AUTHOR ""
TITLE "Meaning Through Language Contrast: Volume 1"
SUBJECT "Pragmatics & Beyond, New Series, Volume 99"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Meaning Through Language Contrast
Pragmatics & Beyond New Series Editor Andreas H. Jucker University of Zurich, English Department Plattenstrasse 47, CH-8032 Zurich, Switzerland e-mail:
[email protected]
Associate Editors Jacob L. Mey University of Southern Denmark
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
Catherine Kerbrat-Orecchioni
Hebrew University of Jerusalem
University of Lyon 2
Jean Caron
Claudia de Lemos
Université de Poitiers
University of Campinas, Brazil
Robyn Carston
Marina Sbisà
University College London
University of Trieste
Bruce Fraser
Emanuel Schegloff
Boston University
University of California at Los Angeles
Thorstein Fretheim
Deborah Schiffrin
University of Trondheim
Georgetown University
John Heritage
Paul O. Takahara
University of California at Los Angeles
Kansai Gaidai University
Susan Herring
Sandra Thompson
University of Texas at Arlington
University of California at Santa Barbara
Masako K. Hiraga
Teun A. Van Dijk
St.Paul’s (Rikkyo) University
Pompeu Fabra, Barcelona
David Holdcroft
Richard J. Watts
University of Leeds
University of Berne
Sachiko Ide Japan Women’s University
Volume 99 Meaning Through Language Contrast: Volume 1 Edited by K.M. Jaszczolt and Ken Turner
Meaning Through Language Contrast Volume 1 Edited by
K.M. Jaszczolt University of Cambridge
Ken Turner University of Brighton
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 Meaning through language contrast / edited by K.M. Jaszczolt and Ken Turner. p. cm. (Pragmatics & Beyond, New Series, issn 0922-842X ; new ser. 99-100) Includes bibliographical references and indexes. 1. Contrastive linguistics. 2. Semantics. 3. Pragmatics. 4. Grammar, Comparative and general. I. Jaszczolt, Katarzyna. II. Turner, Ken, 1956- III. Series. P134 M35 2002 401’.43--dc21
2002074427
isbn 90 272 5119 3 (Eur.) / 1 58811 206 3 (US) (Hb; alk. paper, vol. 1. 99) isbn 90 272 5120 7 (Eur.) / 1 58811 207 1 (US) (Hb; alk. paper, vol. 2. 100) isbn 90 272 5349 8 (Eur.) / 1 58811 332 9 (US) (Hb; alk. paper, set volumes 1-2.) © 2003 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
Table of contents
Acknowledgements
ix
Editorial preface Katarzyna M. Jaszczolt and Ken Turner
xi
Negation Distributional restrictions on negative determiners L. M. Tovena
3
Towards a comprehensive view of Negative Concord João Peres
29
Temporality On temporal constructions involving counting from anchor points – Semantic and pragmatic issues Telmo Móia
45
On the semantics and pragmatics of situational anaphoric temporal locators in Portuguese and in English Ana Teresa Alves
61
Remarks on the semantics of eventualities with measure phrases in English and Romanian Ilinca Cr˘ainiceanu
75
The present perfect in English and in Catalan Hortènsia Curell A contrastive reading of temporal-aspectual morphemes in Swahili: The case of ‘-li’ and ‘-me’ Frederick Kang’ethe Iraki
101
117
Table of contents
Modality Semantic and pragmatic constraints on mood selection Rui Marques Dilemmas and excogitations: Further considerations on modality, clitics and discourse A. Capone
129
147
Evidentiality Inferred evidence: Language-specific properties and universal constraints Sergei Tatevosov
177
Extension of meaning: Verbs of perception in English and Lithuanian Aurelia Usonien˙e
193
Perspectives on eventualities Information structure, argument structure, and typological variation Márta Maleczki
223
The network of demotion: Towards a unified account of passive constructions Andrea Sansò
245
Valence change and the function of intransitive verbs in English and Japanese Mayumi Masuko
261
The transitive/intransitive construction of events in Japanese and English discourse Patricia Mayes
277
Topics in grammar and conceptualization Towards a universal DRT model for the interpretation of directional PPs within a reference frame Didier Maillat
295
Table of contents
The interaction of syntax and pragmatics: The case of Japanese ‘gapless’ relatives Akiko Kurosawa
307
Constraint interaction at the semantics/pragmatics interface: The case of clitic doubling Javier Gutiérrez-Rexach
335
Cross-language commutation tests and their application to an error-prone contrastive problem – Ger. einige, Fr. quelques, Sp. algunos Eva Lavric
355
Language index
371
Name index
373
Subject index
375
Contents of Volume 2
381
TSL[v.20020404] Prn:23/08/2002; 12:08
F: PBNSACK.tex / p.1 (ix)
Acknowledgements
We are indebted to a number of people. Firstly, we would like to express our thanks to the Conference Secretary Marilyn Dowling for her invaluable help with organizing the conference at Newnham College, Cambridge of which this volume is an outcome. To Patrick Welche we are grateful for designing and maintaining the conference web site. Our thanks also go to Heather Wynn for her administrative assistance. To the Ph.D. students Luna Filipovic, Rachel Smith and Marina Terkourafi we are grateful for their help with running the event. The co-opted members of the Programme Committee, Robyn Carston and Victoria Escandell-Vidal contributed their excellent work selecting conference contributions: thank you! To Elsevier Science we are grateful for their sponsorship of a social event accompanying the conference. We would also like to thank Newnham College at Cambridge for hosting the conference and all the College staff for their contribution to the tremendous success of the event. Last but not least, our gratitude goes to Isja Conen of J. Benjamins for her encouragement and invaluable editorial assistance.
TSL[v.20020404] Prn:23/08/2002; 12:18
F: PBNSPR.tex / p.1 (xi)
Editorial preface Katarzyna M. Jaszczolt and Ken Turner
This collection brings together selected papers from the Second International Conference in Contrastive Semantics and Pragmatics that took place in Cambridge from 11 to 13 September 2000 and was organized by the editors. The five-year period that separated this event from the first conference on this theme (see Jaszczolt & Turner 1996) witnessed a significant turn of orientation. There is a conspicuous shift of emphasis from speech-act based studies predominant in the past to semantics-based approaches, of the truth-conditional or cognitive orientation. Research that adheres to speech act theory is in a significant minority and appears progressively ‘detheorized’ in its selective utilization of those aspects of the theory that facilitate discourse about language use. Furthermore, this collection testifies to the growing importance of diachronic analysis. There are numerous contributions in the category of historical semantics and pragmatics and for those, predictably, grammaticalization is the focal object of research. The contrastive, cross-linguistic perspective proves to be a successful way of researching historical processes in that it sheds more light on the universality and language-relativity of aspects of language change (see e.g., Nicolle, Visconti, this collection). The historical perspective is also documented in speech act theory research (see Jung & Schrott, this collection). Another significant shift of emphasis can be observed at the lexical/sentence semantics interface. The generally acknowledged return to studies of word meaning is corroborated in this collection in the papers that approach the lexicon from the cognitive perspective or the perspective of a generative lexicon, frequently through the use of electronic lexical databases (see e.g., Viberg). The role of information structure for truth conditions is the cutting edge of contemporary semantic research. This avenue is also represented in the collection (see e.g., Maleczki). Descending to more detailed shifts of emphasis, the noun phrase seems to be gradually giving way to the studies of the verbal phenomena such as modality, tense, aspect and voice. These topics are numerously and strongly represented (see e.g., Marques, Masuko, Sanso, Usoniene).
TSL[v.20020404] Prn:23/08/2002; 12:18
F: PBNSPR.tex / p.2 (xii)
Katarzyna M. Jaszczolt and Ken Turner
The discussion of the boundary between semantics and pragmatics has also undergone various changes of emphasis and style. In the 1970s, sense-generality and pragmatic inference were brought to the fore (see e.g., Cole 1981; Atlas 1989; Turner 1999; Jaszczolt 1999). Almost two decades later, the dynamic perspective in semantics allowed for contextual information to be semanticized (see Kamp & Reyle 1993). Subsequent developments of the idea of underspecification (see e.g., van Deemter & Peters 1996) demonstrated that pragmatics, intentions and intentionality are frequently irreducible and do not yield to formalizations (see Blutner & van der Sandt 1998; van Deemter 1998; Dekker 2002; Jaszczolt & Turner 2002). The predominance of semantic analyses strongly suggests that (i) contrasting meaning in various natural languages requires firm foundations, strict modelling and some degree of formalization; (ii) both (a) cognitive semantics and (b) Tarskian, post-Montagovian semantics supplemented with post-Gricean pragmatics are more productive than the offshoots of the ordinary language philosophy. Finally, to address the empiricism-rationalism dilemma, it can be observed that inferring from quantitative analyses and supporting theories by unquantified data constitute equally successful directions in semantic and pragmatic research.
References Atlas, J. D. (1989). Philosophy Without Ambiguity: A Logico-Linguistic Essay. Oxford: Clarendon Press. Blutner, R., & R. van der Sandt (1998). Editorial Preface to Special Issue on Underspecification and Interpretation. Journal of Semantics, 15, 1–3. Cole, P. (Ed.). (1981). Radical Pragmatics. New York: Academic Press. Dekker, P. (2002). Pronouns in a pragmatic semantics [Special Issue on Conceptual Contours at the Semantics/Pragmatics Interface]. Journal of Pragmatics, 34, 815–827. Jaszczolt, K. M. (1999). Discourse, Beliefs, and Intentions: Semantic Defaults and Propositional Attitude Ascription. Oxford: Elsevier Science. Jaszczolt, K. M., & Turner, K. (Eds.). (1996). Contrastive Semantics and Pragmatics, 2 volumes. Oxford: Pergamon Press. Jaszczolt, K. M., & Turner, K. (2002). Editorial Introduction to Special Issue on Conceptual Contours at the Semantics/Pragmatics Interface. Journal of Pragmatics, 34, 811–813. Kamp, H., & Reyle, U. (1993). From Discourse to Logic: Introduction to Modeltheoretic Semantics of Natural Language, Formal Logic and Discourse Representation Theory. Dordrecht: Kluwer. Turner, K. (Ed.). (1999). The Semantics/Pragmatics Interface from Different Points of View. Oxford: Elsevier Science. van Deemter, K. (1998). Ambiguity and idiosyncratic interpretation. Journal of Semantics, 15, 5–36. van Deemter, K., & Peters, S. (Eds.). (1996). Semantic Ambiguity and Underspecification. Stanford: CSLI Publications.
TSL[v.20020404] Prn:23/08/2002; 12:04
P I
Negation
F: PBNSP1.tex / p.1 (1)
Distributional restrictions on negative determiners L. M. Tovena Université de Lille
This paper discusses restrictions in the distribution and interpretation of negative determiners in English, Italian and French. The typology of determiners defined by Chierchia (1998) is taken as a starting point, and it is tested against new data. Several shortcomings are highlighted. In particular, it is shown that some determiners are sensitive to differences among members of the class of uncountable nouns, and that the distribution of a number of singular determiners cannot be captured by restricting their domain of quantification to all and only singular countable nouns.
.
The issue
This paper discusses restrictions in the distribution and interpretation of negative determiners in English, Italian and French. We take the typology of determiners defined by Chierchia (1998) as our starting point, and we test it against new data. In Chierchia (1998) a unique lattice model is built for the denotation of countable and uncountable nouns, under the name of Inherent Plurality Hypothesis. The main thesis of this unified analysis of nouns is that mass nouns differ from count ones (only) insofar as they come out of the lexicon with plurality already built in. A singular countable noun is associated with a set of atoms, and the set-forming operator PL is used to enable us to talk about sets of them. The basic lexical entry of a mass noun does not single out the set of atoms, but a sublattice. In this case, the difference between singular and plural is neutralised, for the noun applies equally to atoms and sets thereof. The atomic texture is foregrounded in countable nouns, as by definition its extension singles out a set of atoms. On the contrary, in uncountable nouns this structure is present but only implicitly. This gives one a single structure for describing the denotation of countable and mass nouns. This unified model is then used by Chierchia to analyse the distribution of determiners. Attention is paid to the issue of how to compose information coming
L. M. Tovena
from the morphological number of the N, the count/noncount distinction and the determiner characteristics in order to obtain truth conditional satisfaction in the appropriate cases. The structure of the paper is as follows. We will first look at the case of the English unrestricted determiner no, then at the singular Italian determiner nessuno (no) together with French aucun (no). Then, determiner-like uses of Italian niente (no) will be discussed. Finally, this type of occurrence of niente will be contrasted with uses of nessuno-phrases in verbless clauses.
. Unrestricted determiners The standard definition of the English determiner no in Generalized Quantifier Theory is as in (1). (1) NO (X)(Y) = 1 iff X ∩ Y = Ø This determiner is called unrestricted (Chierchia 1998) because it applies to singular and plural count nouns, as well as to uncountable nouns. In this case, the variation in number and mass/countable nature is said not to affect the interpretation of the determiner, which is always expressed as a constraint of empty intersection of the first argument with the second. In contrast, in the case of the, another unrestricted determiner, the same variation has a broader impact. There is a uniqueness presupposition when this determiner combines with a singular noun, and a presupposition that there be more than one element when it combines with a plural noun. Chierchia ascribes to Schwarzschild (1991) the claim that a theory of plurality which uses set theory gives the wrong results in the case of (1). For a sentence such as no men lifted the piano, such an analysis would require that no men did, but leave open the possibility that single men did. Similar considerations apply for cases where no combines with a singular noun. In sum, if the head noun of the quantified NP is singular and count, as in (2), the proposition will be false when a student hit a lorry, but it should also turn out to be false when a group of students does. Similarly, (3) is expected to be false also when a single student hit the lorry. The two sentences are truth conditionally equivalent, and sentence (4), for instance, is contradictory. (2) No student hit a lorry (3) No students hit a lorry (4) *No students hit a lorry but Daniel did In order to treat the question of the unrestricted distribution of no and its seeming insensitivity to morphological information, that he believes to be the cause of the
Distributional restrictions on negative determiners
contradictoriness of (4), Chierchia (1998) puts forward the definition in (5), which is cast under his Inherent Plurality Hypothesis. (5) NO (X)(Y) = π(X) ∩ Y = Ø For any set u in U, the ideal π(u) generated by u is {x : x ≤ u}, where ≤ is the ‘component of ’ relation that orders the structure used to represent nouns.1 For any set X, π(X) is the set of all the elements which are components of the supremum of X. So, if X is the denotation of a mass noun, π(X) = X, because X already contains all the components of its supremum. On the contrary, if X is singular count, π(X) yields the closure of X under the operation of ‘sum’ defined in terms of ≤. Finally, if X is plural, π(X) will add the atoms to it. In any case, π(X) yields a complete atomic sublattice of the domain. In this way, (2) and (3) are accounted for as well as the fact that no works on uncountable, see (6) and (7). As a short aside, note that Chierchia touches only very briefly on the distinction between concrete and abstract mass nouns. For the latter, he proposes quanta identified with instantaneous states in the case of ‘honesty’ or with unspecified units in the case of ‘sense’, and says that abstract mass nouns denote sets of these quanta closed under sum. (6) No water flooded this area (7) There was no honesty in his words In other words, π(u) allows one to get all the components of u independently of the constraints that number and the count/uncountable distinction impose on the denotation of N. It is as if information provided by morphological number were no longer available at the level of the quantified NP. Actually, Delfitto (1998) goes as far as claiming that the difference in number in a no-phrase is devoid of any interpretive import. However, there is some evidence supporting the interpretive relevance of morphological number in unrestricted negative determiners such as no.
. The relevance of number In this section, we review some evidence in favour of the interpretive relevance of number information. First, it is true that no combines with singular and plural count nouns, but it is also true that the resulting quantified NPs are not always free alternates. (8) No man is immortal (9) No buses are running today Examples (8) and (9) are presented to highlight the fact that there is a preference for using no Nsingular when making a generic statement, whereas no Nplural is used in
L. M. Tovena
contingent statements. It is not obvious how this situation may be captured if one takes (5) as saying that the denotation of the first argument is always ‘normalised’ by π and number differences are hidden as a result. A second case where number differences cannot be claimed to be devoid of interpretive import has to do with the individuation of references and presupposition. Consider examples (10)–(11). (10) There will be no train tomorrow (11) There will be no trains tomorrow Sentence (10) contains a no-phrase with a singular countable noun and can be used felicitously only to deny tomorrow’s running of a single train. Furthermore, the existence of a train is presupposed and its ‘identity’ is discourse old and accessible. On the contrary, (11) cannot be used to deny the running of a single train, and although it may be taken to presuppose the existence of trains, as it is the case for the restrictor of a generalised quantifier, it does not impose requirements with respect to their identification. A third case is about coreference and discourse anaphora. Consider the sentences in (12). (12) a.
No students attended the meeting. *Each of them had something else to do. b. No student attended the meeting. Each of them had something else to do. c. No students attended the meeting. All of them had something else to do.
The anaphoric link established by the quantified NP in the second clause cannot be resolved if no coreference can be established with respect to the denotation of the N in the quantified NP in the first clause. In (12a) coreference does not seem to be possible, whereas it does in (12b, c). This variation in acceptability confirms the observation that number remains visible. Only no Nsingular gives access to the set of individual students, which is the restricted complement of the denotation of the quantified NP. Finally, no-phrases behave differently with respect to collective predicates, as shown in (13). As discussed in Löbner (2000), quantified NPs with plural noncollective nouns admit collective predicates in the second argument, as shown in (13b), whereas quantified NPs with singular nouns do not, as shown in (13a). In Löbner’s analysis this difference follows from the fact that only the former type of phrase yields a quantification for which the predicate in the second argument is defined.
Distributional restrictions on negative determiners
(13) a. *No student gathered in the courtyard. b. No students gathered in the courtyard. The visibility of number information from the head noun could be enhanced within Chierchia’s framework by assuming that the part of structure made visible by π remains backgrounded, and that the original denotation is the only foregrounded part. However, giving different statuses to the original denotation of the noun and to the part added by π goes against the original idea of ‘normalising’ all the cases, and nullifies the purpose of π.
. Singular determiners Not all negative determiners are unrestricted. Some of them exhibit restrictions related to the structure of the domain. The class of singular determiners gathers together determiners that require a domain composed only of atoms. Chierchia defines the function SG to check whether a predicate foregrounds a set of atoms or not. SG maps each possible noun denotation into its presupposed singularities. If A is already a set of singularities, then SG(A) = A. Otherwise, SG(A) is the set of atoms At that generates A via the set-forming operator PL, if such a set exists. Therefore, since SG(A) is undefined if A is the denotation of a mass noun, SG works as a tool for setting apart countable from uncountable nouns. The function SG acts as domain regulator for determiners according to the schema in (14a), abbreviated as (14b). (14) a. D(SG(X))(Y) b. DSG (X)(Y) Then, S is defined as a restriction of SG to atoms. It sets apart singular from plural countable nouns. For any subset X of the domain, S(X) = X if X ⊆ At, and otherwise S(X) is undefined. Using S, one can capture the distribution of determiners such as the English every, classified as an instance of singular determiner and defined as follows. (15) EVERYS (X)(Y) = S(X) ⊆ Y Chierchia assigns also the Italian negative determiner nessuno2 (no) to the group of singular determiners, together with positive qualche (some), but does not provide a definition for the former. The data in (16) show that nessuno combines with singular count nouns but not with plurals or uncountables and seem to offer support to his classification.
L. M. Tovena
(16) a.
nessuno studente no student b. *nessun libri no books c. *nessun vino no wine
Considering the data in (16) and the contradictory status of (17), and working from (5) and (15), one can get (18) as a potential definition for nessuno. (17) *Nessuno studente ha colpito il camion ma Mario e Daniele sì no student hit the lorry but Mario and Daniel did (18) NESSUNOS(X)(Y) = π(S(X)) ∩ Y = Ø
(putative)
However, as soon as one takes into account a little more Italian data, it appears not only that (18) is problematic, but also that a characterisation of nessuno as standard singular determiner is not as straightforward as it would appear from the discussion in Chierchia (1998).
