, die
Logical concepts:
not, maybe, can, because, if
Time:
when/time, now, after, before, a long time, a short time, for some time
The empirical investigations from which the above table of universal human concepts has emerged have been extended over a considerable number of languages. The first large-scale report on these investigations was published in 1994 in the form of a collective volume entitled Semantic and Lexical Universals (Goddard and Wierzbicka 1994). In this collection of studies, conceptual
240 Anna Wierzbicka
primitives posited initially on the basis of a mere handful of languages were subjected to systematic study across a wide range of languages from different families and different continents. The languages investigated in this volume included: Ewe (of the Niger-Congo family in West Africa), Mandarin Chinese, Thai, Japanese, the Australian languages Yankunytjatjara, Arrernte (Aranda), and Kayardild, three Misumalpan languages of Nicaragua, the Austronesian languages Acehnese (of Indonesia), Longgu (Solomon Islands), Samoan, and Mangap-Mbula (New Guinea), the Papuan language Kalam, and — the only European language beside English — French. On the whole, (except for one or two grey areas requiring further investigation) these studies strongly supported the hypothesized set of primitives. In most cases, words (or bound morphemes) for the proposed primitives (e.g. ‘I’ and ‘you’, ‘someone’ and ‘something’, ‘where’ and ‘when’, ‘big’ and ‘small’, ‘good’ and ‘bad’, or ‘do’ and ‘happen’) could be readily identified. Another collective volume investigating the use of the postulated conceptual primitives over a number of unrelated languages was published as a special issue of the journal Language Sciences (Goddard 1997b), and yet another, entitled Meaning and Universal Grammar is currently in press (Goddard and Wierzbicka in press). The set of concepts postulated as universal, will of course be differently realized in different languages. Within any one language, every element occupies a particular place in a unique network of relationships. When we compare two or more languages we cannot expect to find identical networks of relationships. We can, nonetheless, expect to find corresponding sets of indefinables. For example, the English words big and small correspond in meaning to the Russian words bol’šoj and malen’kij, even though in English, small has also a special relationship with little, and even though in Russian, malen’kij — formally a dimunitive — has a special relationship with dimunitive adjectives such as belen’kij (from belyj ‘white.dim’) or kruglen’kij (from kruglyj ‘round.dim’), where -enk is a dimunitive suffix and -yj/ij the inflection. But whatever the differences in “resonance” between small and malen’kij are, these differences cannot be shown through definitions; and so, from a definitional point of view, these words constitute a “perfect” match (in the systems of English and Russian indefinables, they occupy the same slot). Similarly, regardless of any differences in “resonance” (any use), the Japanese words ookii and tiisai constitute a perfect semantic match for big and small, and the Japanese words ii and warui, for good and bad (cf. Onishi 1994).
Leibnizian linguistics 241
It is only the postulated isomorphism of exponents of conceptual primitives which allows us to compare different semantic systems at all. For any comparison requires a tertium comparationis, a common measure. The hypothesized set of universal semantic primitives offers us such a common measure and thus makes it possible to study both the semantic differences between languages and the unity of human thoughts which underlies them.
6. Diversity of languages and “the psychic unity of humankind” The fact that it has now been possible to establish a list of concepts which appear to be lexically encoded in all languages vindicates Leibniz’s life-long belief in the existence of “a universal semantic system, immediately accessible to all people, and based in a set of pan-human conceptual primitives” (Steiner 1992: 781). It also vindicates Leibniz’s conviction that this universal semantic system is reflected in languages (“the best mirror of the mind”). In modern linguistics — as in the philosophy of language — two radically opposed points of view concerning language universals have often been asserted. As Steiner (1992: 76) put it: The one declares that the underlying structure of language is universal and common to all men. (…) Translation is realizable precisely because those deep-seated universals (…) can be located and recognized as operative in every human idiom. The contrary view can be termed ‘monadist’. It holds that (…) the actual workings of human speech (…) are so diverse (…) that universalist models are at best irrelevant and at worst misleading.
The extreme universalists have often attacked the view that different languages reflect different conceptual systems, identifying it with “the idea that thought is the same thing as language”, which they reject as an “absurdity” (Pinker 1994: 57). On the other hand, extreme relativists have often attacked the view that beneath all the differences there may be some conceptual universals, reflected in universals of language. In particular, the distinguished British anthropologist Rodney Needham criticized Leibniz for recommending comparative study of languages as “a path to discovering the universal inner essence of men” and “the workings of the human mind” (Leibniz 1981 [1765]: 326). Discussing Leibniz’s “grand proposal” that a search for a universal lexicon and universal grammar should be linked with a thorough examination and comparison of languages, Needham (1972: 220) commented critically:
242 Anna Wierzbicka
This bold suggestion … was based on the tacit premise that the human mind was everywhere the same … . Methodologically, Leibniz was thus proposing a comparative analysis of the kind that Lévy-Bruhl was to put into effect almost exactly two centuries later, and even in terms that find ready agreement today; but it is not premises, not the type of research that he recommended, that have since been called into renewed question. Underlying his proposal was the conviction that human nature was uniform and fixed, and it is precisely this idea that more recent conceptual analyses have made difficult to accept.
Thus, scholars of Pinker’s orientation dismiss the study of differences between languages as a possible source of insight into social cognition because they regard such differences as superficial. On the other hand, scholars of Needham’s orientation dismiss the search for linguistic universals as a possible guide to conceptual universals because in their view there can’t be any conceptual universals: to admit that such universals could exist would mean to accept the possibility that some aspects of “human nature” and human cognition may be constant (“uniform and fixed”). What both sides seem to fail to consider is the possibility that languages, and the ways of thinking reflected in them, exhibit both profound differences and profound similarities; that the study of diversity can lead to the discovery of universals; that some hypotheses about universals are indispensable for the study of the diversity; and that all hypotheses about conceptual universals have to be checked and revised in response to empirical findings emerging from systematic cross-linguistic investigations. In fact, there is no conflict between an interest in linguistic and conceptual universals on the one hand and an interest in the diversity of language-andculture systems on the other. On the contrary, to achieve their purpose, these two interests must go hand in hand. If the meanings of all words were culture-specific, then cultural differences could not be explored at all. The “hypothesis of linguistic relativity” makes sense only if it is combined with a well thought-out “hypothesis of linguistic universality”: only well-established linguistic universals can provide a valid basis for comparing conceptual systems entrenched in different languages and for elucidating the meanings which are encoded in some languages (or language) but not in others. Leibniz’s conception of an “alphabet of human thoughts” provides a way out of an otherwise impossible dilemma, for it allows us to combine radical universalism with thorough-going relativism. “Leibnizian linguistics” accepts the uniqueness of all language-and-culture systems, but posits a set of shared
Leibnizian linguistics 243
concepts in terms of which differences between these systems can be assessed and understood; and it can interpret the most idiosyncratic semantic structures as culture-specific configurations of universal semantic primitives — that is, of innate human concepts.
7. Si nihil per se concipitur… Discussions of semantic primes are often vague and purely theoretical. The value of different possible ingredients for the pudding is pondered and debated, without an edible product ever being placed on the table. Yet the proof of the pudding is in the eating. The value of the Leibnizian principles of linguistic analysis can be best appreciated when one attempts to apply these principles in actual descriptive work — in large scale lexicographic projects, in grammatical typology, in comparison of cultural key words, or in the comparative study of ethnobiology, ethno-psychology, or ethno-philosophy. Only then can one see what conception of primes can be fruitful and relevant. Leibniz’s idea that the primes must be self-explanatory is of fundamental importance in this respect, and it is far from being generally appreciated. To take an example; two prominent researchers into child language and the authors of a very valuable study on the acquisition of meaning, Lucia French and Katherine Nelson (1985: 38), start their discussion of the concept ‘if ’ by saying: “it is difficult to provide a precise definition of the word if”. Then, after some discussion, they conclude: “the fundamental meaning of if, in both logic and ordinary language, is one of implication”. Two common assumptions are reflected in these statements. First, that it is possible to define all words — including if — and second, that if a word seems difficult to define, one had better reach for a scientific word of Latin origin (such as implication) to ‘define’ it with. These assumptions are not only false; they jointly constitute a stumbling block for semantic analysis. One cannot define all words, because the very idea of ‘defining’ implies that there is not only something to be defined (a definiendum) but also something to define it with (a set of definables). Another researcher into child language, Eve Clark (1971: 273), tried to analyze the simple concepts before and after as follows: before: “At a time preceding the time at which.” after: “At a time following the time at which.”
244 Anna Wierzbicka
But as noted by Tien (1997), the words precede and follow are hardly clearer or simpler than the notions that they are intended to define. A definition which attempts to explain the simple and intuitively understandable word if via the complex and technical word implication cannot advance our understanding. As Leibniz put it, “if nothing could be understood in itself, nothing at all could ever be understood” (si nihil per se concipitur, nihil omnino concipietur). But a word can only be truly understood in the context of a sentence. For example, when we say that if is “self-explanatory”, this is intended as an abbreviation. If is “self-explanatory” not as an isolated word but as a word used in the context of sentences (e.g. “If you do this something bad can happen.”). This means that the notion of an “alphabet of human thoughts” is fruitful only if it is understood as a lexical core of a universal mini-language — the “language of human thoughts”, the universal “lingua mentalis” (cf. Wierzbicka 1980). This is why the project of identifying the set of universal human concepts (the shared lexical core of all languages) is inseparable from the project of discovering the universal grammar (the shared grammatical core of all languages). For example, the primitive status of the concept IF hinges not only on the status of IF as a lexical universal but also on the universality of bi-clausal IF sentences. Thus, the Leibnizian principle of “intelligibility” of the primes means that the “common measure” provided by the primes must be understood in the sense of a universal mini-language, accessible to intuition via a set of language-specific mini-languages. For example, the universal prime expressible in English as if and in Russian as esli is accessible to English speakers’ intuition via sentences like If you do this, something bad can happen.
and to Russian speakers’ intuition via matching Russian sentences like Esli ty ˙eto sdelaeš ˇcto-to ploxoe slucˇit’sja. if you-sg this-acc do-2sg something bad happen-3sg
The bi-clausal construction illustrated by these two sentences can be regarded as part of universal grammar. It is that language-like character of the postulated set of primes which makes semantic formulae based on primes verifiable by substitution. The principle of substitutability as a criterion of empirical validity in semantics was proposed by Leibniz (not the celebrated substitutability “salva
Leibnizian linguistics 245
veritate”, but the linguistically much more relevant substitutability “salvo sensu”). The semantic representation should be in principle substitutable for the utterance or part of the utterance whose meaning it is supposed to represent, and this applies not only to nouns, adjectives and verbs, but even to conjunctions, particles and grammatical inflections: An abstract account of the kind we have just given here is not enough for a thorough explication of the particles: what we need is a paraphrase which can be substituted for the particle, just as a definition can be put in place of the defined expression. If we devote ourselves to seeking and formulating these ‘substitutable paraphrases’ for all particles in so far as they admit of them, then we shall have dealt with the significations of particles. (Leibniz 1981 [1765]: 332)
Similarly, elsewhere: All inflections and particles should be explicated and reduced to the simplest possible explications which could always be substituted [for what is being explicated] with the sense preserved, in the same place. (Couturat (ed.) 1903: 244)12
Obviously, this kind of substitutability has its limitations. If a tiny particle or inflection is replaced with a long formula in which all the assumptions and all the relations that it stands for are made explicit, the resulting paraphrase will generally become so long that from a pragmatic point of view it will be anything but equivalent to the original. For this reason, semantic substitutability does not entail pragmatic substitutability. Furthermore, semantic formulae are often not only long, but also awkward and unidiomatic. Who, for example, would ever formulate a request along the following lines: I want you to do something. I say it because I want you to do it. I know that you don’t have to do it.
On the other hand, if someone said: “imp you say something” (using the artificial symbol imp for “imperative”) they would not be understood at all. There is, therefore, a difference in kind, not a difference in degree, between the pragmatic non-substitutability of natural-language explications and the fundamental non-substitutability of “features”, “markers”, “abstract predicates” or logical formulae. In order to be even semi-substitutable (and consequently, verifiable) semantic representations have to be formulated in a metalanguage based on natural language.
246 Anna Wierzbicka
As Goddard (1998) writes: For many linguists and logicians working in other frameworks, nothing is more mysterious and intangible than meaning. But adopting reductive paraphrase as a way of grasping and stating meanings makes meanings concrete, tangible. Above all, it makes statements about meanings testable — because explications couched in natural language can be directly substituted in place of the expressions they are intended to represent, and so can be submitted to the test of substitution salvo sensu. It all depends on the fact that explications in ordinary language are intelligible to native speakers.
8. Conclusion It is an essential feature of “Leibnizian linguistics” that it proceeds by trial and error. Leibniz’s program has often been criticized or even dismissed because this trial and error aspect of it was not properly understood. For example, Rodney Needham (1972: 222) branded Leibniz’s program as “self-contradictory”, arguing as follows: The aim … of an ideal or universal language is to construct verbal concepts that … must be identically comprehensible to speakers of different natural languages … Yet … this result cannot be attained except on condition that it be already attained: the ideal language is to be the product of successful comparison and accommodation of differing cultural concepts, but these cannot in the end be objectively and precisely compared except by means of the ideal language itself.
The “Natural Semantic Metalanguage” used currently in cross-linguistic investigations is of course not the same thing as Leibniz’s ideal and universal language. But it is an embodiment of the same basic idea, and Needham’s comments on the apparently “self-contradictory” nature of Leibniz’s project apply also to the “NSM” project. For the set of universal and presumably innate concepts postulated in current NSM work and enumerated in Table 1 could only be arrived at on the basis of intensive explorations of many diverse languages; and yet such explorations could only proceed in the first place on the basis of a hypothetical set of universal concepts postulated prior to any wideranging cross-linguistic investigations. In fact, however, Needham’s negative conclusions are not justified, for he mistakes trial and error for an insuperable internal contradiction. The “NSM”
Leibnizian linguistics 247
project has proceeded all along according to the Leibnizian methodology of trial and error: first, a minimal set of hypothetical universal concepts was posited on the basis of speculation and tradition (going back to Aristotle’s Categories (1963)). Then it was continually expanded and modified as the empirical basis of the study broadened to include more and more languages, more and more conceptual domains, and more and more aspects of language structure and language use. As a result of this process of continual revision, the number of postulated universal concepts has increased from fourteen (see Wierzbicka 1972) to sixty, and three elements from the original set (IMAGINE, WORLD, and DON’T WANT) have been definitely removed from the list. At the same time, the lexical focus of the search (that is, the focus on the “alphabet of human thoughts” has been replaced by a more broadly-based search for a universal “language of human thoughts”, embracing both a universal lexicon and a universal grammar. Since the universal grammar is conceived in NSM work as a universal combinatorics of the lexically embodied universal concepts, the work on universal grammar constitutes a natural continuation of the work on the universal lexicon (see Goddard 1997a,b; Goddard and Wierzbicka in press). Although the construction of the Natural Semantic Metalanguage has taken more than thirty years and is still incomplete, there is already a very substantial body of descriptive empirical work carried out within the NSM framework — into aspects of lexical, grammatical and illocutionary semantics, cultural pragmatics, and now into universal grammar. For reasons of space this work cannot be reviewed here, but representative works include: Ameka (1990a, 1990b); Chappell (1986, 1991); Goddard (1991, 1994, 1995, 1996); Harkins (1996); Hasada (1996, 1997); Kornacki (1995); Onishi (1997); Peeters (1993); Peeters and Eiszele (1993), Rieschild (1996), Travis (1997, 1998), Wierzbicka (1988, 1992, 1996, 1997, 1999); Wilkins (1986). The languages involved include English, Russian, French, German, Polish, Italian, Ewe, Malay, Japanese, Chinese, Arabic, Yankunytjatjara, Arrernte, and Maori, among others. “Leibnizian linguistics” as represented in the NSM movement is very different from the “Cartesian linguistics” of Chomsky’s generative grammar. But there was no real conflict between Descartes’ ideas about language and thought and those of Leibniz. In particular, the fundamental idea that at the basis of language and thought there must be a set of innate conceptual primitives expressed in indefinable words or natural languages was fully shared by Leibniz and Descartes (see Descartes 1931 [1701], Arnauld 1964 [1662]). In this sense, the NSM linguistics, too, could rightfully be called “Cartesian
248 Anna Wierzbicka
linguistics” — perhaps more rightfully than generative grammar, in which the idea of innate conceptual primitives has never played any role. But Leibniz’s program of linguistic and conceptual inquiry was much more fully developed than that of Descartes, and was supported by numerous examples of linguistic analysis, both lexical and grammatical. It was also linked with a deep interest in and appreciation of languages — “the most ancient monuments of mankind” (Leibniz 1981 [1765]: 336). It is Leibniz, therefore, who should be seen as the progenitor of any linguistics based on empirically established conceptual primes (a “universal lexicon”) and their combinatorial possibilities (“universal grammar”). To those interested both in the diversity of languages as conceptual and cultural universals and in the underlying unity of human thoughts it is Leibniz who points the way.
Notes 1. For a representative sample, see the references to the works of Ameka (1990a,b), Bugenhagen (1990), Chappell (1986, 1991), Durst (forthcoming), Goddard (1991, 1994, 1995, 1996, 1997a,b, 1998), Harkins (1996), Hasada (1996, 1997), Kornacki (1995), Peeters (1993), Rieschild (1996), Travis (1998a, 1998b), Wierzbicka (1972, 1980, 1987, 1988, 1991, 1992, 1996, 1997, 1999) and D. Wilkins (1986). 2. Alphabetum Cogitationum Humanarum est catalogus earum quae per se concipiuntur, et quorum combinatione ceterae ideae nostrae exurgunt. (Couturat (ed.) 1903: 430) 3. quae consideratio [quomodo notiones aliae ex allis oriantur] nos faciet evitare inanes circulos, quos plerumque in hic rebus decurrimus, figeturque animus et ad certas quasdas constantesque notiones constituendas cogetur. Quod quanti momenti sit, pauci capiunt quia pauci considerant quanti sit prima in omnibus elementa constituisse. (Couturat (ed.) 1903: 160) 4. Les premiers terms indéfinibles ne se peuvent aisement reconnoisse de nous, que comme les nombres premiers: qu’on ne sçauroit discerner jusqu’icy qu’en essayant la division [par tous les autres qui son moindres]. (Couturat (ed.) 1903: 187). 5. Si nihil per se concipitur, nihil omnino concipietur. Nam quod ·non nisiÒ per alia concipitur, in tantum concipietur in quantum alia illa concipiuntur et hoc rursum ita: ac proinde tum demum ·actu ipsoÒ aliquid concipere dicemur, cum in ea quae per se concipiuntur incidemus. (Couturat (ed.) 1903: 430) 6. J’avois consideré cette matiere avant le livre de Mr Wilkins, quand j’estois un jeune homme de dix-neuf ans, dans mon petit livre de Arte combinatoria, et mon opinion est que les Caracteres veritablement réels et philosophiques doivent repondre à l’Analyse des pensées. (Couturat (ed.) 1901 [1697]: 54)
Leibnizian linguistics 249
7. Scopus nostrae Characteristicae est tales adhibere voces, ut omnes consequentiae quae institui possunt statim ex ipsis verbis vel characteribus emanantur …. Haec consequentia ex his vocabulis latinis nisi resolvantur in alia aequipollentia demonstrari non potest; in lingua generali debet ex vocabulorum analysi in suas literas demonstrari posse. (Couturat (ed.) 1901: 75n) 8. Über die einfachen Begriffe und Definitionen sind eine Unzahl Manuscripte unter den Leibnizischen Papieren vorhanden … . Durchdrungen von der Überzeugung, dass er die riesige Arbeit nicht allein zu Stande bringen könne, wie oft hat er sich nach Hülfe gesehnt! Endlich im vorletzten Jahrzehnt seines Lebens liess er durch seinen Secretär Johann Friedrich Hodann eine noch vorhandene Sammlung von Definitionen ausstellen … . (Gerhardt (ed.) 1960–61, Vol. 7: 30) 9. For other tables of definitions, see also Leibniz (1960) and Grua (1948). For discussion see Dascal (1987). 10. Il y a un abîme qui la sépare [la sémantique structurale] des anciens essais pour établir une sémantique universelle ou ars magna, culminant dans la scientia generalis ou characteristica generalis de G. W. Leibniz, mais reposant, pour le principe de la méthode et pour l’essence de son idée, sur l’Ars generalis de Raymond Lullus … . (Hjelmslev 1957: 98) 11. Quite a few of these empirically established universals have in fact appeared, at one time or another, in various lists compiled by Leibniz in his unpublished manuscripts. Among them, we will note think (cogitatio), feel (sensio), one (unitas), many (multitudo), the same (identitas), for some time (duratio), can (possibilitas), this (hoc) (Couturat 1903: 281); before, after (prius, posterius) (Grua 1948: 2, 542 quoted in Eco 1995: 276). 12. Explicandae omnes flexiones et pasticulae; reducendaque omnia ad simplicissimas explicationes, quae semper salvo sensu in locum substitui possunt. (Couturat (ed.) 1903: 244)
References Ameka, Felix 1990a “The grammatical packaging of experiences in Ewe: A study in the semantics of syntax”. Australian Journal of Linguistics [Special issue on the semantics of emotions, A. Wierzbicka (ed.)] 10 (2): 139–181. 1990b “How discourse particles mean: The case of the Ewe ‘terminal’ particles”. Journal of African Languages and Linguistics 12 (2): 143–170. Aristotle 1963 Categories. J. L. Ackrill (trans). Oxford: Clarendon. Arnauld, Antoine 1964 [1662] The art of thinking (Port-Royal Logic). Translated by James Dickoff and Patricia James. Indianapolis: Bobs-Merrill. Bogusławski, Andrzej 1966 Semantyczne pojecie liczebnika i jego morfologia w j,ezyku rosyjskim. WrocRaw: Ossolineum.