. Nessuno with mass nouns First, contra Chierchia’s claim, nessuno applies to uncountable as well as count nouns, cf. (19). (19) Non ha mostrato nessun coraggio She didn’t show any courage at all (20) ?Nessun’acqua laverà il suo sangue dalle tue mani no water will wash her blood from your hands Sentence (20) is admittedly marginal, and has required a careful wording in order to reach so much as this level of acceptability. On the contrary, (19) is perfect in the reading ‘not the least bit of ’, i.e. a truly mass reading. In both cases we leave aside the taxonomic readings, which seem always to be possible, although rather awkward in the case of (19). Therefore, the restriction S does not seem to be appropriate for capturing the distribution of nessuno, because it rules out acceptable cases such as (19), where nessuno combines with an uncountable noun. The problem is that S cannot be defined in a different way, in the system as it is. This is not a minor point. Indeed, Chierchia claims that there are no functions restricted to singular count and mass noun denotations in natural language, and says that the existence of such a gap follows from his system because there is no natural domain regulator that would have the effect of restricting the left argument of a determiner in this way. Recall that nessuno does not combine with plurals,
Distributional restrictions on negative determiners
and that in his analysis plural and uncountable nouns come down to the same for the issue in hand. In the light of the evidence provided in (16) and (19), his classification appears to break down. Second, at first sight the distribution of nessuno does not seem to correspond to a consistent behaviour. This determiner sometimes combines with a mass noun, as in (19), and sometimes does not, as in (16c). This variation rules out the possibility of capturing the distribution of nessuno by referring to a constraint expressed purely in terms of morphological number. However, on a closer look, it can be seen that the line separating acceptable and unacceptable instances follows rather closely the abstract vs concrete divide. There is a preference for abstract mass nouns. As noted above, (20) is not perfect. In general nessuno does not combine with concrete mass nouns of the substance mass type, e.g. ‘milk’ and ‘water’, or of the collective mass type (Bosveld-de Smet 1998), i.e. ‘mail’ and ‘furniture’. (21) a. *Non ha messo nessuno zucchero nella spremuta she didn’t put any sugar in the juice b. *Non ha ricevuto nessuna posta She didn’t get any mail However, this distinction among types of mass nouns cannot find a place in Chierchia’s classification. Note that, with respect to uncountable nouns, the behaviour of nessuno clearly differs from qualche, the other Italian singular determiner mentioned in Chierchia (1998), as the latter cannot combine with them at all, as shown in (22). (22) a. *Questa impresa ha richiesto qualche coraggio this entreprise required some courage b. *Ha messo qualche sale nella minestra she put some salt in the soup It could be objected that these data do not really falsify Chierchia’s classification. It could be said that in (19) there is an implicit occurrence of a type-shifting device, that makes available a reading of the type ‘degree-of ’ or ‘act-of ’. However, this remark would not give any contribution towards clarifying the contrast between (19) and (22), since it would not explain why a similar device is not equally available for (22a). Third, note also that nessuno is not the only determiner to exhibit a non homogeneous behaviour with respect to uncountable nouns, to split this class into two groups, and to cluster abstract uncountable nouns together with singular countable nouns. French provides more examples, as shown in (23) for the negative determiner aucun.3
L. M. Tovena
(23) a.
Il n’a pris aucun livre He did not take any book b. Il n’a montré aucune pitié He didn’t show any mercy c. *Il n’a vu aucun étudiants He didn’t see any students d. *Il n’utilise aucun sable He uses no sand
Furthermore, this sensitivity to the typology of uncountable noun is not a peculiarity of negative determiners. As shown in (24), the negative polarity and free-choice French item le moindre4 (the least) behaves in the same way in this respect. (24) a. b. c. d. e.
Il n’a pas lu le moindre livre he did not read any books Il connaît le moindre recoin du village he knows any corner Il n’a pas montré le moindre courage He did not show the least bit of courage *Il n’a pas lu le moindre livres/les moindres livres He didn’t read any books *Il n’a pas bu la moindre eau He did not drink a drop of water
A first conclusion that can be drawn from the preceding discussion is that Chierchia’s domain restrictor S does not work for the cases in hand, while a restrictor predicted to be impossible seems to be needed. Then, this need for an impossible S could be taken to point out that the whole model does not work since it is unable to distinguish among different types of mass nouns. Otherwise, less drastically, it could be taken to suggest that the interaction with the mass/count distinction – i.e. Chierchia’s noun properties 4 to 7 – is not sufficient for a proper classification of determiners. The idea of modifying the basic ontology is too costly a solution, in our opinion, to be further explored as the first choice, and we would rather leave it as a last resort option. The definition of S and the interaction between determiners and the mass/count distinction can be considered as two sides of the same coin. This section closes with the question of how to characterise the subset of uncountable nouns that combines with nessuno, aucun and le moindre. The idea of abstract noun, that provides a first intuition for singling out this subset, cannot be captured simply by requiring that the members of this subset be devoid of the spatial dimension, as shown by the unacceptability of (25).
Distributional restrictions on negative determiners
(25) a. *Non ho nessun tempo I have no time b. *Je n’ai aucun temps Tovena (2001) puts forwards an account for the split within mass nouns and the association of part of it with count nouns which can be exploited for the data in (16)–(19) and would come at a minimal cost to Chierchia’s model. We discuss it in the next section.
. Abstract nouns The subset of mass nouns compatible with nessuno and aucun can be characterised via the notion of grandeur intensive (intensive quantity)5 (Van de Velde 1996), i.e. entities that manifest themselves in degrees. Van de Velde provides a set of contexts where this type of nouns behave differently from the rest of the class. For instance, they are compatible with un certain (a certain), like countable nouns but unlike the rest of uncountable nouns. They cross over the quantity/quality distinction, cf. (26), – which suggests that degrees do not qualify as traditional quanta and are not like measures that map parts into numbers. (26) a.
quanto coraggio! = che coraggio! what courage b. quanto burro! = che burro! what a lot of butter = what (good) butter
Furthermore, they are only indirectly quantized. For instance, courage is a property expressed via courageous acts, which, in turn, are quantized into eventualities instantiating them. Similarly, the quanta ‘instantaneous state’ proposed by Chierchia applies to ‘honest behaviour’, i.e. a manifestation of ‘honesty’, and not to ‘honesty’ itself. Tovena (2001) explains the double behaviour of intensive quantities by building on these observations. The classification of intensive quantities as mass nouns is derived from the fact that their lexical entries are not directly associated with the atoms in their denotations. Degrees of intensive quantities are quanta that do not discretise the domain into individuals, nor into measurable parts. As Kant says: Thus a certain expansion which fills a space, for instance, heat, and every other kind of phenomenal reality, may, without leaving the smallest part of space empty, diminish by degrees in infinitum, and nevertheless fill space with its smaller, quite as much as another phenomenon with greater degrees. (Kant 1881: 153)
L. M. Tovena
So, a lot of courage is not a bigger (extensive) quantity of courage, it may well be the same quantity at that, but it is a bigger intensive quantity, i.e. a higher degree of intensity of courage. At the same time Tovena also derives their distribution somehow close to countable nouns, as far as the combination with determiners is concerned, because manifestations, such as courageous acts, are sort of instances of the entity ‘courage’. A high degree or, for that, any given degree of courage are different types of courage. The domain is discretised by weakly discrete units interpreted as degrees, and this is the default way for intensive quantities. For countable nouns the default way is a discretisation by individuals, which are strongly discrete units. For the rest of uncountable nouns, the default strategy is by parts, defined with respect to units of measure, and by species only as a secondary option that applies to rescue a phrase when no measure is provided or recoverable from the context. For intensive quantities, no units of measures are available, because measures apply to extensions. It appears that the level of discreteness provided by weakly discrete units is enough to satisfy the requirement of a singular determiner such as nessuno and aucun. On the one hand, it provides a partitioning that can be exploited by the domain restrictor S. On the other hand, strongly discrete units are not needed because we are dealing here with negative quantifiers, whose witness set (Barwise & Cooper 1981; Szabolcsi 1997) is empty by definition. Hence, at no stage of the interpretation will a particular occurrence have to be singled out. Therefore, the acceptability of (19) and (23b) is the result of the interaction between a weakly discretised domain and a determiner whose denotation can be represented by a constant function. The situation changes if the determiner is not a constant function. In the case of the positive singular Italian determiner qualche, for instance, some occurrences would have to be singled out despite the fact that they are not lexically accessible, because the witness sets have cardinality greater than zero. As a result the quantified NP cannot be built, cf. (27). (27) *qualche coraggio some courage
. First partial conclusion In conclusion, by assuming that the mass/count distinction is based on strongly discrete units, and that quantifiers may be satisfied with weakly discrete units in certain cases, one can rescue Chierchia’s domain restrictor S. However, his description of the interaction of the determiner system with the mass/count distinction is confirmed to be insufficient for a proper classification of determiners.
Distributional restrictions on negative determiners
The line of reasoning adopted in the present paper seems promising also with respect to the problem raised by sentences such as (28), predicted to be impossible by Chierchia’s account. (28) She still had every confidence in him as a mathematician. In this case, the quantified NP has a unique witness set that includes the whole denotation of the noun. Such a set may be identified without making use of strongly discrete units as it doesn’t partition the denotation of the noun. With respect to le moindre, cf. the sentences in (24), although this determiner is not a constant function strictly speaking, it points to an end-of-scale position and can be equated with a universal quantification on the whole scale and, depending on the directions of the inferences on the scale, turns out to be a sort of universal or negative quantifier.
. Niente N phrases In the remainder of this paper we will discuss more data on negative determiners in Italian. We will start off from a remark on nessuno made by Chierchia while discussing the predictions of his classification of determiners. He observes that there is no principled way to predict if a language has an unrestricted function or a set of restricted functions to cover a given space. Furthermore, there may be gaps. He says that Italian has no mass or plural counterpart of nessuno. The ‘mass side’ of nessuno has been touched upon in Section 3, where it has been shown that the distribution of this determiner is broader than predicted by Chierchia. However, a second option is sometimes exploited in Italian. The expression niente (nothing), usually working as a quantifier, in certain cases seems to work also as a determiner. In these cases it is interpreted as no. Some examples are provided in (29) to (31). (29) a.
Niente età limite per le adozioni no maximal age threshold for adoptions b. Niente pensione a chi risiede all’estero no pension to the citizens living abroad
(30) a.
(25/7/1996IM)
Per questa torta ci vogliono tre etti di farina, un uovo, due mele e niente zucchero for this tart one needs 300gr of flour, one egg, two apples and no sugar b. Daniele mangia pesche, pere, niente mele e poca uva Daniel eats peaches, pears, no apples and a few grapes
L. M. Tovena
(31) a.
Non ho niente voglia I have no desire (i.e. I don’t feel like it at all) b. Non fa niente freddo it is not the least bit cold
In these cases, niente combines with bare nouns or N , as shown in (32) with an adjective in pre and post nominal position, and does not tolerate the presence of any other determiner, see the selection of determiners in (33). (32) a.
niente piccoli passi no small steps b. niente palloni rossi no red balls
(33) a. *niente l’acqua no the water b. *niente un libro no a book c. *niente questo libro no this book d. *niente tre libri no three books The examples in (29) to (31) might involve some more structure than a simple DP. The literature, to the best of our knowledge, has largely ignored these cases. In this paper, pending a syntactic analysis of these structures, we will adopt the minimal option consisting of treating them as quantified NPs, i.e. DPs. In this view, niente behaves as an unrestricted determiner, as it combines with singular and plural countable nouns as well as uncountable ones. In these uses, it may be taken to complete the coverage of negative quantification in Italian. Examples (29) to (31) instantiate three different cases. By far, the most common case is the one presented in (29). A first important question raised by this type of occurrences of niente is why should the distribution of this phrase be restricted almost exclusively to verbless sentences. Again, the literature does not offer an explanation for this behaviour. Examples such as (29a) and (29b) are very shortly, if ever, discussed. It has been proposed in Manzotti (1991) to treat them as cases of ellipsis of the verb. It is said that (29a) is equivalent to something close to (34), but the reason motivating the relation is not discussed. This analysis would require the additional step represented by establishing a relation of transformation or deletion that would hold between niente and the determiner nessuno or a surface null element, so that (29a) can be related to (34), or to (35) where there is no determiner. This is a crucial step which is left unexplained.
Distributional restrictions on negative determiners
(34) Non ci sarà nessuna età limite per le adozioni there will be no limit of age for adoptions (35) Non ci sarà età limite per le adozioni there will be no limit of age for adoptions It is also not explained why verb ellipsis is obligatory in (29). For these ‘reduced’ clauses no ‘full’ clauses are possible. Furthermore, an ellipsis implies that the thing being omitted has already been expressed, which does not seem to be the case for the verbs in these clauses. Third, niente and nessuno phrases in verbless clauses undergo a particular thematic restriction which so far seem to have gone unnoticed. Neither of these phrases can discharge what would be the agent role in the corresponding clause with no verb ellipsis. When needed, this role is discharged by a PP, as shown in (36) and (37), which makes us suppose that a passive verb form was ‘left out’. Such a restriction does not follow from ellipsis. (36) Dal Csm nessuna censura al procuratore Vigna no reproof [is made] to attorney Vigna by the CSM
(19/2/1998IM)
(37) Niente scherzi dalla Bundesbank no tricks [were played] by the Bundesbank
(22/3/1998IM)
The second case, exemplified in (30)–(31), involves niente-phrases inside a conjunction. Depending on the preferred analysis of coordination, they might also be taken as more instances of ellipsis, if one stress the following fact. On the one hand, in both (30a) and (30b) niente occurs within a sequence of NP conjuncts6 that are all acting as direct objects of the verb occuring in the same sentence. On the other hand, however, it is the whole conjunct structure that instantiates the argument position and not the single conjuncts, and there is no verb inside the sequence. Note that there is no constraint on the linear position of the conjunct containing niente in the sequence. Middle and final position are possible, as we have seen in (30b) and (30a) respectively, but the initial position is not excluded either, as shown in (38). (38) Se si passa poi dalla quantità dei lavoratori alla qualità della loro vita, ci si addentra in un roseto niente petali e tutto spine: è impossibile programmarsi serenamente la vita sui tempi lunghi, essendo il rinnovo dei loro contratti – in media ogni anno – legato necessariamente agli appalti che le imprese o cooperative da cui dipendono riescono a vincere. (13/12/2000IM) Then, if [the discussion] moves from the number of workers to the quality of their lifes, we enter a rose garden [where there are] no petals and all thorns: it is impossible for them to plan their lives calmly on a long time
L. M. Tovena
span, as the renewal of their contract – about once a year – is necessarily linked to the contracts by tender that the companies or cooperatives employing them manage to get. The sentences in (30) and (38) are special insofar as they do not satisfy the constraint of negative concord. Simplifying a little, negative concord in Italian requires negation to be expressed in a preverbal position for the sentence to be allowed to contain N-words in postverbal positions (Corblin & Tovena 2001). However, it appears that this constraint is waived when a conjunction in postverbal position contains positive and negative conjuncts, in which case the verb would be expected to be both negated and not negated depending on the polarity of the conjunct, as discussed in Tovena (1996). Example (38) brings in a novelty element in the discussion because it contains an instance of negative conjunct occurring as the first element of a sequence. A constraint that appears to be satisfied in all these cases of conjunction, and for which we cannot offer an explanation so far, applies to the type of noun that enters these niente-phrases. This type of context does not seem to be suitable for a niente Nsingular count phrase, as shown in (39), contrary to the general situation described in Section 6. (39) ??Daniele mangia pesche, pere, niente mela e poca uva Daniel eats peaches, pears, no apple and a few grapes If interpretable at all, (39) talks about the type of food ‘apple’, and not about the fruits, as it is done for pears and peaches. Finally, as for examples such as (31), they could be dismissed as cases where niente works as a degree adverb that strengthens the predication, in alternative to the analysis as determiner. In other words, (40a) would be considered as a paraphrase of (40b) – a hypothesis that would go against Molinelli’s (1988) claim that modern Italian does not have this type of adverbial use – rather than of (40c). So far, we do not see strong evidence in favour of either analyses. (40) a.
Non ho niente fame I am not a bit hungry b. Non ho affatto fame I am not hungry at all c. Non ho neanche un po’ di fame I am not a bit hungry
We leave structural restrictions on the distribution of the determiner-like niente for further research. In the remainder of this paper we concentrate on verbless clauses, because they are the most frequent cases. They are often used in headings and titles, but may occur in the body of an article too, as shown in (41).
Distributional restrictions on negative determiners
(41) Il testo più bello comunque è quello che elogia il cattivo carattere sul luogo di lavoro. Niente sorrisi e cortesie, ma solo espressioni di malcelato fastidio per un ambiente dove le buone maniere sono solo la manifestazione di un’adesione alla gerarchia e l’accettazione di una espropriazione del proprio tempo da parte del capitale. (12/4/1998IM) Anyway, the best text is the one that celebrates the fact of being an illnatured person on the working place. No smiles or courtesies, but only expressions of half hidden annoyance for an environment where good manners are only a way to show that one endorses the hierarchy and accepts to lose control over one’s own time in favour of the capital. Furthermore, verbless clauses are the type of construction where niente can occur in partial alternation with nessuno. In the examples seen so far, both forms have been glossed as ‘no’, and are interpreted as imposing a constraint of empty intersection between the first and second argument. For want of a full analysis of niente-phrases that would expose its peculiarities, the study of its alternance with nessuno-phrases may provide some help to understand its functioning.
. Negative phrases in verbless clauses As it will soon be apparent, there are partial overlaps in the distribution and use of nessuno N and niente N phrases. One of the aims of this section is to discuss the cases for which often native speakers point out that occurrences of nessuno can be replaced with niente and vice-versa, with no apparent change in meaning, despite possible number modifications, and cases where the two types of phrase are not alternate. What makes this relatively free alternation of particular interest for us is the fact that it does not show up where one might expect on the basis of the equivalence between uncountable and countable plural nouns expressed in Chierchia’s system. In the following, first we will review the cases where there are overlaps, and then we will discuss the differences that are perceived in the use of the remainder of the forms.
. Overlaps in the distribution We saw that nessuno combines only with singular nouns. Niente is unrestricted. Tests on native speakers have shown a moderate possibility of replacing these Nwords with one another in their uses as determiner in verbless clauses. The examples are taken from the national press. Let us start from occurrences of nessuno that can be substituted by niente. In (42) the substitution is possible provided the
L. M. Tovena
countable noun is changed into plural form. In case there is an uncountable noun, such as in (43), the replacement does not force a change in morphological number, but there is a further restriction discussed below with respect to examples (65) and (66). (42) Nessuna reazione invece dalla presidenza della Repubblica(19/4/1998IM) but no reactions from the president of the republic (43) Nessuna meraviglia quindi se il discorso ha mandato su tutte le furie i repubblicani e si è guadagnato le lodi di non pochi esperti che hanno parlato di capolavoro politico concepito per la conquista del centro. (29/01/1997L’U) Then no surprise that the speech infuriated the republicans and succeeded in being praised by not a small number of experts who said it was a political masterpiece conceived for conquering the centre. The picture that emerges from the answers provided by native speakers can be summarised via the following schema. Nessuno-phrases might be changed into – niente+plural N – and into niente+singular N if the head noun is abstract (some type) Next, let us consider substitution in the opposite direction. Native speakers do not accept substituting nessuno for any naturally occurring instances of niente combined with singular nouns. The substitution in (44) would cause a shift from a sentence that conveys a command forbidding the envoy of the army, to a sentence that would record a situation of absence of any of the armies, with a partitive interpretation not available in the original form. In the case of (45), substituting the forms is not possible simply because nessuno does not combine with concrete uncountable nouns. (44) ANDREATTA Niente esercito in Puglia, altri blitz se sarà necessario (7/3/1997IM) Andreatta [said]: no army in Puglie, more blitz if needed (45) Ieri è stata rispettata la tregua, niente letame sulle autostrade o trattori sui binari (29/11/1997IM) [...] no manure on the motorways nor tractors on the rails In the case of niente combined with a countable plural noun, the substitution is possible. Example (46) would be equally possible with nessuno. (46) ONU Francia, Russia e Cina: niente mozioni contro l’Iraq (7/9/1996IM) Onu France, Russia and China: no motions against Iraq
Distributional restrictions on negative determiners
The picture that emerges from the answers of native speakers with respect to this second type of replacement can be summarised via the following schema. – –
Niente+singular N cannot be replaced by nessuno-phrases, except for abstract Ns (some type) Niente+plural N can be changed into nessuno-phrases
The first striking observation that can be made on the basis of these schemata is that meaning overlaps do not concern niente-phrases containing singular countable nouns. This is all the more surprising if one recalls that singular countable nouns make up the largest area of overlap in the distribution of the two items from the morphosyntactic point of view. A second observation, connected with the previous one, is that, when applied to a domain of atoms, these two negative quantifiers contribute different facets of meaning. In the following subsection, we give a closer look at the differences in the distribution to see if they provide hints for the characterisation of such facets. But, before that, let us add just one more observation on coordination. It was noted above, with respect to (30), that niente-phrases can occur as conjuncts. In the case where the coordination is inside the phrase, that is to say that the coordinated elements are the nouns niente applies to, as in (41), we observe a shift from conjunction to disjunction whenever niente is replaced with nessuno, as shown in (47). (41) Niente sorrisi e cortesie, ma solo espressioni di malcelato fastidio [. . . ] (12/4/1998IM) No smiles or courtesies, but only expressions of half hidden annoyance (47) Nessun sorriso o cortesia . . . At the present moment, it is not clear whether this shift is due to the constraint that nessuno, as a singular determiner, does not combine with a plurality – a hypothesis that requires the supplementary assumption that conjunctions denote only pluralities. It could also be ascribed to structural constraints, given the unacceptability of (48), but note that (49) is a little marginal but still possible, and has only the interpretation whereby there isn’t an entity that shares both properties of being white and being black, with e expressing set intersection as expected. On the contrary, no noticeable differences are reported between the cases of conjunction and disjunction with niente, cf. (41) and (50), as if intersection and union made no difference. Thus, nessuno appears to sharpen the logical interpretation of the connectives. (48) *nessun sorriso e cortesia (49) ??nessun bianco e nero no white and black (e.g. chequered or striped combination) (50) niente sorrisi o cortesie
L. M. Tovena
. Differences This section looks at the differences in the distribution of nessuno and niente in verbless clauses. We begin by comparing phrases with countable nouns, then we look at cases with countable nouns with both countable and uncountable readings, and we conclude with mass nouns.
.. Countable nouns When discussing sentence (44), it has been pointed out that different presuppositions arise from the use of nessuno and niente. The examples in (51) and (52) replicate the same opposition. The difference is not simply between presence versus absence of a presupposition of existence. Rather, in the case of presence, the presupposition of existence seem to be more precisely characterised as a presupposition of plurality, in the sense that the cardinality of the domain of quantification is expected to be > 1. In (44), as well as in (51), there is a presupposition of unicity with respect to the denotation of the noun. No such presupposition arises in the case of (52). On the contrary, here nessuno is used to exclude any interpretation of the resolution that could be taken as providing a mandate for military intervention in (52a) and (52b). (51) Bossi: “Niente Lega alle elezioni padane” Bossi said: No Lega at the elections in ‘Padania’
(20/7/1997IM)
(52) a.
KOFI ANNAN: Nessun mandato per colpire l’Iraq. Il segretario generale dell’Onu ha gettato acqua sugli ardori bellici degli Stati Uniti sostenendo che in caso di una infrazione irachena dell’accordo del 22 febbraio scorso sulle ispezioni ai siti presidenziali, la risoluzione approvata dal Consiglio di sicurezza non concede affatto agli Usa il cosiddetto “grilletto automatico”. (10/3/1998IM) Kofi Annan says that there is no mandate for hitting Iraq. The UN secretary-general threw some water on the United States’ burning war fervour, claiming that were Iraq to infringe the agreement on the inspections to the presidential sites dated 22 February, the resolution approved by the Security Council does not give the USA any so-called ‘automatic trigger’ b. Russia, Francia e Cina: “Nessun automatismo” (4/3/1998IM) Russia, France and China: No automatic right [to intervene]
Second, nessuno is marginal with nouns identifying unique entities, what we could call Russellian iotas. Sentence (53a) is the standard way of talking about a given type of weather. In our galaxy there is only one sun, and niente can be used to deny the presence of this entity in a particular context. On the contrary, (53b)
Distributional restrictions on negative determiners
brings in the presupposition that there is more than one sun, which, in order to be accommodated, forces a reading whereby the sentence conveys a claim about an astronomical situation that exceeds our galaxy. (53) a.
Niente sole quest’estate no sunshine this summer b. ??Nessun sole quest’estate none of the suns this summer
There is a similar contrast in French. (54) a.
??Aucun soleil cet été
none of the suns this summer b. Pas de soleil cet été no sunshine this summer It could be objected that (53b) and (54b) are not real occurrences of a countable noun, since their meaning is close to ‘sunshine’ or ‘sunny weather’. But even if one assumes that they contain instances of an uncountable noun, this does not explain why the interpretation as uncountable is the only one available. Furthermore, it would leave unexplained the fact that the shift to this type of reading cannot be escaped. The acceptability of (55) suggests that the presupposition of non unicity need not be strictly enforced considering the spatial and temporal dimensions. In fact, (55), like (53a), introduces a unique entity in the domain of discourse, but, unlike (53a), it is possible to exploit a variation along the temporal dimension. Since the pope is an entity who has been ‘instantiated’ by more than one individual across the centuries, there is no strict uniqueness and nessuno becomes acceptable. (55) Nessun papa ha difeso questa causa no pontiff defended this cause If the temporal dimension is also closed, as in (56a) where the progressive verbal form blocks the generic reading, then we record a difference in the function of nessuno. In (56a), it adds emphatic force, and the sentence is used felicitously as a reply. In (56b), which is a verbless clause, it is interpreted as a quantifier with its presupposition of non-uniqueness, and it is awkward. (56) a.
Nessun papa sta difendendo questa causa no pontiff is defending this cause b.??*Nessun papa per questa causa no pontiff for this cause
L. M. Tovena
The felicity of (56a) suggests that the presupposition of non-uniqueness conveyed by nessuno, what may also be called its partitive reading, is a pragmatic constraint, possibly triggered by the verbless construction. It is interesting to note that (19), the other instance of emphatic interpretation of nessuno we have met in this paper, also occurs in a tensed clause. Finally, niente is selected for non referential uses and event names. For instance, the issue in (57) is whether the person who is asked has or hasn’t an instance of authorisation, knowing that any instances would do. The question in the negative form is asked if no authorisation is produced (about presuppositions in negative questions see Ladd 1981; Gutiérrez-Rexach 1996), but only niente can be used felicitously in this context. (57) A policeman asking people at the entrance of a building under surveillance: a. ??Nessun permesso? none of the authorisations? b. Niente permesso? don’t you have the authorisation? In (58) and (59) too the effect of the use of niente is to block any specific reading of the noun. (58) a.
Decisione d’un parroco. Gay in chiesa: niente Messa (13/01/1997ICDS) the decision of a priest: if there are gay people in the church, there won’t be Mass
(59) a.
Il Papa è malato, niente udienza the Pope is ill, no audience
(06/02/1997LS)
Niente-phrases can be perceived as emphatic in the sense that the denial of existence that they can express may be used to convey an interdiction. Nessuno-phrases are preferred for expressing contingent information. This difference is highlighted by the minimal pair presented in (60). Sentence (60a) reports Cernomyrdin’s successful opposition to Lebed’s getting an office of vice-president in Russia. Recall that the office of vice-president does not exist in Russia. On the contrary, (60b) just says that the (existing) presidency of the committee has not (yet) been assigned. (60) a.
Cernomyrdin a Lebed: niente vicepresidenza (05/07/1996LU) Cernomyrdin [said] to Lebed: no vice-presidency b. Commissione di Vigilanza: nessuna presidenza (06/08/1997RR1) [as for the] Commissione di Vigilanza: no presidency [has been assigned]
Distributional restrictions on negative determiners
.. Double readings As we have seen, the substitution of nessuno for niente, and vice versa, may produce a switch between two possible interpretations, namely an interpretation of the noun as countable or uncountable. A regular feature in this alternation is the constant pairing of niente with the uncountable reading, as shown in (61a), and of nessuno with the countable reading, as shown in (61b). (61) a.
Niente formaggio alla fine del pasto no cheese at the end of the meal b. Nessun formaggio tra i cibi vietati no type of cheese among the food that has been banned
The difference in the reading associated with the noun has consequences on the interpretation of the whole clause. If we replace the determiners in (61) as done in (62), this causes a change in the interpretation that is not always easily rendered by the glosses. The English gloss of (62a) is slightly imprecise, because the example is not really about types of cheese, rather it is about instances of cheese, which possibly but not necessarily correspond to different types. While (61a) is used to exclude the presence of cheese, (62a) is used to report on its absence. Conversely, while (61b) reports the fact that the interdiction of eating certain instances/types of food does not concern cheese, (62b) forbids the interdiction of eating cheese. (62) a.
Nessun formaggio alla fine del pasto no type of cheese at the end of the meal b. Niente formaggio tra i cibi vietati no type of cheese among the food that has been banned
The restrictions coming from the determiners are so strong that, when they are not met, they may cause the discourse to become incoherent. So, only nessuno, that reports the absence of any degree/type of fear conceived as possible in a given situation, is compatible with the continuation proposed in (63). On the contrary, niente, that seems to negate the type itself, is unfelicitous in a contingent statement. (63) a.
nessuna paura tra gli ostaggi, l’operazione si è svolta con ordine no fear among the hostages, the operation progressed in an orderly way b. niente paura tra gli ostaggi, *l’operazione si è svolta con ordine
The interaction between the lack of existential import proper to niente-phrases and the properties of uncountable abstract nouns such as names of feelings gives rise to another phenomenon, discussed in the following subsection.
L. M. Tovena
.. Uncountable In Sections 3 and 4 the restricted possibilities for nessuno to combine with uncountable nouns were discussed. It has appeared that only intensive quantities are suitable. On the contrary, niente can combine with all sorts of uncountable nouns. When discussing (63), it was noted that names of feelings are a subset of abstract uncountable nouns with which both items can combine. It was also noted that the interpretation of the combination differs, in particular the different presuppositions associated with the items are satisfied in different contexts. Nientephrases have an intensional flavour that tends to make them infelicitous in contingent or episodic statements, where the entities in the restriction of the generalized quantifier are presupposed to exist or a specific spatio-temporal location is identified. In other words, the requirements coming from the item must not mismatch with the constraints coming from the interpretation of the predicate or VP. A better continuation for niente paura than the one in (63) is provided in (64), where the event described by the second clause is distinct. Actually, in (64) the first clause does not describe an eventuality, rather it expresses an exhortation, as captured by the imperative negative in the gloss. This wish-form applies to the situation holding at utterance time. (64) niente paura, sta arrivando un mediatore della nostra ambasciata do not be afraid, a mediator from our embassy is on her way Examples of this exhortative function of niente-phrases are frequent, cf. (65). (65) Niente paura Lazzeretti, la missiva craxiana ha raggiunto tutti i 108 delegati al consiglio generale dell’Internazionale. (23/1/1997IM) Do not be afraid Lazzaretti, the message from Craxi reached all the 108 delegates in the general council of the International Going back to example (43), the possibility of replacing niente with nessuno in this particular case, with no relevant shift in meaning, can be ascribed to the fact that the nessuno-phrase describes a state that also holds at utterance time. The difference between the two items is mainly a matter of whether the hearer/addressee is included in the set of people being surprised, as in the case with niente, or the inclusion is left unspecified, as in the case with nessuno. (43) Nessuna meraviglia quindi se il discorso ha mandato su tutte le furie i repubblicani [. . . ] Then no surprise that the speech infuriated the republicans Accordingly, (66a) is interpreted as providing an account of facts and (66b) describes the absence of feelings of sorrow by the terrorist.
Distributional restrictions on negative determiners
(66) a.
Un microscopico mondo popolato da feroci clan: nessuna pietà per i nemici, ma dedizione totale alla propria colonia. (31/7/1997IM) a microscopic world inhabited by fierce clans: no mercy towards the enemies, but total dedication to their own colony b. Nessun pentimento per Noam Friedman, l’attentatore israeliano. (3/7/1997IM) no regret by Noam Friedman, the Israeli terrorist
Finally, as a last difference between the two items with respect to the spectrum of uncountable nouns, let us recall that niente can combine with concrete mass nouns, cf. (67), while nessuno cannot, cf. (16c). (67) a.
niente latte no milk b. niente posta no mail
. Second partial conclusions If we consider the distribution of nessuno and niente from the point of view of the constraints imposed by the domain of quantification, we can recall the following points. –
–
–
When the structure of the domain does not contain atoms, or they are not clearly individuated, the use of niente allows the possibility of expressing a negation on the whole structure via a ‘negator of essence’ (Horn 1989). The expression can be interpreted as future oriented and having nomic character. It can have a modal flavour and it offers the possibility of eliminating the presupposition of existence that usually characterises the first argument of a generalised quantifier. Actually, niente allows the possibility of linking the ‘existence’ of entities in the restrictor with the will of the speaker. With respect to the same domain, the use of nessuno forces individuation, or at least suggests the possibility of a spatio-temporal localisation of discrete entities. Otherwise it prompts a taxonomic reading, which also implies reference to a set of discrete entities, as with concrete uncountable nouns. The ‘degree/quantity’ reading available with intensive quantities might fall in the same group, in the sense that degrees are closer to species than to extensive quantities. When the structure of the domain contains atoms, the use of nessuno allows the possibility of individuating them. This may motivate the contingent character of this type of quantification, and its use in many cases describing eventualities
L. M. Tovena
located in the past. In many cases, it gives the clause an orientation towards the past even in the absence of overt information. If we consider the constraints that seem to come from the determiners, we can recall the following points. –
–
nessuno seems to suggest a cardinality > 1 of the domain, i.e. in verbless clauses it has a partitive reading. From there it may follow the presupposition of existence that is particularly evident in verbless clauses, and the sensitivity to the presence of weakly or strongly discrete entities in the quantificational domain. niente does not seem to impose particular constraints. In return, it allows a greater freedom with respect to the presuppositional status of the entities in the quantificational domain.
In the light of this short recapitulation, we can look at the mixed sequence contained in example (68). (68) Berlusconi invece [. . . ] ha annunciato “vado ai Caraibi”. Ma niente bermuda alle Bermuda, niente Galliani, Confalonieri e compagnia in calzoncini corti, come in passato. Niente jogging con i collaboratori più stretti. Nessun rischio di fotografi nelle vicinanze. (28/12/1996LS) ‘Instead, Berlusconi announced “I go to the Caraibes”. But no bermuda shorts at the Bermuda, no Galliani, Confalonieri and the company with short trousers, as in the past. No jogging with the closest members of his team. He won’t run the risk of finding photographers nearby’ Niente-phrases can have a volitional flavour. The items in (68), be they things or people, are ruled out by Berlusconi himself. Their presence is intentionally barred as a consequence of the will of the speaker, all of them except for the last one which is beyond his direct control. In this particular case, the entity is introduced by nessuno.
. Summing up In this paper, we have discussed restrictions in the distribution and interpretation of negative determiners, in particular English no, Italian nessuno and French aucun, and determiner-like uses of Italian niente. The typology of determiners defined by Chierchia (1998) has been shown to be too coarse-grained for an accurate classification of negative determiners. In the case of no, evidence has been provided in support of the relevance of morphological number in its interpretation and use, while Chierchia acknowledges this relevance just for the.
Distributional restrictions on negative determiners
Then, it has been shown that some determiners are sensitive to differences among members of the class of uncountable nouns, and that the distribution of a number of singular determiners cannot be captured by restricting their domain of quantification to all and only singular countable nouns. The notion of intensive quantity (Van de Velde 1996) is used in Tovena (2001) to characterise the subset of uncountable nouns with which these determiners can combine. Finally, the behaviour of nessuno in verbless clauses has been compared with determiner-like occurrences of niente. Although both items could be defined as generalised quantifiers imposing a constraint of empty intersection between their first and second arguments, this characterisation is not sufficient. There are other features of their interpretation that must be captured if one wants to account for the differences in their use. In particular, they seem to carry different presuppositions on their quantificational domain.
Notes . In short, this relation covers the two cases of ‘greater or equal’ and ‘part of ’. It is used to define the operation of ‘sum’ (union or join). . The same form can be used also as a quantifier, cf. (i). (i)
Nessuno ha parlato. nobody spoke
. This form can be used also as a quantifier, in which case it is usually followed by a partitive, cf. (i). (i)
Aucun de nous n’a parlé. none of us spoke
It is analysed as anaphoric, a point made to explain the unacceptability of (ii) when uttered out of the blue. (ii) #Aucun n’a parlé. nobody spoke . For a discussion of the polarity sensitivity and free-choiceness of this item, see Tovena and Jayez (1999); Jayez and Tovena (2000). . Van de Velde borrows this term and the notion it names from Kant’s Critique of pure reason. The English rendering given in the text comes from F. Max Müller’s translation dated 1881. . We can call them DPs, the difference is not relevant for the issue under discussion.
L. M. Tovena
References Barwise, J., & Cooper, R. (1981). Generalized quantifiers and natural language. Linguistics and Philosophy, 4, 159–219. Bosveld-de Smet, L. (1998). On Mass and Plural quantification. Ph.D. thesis, University of Groningen. Chierchia, G. (1998). Plurality of mass nouns and the notion of ‘semantic parameter’. In S. Rothstein (Ed.), Events and Grammar (pp. 53–103). Dordrecht: Kluwer. Corblin, F., & Tovena, L. (2001). On the multiple expression of negation in Romance. In Y. D’Hulst, J. Rooryck, & J. Schroten (Eds.), Romance Languages and Linguistic Theory 1999 (pp. 87–115). Amsterdam: John Benjamins. Delfitto, D. (1998). When plural does not mean plural: Some remarks on ‘notional’ and ‘grammatical’ plurality. In Recherches de linguistique française et romane d’ Utrecht XVII (pp. 26–39). OTS. Gutiérrez-Rexach, J. (1996). Negative polarity licensing and the rhetorical interpretation of questions. In Proceedings of the Western States Conference on Linguistics. Horn, L. R. (1989). A natural history of negation. Chicago: Chicago University Press. Jayez, J., & Tovena, L. M. (2000). Strategies of Free–Choiceness. Ms., EHESS and Université de Lille. Kant, I. (1881). Critique of Pure Reason (translated by F. Max Müller). London: MacMillan. Ladd, R. D. (1981). A first look at the semantics and pragmatics of negative questions and tag questions. In Seventeenth Regional Meeting of the Chicago Linguistics Society. Löbner, S. (2000). Polarity in natural language: Predication, quantification and negation in particular and characterizing sentences. Linguistics and Philosophy, 23, 213–308. Manzotti, E. (1991). La negazione. In L. Renzi & G. Salvi (Eds.), Grande grammatica italiana di consultazione, Vol. 2, Ch. 5 (pp. 245–318). Bologna: il Mulino. Molinelli, P. (1988). Fenomeni della negazione dal latino all’italiano. Firenze: La nuova Italia. Schwarzschild, R. S. (1991). On the meaning of definite plural phrases. Ph.D. thesis, University of Massachusetts. Szabolcsi, A. (1997). Ways of scope taking. Dordrecht: Kluwer. Tovena, L. M. (1996). Studies on polarity sensitivity. Ph.D. thesis, University of Edinburgh, published by Garland 1998. Tovena, L. M. (2001). Between mass and count. Proceedings of WCCFL 20 (pp. 565–578). Cascadilla Press. Tovena, L. M., & Jayez, J. (1999). Any: From scalarity to arbitrariness. In F. Corblin, J.-M. Marandin, & C. D.-Sorin (Eds.), Empirical issues in formal syntax and semantics II (pp. 39–57). The Hague: Holland Academic Graphic. Van de Velde, D. (1996). Le spectre nominal. Paris: Peeters.
TSL[v.20020404] Prn:29/08/2002; 15:24
F: PBNS02.tex / p.1 (29)
Towards a comprehensive view of Negative Concord João Peres Faculdade de Letras da Universidade de Lisboa
.
Refreshing the notion of Negative Concord
The nuclear fact that characterises the notion of negative concord (henceforth, NC) is illustrated in the sentences in (1): (1) a. Pietro *(non) ha visto nessuno. b. O Pedro *(não) viu ninguém. c. Pedro *(no) vio a nadie. ‘Peter not has seen/saw (at) nobody’ Peter saw nobody/didn’t see anyone.
[italian] [portuguese] [spanish]
The salient phenomenon in these Romance data is the co-occurrence of a standard (arguably sentential) negation operator and a distinguished expression (nessuno, ninguém, nadie – English nobody), commonly named “n-word” (after Laka 1990). The well-known peculiarity of such kind of expressions resides in that, in contexts like (2) (2) a. Nessuno (*non) ha visto Pietro. b. Ninguém (*não) viu o Pedro. c. Nadie (*no) vió a Pedro. ‘Nobody (not) has seen/saw Peter’ Nobody saw Peter.
[italian] [portuguese] [spanish]
they can be proven to express by themselves, without any overt negation operator, some sort of negative value, while in contexts like (1), if, following the standard analyses of such sentences, the sentential negation operator is to be assigned its full semantic capacity, then the negative value of the n-word is, so to speak, inert.
TSL[v.20020404] Prn:29/08/2002; 15:24
F: PBNS02.tex / p.2 (30)
João Peres
This is the core notion of what might be called strong negative concord: an expression that is negative in some sense – in the cases under scrutiny, in the sense of involving contradiction and being antiadditive1 – is deprived of its negative value, while keeping its form, even if it involves a (more or less recognisable by present day speakers) negative morpheme. As is well known, this is not the exact pattern found in French, where the weak negative particle (ne) is always required, and, more especially, in Slavic languages or Hungarian, where a strong negation particle always accompanies n-words. Of course, the significance of a notion of NC where the concordant element does not autonomously possess some degree of negative capacity deserves serious reflection. I will skip this issue respecting the variety of patterns covered by the label NC in the literature. At this initial stage, it should also be stressed that more relaxed patterns of NC can be found in several languages (including those being considered here), namely cases where the negative value that justifies concordant elements like n-words is not an overt negative operator or even a negative operator as was the case in the data in (1). Witness examples like the following, which will not be scrutinised in this paper: (3) Credi che si presenterà nessuno? IT ‘believe that himself show-up nobody?’ Do you believe that anyone will show up?
[Rizzi 1982: 126]
(4) Dubito che venga nessuno. IT ‘doubt that comes nobody’ I doubt that nobody comes. (5) A Ana sabe o que fazer melhor do que ninguém. PT ‘the Ana knows the what do better than what nobody’ Ana knows what to do better than anyone. (6) Es la ultima vez que te digo nada.2 SP ‘is the last time that you say1st sg say nothing’ This is the last time I tell you anything. (7) En lugar de habérselo dicho a nadie, te lo deberías haber callado. SP ‘in place of have-it said to nobody, you it should have concealed’ Instead of having said it to someone, you should have kept it secret.
TSL[v.20020404] Prn:29/08/2002; 15:24
F: PBNS02.tex / p.3 (31)
Towards a comprehensive view of Negative Concord
. The extended domain of Negative Concord . The licensees of Strong Negative Concord At least in the restricted group of languages where the patterns in (1) and (2) can be found – which includes Italian, Portuguese and Spanish – NC is a much more widespread phenomenon than what is generally acknowledged in the literature. In fact, different subtypes of this construction can be found. The subsequent data are from Portuguese, but most of the constructions are available in the three mentioned Romance languages, with minor variation.
A. Non-coordinate (implicative) operators – even type A.1. Implicative NP’s/PP’s/AdvP’s/. . . ([not even XP] subtype). This subtype comprises at least different definite or generic NP’s, argument PP’s and different sorts of adverbials. The particular negative construction conveys an implicature that is abundantly described in the literature, namely as regards its positive counterpart (with even). A.1.1. Implicative definite NP’s (8) a.