250 Anna Wierzbicka
1970
“On semantic primitives and meaningfulness”. In Sign, Language and Culture, A. Greimas, R. Jakobson, M.R. Mayenowa and S. Z˙ółkiewski (eds), 14–152. The Hague: Mouton. Bruner, Jerome 1990 Acts of meaning. Cambridge, Mass.: Harvard University Press. Bugenhagen, Robert 1990 “Experiential constructions in Mangap-Mbula”. Australian Journal of Linguistics [Special issue on the semantics of emotions, A. Wierzbicka (ed.)] 10 (2): 183–215. Chappell, Hilary 1986 “The passive of bodily effect in Chinese”. Studies in Language 10 (2): 271–296. 1991 “Strategies for the assertion of obviousness and disagreement in Mandarin: A semantic study of the modal particle me”. Australian Journal of Linguistics 11: 39–65. Chomsky, Noam 1966 Cartesian linguistics: A chapter in the history of rationalist thought. New York: Harper & Row. Clark, Eve 1971 “On the acquisition of the meaning of ‘before’ and ‘after’”. Journal of Verbal Learning and Verbal Behaviour. 10: 266–275. Couturat, Louis 1901 La logique de Leibniz. Paris: Presses Universitaires de France. [Reprinted 1961, Hildesheim: George Olms.] 1903 Opuscules et fragments inédits de Leibniz. Paris: Presses Universitaires de France. Croft, William 1990 Typology and Universals. Cambridge: Cambridge University Press. Dascal, Marcelo 1987 Leibniz: Language, signs and thought. A collection of Essays. Amsterdam: John Benjamins. Descartes, René 1931 [1701] “The search after truth by the light of nature”. In The philosophical works of Descartes Vol. 1. Trans. E. S. Haldane and G.R.T. Ross, 305–327. Cambridge: Cambridge University Press. Durst, Uwe forthcoming“Why Germans don’t feel ‘anger’”. In Emotions in a cross-linguistic perspective, A. Wierzbicka and J. Harkins (eds). Berlin: Mouton de Gruyter. Eco, Umberto 1995 The Search for the Perfect Language. Oxford: Blackwell. French, Lucia A. and Katherine Nelson 1985 Young children’s knowledge of relational terms: Some ifs, ors and buts. New York: Springer Verlag. Gensini, Stefano 1990 Gottfried W. Leibniz: Dal segno alle lingue. Profilo, testi, materiale. Casale Monferrato: Marietti Scuola.
Leibnizian linguistics
1991 Il naturale e il simbolico. Rome: Bulzoni. Gerhardt, C.I. 1960–61 [1890]Die philosophischen Schriften von Gottfried Wilhelm Leibniz, vols 1–7. Hildesheim: Georg Olms. Goddard, Cliff 1991 “Anger in the Western Desert: A case study in the cross-cultural semantics of emotion”. Man 26: 602–19. 1994 “The meaning of lah: Understanding “emphasis” in Malay (Bahasa Melayu)”. Oceanic linguistics 33 (1): 145–165. 1995 “Who are ‘we’? The natural semantics of pronouns”. Language Sciences 17 (1): 99–121. 1996 “The “social emotions” of Malay (Bahasa Melayu)”. Ethos 24 (3): 426–464. 1997a “Grammatical categories and semantic primes”. Australian Journal of Linguistics 17 (1): 1–43. 1997b Studies in the Syntax of Universal Semantic Primitives [Special issue of Language Sciences 19 (3)]. 1998 Semantic Analysis: A Practical Introduction. Oxford: Oxford University Press. Goddard, Cliff and Wierzbicka, Anna 1994 Semantic and Lexical Universals: Theory and empirical findings. Amsterdam: John Benjamins. in pressMeaning and Universal Grammar. Amsterdam: John Benjamins. Grua, Gaston 1948 Gottfried W. Leibniz: Textes inédits de la Bibliotèque provinciale de Hanovre. Paris: PUF. Harkins, Jean 1996 “Linguistic and cultural differences in concepts of shame”. In Shame and the Modern Self, D. Parker, R. Dalziell and I. Wright (eds), 84–96. Melbourne: Australian Scholarly Publishing. Harré, Rom and Gillett, Grant 1994 The Discursive Mind. Thousand Oaks, California: Sage Publications. Hasada, Rie 1996 “Some aspects of Japanese cultural ethos embedded in nonverbal communicative behaviour”. In Nonverbal Communication in Translation, F. Poyatos (ed.), 83–103. Amsterdam: John Benjamins. 1997 “Conditionals and counterfactuals in Japanese”. Language Sciences 19 (3): 277–288. Hjelmslev, Louis 1953 Prolegomena to a Theory of Language. Translated by F.J. Whitfield. Madison, Wis.: University of Wisconsin Press. 1957 “Pour une sémantique structurale”. Reprinted in Essais linguistiques, 1959, 96–112. Copenhagen: Nodisk Sprog-og Kulturforlag. Katz, Jerrold J. and Fodor, Jerry A. 1963 “The structure of a semantic theory”. Language 39: 170–211.
251
252 Anna Wierzbicka
Kornacki, Pawel 1995 Aspects of Chinese cultural psychology as reflected in the Chinese lexicon. Doctoral dissertation, Australian National University. Lakoff, George 1970 “Natural logic and lexical decomposition”. In Papers from the Sixth Regional Meeting of the Chicago Linguistic Society, M. A. Campbell et al (eds), 340–362. Chicago: Chicago Linguistic Society. Leibniz, Gottfried Wilhelm 1960 Fragmenten zur Logik. Edited by Franz Schmidt. Berlin: Akademie-Verlag. 1981 [1765]New essays on human understanding. Peter Remnant & Jonathan Bennett (trans. and eds). Cambridge: Cambridge University Press. Martin, Gottfried 1964 Leibniz: Logic and metaphysics. K.J. Northcott and P.G. Lucas (trans.). Manchester: Manchester University Press. McCawley, James D. 1973 Grammar and meaning. New York: Academic Press. Needham, Rodney 1972 Belief, language and experience. Oxford: Blackwell. Onishi, Masayuki 1994 “Semantic Primitives in Japanese”. In Semantic and lexical universals: theory and empirical findings, C. Goddard and A. Wierzbicka (eds), 361–385. Amsterdam: John Benjamins. 1997 “The grammar of mental predicates in Japanese”. Language Sciences 19 (3): 219–233. Peeters, Bert 1993 “Commencer et se mettre à: une description axiologico-conceptuelle”. Langue française 98: 24–47. Peeters, Bert and Eiszele, Aileen 1993 “Le verbe prendre pris au sérieux”. Cahiers de lexicologie 62: 169–184. Pinker, Steven 1994 The language instinct. New York: William Morrow. Rieschild, Verna 1996 Lebanese-Arabic Discourse: adult interaction with young children (with reference to Australian-English situations). PhD Dissertation, Australian National University. Sapir, Edward 1949 Selected Writings of Edward Sapir in Language, Culture and Personality, David Mandelbaum (ed.). Berkeley: University of California Press. Searle, John R. 1994 The rediscovery of the mind. Cambridge, Mass.: MIT Press. Shweder, Richard A. 1990 “Cultural psychology: What is it?” In Cultural psychology: Essays on comparative human development. J.W. Stigler, R.A. Shweder and G.Herdt (eds), 1–46. Cambridge: Cambridge University Press.
Leibnizian linguistics 253
Sørensen, Holger Steen 1958 Word-classes in modern English. Copenhagen: Gad. Steiner, George 1992 After Babel. Oxford: Oxford University Press. Tien, Adrien 1997 NSM and language acquisition. MA. thesis, Australian National University. Travis, Catherine 1998a “Omoiyari as a core Japanese value: Japanese-style empathy?” In Speaking of Emotions: Conceptualization and expression, A. Athanasiadou and E. Tabakowska (eds), 55–82. Berlin: Mouton de Gruyter. 1998b “Kind, Considerate, Thoughtful: A semantic analysis”. Lexikos 7: 130–152. Wierzbicka, Anna 1972 Semantic Primitives. Frankfurt: Athenäum Verlag. 1980 Lingua Mentalis: The Semantics of Natural Language. Sydney: Academic Press. 1987 English Speech Act Verbs: A semantic dictionary. Sydney: Academic Press. 1988 The Semantics of Grammar. Amsterdam: John Benjamins. 1991 Cross-cultural Pragmatics: The semantics of social interaction. Berlin: Mouton de Gruyter. 1992 Semantics, Culture, and Cognition. Oxford: Oxford University Press. 1996 Semantics: Primes and Universals. Oxford: Oxford University Press. 1997 Understanding Culture through their Key Words. Oxford: Oxford University Press. 1999 Emotions across Languages and Cultures. Cambridge: Cambridge University Press. Wilkins, David 1986 “Particles/clitics for criticism and complaint in Mparntwe Arrente (Aranda)”. Journal of Pragmatics 10: 575–596. Wilkins, John 1668 An essay towards a real character and a philosophical language. London: Gellibrand.
DOCINFO
AUTHOR ""
TITLE "Dialogue and Thematic Structure"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Part IV
Dialogue and thematic structure
AUTHOR "Kerstin Jonasson"
TITLE "Naming conventions, focalization, and point of view in Balzac’s La Peau de chagrin"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Naming conventions, focalization, and point of view in Balzac’s La Peau de chagrin* Kerstin Jonasson Uppsala University
The famous beginning of Balzac’s novel La Peau de chagrin (originally published in 1830) is usually cited by literary critics (Genette 1983: 46) as a typical example of nineteenth century narrative technique: (1) Vers la fin du mois d’octobre dernier, un jeune homme entra dans le Palais-Royal au moment où les maisons de jeu s’ouvraient, conformément à la loi qui protège une passion essentiellement imposable. (Balzac 1989: 19) ‘Towards the end of the month of October 1829 a young man entered the Palais-Royal just as the gaming-houses opened, agreeably to the law which protects a passion by its very nature easily excisable.’ (Marriage 1998)
The hero is introduced here with an indefinite NP, un jeune homme (‘a young man’), as an unknown person who is seen from the outside and whose identity and proper name are kept secret from the reader for no less than around forty pages. As Genette puts it (1972: 208 n.1) when discussing the beginning of Flaubert’s L’Education sentimentale: “Tout se passe comme si, pour l’introduire, l’auteur devait feindre de ne pas le connaître” (‘Everything happens as if the author, when introducing him, must pretend that he does not know him’). Genette also mentions Zola’s novel Germinal, where the hero continues to be called un homme (‘a man’) until he introduces himself: “Je me nomme Etienne Lautier” (loc.cit., ‘My name is Etienne Lautier’). From that moment on Zola calls him Etienne. In La Peau de chagrin, we learn that the protagonist’s name is Raphaël, when he stumbles on some friends who call him by his first name:
258 Kerstin Jonasson
(2) En s’élançant de la porte du magasin sur la chaussée, il heurta trois jeunes gens qui se tenaient bras dessus bras dessous. […] – Eh ! c’est Raphaël. – Ah ! bien, nous te cherchions. – Quoi ! c’est vous ? (Balzac 1989: 56) ‘As he rushed out of the door into the street, he ran up against three young men who were passing arm-in-arm. […] “Why, it is Raphaël!” “Good! we were looking for you.” “What! It is you, then?”’ (Marriage 1998)
After having described his hero in terms of definite NPs, such as le jeune homme (‘the young man’) and l’inconnu (‘the stranger’), or designated him by the personal pronoun il (‘he’), from here on Balzac’s narrator uses the proper name as the standard reference. This type of “introït énigmatique” (Genette 1972: 207), which Genette classifies as an external focalization, has generally been abandoned nowadays and replaced by a narrative technique which allows the reader to get immediate access not only to the protagonist’s identity, either by using the proper name directly or even by using a referentless or antecedentless pronoun (Emmott 1997), but also to his/her thoughts and feelings. An example of this type of internal focalization is the beginning of André Malraux’ novel La Condition humaine from 1933: (3) Tchen tenterait-il de lever la moustiquaire ? Frapperait-il au travers ? L’angoisse lui tordait l’estomac ; il connaissait sa propre fermeté, mais il n’était capable en cet instant que d’y songer avec hébétude, fasciné par ce tas de mousseline blanche qui tombait du plafond sur un corps moins visible qu’une ombre, et d’où sortait seulement ce pied à demi-incliné par le sommeil […] (Malraux 1933) ‘Should Tchen try to lift up the mosquito net? Should he strike through it? The anguish twisted his stomach; he knew his own firmness, but at this moment he was not capable of thinking of it clearly, so fascinated was he by the white muslin that hung from the ceiling on a body less visible than a shadow, and from which stuck out only this foot half-tilted by sleep […]’
Genette’s focalization theory tries to give answers to the questions Who speaks? and Who sees?, i.e. to define the narrative perspective or the point of view in a story. To put it simply, the distinction between external and internal focalization
Focalization in Balzac’s La Peau de chagrin 259
is based on the difference between seeing from without or from within. “Intern fokalisation avslöjar vad en gestalt tänker och känner, extern bara det yttre beteendet” (Skalin 1991: 206, ‘Internal focalization reveals what a character thinks and feels, external only the external behavior’). According to Genette (1972: 208) however, an external focalization on one character may at the same time be an internal focalization on another, which seems to indicate that the model is somewhat problematic, since it can contain two types of focal characters. Incidentally, this ambivalence is, according to Genette, present at the beginning of La Peau de chagrin, where the witness is not “personalized” but remains a vague and floating observer. The external focalization on the young man is simultaneously an internal focalization on an anonymous witness who is seeing him. Gradually, however, the narrator abandons the external focalization on the protagonist, who according to Genette is internally focalized in sentences like “quant au jeune homme, il ne comprit sa ruine qu’au moment où le râteau s’allongea pour ramasser son dernier napoléon” (‘The young man only understood his calamity when the croupier’s rake was extended to sweep away his last napoleon’), or “il affecta l’air d’un Anglais pour qui la vie n’a plus de mystères” (Balzac 1989: 26, Genette 1972: 213, ‘he affected the airs of an Englishman, to whom life can offer no sensation’, Marriage 1998). Elsewhere in the story it is also quite clear that the focal character is personalized and internally focalized. Such is the case in the following excerpt, where Raphaël’s point of view is expressed again: (4) Il [=Raphaël] marchait comme au milieu d’un désert, coudoyé par des hommes qu’il ne voyait pas, n’écoutant à travers les clameurs populaires qu’une seule voix, celle de la mort (Balzac 1989: 27) ‘He walked as if he were in some desert, elbowed by men whom he did not see, hearing through all the voices of the crowd one voice alone — the voice of death.’ (Marriage 1998)
If Raphaël is focalized externally in the beginning of the sentence, the perception verbs voyait (‘saw’) and écoutait (‘listened to’) change the focalization into an internal one, where we gain access to Raphaël’s experience (incidentally, long before we get access to his identity through his name). A little earlier in the story Balzac’s narrator adopts the point of view of the gamblers: (5) Au premier coup d’oeil les joueurs lurent sur le visage du novice quelque horrible mystère, ses jeunes traits étaient empreints d’une grâce nébuleuse, son regard attestait des efforts trahis, mille espérances trompées ! [—] Quelque secret génie scintillait au fond de ces yeux voilés
260 Kerstin Jonasson
peut-être par les fatigues du plaisir. Etait-ce la débauche qui marquait de son sale cachet cette noble figure jadis pure et brillante, maintenant dégradée ? (Balzac 1989: 24) ‘The gamblers saw at a glance a dreadful mystery in the novice’s face. His young features were stamped with a melancholy grace, his looks told of unsuccess and many blighted hopes! […] Some sort of demon sparkled in the depths of his eyes, which drooped, wearied perhaps with pleasure. Could it have been dissipation that had set its foul mark on the proud face, once pure and bright, and now brought low?’ (Marriage 1998)
The narrator here describes how the gamblers take in Raphaël when he enters the gambling-house. They constitute the subject of the first sentence, and the expression chosen to designate Raphaël, le novice (‘the beginner’), also reveals where the point of view is located. The adverb peut-être (‘perhaps’) indicates the tentative attempts of the gamblers to interpret what they are seeing, and their curiosity is further underlined by the form of a direct question. Thus this passage is full of subjective expressions or empathy markers, i.e. linguistic expressions that mark a certain character’s point of view or perspective (cf. Banfield 1982; Hellberg 1984). We have here another example of an internal focalization on one character (here several characters) that is at the same time an external focalization on another. The internal focalization on the gamblers makes another character, Raphaël, stand out as seen from their point of view with his air of mystery, and their curiosity is of course intended to arouse that of the reader. The ambivalence problem caused by the simultaneous presence of several types of focal characters has led other narratologists to redefine Genette’s terms to some extent. Mieke Bal makes the distinction between internal focalization having a subject that can see, and external focalization having an object that is seen (apud Skalin 1991: 291). This distinction has been adopted by Manfred Jahn (1996), who takes the window as a metaphor for a revised theory of focalization, based on a model of vision. According to Jahn’s account, Focus 1 stands metonymically for the eye’s owner, i.e. the subject of focalization, and Focus 2 for its object, i.e. the object or area in the focus of interest or attention. Thus Focus 1 is the character(s) that see(s), perceive(s), hear(s), think(s), remember(s), etc., and whose consciousness the narrator makes accessible to us, while Focus 2 represents that which is seen, heard or in some other way is paid attention to by Focus 1 (or the narrator!). A similar model has also been proposed by two American linguists (Zubin and Hewitt 1995) within the framework of Deictic Shift Theory, elaborated by an interdisciplinary research
Focalization in Balzac’s La Peau de chagrin 261
group at the State University of Buffalo N.Y. (cf. Duchan et al. 1995). In their model Focus 1 corresponds to the deictic center or origo (cf. Bühler 1934) from which the story’s events are described. This center can either coincide with one (or several) character(s) in the story, and represent a subjective perspective, or be located outside the story in a narrator, who may be either covert (in modernist narratives) or intrusive and opinionated like Balzac’s, giving rise respectively to an objective or subjective perspective. As a linguist I have become interested in focalization theory and narratological models of perspective and point of view in connection with my research on referential expressions and the definiteness and indefiniteness of noun phrases. Since the question of definiteness is related to that of identifiability, and identifiability is a cognitive stance relative to some consciousness, the question that arises is, whose consciousness? In a verbal interaction type like ordinary conversation we usually adapt to the presumed knowledge of the person we are talking to and choose an indefinite or a definite form after having judged that person’s possibilities of identifying the person or thing we want to talk about. But in a story the author can use this distinction in order to adopt a certain perspective. For instance to present new information as already known is an example of empathy marking mentioned by Hellberg (1984). To use an indefinite NP to designate a referent that has already been introduced and thus should in principle be an identifiable entity can have the effect of presenting this referent in another light or from another perspective (cf. Ushie 1986). Especially when it comes to the description of demonstrative NPs like cet homme (‘this man’), ce vieillard (‘this old man’), cette charmante personne (‘this charming person’), the focalization theories and the models of narrative perspectives have proven to be of great interest. According to Philippe (1998: 60), the French demonstrative determiner (ce, cette, ces) is a very strong empathy marker in narrative texts and always establishes a specific deictic center or origo. This center may be a character within the story who assumes the Focus 1 position, or the narrator. Chafe, whose study on discourse and consciousness (1994) also contributes greatly to elucidating these matters, describes this distinction in terms of “represented consciousness” and “representing consciousness” respectively. I have argued elsewhere (Jonasson 2000a) that in terms of Jahn’s focalization theory the referent introduced by a demonstrative determiner is always Focus 2, never Focus 1. He/she/it is by means of such a description seen from the outside, an object being seen, not a subject that sees, and can not in that context have the empathy.