Nem o Pedro respondeu à pergunta. ‘not-even the Pedro answered to-the question’ b. Não respondeu à pergunta nem o Pedro. ‘not answered to-the question not-even the Pedro’ Not even Pedro answered the question. Este Verão nem em Londres choveu. ‘this summer not-even in London rained’ b. Este Verão não choveu nem em Londres. ‘this summer not rained not-even in London’ This summer it didn’t rain even in London.
[NC version]
(9) a.
[NC version]
A.1.2. Implicative generic NP’s (10) a.
Nem um elefante derrubava este muro. ‘not-even an elephant would-knock-down this wall’ b. Não derrubava este muro nem um elefante. [NC version] ‘not would-knock-down this wall not-even an elephant’ Not even an elephant would knock this wall down.
A.1.3. Implicative NP’s with numerals (11) a.
Nem vinte estudantes foram à festa. ‘not-even twenty students went to-the party’
TSL[v.20020404] Prn:29/08/2002; 15:24
F: PBNS02.tex / p.4 (32)
João Peres
b. Não foram à festa (nem) vinte estudantes. [NC version; ambiguous without nem] ‘not went to-the party not-even twenty students’ Not even twenty students attended the party. (12) a.
Nem vinte estudantes couberam na sala. [simple past; episodic reading 3 ] ‘not-even twenty students fit in-the room’ b. Não couberam na sala (nem) vinte estudantes. [NC version; ambiguous without nem] ‘not fit in-the room not-even twenty students’ Not even twenty students fit in the room.
(13) a.
Nem vinte estudantes cabem na sala. [present tense; generic reading; implicatum: 20 students would fit; inference: possibly, < 20 will fit] ‘not-even twenty students fit in-the room’ b. Não cabem na sala (nem) vinte estudantes. [NC version; ambiguous without nem] ‘not fit in-the room not-even twenty students’ Not even twenty students fit in the room.
(14) a.
Nem vinte homens levantam esta pedra. [present tense; generic reading] ‘not-even twenty men can-lift this stone’ b. Não levantam esta pedra (nem) vinte homens. [NC version; ambiguous without nem; implicatum: 20 men would lift; inference: possibly, > 20 can lift] ‘not can-lift this stone not-even twenty men’ Not even twenty men can lift this stone.
A.1.4. Implicative NP’s with a universal quantifier plus group and generic reading (15) a.
Nem todos os membros desta equipa levantariam esta pedra. [ambiguity DR/GR] ‘not-even all the members of-this team would-lift this stone’ b. Não levantariam esta pedra (nem) todos os membros desta equipa. [GR; preferred NC; implicatum: all the members of the team together would lift] b . Não levantariam esta pedra (*nem) todos os membros desta equipa. [DR; *NC; implicatum: all the members of the team would each lift] ‘not would-lift this stone not-even all the members of-this team’ Not even all the members of this team would lift this stone.
TSL[v.20020404] Prn:29/08/2002; 15:24
F: PBNS02.tex / p.5 (33)
Towards a comprehensive view of Negative Concord
A.2. Implicative sentences (a subgroup of concessive conditionals – even if type) (16) a.
Nem que me matem eu conto tudo o que sei. ‘not-even that kill3rd pl me I tell all the what [I] know’ b. Eu não conto tudo o que sei, nem que me matem. [NC version] ‘I not tell all the what [I] know not-even that kill3rd pl me’ I won’t tell everything I know, even if they kill me.
B. Disjunctive coordinate structures (arguably NP, PP, AdvP, . . . coordinations) (17) a.
Nem o Pedro nem a Ana foram à festa. ‘neither the Pedro nor the Ana went to-the party’ b. Não foram à festa (nem) o Pedro nem a Ana. [NC version] ‘not went to-the party (neither) the Pedro nor the Ana’ Neither Pedro nor Ana went to the party.
. Identifying the class of expressions licensed under Strong Negative Concord Table 1 groups all the concordant NP types presented in the previous data, with a cross-reference to the sentences. Since I assume as an empirical evidence that this Table exhausts the set of subclasses of NP’s that enter in strong negative concord,4 the obvious question to be raised at this point concerns the characterisation of the group. From direct observation, the first striking fact is the absence of a number of NP types containing a negative element, namely, among others, those formed with the equivalents of not many, not few, not less than n, not more than n and not all (in the latter case, in the distributive reading usage) – which are frequently called Table 1. NP’s involved as concordant elements in NC in Portuguese5
a b c d e f g h i
Portuguese structures
English equivalents
nenhum nem [o]DEF N nem [o/um/os N]generic [nem n N]dr [nem n N]gr/episodic (1st arg. of fit V’s) [nem n N]gr/generic (1st arg. of fit V’s) [nem n N]gr/generic (1st arg. of lift V’s) [nem todos os N]gr (1st arg. of lift V’s) nem o X nem o Y
no not even [the] + N/Proper Noun not even [[the/a] + N]/bare plural not even n not even n not even n not even n not even all neither X nor Y
cf. (20) cf. (8)–(9) cf. (10) cf. (11) cf. (12) cf. (13) cf. (14) cf. (15) cf. (17)
TSL[v.20020404] Prn:29/08/2002; 15:24
F: PBNS02.tex / p.6 (34)
João Peres
“negated phrases”. In Horn’s terms applied to not many, these are forms of “special negation, despite their association with the semantics of contradictory opposition” (1989: 507). In the face of this partition, three questions, which I cannot address here directly, deserve a systematic scrutiny, since, to my knowledge, they have not yet been considered together in an integrated manner: a.
Given some language, which quantifying expressions can become negative (that is, acquire the status of n-words) and/or negated? b. What is the status, in terms of degree of negativity, of the Italian neanche, Portuguese nem, and Spanish ni (English not even) subclass and the Italian non, Portuguese não, and Spanish no (English not) type of constructions? c. Which are the relevant syntactic differences between the two groups (part of the answer presumably being that the not type always corresponds to constituent negation)?
Engaging in a more modest endeavour than trying to answer these questions, I will just attempt to draw a semantic dividing line between the expressions that enter NC chains as licensing elements6 and those that don’t, in the languages under analysis. Before that, let me point out two interesting syntactic facts. The first is that the two classes are in complementary distribution with respect to their occurrence as verb complements in the presence or the absence of standard negation: (18) a.
O Pedro (*não) escreveu não mais de cinco/não muitos artigos. [non-licensee] ‘the Pedro wrote not more of five/not many papers’ Pedro wrote not more than five/not many papers. b. O Pedro *(não) escreveu nem cinco artigos. [licensee] ‘the Pedro not wrote not-even five papers’ Pedro wrote not even five papers.
In contrast to this absolute syntactic discrepancy, the second interesting fact is that some of the NC licensees are totally incompatible with a computable negation operator to their right – cf. (19a–b) – while others share this possibility with non-licensees – cf. (19c–f): (19) a. *Nenhum estudante NÃO foi à festa. ‘no student not went to-the party’ b. *Nem o Pedro nem a Ana NÃO foram à festa. ‘neither the Pedro nor the Ana not went to-the party’ c. Nesta festa, nem o Pedro NÃO dançou. ‘in-this party, not-even the Pedro not danced’ At this party, not even Pedro didn’t dance.
[licensee] [licensee] [licensee]
TSL[v.20020404] Prn:29/08/2002; 15:24
F: PBNS02.tex / p.7 (35)
Towards a comprehensive view of Negative Concord
d. Nem cinco estudantes NÃO foram à festa. ‘not-even five students not went to-the party’ Not even five students didn’t attend he party. e. Nem todos os estudantes NÃO foram à festa. ‘not all the students not went to-the party’ Not all the students didn’t attend the party. f. Não poucos estudantes NÃO foram à festa. ‘not few students not went to-the party’ Not few students didn’t attend the party.
[licensee]
[non-licensee]
[non-licensee]
It is quite obvious that at least the syntactic property just considered doesn’t constitute solid grounds for defining the two classes of operators we have been observing. I will not tackle these issues here, concentrating instead on some semantic properties that present themselves as candidates for characterising the class of NC licensees. The first semantic property that comes to mind is decreasing monotonicity, a property that became of paramount importance in the treatment of negation related issues in particular after the work of Ladusaw (1979). In fact, the NP’s that are licensed in NC involve quantifying operations that result in left downward entailment (that is, DE with respect to the first argument of the quantifier), except for the special case of definite descriptions, which can only trivially be so considered. In order to facilitate the computation of the monotonicity of the classical n-words, a sentence like (20) below, where a determiner n-word (nenhum/nessun/ningún – English no) is used instead of a full NP (ninguém/nessuno/nadie – English nobody) should be observed instead of (2) above: (20) O Pedro não leu nenhum livro. ‘the Pedro not wrote no book’ Pedro didn’t write any book. The uniformity of left monotonicity could give support to Dowty’s (1994) idea that NC acts as a special marker for DE positions, to which I adhered in Peres (1997). However, it can and has been objected that several DE positions, namely those that are not induced by a standard negation operator, are not assigned any special marking. Accordingly, it seems more reasonable to think of left DE as a necessary but not a sufficient condition for NC licensing. The second noticeable semantic fact concerns the logical properties that have been associated with the different manifestations of negation and the degrees of negativity. In this respect, the first important verification is that both the expressions in Table 1 and those that reject NC – namely, expressions like not many, not few, not less than n, not more than n and not all (in DR) – are associated with con-
TSL[v.20020404] Prn:29/08/2002; 15:24
F: PBNS02.tex / p.8 (36)
João Peres
tradiction, inasmuch as all of them play a crucial role in the generation of the contradictory of another proposition. Accordingly, this feature, exactly like decreasing monotonicity, is not responsible for concordant elements in NC, other ingredients of negation having to be scrutinised. The most obvious candidates are the properties that De Morgan’s laws reveal about negation. A standard negative operator of the kind of not and its equivalents is, as is well-known from the literature, antimorphic, which means that it constitutes a function f obeying the two De Morgan laws generalised from set theory and the propositional calculus. This is their set-theoretical formulation: (21) a. f (X ∪ Y) ↔ f (X) ∩ f (Y) b. f (X ∩ Y) ↔ f (X) ∪ f (Y)
(anti-additivity) (anti-multiplicativity)
In Table 2, the two groups of expressions under scrutiny are classified, in columns 2–4, with respect to monotonicity and the two above properties, regarding both arguments of the quantifying expression. The group of expressions in lines 8–12 including only non-licensees of NC, once again it becomes indisputable that in neither argument decreasing monotonicity is a distinctive factor in the matter. As for anti-additivity and anti-multiplicativity, it is also observable that the first argument position is totally unrevealing. However, the same cannot be said about anti-additivity in the second argument: except for the vexing case of nem n N (not even n N) in distributive reading, all the licensed n-phrases are anti-additive in the second argument, which means that, besides “their association with the semantics of contradictory opposition” (Horn 1989: 507), they hold a considerable share of the logical properties of what Zwarts (1986) termed classical negation (epitomised by the not and without sort of operators). To this group of weaker (than classical negation) negative operators, the same author assigned the label regular negation. Table 2. Negation related properties of (Portuguese) negative and negated n-phrases
nenhum N nem o X nem o Y nem [o]DEF N nem [o/um/os N]GENERIC [nem n N]DR [nem n N]GR [nem todos os N]GR [nem todos os N]DR não muitos (not many) N não poucos (not few) N não mais de (not more than) n N não menos de (not less than) n N
Monot.
Anti-addit.
Anti-mult.
Anti-verid.
↓ MON ↓ n.a. MON ↓ Ø MON ↓ ↓ MON ↓ ↓ MON ↓ ↓ MON ↓ ↓ MON ↓ ↑ MON ↓ Ø MON ↓ Ø MON ↑ ↓ MON ↓ ↑ MON ↑
+ A-ADD + n.a. A-ADD + – A-ADD + + A-ADD + – A-ADD – – A-ADD + – A-ADD + – A-ADD – – A-ADD – – A-ADD – – A-ADD – – A-ADD –
– A-MULT – n.a. A-MULT – – A-MULT + – A-MULT – – A-MULT – – A-MULT – – A-MULT + – A-MULT + – A-MULT – – A-MULT – – A-MULT – – A-MULT –
+ A-VERID + A-VERID + A-VERID + A-VERID + A-VERID + A-VERID + A-VERID – A-VERID – A-VERID – A-VERID – A-VERID – A-VERID
TSL[v.20020404] Prn:29/08/2002; 15:24
F: PBNS02.tex / p.9 (37)
Towards a comprehensive view of Negative Concord
We can therefore state that, except for the already mentioned exception, regular negation operators are licensees in NC. Giannakidou 1998 has brought considerable attention to the properties of veridicality, non-veridicality and anti-veridicality in relation with the licensing of the expressions that have for long been called “negative polarity items”, showing that this kind of expressions can be sensitive to that variety of values. The value that matters here is anti-veridicality, which grosso modo is the property of a context where some proposition is assumed to be false by some entity relevant for the discourse, either directly, by means of an overt antiveridical operator (like classical negation) or indirectly, by implicature. It is the latter case that may shed some light on the picture given in Table 2. In fact, all the operators in lines 3–7 of the Table are implicative in nature, each conveying the denial of an implicit proposition (for instance, not even twenty people attended the meeting, carries the implicatum that at least twenty people would attend the meeting, which is being denied, the same going for the equivalents in the NC languages being considered). Given the account of the relevant facts just sketched, it now becomes quite clear why nem todos os N (English not all N, in distributive reading), contrary, for instance, to nem n N (English not even n N), is not licensed in a NC chain: (22) a.
Nem todos os estudantes foram à festa. ‘not all the students went to-the party’ b. Não foram à festa (*nem) todos os estudantes. [*NC version; ambiguous without nem] ‘not went to-the party not twenty students’ Not all the students attended the party.
In fact, this operator not only lacks the apparently necessary condition of decreasing monotonicity in its first argument – which precludes the emergence of a stronger property of negation like anti-additivity – but also lacks the either direct or implicative anti-veridical effect that other operators exhibit. The claims that were just made should not be interpreted as sustaining the idea that only under anti-veridicality – verified in the contexts analysed here – can n-phrases occur in the languages under consideration. This is not the case, as I have tried to prove in my 1998 Salford paper and in Peres (2000), where I paid close attention to the cross-sentential licensing of n-words, trying to prove that Portuguese n-words are endowed with a multifactorial sensitivity, an idea that, at least in my reading of her work, plays a central role in the view of the field that was offered in Tovena (1998).
TSL[v.20020404] Prn:29/08/2002; 15:24
F: PBNS02.tex / p.10 (38)
João Peres
. Mutual licensing of n-phrases It is a well-known fact that classical n-words (of the Italian nessun, Portuguese nenhum, and Spanish ningún – English no – type) and n-phrases (of the Italian nessuno, Portuguese ninguém, and Spanish nadie – English nobody – or Italian niente, Portuguese and Spanish nada – English nothing – type, etc.) are mutual licensers. Having now contemplated a wider domain of n-phrases, the question is in order whether or not all the operators that engage in a NC chain headed by anti-morphic negation can license each other. The observation of the data leads to the following description: A. In general, classical n-phrases can act as a licensers of one another, as is abundantly documented in the literature: (23) a. a . b. b . c. c .
Nessuno ha fatto niente. IT Nessun studente conosce né Parigi né Londra. Ninguém fez nada. PT Nenhum estudante connhece (nem) Paris nem Londres. Nadie hizo nada. SP Ningún estudiante conoce (ni) Paris ni Londres. Nobody did anything. No student knows Paris or London.
B. Disjunctive n-phrases, exemplified above in (17), operate as NC licensers just like classical n-phrases: (24) a. a . b. b . c. c .
Né Pietro né Maria hanno letto nessun libro. IT Né Pietro né Maria sono stati né a Parigi né a Londra. Nem o Pedro nem a Maria leram nenhum livro. PT Nem o Pedro nem a Maria estiveram (nem) em Paris nem em Londres. Ni Pedro ni Maria han leído ningún libro. SP Ni Pedro ni Maria estuvieron (ni) en Paris ni en Londres. Neither Peter nor Mary read any book. Neither Peter nor Mary were in Paris or London.
C. Definite n-phrases are in general accepted as licensers, apparently with some variation: (25) a.
Neanche Pietro há fatto niente per aiutarmi/avveva letto *nessun libro. IT ‘not-even Pietro has done nothing for help-me/had read no book’
TSL[v.20020404] Prn:29/08/2002; 15:24
F: PBNS02.tex / p.11 (39)
Towards a comprehensive view of Negative Concord
b. Nem (sequer) o Pedro fez nada para me ajudar/tinha lido nenhum livro. PT ‘not-even the Pedro did nothing for me help/had read no book’ c. Ni (siquiera) Pedro hizo nada para ayudarme/leyó ningún libro. SP ‘not-even Pedro did nothing for help-me/read no book’ Not even Peter did anything to help me/(had) read any book. D. Also group reading implicative operators are in general accepted as licensers, again apparently with some variation. (26) a.
Nemmeno cinque uomini potrèbbero sollevare *nessuna/neanche una di queste pietre. IT ‘not-even five men could lift none/not-even one of these stones’ b. Nem cinco homens levantariam nenhuma destas pedras. PT ‘not-even five men would-lift none of-these stones’ c. Ni (siquiera) cinco hombres levantarían ninguna de estas piedras. ES ‘not-even five men would-lift none of these stones’ Not even five men could lift any of these stones.
(27) a.
Nemmeno tutti i lavoratori potrèbbero sollevare *nessuna di queste pietre. IT ‘not-even all the workers could lift none of these stones’ b. Nem todos os operários levantariam nenhuma destas pedras. PT ‘not-even all the workers would-lift none of-these stones’ c. Ni (siquiera) todos mis obreros (juntos) serian capaces de levantar ninguna de estas piedras. ES ‘not-even all my workers (together) would-be able of lift none of these stones’ Not even all my workers together would be able to lift any of these stones.
E. Numeral distributive n-phrases don’t play the licenser role: (28) a. *Neanche due studenti hanno fatto niente per aiutarmi/hanno letto nessun libro. IT ‘not-even two students have done nothing for help-me/have read no book’ b. *Nem dois estudantes fizeram nada para me ajudar/leram nenhum livro. PT ‘not-even two students did nothing for me help/read no book’
TSL[v.20020404] Prn:29/08/2002; 15:24
F: PBNS02.tex / p.12 (40)
João Peres
c. *Ni diez estudiantes hicieron nada para ayudarme/han leído ningún libro. SP ‘not-even two students did nothing for help-me/have read no book’ Not even two students did anything to help me/have read any book. F. Finally, any combination of implicative operators is rejected, that is, nothing of the following format is accepted in any of the languages under scrutiny: (29) not-even X – V – not-even Y (30) *Nem vinte estudantes fizeram nem dois trabalhos. ‘not-even twenty students made not-even two papers’ *Not even twenty students wrote not even two papers. I assume this fact corresponds to a very general – possibly, universal – constraint on the implicative even-constructions at stake. Apparently, and in accordance with all the facts that were discussed before, the only possible explanation for the distribution shown by the data in A–E above is that, in order to be able to license other n-phrases, these expressions must have the capacity to generate a context that shares with classical negation not only a contradictory semantics, but also other logical properties. This is precisely the case with the operators in A–D, which are all anti-additive in their second argument, inside which the licensee is supposed to occur. Consistently, this is not the case with the negated numerals in E, which only carry a weaker negative value, namely, decreasing monotonicity. This is shown in Table 3, where values are marked for n-phrases that act as licensers in Portuguese (and, for that matter, for their equivalents in other languages), regarding both arguments and with respect to monotonicity, anti-multiplicativity and anti-additivity.
Table 3. Negation related properties of (Portuguese) licenser n–phrases
nenhum N nem o X nem o Y nem [o]DEF N nem [o/um/os N]GENERIC [nem n N]GR [nem todos os N]GR [nem n N]DR
Monot.
Anti-addit.
Anti-mult.
↓ MON ↓ n.a. MON ↓ Ø MON ↓ ↓ MON ↓ ↓ MON ↓ ↓ MON ↓ ↓ MON ↓
+ A-ADD + n.a. A-ADD + – A-ADD + + A-ADD + – A-ADD + – A-ADD + – A-ADD –
– A-MULT – n.a. A-MULT – – A-MULT + – A-MULT – – A-MULT – – A-MULT + – A-MULT –
TSL[v.20020404] Prn:29/08/2002; 15:24
F: PBNS02.tex / p.13 (41)
Towards a comprehensive view of Negative Concord
. General conclusion The domain of analysis of the present work is defined by the structures that were classified in Part 1 as instances of “strong negative concord”, that is, those structures where an expression that can per se convey negative values occurs under the licensing effect of another negative expression, whose negative value is the only one to be computed. Two issues were addressed: firstly, which phrases are licensed in a Negative Concord chain (the analysis having been restricted to noun phrases), and, secondly, which noun phrases that are licensed in negative concord chains can assume the role of licensers in the same kind of configuration. The conclusions that were reached can be summarised as follows: (i) as should be expected, all licensees and licensers in NC chains involve a semantics of contradiction; (ii) a strong logical property of classical negation (again, resorting to Zwarts’ 1986 terminology), namely anti-additivity, is a sufficient condition for the fulfilment of both roles, and a necessary condition only for the licenser role; (iii) the licensee role is less constrained than that of licenser, inasmuch as the conveyance of anti-veridicality, either directly or by implicature, acts as a sufficient condition. Crucially, while the licensed elements can have the required negative values realised indirectly, via implicature, licensers must directly communicate strong negative values.