262 Kerstin Jonasson
To what extent do the referential expressions chosen by Balzac in La Peau de chagrin contribute to creating or marking a certain point of view in the story? I have investigated the first part of the novel, entitled “Le Talisman”, where we follow Raphaël de Valentin into the gambling-house where he loses his last napoleon, out on his suicidal walk down to the Seine, into the old curiosity shop where he buys the magic skin, and eventually to the orgy at Taillefer’s house. During these dramatic hours he meets a lot of people, about one hundred, who are introduced by Balzac and then designated by way of different NPs. I shall here concentrate my attention on the main character in whom the narrator often places the empathy and who is often Focus 1, the deictic center, even if in the beginning he is Focus 2. There are about 250 references to Raphaël on the 74 pages that constitute the part called “Le Talisman”. The most common expression used by Balzac to designate him is the personal pronoun il (‘he’), which appears 170 times, i.e. two out of three times. This indicates his status as protagonist and thematic referent. Among the rest of the characters, the shop-keeper and Émile, Raphaël’s best friend, rank above the others with 51 and 47 references respectively, 11 and 17 respectively in the form of il. The proper name Émile appears 23 times, which makes it almost as frequent as the name Raphaël (that of the hero), which is used 26 times. As already mentioned, the Balzacian narrator often uses descriptive definite NPs to designate Raphaël, of which le jeune homme (‘the young man’) is by far the most frequent with its 33 occurrences, whereas l’inconnu (‘the stranger’) appears 23 times. The definite article is neutral as to perspective, but the noun that follows may indicate a point of view. This seems to be the case with l’inconnu, which describes the referent’s relation to a certain consciousness other than his own. However, this semantic information is not always used to adopt the perspective of some other character(s) than Raphaël. In the first of the following examples, it is used in that way, but not in the other two: (6) Le silence devint en quelque sorte plus profond, et les têtes se tournèrent vers le nouveau venu par curiosité. Chose inouïe ! les vieillards émoussés, les employés pétrifiés, les spectateurs, et jusqu’au fanatique Italien, tous en voyant l’inconnu éprouvèrent je ne sais quel sentiment épouvantable. (Balzac 1989: 23–24) ‘The silence seemed to grow deeper as all heads turned curiously towards the new arrival. Who would have thought it? The jaded elders, the fossilized waiters, the onlookers, the fanatic Italian himself, felt an indefinable dread at the sight of the stranger.’ (Marriage 1998)
Focalization in Balzac’s La Peau de chagrin 263
(7) Dans l’horrible situation où se trouvait l’inconnu, ce babil de cicérone, ces phrases sottement mercantiles furent pour lui comme les taquineries mesquines par lesquelles des esprits étroits assassinent un homme de génie. (Balzac 1989: 32) ‘In the the stranger’s fearful position this cicerone’s prattle and shopman’s empty talk seemed like the petty vexations by which narrow minds destroy a man of genius.’ (Marriage 1998) (8) L’inconnu, surpris et presque irrité de se voir toujours plaisanté par ce singulier vieillard dont l’intention demi-philanthropique lui parut clairement démontrée dans cette raillerie, s’écria : (Balzac 1989: 55) ‘The stranger was surprised and irritated that this peculiar old man persisted in not taking him seriously. A half philantropic intention peeped so clearly forth from his last jesting observation, that he exclaimed:’ (Marriage 1998)
In the first example, Raphaël is the object of the attention of the persons in the gambling-house, for whom he is the new-comer, the stranger and Focus 2. But in the latter two excerpts, the stranger’s point of view is adopted, and despite his being described in the same way, he is now Focus 1, the empathy-holder. The semantic content of l’inconnu is no longer used to mark that the empathy is placed outside Raphaël but has become an ordinary alternative to le jeune homme in referring to the hero without respect to perspective or point of view. According to Hellberg (1984) relational nouns or NPs like John’s wife also function as empathy markers. An example of this may be found in the following excerpt: (9) Le soupçonneux marchand examina d’un oeil sagace le morne visage de son faux chaland tout en l’écoutant parler. Rassuré bientôt par l’accent de cette voix douloureuse, ou lisant peut-être dans ces traits décolorés les sinistres destinées qui naguère avaient fait frémir les joueurs, il lâcha les mains ; (Balzac 1989: 47) ‘While he spoke, the jealous merchant watched the haggard face of his pretended customer with keen eyes. Perhaps the mournful tones of his voice reassured him, or he also read the dark signs of fate in the faded features that had made the gamblers shudder; he released his hands,’ (Marriage 1998)
In describing Raphaël as the shop-keeper’s “pretended customer” (faux chaland), the narrator uses “the suspicious merchant” (le soupçonneux
264 Kerstin Jonasson
marchand) as Focus 1 or deictic center. This point of view is also underlined by the possessive pronoun son (‘his’) and by the perception verbs examina d’un oeil sagace (‘watched … with keen eyes’) and écoutant (not translated in the English version) in the same sentence. This perspective is maintained in the description of Raphaël, who is here seen from the outside and constitutes Focus 2, which also comes from his being designated by demonstrative NPs like cette voix douloureuse (‘that mournful voice’, translated as a possessive NP), ces traits décolorés (‘those faded features’, translated as a definite NP1). The shop-keeper is here Focus 1, i.e. the deictic center from which Raphaël is observed. Let us now turn to the passages where Raphaël is designated by a demonstrative NP and examine whether he can act as a center of empathy in such contexts. If we except descriptions of parts of his body, as in the above example, and look at the demonstrative NPs that refer to him as a person, they are only four, i.e. not even 2% of the 250 references to Raphaël. This can be compared to the shop-keeper who ten times out of 51 is described with a demonstrative NP, i.e. in 20% of the references to him. Who is Focus 1 in the four cases where Raphaël is Focus 2, i.e. designated by a demonstrative NP? Is it the narrator or one (or several) of the characters in the story that put him in the center of attention? Twice he is described as ce jeune homme (‘this/that young man’): (10) Il entra résolument dans la salle, où le son de l’or exerçait une éblouissante fascination sur les sens en pleine convoitise. Ce jeune homme était probablement poussé là par la plus logique de toutes les éloquentes phrases de Jean-Jacques Rousseau, et dont voici, je crois, la triste pensée : (Balzac 1989: 20–21) ‘He walked boldly into the saloon, where the rattle of coin brought his senses under the dazzling spell of an agony of greed. Most likely he had been drawn thither by that most convincing of Jean-Jacques’ eloquent periods, which expresses, I think, this melancholy thought,’ (Marriage 1998) (11) – J’ai entendu, dit-il, une voix qui me criait dans l’oreille : Le Jeu aura raison contre le désespoir de ce jeune homme. (Balzac 1989: 26) ‘“A voice seemed to whisper to me”, he said. “The luck is sure to go against that young man’s despair.”’ (Marriage 1998)
In the first example it is obvious that it is the narrator that draws attention to Raphaël by using the expression ce jeune homme (‘this young man’2). He even appears as responsible for an assumption (probablement, ‘probably’) about what
Focalization in Balzac’s La Peau de chagrin 265
has driven Raphaël to the gambling-house, and he also takes the responsibility for the quotation from Rousseau, where je crois (‘I think’) explicitly indicates the strength of his belief concerning the correct wording. The choice of the demonstrative determiner instead of the definite article, which would have been quite acceptable here,3 has the effect of focalizing Raphaël, who is picked out of the scenario and isolated in order to be further described from the narrator’s perspective.4 In the second example, the Italian who won the game is speaking. He is the deictic center of a subjective perspective, and with the demonstrative NP he designates Raphaël, who has just left the gambling-house after having lost his last napoleon. The hero is still in the center of attention of the gamblers and thus clearly Focus 2.5 In the third demonstrative NP reference it is not so clear from what perspective Raphaël is seen: (12) Un jeune ramoneur dont la figure bouffie était noire, le corps brun de suie, les vêtements déguenillés, tendit la main à cet homme pour lui arracher ses derniers sous. (Balzac 1989: 30) ‘A little chimney sweeper, with puffed cheeks, all black with soot, and clad in tatters held out his hand to beg for the man’s last pence.’ (Marriage 1998)
If it is still the narrator who points to Raphaël by using cet homme (‘this/that man’), it is not with the same intention as in the first example above. Raphaël is not picked out to be the object of a further description. Rather, it seems as if it is the little chimney sweeper’s perspective that is adopted here. The example is preceded in the text by his appealing cry: “–Ah ! mon bon monsieur, la carita ! la carita ! catarina ! Un petit sou pour avoir du pain !” (‘–Ah, kind gentleman! carita, carita; for the love of St. Catherine! only a halfpenny to buy some bread!’) and ends with an indication of the intention behind his stretched hand (pour lui arracher ses derniers sous, ‘to beg for the man’s last pence’). According to Hellberg (1984: 43) more or less strong empathy is expressed when someone’s intention is communicated in a story. In any case it seems clear that Raphaël is the object (Focus 2) and not the subject (Focus 1) of the focalization. The fourth and last example is the following: (13) En arrivant à l’étalage d’un marchand d’estampes, cet homme presque mort rencontra une jeune femme qui descendit d’un brillant équipage. (Balzac 1989: 30)
266 Kerstin Jonasson
‘As he reached the shop window of a print-seller, this man on the brink of death met a young woman alighting from a showy carriage.’ (Marriage 1998)
Here we have no presumptive Focus 1 or subject of any focalization other than the anonymous witness who was present at the beginning of the story. This would suggest that you could see from the way he looked that Raphaël was half dead, and, as a matter of fact, the rest of the sentence describes visible things, too. But the demonstrative may also be motivated by the fact that the narrator, acting as “the representing consciousness”, wants to give the readers some new information about Raphaël and finds it appropriate to do so by incorporating the information in the form of an adjective phrase (presque mort) in a referring expression that designates the referent unambiguously. With a definite NP this state of half-death would have been presented as already known, which it is not by the readers at this moment in the story, and the identification of the referent might thus have been problematic. This is not the case with demonstrative NPs, which often reclassify a thematic referent or predicate new information about it (cf. Maes and Noordman 1995; Apothéloz and Reichler-Béguelin 1999). This is so because demonstrative NPs rely on contextual clues for the identification of their referent, and not, as definite NPs, on their own lexical material. It is a general rule that referents already introduced in the discourse are subsequently referred to by definite NPs, proper names or pronouns, all of which establish coreference. Thus, after having introduced Raphaël as un jeune homme (‘a young man’), Balzac’s narrator must subsequently use such expressions to designate his hero, and we have seen that this is what he does. Except for the first introductory indefinite NP at the beginning, there are however, in “Le Talisman”, two other indefinite NPs that seem to refer to Raphaël. The first one appears when Émile tells Raphaël about the anxiety he and their friends have felt about Raphaël: (14) – Mon cher, dit l’orateur en continuant, nous sommes à ta poursuite depuis une semaine environ. […] Nous ne t’avons rencontré nulle part, ni sur les écrous de Sainte-Pélagie, ni sur ceux de la Force ! […] nous gémissions sur la perte d’un homme doué d’assez de génie pour se faire également chercher à la cour et dans les prisons. (Balzac 1989: 57) ‘“We have been after you for about a week”, the speaker went on.[…] “We could not find you anywhere. Your name was not in the jailers’ registers at the St. Pélagie nor at La Force! […] We bewailed the loss of a
Focalization in Balzac’s La Peau de chagrin 267
man endowed with such genius that one might look to find him at Court or in the common jails”.’ (Marriage 1998)
It seems quite clear that it is Raphaël who is meant by the italicized indefinite NP, and it presents him in a new light, from another perspective. From having observed and described him as a close friend and addressed him as tu, Émile here gives a pertinent and witty portrait of Raphaël that could have been addressed to someone else. The expression contains new information about Raphaël, and that motivates the indefinite article, which could not be replaced by the definite article here. The demonstrative determiner would not be appropriate either, because Émile is still speaking to Raphaël and cannot refer to his interlocutor by using a demonstrative NP. According to Ushie (1986), these coreferential indefinite expressions, may communicate an interpretation of some person’s behavior. This seems to be the case here, since the indefinite description implies an explanation for the strong reaction of his friends: a man of such genius that you have to look for him both at court and in the prisons must not get lost. The second example appears at the end of a two-page description of Taillefer’s sumptuous apartment, the guests at the banquet, and Raphaël’s admiration for all this: (15) Avant de quitter les salons, Raphaël y jeta un dernier coup d’oeil. […] Tout jusqu’aux draperies respirait une élégance sans prétention ; enfin, il y avait en tout je ne sais quelle grâce poétique dont le prestige devait agir sur l’imagination d’un homme sans argent. (Balzac 1989: 64) ‘Raphaël took a last look round the room before he left it. […] Everything, even the curtains, was pervaded by elegance without pretension, and there was a certain imaginative charm about it all which acted like a spell on the mind of a needy man.’ (Marriage 1998)
Once more it is Raphaël who is described with an indefinite NP, un homme sans argent (‘a needy man’). His proper name could also have been used here, and in the following sentence the narrator refers to him with a personal pronoun. One could once more agree with Ushie (1986) that when choosing this indefinite description of Raphaël instead of his proper name, the narrator gives us an explanation of why Raphaël is so fascinated by his surroundings. This time, however, the information about Raphaël is not new: the reader knows already that he has no money. One could therefore perhaps have expected to find a definite description here, l’homme sans argent, (‘the needy man’), but such an
268 Kerstin Jonasson
expression would probably prompt a generic and not a coreferential interpretation in this context and would thus perhaps not be taken as referring to Raphaël. On the other hand, a demonstrative NP, cet homme sans argent (‘this needy man’) could have been used without much difference. But with the indefinite form Raphaël’s poverty is enhanced as a new and, for the situation, particularly important property. According to Ushie (1986: 429) the indefinite expression functions as an NP with a non-restrictive apposition, i.e. as Raphaël, un homme sans argent. These two cases of what Ushie (1986) calls “co-representation”, i.e. a coreferential use of indefinite expressions, may thus be motivated by the fact that the events are here “interpreted” and not “merely presented”. As Ushie suggests, another factor that might explain this use of indefinite expressions is a change in point of view. There is an example of this at the beginning of the description of the shop-keeper that immediately follows the narration of Raphaël’s meeting with him, where he has been presented as un petit vieillard, (‘a little old man’), cette apparition (‘this apparition’), ce personnage (‘this figure’), cette espèce de fantôme (‘this sort of spectre’): (16) Figurez-vous un petit vieillard sec et maigre, vêtu d’une robe en velours noir, serrée autour de ses reins par un gros cordon de soie. (Balzac 1998: 42) ‘Imagine a short old man, thin and spare, in a long black velvet gown girded round him by a thick silk cord.’ (Marriage 1998)
Whereas the first two examples each appear at the end of a passage and imply an interpretation of the events described in that passage, this last one introduces a description of the curiosity-dealer in which the narrator directly addresses the readers after having given Raphaël’s first impression of him. So here we have an explicit shift in point of view that extends over a page and a half, after which Raphaël is restored as the center of empathy: Tel fut le spectacle étrange qui surprit le jeune homme au moment où il ouvrit les yeux […] (1989: 44, ‘This was the strange spectacle that startled the young man’s returning sight […]’, Marriage 1998). We have seen that Balzac’s choice of different referential expressions in referring to the characters and especially to the hero in La Peau de chagrin contributes to the establishment of and shifts in narrative perspective and point of view. Not only is the description offered by the noun in the NP, e.g. inconnu (‘stranger’), novice (‘beginner’), chaland (‘customer’) important here, the determiner itself is highly indicative, too, pointing out the referent’s status in
Focalization in Balzac’s La Peau de chagrin 269
the situation of utterance. It can therefore be related to a number of contextual aspects, such as the persons involved (narrator, readers, and fictive characters) and their respective points of view. Thus, by choosing a determiner that reflects a certain cognitive stance towards the referent, Balzac gives the reader the opportunity to empathize with the narrator or with one of the characters in the story.
Notes * I am most thankful to Manfred Jahn, whose comments on an earlier version of this paper largely improved both its content and form. 1. On the translation of this type of French demonstrative NPs into Swedish (cf. Jonasson 2000b). 2. Once more we find a demonstrative NP that the English translator has not bothered to render literally, choosing instead a pronominal reference, he. 3. And which was chosen by the Swedish translator: ‘den unge mannen’. 4. On this thematizing function of ce, see Kleiber 1986; De Mulder 1997. 5. In the English and Swedish translations of Balzac’s novel this ce is rendered by the distal demonstrative determiner ‘that’ and “den där” respectively: ‘that young man’s despair’, “den där förtvivlade unge mannen”. This explicit deprecation is absent in the French original where the demonstrative determiner is neutral as to the proximal/distal distinction expressed by Swedish and English demonstratives.
References
Primary sources Balzac, Honoré de 1989 (1830)La Peau de chagrin. Paris: Pocket (Collection Lire et Voir les Classiques). Malraux, André 1933 La condition humaine. Paris: Gallimard. Marriage, Ellen 1998 The magic skin, translation of Balzac 1989, available via project Gutenberg at http://promo.net/pg/.
Secondary sources Apothéloz, Denis and Reichler-Béguelin, Marie-José 1999 “Interpretations and functions of demonstrative NPs in indirect anaphora”. Journal of Pragmatics 31: 363–397.
270 Kerstin Jonasson
Banfield, Ann 1982 Unspeakable Sentences. Narration and Representation in the Language of Fiction. London: Routledge. Bühler, Karl 1934 Sprachtheorie. Die Darstellungsfunktion der Sprache. Jena: Fischer. Chafe, Wallace 1994 Discourse, Consciousness, and Time. The Flow and Displacement of Conscious Experience in Speaking and Writing. Chicago & London: The University of Chicago Press. De Mulder, Walter 1997 “Les démonstratifs : des indices de changement de contexte”. In Entre général et particulier : les déterminants, N. Flaux, D. Van de Velde and W. De Mulder (eds), 137–200. Arras, France: Artois Presses Université. Duchan, Judith, Bruder, Gail, and Hewitt, Lynne 1995 Deixis in Narrative. A Cognitive Science Perspective. Hillsdale, New Jersey: Lawrence Erlbaum Associates. Emmott, Catherine 1997 Narrative Comprehension: A Discourse Perspective. Oxford: Clarendon. Genette, Gérard 1972 Figures III. Paris: Seuil. 1983 Nouveau Discours du Récit. Paris: Seuil. Hellberg, Staffan 1984 “Satsens subjekt och textens”. Nysvenska studier 64: 29–82. Jahn, Manfred 1996 “Windows of Focalization: Deconstructing and Reconstructing a Narratological Concept”. Style 30: 241–267. Jonasson, Kerstin 2000a “Référence et perspective”. Actes du XXIIe Congrès international de linguistique et de philologie romanes, Bruxelles, 23–29 juillet 1998, Vol. VII, 305–313. Tübingen: Max Niemeyer Verlag. 2000b “Translation and Deixis”. översättning och tolkning. Rapport från ASLA’s höstsymposium. (Proceedings from the ASLA autumn Symposium on Translation and Interpretation), Stockholm, 5–6 November 1998, B. Englund Dimitrova (ed.), 173–181. Uppsala: Universitetstryckeriet. Kleiber, Georges 1986 “Adjectif démonstratif et article défini en anaphore fidèle”. In Déterminants: Syntaxe et sémantique, J. David and G. Kleiber (eds), 169–185. Paris: Klincksieck. Maes, Alfons and Noordman, Leo 1995 “Demonstrative nominal anaphors: a case of nonidentificational markedness”. Linguistics 33: 255–282. Philippe, Gilles 1998 “Les démonstratifs et le statut énonciatif des textes de fiction : l’exemple des ouvertures de roman”. Langue française 120: 51–65. Skalin, Lars-Åke
Focalization in Balzac’s La Peau de chagrin 271
1991
Karaktär och perspektiv. Att tolka litterära gestalter i det mimetiska språkspelet. Uppsala & Stockholm: Almqvist & Wiksell International.
Ushie, Yukiko 1986 “‘Corepresentation’ — A textual function of the indefinite expression”. Text 6 (4): 427–446. Zubin, David, and Hewitt, Lynne 1995 “The Deictic Center: A Theory of Deixis in Narrative”. In Deixis in Narrative. A Cognitive Science Perspective, J. Duchan et al. (eds), 129–155. Hillsdale, New Jersey: Lawrence Erlbaum Associates.
AUTHOR "Monika Doherty"
TITLE "Discourse theory and the translation of clefts between English and German"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Discourse theory and the translation of clefts between English and German Monika Doherty Humboldt University, Berlin
1.
Introduction
The following text presents a close analysis of a couple of examples which are prototypical for the translation of clefts between German and English. It is one of four papers written in connection with a two-year research project, which did not aim at a statistical survey but rather at in-depth analyses of the discoursefunction of clefts and their substitutes (mainly marked word order). The conclusions drawn from the data are: (1) clefts are used if less explicit focusing devices are not available (which means different things in German and English); and (2) the presuppositional semantics of clefts indicates macrostructural relevance, which marked word order does not. The use of clefts in translations between English and German differs widely. As a rule, English uses clefts much more often than German, which may or may not use substitutes for the clefts.1 There are some obvious formal differences between German and English clefts, the major one being the German constraint against daß/that-complement clauses. While English can say: “It was through her that he …”, German: *Es war durch sie, daß er … is ungrammatical. But as there is the acceptable German paraphrase: Es war sie, durch die er … (“*It was she through whom he …”), the lower frequency of German clefts cannot be blamed on this formal difference. Theoretically, we could assume that there are different stylistic conventions in English and German concerning the use of clefts. But different conventions can have linguistic or extra-linguistic reasons and a detailed analysis of the translational evidence should help us to identify some of the conditions for the use of clefts in both languages.2 As a rule, one distinguishes between it-clefts and wh-clefts (including socalled ‘reversed’ wh-clefts). The formal and functional characteristics of copular
274 Monika Doherty
matrix clauses with a special focusing effect are shared by structures like: “The (major) reason limiting the applications of these techniques is that …” But their complexity deserves special treatment.3 The canonical types of clefts can be classified further according to the hierarchical origin of the focused element (as a verb phrase internal or external element), and the syntactically dependent or independent nature of the cleft. These syntactic differences are suggested by the main substitute German uses instead of the English cleft: marked word order. There are two variants of marked order to be distinguished according to the embeddedness or unembeddedness of the cleft: (1) topicalization into the highest possible position in the syntactic hierarchy of an independent sentence (durch sie war er … “*Through her he was …”); (2) use of an equivalently marked word order in an embedded structure.4 (The second case will be taken up in some detail below.) Reordering verb phrase internal elements can focus the element reordered and in many cases English clefts correspond to marked word order in German. But subjects occupy a verb phrase external position anyway, which means that they cannot be focused by topicalization. Therefore, subjects are more likely to be focused by clefts — if their focus is marked at all (which, again, is the case more often in English than in German).5 There is a small set of English clefts where the German versions do not use topicalization although they could use it. (Ich sehne mich danach / “That was what I craved” Bernhard 1997: 153.)6 Whether there is anything systematic behind these exceptions will have to remain an open question for the time being. The following considerations will concentrate on the prototypical dichotomy in the German versions of English clefts: marked word order vs. clefts. I will proceed in the inductive way, present a (minimal) pair of examples first and then try to explain the phenomena along the theoretical lines suggested in the literature on clefts and related means. The explanation will include assumptions about discourse relations and language processing, to be taken up in due course.
2. Two examples Shortly before the climax of Süskind’s story of a murderer, Antoine Richis, the father of the most beautiful Laure, who had so far kept a cool head, began to be worried. “While others publicly celebrated the end of the rampage as if the
Discourse theory and the translation of clefts between English and German 275
murderer were already hanged and had soon fully forgotten about those dreadful days, fear crept into Antoine Richis’s heart like a foul poison.” (E1) For a long time he would not admit that it was fear that caused him to delay trips that ought to have been made some time ago, or to be reluctant merely to leave the house, or to break off visits and meetings just so that he could quickly return home. (209)
The last sentence is our first example, an it-cleft embedded into an attitudinal matrix clause and the translation stays very close to the German original: Während nämlich die Menschen draußen, als hätten sie den Mörder schon gehenkt, das Ende seines Treibens feierten und die unselige Zeit bald ganz vergaßen, kehrte in das Herz Richis’ die Angst ein wie ein häßliches Gift. (G1) Lange Zeit wollte er sich’s nicht zugeben, daß es die Angst war, die ihn bewog, längst fällige Reisen hinauszuzögern, ungern das Haus zu verlassen, Besuche und Sitzungen abzukürzen, damit er nur rasch wieder heimkehren könne. (256f)
But one page later we find Richis waking up from a nightmare: (E2) He went back to his chamber, bathed in sweat and trembling with agitation, no, not with agitation, but with fear, for he finally admitted it to himself: it was naked fear that had seized him, and in admitting it he grew calmer and his thoughts clearer. (210)
And here the German original is no cleft: (G2) Er ging zurück in sein Gemach, schweißnaß und bebend vor Aufregung, nein, nicht vor Aufregung, sondern vor Angst, jetzt endlich gestand er es sich ein, daß die schiere Angst ihn gepackt hatte, und indem er es sich eingestand, wurde er ruhiger und klarer im Kopf. (257f)
The question is: is the converging result in the English translations merely accidental, due to the translator’s whims, or is there anything regular behind it? To answer this question we can ask ourselves what would happen if we used an alternative version in either case. Let me begin with (2), the shorter sentence. If we compare the non-cleft version: (E2) a.
… for he finally admitted it to himself: naked fear had seized him,
with the cleft:
276 Monika Doherty
(E2) i.
… for he finally admitted it to himself: it was naked fear that had seized him,
the difference seems quite obvious: the first sentence is a neutral sentence. Except for the pronominal and thus inherently given element “him”, the whole sentence could contain new and hence focused information. The second version, however, restricts the focus to “naked fear”, presupposing the remaining part of the sentence, namely that there was something that had seized Richis. This is what we can read off from the sentences themselves. And when we look at the sentence-internal and sentence-external context, we can confirm the interpretation of (2): there had, indeed, been something that had seized Richis and caused him to behave strangely, even if he had for a long time not admitted it to himself what it was. In contrast to (2i), (2a) does not suggest a narrow focus on the subject, but it does not exclude it, either. And as all the contextual information is there whether we use the cleft or the non-cleft version, we know that something had seized Richis and that this something was fear whether we use the cleft or the non-cleft version. On the other hand, the sentence under discussion is embedded into an attitudinal matrix clause, and the attitudinal subject is not we, the readers, but Richis. It is Richis’s perspective that determines what is presupposed, and Richis knows that there is something wrong with him, but he does not admit to himself what it is. Nevertheless, knowing what we do about Richis’s state of mind, could we not interpret the non-cleft version in the same way as the cleft? The answer is: perhaps we could, but it would take us an extra effort because the linguistic form of the non-cleft does not suggest this reading. To show what this means, I will now try to spell out some of the linguistic assumptions about information structuring and focus identification in clefts and non-clefts, and look at the process of sentence interpretation from the psycholinguistic point of view of sentence processing.
3. Discourse relations of clefts Let me localize the difference between clefts and non-cleft versions in a bird’s eye view of sentence meaning. Sentence meaning consists of various layers, which could be said to form something like a hierarchy: elements of one layer are in the scope of elements of another layer. Roughly seen, it could be said that
Discourse theory and the translation of clefts between English and German 277
there is the propositional, referential meaning, subject to propositional attitudes, which evaluate the proposition in terms of truth and emotions, and then there is the illocutionary meaning, which allows us to use the evaluated propositional meaning in various speech acts.7 The most neutral speech act is that of informing, which is subject to certain information structuring conditions concerning topic/theme vs. comment/ rheme and focus vs. background. Clefts have traditionally been considered focusing elements, that is information structuring elements and hence the difference between the cleft and non-cleft version should pertain to this layer of sentence meaning. But what is the information structuring effect of clefts? Since Prince (l978) three properties have become generally accepted: first, it-clefts can have a focus in the matrix clause and/or in the subclause; secondly, the subclause carries a presupposition; and, thirdly, the copula expresses an exhaustively identifying, resp. specifying relation (cf. Prince 1978; Declerck 1988; Delin 1995 among others). If we apply these concepts to (E2) i.
… for he finally admitted it to himself: it was naked fear that had seized him,
we would say that “something had seized him” is presupposed and that the copula identifies what had seized him with “naked fear”. The identifying relation being exhaustive, any other possible explanation of what had seized him is excluded. The exclusion of alternative possibilities amounts to a contrastive focus interpretation of the identifying element. This interpretation fits in nicely with the contextual reading, as example (2) is preceded by the alternative idea of “agitation” (as a possible specification of what had seized him/made him tremble…), which is explicitly rejected: (E2) ii. He went back to his chamber, bathed in sweat and trembling with agitation, no, not with agitation, but with fear, …
But the use of “finally” indicates that there is also a contrastive relation that leads us far beyond the immediate, ‘local’ context, back to the comments following the first cleft: “He gave himself the excuse that he was out of sorts or overworked, but admitted as well that he was a bit concerned, as every father with a daughter of marriageable age is concerned, a thoroughly normal concern …” (209) When Richis in (2) finally admits to himself that the reason for his nervousness and nightmare is naked fear, fear clearly stands in a contrastive relation to everything he had thought before. This extends from agitation to normal
278 Monika Doherty
concern, to merely being overworked — all the excuses with which we got acquainted from the end of the last section, altogether three paragraphs earlier, up to the immediately preceding context. In fact, (1) itself stands in a contrastive relation to (2). Compare: (E1) i.
For a long time he would not admit that it was fear that …
with (E2) i.
… he finally admitted it to himself: it was naked fear that …
Thus, we have something like a close and a wide (sentence external) discourse contrast concerning the identifying element in the presupposition. As the presupposition in (2i), “something had seized him”, sums up all the preceding events revealing Richis’ sorrows, the presupposition, too, stands in a discourse relation (say: a resumptive relation) with a larger segment of the sentence external context. Let me call the discourse relations covering larger segments of the context ‘macrostructural discourse relations’. If we study the non-cleft version in this context, we should be able to reach the same interpretation, but this would then only be due to the context. In the non-cleft version itself, there is nothing that indicates that there is a macrostructural discourse relation associated with the contrastive focus on “fear”. In fact, the non-cleft version itself does not even restrict the focus interpretation to the subject. Only when we merge the interpretation of the sentence with the immediate context can we realize that there exists a contrastive discourse relation which requires us to restrict the focus to the subject. Thus, the difference between the cleft and the non-cleft version is simply that the cleft tells us right away that the identifying element stands in a macrostructural contrastive discourse relation to the preceding context, and the noncleft version does not do this. In other words, the cleft with its presupposition and exhaustively identifying relation is a formal indicator of a discourse relation between its identifying element and some larger segment of the preceding discourse. It is this formal indicator that is missing in the non-cleft version. As the context will always supply the relevant pieces of information to figure out the discourse relation without a linguistic indicator, the difference between the cleft and non-cleft versions may seem to be negligible. But if we look at the difference from the perspective of language processing we will arrive at another conclusion.
Discourse theory and the translation of clefts between English and German 279
4. Processing clefts and non-cleft versions There are various approaches to language processing, which — taken to their extremes — may even exclude each other. But the idea that processing takes place in various specialized components and subcomponents (modules and submodules) of our ‘processor’, corresponding to the various levels of the linguistic structures processed, is common to all approaches. What it is that these components are specialized on, which principles of processing are at work in which component, and how the various components interact with each other and with other cognitive capacities outside the linguistic capacity, have been hotly debated issues — for about thirty years of intensive research. (For a more recent, comprehensive survey on questions of interpretive processing, cf. Frazier 1999, esp. Chapter 1.) To cut a long story short, let me spell out my basic assumptions. I have opted for this partially modular, incremental approach (along the lines of Crocker 1996) because it can explain many of our intuitive assessments of translational (in)adequacies as (temporary) garden paths, which require reanalysis to match their (wider) context. If the processor were aware of the contextual conditions all the time, it would not miss the contextually adequate interpretation in its first-pass analysis — and it should do equally well without any processing ‘aids’, such as clefts, topicalizations and the like. I am quite aware that the following assumptions are too simplistic and that most questions are left open this way, but it should become clear that the principal line is compositional, that is, interpretation proceeds stepwise, computing the information bit by bit from the linguistic structure before it is integrated with the information stored from previous interpretations. 1.Sentences are first processed according to their linguistic structures and only afterwards embedded into the discourse model we have created from the preceding context. That is, as a rule, sentence internal processing takes place without recourse to the sentence external context.8 2.Interpretation of the sentence structure proceeds incrementally, from the beginning of a sentence to its end. The size of the increments is determined by our syntactic knowledge about grammatical categories, function words, valency (argument structure) and the like.9 In a simplified version we could equate increments with words.
280 Monika Doherty
3.Interpretation sets in an increment or so after the syntactic analysis (that is after attachment of the incoming increment to the current phrase structure); the output is interpreted on the basis of linguistic and non-linguistic knowledge.10 4.If the first-past analysis (including the focus structure of the sentence) turns out to be wrong, ending at some point in incompatible results (for example in a mismatch with the discourse relations of the context), the structure will be reanalyzed. 5.Processing of language is controlled by the principle of economy, which makes us prefer linguistic forms securing optimal conditions for maximal cognitive gains and a minimum of processing efforts.11 Let me call the fifth assumption the principle of optimal processing conditions, PC. It requires us to avoid structures which have to be reanalyzed. If we apply PC to (2) and (2a), the examples under discussion, we can see that it is met by the cleft and violated by the non-cleft version. While the cleft-version (2) signals a contrastive interpretation by the very meaning of the cleft-structure, the noncleft version (2a) will be attributed a normal, non-contrastive focus interpretation. In line with the normal word order of (2a) and the (in)definite character of its two arguments, the processor will identify the subject, “naked fear”, as a normal, projecting focus, including the entire clause except for the pronoun: (E2) a.
… that [naked fear had seized] him. (small caps symbolize the focus exponent, square brackets the focus projection)
This has to be corrected, i.e. reanalyzed afterwards, when the result of the first processing phase is embedded into the context, resp. when the translation is matched against the original, which does have a contrastive reading (see below). The contrast concerns a larger passage of the text (reaching three paragraphs back), and is — as I said in the previous section — a macrostructural discourse relation; hence the discourse relevance of the contrast is quite high. But as the relation to a distant antecedent will be more difficult to discover if nothing reminds us of it, the non-cleft version could make us miss the macrostructural contrast altogether. If we need an extra effort to reanalyze the sentence and compute the contrastive interpretation merely on the basis of what we remember from the context, i.e. on the basis of the cognitive model available to us at a certain point of processing, this would clearly have to be considered a processing disadvantage, to be avoided if possible. But if we fail to notice the contrastive relation at
Discourse theory and the translation of clefts between English and German 281
all, this would even amount to an illicit deviation from the original, which would not be justified. Unless we have a special reason not to do so, we should retain the information of the original — as long as this is possible at a reasonable processing ‘price’. And possible it is in this case: if we use a cleft, the matrix sentence signals a contrastive focus right away. Of course, there is the price of a bit of an extra processing effort, as we have to work our way through the additional elements of the matrix clause and the element connecting the two clauses. But these are all functional words with little or no meaning beyond their grammatical functions, which means that their analysis is restricted to the first and most shallow level of processing and does not require any of the more complex types of effort. As a result of this first, shallow phase of processing, we will have achieved something like a constituent structure where the predicative of the copula, “naked fear”, is marked by a contrastive focus feature (for a nonprojecting, narrow focus): (E2i) it was [naked fear] that had seized him.
and the background is interpreted as presupposed, which amounts to something like: Presupposition: $x (x had seized him) Assertion: (x = naked fear)12. So far, so good. In English. If we look back at the German original we find a non-cleft version, suggesting either that the translation is better than the original or that the conditions for optimal processing differ between English and German. As it is the second explanation that will be more likely, we will now take a closer look at the German cleft- and non-cleft structures.
5. Marked word order in German There is no question that we could use a cleft in the German context, too: (G2) a.
… jetzt endlich gestand er es sich ein, daß es die schiere Angst war, die ihn gepackt hatte.
But the original (G2) i.
jetzt endlich gestand er es sich ein, daß die schiere Angst ihn gepackt hatte.
282 Monika Doherty
is better. To begin with, German is quite allergic against repetitions at short distance, and the expletive es in the matrix clause of the cleft, which follows another expletive es in the dominating clause would be reason enough to search for another structural possibility. But we could drop the first es, or cliticize it, and would still consider the cleft less good than the non-cleft version: (G2) c.
… jetzt endlich gestand er’s sich ein, daß es die schiere Angst war, die ihn gepackt hatte.
(G2) d. …; jetzt endlich gestand er’s sich ein, daß die schiere Angst ihn gepackt hatte.
The non-cleft version seems to be sufficient, the cleft unnecessarily clumsy, simply redundant. But if we appreciate the explicit indicator of a discourse relevant contrastive focus in the English version, why should we consider it redundant in German? The answer is that the German non-cleft structure does already have a formal indicator for the contrastive focus. The device is, however, far less visible than the cleft; to be able to discern it, we have to take a look at the specific information structuring conditions of German. English word order is mainly controlled by grammatical rules paying tribute to a configurational language, which identifies grammatical relations by the position of phrases relative to the verb (basically subject before, object after the verb). This amounts to a relatively rigid word order, with little variations from the basic word order, which is determined by the conceptual relations between the verb and its partners. As far as lexical projection is concerned, the basic word order of English and German is similar, except for the grammatical parameter of directionality, which extends the verb phrase to the right in English and to the left in German. But German is not restricted to the basic word order; on the contrary, German word order is controlled by information structuring aspects, which requires background information to be presented before focused material. If this requirement is not met, the resulting structure will be interpreted as marked, i.e. the constituent that turns up in the ‘wrong’ place will be associated with a narrow contrastive focus.13 This is the case in the German original, the non-cleft version of (G2i). The pronominal object ihn is a given element by its very nature, while the subject schiere Angst is a focusable element. Thus the neutral order of things would require us to present the object before the subject:
Discourse theory and the translation of clefts between English and German 283
(G2) e.
daß ihn die schiere Angst gepackt hatte
securing a normal focus interpretation: (G2) e.
daß ihn [die schiere Angst gepackt hatte].
But the original sentence presents subject and object in the order violating the inherent information structural requirements (G2) i.
daß die schiere Angst ihn gepackt hatte.
and thus assigns a contrastively marked focus to the subject: (G2) i.
daß die [schiere Angst] ihn gepackt hatte.
As a contrastive focus structures its proposition very much like a cleft into a background/presupposition and an identifying relation excluding alternatives, the interpretation of the German original is roughly equivalent to the English cleft. And as the marked word order is a more economical device than the cleft, PC makes us prefer the marked order to the cleft. (G2i) expresses a contrastive discourse relation, but — missing the presupposition of the cleft — the marked word order does not indicate the macrostructural relevance of the contrast. Even so, the author’s choice suggests that PC makes us prefer the more economical device to the more ‘explicit’ form and rely on inferences from the context instead. But the English translator does not have this choice. Due to its configurational restrictions, English cannot use the economical word order device; the neutral (E2) a.
… naked fear had seized him
is the only grammatically acceptable order. Hence the cleft.
6. One focus or more The first example, (1), as you will remember, had a cleft in both languages. If we decleft the English version (E1) For a long time he would not admit that it was fear that caused him to delay trips that ought to have been made some time ago, or to be reluctant merely to leave the house, or to break off visits and meetings just so that he could quickly return home.
284 Monika Doherty
into (E1) a.
For a long time he would not admit that fear caused him to delay trips that ought to have been made some time ago …
we get the same difference as between (E2) and (E2a). (E1a) is a neutral sentence, where everything (except for the inherently given “him”) could be new and hence focused information: (E1) a.
… [fear caused him to delay trips that …]
If we look into the context, where the last clause was: “fear crept into Antoine Richis’s heart like a foul poison”, we have to reanalyze “fear” as background. With (E1i) things are different as the cleft assigns a focus to “fear”. But this time the focused subject is not the only focus we can make out in the cleft. The predicate of the subclause is followed by something like three complex, albeit reduced clauses. Each of these non-finite complements contains focusable material, and if we look at the context, we know that the complements consist of new information throughout. Thus (E1i) is an ‘informative presupposition cleft’, as Prince (1978) has called them. (E1) i.
it was [fear] that [caused him to delay trips that …]
We have a focused complement also in (E1a), the non-cleft version. But (E1a) does not assign a narrow focus to the subject. And contextual reinterpretation is more difficult for (E1a) than for (E2a): unlike (E2i), (E1i) does form a macrostructural contrast with earlier sentences, but it does not participate in a contrastive discourse relation with its immediate context. In (E2), one of the possible alternatives to “fear”, viz. “agitation”, was explicitly introduced, rejected and corrected in the sentence preceding the colon. (Colons and semicolons may be taken to signal autonomous intonational units and hence as separating sentences.14) But such a ‘sentence-external’ context can still be considered ‘immediate’ compared with the wider context needed for the contrastive discourse relation of (1), which goes back three sentences to a passage that depicted Richis as “one of the few people in the town who were immune to the fever of fear and kept a cool head”. The change from someone immune to fear to someone whom fear caused to alter his normal behavior noticeably is what the contrastive focus on “fear” is about. It is a turning point in the story, definitely worth an extra focus. But without a formal indicator, as in (E1a), we would certainly have to correct our
Discourse theory and the translation of clefts between English and German 285
first-pass analysis, and if our memory fails us, we would not even be able to recognize the macrostructural importance of “fear” at this point.
7. The relativity of focus interpretation If we decleft the original (G1) i.
Lange Zeit wollte er sich’s nicht zugeben, daß es die Angst war, die ihn bewog, längst fällige Reisen hinauszuzögern … (256)
into a non-cleft version: (G1) a.
Lange Zeit wollte er sich’s nicht zugeben, daß die Angst ihn bewog, längst fällige Reisen hinauszuzögern …
we could opt for the same marked word order with the subject before the pronominal object. This was itself a sufficient device in (2) to mark the subject as a contrastively focused element. But if we compare the marked order with the neutral order in: (G1) a.
daß die Angst ihn bewog, längst fällige Reisen hinauszuzögern …
(G1) b. daß ihn die Angst bewog, längst fällige Reisen hinauszuzögern …
the difference between the two interpretations is not the same as the one we had encountered in (2): (G2) e.
daß ihn die schiere Angst gepackt hatte.
(G2) i.
daß die schiere Angst ihn gepackt hatte.
Clearly, the presence of the long infinitival complement in (1) creates a different condition. The syntactic properties of the verb bewog differ from those of gepackt hatte. While the latter has only two arguments, both of which have already been processed by the time we reach the verbal complex, jemand zu etwas bewegen, has three arguments, and the prepositional object/infinitival complement is still ahead when the verb is being processed. Thus bewog makes us expect a focusable complement after it; when we read on, this expectation is met by a whole series of foci. Even if the marked order of subject and pronominal object did signal a contrastive focus for the subject, this focus seems to disappear in the presence of all the foci after the verb. There is a good chance that we will miss the contrastive focus on the subject:
286 Monika Doherty
(G1) a.
… daß die Angst ihn [bewog, längst fällige Reisen hinauszuzögern …]
Compared with (G2i), the ‘downtoning’ is all the more noticeable as the subject in (G1a) is merely a definite noun, die Angst, unlike the subject in (G2i), which is reinforced by an intensifying adjective, die schiere Angst. However if we compare the non-cleft version with the cleft: (G1) b. daß die Angst ihn bewog … (G1) i.
daß es die Angst war, die ihn bewog …
we can see that the relevance of the subject increases noticeably. The subject is assigned the main focus of all foci. Being relativized by the series of foci after it, the marked word order of (G1b) is too weak to indicate the contrastive focus of the subject (let alone the macrostructural significance of the contrast); thus, German also requires a cleft.
8. Theoretical implications Comparing the original versions and their translations with the alternative possibilities, in terms of clefts and non-cleft versions, we have found that the published versions provide optimal processing conditions for focus identification. To explain this intuition, we have had to make a number of hypotheses, which should be applicable to a wider set of empirical data and consistent with theoretical assumptions held elsewhere. The two criteria are to a certain extent interdependent, as the theoretical assumptions held elsewhere have been developed on the basis of other empirical evidence. Since a critical assessment requires a systematic presentation of the assumptions, let me repeat the claims I have made in discussing the examples and their variations. Clefts are means to focus the identifying elements of their matrix clauses; they presuppose their subclauses, which may or may not contain focused material. Due to the presupposition, the focus on the identifying element of the matrix cleft participates in a contrastive discourse relation which is of macrostructural relevance, that is, it is a discourse relation that involves more than the immediate propositional context. The macrostructural contrast is part of the conventionalized meaning of the cleft. In German, the focusing function of the cleft (but not its macrostructural relevance) can be achieved compositionally by marked word order — provided the word order marking is not ‘flouted’ by other focusing means. The reordering device is preferred to the cleft according to PC.
Discourse theory and the translation of clefts between English and German 287
Now the question is, does this analysis hold for other examples as well and is it compatible with other assumptions about clefts, focus, discourse relations and language processing? As there is no space for a proper presentation of other examples, let alone their detailed analysis, I can now only declare that all clefts I have been studying in the last two years in connection with my research project on the translation of clefts participate in a macrostructural discourse relation.15 Many of the clefts participated also in a contrast with the immediate context, often a sentenceinternal contrast. Compare, for example, the following cases, where the identifying element in the matrix clause is negated and followed by its correction directly: (E3) he suspected that it was not he who followed the scent, but the scent that had captured him and was drawing him irresistibly to it. (42)
or (E4) Grenouille found his heart pounding, and he knew that it was not the exertion of running that had set it pounding, but rather his excited helplessness in the presence of this scent. (41)
The sentence internal contrast has the effect of downtoning the focus on the identifying element in the first matrix clause in favor of the identifying element in the second, the elliptical clause: “exertion of running/excited helplessness”. We could consider this type of cleft as yet another case of focus relativization, associated with the special meaning of the connecting form: “not … but”. It seems to restrict the contrastive relation to a macrostructural, sentence internal contrast.16 But if we look into the wider context of the example, we can identify the macrostructural relevance in the correcting, adversative part after sondern/but and its discourse relation to the preceding description of Grenouille’s pastime. Though both sentences follow each other at the distance of a page, they express basically the same idea: Grenouille is overwhelmed by the scent that makes him follow it with pounding heart, driving him into his first murder. What constitutes the macrostructural contrast is the shift from the active hunter of scents, strolling through Paris to register indiscriminately any scent the world offers, that is from a person acting by free will to a person driven by passion, by the passion for one particular scent. It may be interesting to note that the two clefts correspond, again, to a cleft and a non-cleft original:
288 Monika Doherty
(G4) Grenouille spürte, wie sein Herz pochte, und er wußte, daß es nicht die Anstrengung des Laufens war, die es pochen machte, sondern seine erregte Hilflosigkeit vor der Gegenwart dieses Geruchs. (52)
and (G3) … denn er ahnte, daß nicht er dem Duft folgte, sondern daß der Duft ihn gefangengenommen hatte und nun unwiderstehlich zu sich zog.