Notes . For an n-word like English no, asserting that this operator is anti-additive is tantamount to stating that the following equivalence (where N1 and N2 stand for any nouns and X stands for any predicative expression) holds (see Zwarts 1995 for deeper analysis): (ii) [no (N1 or N2 ) X] if and only if [no (N1 ) and no (N2 ) X] . To my knowledge, this kind of structure as well as the subsequent one was not yet acknowledged in the literature. I am indebted to León Acosta for having called my attention to them. . If subject to appropriate lexical selection or given an adequate context, all the sentences in (12)–(14) are ambiguous between a distributive (DR) and a group reading (GR). Witness the following clearly ambiguous sentence (given a situation where someone is testing the capacity of a briefcase for containing different folders separately or as a group): (i)
Nem cinco cadernos couberam na minha pasta. ‘not-even five folders fit in-the my briefcase’ Not even five folders fit in my briefcase.
However, since this variation does not impinge on the licensing of NC, it is being skipped here. Quite differently, it is relevant with the universal quantifier – cf (15). . Several authors – among them van der Wouden (1994) – take phrases like almost nobody to be quantifiers. This is a debatable view, to which I prefer considering almost as an operator
TSL[v.20020404] Prn:29/08/2002; 15:24
F: PBNS02.tex / p.14 (42)
João Peres
on propositions. If this trend is adopted, then the concordant element is still just nobody and related expressions. . In lines e–h in Table 1, the reference to specific verbs implies the acknowledgement that a semantic difference of the kind established between upward and downward scalar predicates in Beck and Rullmann (1996) may play an important role in the semantic characterisation of nem (not even) phrases. . I won’t discuss here the status of licensers and licensees, since I assume that in sentences like (1) above the n-word is the licensed element. Accordingly, I reject the idea – advocated in some literature – that the standard negative operator can be a doubled element triggered by the n-word. As I tried to prove elsewhere, such an idea is not compatible with NC chains where a standard negative element occurs, as a licenser, in a sentential domain higher than that of the licensed n-word and cannot be dispensed with in the meaning computation.
References Beck, S., & Rullmann, H. (1996). Degree Questions, Maximal Informativeness, and Exhaustivity. In P. Dekker & M. Stokhof (Eds.), Proceedings of the Tenth Amsterdam Colloquium, December 18–21, 1995 (pp. 73–92). Amsterdam: ILLC. Dowty, D. (1994). The Role of Negative Polarity and Concord Marking in Natural Language Reasoning. In M. Harvey & L. Santelmann (Eds.), Proceedings from Semantics and Linguistic Theory IV (pp. 114–144). Ithaca, NY: DMLL Publications. Giannakidou, A. (1998). Polarity Sensitivity as (Non)Veridical Dependency. Amsterdam: John Benjamins. Horn, L. (1989). The Natural History of Negation. Chicago: University of Chicago Press. Ladusaw, W. (1979). Polarity Sensitivity as Inherent Scope Relations. Ph.D. dissertation, University of Texas at Austin. Laka, I. (1990). Negation in syntax: on the nature of functional categories and projections. Ph.D. dissertation, MIT, Cambridge. Peres, J. A. (1997). Extending the Notion of Negative Concord. In D. Forget, P. Hirschbühler, F. Martineau, & M.-L. Rivero (Eds.), Negation and Polarity, Syntax and Semantics, Selected Papers from the Colloquium Negation: Syntax and Semantics, Ottawa, 11–13 May, 1995 (pp. 289–310). Amsterdam: John Benjamins. Peres, J. A. (1998). On Romance Sensitivity to Non-veridicality. Ms., paper presented to the Negation: Syntax, Semantics and Pragmatics Conference, University of Salford, Manchester, October 30–November 1, 1998. Peres, J. A. (2000). On the Nature and Licensing Conditions of N-phrases in Portuguese. DELTA (J. of the Brazilian Association of Linguistics), 16 (special issue), 165–199. Rizzi, L. (1982). Issues in Italian Syntax. Dordrecht: Foris. Tovena, M. L. (1998). The Fine Structure of Negative Polarity. New York: Garland Publishers. Wouden, T. van der (1994). Negative Contexts. Ph.D. dissertation, Rijksuniversiteit Groningen. Zwarts, F. (1986). Categoriale Grammatica en algebrasche semantiek. Eeen studie naar negatie en polariteit in het Nederlands. Ph.D. dissertation, Rijksuniversiteit Groningen. Zwarts, F. (1995). Nonveridical Contexts. Linguistic Analysis, 25, 286–312.
TSL[v.20020404] Prn:23/08/2002; 12:19
P II
Temporality
F: PBNSP2.tex / p.1 (43)
TSL[v.20020404] Prn:29/08/2002; 15:26
F: PBNS03.tex / p.1 (45)
On temporal constructions involving counting from anchor points Semantic and pragmatic issues Telmo Móia Universidade de Lisboa, Portugal
Introduction This paper focuses on two closely related subtypes of temporal expressions, illustrated in (1) and (2) below. The property that brings them together is the fact that they both involve counting of temporally ordered entities (such as weekdays – Sundays, in the examples below – calendar years or sport seasons, for instance). I will concentrate on Portuguese and English, although parallel expressions in other languages, such as French or Spanish, seem to behave similarly with respect to the issues at stake. (1) a.
O Paulo não vai à igreja há três domingos. “the Paulo not goes to-the church there-is three Sundays” b. Paulo hasn’t been to church for three Sundays (now).
(2) a.
O Paulo não vai à igreja desde há três domingos (atrás). “the Paulo not goes to-the church since there-is three Sundays (behind)” b. Paulo hasn’t been to church since three Sundays ago.
Notice that the two Portuguese sentences above – sentences a – contain the expression há (an inflected present tense verb form, similar to French il y a), differing only with respect to the temporal preposition desde (the counterpart of English since).1 However, their approximate English counterparts – sentences b – symptomatically involve different temporal particles: for in the first case, and a combination of since and ago in the second case.
TSL[v.20020404] Prn:29/08/2002; 15:26
F: PBNS03.tex / p.2 (46)
Telmo Móia
Now, the sentences in (1) and (2) – which describe a negative habitual state – are not equivalent. In a scenario where Paulo went to church on a Wednesday the week before the utterance time, for instance, the statements in (1) would not necessarily be false, whereas those in (2) would. The reason is that the assertions in (1) merely involve the three Sundays that precede the utterance time, while those in (2) involve the whole interval between the utterance time and the third Sunday in its past, including all weekdays (cf. Section 3 for elaboration on this issue). In schema (3), the relevant intervals for the location of the described state, which are discontinuous for (1), are represented by the thicker lines below the time axis: (3)
há três domingos / three Sundays ago, in (2) Sunday3
Sunday2
utterance time Sunday1 há três domingos / for three Sundays, in (1)2 desde há três domingos / since three Sundays ago, in (2)
Given the asymmetry portrayed in (3), I will assume that the relevant temporal expressions in (1) and (2) belong in different semantic categories, Portuguese háexpressions being genuinely ambiguous,3 approximately in the following terms: i.
In sentences like (1a), Portuguese há-expressions are associated with the sum of n periods of the stated type (in this case, three Sundays) in the past of the utterance time, that is, they involve a set of location times, rather than a single location time. In this respect, they are akin to the so-called ‘adverbials of temporal quantification’ like todos os domingos (every Sunday), em três domingos (on three Sundays) and nos últimos três domingos (on the last three Sundays), and directly express what could perhaps be termed location relative to a set of intervals (cf. Móia 2000). In the use at stake, Portuguese há-expressions behave as the counterpart of English for-expressions (like those in (1b)). ii. In sentences like (2a) – or (4a) below – Portuguese há-expressions represent the nth period of the stated type (in this case, the third Sunday) in the past of the utterance time;4 in fact, they are time-denoting expressions, which can be combined – though need not be, as illustrated in (4a) – with a locating preposition like desde to form a (strict) temporal locating adverbial, i.e. a phrase defining a single location time (cf. end of Section 1 for a development on this issue). In the use at stake, Portuguese há-expressions behave as the counterpart
TSL[v.20020404] Prn:29/08/2002; 15:26
F: PBNS03.tex / p.3 (47)
Temporal counting from anchor points
of English ago-expressions (like those in (2b) or (4b)), or before-expressions, in other contexts (such as Paulo had been in church three Sundays before (that)). (4) a.
O Paulo esteve na igreja há três domingos (atrás). “the Paulo was in-the church there-is three Sundays (behind)” b. Paulo was in church three Sundays ago.
As said above, the property shared by the temporal expressions in (1) and (2) is the involvement of an operation of counting temporally ordered entities from an anchor point. Before proceeding to the semantic analysis of these expressions, which is the focus of this paper, it is worth noticing that the same temporal particles – há in Portuguese, for or ago in English – may be used with predicates of amounts of time (e.g. three hours, three months) as their complements, in which case an operation of time measurement (rather than counting) is involved. Witness the following examples: (5) a.
O Paulo não vai à igreja há três meses. “the Paulo not goes to-the church there-is three months” b. Paulo hasn’t been to church for three months (now).
(6) a.
O Paulo não vai à igreja desde há três meses (atrás). “the Paulo not goes to-the church since there-is three months (behind)” b. Paulo hasn’t been to church since three months ago. [odd, or not used]
Sentences (5a) and (6a) arguably contain two different há-homonyms (parallel to those in (1a) and (2a), respectively). However, these two Portuguese sentences – contrary to what is the case with (1a) and (2a) – are equivalent. In Section 2, these facts will be considered, together with the possible oddity of English sentences like (6b).
.
Time-denoting expressions involving counting from anchor points
Time-denoting phrases which define intervals by measurement from an anchor point, like those in (6) (há três meses/three months ago), have often been mentioned in the literature. For instance, Bras (1990) considers a class of “adverbiaux qui désignent la zone temporelle en operánt un report de mesure” (p. 199, my emphasis), and Asher et al. (1995) – referring back to Molinès (1989) and Bras and Molinès (1993) – state that (. . . ) [locating time adverbials] fall into (. . . ) classes, depending on (. . . ) whether or not the identification of the referent depends on the projection of
TSL[v.20020404] Prn:29/08/2002; 15:26
F: PBNS03.tex / p.4 (48)
Telmo Móia
a length of time on the temporal axis (from some given point). For example, for the adverbials il y a 3 jours, 3 jours avant Noël, depuis 3 jours (3 days ago, 3 days after Christmas, for the last 3 days) the measure “3 days” is projected on the temporal axis. (p. 109, my emphasis)
To my knowledge, however, the possibility illustrated in (2), where the same temporal particles occur in constructions involving counting rather than measuring (há três domingos/three Sundays ago), has not been discussed in the literature (in connection with particles of the type at stake). From a descriptive point of view, it is worth noticing that at least three different sorts of entities – to wit: intervals, eventualities and “ordinary” individuals/objects – may be involved in the counting operation associated with these time-denoting expressions (see Table 1). In connection, three different types of expressions may occur, in the relevant position, in combination with the temporal particles under analysis: (common) predicates of times, situational predicates, and object/individual-denoting predicates. In general, the use of this relatively complex type of temporal expressions appears to be particularly constrained, their counterparts with predicates of amounts of time being possibly more common. One fact that seems to favour the use of these expressions (though it is not crucial) is the relevance of the mentioned type of interval in connection with the type of described eventuality (world knowledge obviously interfering). So, for instance, sentences (7) and (8) are possibly less likely to appear with Tuesdays than with Sundays or weekends, in (7) and (8) respectively (or, for that matter, with measure nouns like weeks).
Table 1. Time-denoting expressions involving counting from anchor points Type of counted entity A intervals
B eventualities
C ordinary individuals/objects
Type of relevant com- Illustrative sentences in Portuguese and plement English predicates of times (three weekends/nights/ Sundays/summers. . . ) situational predicates (three classes/elections/ meals/Olympic Games. . . ) individual/objectdenoting predicates (three cigarettes/ papers. . . )
O Paulo esteve nos Alpes há três fins-desemana (atrás). Paulo was in the Alps three weekends ago. O professor começou a analisar a obra de Goethe há três aulas (atrás). The teacher started to analyse Goethe’s work three classes ago. O Paulo está a fumar cigarros uns atrás dos outros. Há três cigarros (atrás) começou a tossir. Paulo is smoking cigarettes one after the other. Three cigarettes ago he started to cough.
TSL[v.20020404] Prn:29/08/2002; 15:26
F: PBNS03.tex / p.5 (49)
Temporal counting from anchor points
(7) a.
O Paulo não esteve na igreja há três {domingos/terças-feiras} (atrás). “the Paulo not was in-the church there-is three {Sundays/Tuesdays} (behind)” b. Paulo was not in church three {Sundays/Tuesdays} ago.
(8) a.
O Paulo teve de trabalhar há três {fins-de-semana/terças-feiras} (atrás). “the Paulo had to work there-is three {Sundays/Tuesdays} (behind)” b. Paulo had to work three {weekends/Tuesdays} ago.
The first group of expressions in Table 1 – that is, those involving (common) predicates of times – essentially involve what some authors term “calendar nouns”, i.e. expressions which represent recursive intervals of the time axis, like parts of the day (afternoons, evenings, nights), days of the week, weekends, months or seasons of the year, for instance. Searching the British National Corpus (henceforth, BNC), numerous instances of these expressions were found – e.g. nights ago, summers ago or Saturdays ago (39, 10, and 5 instances, respectively). For obvious pragmatic reasons, the commonest forms are those with low cardinals (two, three, or four) and vague quantifiers (like a couple of or a few). There are, however, mainly in sport reports, a few cases with higher exact numerals, like the following: (9) a.
“Forty three summers ago, England suffered their greatest humiliation when the likes of Tom Finney, Billy Wright and Stan Mortensen lost 1-0 to the USA in the 1950 World Cup.” (K3A 1256) b. “If those boys in Karachi had found an 8 to hook on the scoreboard those 33 long winters ago, instead of rummaging fruitlessly about on all fours (. . . ).” (CU1 534)
It must be noted that common nouns like semana/week, mês/month or ano/year, when used as calendar terms (and not as measure nouns), form expressions of this type as well. The difference is that – contrary to what happens with nouns like domingo/Sunday or fin-de-semana/weekend – the relevant counted intervals are adjacent. Thus, time-denoting expressions like há três anos/three years ago are ambiguous: they may involve time measurement – just like in (6) – or counting (of calendar years) – just like in (2) – the latter case being illustrated in (11) below: (10) há três anos/three years ago – time measurement – “a moment surrounding the 1095th day (approximately) in the past of the utterance time”; – counting – “the third calendar year in the past of the utterance time.”
TSL[v.20020404] Prn:29/08/2002; 15:26
F: PBNS03.tex / p.6 (50)
Telmo Móia
(11) a.
No ano passado, foram construídos 250 km de novas estradas, em contraste com 110 km há dois anos e 95 km há três anos. “in-the year past, were built 250 km of new roads, in contrast with 110 km there-is two years and 95 km there-is three years” b. Last year, 250 km of new roads were built, in contrast with 110 km two years ago, and 95 km three years ago.
Schematically, the difference can be depicted as follows: (12)
há três anos / three years ago
COUNTING
[ano / year as CALENDAR NOUN]
utterance time
| (calendar) year3 || (calendar) year2 || (calendar) year1 |
|
1095 days há três anos / three years ago
MEASUREMENT
[ano / year as MEASURE NOUN]
As for the cases in line B of Table 1, the relevant situational predicates obviously refer to recurrent eventualities. These eventualities may occur with a (more or less fixed) regular cycle or not – compare, for instance, the differences between Jogos Olímpicos/Olympic Games, refeições/meals, eleições/elections, aulas/classes or viagens/trips, all of which may occur in the relevant environment. The use of this type of complements in the time-denoting expressions under discussion seems to be strongly restricted, although it may take place if an adequate context is supplied, as in the examples given in Table 1, or in the following ones taken from the BNC (where 85 and 6 instances of seasons ago and games ago, respectively, are registered): (13) a.
“Allison, who took over Rovers four games ago after the sacking of Dennis Rofe, has already hauled them off the bottom of the table (. . . ).” (CBG 8775) b. “It was Robson’s return four games ago that helped spark a maximum haul of 12 points.” (CBG 10414)
As for the (apparently, not very frequent) expressions in line C of Table 1, the designated objects are also associated with given intervals of the time axis (in the examples given, cigarettes with the moment they are smoked) and, by way of this association, ordered in time. Obviously, a special context, where the counted objects are involved in some form of temporal ordering, is required, as in the following example from the BNC:
TSL[v.20020404] Prn:29/08/2002; 15:26
F: PBNS03.tex / p.7 (51)
Temporal counting from anchor points
(14) “There had been no boys left to help Jimmy because since five minutes and three hundred pounds ago, three of them had started working for us and a fourth one was lying in the toilet presently not working for anybody.” (H80 1899) Before proceeding to a brief discussion of some restrictions affecting the use of these expressions, two side-notes are in order at this point. The first is that, although I am concentrating here on expressions with há in Portuguese and ago in English, many other phrases (with different particles) belong in this group of timedenoting expressions as well – see (15)–(16) below. In fact, as can be easily seen, the class of expressions under analysis may be subdivided according to different parameters such as (i) the direction of the operation – backwards or forwards in time – and (ii) the type of anchor point – a deictic or anaphoric temporal perspective point, or a time set by an explicit (referentially independent) complement (see Móia 2000, Chapter 7, for a more thorough analysis): –
forward counting or measurement from a deictic temporal perspective point: (15) a.
daqui a três {fins-de-semana/horas} “from-here to three {weekends/hours}” b. three {weekends/hours} from now
–
backward or forward counting or measurement from a time set by an explicit (referentially independent) complement: (16) a.
três {fins-de-semana/meses} {antes/depois} das eleições “three {weekends/months} {before/after} of-the elections” b. three {weekends/months} {before/after} the elections
The second note is to briefly justify why I am classifying all the relevant Portuguese há-expressions and the English ago-expressions as time-denoting phrases, and not as locating adverbials, contrary to the usual categorisation in the literature. As is known, a clear dividing line between time-denoting expressions and temporal locating adverbials is not easy to draw, given that some expressions – like yesterday or last week, for instance – may occur with the same superficial form in the typical contexts of both categories (cf. e.g. yesterday was a nice day vs. John left yesterday). The relevant há- and ago-expressions belong in this group of ambivalent phrases, as illustrated in (17)–(19). The first two of these examples contain occurrences of such phrases in the typical contexts of time-denoting expressions – verb complement position, in (17), and complement of an explicit temporal preposition (até/until), in (18). Conversely, in (19), they surface as ‘full’ adverbials.
TSL[v.20020404] Prn:29/08/2002; 15:26
F: PBNS03.tex / p.8 (52)
Telmo Móia
(17) a.
A reunião estava marcada para há duas horas (atrás). “the meeting was scheduled for there-is two hours (behind)” b. The meeting was scheduled for two hours ago.
(18) a.
O Paulo esteve na sala até há três horas (atrás). (cf. (2a)) “the Paulo was in-the room until there-is three hours (behind)” b. Paulo was in the room until three hours ago. (cf. (2b))
(19) a.
O Paulo saiu da sala há três horas (atrás). “the Paulo left of-the room there-is three hours (behind)” b. Paulo left the room three hours ago.
Now, by postulating a null locating preposition with a value close to that of in, on or at, in these latter contexts – as I advocate in Móia (2000) and assume here – a uniform analysis is achieved, in which the relevant há- and ago-expressions are always taken as time-denoting expressions. Accordingly, (19) will be analysed as follows: (19 ) a. O Paulo saiu da sala Øem há três horas (atrás). b. Paulo left the room Øat three hours ago.
. Some particular restrictions on the combination desde há in Portuguese and since . . . ago in English In this section, I will concentrate on a particular instance of the time-denoting expressions described in the previous section, namely that with the combination desde há in Portuguese, and since. . . ago in English, which seems to be subject to specific restrictions. As was said before, concerning sentences (1) and (2), the difference between há (n periods) and desde há (n periods) in Portuguese, as well as that between for (n periods) and since (n periods) ago in English (where “n periods” stands for a predicate of the groups described in Table 1 above), lies essentially in the relevance of the intervals between the counted periods. Now, this difference vanishes – as was briefly mentioned, in connection with sentences (5) and (6) – in structures where predicates of amounts of time (x-time, henceforth) occur in the same context as “n periods”. This seems due to the fact that continuous intervals are associated with predicates of amounts of time in these contexts. Observe the following sentences (which are of the same type as (5) and (6)): (20) a.