The use of these cleft and non-cleft versions in English and German can be explained along the lines developed for the examples under (1) and (2). While the marked order: nicht er suffices to indicate the contrast, nicht die Anstrengung des Laufens machte es pochen is burdened with a number of processing difficulties arising already at the syntactic level. I cannot pursue the analysis any further here, but it is clear that PC will make us prefer a cleft or a non-cleft version not only in the interest of optimal focus-identification but also in the interest of any other aspect of sentence interpretation. Now, even if there were clefts that did not participate in macrostructural contrasts, the two macrostructural types described above are representative of many uses of clefts. Thus it should be worthwhile to extend the theoretical inventory of focus interpretation onto the presentation of macrostructural discourse relations. Asher’s theory on SDRSs (segmented discourse representations) could serve as a theoretical starting point. It allows us to embed macrostructural discourse relations into larger segments and interpret the representations in model-theoretic terms. Asher himself provides a token-reflexive on the semantic implications of clefts: (5) It is the embedding of DRSs, small, partial models of what is being talked about, into large, total models that gives us truth conditions (Asher 1993: 65)
Asher’s cleft/theory places the immediate contrast between DRS inscriptions and DRS representations-by-embedding into the macrostructural context of a more comprehensive theory of verbal communication (Asher 1993: 64). If clefts indicate macrostructural discourse relations, the focus structure of the cleft has to be interpreted accordingly. That is, in the case of the contrastive relation, we have to construe the alternative(s) to the identifying element from the wider context so that it/they fit(s) into the presupposition of the cleft (or a related open proposition subsuming the presupposition and relevant ideas of the context). The result could be more or less straightforward, as in the examples
Discourse theory and the translation of clefts between English and German 289
topicalizing “fear”, or it could be more indirect as in the examples with the sentence-internal contrast. In any case, we can assume that clefts function somehow like an instruction for information packaging at the macrostructural level, bracketing a larger segment of the discourse. Although we are basically free to discover any macrostructural discourse relation also without explicit indicators, we may easily fail to do so because of our limited storing capacities in general and because of the extra effort we have to invest into the reanalysis of our primary reading of non-cleft versions, in particular. Reanalysis involving focus structures requires prosodic reanalysis, which amounts to an extra effort as M. Bader (1998) has shown convincingly. If we assume with Sperber and Wilson (1986) that our processing strategies are guided by the ‘presumption of optimal relevance’, which makes us expect a maximum of cognitive gains for a minimum of effort, it is more likely than not that the reanalysis will not take place at all. That is, a linguistic cue like the cleft is no less in our own interest than in the interest of our readers. But clefts belong to a set of complementary focusing devices which are used differently in different languages. Marked word order and focusing particles are among those linguistic forms competing with clefts. Yet even in German, where reordering and particles have fewer constraints than in English, we will use clefts if the more economical devices are burdened by more processing disadvantages than the clefts. Specifying optimal processing conditions for clefts, marked word order, focusing particles and the like is — as the discussion of the examples has shown — a real challenge and the closer we get to the theoretical details, the more questions will have to be answered. But something along the lines I have sketched, linking focus interpretation, discourse relations and processing ease, will certainly prove useful.
Notes 1. In a sample of 60 examples taken from one novel (Auslöschung/Extinction by Thomas Bernhard, henceforth: B), 19 sentences, i.e. almost a third of the clefts in the English translation did not contain any formal substitute in the original, and altogether only 14, i.e. not even a quarter of the English clefts had had clefts in the German original. The figures and proportions varied from corpus to corpus, but the basic patterns were the same in all the translational corpora we looked at in the research project.
290 Monika Doherty
2. P. Erdmann (1990) was the first to compare the use of German and English clefts on the basis of translations. Comprising almost 500 examples, his study arrives at different figures and proportions than ours, but the patterns of translation are basically the same. Erdmann looks at the examples in isolation, but explains the different use of German and English very much along the same lines as we do — albeit in a purely contrastive way where none of the structural and contextual details of the different focus structures are spelled out. 3. Cf. Doherty (1991) for a detailed discussion of this example. Cleft-like structures like these will be dealt with systematically in a special issue of Linguistics (forthcoming). 4. For a detailed presentation of the differences between topicalization and verb phrase internal reordering (scrambling) in German compare Rosengren (1993). 5. Erdmann (1990) finds a majority of it-clefts translated analogously into German in the case of focused subjects and concludes that the clefts in this case compensate the natural constraints on subject-topicalization. Though I would subscribe to his explanation, I cannot confirm the data as subject-clefts in English make up only about 10% of all my clefts and most of them are declefted in the German versions. 6. Reference to German/English pages in source. References to the German and English examples from Süskind (below) follow quotations directly. 7. I am using the traditional term: proposition for the sake of simplicity because it is easier to handle than ‘attitude objects’ — although I would subscribe to Asher’s idea (1993) of relativizing such cognitive objects to their individual bearers. 8. Instead of sentences, I should probably speak of intonational phrases, characterized by one contour, breaks, boundary tones and the like. Intonational phrases could also be parts of sentences, as for example non-defining relative clauses, preposed adverbials etc. But as they participate in one sentence, they will in the end be interpreted as part of one informational unit, too. 9. Syntactic knowledge allows us to unfold a part of the structural hierarchy, into which we can then fill the subsequent input. 10. The critical point here is that some aspects of the semantic structure, in terms of semantic roles, coreference and focus features, for example, are already part of the syntactic information. 11. Sperber and Wilson’s (1986) Principle of Optimal Relevance as a universal of language processing. 12. In fact, x ought to be an exhaustive list of things, X, with the property, P, assigned to them by the presupposition; cf. Rooth (1999), 234f. 13. For a classical, concise description of the set of factors determining basic and derived word order in general, cf. Zimmermann (1999), 43f. 14. Cf. Fabricius-Hansen (1999) for a detailed presentation of incremental discourse organization in terms of sentences. 15. I have presented a first in-depth analysis (Doherty 1999) on the translation of English clefts into German, which yields similar results.
Discourse theory and the translation of clefts between English and German 291
16. Actually, the lexically explicit contrastive clefts could be considered a mere variant of the purely affirmative clefts with their contrastive implications. Both types together belong to the class of alternative clefts, which are opposed to wh-clefts very much like alternative questions to wh-questions. This idea is certainly worth while being spelled out in detail, but such an endeavor would go far beyond the scope of this paper.
References Asher, Nicholas 1993 Reference to Abstract Objects in Discourse. Dordrecht, Boston, London: Kluwer. Bader, Markus 1998 “Prosodic Influences on Reading Syntactically Ambiguous Sentences”. In Reanalysis in Sentence Processing, J. D. Fodor, F. Ferreira (eds), 1–46. Dordrecht, Boston, London: Kluwer. Crocker, Matthew W. 1996 Computational Psycholinguistics: An Interdisciplinary Approach to the Study of Language. Dordrecht, Boston, London: Kluwer. Declerck, Renaat 1988 Studies on Copular Sentences, Clefts and Pseudo-clefts. Leuven: Leuven University Press. Delin, Judy 1995 “Presuppositions and shared knowledge in it-clefts”. Language and Cognitive Processes 10 (2): 97–120. Doherty, Monika 1991 “Informationelle Holzwege”. Zeitschrift für Literaturwissenschaft und Linguistik 84: 30–49. 1999 “Clefts in translations between English and German”. Target 11 (2): 289–315. Erdmann, Peter 1990 “Fokuskonstruktionen im Deutschen und Englischen”. In Kontrastive Linguistik [Forum Angewandte Linguistik 19], C. Gnutzmann (ed.), 69–83. Frankfurt am Main: Lang. Fabricius-Hansen, Cathrine 1999 “Information packaging and translation: aspects of translational sentence splitting (German-English/Norwegian)”. Studia Grammatica 47: 175–214. Frazier, Lyn 1999 On Sentence Interpretation. Dordrecht, Boston, London: Kluwer. Prince, Ellen 1978 “A comparison of WH-clefts and it-clefts in discourse”. Language 54: 883–906. Rooth, Mats 1999 “Association with Focus or Association with Presupposition?”. In Focus. Peter Bosch and Rob van der Sandt (eds), 232–246. Cambridge: Cambridge University Press.
292 Monika Doherty
Rosengren, Inger 1993 “Wahlfreiheit mit Konsequenzen — Scrambling, Topikalisierung und FHG im Dienste der Informationsstrukturierung”. In Wortstellung und Informationsstruktur [Linguistische Arbeiten 306], Marga Reis (ed.), 251–312. Tübingen: Niemeyer. Sperber, Dan and Wilson, Deirdre 1986 Relevance. Oxford: Blackwell. Zimmermann, Ilse 1999 “Die Integration topikalischer DPs in die syntaktische und semantische Struktur von Sätzen”. Studia Grammatica 47: 41–60. Examples from: Bernhard, Thomas 1989 Auslöschung. Ein Zerfall. Berlin: Volk und Welt. 1995 Extinction. London: Penguin. Süskind, Patrick 1994 Das Parfum. Zürich: Diogenes. 1987 Perfume. London: Penguin.
AUTHOR "Eva Haji cˇová, Jarmila Panevová and Petr Sgall"
TITLE "Tectogrammatics in corpus tagging"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Tectogrammatics in corpus tagging Eva Hajicˇová, Jarmila Panevová and Petr Sgall Charles University, Prague
A semi-automatic syntactic annotation of a part of the Czech National Corpus in the Prague Dependency Treebank is being carried out, based on the Praguian Functional Generative Description, the core of which is a dependency based account of underlying sentence structures. The first phase of the tagging procedure (see Hajicˇ 1998) consists of morphemic and “surface” annotations, during which the intermediate ‘analytic level’ (AL) is achieved; the analytic tree structures (ATSs) contain a node for every token of a word, and even of a punctuation mark, as is often the case in tagging procedures. The aim of the present paper is to characterize the second phase, i.e. the main step of syntactic tagging, the procedure prepared for transducing from AL to syntax itself, i.e. to underlying, tectogrammatical representations (TRs, which are technically modified for the given purpose, as will be seen in the sequel).
1.
Main differences
Dependency trees are present both on AL and on the level of TRs. However, in the TRs only the nodes corresponding to lexical (autosemantic) units are present; function words (or, more exactly, their functions) are represented by indices of the lexical labels, i.e. by syntactic functors and by grammatemes (which mark values of tense, aspect, modalities, number, and of other grammatical categories). While TRs are underlying structures (basically appropriate to serve as input to semantic interpretation, see Sgall et al. 1986; Sgall 1992) and distinguish at least about 40 kinds of syntactic relations (classified in the valency grids
294 Eva Hajicˇová, Jarmila Panevová and Petr Sgall
included in the lexical entries of the head words as arguments or adjuncts, and obligatory or optional, see Panevová 1974, 1998), in ATSs syntactic relations are classified only on a “surface” layer, and without more subtle differences, such as those between types of objects or of adverbials. One aspect of the TRs is their topic–focus articulation with a scale of communicative dynamism, represented as underlying word order; e.g. an adjective is prototypically more dynamic than its head, even if preceding it in the surface, i.e. in the word order of the morphemic representation (a string without parentheses), cf. malý ‘small’ in (1); see Sgall (1967), Hajicˇová (1984, 1993).
2. Coordinating constructions For technical reasons, in tagging we use nodes for coordinating conjunctions (as heads of the coordinated items), although this does not exactly correspond to the theoretical specification of TRs (a formal treatment of TRs including all combinations of dependency and coordination and based on the detailed specification of the linguistic approach in Sgall et al. 1986 was presented by Petkevicˇ 1995). Therefore we distinguish between TRs proper and Tectogrammatical Tree Structures (TGTSs), see Hajicˇová (1998); cf. Fig. 1, i.e. a (highly simplified) underlying tree for example (1). (1) Marie a Jan, kterˇí mají malého syna, žijí v Lomnici. Mary and John who have small son live in Lomnice ‘Mary and John, who have a small son, live in Lomnice.’ z¦ ít a.conj Marie.act.coref
Lomnice.loc
Jan.act.coref mít.descr který.act
syn.pat malý.rstr
Figure 1.A highly simplified TGTS of (1), with functors attached to dependent nodes (Conjunction, Actor/Bearer, Patient, Descriptive adjunct).
Tectogrammatics in corpus tagging 295
3. Linearized underlying representations TRs can be unambiguously linearized; e.g. the primary TR of (1) can be written as (1¢), with each dependent item and each coordinated construction closed into parentheses; the subscripts (at the parenthesis oriented to the head word) indicate functors: (1¢) ((Marie Jan)Conj (Descr (který.Plur)Actor mít (Obj syn.Plur (Restr malý)))) Actor žít (Loc.in Lomnice)
Unmarked grammatemes (Sing, Pres, Declar, etc.) are not written here. A further example: (2) Iniciátorˇi dosud nesehnali potrˇebných trˇicet podpisu˚ initiators hitherto have.not.gathered necessary thirty signatures poslancu˚. of.MPs
‘The initiators have not yet gathered the necessary thirty signatures of MPs.’ (2¢) ((Iniciátor.Plur (Pat on))Act (dosud)Temp.on (Neg)Rhem sehnat.Pret (Pat podpis.Plur (Appurt poslanec.Plur) (Restr trˇicet) (Restr potrˇebných))
Note that such a deverbal noun as iniciátor has an obligatory Patient. With cases of coreference (anaphora) the data on the antecedent are registered in the label of the coreferential node (see Section 4 (2c) below).
4. The automatic part of the transduction of ATSs to TGTSs A part of the transduction procedure (see Hajicˇová 1998) can be formulated as general steps, carried out automatically: 1.In an automatic ‘pre-processing’ module, the input of which are the ATSs, the tree structures are pruned, i.e. the nodes that are marked as auxiliary items in the ATSs get deleted, without losing any important pieces of information these auxiliary items carry. During this pre-processing, most of the analytical morphemic forms are put together (being placed in the position of the ‘highest’ of their parts), and the information they convey is added in the form of indices
296 Eva Hajicˇová, Jarmila Panevová and Petr Sgall
(esp. grammatemes) of the TGTS complex tags. This concerns the values of morphological categories such as tense (Preterite, Future), verbal modality (Conditional), deontic modality (with musí ‘must’, mu˚že ‘can, may’ and other modal verbs), diathesis, etc. and aspect, or gender and number with nouns, and degrees of comparison with adjectives and adverbs; they get their values on the basis of their morphemic tags (some asymmetries between forms and their respective functions will be solved later, during the manual procedure). The grammateme of sentential modality (with the values enunc, interr, imp, desid) is specified automatically with all heads of main clauses on the basis of the node standing for the final sentence boundary and of other data (esp. particles) present in the analytical tree. The analytical function Subject with an active verb is converted into the tectogrammatical functor act (Actor/Bearer). The analytical function AuxR denoting the particle of reflexive passive is converted into a node with the lexical value General and the functor act. 2.After the ‘manual’ handling of TGTSs (see Section 5 below), another automatic module is being prepared, which will serve to add information that can be ‘retrieved’ automatically now in the preliminary version of TGTSs: a.the gender and number values are cancelled with word tokens with which they only indicate agreement (adjectives in most positions, certain pronouns, numerals, etc.); thus, an adjective retains its gender value only if it does not depend on a noun (e.g. a superlative); b.the sentence modality value with ‘content’ clauses (indirect speech and similar) is added into the respective grammateme of the head verbs of these clauses in accordance with the conjunction present, e.g. enunc (že ‘that’), imp (atˇ, nechtˇ, aby ‘that’), inter (zda ‘whether’ and other interrogative words); c.certain additions are carried out which can be specified in this phase of the procedure, e.g.: –
–
the lemma of the node carrying the functor value act is assigned to the grammateme coref of an occurrence of se that has not yet been treated (i.e. the pat of an active verb in the prototypical case); the remaining nodes without lemmas (in coordinated constructions or in apposition) are assigned the lemmas of their counterparts in the given construction; e.g. in (3) Jirka pozval Marii a Karel Milenu. Jirka invited Mary and Karel Milena ‘Jirka invited Mary and Karel (invited) Milena.’
Tectogrammatics in corpus tagging 297
the node corresponding to the deleted second occurrence of the verb (which has been added “by hand” as governing both Karel.act and Milenu.pat) gets a lemma identical to that of the lefthand-side coordinated item; –
–
the secondary values of syntactic grammatemes (cf. Section 5 below) are added there, where a preposition allows for a reliable choice: accompaniment.without (bez ‘without’), benefactive.neg (proti ‘against’), dir3.in (do ‘into’), etc.; the remaining nodes corresponding to commas, dashes, quotes, etc. are deleted.
In the next stages, the automatic procedure is supposed to be enriched in various respects, to cover at least the most regular phenomena of subdomains such as: –
–
–
word derivation (up to now only the deverbal adjectives, possessive adjectives and pronouns, and adverbs derived from adjectives are handled on the basis of the lemmas of the source words), certain elementary ingredients of the build-up of the lexicon, which should contain several kinds of grammatical data especially including the valency frames or grids, the development of the degrees of activation of the ‘stock of shared knowledge’ (see Hajicˇová 1993) as far as derivable from the use of nouns in subsequent utterances in a discourse.
5. Intellectual part of underlying tagging The following operations can only be performed intellectually (before further analysis helps to find reliable criteria to identify specific contexts in which secondary functions occur): 1.the analytic functions (such as Subject, Object, Adverbial, Attribute), expressed by case endings, subordinating conjunctions and prepositions, are changed into corresponding functors; e.g. Dative with an AL object primarily yields addressee, with an adverbial it yields benefactive, Cz. aby ‘that, to’ or na ‘on’ yields Objective with AL objects and aim or loc, respectively, with adverbials; the syntactic grammatemes accompanying loc (corresponding to the primary functions of prepositions such as v ‘in’, na ‘on’, pod ‘under’, mezi ‘between’, and so on) are left for further treatment (the original preposition is retained as the value of a specific attribute in the complex symbol of the noun; cf. Section 4 (2c) as for the subsequent automatic step);
298 Eva Hajicˇová, Jarmila Panevová and Petr Sgall
2.nodes for the deleted items are ‘restored’ either as pronouns (including specific symbols for a ‘General Participant’, for a ‘Controllee’ and for an ‘Empty Verb’ (with the non-verbal heads of sentences that are neither Vocatives, nor pure denominations, such as nominal headings)) or the attribute ‘lemma’ is left vacant for further treatment (e.g. in coordinations, see Section 4 (2) (c) above); 3.the topic–focus articulation of the sentence is accounted for by means of three values of the corresponding attribute, namely f for ‘focus’ (more exactly: contextually non-bound), t for non-contrastive (part of) topic (contextually bound) and c for ‘contrastive’ (part of) topic; 4.with possessive adjectives and pronouns dependent on nouns, the number and gender values of their basis are taken as the values of their respective grammatemes: – – – – –
jeho ‘his’ gets the values sing, animate (or inanimate or neuter, according to the context, i.e. to the gender of the antecedent, její ‘her’ gets sing, feminine, jejich gets pl and the appropriate gender, mu˚j ‘my’ gets sing and either anim or fem, matcˇin ‘mother’s’ gets sing, fem, and so on.
The annotators use the help of a ‘user-friendly’ software that enables them to work with diagrammatic shapes of trees.
6. Concluding remarks Almost 100000 sentences from the Czech National Corpus have obtained their ‘analytical’ annotations, and we expect to get about 1000 sentences annotated by their TRs before the end of the year 2000. Neither the automatic nor the manual part of the tagging can achieve a complete formulation of tectogrammatical representations. Several types of grammatical information will be specified only after further empirical investigations. Thus, e.g., the disambiguation of the functions of prepositions and conjunctions can only be completed after lists of nouns and verbs with specific syntactic properties are established. However, the annotated corpus will offer a suitable starting point for monographic analysis of the problem concerned. Whenever possible, also statistical methods will be used; specific combined procedures are being tested, based on statistical and structural approaches.
Tectogrammatics in corpus tagging 299
In this way a theoretically substantiated labeling of the TRs can be gained, distinguishing between different kinds of objects and adverbials, between meanings of function morphemes, topic and focus, and so on. The result will be much more complex than that of a parser or tagger of the usual kinds: not only the grammatical well-formedness will be checked, but disambiguated representations of sentences will be achieved, which (although underspecified in the points in which the sentence structure is not fully specific — indistinctness, “systematic ambiguity”, scopes of quantifiers) would constitute an appropriate input for a procedure of semantic-(pragmatic) interpretation.
References Hajicˇ, Jan 1998
“Building a syntactically annotated corpus: The Prague Dependency Treebank”. In Issues of Valency and Meaning. Studies in Honour of Jarmila Panevová, E. Hajicˇová (ed.), 106–132. Prague: Karolinum.
Hajicˇová, Eva 1984 “Presupposition and allegation revisited”. Journal of Pragmatics 8: 155–167; amplified in Contributions to functional syntax, semantics and language comprehension, P. Sgall (1984) (ed.), 99–122. Amsterdam: Benjamins — Prague: Academia. 1993 Issues of sentence structure and discourse patterns. Prague: Charles University. 1998 “Prague Dependency Treebank: From analytic to tectogrammatical annotations”. In Text, speech, dialogue. Proceedings of the Conference TSD 98, P. Sojka et al. (eds), 45–50. Brno: Masaryk University. Panevová, Jarmila 1974 “On verbal frames in Functional Generative Description”. Prague Bulletin of Mathematical Linguistics 22: 3–40 and 23: 17–52; a revised version in Prague Studies in Mathematical Linguistics 6: 227–254. 1998 “Ješteˇ k teorii valence [Valency theory revisited]”. Slovo a Slovesnost 59: 1–14. Petkevicˇ, Vladimír 1995 “A new formal specification of underlying structures”. Theoretical Linguistics 21: 7–61. Sgall, Petr 1967 “Functional sentence perspective in a generative description”. Prague Studies in Mathematical Linguistics 2: 203–225. 1992 “Underlying structure of sentences and its relations to semantics”. Wiener Slawistischer Almanach [Sonderband 33], T. Reuther (ed.), 273–282. Sgall, Petr, Hajicˇová, Eva and Panevová, Jarmila 1986 The meaning of the sentence in its semantic and pragmatic aspects. Dordrecht: Reidel — Prague: Academia.
AUTHOR "Jens Allwood"
TITLE "Capturing differences between social activities in spoken language"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Capturing differences between social activities in spoken language Jens Allwood University of Göteborg
1.
Introduction
Wittgenstein (1953) introduced the concept of a “language game” and claimed that what we often think of as a language really consists of a collection of more or less diverse “language games”. Standard or national languages can therefore only be the result of some form of abstraction performed by linguists (cf. Harris 1980) or other persons, mostly in the service of socio-political interests. It has, therefore, both theoretical and practical interest to investigate to what extent “language games” actually are different and to what extent there exist linguistic phenomena which play an equally important role in all “language games”. Another issue brought up by Wittgenstein’s challenge is the question of how we should conceive of “language games”. His own remarks are inspiring but require specification to enable empirical investigation of the issues involved. There have been several proposals for how to do this. One proposal is to identify language games with what in the study of literature is called “genres”. This is, for example, the approach assumed in Biber (1988). A problem with this approach, if one is mainly interested in spoken language, is that it tends to make analysis of spoken language dependent on what has worked well for written language. Another approach is therefore to take as one’s point of departure spoken language. One such approach is the approach suggested in Allwood (1980 and 1995), where variation in spoken language is seen as dependent on what social activity spoken language is serving as an instrument for. The purpose, the roles, the artefacts, environment and domains of the activity are claimed to have a crucial influence on the language (lexicon, grammar and functions) of the activity.
302 Jens Allwood
2. Data To investigate this claim, a spoken language corpus which today comprises about one million words of spoken Swedish has been assembled. The corpus has been collected with the goal of representing many different social activities. The following activity-types (Table 1) are represented in the corpus. Table 1.The Göteborg spoken language corpus Activity
Recordings
Words
Shop Occupational therapy Auction Discussion Court Formal meeting Quarrel Hotel Informal conversation Interview Classrom interaction Consultation Dinner Trade fair Sermon Radio talk show Role play Conversation in factory Seminar discussion Physical therapy Phone Market TV Task-oriented dialogue Retelling of article
Capturing differences between social activities in spoken language 303
3. Method The characterization of the linguistic differences between the different social activities has been done in two ways: 1. manual coding of certain functions of communication 2. development of automatic routines to capture linguistic properties of languages in different social activities In this paper I will only discuss the automatically derivable properties. What can be derived automatically is, of course, dependent on what is given as input, i.e., on what properties are captured in the transcriptions of our recorded activities. I will therefore briefly present the transcription standard we have used. Consider (1) below: (1) Transcription according to the MSO4 standard with translation. §1. Small talk $D: säger du de0 ä0 de0 ä0 de0 så0 besvärlit då $P: ja0 ja0 $D: m0 // ha0 / de0 kan ju bli0 så0 se1 du $P: < jaha > @ $D: du ta1 den på morronen $P: nej inte på MORRONEN kan ja1 ju tar allti en0 promenad på förmiddan [1 å0]1 då vill ja0 inte ha0 [2 den] 2 medicinen å0 sen1 nä1 ja0 kommer hem möjligtvis $D: [1 a0] 1 $D: [2 nä0] 2
$D: oh I see is it it is so troublesome then $P: yes yes $D: m // yes / it can be that way you see $P < yes > @ $D: you take it in the morning $P: no not in the MORNING I always take a walk before lunch [1 and] 1 then I don’t want [2that] 2 medicine and then when I get home possibly $D: [1 yes] 1 $D: [2 no] 2
As we can see the transcription has the following properties 1. Section boundaries paragraph sign (§). These divide a longer activity up into subactivities. A doctor-patient interview can, for example have the following subactivities: (1) greetings and introduction, (2) reason for visit, (3) investigation, (4) prescribing treatment. 2. Words and space between words.
304 Jens Allwood
3. Dollar sign ($) followed by capital letter, followed by colon (:) to indicate a new speaker and a new utterance. 4. Double slash (//) to indicate pauses. 5. Capital letters to indicate contrastive stress. 6. Word indexes to indicate which written language word corresponds to the spoken form given in the transcription (de0 corresponds to written language det). 7. Overlaps are indicated using square brackets ([ ]) with indices which allow disambiguation if several speakers overlap simultaneously. 8. Comments can be inserted using angular brackets (· Ò to mark the scope of the comment and @· Ò for inserting the actual comment). These comments are about events which are important for the interaction or about such things as voice quality and gestures. By using this information, which, in comparison with some transcription systems that provide detailed information about prosody, is relatively close to written language, we have defined a set of automatically derivable properties which include the following: 1. Volume: Volume comprises measures of the number of words, pauses, stresses, overlaps, utterances, turns relative to speaker, activity and subactivity. 2. Ratios: Various ratios can then be calculated based on the volume measures. (2) mean length per utterance (=MLU) % pauses % stress % overlap
Alternatively, pause, stress and overlap can be given per utterance. All of these measures can then be relativized to speaker, activity or subactivity. 3. Special measures: One example of a special type of measure is “vocabulary richness” as measured through type/token, Guiraud, Über, Herdan or “theoretical vocabulary”, cf. van Hout & Rietveld (1993). Another measure we have constructed is “stereotypicality”, which looks at how often words and phrases are repeated in an activity. 4. Lemma: We have also implemented a simple stemming algorithm, which enables us to collect regularly inflected forms together with their stems. 5. Parts of speech: Parts of speech are assigned using a probability based algorithm and adjusted probabilities originally based on a Swedish written
Capturing differences between social activities in spoken language 305
language corpus tagged for part of speech as input. Words subdivided according to part of speech can then be assigned to speaker, activity or subactivity. 6. Collocations: All speakers, activities and subactivities can be characterized in terms of their most frequent collocations. 7. Sequences of parts of speech: Utterances of different length can be characterized as to sequence of parts of speech. This allows a first analysis of grammatical differences between speakers, activities and subactivities. 8. Similarities: Similarities between activities are captured by looking at the extent to which words and collocations are shared between activities.
4. Examples of the measures applied to six activity types Let us now consider some examples. They will consist in demonstrating how the measures are applied to six activity types chosen from our corpus (Informal talk at home, Interaction in a computer shop, Doctor-patient consultation, Demonstration at a trade fair, Auction, Sermon in a church). 4.1 Volume and Ratios in the six activities The percentages are calculated on the basis of the subcorpus created by the six activities (Table 2). Table 2.Volume of subcorpus
There are also measures of contrastive stress and pausing (Table 3). The percentages are calculated relative to the number of tokens in the corpus.
306 Jens Allwood
Table 3.Pauses and stress
Pause Stress Str%Tok Pau%Tok
Informal
Shop
2851. 360. 0.5 4.0
3802. 606. 1.5 9.2
Consultation Trade Fair 1013. 116. 0.6 5.4
1096. 14. 0.1 7.6
Auction
Sermon
2136. 2111. 17.1 17.3
1457. 621. 6.3 14.9
Another property directly based on the transcriptions is “overlap of utterances”. In Table 4, I give a few measures based on “overlap”. Table 4.Overlap
No of Overl TokOverl Overl % Utt Ovetok %Tok
Informal
Shop
2461. 5166. 37.6 7.3
1276. 2526. 23.1 6.1
Consultation Trade Fair 926. 1630. 39.7 8.7
12. 38. 9.7 0.3
Auction
Sermon
18. 28. 11.7 0.2
6. 22. 31.6 0.2
On the basis of the volume and ratio data alone, we can see that the activities fall into two groups. The first group is more interactive and is characterized by more utterances per words, i.e., shorter mean length per utterance (=MLU) and more overlap. The second group is more monological and characterized by fewer utterances and thus by very much longer MLU as well as by less overlap. The sermon and the auction are also characterized by many more pauses and stressed words. 4.2 Special measures 4.2.1
We can see that the different measures of vocabulary richness (Table 5) give slightly different results. As has been pointed out in the literature, many of them
Capturing differences between social activities in spoken language 307
are sensitive to text length. The most neutral measure seems to be Vocab (theoretical vocabulary, cf. van Hout & Rietveld 1993). According to this measure, the sermon is richest in vocabulary. 4.2.2 Stereotypicality Table 6.Stereotypicality Tokens in Collocation 1 2 3 4 5 6 7 8 9 10
The measure of stereotypicality (Table 6) shows that the auction is the most stereotypical activity with many repeated collocations. We also see that for collocations with many words all the three activities that involve a greater amount of one-way communication are also more stereotypical. In the case of the sermon, we can see that it is actually the least stereotypical activity on the word and word-pair level, but becomes more stereotypical as the number of words in a collocation increase. This is probably related to the fact that both auctions and sermons involve the use of long standardized phrases. 4.3 Lemmatized Words and Collocations In order to derive the vocabulary of an activity, we have employed three methods: (1) overall vocabulary frequency, pulling out unique words, (2) only categorematic terms (nouns, verbs, adjectives and adverbs), and (3) lemmatizing categorematic words and collocations. In Table 7, I give a few examples of the third procedure.
308 Jens Allwood
Table 7.Lemmatized words and collocations Informal Words
Shop
Consultation
Trade Fair
Auction
Sermon
åka ‘go’
köpa ‘buy’
ont ‘hurt’
kronor ‘crowns’
kronor ‘crowns’
gud ‘god’
dag ‘day’
spelar ‘play’
dag ‘day’
bilen ‘the car’
etthundra ‘one ande ‘spirit’ hundred’
jobbar ‘work’
kronor ‘crowns’
tabletter ‘pills’
kniven ‘the knife’
nummer ‘number’
herre ‘lord’
år ‘year’
kostar ‘cost’
barn ‘child’
år ‘year’
femtiolappen ‘the fifty note’
ord ‘word’
hör ‘hear’
heter ‘is called’ titta ‘look’
idag ‘today’
sjuttiofem ‘seventyfive’
heliga ‘holy’
kronor för ‘crowns for’
helige ande ‘holy spirit’
Pairs tala om ‘talk about’
titta på ‘look at’ titta på ‘look at’ här vise ‘this way’
till exempel ‘for som heter example’ ‘called’
ont i ‘pain in’
års garanti femtio kronor ‘year’s guaran- ‘fifty crowns’ tee’
den helige ‘the holy’
i morgon ‘tomorrow’
spel och ‘game and’
jag skriver ‘I write’
vår bil ‘our car’ såld för ‘sold for’
vi ber ‘we pray’
i dag ‘today’
köpa en ‘buy a’ på morgonen ‘in the morning’
kronor för ‘crowns for’
femtiolappen för ‘fifty note for’
jesus kristus ‘Jesus Christ’
fråga om ‘ask about’
heter det ‘is called’
kniv som ‘knife åttiosju which’ nummer ‘eighty seven items’
ber för ‘pray for’
i benet ‘in the leg’
Triples vad heter det spel och sådant det gör ont ‘what’s it called’ ‘games and ‘it hurts’ such’
det här vise ‘this way’
och hälsade på ‘and greeted’
titta på den ‘look at it’
under dagens femtio kronor hushåll för ‘fifty crowns ‘during the fair’ for’
vi ber för ‘we pray for’
över huvud taget ‘at all’
vad heter det vi titta på ‘what’s it called’ ‘we look at’
på detta viset ‘in this way’
låt oss bedja ‘let us pray’
är du rädd ‘are you afraid’
kronor för den ‘crowns for it’
viktor åttiosju nummer ‘viktor eightyseven number’
den helige ande ‘the holy spirit’
Capturing differences between social activities in spoken language 309
kommer du ihåg ‘do you remember’
lämnade in den titta på det ‘delivered it’ ‘look at it’
på det viset ‘in this way’
femtiolappen för hör vår bön den ‘fifty note ‘hear our for it’ prayer’
är frågan om ‘a question of ’
ha en påse ‘have a bag’
vaxar din bil ‘wax your car’
femtiolappen på fadern och den ‘fifty note on sonen ‘the it’ father and the son’
och titta på ‘and look at’
The translations fairly clearly indicate that the distinct domains of the six activities are reflected in the words and collocations that are selected. The reason for this is that many activities coincide with particular conceptual domains. We could talk about religion in a shop, but mostly we don’t, so that even when there is no necessary link between a particular activity and a particular conceptual domain, there might be such a link in practice and this will be empirically observable. 4.4 Parts of Speech In Table 8, data on parts of speech in the six activities are presented. The table also gives data on the total number of tokens in the activities and the relative share of the parts of speech in all activities. Note that we have introduced “feedback words” (fb), “own communication managent words”, e.g. hesitation sounds as a parts of speech. There are also the two categories of “phrases which function as single units” (phras) and “delimiters” (del) which are the signs we use to mark pauses, overlaps and comments with. See the description of the transcription conventions above.
310 Jens Allwood
Table 8.Parts of speech Informal Shop
Consul- Trade tation Fair
Auction Sermon Tot%
pron verb adv noun conj prep fb adj num del ocm int phras
As we can see, pronouns are totally the most frequent part of speech, followed by verbs and adverbs. This is a fairly stable result for spoken language. This is also, perhaps not surprisingly, the order we find in the three first activities which are the most interactive. In the three last activities, the order is different. In the Trade fair, adverbs and nouns are the two most frequent. This can probably be explained by the nature of the activity where a person is demonstrating various objects to an audience. In the Auction, adverbs and numerals are the most frequent and pronouns and nouns are third and fourth most frequent. Again, this can probably be explained by the nature of the activity, which involves displaying and bidding for displayed objects. Finally, the sermon has nouns and adverbs as the most frequent parts of speech. It also has an unusually high share of adjectives — higher than any other activity. It is likely that these features can be related to the role that biblical quotations and descriptions play in this activity. Let us now turn our attention to more particular data about each part of speech. The most common words in four parts of speech are presented for each activity in Table 9. Of the three parts of speech, nouns show the most interesting differences between the activities. (English translations are given only the first time that a word occurs in the tables). The table shows that syncategorematic function words, e.g., pronouns and
vi
en
de
som
man
vad ‘what’
1271
1013
765
701
676
614
561
541
du ‘you’
de ‘they’
man ‘one’
den ‘it’
en ‘a’
vi ‘we’
han ‘he’
som ‘that’
är
har
ska
kan
var
ha
får
vet
tror ‘thinks’
köpa ‘buy’
1756
937
805
521
454
354
331
325
315
248
har ‘has’
var ‘was’
kan ‘can’
ska ‘will’
vet ‘knows’
får ‘gets’
hade ‘had’
skulle ‘should’
ha ‘have’
du
är ‘is’
Verbs
den
2254
jag
4221
det
Shop
jag ‘I’
Freq
det ‘it’
Pronouns
Informal
man
742
112
125
164
167
227
273
316
335
629
1290
279
284
344
443
473
tar ‘takes’
ta ‘take’
får
gör ‘does’
ha
kan
var
ska
har
är
vad
vi
som
de
en
den
jag
765 473
du
det
Consultation
1017
2519
Freq
Table 9.Pronouns, verbs, nouns and feedback
66
78
86
95
118
154
157
170
338
511
78
103
105
113
137
152
168
46
50
57
119
142
204
366
52
68
75
83
101
109
163
168
291
650
Freq
vet ‘knows’
blir ‘becomes’
31
38
kommer ‘comes’ 45
gör
ta
tar
ska
kan
har
är
alla ‘all’
som
jag
du
man
en
den
ni ‘you’
vi
633
det
657
Trade Fair
1309
Freq
27
43
82
84
108
129
219
392
768
Freq
får
ska
ger ‘gives’
går ‘walks’
börjar ‘begins’
kommer
börja ‘begin’
hade
är
har
52
57
59
66
71
109
112
142
147
470
någon ‘some-one’ 25
alla
ett ‘a’
jag
ingen ‘no one’
vem ‘who’
en
det
den
vi
Auction
hör ‘hears’
ber ‘prays’
vill ‘wants’
ska
hade
kommer
kan
älskar ‘loves’
har
är
honom ‘him’
dem ‘them’
som
den
din ‘your’
dig ‘you’
han ‘he’
oss ‘us’
vi
det
Sermon
21
22
22
26
27
28
29
33
83
148
48
49
49
51
51
52
63
86
149
213
Freq
Capturing differences between social activities in spoken language 311
datorn ‘the comp- 40 uter’ 35
52
46
vecka ‘week’
jobbet ‘the job’
mats (name)
42
morgon ‘morning’ 42
priset ‘the price’
bilden ‘the picture’
spänn ‘buck’
43
tid ‘time’
vecka ‘week’
klockan ‘the time’ 45
mats
29
31
31
32
43
spel ‘game’
90
gång ‘time’ 34
36
ögat ‘the eye’
medicinen ‘the medicine’
doktorn ‘the doctor’
år ‘year’
22
25
26
28
blodtryck ‘blood 31 pressure’
gång ‘time’
vecka ‘week’
39
benet ‘the leg’
60
fall ‘case’
101
49
dag ‘day’ tabletter ‘tablets’ 48
175 84
år ‘year’
kronor ‘crowns’
119
dag ‘day’
tack ‘thanks’
123
fall ‘case’
Nouns man ‘one’
128 49
51
burken ‘the jar’
dag ‘day’
munstycke ‘mouth-piece’
viset ‘the way’
gång ‘time’
år ‘year’
20
21
21
22
23
24
kniven ‘the knife’ 43
bil ‘car’
kronor ‘crowns’
kronor ‘crowns’
336
24
27
31
44
58
78
vatten ‘the water’ 23
lars (name)
oljan ‘the oil’
sigvard (name)
viktor (name)
lådan ‘the box’
hundralapp ‘hundred note’
femtiolappen ‘the 162 fifty note’
nummer ‘number’ 163
gud ‘god’
103
48
55
67
jesus (name)
kärlek ‘love’
man ‘man’
människor ‘people’
kyrka ‘church‘
30
30
32
33
35
fadern ‘the father’ 38
ord ‘word’
herre ‘lord’
ande ‘spirit’
312 Jens Allwood
ja
nä
m
va
okey
jaha ‘yes’
nej
jo
jaså
jaja
700
468
294
222
143
111
91
73
57
nä ‘no’
m
va ‘what’
hm
jo ‘yes’
jaja ‘yesyes’
nej ‘no’
jaså ‘I see’
jaja ‘yesyes’
Shop
2252
Freq
ja ‘yes’
Feedback
Informal
17
41
52
56
92
107
302
409
475
1453
Freq
aha
jaså
hm
jo
va
jaha
nej
nä
m
ja
Consultation
11
22
25
29
42
45
98
142
176
867
Freq
nähä
a
m
ne ‘no’
okey
nä
nej
jo
va
ja
Trade Fair
1
2
2
2
3
7
8
9
84
110
Freq
jae ‘yes’
jaså
jo
va
niao
nä
jaha
ja
Auction
1
1
1
1
2
6
7
90
Freq
a ‘yes’
javisst ‘yes sure’
jo
m
nä
va
ja
Sermon
1
1
1
1
1
1
13
Freq
Capturing differences between social activities in spoken language 313
314 Jens Allwood
feedback words, are shared to a much greater extent than categorematic words, e.g., most verbs and nouns. Only in the case of nouns is there a big difference, in accordance with the activity domains, between what words are frequent in the six activities. The verbs do not show an equal differentiation, since most of the highly frequent verbs are auxiliaries, i.e., syncategorematic.
5. Sequence of parts of speech in specific utterance types Another measure of distinctiveness can be obtained by studying what sequences of parts of speech are typical of each activity. Space allows me to exhibit only the sequences of utterances found in the shop activity (Table 10). As we can see, these utterances mostly are what can be expected in a shop context. Table 10.Collocations Shop Part of speech
Frequency
Most common example
Freqency of example
1-word utterances feedback (=fb) noun int adv verb ocm num pron adj conj
va3 de0 bra så0 ‘was it good like that’ nä0 ja1 har de0 ‘no I have that’ näe han sa0 ju ‘no he said you know’ vi0 bytte en0 grej ‘we changed a thing’ du tror de me ‘you think so too’
9
okej ja1 tar de0 då ‘ok I will take that then’ va1 tar ni för dom2 ‘what do you charge for them’ så0 kan ja1 ta0 dom2 ‘so can I take them’ ä0 de0 e0 nån häst ‘is that a horse’ ä0 re0 hä1 bra då ‘is this good then’
1
ska du få0 en0 påse också ‘would you like a bag too’ m0 man kan nästan tro0 de0 ‘you can almost believe that’ okej ja1 kikar in i morron ‘ok I will look in tomorrow’ nähe de0 va0 de0 ja1 trodde ‘no that’s what I thought’
ja0 de0 ska den ju säkert göra ‘yes it should certainly do 1 that’ nä0 ja1 vet inte va3 de0 va3 1 ‘no I don’t know what it is’
315
316 Jens Allwood
6. Similarities between activites Let us now turn to the question of the extent to which words and collocations are shared between activities (Table 11). Only words and pairs are translated. In Swedish spoken language, it seems as if constructions involving de (it) as a topic marker with a copula e (is) or a conjunction å (and) are among the linguistic sine qua non of most social activities. Table 11.Words and collocations Number of activities
å0 de0 e0 ‘and it is’ ja0 de0 e0 ‘yes it is’ nu ska vi0 ‘now will we’ de0 e0 ju ‘it is you know’ i alla fall ‘in any case’
Quadruples 4 4 4 4 4
2.1437 2.0090 1.9291 1.8027 1.8027
18 17 15 14 14
de0 e0 de0 som ‘it is what which’ de0 e0 ju en0 ‘it is a’ då ska vi0 se0 ‘then let us see’ ja0 de0 e0 de0 ‘yes it is it’ de0 e0 inte så0 ‘it is not so’
Capturing differences between social activities in spoken language
Quintuples 3 3 3 3 3
7 6 5 3 4
1.0595 1.0343 0.9336 0.8819 0.8604
ja1 vet inte om0 de0 ‘I dont´ know if it’ de0 e0 de0 som e0 ‘it is what which is’ de0 e0 ju så0 att ‘it is you know so that’ ja1 vill att du ska ‘I want you to’ i å0 me0 att du ‘since you’
We have also developed a measure of the relative amount of words and collocations occurring in one or more of the activities in a corpus. This is presented in Table 12. Table 12.Percentage of words and collocations occurring in a given number of activities Length Tok of coll. 1 2 3
The table shows that 76% of all words are confined to only one activity, 12 % occur in two activities, 5% in three activities etc. Only a very small part of the vocabulary, 1%, is shared between all six activities. Turning to collocations, we see that they are, to an even greater extent, connected with a single activity. This is so because most of the collocations only have a frequency of one.
7. Discussion This paper presents a number of ways in which linguistic properties of distinct social activities can be derived from transcriptions made according to a format with a relatively small number of special requirements. The properties extracted show that the activities differ from each other in volume and various ratios. MLU is an example of a ratio that directly measures mean length of utterance but can also be seen as a measure of interactivity. There are also differences as measured by “vocabulary richness” and “stereotypicality”. Furthermore, there are differences in vocabulary and in the parts of speech which are typically employed in an activity. Finally, we have been able to demonstrate differences in the collocations and in the grammatical structures
317
318
Jens Allwood
as indicated by the sequences of parts of speech which are typical of utterances with a given length. Turning to similarities, we have been able to show that certain words and collocations seem to be employed in all the activities that were studied. Returning to the original question posed in this paper, it seems that language to some extent can be seen as a collection of “language games”, which are distinct from each other in vocabulary, structure and function. But there also seem to be a very small number of certain highly frequent words and structures used in all activities. By and large, these words are all syncategorematic or function words, i.e., words helping to structure what we say. One question that this raises is whether the differences between the activities can be compared to differences between “national standard languages”. Can whole national standard language communities be characterized by some of the measures employed above? Could there be, for example, typical patterns of MLU, or typical uses of pauses and overlaps, preferred levels of vocabulary richness and word frequencies which differentiate national standard languages? For some of these measures, we have reason to believe that this is so. Concerning overlaps, for example, Fant (1995) have demonstrated that Swedish has a much lower rate than Spanish in the activity of business negotiation — a difference between Swedish and Spanish that might be repeated in other activities. Concerning word frequencies, it is also rather well-known that there are interesting differences. It remains an open question, however, if the range of linguistic variation between the activities of a single national language community is as great as what can be found between different national languages.