O Paulo está no hospital há dois meses. “the Paulo is in-the hospital there-is two months” b. O Paulo está no hospital desde há dois meses. “the Paulo is in-the hospital since there-is two months”
⇔
TSL[v.20020404] Prn:29/08/2002; 15:26
F: PBNS03.tex / p.9 (53)
Temporal counting from anchor points
(21) a. Paulo has been in the hospital for two months (now). ⇔ b. Paulo has been in the hospital since two months ago.[odd, or not used] The two Portuguese sentences in (20) – the second containing counterparts of since and ago – are equivalent and equally grammatical. The sentences in (21) illustrate a curious fact about English, to wit: that the logically admissible combination of since and an ago-expression with predicates of amounts of time is normally not used and/or accepted. In fact, sentence (21b) is considered odd by most speakers. In order to express the temporal relations at stake, English normally resorts to a forconstruction, as in (21a), which, though formally involving a temporal expression of a different category, conveys the same temporal information. Now, the oddity of the combination since. . . ago in (21b) is somehow unexpected, given that, once we assume (as I do) a time-denoting analysis for ago-phrases, they are expected to freely combine with temporal locating prepositions, such as since or until. As for the combination until. . . ago, it seems unproblematic in English, like in Portuguese for the counterparts of these particles: (22) a.
O Paulo esteve no hospital até há dois meses. “the Paulo was in-the hospital until there-is two months” b. Paulo was in the hospital until two months ago.
As for the combination since x-time ago, its oddity seems, at a closer look, more a matter of use than of strict grammaticality. In fact, the BNC contains several records of this combination: (23) a.
“With exasperation, he said, ‘Since when? Since Christmas?’ Doone said stolidly, ‘Since ten days ago.’ ” (ADY 679) b. “Nothing in the metal’s fundamentals has changed since a month ago, when the price languished at a seven-year low of $126 an ounce.” (CR7 2794) c. “I haven’t played with anyone for like since months ago!” (KE1 805, KE1 1418) d. “Every room here has been booked since a year ago, and I was dearly hoping Donna would screw up the nerve to send her packing (. . . )” (JY6 2805) e. “(. . . ) and they’ve probably been isolated since fourteen thousand years ago, with that every population has gone slightly different to, to the next one.” (F8H 74) f. “(. . . ) I well recall the Second Reading debate of the Shops Bill 1986 which followed upon the discussions and deliberations of the Auld committee which, since as long ago as 1985, had been engaged in a series of debates about Sunday trading.” (HHX 11597)
TSL[v.20020404] Prn:29/08/2002; 15:26
F: PBNS03.tex / p.10 (54)
Telmo Móia
g. “Animals have been anthropomorphised since as long ago as Egyptian times.” (BMD 337) In contrast with these cases, others exist where (according to my informants) the combination of since and ago appears to be generally accepted. This is the case of (2b) above, for instance. As far as I can judge from the information I gathered, the non-problematic cases in English seem to require the concurrence of three facts:5 i.
involvement of discontinuous intervals (e.g. Sundays, weekends, nights, seasons), corresponding to temporal predicates like those in Table 1 above (given that the combination with predicates of amounts of times, which are associated with continuous intervals, is seldom used) – cf. the contrast between since three Sundays ago, in (24a) and since three weeks ago, in (24d) below; ii. relevance of the periods between those intervals (otherwise, a for-construction is preferred) – cf. since three Sundays ago, in (24a) vs. for three Sundays, in (24b); iii. (contextual) salience of the period marking the starting point of the location time, which justifies the choice of the relatively complex expression “n periods ago” (otherwise, a construction with “for x-time” is preferred) – cf. since three Sundays ago, in (24a) vs. for three weeks, in (24c). (24) a.
Paulo hasn’t been to church since three Sundays ago. (This was the last time the priest saw him.) b. Paulo hasn’t been to church for three Sundays (now).[different truthb . Paulo hasn’t been to church for the last three Sundays. conditions] c. Paulo hasn’t been to church for three weeks (now). [relevant Sunday c . Paulo hasn’t been to church for the last three weeks. is not salient] d. (?)Paulo hasn’t been to church since three weeks ago.
When these facts concur, there seems to be no alternative construction to “since n periods ago” that is formed by a single temporal preposition and a time-denoting complement, of the type “prep x-time” or “prep the last x-time”. The nonexistence of such an alternative apparently closely correlates with the acceptance of the combination since. . . ago. The examples given above contain atelic descriptions. In combination with telic descriptions, constructions with since. . . ago behave similarly, the main difference being that the temporal preposition now occurring in the approximately equivalent constructions with predicates of amounts of time is not for, but rather within, in/on or during (together with the last). Compare, for instance, (25) below, which represents a sum of telic events, with (26), which basically differs in that it does not make the mentioned third Sunday salient:
TSL[v.20020404] Prn:29/08/2002; 15:26
F: PBNS03.tex / p.11 (55)
Temporal counting from anchor points
(25) Thirty weddings have been celebrated in this church since three Sundays ago. (This was the day weddings restarted to be celebrated here, after a break of nearly three years.) (26) Thirty weddings were celebrated in this church {within/in/during} the last three weeks. (27) Thirty weddings were celebrated in this church since three weeks ago. [odd, or not used] Likewise, we observe that constructions with “since n periods ago” – as (25) – and those with e.g. “within the last n periods” – as (28) below – are not equivalent, differing – like (2) differs from (1) – in that the latter only involves the mentioned three Sundays and not the intervening periods:6 (28) Thirty weddings have been celebrated in this church {within/on/during} the last three Sundays.
. Semantic-pragmatic differences between constructions involving counting of ordered entities The constructions involving counting of ordered entities of the types illustrated in (1) and (2) exhibit several semantic-pragmatic differences. In this section, I will elaborate a bit more on two of them. One difference, which has already been mentioned several times, concerns the relevance of periods intervening between the counted intervals. The combination with ‘adverbs of temporal quantification’ – illustrated in (29) and (30) – shows that different intervals are involved, in accordance with schema (3): (29) a.
Um avião etíope aterra em Lisboa todos os dias desde há três domingos (atrás). “an aeroplane Ethiopian lands in Lisbon all the days since there-is three Sundays (behind)” b. An Ethiopian aeroplane has been landing in Lisbon everyday since three Sundays ago.
(30) a. *Um avião etíope aterra em Lisboa todos os dias há três domingos. “an aeroplane Ethiopian lands in Lisbon all the days there-is three Sundays” b. *An Ethiopian aeroplane has been landing in Lisbon everyday for (the last) three Sundays.
TSL[v.20020404] Prn:29/08/2002; 15:26
F: PBNS03.tex / p.12 (56)
Telmo Móia
It was also briefly mentioned that if the periods between the discontinuous mentioned intervals are immaterial (world knowledge, for instance, interfering), then constructions with for (in English) and há (in Portuguese) are preferred, this preference being possibly stronger in English than in Portuguese.7 Therefore, assuming – in accordance with current knowledge – that soup eating only takes place during meals, while wine drinking may occur both during and between meals, the contrasts in (31) and (32) are predicted. Sentences in (31) are slightly odd with desde há and since. . . ago, while those in (32) are not. This seems due to the fact that sentences in (32) with desde há and since. . . ago are not equivalent to those with há or for, respectively, whereas in (31) this equivalence exists (given the irrelevance of the periods between meals). (31) a.
O Paulo não come sopa {?desde há duas refeições (atrás)/há duas refeições}. “the Paulo not eats soup {since there-is two meals (behind)/there-is two meals” b. Paulo hasn’t eaten soup {?since two meals ago/for two meals}.
(32) a.
O Paulo não bebe vinho {desde há duas refeições (atrás)/há duas refeições}. “the Paulo not drinks wine {since there-is two meals (behind)/there-is two meals” b. Paulo hasn’t drunk wine {since two meals ago/for two meals}.
A second difference between the constructions like those in (1) and (2) has to do with the pertinence – for counting purposes – of the period containing the utterance time. Take for example the Portuguese expressions desde há três domingos/há três domingos, or their English counterparts since three Sundays ago/for three Sundays, and consider a scenario where the utterance takes place on a Sunday. In the constructions with desde há and since. . . ago, the general consensus among speakers seems to be (despite some hesitation) that the “Sunday of utterance” is never one of the three that are being counted. Thus, sentences like (33) below always count back three entire Sundays (irrespective of whether any landing occurred on the Sunday of utterance or not): (33) a.
Um avião etíope aterra neste aeroporto desde há três domingos. “an aeroplane Ethiopian lands at-this airport since there-is three Sundays” b. An Ethiopian aeroplane has been landing at this airport since three Sundays ago.
Constructions with há and for are different, one of two situations possibly holding: (i) if the “Sunday of utterance” contains (prior to the utterance) an instance of the
TSL[v.20020404] Prn:29/08/2002; 15:26
F: PBNS03.tex / p.13 (57)
Temporal counting from anchor points
events of the type described in the main clause (landings of Ethiopian aeroplanes, here), then it is taken as one of the three that are being counted – see (34), and its possible continuation in a and b . Otherwise, that is, if no landing has yet occurred on that Sunday, by the utterance time, then it is irrelevant for counting purposes – see (34) again, and its continuation in a and b . (34) a.
Um avião etíope aterra neste aeroporto há três domingos. “an aeroplane Ethiopian lands in-this airport there-is three Sundays” b. An Ethiopian aeroplane has been landing at this airport for three Sundays (now).
a . . . . Hoje chegou mais cedo que das outras duas vezes. “today [it] arrived more soon than of-the other two times” b . . . . Today it arrived sooner than the other two times. a . . . . Vamos ver se hoje também aterra e no mesmo sítio das outras três vezes. “let-us see if today [it] also lands and at-the same place of-the other three times” b . . . . Let us see if it also lands today, and at the same place as the other three times.
. Conclusion In this paper, two different temporal constructions involving counting from anchor points – which, to my knowledge, have not been analysed in the literature – were discussed, their semantic and pragmatic specificity being partially explored. Concurrently, we argued for the need to distinguish between two categories of temporal phrases occurring in these constructions (which, interestingly, have homonymous elements in Portuguese): (i) one whose members count the number of contiguous intervals of a given type containing instances of given events, illustrated in (1); (ii) another one whose members are merely time-denoting phrases, illustrated in (2). The semantic and pragmatic differences that were pointed out here should be taken as a contribution to a more thorough study – pending further research – of the vast subclass of anchor-dependent temporal expressions in natural languages, of which they are distinguished instances.
TSL[v.20020404] Prn:29/08/2002; 15:26
F: PBNS03.tex / p.14 (58)
Telmo Móia
Notes . In European Portuguese, the adverbial form atrás (literally ‘behind’) optionally co-occurs with há – without any semantic variance – in sentences of type (2a). In Brazilian Portuguese, there seems to be a growing trend to avoid the verbal form (há) and only use the adverbial atrás. This is more conspicuous in sentences like (2a), where the temporal expression is dependent on an explicit preposition (e.g. desde [since], até [until] or de [from/of ]), than in sentences like (4a) – cf. Móia and Alves (2000). In this paper, only European Portuguese is taken into account. . As will be mentioned in Section 3, the utterance time may overlap “Sunday1 ” in structures like (1), but not in those like (2). . Portuguese time-denoting há-expressions – as those in (2a) and (4a) – contrast with the type of há-expressions in (1a) in at least the following three properties: (i) they optionally co-occur with the adverbial form atrás, (ii) they may be preceded by explicit temporal prepositions, like the counterparts of since and until, and (iii) they combine both with atelic and telic descriptions (whereas those in (1a) only combine with atelics). . I take these expressions to denote, as a whole, the entire nth instance of the relevant period, counting backwards from the anchor point, given the interpretation of sentences like the following: (i)
O número de bilhetes (para este museu) vendidos aos fins-de-semana é o seguinte: 1200 no fim-de-semana passado, 1100 há dois fins-de-semana, 800 há três fins-desemana. “the number of tickets (for this museum) sold on weekends is the following: 1200 in-the weekend past, 1100 there-is two weekends, 800 there-is three weekends”
(i ) The number of tickets (for this museum) sold on weekends is as follows: 1200 last weekend, 1100 two weekends ago, 800 three weekends ago. (ii) 115 aviões aterraram neste aeroporto há três fins-de-semana, o que constituiu um novo recorde. “115 planes landed in-this airport there-is three weekends, the what constitutes a new record” (ii ) 115 planes landed at this airport three weekends ago, which sets a new record. The processing of these sentences requires consideration of all the relevant events (ticketsales or landings) that occurred within a given stretch of time. In these cases, this stretch is obviously a whole weekend (no part of it being irrelevant). . The requirements (ii) and (iii) apply, with the relevant adaptations, to Portuguese constructions with “desde há n periods” as well (although the second one – discussed below, in Section 3 – appears to be less compulsory in Portuguese). . Facts in Portuguese are similar for desde há constructions. See the examples below (where se is a clitic pronoun, marking a passive construction, which is not translated in the glosses):
TSL[v.20020404] Prn:29/08/2002; 15:26
F: PBNS03.tex / p.15 (59)
Temporal counting from anchor points
(i)
Celebraram-se trinta casamentos nesta igreja desde há três domingos (atrás). [like (25)] “Celebrated se thirty weddings in-this church since there-is three Sundays (behind)”
(ii) Celebraram-se trinta casamentos nesta igreja nas últimas três semanas. “Celebrated se thirty weddings in-this church in-the last three weeks”
[like (26)]
(iii) Celebraram-se trinta casamentos nesta igreja desde há três semanas (atrás). [not odd, unlike (27)] “Celebrated se thirty weddings in-this church since there-is three weeks (behind)” (iv) Celebraram-se trinta casamentos nesta igreja nos últimos três domingos. [like (28)] “Celebrated se thirty weddings in-this church in-the last three Sundays” . See, however, the following example from the BNC: “I’ve been coming to games at the Manor since I was 10 (. . . ) since three seasons ago I haven’t missed a match – I’ve been to them all” (ECN 1624).
References Asher, N. et al. (1995). Spatial, Temporal and Spatio-Temporal Locating Adverbials in Discourse. In P. Amsili, M. Borillo, & L. Vieu (Eds.), Workshop Notes of the 5th International Workshop on Time, Space and Movement TSM’95 (pp. 101–119). Bras, M. (1990). Calcul des Structures Temporelles du Discours. Ph.D. thesis, Université Paul Sabatier de Toulouse. Bras, M., & Molinès, F. (1993). Adverbials of Temporal Location: Linguistic Description and Automatic Processing. In Proceedings of the 26th Colloquium of Linguistics. Niemeyer, Tübingen: Linguistiche Arbeiten. Kamp, H., & Reyle, U. (1993). From Discourse to Logic. Introduction to Modeltheoretic Semantics of Natural Language, Formal Logic and Discourse Representation Theory. Kluwer: Dordrecht. Móia, T. (2000). Identifying and Computing Temporal Locating Adverbials with a Particular Focus on Portuguese and English. Ph.D. thesis, Universidade de Lisboa. Móia, T., & Alves, A. (2000). Sobre a Expressão de Distâncias Temporais no Português Europeu e no Português Brasileiro. Actas do XVI Encontro Nacional da Associação Portuguesa de Linguística (Colóquio Português Europeu e Português Brasileiro Unidade e Diversidade na Passagem do Milénio – PEPB-2000) (pp. 699–713). Lisboa: APL. Molinès, F. (1989). Acceptabilité et Acceptation des Adverbiaux de Localisation Temporelle: Grammaire ou Distionnaire. Mémoire de DEA en Sciences du Langage, Université de Toulouse Le Mirail.
TSL[v.20020404] Prn:17/10/2002; 10:36
F: PBNS04.tex / p.1 (61)
On the semantics and pragmatics of situational anaphoric temporal locators in Portuguese and in English Ana Teresa Alves Universidade dos Açores, Portugal
Introduction Several authors, among whom Kamp and Reyle (1993) and Móia (2000), distinguish between two types of temporal expressions: those that directly represent intervals of time, e.g., 2001, next week, last month, today, and those that represent intervals of time indirectly, such as eventuality descriptions like John visited Paris, Mary was hospitalised, or the school will hire a new teacher. In the Kamp and Reyle (1993) framework, the former directly introduce discourse referents for temporal expressions, i.e., discourse referents of type t, while the latter introduce discourse referents for eventualities, that is, discourse referents of type e(vent) or s(tate). To account for the fact that eventuality descriptions also represent time, Kamp and Reyle introduce a function loc, which associates each eventuality with the smallest time interval it occupies. The distinction between these two types of temporal expressions – I’ll dub the former as direct and the latter as indirect – proved to be important in the study of temporal anaphors, inasmuch as some of these constrain their antecedents to be either a direct or indirect temporal expression. In this paper, I’ll focus on anaphoric temporal expressions that refer back to time intervals that are defined by eventuality descriptions, and try to describe their semantic and pragmatic licensing conditions.
TSL[v.20020404] Prn:17/10/2002; 10:36
F: PBNS04.tex / p.2 (62)
Ana Teresa Alves
.
A tripartite classification of anaphoric temporal locators: Direct, indirect, and ambivalent
Regarding the type of temporal expression by means of which anaphoric temporal locators define a location interval, these locators might be sub-classified as direct, indirect and ambivalent, as illustrated in the Table 1. Table 1. Subcategorisation of anaphoric locators wrt. the type of possible antecedent Direct nesse mês (in) that month no mesmo ano/ (in) the same year ...
{depois/antes} disso {after/before} that nessa altura/ at that time até {lá, então}/ until then ...
Indirect durante esse tempo/ during that time entretanto/ (in the) meantime enquanto isso/ (in the) meanwhile ...
examples to be considered in this paper
Direct temporal locators define a location interval by means of direct timedenoting expressions only. On the other hand, indirect locators define location intervals by means of eventuality descriptions only. Ambivalent locators are those that might define location intervals by means of expressions of both kinds. Direct and indirect anaphoric locators are illustrated in appropriate context (1) and (2), respectively. (1) O Paulo chegou a Paris em [Maio de 1999]i . A Maria também chegou a Paris n[esse mês]i . Paul arrived in Paris in [May, 1999]i. Mary also arrived in Paris [that month]i . (2) a.
[O Paulo visitou Paris]i em [1980]*i. Durante [esse tempo]i , a Maria tomou conta do bebé. [Paul visited Paris]i in [1980]*i . During [that time]i Mary looked after the baby. b. [A noite passada]*i o [Paulo corrigiu dez testes]i . Enquanto [isso]i a Maria escreveu duas cartas. [Last night]*i [Paul corrected ten tests]i . [Meanwhile]i Mary wrote two letters.
In (1), the interval associated with esse mês/that month coincides with the time interval associated with the direct temporal expression Maio de 1999/May, 1999. In (2a) and (2b), the location interval associated with durante esse tempo/during that
TSL[v.20020404] Prn:17/10/2002; 10:36
F: PBNS04.tex / p.3 (63)
On the semantics and pragmatics of situational anaphoric temporal locators
time and enquanto isso/meanwhile corresponds, respectively, to the running time of the eventualities ‘Paul visit Paris’ and ‘Paul correct ten tests’; Clearly, the expressions 1980 and a noite passada/last night do not provide appropriate antecedents for the anaphoric expressions. See, for the sake of illustration, the simplified DRSs for sentences (1) – where, besides other simplifications, the contribution of also is ignored – and (2a), respectively. x y t n e e1 t 1 Paul (x)
x y t n e s t1
Mary (y)
Paul (x)
May, 1999 (t)
Mary (y)
e
1980 (t)
eÍt
e
e: x arrive in Paris month (t1) t1 = t
t1 = loc (e)
e1 < n
s
e1 Í t1
s Í t1
e1: y arrive in Paris (DRS-1)
e: x visit in Paris
s: y look after the baby (DRS-2a)
Ambivalent expressions are illustrated in sequences (3) and (4), which show that the expression then can pick up both a direct time-denoting expression and a situational/indirect temporal expression. (3) [A Maria muda-se para Paris]i em 2004. Fica em Lisboa até [lá]i . [Mary will move to Paris]i in 2004. She’ll stay in Lisbon until [then]i . (4) A Maria marcou o encontro para [a próxima segunda-feira]i . Está fora até [lá]i . Mary scheduled the meeting for [next Monday]i . She’ll be out of town until [then]i . In (3), lá/then refers back to the running time of the moving eventuality and in (4) to the time interval represented by next Monday. The DRT conditions associated to each of these two possibilities are given in (3 ) and (4 ), respectively, where t is the discourse referent introduced by the anaphor and t the discourse referent associated with its antecedent.
TSL[v.20020404] Prn:17/10/2002; 10:36
F: PBNS04.tex / p.4 (64)
Ana Teresa Alves
(3 ) [loc (ev) = (t )], [t = t ] (4 ) [next Monday (t )], [t = t ]
. Formal account of the expressions durante esse tempo/during that time, enquanto isso/(in the) meanwhile, entretanto/(in the) meantime within DRT . Differences regarding the type of anaphora The anaphoric expressions at stake exhibit differences wrt. the type of anaphora in which they might occur giving rise to three types of cases: 1. Cases in which the anaphor refers back to the time interval corresponding to the “running time” of an eventuality. (6) O Paulo visitou Paris em 1980. Durante esse tempo a Maria tomou conta do bebé. Paul visited Paris in 1980. During that time Mary looked after the baby. (7) Ontem à noite o Paulo fez o jantar. Enquanto isso a Maria limpou o quarto dela. Last night Paul cooked dinner. Meanwhile Mary cleaned her bedroom. In these sentences, the anaphors esse tempo, isso, that time and meanwhile take as their antecedent the discourse referent representing the running time of the eventualities described in the first clause of each sequence. 2. Cases in which the anaphor refers back to a time interval stretching from the speech time to the “running time” of an eventuality. (8) No próximo ano contrataremos um novo professor. Entretanto a Maria ensina o curso 230. (9) Next year the school will hire a new teacher. In the meantime Maria will be teaching course 230. In cases like (8)–(9), the anaphors entretanto and the meantime refer back to time intervals whose left boundary is the speech time and whose right boundary is defined by the eventuality described in the first clause. 3. Cases in which the anaphor refers back to a time interval stretching from the time of an eventuality to the speech time.