References Allwood, Jens 1980 “On Power in Communication”. In ALVAR: a Festschrift to Alvar Ellegård [SPELL I], J. Allwood and M. Ljung (eds), 1–20. University of Stockholm, Department of English. 1995 “An Activity Based Approach to Pragmatics”. Gothenburg Papers in Theoretical Linguistics 76: 1–38. Biber, Douglas 1988 Variation Across Speech and Writing. Cambridge: Cambridge University Press. Fant, Lars 1995 “Negotiation discourse and interaction in a cross-cultural perspective: The case of Sweden and Spain”. In The Discourse of International Negotiations, K. Ehlich and J. Wagner (eds), 177–201. Berlin-New York: Mouton de Gruyter.
Capturing differences between social activities in spoken language 319
Harris, Roy 1980 The Language Makers. London: Duckworth. Van Hout, Roeland and Rietveld, Toni 1993 Statistical Techniques for the Study of Langauge and Language Behaviour. Berlin & New York: Mouton de Gruyter. Wittgenstein, Ludwig von 1953 Philosophical Investigations. Oxford: Basil Blackwell.
AUTHOR "Bruce Fraser"
TITLE "An account of innuendo"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
An account of innuendo Bruce Fraser Boston University
1.
Introduction
The term innuendo is frequently found in the popular press, especially in an election year. Claims of innuendo, rumor and innuendo, gossip and innuendo, veiled innuendo, gossamer innuendo, unsubstantiated innuendo, and rank innuendo can be found attributed to both written and oral comments, but seldom is the offending comment cited.1 It seems to be assumed that the audience has a vague familiarization with what personal innuendo is in its various manifestations (hereafter only innuendo), and it is treated like a buzz word, intended to evoke an impression, much like reporting on a bad smell. People, in fact, have difficulty talking about innuendo. Whereas we have no difficulty reporting that “It was rumored that…”, “John hinted that…”, “Jane alluded to the fact that…”, or “Jake insinuated that…”, we do not use innuendo as a verb.2 If we are specific at all, we usually make a report, saying something like, “Through innuendo, Jones suggested that…”, “By innuendo, Jim alluded to the fact that…”, or “His reputation was besmirched by rumor and innuendo.” This paper is an investigation into innuendo: what it is, where it fits into linguistics, and how it is performed. After presenting a few representative examples of innuendo, I will examine what others have said about the concept, and following this, I will present my view of innuendo.
2. Examples of innuendo The first example of innuendo comes from the TV program “60 Minutes”. In 1991 Sen. Alan Simpson had alleged that a CNN reporter Peter Arnett was a sympathizer of Saddam Hussein, an allegation he later retracted. In his “60 Minutes” contribution on the March 24, 1991, Andy Rooney commented:
322 Bruce Fraser
(1) Well, Senator Simpson wrote a letter to The New York Times this week apologizing to Peter Arnett. In his letter, Senator Simpson said he was wrong repeating the rumors about Mr. Arnett’s family connection to the Viet Cong because he couldn’t prove them. “I said from the outset”, Senator Simpson wrote, “that if it couldn’t be proven, I would apologize.” In that same spirit, I’d like to apologize to the Senator tonight. For some time now, at our dinner table and when I am among friends, I have referred to Senator Simpson as Saddam Hussein’s friend. Senator Simpson did go to Baghdad to see Hussein last April 13th and at that time, he comforted Hussein for things being written about him in our newspapers by saying that American reporters were “pampered and haughty.” That’s why I’ve been calling Senator Simpson “Saddam Hussein’s friend.” Well, now I feel sort of bad about it. I shouldn’t have done that. Senator Simpson says that “The Wall Street Journal” has suggested he’s a racist, too. I certainly wouldn’t suggest he’s a racist because I simply don’t know. I’ve heard rumors that, if he could, he’d repeal the 1st Amendment guaranteeing freedom of the press. I’ve heard rumors that he’s one of our dumbest senators. It would be unfair of me to repeat those rumors because I’m not sure they’re true. I’ve never even met him. Neither can I prove that Senator Simpson is a friend of Saddam Hussein. It is not certain that they’re friends and, unless the facts prove otherwise, I apologize to him for having said they are friends. I hope you take this apology in the spirit in which it’s intended, Senator — unless you can prove otherwise. (CBS News Transcript, 60 Minutes, vol. 23, no. 28, March 24, 1991, pp. 17–18)
A second example comes from Fischer (1970). A ship captain, annoyed by his first-mate’s drunkenness, wrote in the daily log, “The first-mate was drunk all day.” The first-mate took issue with this entry, and the next day wrote, “The captain was sober all day.” (272) A third example comes from the comic strip Mallard Fillmore. On January 11, 2000, Mallard was commenting on what Clinton had characterized as “bloated tax cuts”. He ended his commentary with the following: “Whether you’re for or against ’em, cuts, by definition, just can’t be ‘bloated’. On the other hand, for someone who can’t define “is”, two syllable words must really be tough.”
An account of innuendo 323
A fourth example was overheard during a conversation between administrators over the merits of having a particular teacher serve as a guidance counselor on sex education for high school boys. One said, “Well, I would certainly agree that Mr. Jenkins is quite qualified to discuss all aspects of homosexual behavior.” A fifth example comes from a debate in early January, reported in the New York Times, January 15, 2000. Bill Bradley, after being attacked by Al Gore said, “Let me explain to you, Al, how the private sector works, O.K.?” And a final example involves former President George Bush talking about the Presidential campaign. Calling for new leadership, he said, “Here’s what I want; I want to see us return to a society with fundamental values.” All of these cases convey an innuendo: an implied message in the form of an allegation whose content constitutes some sort of unwanted ascription towards the target of the comment. In the case of the ship’s captain, the allegation is that the captain was not usually sober; in the case of Mallard Fillmore, the allegation is that President Clinton was disingenuous in defining “is” in connection with the Monica Lewinsky scandal; and in the case of former President Bush, it is that Vice President Gore, who is running against Bush’s son, was partly responsible for turning our society away from these fundamental values. Before examining the previous research, I want to point out that the term innuendo has been used in a broader way than that illustrated above. There is the two-word expression, “sexual innuendo”, which refers to allusions made of a sexual nature, typically to a member of the opposite sex. The comment, “I want to see more of you” when made by a man to a woman will often be construed as a sexual innuendo, as was the question made several years ago by model Brooke Shields who, in a provocative pose in an advertisement, asked, “Want to know what comes between me and my Calvins?” In the law of defamation, an innuendo is the plaintiff’s explanation, drawing on external facts, of the defamatory meaning of what was said or written, when that meaning is not apparent from the statement’s face. For example, the claim that “David burned his house” is not defamatory on its face, but it may be if it is further explained that David had insurance and that the words were understood to mean that he was defrauding the insurance company. Other examples show that innuendo is considered to be synonymous with rumor and there need be nothing derogatory associated with the use. For example, the Boston Globe reported that
324 Bruce Fraser
(2) Gretzky, 38, finally made official what has been swirling around as innuendo for a week, then as fait accompli for the last few days: that he will play one more game for the Rangers, tomorrow afternoon’s season finale against Pittsburgh at the Garden.
Earlier this year, the Globe reported that (3) Week two of the Bill Belichick saga got off to an early start last night with more rumor and wild innuendo. Reports of Bill Belichick being named Patriots coach and general manager today — with a second-round pick going to the Jets as compensation — were not accurate, according to Patriots spokesman Stacey James.
When Ellen DeGeneres, who portrayed the title character in “Ellen”, was about to reveal her sexual preference, the report was that “The groundbreaking story line came at the end of a long series of innuendo-riddled trailers, interviews and articles.” Finally a reporter commenting on the revival of “Godspell”, first performed in 1971, stated that “…when a young cast brings the musical to the stage…, its ’70s innuendo will be replaced with some 1990s references everyone can identify with”. These examples notwithstanding, I will treat innuendo as I indicated above: an implication in the form of an unwanted allegation about a person.3
3. Previous research on innuendo Austin (1962) who briefly discusses insinuating (which I suggest below is closely related to innuendo), writes that he is concerned that there are speech acts which defy his classification [of illocutionary and perlocutionary acts]: (4) …There may be some things we ‘do’ in some connection with saying something which do not seem to fall, intuitively at least, exactly into any of these roughly defined classes, or else seem to fall vaguely into more than one;… For example, insinuating, as when we insinuate something in or by issuing some utterance, seems to involve some convention, as in the illocutionary act; but we cannot say ‘I insinuate…’, and it seems like implying to be a clever effect rather than a mere act.” (104–5)
About the same time, Strawson (1964) addressed the alleged lack of overtness in insinuation when he wrote:
An account of innuendo 325
(5) An essential feature of the intentions which make up the illocutionary complex is their overtness. They have, one might say, essential avowability. This is, in one respect, a logically embarrassing feature…At the same time it [the criterion of avowability] enables us easily to dispose of insinuating as a candidate for the status of a type of illocutionary act. The whole point of insinuating is that the audience is to suspect but not more than suspect, the intention, for example to induce or disclose a certain belief. The intention one has in insinuating is essentially nonavowable. (33–4)
Strawson makes it clear that he does not see insinuating as an illocutionary act because, unlike an illocutionary act, the speaker does not have the intention that the utterance be recognized as intended as an insinuation; the hearer is merely to suspect it. Bach & Harnish (1979: 96–103) write of “collateral acts”, conversational (not illocutionary) acts that can be performed in conjunction with or in lieu of illocutionary acts. They suggest, analogous to perlocutionary effects, that the intentions behind these collateral acts may or may not be intended to be recognized. They propose three types. One type, unnamed, consists of acts which include circumlocutions, changing the subject, and make small talk, where the success of each does not depend on the recognition of the intention with which the act is performed. One can engage in small talk, be aware of doing so, without being aware whether the other intended this or not. A second type they label “overt acts”, such as kidding, storytelling, joking, punning, mimicking, and reciting. These require the hearer to recognize the intent with which the act was performed, although as they point out, you can kid or mimic someone without intending the person to recognize it. The third type they label “covert acts”, acts such as deliberate ambiguity, sneaky presupposition, innuendo (I presume insinuation, though this act is not specified), where they claim that: (6) Covert collateral acts are performed with intentions that are intended not to be recognized. Generally, these are acts of manipulation, including such devious acts as innuendo…Covert acts succeed (the intention with which they are performed is fulfilled) only if their intent is not recognized, or at last not recognized as intended to be recognized. The idea is to get someone to think you think something and thereby to get him to think it without recognizing that that’s what you want him to do. Whereas an indirect act is performed with an intention that can be reasonably expected to be recognized (on the basis of the utterance and the
326 Bruce Fraser
context) so that the speaker cannot, if challenged, plausibly deny that he intended the hearer to infer his intention, the key to innuendo is deniability. (101)
It is clear that Bach and Harnish also reject illocutionary act status for innuendo, and hence it is not considered to be an act of communication. In fact, they characterize an innuendo as “devious”. Bell (1997), the most detailed of the writers, defines innuendo as follows:4 (7) On the basis of examples like this, I want to define innuendo as follows: An innuendo is a non-overt intentional negative ascription, whether true or false, usually in the forms of an implicature, which is understood as a charge or accusation against what is, for the most part, a non present party…[it] is a verbal or non-verbal strategy ‘for critical junctures, that is, for situations in which overt comment or criticism would be improvident or improper but which demands some action on the speaker’s part.’ (36)
Bell follows Bach & Harnish (1979) claiming that an innuendo is an intentional act; it can not be accidentally performed as, for example, can the perlocutionary act of embarrassing someone. Moreover, he agrees that it is a non-overt act (his term rather than Bach and Harnish’s covert act), therefore not an illocutionary act, since it requires that “the speaker’s [communicative] intent not be recognized (or at least not recognized as intended to be recognized)” (37).5 According to Bell, an innuendo “can be seen as one type of conversational implicature” (46), though he never indicates what other types might be.
4. A different account of innuendo I will treat innuendo as a restricted type of insinuation, which in turn is a restricted type of implication. I shall begin this account of innuendo by briefly discussing the act of implication, then turning to the act of insinuating, and finally to the act of innuendo. 4.1 Implication Communication is generally considered to be effected by messages falling into one of two types: direct, and implied. In the case of a direct message, a speaker intends the message to be recognized on the basis of what has been said in conjunction with the context in which it is uttered. In the case of an implied
An account of innuendo 327
message, the speaker intends the audience to recognize the implication on the basis of the direct message in conjunction with the context. In both cases, the hallmark of communication is the fact that the speaker intends for the audience to recognize a message and intends that the audience realize that it was supposed to be recognized. Depending on the ideology embraced, an “implication” as it is used in pragmatics (as opposed to logic) may be referred to as an “implicature” (Grice 1975) or an “indirect speech act” (Bach & Harnish 1979). However, for my purposes here, I shall simply use the general term “implication” and mean by it a message inferred on the basis of the direct, explicit message in conjunction with its context. Examples of implication abound. The speaker uttering, “It’s cold in here” under the circumstances where the temperature in the room is below normal may intend to imply any of the following: “Please shut the window”, “Please bring me my sweater that is hanging over the chair”, Please turn down the air conditioner” or “You are insensitive not to turn up the heat”, depending on yet other contextual facts. The speaker complaining that “Unfortunately, they started late”, may be implying “So, they are going to arrive here late”, while the speaker commenting that “John is ideally suited for the center on our basketball team”, may be implying that “John is relatively tall, compared to the other players”.6 4.2 Insinuation I define an insinuation as an implication in the form of an allegation which asserts an unwanted ascription about someone. The details of this definition will come later. The examples given at the beginning of the paper, designated there as conveying an innuendo, may also be seen as conveying an insinuation. By examining these examples, it is clear that through insinuation, as opposed to expressing the message directly, the speaker can convey a message with a derogatory import, and, because the message is deniable, can avoid taking the responsibility for the negative materials. There is no disagreement among researchers that the act of insinuation is intentional, but there is a question of whether or not it is an implication, and hence communicative. I will first characterize the content of an insinuation, and then make the case that insinuation is a type of implication, hence a communicative act, contrary to the writers cited above. The content of an insinuation must meet the following five conditions. First, an insinuation must have a target, a person or institution which the
328 Bruce Fraser
content of implications is about. When you make the statement to John, “You have trouble telling reality from fiction”, and insinuate that “John is a liar”, John is the target.7 It is not necessary for the target to be explicit in the comment. Suggesting that, “We could use a leader in this state who knows how to handle finances”, could convey an insinuation about the present governor, Paul Cellucci, without naming him. Second, an insinuation occurs only in the form of an assertion which alleges an unproven claim about the target: an allegation. Implications in the form of imperatives or interrogatives do not qualify as insinuations. Third, the content of the allegation must involve some aspect of the target such as the target’s integrity, morals, sobriety, intelligence, behavior, sexual orientation, height, weight, and so forth. Virtually any aspect of the target can serve, as can be seen by the challenge: “Are you insinuating that I am a liar (that I am gay, ugly, dishonest, unhappy, a drunk, immoral, etc.)?” Fourth, and related to the third condition, the allegation must represent the aspect of the target in an unfavorable light, what I shall refer to as an unwanted attribution. For example, whereas we accept an insinuation that someone is a drunk (The Captain was sober all day), a liar, gay, ugly, etc., presumably because they consider the attribution to be unwanted, we do not accept as an insinuation that the target is sober, happy, beautiful, a heterosexual, or moral. Consider the humorous, if not unacceptable challenge, “Are you insinuating that I am the most intelligent student in the class?” or “Are you insinuating that I am sober, beautiful, and happy?” There are cases, though, where the insinuation captures an unwanted though not exactly derogatory condition. For example, the statement by a coach, “Let’s face it, Jackson, there are some people who are not going to make the squad”, who intends to insinuate “Jackson, you should not expect to make the team”, or the comment made to a feminist who became radicalized after 21, “Carla, you are so beautiful. Is it true that you entered beauty contests when you were a young woman?” insinuating that she is enamoured with her physical attributes. There are some situations where the allegation is true, but because of the negative import, the speaker considers the content unwanted. For example, when a speaker asked, “Is it true that Henry has AIDS?” intending to insinuate that “Henry has AIDS”, which is true, but something Henry would just as soon not make public. On the other hand, when either the target or the audience accepts the truth of a potential insinuation and does not consider the attribution unwanted, the resulting implication fails to be an insinuation. If “The
An account of innuendo 329
Captain was sober all day”, gives rise to the implication, “The Captain is typically drunk”, it may be that the Captain is drying out in a detox center and this news is welcome. Fifth, an insinuation may be conveyed to the target directly (either by speaking or letter), or to an audience, a group of people which may or may not include the target.8 The target and the audience may hear the comment as conveying an insinuation, neither of them may hear it as such, or one but not the other may. The designation of an implication as an insinuation rests with the receiving party. For example, if Jesse Jackson were to make the comment, “It’s a shame that some Americans can’t put aside skin color as a consideration for determining the quality of a human being”, in the midst of a debate between himself and the Grand Wizard of the Ku Klux Klan, the Klan members in the audience would recognize the intended allegation, namely, “The Grand Wizard is biased against blacks.” But, while they would consider the comment as designed to malign the Wizard, they would not consider the remark to convey an insinuation. The content of the remark was true, and it was exactly what they espouse. On the other hand, liberals in the audience would consider the remark as conveying an insinuation, given that they find it inconceivable that anyone would “want” the attribution of being a racist. Typically, for a comment to be considered an insinuation, the target/ audience has to believe that the speaker is unwilling or unable to state the insinuation as an explicit message under the circumstances. For an insinuation which is true, the avoidance of explicitness may stem from political or politeness considerations; for an insinuation which is false, the avoidance is a necessity, since to make an explicit allegation about someone requires proof. Let us now turn to the question of the pragmatic status of an insinuation. By analyzing insinuation as a species of the genus implication, it follows that it is a communicative act. It follows that if you comment, “Henry has trouble telling fact from fiction”, and intend to insinuate to the audience that “Henry is a liar”, I submit that you intend the audience to recognize that you were intending to insinuate, not merely to suspect as Strawson and Bach & Harnish suggest. This is analogous to the case of complaining, “It’s cold in here”, and intending the audience to recognize the message, “Please shut the window”. Let us examine the evidence opposed to this view. Austin (1962) was concerned that insinuation could not occur as a performative expression, for example, “I insinuate that” or “I innuendo that”, and concludes that an insinuation could not be an illocutionary act, therefore not a communicative act. He was basing this comment on the fact that the hallmark of
330 Bruce Fraser
an illocutionary act was its ability to take the syntactic form of a performative. But the fact that there is no performative for insinuate does not dispose of the matter. There are acts such as expressing disgust and expressing happiness for which there is no illocutionary verb, thus no performative expression, but nevertheless these are analyzed as illocutionary acts. Moreover, there is at least one group of verbs, including admonish, blame, complain, condemn, criticize, reprimand, reproach, scold, and threaten, which name illocutionary acts, but do not occur in performative expressions (e.g. “I hereby criticize you for…), possibly because the verbs all convey negative import towards the audience. What we have with insinuation is an act which can be performed only by implication. This fact places insinuation among a small minority of acts, but doesn’t disqualify it from being treated as a communicative act. Both Strawson (1964) and Bach & Harnish (1979) suggested that the failure of insinuation to achieve the status of an illocutionary act is tied up with its lack of avowability, the latter writing that covert acts are performed with intention that is intended not to be recognized. I think they both are mistaken about this aspect of insinuation.9 They only voiced their suspicions on the matter, and the single example in Bach & Harnish does not convince me that the insinuation in that case was not intended as an act of communication. Those hearing an insinuation are expected to do more than suspect; they are expected to recognize an insinuation as a type of implication. If one addresses a comment conveying an insinuation to the target directly, for example, there is no reason to conclude that the speaker does not intend the target to recognize the unwanted implication. When Bradley said to Gore, “Let me explain to you, Al, how the private sector works”, he surely intended Gore to catch his insinuation. And, since there was an audience, he had a similar intention towards them. When an angry faculty member stated that “The university believes on rewarding faculty for their outstanding performance as determined by the Dean”, she surely intended her audience, which did not include the Dean, to recognize her implied message, namely, “The Dean has his favorites and acts on them when it comes to salary increase.” In reviewing dozens of cases of what have been labeled insinuations by others, I have found no case where the speaker did not arguably intended to communicate. Of course, because of unwanted import the speaker may not want to take responsibility for the insinuation. Rejection of this responsibility is within the speaker’s ability. If, for example, Gore were to have challenged Bradley with, “Bill, what the hell are you talking about; I’m fully aware of the workings of the private sector”, Bradley could have responded with, “I didn’t mean to imply that
An account of innuendo
you were not.” It is an issue separate from that of the communicative status of an insinuation if, for political or other reasons, the speaker does wish to be explicit and direct with the content of the insinuation and takes the implied route. 4.3 Innuendo What is the difference between an insinuation and an innuendo? Very little, it turns out. They both are implications which require a target, and both take the form of an unwanted allegation about some aspect of the target. The difference lies at whom the comment is directed and who determines the allegation is unwanted. Whereas an insinuation can be directed either to the target, the audience, or to both, an innuendo is directed only to an audience. An innuendo is a special type of insinuation in which the implied message is intended for the audience and it doesn’t matter if the target is even aware of the comment. To be sure, the comment giving rise to an innuendo may be spoken to the target, for example, when Bradley said to Gore, “Let me explain to you. Al, how the private sector works”, but the innuendo was meant for the audience, who, at that time, were potential voters for Bradley. The response to an innuendo is illustrative. If someone made the comment to you, “You have trouble keeping fact and fiction separate”, you could reasonably ask, “Are you insinuating that I am a liar?” but not, “Are you implying (suggesting) by innuendo that I am a liar?” On the other hand, if I heard the same comment, I could report, “He was suggesting through innuendo that you were a liar.” Innuendo is an act reportable by anyone in the audience, on the basis of their perception of the import of the implication, but not by the target of the innuendo.10 The dispute between comedian Roseanne Arnold and TV critic Matt Roush reported in a column in the Washington Post, cited in Bell (1997), illustrates this point. In writing to Roush to protest his criticism of a comedy special starring her husband, Tom Arnold, she ended her letter with, (8) PS. You are not in a position to understand or criticize anything about heterosexuals.
Since an innuendo has an audience separate from the target, as long as this correspondence remained solely between Arnold and Roush, the postscript would not be interpreted as an innuendo. Roush could have responded to Arnold with “I didn’t really appreciate your insinuation”, if he was not gay, or
331
332 Bruce Fraser
“I didn’t appreciate your snide remark”, if he was, but he could not say “I didn’t really appreciate your innuendo”. But the audience, on reading of the comment by Roush, could think, “Your innuendo was not appropriate”, reflecting the audience’s perception of the ascription. It is the audience which decides if the implied allegation is an innuendo, based upon its understanding of the target’s sensitivity to the ascription and not on the actual facts, whatever they might be. Suppose, for example, the same postscript were addressed to Truman Capote, a writer who flaunts his homosexuality. For members of the audience who are aware that the implied allegation would not be considered offensive to Capote (in fact, he might very well have endorsed the innuendo), they would not ascribe an innuendo to the writer. A sniping barb, perhaps, designed to make malign him, but because they know that he didn’t care that the public knew the facts, it would not be an innuendo. For others, not in the know, it might very well be considered innuendo. One of the characteristics commonly associated with innuendo (but less so with insinuation, for reasons I am unaware of) is the fact that, true or not, it tends to stain, to remain associated with the target. Bell suggests that the semantic content but not innuendo can be cancelled, and thus the stain remains. He suggests that if the first mate were to say, “But I don’t mean to imply that he was drunk before”, referring to the ship’s Captain, he would be canceling the content of the innuendo but leaving the innuendo effect intact. In fact, he suggests that the effect of the unwanted allegation might even be enhanced by the explicit denial of that which heretofore had only been implied. Whereas I agree with his observations, this “staining” effect is not unique to innuendo or insinuation; it holds true for all implications. Suppose I made the statement, “You are a fine example of American youth”, implying thereby that “I admire you”. If I subsequently said, “But I don’t admire you”, I will have denied the semantic content of the implied message but I will not have erased the effect that there is something admirable about you. Why else would I have made my initial comment? Similarly, if I ask, “Did you paint that wonderful watercolor on the wall over there?”, implying that “I like the painting very much”, and then deny the implication, there is still the understanding that I did like the painting. Innuendoes do not differ from other implications in terms of defeasibility. Even though the implication has been denied, the original implied message — positive or negative — has been communicated to the audience and, like a direct message, a denial does not totally erase the effect. Moreover, the alleged more powerful staining effect of innuendo as opposed to explicit accusation is, in my opinion, greatly overrated. If I comment, “I hear
An account of innuendo 333
The Harrison Company has two sets of books”, intending to imply to my audience that “The Harrison Company is defrauding the IRS”, I don’t find that I have maligned the honesty of the Company more than if I had simply stated, “The Harrison Company is defrauding the IRS.” In fact, quite the contrary. This accounts for the enhanced effects of what was initially only innuendo, for example, when Mary Matalin, during the 1992 presidential campaign, denied the charge of sleaze by innuendo by explicitly denying it, “We’ve never said to the press that he’s [Clinton’s] a philandering, pot smoking draft dodger.”11 Bell has suggested that there is a difference between indirect speech acts and innuendo, namely, that (9) in an innuendo there is likely to be a greater need to attend to the seriousness and plausibility of the literal meaning than in an indirect speech act. Indirect speech acts function, in part, because of the implausibility of their literal meaning alone. (43)
I don’t find this to be the case. There is no reason to think that “The Captain was sober today” or “Let me explain to you, Al, how the private sector works”, are more plausible than “It’s cold in here”, or “There is a gas station around the corner”, implying that you can buy gas there. I find no basis for concluding that an innuendo is any different from any other implication in its plausibility.
5. How is an innuendo (or insinuation) is conveyed? For most implications, it is the content rather than the syntactic form of the explicit utterance that plays the greatest role. We see this to be true of innuendo as well. Thus, we find a wide range of forms conveying innuendo. There are explicit assertions, (10) a. The Captain was sober today. b. You have trouble separating fact from fiction. c. We were never able to convict him of any violation of the rules.
questions, (11) a.
At the time when you decided to file the lawsuit against the Company, did you know that it was involved in other lawsuits of the same sort? b. Did the police ever conclude that it was John who was looking in the girls’ windows?
334 Bruce Fraser
c. Does Bob have AIDS? d. Do you think Hillary is seeing someone?
as well as directives (12) a. Let me explain to you how the private sector works, Al. b. Bring your boyfriend with you, Harry.
A second means is through pragmatic presuppositions, where the presupposed proposition serves as the source for the innuendo. (13) a.
Well, fortunately, the CIA is no longer involved in political assassinations. b. Have you stopped beating your wife?
A third means is through what I shall dub “creative ambiguity”, cases where the dominant reading of the utterance is appropriate but the less dominant reading, once recognized, sets up the innuendo. Perhaps the best examples are the humorous recommendation statement: (14) a.
A man like him is hard to find. [He is always disappearing] b. You would indeed be fortunate to get this person to work for you. [He doesn’t do much work.] c. I can assure you that no person would be better for the job. [You might as well hire no one than him.] d. I would urge you to waste no time in making this candidate an offer of employment. [You have better things to do than offer him the job.]
Fourth, there are the relatively fixed expressions in which the speaker denies exactly the content of the innuendo. (15) a. I don’t want you to think I’m calling John a liar. b. I categorically deny that Herb is a racist; why I never even considered it. c. I don’t mean to imply that Helen is running around with young boys. d. No one has ever suggested that you were anything less than completely honest.
In addition, an innuendo can be conveyed by a unit larger than an utterance and some innuendo does not get conveyed until the entire paragraph is finished:
An account of innuendo 335
(16) If you knew that one of the candidates in this race was receiving money from illegal sources, would that affect your voting decision? Look into the matter and see where the campaign funds of my opponent are coming from. The facts might surprise you.
6. Conclusion In this paper I have claimed that innuendo is a special type of insinuation, and, moreover, that both acts are implications, thus communicative acts. This is in contrast to the literature which argues that a speaker who conveys an innuendo does so intentionally but without the intention that the intention be recognized. Obviously, this is a preliminary study and I recognize that some readers may disagree with my judgements and conclusion. In that case, it will be necessary to determine whether my analysis needs revision or whether there is a wide variation on how such acts are construed. I hope in the future to investigate related acts, including, allusion, character assassination, gossip, hinting, imputation, insulting, spreading rumor, and snide remarks, all of which lie for one reason or another on the periphery of communicative acts and are more complicated than the acts of requesting, apologizing, and claiming.
Notes 1. The terms “sexual innuendo” and “legal innuendo” appear to be related but are actually quite different concepts from “innuendo.” I will mention them briefly below but not analyze them in this paper. 2. There is a verb “to innuendo” which is both transitive and intransitive but it is not currently used. 3. Personal innuendo can be extended to cover institutions as well. 4. Bell claims that an innuendo is an implicature yet also claims it is a non-overt act. This is a contradiction that is never satisfactorily resolved. 5. He subsequently adopts the term “pragmatic act”, taken from Mey (1993). 6. In all cases, I use “may be implying” rather than “is implying”, since only the speaker knows what non-direct messages are intended. 7. An insinuation cannot be self-oriented. The statement “I usually shoot par” could be construed to imply that “I am a very good golfer” but I do not hear it as an insinuation. A
336 Bruce Fraser
challenge to the statement would take the form, “Are you suggesting that you are a good golfer?” rather than “Are you insinuating that you are a good golfer.” 8. Some comments which convey an insinuation are not designed for a one-on-one interchange, for example, Mallard Fillmore’s comment cited above. 9. Bach & Harnish did not treat insinuation but only innuendo in the quote above. I am assuming that they would have included insinuation. 10. A target of innuendo could complain, “My reputation was ruined by rumor and innuendo”, but this is a report based on other people’s assessment of the comments made about the target. 11. Cited in Bell (1997).
References Austin, John 1962 How do things with words. Cambridge: Cambridge University Press. Bach, Kent and Harnish, Robert M. 1979 Linguistic communication and speech acts. Cambridge, Mass.: MIT Press. Bell, David 1997 “Innuendo” Journal of Pragmatics 27: 35–59. Fischer, David 1970 Historians’ fallacies. New York: Harper & Row. Grice, Paul 1975 “Logic and conversation” In Pragmatics, S. Davis (ed.), 305–315. New York: Oxford University Press. Strawson, Peter F. 1964 “Intention and convention in speech acts”. In Pragmatics, S. Davis (ed.), 290–304. New York: Oxford University Press.
AUTHOR ""
TITLE "Name index"
SUBJECT "Pragmatics & Beyond, New Series, Volume 90"
V vagueness 125, 178–180, 182, 183 valency 279, 293, 297, 299 vocabulary richness 304, 306, 317, 318 volitionality 68 W Wh-interrogative 207, 209 what is said xi, xv, 147–161, 163, 164, 166–170 wide scope 213–217 word order vii, xix, 83, 203, 273, 274, 280–283, 285, 286, 289, 290, 294
In the PRAGMATICS AND BEYOND NEW SERIES the following titles have been published thus far or are scheduled for publication: 1. WALTER, Bettyruth: The Jury Summation as Speech Genre: An Ethnographic Study of What it Means to Those who Use it. Amsterdam/Philadelphia, 1988. 2. BARTON, Ellen: Nonsentential Constituents: A Theory of Grammatical Structure and Pragmatic Interpretation. Amsterdam/Philadelphia, 1990. 3. OLEKSY, Wieslaw (ed.): Contrastive Pragmatics. Amsterdam/Philadelphia, 1989. 4. RAFFLER-ENGEL, Walburga von (ed.): Doctor-Patient Interaction. Amsterdam/Philadelphia, 1989. 5. THELIN, Nils B. (ed.): Verbal Aspect in Discourse. Amsterdam/Philadelphia, 1990. 6. VERSCHUEREN, Jef (ed.): Selected Papers from the 1987 International Pragmatics Conference. Vol. I: Pragmatics at Issue. Vol. II: Levels of Linguistic Adaptation. Vol. III: The Pragmatics of Intercultural and International Communication (ed. with Jan Blommaert). Amsterdam/Philadelphia, 1991. 7. LINDENFELD, Jacqueline: Speech and Sociability at French Urban Market Places. Amsterdam/Philadelphia, 1990. 8. YOUNG, Lynne: Language as Behaviour, Language as Code: A Study of Academic English. Amsterdam/Philadelphia, 1990. 9. LUKE, Kang-Kwong: Utterance Particles in Cantonese Conversation. Amsterdam/Philadelphia, 1990. 10. MURRAY, Denise E.: Conversation for Action. The computer terminal as medium of communication. Amsterdam/Philadelphia, 1991. 11. LUONG, Hy V.: Discursive Practices and Linguistic Meanings. The Vietnamese system of person reference. Amsterdam/Philadelphia, 1990. 12. ABRAHAM, Werner (ed.): Discourse Particles. Descriptive and theoretical investigations on the logical, syntactic and pragmatic properties of discourse particles in German. Amsterdam/Philadelphia, 1991. 13. NUYTS, Jan, A. Machtelt BOLKESTEIN and Co VET (eds): Layers and Levels of Representation in Language Theory: a functional view. Amsterdam/Philadelphia, 1990. 14. SCHWARTZ, Ursula: Young Children’s Dyadic Pretend Play. Amsterdam/Philadelphia, 1991. 15. KOMTER, Martha: Conflict and Cooperation in Job Interviews. Amsterdam/Philadelphia, 1991. 16. MANN, William C. and Sandra A. THOMPSON (eds): Discourse Description: Diverse Linguistic Analyses of a Fund-Raising Text. Amsterdam/Philadelphia, 1992. 17. PIÉRAUT-LE BONNIEC, Gilberte and Marlene DOLITSKY (eds): Language Bases ... Discourse Bases. Amsterdam/Philadelphia, 1991. 18. JOHNSTONE, Barbara: Repetition in Arabic Discourse. Paradigms, syntagms and the ecology of language. Amsterdam/Philadelphia, 1991. 19. BAKER, Carolyn D. and Allan LUKE (eds): Towards a Critical Sociology of Reading Pedagogy. Papers of the XII World Congress on Reading. Amsterdam/Philadelphia, 1991. 20. NUYTS, Jan: Aspects of a Cognitive-Pragmatic Theory of Language. On cognition, functionalism, and grammar. Amsterdam/Philadelphia, 1992. 21. SEARLE, John R. et al.: (On) Searle on Conversation. Compiled and introduced by Herman Parret and Jef Verschueren. Amsterdam/Philadelphia, 1992.
22. AUER, Peter and Aldo Di LUZIO (eds): The Contextualization of Language. Amsterdam/Philadelphia, 1992. 23. FORTESCUE, Michael, Peter HARDER and Lars KRISTOFFERSEN (eds): Layered Structure and Reference in a Functional Perspective. Papers from the Functional Grammar Conference, Copenhagen, 1990. Amsterdam/Philadelphia, 1992. 24. MAYNARD, Senko K.: Discourse Modality: Subjectivity, Emotion and Voice in the Japanese Language. Amsterdam/Philadelphia, 1993. 25. COUPER-KUHLEN, Elizabeth: English Speech Rhythm. Form and function in everyday verbal interaction. Amsterdam/Philadelphia, 1993. 26. STYGALL, Gail: Trial Language. A study in differential discourse processing. Amsterdam/ Philadelphia, 1994. 27. SUTER, Hans Jürg: The Wedding Report: A Prototypical Approach to the Study of Traditional Text Types. Amsterdam/Philadelphia, 1993. 28. VAN DE WALLE, Lieve: Pragmatics and Classical Sanskrit. Amsterdam/Philadelphia, 1993. 29. BARSKY, Robert F.: Constructing a Productive Other: Discourse theory and the convention refugee hearing. Amsterdam/Philadelphia, 1994. 30. WORTHAM, Stanton E.F.: Acting Out Participant Examples in the Classroom. Amsterdam/Philadelphia, 1994. 31. WILDGEN, Wolfgang: Process, Image and Meaning. A realistic model of the meanings of sentences and narrative texts. Amsterdam/Philadelphia, 1994. 32. SHIBATANI, Masayoshi and Sandra A. THOMPSON (eds): Essays in Semantics and Pragmatics. Amsterdam/Philadelphia, 1995. 33. GOOSSENS, Louis, Paul PAUWELS, Brygida RUDZKA-OSTYN, Anne-Marie SIMONVANDENBERGEN and Johan VANPARYS: By Word of Mouth. Metaphor, metonymy and linguistic action in a cognitive perspective. Amsterdam/Philadelphia, 1995. 34. BARBE, Katharina: Irony in Context. Amsterdam/Philadelphia, 1995. 35. JUCKER, Andreas H. (ed.): Historical Pragmatics. Pragmatic developments in the history of English. Amsterdam/Philadelphia, 1995. 36. CHILTON, Paul, Mikhail V. ILYIN and Jacob MEY: Political Discourse in Transition in Eastern and Western Europe (1989-1991). Amsterdam/Philadelphia, 1998. 37. CARSTON, Robyn and Seiji UCHIDA (eds): Relevance Theory. Applications and implications. Amsterdam/Philadelphia, 1998. 38. FRETHEIM, Thorstein and Jeanette K. GUNDEL (eds): Reference and Referent Accessibility. Amsterdam/Philadelphia, 1996. 39. HERRING, Susan (ed.): Computer-Mediated Communication. Linguistic, social, and cross-cultural perspectives. Amsterdam/Philadelphia, 1996. 40. DIAMOND, Julie: Status and Power in Verbal Interaction. A study of discourse in a closeknit social network. Amsterdam/Philadelphia, 1996. 41. VENTOLA, Eija and Anna MAURANEN, (eds): Academic Writing. Intercultural and textual issues. Amsterdam/Philadelphia, 1996. 42. WODAK, Ruth and Helga KOTTHOFF (eds): Communicating Gender in Context. Amsterdam/Philadelphia, 1997. 43. JANSSEN, Theo A.J.M. and Wim van der WURFF (eds): Reported Speech. Forms and functions of the verb. Amsterdam/Philadelphia, 1996. 44. BARGIELA-CHIAPPINI, Francesca and Sandra J. HARRIS: Managing Language. The
discourse of corporate meetings. Amsterdam/Philadelphia, 1997. 45. PALTRIDGE, Brian: Genre, Frames and Writing in Research Settings. Amsterdam/Philadelphia, 1997. 46. GEORGAKOPOULOU, Alexandra: Narrative Performances. A study of Modern Greek storytelling. Amsterdam/Philadelphia, 1997. 47. CHESTERMAN, Andrew: Contrastive Functional Analysis. Amsterdam/Philadelphia, 1998. 48. KAMIO, Akio: Territory of Information. Amsterdam/Philadelphia, 1997. 49. KURZON, Dennis: Discourse of Silence. Amsterdam/Philadelphia, 1998. 50. GRENOBLE, Lenore: Deixis and Information Packaging in Russian Discourse. Amsterdam/Philadelphia, 1998. 51. BOULIMA, Jamila: Negotiated Interaction in Target Language Classroom Discourse. Amsterdam/Philadelphia, 1999. 52. GILLIS, Steven and Annick DE HOUWER (eds): The Acquisition of Dutch. Amsterdam/ Philadelphia, 1998. 53. MOSEGAARD HANSEN, Maj-Britt: The Function of Discourse Particles. A study with special reference to spoken standard French. Amsterdam/Philadelphia, 1998. 54. HYLAND, Ken: Hedging in Scientific Research Articles. Amsterdam/Philadelphia, 1998. 55. ALLWOOD, Jens and Peter Gärdenfors (eds): Cognitive Semantics. Meaning and cognition. Amsterdam/Philadelphia, 1999. 56. TANAKA, Hiroko: Language, Culture and Social Interaction. Turn-taking in Japanese and Anglo-American English. Amsterdam/Philadelphia, 1999. 57 JUCKER, Andreas H. and Yael ZIV (eds): Discourse Markers. Descriptions and theory. Amsterdam/Philadelphia, 1998. 58. ROUCHOTA, Villy and Andreas H. JUCKER (eds): Current Issues in Relevance Theory. Amsterdam/Philadelphia, 1998. 59. KAMIO, Akio and Ken-ichi TAKAMI (eds): Function and Structure. In honor of Susumu Kuno. 1999. 60. JACOBS, Geert: Preformulating the News. An analysis of the metapragmatics of press releases. 1999. 61. MILLS, Margaret H. (ed.): Slavic Gender Linguistics. 1999. 62. TZANNE, Angeliki: Talking at Cross-Purposes. The dynamics of miscommunication. 2000. 63. BUBLITZ, Wolfram, Uta LENK and Eija VENTOLA (eds.): Coherence in Spoken and Written Discourse. How to create it and how to describe it.Selected papers from the International Workshop on Coherence, Augsburg, 24-27 April 1997. 1999. 64. SVENNEVIG, Jan: Getting Acquainted in Conversation. A study of initial interactions. 1999. 65. COOREN, François: The Organizing Dimension of Communication. 2000. 66. JUCKER, Andreas H., Gerd FRITZ and Franz LEBSANFT (eds.): Historical Dialogue Analysis. 1999. 67. TAAVITSAINEN, Irma, Gunnel MELCHERS and Päivi PAHTA (eds.): Dimensions of Writing in Nonstandard English. 1999. 68. ARNOVICK, Leslie: Diachronic Pragmatics. Seven case studies in English illocutionary development. 1999.
69. NOH, Eun-Ju: The Semantics and Pragmatics of Metarepresentation in English. A relevance-theoretic account. 2000. 70. SORJONEN, Marja-Leena: Recipient Activities Particles nii(n) and joo as Responses in Finnish Conversation. n.y.p. 71. GÓMEZ-GONZÁLEZ, María Ángeles: The Theme-Topic Interface. Evidence from English. 2001. 72. MARMARIDOU, Sophia S.A.: Pragmatic Meaning and Cognition. 2000. 73. HESTER, Stephen and David FRANCIS (eds.): Local Educational Order. Ethnomethodological studies of knowledge in action. 2000. 74. TROSBORG, Anna (ed.): Analysing Professional Genres. 2000. 75. PILKINGTON, Adrian: Poetic Effects. A relevance theory perspective. 2000. 76. MATSUI, Tomoko: Bridging and Relevance. 2000. 77. VANDERVEKEN, Daniel and Susumu KUBO (eds.): Essays in Speech Act Theory. n.y.p. 78. SELL, Roger D. : Literature as Communication. The foundations of mediating criticism. 2000. 79. ANDERSEN, Gisle and Thorstein FRETHEIM (eds.): Pragmatic Markers and Propositional Attitude. 2000. 80. UNGERER, Friedrich (ed.): English Media Texts – Past and Present. Language and textual structure. 2000. 81. DI LUZIO, Aldo, Susanne GÜNTHNER and Franca ORLETTI (eds.): Culture in Communication. Analyses of intercultural situations. n.y.p. 82. KHALIL, Esam N.: Grounding in English and Arabic News Discourse. 2000. 83. MÁRQUEZ REITER, Rosina: Linguistic Politeness in Britain and Uruguay. A contrastive study of requests and apologies. 2000. 84. ANDERSEN, Gisle: Pragmatic Markers and Sociolinguistic Variation. A relevance-theoretic approach to the language of adolescents. 2001. 85. COLLINS, Daniel E.: Reanimated Voices. Speech reporting in a historical-pragmatic perspective. 2001. 86. IFANTIDOU, Elly: Evidentials and Relevance. n.y.p. 87. MUSHIN, Ilana: Evidentiality and Epistemological Stance. Narrative Retelling. n.y.p. 88. BAYRAKTAROGLU, Arin and Maria SIFIANOU (eds.): Linguistic Politeness Across Boundaries. Linguistic Politeness Across Boundaries. n.y.p. 89. ITAKURA, Hiroko: Conversational Dominance and Gender. A study of Japanese speakers in first and second language contexts. n.y.p. 90. KENESEI, István and Robert M. HARNISH (eds.): Perspectives on Semantics, Pragmatics, and Discourse. A Festschrift for Ferenc Kiefer. 2001. 91. GROSS, Joan: Speaking in Other Voices. An ethnography of Walloon puppet theaters. n.y.p. 92. GARDNER, Rod: When Listeners Talk. n.y.p. 93. BARON, Bettina and Helga KOTTHOFF (eds.): Gender in Interaction. Perspectives on feminity and masculinity in ethnography and discourse. n.y.p.