TSL[v.20020404] Prn:17/10/2002; 10:36
F: PBNS04.tex / p.5 (65)
On the semantics and pragmatics of situational anaphoric temporal locators
(10) O Paulo licenciou-se em 1980. Entretanto ganhou a lotaria, mudou-se para o Hawaii, e desistiu da carreira. (11) ?Paul graduated in 1980. In the meantime, he won the lotto, moved to Hawaii, and gave up his career. In (10)–(11), the anaphors entretanto and the meantime refer back to a time interval whose initial boundary is set by the eventuality described in the initial clause and whose final boundary is defined by the speech time. Table 2 shows the type of antecedent accepted by each anaphor. Table 2. Correspondence between types of antecedents and anaphors Types of antecedents 1 cf. (6)–(7)
2 cf. (8)–(9) 3 cf. (10)–(11)
t
Anaphors
= loc(ev)
(durante) esse tempo, (enquanto) isso, entretanto (during) that time, (in the) meantime, (in the) meanwhile t , [beg(t ) = n], [end(t ) ⊆ loc(ev)] entretanto, (enquanto) isso, (in the) meantime, (in the) meanwhile t , [beg(t ) ⊆ loc(ev)], [end(t ) = n] entretanto, ?(in the) meantime
There is a major difference between cases presented in 1, on the one hand, and those in 2 and 3, on the other hand: in the latter, unlike what happens in the former, the discourse referent associated with the antecedent of the anaphor is not introduced in the DRS by any expression in the previous discourse. It must be introduced in the DRS as the result of an inferential process: the mention of an eventuality occurring in a time anterior or posterior to the speech time allows the speaker to refer back to the time interval stretching from then to the speech time, or from the speech time to then. As it is not the object of this paper, I won’t elaborate here on this type of inference. In this paper, the cases on which I’ll mainly focus are those of type 1. It is on the basis of cases of type 1 – anaphora with explicit antecedents – that the anaphoric expressions at stake are classified in Table 1 as situational.
. Aktionsart constraints on the eventuality descriptions providing the anaphora antecedent Immediately above, I drew attention to the fact that in cases 1, unlike those in 2 and 3, the anaphora antecedent is provided by an expression in the previous discourse. These expressions, which are eventuality descriptions, must verify certain
TSL[v.20020404] Prn:17/10/2002; 10:36
F: PBNS04.tex / p.6 (66)
Ana Teresa Alves
constraints, namely constraints related to the aktionsart of eventualities. The categorisation of eventuality descriptions in classes such as states, activities, accomplishments, and achievements (cf. Vendler 1967), on the basis of properties such as temporal duration, homogeneity and nuclear structure, has proved to be pertinent in the analysis of the linguistic distribution of certain expressions, in particular time measure adverbials (e.g., for three hours, in five minutes). Temporal location anaphors seem to be sensitive as well to the aktionsart type of the expressions that provide their antecedents. Consider the examples below – namely the acceptability contrast between (12)–(14) and (15)–(16) – which show that the eventualities providing the anaphoric antecedents must be durative/extended. (12) John was hospitalised last year. During that time/Meanwhile Ana looked after their kid. (13) John wrote a novel in 1999. During that time/Meanwhile Ana looked after their baby. (14) John swam the whole morning yesterday. During that time/Meanwhile Ana played ping pong. (15) #Johnny hiccupped. During that time his mother did not feed him. (16) #John won the Wimbledon men’s final. During that time he was happy. In cases under 2 and 3, in which the eventuality in the first clause just sets the initial or the final boundary of the time interval referred to by the anaphor, similar aktionsart constraints do not arise. These eventualities may be extended or punctual, as shown below: (17) John only leaves for Lisbon on Sunday. In the meantime he can still use his office. (18) ?Paul graduated in 1980. In the meantime, he won the lotto, moved to Hawaii, and gave up his career. (19) John will be the president of the company from December on. But meanwhile a number of steps will have to be taken.
..
Blocking of aspectual transitions in some constructions with durante esse tempo/during that time, enquanto isso/(in the) meanwhile, entretanto/in the meantime Although the aktionsart constraints affecting the anaphoric complements in cases of anaphora like those in 1 were (to my knowledge) not mentioned before in the literature, they should not be surprising. They are similar to those affecting non-
TSL[v.20020404] Prn:17/10/2002; 10:36
F: PBNS04.tex / p.7 (67)
On the semantics and pragmatics of situational anaphoric temporal locators
anaphoric complements of the operators during and while. Consider the examples below, where the relevant reading of the subordinate clause is the one in which there is no iteration of the described eventuality: (20) *Ana did not feed the baby while he hiccupped.
[episodic reading]
(21) *John was happy while he won the Wimbledon Tournament. [episodic reading] These same two sentences, given above as unacceptable, are, however, acceptable in the reading in which there is an iteration of the eventuality described in the subordinate clause. Moens and Steedman (1988), for instance, account for the second reading by means of an aktionsart transition: via iteration a non-extended/atomic eventuality is transformed into an extended eventuality. (22) Ana did not feed the baby while he hiccupped.
[iterative reading]
(23) John was happy while he won the Wimbledon Tournament. [iterative reading] If we turn to cases of anaphora again – cf. (24)–(25), where the reading associated with the non-temporal value of the anaphoric locator is to be disregarded – we notice that transitions of this type are blocked when the complement of the temporal operator is an anaphor. This is particularly clear in Portuguese, where the counterpart of meanwhile – enquanto isso – literally corresponds to ‘while that’. Compare sequences a. with sequences b. below, noticing the blocking in b. of aktionsart transitions equivalent to those that in a. transform the non-extended eventualities described in subordinate clauses into extended ones. (24) a.
Ana lost six kilos during the time in which she had Japanese food for lunch. [iter. read.] b. *Ana had Japanese food for lunch. During that time she lost six kilos. [iter. read.]
(25) a.
A Monica foi a segunda do ranking enquanto a Stefi ganhou o torneio de Wimbledon. [iter. read.] Monica held the number two ranking while Stefi won the Wimbledon tournament. [iter. read.] b. *A Stefi ganhou o torneio de Wimbledon. Enquanto isso a Monica foi a segunda do rank. [iter. read.] ‘The Stefi won the tournament of Wimbledon. While that the Monica was the second of the rank.’ *Stefi won the Wimbledon tournament. In the meantime/meanwhile Monica held the number two ranking. [iter. read.]
TSL[v.20020404] Prn:17/10/2002; 10:36
F: PBNS04.tex / p.8 (68)
Ana Teresa Alves
To account for the blocking in b. of the aktionsart transitions happening in a., I hypothesise the existence of a general principle according to which the internal features of the antecedents are frozen in the anaphors (e.g. PT isso (=that)) and therefore cannot be changed. This explains why in sentences (24b) and (25b) temporal operators can “read” the aspectual features of their anaphoric complements, but, unlike what happens with their non-anaphoric complements, cannot change them (cf. sentences (24a) and (25a)).
. DRSs and construction rules In this section, I present DRSs for sentences (26), (27) and provide (some of) the rules to build them. I use the Kamp and Reyle (1993) framework with the changes proposed in Móia (2000). (26) A noite passada o Paulo fez o jantar. Entretanto a Maria limpou o quarto. Last night Paul cooked dinner. Meanwhile Mary cleaned her room. (27) O Paulo visitou Paris o ano passado. Durante esse tempo a Maria visitou Londres. Paul visited Paris last year. During that time Mary visited London. (DRS-(27)) x y w z t¢ t t1 tc t1c n e e1
(DRS-(26)) x y t¢ t t1 tc t1c n e e1 Paul (x) Mary (y) last night (tc) e
Paul (x) Mary (y) Paris (w) London (z) last night (tc) e
The first two rules given below are not different than those proposed in Móia (2000) to deal with non-anaphoric adverbial temporal locators. The first rule deals with the S bar node, and the following with the PP node.
TSL[v.20020404] Prn:17/10/2002; 10:36
F: PBNS04.tex / p.9 (69)
On the semantics and pragmatics of situational anaphoric temporal locators
CR.S¢ (adapted from Móia 2000 rule for non-anaphoric adverbials) Triggering S¢
configuration: ¡ Í ¡¢ Î Conk
PP [+temp loc.] P
S
NP
{in, during, Æ}
(the) meantime, (the) meanwhile, that time
Introduce in Uk:
new discourse referents t and ev.
Replace ¡ by:
S (ev) [t]
and
PP[+temp loc.] (ev) [t]
CR.PP (adapted from Móia 2000 rule for non-anaphoric adverbials) Triggering PP (ev) [t]
configuration: ¡ Í ¡¢ Î Conk
Introduce in Uk:
P
NP
in, during, Æ
(the) meantime, (the) meanwhile, that time
new discourse referent t1c
Introduce in Conk: new condition: [t1 = t1c] Introduce in Conk: new condition: [ev Í t1] or [ev O t1] Replace ¡ by: NP(t1c) (the) meantime, (the) meanwhile, that time
The rule below deals with the anaphoric noun phrases.
TSL[v.20020404] Prn:17/10/2002; 10:36
F: PBNS04.tex / p.10 (70)
Ana Teresa Alves
CR.NP Triggering configuration: ¡ Í ¡¢ Î Conk (i)
NP(t1c)
(ii)
NP(t1c)
that time
(the) meantime
Introduce in Uk:
new discourse referent t¢.
(iii)
NP(t1c)
(the) meanwhile
Introduce in Conk: t1c = t¢ Choose a suitable antecedent ev, and introduce a condition t¢ = loc(ev)
. Further constraints on the use of the adverbials enquanto isso, entretanto, (in the) meantime, and meanwhile The data given below clearly show, I believe, that the adverbials at stake in this paper are not always interchangeable. The sequences (28) and (30) illustrate cases in which they are not interchangeable and those in (29) and (31) illustrate cases where they are interchangeable. Notice that the sentence providing the anaphora antecedent is identical in (28) and (29), and in (30) and (31). Consider first the pair (28)–(29). (28) a.
A Ana esteve hospitalizada durante uma semana o ano passado. Durante esse tempo não foi capaz de dormir. Ana was hospitalised for a week last year. During that time she was unable to sleep. vs. b. *A Ana esteve hospitalizada durante uma semana o ano passado. {Entretanto/enquanto isso} não foi capaz de dormir. *Ana was hospitalised for a week last year. {In the meantime/meanwhile} she was unable to sleep.
(29) A Ana esteve hospitalizada durante uma semana o ano passado. {Durante esse tempo/entretanto/enquanto isso} a mãe tomou-lhe conta do bebé. Ana was hospitalised for a week last year.{During that time/in the meantime/meanwhile} her mother looked after her kid.
TSL[v.20020404] Prn:17/10/2002; 10:36
F: PBNS04.tex / p.11 (71)
On the semantics and pragmatics of situational anaphoric temporal locators
Consider now (30) and (31): (30) a.
O Paulo esteve em Paris no Verão passado. Durante esse tempo telefonou para casa todos os dias. Paul was in Paris last summer. During that time he phoned home every day. vs. b. *O Paulo esteve em Paris no Verão passado. {Entretanto/enquanto isso} telefonou para casa todos os dias. *Paul was in Paris last summer. {In the meantime/meanwhile} he phoned home every day.
(31) Paul was in Paris last summer.{During that time/in the meantime/meanwhile} his house was robbed. O Paulo esteve em Paris no Verão passado. {Durante esse tempo/entretanto/ enquanto isso} telefonou para casa todos os dias. According to these data, the use of entretanto, enquanto isso and in the meantime/meanwhile is more restricted than the use of durante esse tempo and during that time. What seems to be relevant here is the relationship between the two eventualities in each sequence. The use of meanwhile and meantime is ruled out in the cases in which the second eventuality is interpreted as being part of the first, as a sub-event of the first: According to our knowledge about trips and hospitalisations, we may think that to sleep bad and to phone home daily may be part of what a person does when she is hospitalised or abroad, respectively. The question arising now is how to account for these constraints. The version of DRT formulated in Kamp and Reyle (1993) does not deal with such questions as the internal structure of eventualities, identity of eventualities, and so on. If in DRT information about the internal structure of eventualities was available and could be checked for every pair of eventualities, then the restrictions under discussion could easily be encoded, being taken as part of the suitability requirements of the antecedent eventuality description (for a discussion about this point, see Alves and Txurruka (2001). If we move to SDRT (cf. Asher 1993), a framework enriched with discourse relations, an alternative analysis is to classify entretanto, enquanto isso and their English counterparts as discourse markers, introducing a discourse relation of temporal parallelism, with an associated post-condition that the second eventuality is not part of the first. Rule 1 <τ, α, β & {entretanto/enquanto isso} (β)> Temporal Parallelism Rule 2 Temporal Parallelism (α, β) → the main eventuality in β is not part of the main eventuality in α. Sentence (30b) is incoherent then, because of the existing conflict between Rule 1 and Rule 3 below, the Elaboration rule presented in Lascarides and Asher (1993).
TSL[v.20020404] Prn:17/10/2002; 10:36
F: PBNS04.tex / p.12 (72)
Ana Teresa Alves
The conflict between the two rules cannot be solved as the antecedent parts of the two rules are not related (cf. the Nixon Diamond in Lascarides and Asher 1993). Rule 3 <τ, α, β> & the main eventuality in β is part of the main eventuality in α Elaboration (α, β) It is worth noticing that non-anaphoric counterparts of enquanto isso/meanwhile, i.e. enquanto-clauses/while-clauses do not exhibit the above-mentioned constraints regarding event structure (or discourse structure, depending on the analysis adopted). This is particularly surprising in Portuguese, where the opposition between enquanto . . . (while . . . ) and enquanto isso (meanwhile) literally corresponds to while . . . and while that, as said before. Compare (32) with (33) and (34) with (35) below: (32) A Ana partilhou um apartamento com o Michel enquanto viveu em Paris na década de oitenta. ‘the Ana shared-pps an apartment with the Michel while [she] lived-pps in Paris in the eighties.’ Ana shared an apartment with Michel while she lived in Paris in the eighties. vs. (33) *A Ana viveu em Paris na década de oitenta. Enquanto isso partilhou um apartamento com o Michel. ‘the Ana lived-pps in Paris in the eighties. While that she shared an apartment with Michel.’ *Ana lived in Paris in the eighties. Meanwhile she shared an apartment with Michel. (34) O Paulo telefonou para casa todos os dias enquanto esteve em Paris no Verão passado. ‘the Paulo phoned to home all the days while [he]was-pps in Paris in-the Summer last.’ Paul phoned home every day while he was in Paris last Summer. vs. (35) *O Paulo esteve em Paris no Verão passado. Enquanto isso telefonou para casa todos os dias. ‘The Paulo was-pps in Paris in-the Summer last. While that [he] phoned home all the days.’ *Paul was in Paris last Summer. Meanwhile he phoned home every day. This discrepancy of constraints contrasts with the uniformity exhibited, for instance, by the following two pairs of anaphoric (cf. a) and non-anaphoric (cf. b) locators. In both cases, a reading according to which the second eventuality is interpreted as a sub-eventuality of the first is ruled out, leading to the oddity of (36).
TSL[v.20020404] Prn:17/10/2002; 10:36
F: PBNS04.tex / p.13 (73)
On the semantics and pragmatics of situational anaphoric temporal locators
(36) a. ?John went to Paris in 1980. He visited the Louvre the same year. b. ?John visited the Louvre the same year he went to Paris. (37) a.
John went to Paris in 1980. He visited a famous museum the same year. b. John visited a famous museum the same year he went to Paris.
. Conclusions In this paper, I proposed a tripartite classification of temporal anaphors according to the type of expression they refer back to. I dubbed as direct those that refer back to expressions that directly represent time intervals – in Kamp and Reyle (1993) these expressions directly introduce discourse referents of type t in a DRS; I dubbed as indirect those that refer back to expressions that represent time intervals in an indirect way, i.e., eventuality descriptions. In Kamp and Reyle (1993) eventuality descriptions directly introduce in a DRS discourse referents of type e or s. It is through the function loc that the time interval corresponding to their duration is provided. I termed as ambivalent those that admit the two possibilities. Among the group of indirect anaphors, a three-cases distinction was still pointed out: (1) cases where the anaphor refers back to the running time of an eventuality; (2) cases where the anaphor refers back to a time interval stretching from an eventuality and the speech time; (3) cases where the anaphor refers back to a time interval stretching from the speech time to the time of an eventuality. A crucial distinction exists between cases (1) and (2)–(3). In (1) – but not in (2) and (3) – the antecedent of the anaphor is introduced in the DRS by an expression in the previous linguistic expression, on which certain constraints are imposed. In this paper, I mentioned two types of constraints, aktionsart constraints and, for some of the anaphors, constraints related to event structure or, depending on the analysis, discourse structure. I provided representations within the DRT framework and provided some partial rules of DRS construction.
References Alves, A., & Txurruka, I. (2001). The meaning of same in anaphoric temporal adverbials. In M. Bras & L. Vieu (Eds.), Semantic and Pragmatic Issues in Discourse and Dialogue: Experimenting with Current Dynamic Theories (Current Research in the Semantics/Pragmatics Interface, 9) (pp. 147–181). Oxford: Elsevier. Asher, N. (1993). Reference to Abstract Objects in Discourse. Dordrecht: Kluwer Academic Press.
TSL[v.20020404] Prn:17/10/2002; 10:36
F: PBNS04.tex / p.14 (74)
Ana Teresa Alves
Kamp, H., & Reyle, U. (1993). From Discourse to Logic. Introduction to Modeltheoretic Semantics, Formal Logic, and Discourse Representation Theory. Dordrecht: Kluwer Academic Press. Lascarides, A., & Asher, N. (1993). Temporal Interpretation, Discourse Relations and Commonsense Entailment. Linguistics and Philosophy, 16 (5), 437–493. Moens, M., & Steedman, M. (1988). Temporal Ontology and Temporal Reference. Computational Linguistics, 14 (2), 15–28. Móia, T. (2000). Identifying and Computing Temporal Locating Adverbials with a Particular Focus on Portuguese and English. Ph.D. diss., Universidade de Lisboa. Vendler, Z. (1967). Linguistics and Philosophy. Ithaca, New York: Cornell University Press.
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.1 (75)
Remarks on the semantics of eventualities with measure phrases in English and Romanian Ilinca Cr˘ainiceanu University of Craiova, Romania
. Introduction The purpose of this paper is to analyse, in a comparative study of English and Romanian, the aspectual interpretation of state and event descriptions with measure phrases when these descriptions occur in the English Simple Present Perfect (PERF) and its Romanian equivalents, the Prezent (PREZ) and the Perfect Compus (PC). I discuss the following measure phrases: the English for-measure phrases vs. their Romanian equivalents: de/timp de/pentru x amount of time measure phrases. The analysis is conducted in Discourse Representation Theory (DRT) (cf. Kamp & Reyle 1993) and develops a fully compositional analysis of aspect and aspect shift.
.
Background assumptions
Before all else we define the notions we work with: eventuality description, tense operators, aspectual modifiers and the concept of coercion.
. Eventuality descriptions Following Krifka (1992), Verkuyl (1993), Kamp and Reyle (1993), and de Swart (1998) among others, we assume that the aspectual class of a sentence as a whole is determined by the semantic nature of the verb and by the nature of its arguments.1 The aspectual class/Aktionsart of a predication identifies sets of eventualities: states, processes and events. As is well known, processes and events are non-
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.2 (76)
Ilinca Cr˘ainiceanu
stative and form the supercategory of dynamic, quantized eventualities (Taylor 1977; Dowty 1979; Bach 1986; de Swart 1998). States and processes are homogeneous eventualities, they have non-quantized reference and they form the supercategory of homogeneous eventuality descriptions.
. Tenses The Romanian PREZ and PC tenses are taken as tense operators (following de Swart’s proposal (1998) for the French tenses). Tense operators map eventuality descriptions onto the time axes and locate them in time with respect to Speech Time (ST). The PREZ tense presents a condition as holding over some period which surrounds the utterance time. The operator PAST (in our case the PC) precedes the formula, in which case the whole formula is construed as being true at now just in case the part following the PAST operator is true at some previous time.
. Aspectual modifiers Aspectual modifiers map sets of eventualities (of a certain type) onto sets of eventualities (of possibly other type) (cf. de Swart 1998). Tense operators apply after aspectual modifiers have applied. Under these assumptions, the syntactic structure of a sentence is: [Tense [Aspect [eventuality description]]]
(cf. de Swart 1998)
The aspectual selection restrictions are treated as input requirements for the tense operator to apply on. We shall follow de Swart (1998) in treating the English PERF as an eventuality description modifier.
. Coercion Coercion is an eventuality description modifier introduced into a DRS that is governed by contextual reinterpretation or triggered by linguistic material, dictated by the need to solve aspectual conflicts (cf. de Swart 1998). Coercion may be syntactically or morphologically invisible, as it is the case with Romanian tenses, or visible as it is the case with measure phrases. We speak about the following aspectual modifiers: coercion modifiers and the English for-phrases and their Romanian equivalents de/timp de/pentru x amount of time. The main coercion modifiers are: Cse that stands for the coercion of an event into a state and the coercion modifier Ces that maps a state onto an event. If measure phrases are treated as modifiers they directly modify eventuality descriptions as there is no evidence that modifiers have scope (cf. Parsons 1991). In the present study I do not discuss the processes with measure phrases but I shall comment on their aspectual behaviour whenever it proves relevant.
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.3 (77)
Semantics of eventualities with measure phrases
. The aims of the analysis The aims of the analysis are: (a) to prove that Romanian tenses are aspectually sensitive (in the sense of de Swart 1998) and function only as tense operators; (b) to point out the sharp contrast between the semantics of English state eventualities with a for-phrase and the semantics of Romanian state eventualities with de/timp de-phrases (i.e. the Romanian equivalents of the for-phrase); (c) to analyse the aspectual behaviour of certain events in the perf that do occur with for/de-phrases although Kamp and Reyle claim that, in English, for-phrases “never combine with accomplishment/achievement VPs” (1993: 631).
. Romanian tenses function as tense operators As stated above, we want to prove that Romanian tenses are aspectually sensitive tenses and function only as tense operators. The Romanian PREZ and PC are syncretic in nature: the same morphemes are carriers of both tense and aspect markers.
. Unmarked uses of the PREZ and PC Aspectually, these tenses do not change the eventuality description that maps onto them: the PREZ naturally selects for states or processes while the PC simply selects for events. Temporally, the PREZ locates a state or a process at Speech Time (ST) and the PC locates an event in the past with respect to ST. Consider examples (1) and (2) and their grammatical structures in (1 ) and (2 ), respectively: (1) Doamna Ramsey locuie¸ste la Londra Mrs. Ramsey lives (PREZ) in London (1 ) PREZ [Doamna Ramsey a-locui la Londra]2 (Mrs. Ramsey lives in London) (2) Doamna Ramsey alearg˘a Mrs. Ramsey runs (PREZ) (2 ) PREZ [Doamna Ramsey a-alerga] (Mrs. Ramsey is running) The tense operator PREZ does not trigger an aspectual shift in the case of states and processes: it only locates the state and the process at ST. The operator directly applies on the input eventuality (i.e. a state: to live in London/a process: to run) in a strictly compositional way (as in (1 ) and (2 )). The Romanian PC presents the Aktionsart of a predication as including the onset and the outset of an eventuality. This tense is signalled by the discontinuous
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.4 (78)
Ilinca Cr˘ainiceanu
morpheme a avea (to have) plus the past participle of the verb. It naturally selects for event descriptions. Following the perfect compatibility between the temporal location of the PC and the ontological status of events (non-homogeneous, involving an inherent culmination point), the only contribution of the PC is to locate them in the past with respect to ST. Consider example (3) and its grammatical structure given in (3 ): (3) Maria a scris scrisoarea Mary wrote (PC) the letter (3 ) PAST [Maria a-scrie scrisoarea] (Mary has written the letter) These are the unmarked uses of the PREZ and the PC.
.. The English PERF vs. the Romanian PC I closely follow Kamp and Reyle (1993) in characterizing the relevant interpretations of tensed clauses in terms of the two temporal features (Temporal Perspective (TP) and TENSE) and the aspectual features ±Stat and ±PERF. The Temporal Perspective determines the relation between RT and ST which has two values: +PAST (RT ← ST) and –PAST (RT = ST). TENSE is defined as the relation holding between ET and RT: it can be past (ET ← RT) or pres (RT = ET). From the very beginning, it is lucrative to make the distinction between the English PERF and the Romanian PC with the help of the notions defined above. I offer the DRS interpretations of the PERF and the PC with event predicates (in empty context). It can be seen that the major difference between the two tenses is that the PERF designates state descriptions while the PC designates event descriptions. Consider first sentence (4): (4) Mary has written the letter n
s
t
x
e
y
t=n tÍs Mary (x) the letter (y) e ÉÌ s e: x write y
Figure 1. Mary has written the letter
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.5 (79)
Semantics of eventualities with measure phrases
In current studies, the PERF is analysed as expressing anteriority with respect to ST (i.e. ST = RT; ET ← RT). In Kamp and Reyle’s treatment of the PERF, the TENSE is PRES and TP is pres. By positing s as starting at the moment e ends they retrieve the notion of anteriority of the eventuality e with respect to the state s (which has current relevance at ST). The Romanian equivalent of Mary has written the letter is given in (5) and its DRS interpretation is offered in Figure 2. (5) Maria a scris (PC) scrisoarea n
e
t
x
y
t
Figure 2. Maria a scris scrisoarea
Thus, the Romanian PC is a past tense defined as: TENSE = past (ET ← RT) while its TP is pres (–PAST: RT = ST).
. Marked uses of the PREZ and the PC .. Romanian tenses are aspectually sensitive tenses The PREZ and the PC tenses are aspecually sensitive in the sense that they are affected by the distinction homogeneous vs. non-homogeneous/quantized supercategory of eventuality descriptions. If the aspectual requirements of the PREZ/the PC are not satisfied by the eventuality description, the tense operator either does not apply (e.g. in case properties of individuals are predicated of in copular constructions: *Ion a avut ochi alba¸stri/John had blue eyes) or a contextually governed process of reinterpretation comes into play under the form of construction rules which check the input conditions of the tense operator. These construction rules involve the introduction of coercion modifiers that solve the aspectual conflict (i.e. the aspectual shift). For example, if a state or process appears in the PC, the construction rule appeals to the coercion modifiers Ces /Cep which stand for the coercion of a state/process eventuality onto an event description. Consider the following examples:
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.6 (80)
Ilinca Cr˘ainiceanu
(6) Maria a fost îndr˘agostit˘a de Ion Mary was (PC) in love with John (6 ) PAST [Ces [Maria a-fi îndr˘agostit˘a de Ion]] (Mary has been in love with John) (7) A înotat în oceanul Pacific He swam (PC) in the Pacific Ocean (7 ) PAST [Cep [el a-înota în oceanul Pacific]] (He has swum in the Pacific Ocean) The tense operator (PAST) does trigger an aspectual shift of these eventualities: basically, to be in love with John and to swim in the Pacific Ocean are homogeneous eventualities which do not involve an inherent culmination point. Since these sentences are well-formed, for the PAST tense operator to apply, we assume that the construction rule for the PC appeals to the coercion modifiers Ces /Cep which contextually reinterpret the state and the process eventuality descriptions as events. Similarly, the combination of an event description with the PREZ triggers the introduction of the coercion modifier Cse that reinterprets the event as a state. Consider sentences (8)/(9) and their grammatical structures given in (8 )/(9 ): (8) Maria face pr˘ajitura în buc˘atarie (acum) Mary bakes (PREZ) the cake in the kitchen (now) (8 ) PREZ [Cse [Maria a-face pr˘ajitura în buc˘atarie]] (Mary is baking the cake in the kitchen now) (9) Maria împlete¸ste puloverul (acum) Mary knits (PREZ) the pullover (now) (9 ) PREZ [Cse [Maria a-împleti puloverul]] (Mary is knitting the pullover now) Once more, to bake the cake/to knit the pullover are event descriptions which involve an inherent culmination point. The conflict between the eventuality description (i.e. an event) and the PREZ tense operator is solved by the introduction into DRS of the coercion modifier Cse that coerces the event into a state eventuality description. We contend that all past states and processes in the PC are reinterpreted as events (in empty context). Such a treatment of aspectual shift does not overgenerate as restrictions of application are imposed in case further linguistic material is present in the sentence: e.g. by measure phrases in our study (see below). Romanian past events (in empty context) are uniformly rendered in the imperfect (IMP) and reinterpreted as states. We contend there is no inherent aspectual complemen-
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.7 (81)
Semantics of eventualities with measure phrases
tarity between Romanian imperfective tenses (PREZ, IMP) and state eventualities. Again, extra linguistic material present in the PREZ/IMP sentences (e.g. frequency adverbs such as adesea/often, timp de dou˘a ore diminea¸ta / for two hours in the morning, de la 5 la 7 dup˘a amiaza / between 5 and 7 in the afternoon) triggers different shades of meaning of eventualities. An example in point is the habitual interpretation of the PREZ/the IMP sentences with frequency adverbs (e.g. Ion cite¸ste (PREZ) ziarul de la 7 la 8 diminea¸ta / John reads the newspaper from 7 to 8 in the morning, Ion citea (IMP) ziarul de la 7 la 8 diminea¸ta / John used to read the newspaper from 7 to 8 in the morning). Thus, aspectual interpretation is a pragmatically implicit and contextually governed process. All the examples discussed above argue for treating the Romanian PREZ and PC as aspectually sensitive tenses. They are not aspectual operators (or they would be vacuous aspectual operators) but act only as temporal operators and embody grammatical aspect. This treatment of the PREZ/the PC is also motivated on morphological grounds: it is implausible to set apart aspectual and temporal information in morphology (cf. de Swart 1998). We have shown that Romanian tenses are aspectually sensitive to the distinction homogeneous vs. quantized eventuality descriptions. This property of Romanian tenses will plainly come into force when we derive the semantic interpretation of the de-measure phrase with the PREZ and the PC vs. their English PERF equivalents with the for-phrase, in a strictly compositional way. This is the second aim of our study.
. The semantics of state descriptions with for/de-measure phrases . English descriptions in the PERF In English, the semantic effect of the PERF is that of presenting the Aktionsart of v as starting at, but not including the culmination point (cf. Kamp & Reyle 1993). I follow Kamp and Reyle (1993) in assuming that the semantic role of the PERF is that of describing a state s which results from the occurrence of a certain eventuality e of the type described by the verb phrase: (10) Mary has lived in London / Mary has written the letter The PERF of state eventualities closely resembles the perfect of non-stative eventualities as in both cases the PERF relates to the termination of s.
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.8 (82)
Ilinca Cr˘ainiceanu
. The ambiguity of the English stative verbs with a for-phrase However, when the present perfect occurs with stative verbs and a for-phrase, the verb forms acquire two quite distinct readings. Consider the interpretations of sentence (11): (11) Mary has lived in London for 3 years The readings of sentence (11) are: a.
There was somewhere in the past a 3-year period of time during which Mary lived in London
and b. Mary is living in London now and that state of affairs has already been going on for 3 years (cf. Kamp & Reyle 1993).
. On English state eventualities with the for-phrase that designate a bound amount of time The first reading of sentence (11) describes the state of Mary’s living in London as coming about through the termination of the state of her living in London with a for-phrase acting as a “qualification” of the underlying state s . We could explain the semantics of this reading of sentence (11) if we make use of Vet’s (1994) and Krifka’s (1992) observation that the for-phrase brings in a notion of boundedness. Consequently, the for-adverbial should pattern with event descriptions as they make the predicate quantized. On this reading, de Swart (1998) suggests interpreting the for-durational in terms of the function FOR x time: S ∪ P → E. That is, whenever the bound for-phrase acts on states and processes, they are interpreted as eventualities that look very much like event descriptions. For-phrases are thus seen as aspectual modifiers mapping sets of eventualities onto sets of eventualities. They are extensional in nature and so, their interpretation entails the existence of both the input and the output of the eventuality. The grammatical structure of sentence (11) is as in (11 ), and its DRS interpretation is spelled out in Figure 3. (11 ) PRES [PERF [For 3 years [Mary live in London]]] The amount of time introduced by the for-durational is 3 years and the duration of Mary’s living in London lasts at least this amount of time. The application of the bound state variable (e = end of s ) results in a bounded portion of the state which looks like an event as it is quantized. In a strictly compositional way, Mary lives in London is a state eventuality. It maps on the quantized portion of the state (i.e. for 3 years) which counts as an event since it is bounded. The condition e ⊃⊂ s introduced by the PERF means that the result state s starts right at the end of the
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.9 (83)
Semantics of eventualities with measure phrases
n
t
s
x
y
s'
e
mt
t =n sOt Mary (x) London (y) e = end (s') e ÉÌ s three years (mt) dur (s') = mt s': x live in y
Figure 3. Mary has lived in London for three years
eventuality e. The result state s constitutes the output of the PERF modifier and provides the variable for the PRES tense to operate on. “If the result state s holds now, we know that the event took place in the past of the ST and the consequent state has current relevance” (de Swart 1998: 354). As the grammatical structure of the sentence shows (in 11 ), unless we had adopted the compositional procedure and had taken the event semantics of the for-phrase as having scope over the PERF, the tense operator could not have applied on the result state s which is the output of the PERF. Thus, the aspectual interpretation of the sentence would have crashed.
. On Romanian state eventualities with the de-phrase that designate a bound amount of time The Romanian equivalent of this semantic interpretation of the English for-phrase (i.e. as a bound amount of time) is infelicitous. This happens not only with state eventualities but with processes and events as well. As the durational is quantized the amount of time it describes is bound; it follows that all eventualities it occurs with should opt for the aspectually sensitive PC. However, consider a state description with the de-phrase: (12) *A locuit la Londra de 2 ani He lived (PC) in London for 2 years (He has lived in London for 2 years) (12 ) *PAST [de 2 ani [Ces [el a-locui la Londra]]] Why should predication (12) be ill-formed? Take the grammatical structure of the sentence given in (12 ). Compositionally wise, things should work straight-
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.10 (84)
Ilinca Cr˘ainiceanu
forwardly: the state eventuality (i.e. he lives in London) is coerced into an event by the coercion modifier Ces for it to apply on the event semantics of the bound amount of time brought in by the de-phrase. At this point, the PAST tense operator is free to apply and to locate the whole eventuality in the past. However, this is not the case. The incongruity is due to the fact that the de-durational describes the onset of the eventuality thus evincing in a sense the semantics of the English since-phrase. The de α shows that the underlying state s covers the interval stretching from the denotation of α to ST. Due to its semantics the de-durational imposes an important restriction: the tense of the predication it occurs with must be imperfective (the Prezent and the Imperfect). As has already been seen, aspectually sensitive imperfective tenses (in our study the PREZ) always denote state eventualities. Thus, the de-durational must act on the RT of the predication in the PREZ (i.e. TENSE: pres (RT = ET) and TP: pres (RT = ST). This linguistic fact also proves that the adverbs compatible with PC predications act on the ET of the predications. For example: frame adverbs: ieri/yesterday, asear˘a/last night, s˘apt˘amîna trecut˘a/last week; frame anaphoric adverbs: atunci/then, numaidecît/immediately, a doua zi/the following day; bound duration adverbs: de la ora 2 la ora 3/between 2 o’clock and 3 o’clock, între ora 2 ¸si ora 3/from 2 o’clock to 3 o’clock; punctual adverbs: pe 6 august/on the 6th of August, în 1990/in 1990 occur with state, process or event descriptions in the PC and they define the ET (not the RT) of the eventuality as in: (13) Maria a fost aici ieri Mary was (PC) here yesterday (14) Maria a alergat între 2 s¸ i 3 Mary (PC) ran between 2 and 3 o’clock (15) Maria a scris scrisoarea pe 6 august Mary wrote (PC) the letter on the 6th of August Accordingly, the state eventuality (12) with the de-durational becomes well-formed if it occurs in the PREZ tense and its compositional-based interpretation is straightforward: (16) Locuie¸ste la Londra de 2 ani He lives (PREZ) in London for 2 years (He has been living in London for 2 years)3 (16 ) [PREZ [de 2 ani [el a-locui la Londra]]] Kamp and Reyle raised the same problem and came to the conclusion that “this is not so much a question about the perfect [with for-phrases] but rather about the present tense . . . [in languages such as French and German]” to which we should add Romanian (1993: 568, Note 36).
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.11 (85)
Semantics of eventualities with measure phrases
n
t
s
s'
x
y
e
mt
t =n sOt Maria (x) Londra (y) 2 ani (mt) dur (s') = mt s' = beg (s) s': x a-locui in y
Figure 4. Locuie¸ste la Londra de 2 ani
Paraphrasing Kamp and Reyle, we claim that the issue should be regarded in the reversed manner: it is not so much a question about the present tense as it is about the de-durational as far as the semantics of de-phrases selects only the PREZ tense in all types of eventualities. The grammatical structure of sentence (16) is given in (16 ) and its DRS interpretation is spelled out in Figure 4.
. On English state eventualities with the for-phrase that designate an open amount of time On the second reading of sentence (11), repeated here for convenience as (17) (17) Mary has lived in London for 3 years the PERF describes a result state s as starting at the beginning of the underlying state s of the eventuality where state s has not yet come to an end (cf. Kamp & Reyle 1993). The contribution of the for-measure phrase is that of ensuring that s lasts as long as 3 years. “As s and s are concurrent in this case, it is hard to tell whether the for-phrase should be seen as characterizing s or s ” (Kamp & Reyle 1993: 587). Although s and s are concurrent, it is more plausible to interpret the for-phrase as characterizing the duration of s , not s (which is given by the PERF). Consequently, this reading of the state eventuality is rendered by the joint contribution of the PERF and that of the for-phrase. On this reading, the grammatical structure of sentence (17) is given in (17 ): (17 ) [PRES [PERF [For 3 years [Mary live in London]]]] Compositionally (as in 17 ), Mary live in London is a state eventuality; it is the input for the for-phrase, that is s . s counts as a state because it is not bound and has not
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.12 (86)
Ilinca Cr˘ainiceanu
n
t
s
x
y
s'
e
mt
t=n sOt Mary (x) London (y) e = beg (s') e ÉÌ s three years (mt) dur (s') = mt s': x live in y e' Ø e' < n e' = end (s')
Figure 5. Mary has lived in London for three years
yet come to an end. The condition e ⊃⊂ s, introduced by the PERF, means that the result state s, concurrent with s , starts at the end of the eventuality e. It is the result state s which provides the variable for the PRES tense to operate on. The DRS interpretation of predication (17) is given in Figure 5.
. On Romanian state eventualities with the de-phrase that designate an open amount of time The Romanian equivalent of the above English stative eventuality is rendered in the PREZ with a de-phrase. (18) Maria locuie¸ste (PREZ) la Londra de 3 ani The grammatical structure of sentence (18) is given in (18 ): (18 ) [PREZ [De 3 ani [Maria a-locui la Londra]]] We have seen that state eventualities with the de-durational are not ambiguous in Romanian. The DRS interpretation of sentence (18) is that offered in Figure 4 above. Due to its “since” semantics the de-durational selects the PREZ tense (on the present axis of orientation). The de-phrase behaves as a left-hand delimiter open to the right. The aspectually sensitive PREZ tense denotes only state eventualities, and states can indeed be measured by the de-durational. The phrase de 3 ani (for 3
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.13 (87)
Semantics of eventualities with measure phrases
n
t
s
x
y
e
mt
t
Figure 6. Maria a locuit la Londra (timp de) 3 ani
years) in (18) introduces a discourse referent mt that stands for the amount of time that characterises the duration of s (i.e. 3 years), while the preposition de marks the beginning of the state s which overlaps s. We note in passing that this is not the case with the English for-phrase alone that can also occur with the Past Tense (as in He lived in London for 3 years), the Past Tense Progressive (as in It was raining for 2 hours), and even with deictic event verbs in the PERF (as in He has come/gone out for 20 minutes). In compositional terms, if the state eventuality Maria a-locui la Londra maps on the retrospective “since” semantics of the de-durational, the contribution of the PREZ tense operator is to locate the situation at ST.
. Another means of rendering a bound amount of time in Romanian Moreover, Romanian exhibits another possibility of rendering the first interpretation of the English stative eventuality, repeated here for convenience as (19): (19) Mary has lived in London for 3 years viz. there was somewhere in the past a 3-year period of time during which Mary lived in London. This alternative possibility consists in interpreting the dedurational as a bound amount of time (i.e. qualifying as an event) and occurring with PC state eventualities. However, in this latter case, the introductory preposition either disappears altogether or changes into timp de (a close equivalent to the for-phrase) plus a unit of time, preserving the semantics of a container. This second possibility is again available to all types of eventualities (states, processes and events). In this section, I analyse the aspectual interpretation of state eventualities in the PC with this type of measure phrase, postponing the aspectual interpretation of event descriptions until the last part of the analysis.
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.14 (88)
Ilinca Cr˘ainiceanu
For example, a state eventuality such as (20) has the grammatical structure (20 ), and its DRS interpretation is given in Figure 6. (20) Maria a locuit (PC) la Londra (timp de) 3 ani Mary lived (PC) in London for 3 years (20 ) [PAST [(timp de) 3 ani [Maria a-locui la Londra]]] The amount of time introduced by the durational is 3 years and Mary’s living in London lasts precisely this amount of time. As the amount of time is bound, the state has clearly a flavour of an event description located in the past by the tense operator with no explicit current relevance at ST. The timp de + a unit of time measures the duration of the state which is situated entirely in the past by the PC from a PREZ temporal perspective. The linguistic evidence which proves that these state eventualities located in the past behave as events is that, in narrative discourse, they move the story line forward. Consider the piece of discourse in (21): (21) Maria a locuit la Londra (timp de) 3 ani. Apoi s-a m˘aritat s¸ i s-a mutat la Leeds. Mary lived (PC) in London for 3 years. Then, she got married (PC) and moved (PC) to Leeds.
. Conclusions The conclusions we attain with respect to for/de/timp de-phrases with state eventualities are: (1) in English, the for-phrase with PERF eventualities is sensitive to the Aktionsart of the eventuality it occurs with (i.e. the eventuality should be stative). With state eventualities the for-phrase is ambiguous between two distinct readings. (2) in Romanian, the de-measure phrase shares some of the semantic properties of the English since phrase. It exhibits only one reading and opts for imperfective tenses (in our analysis the PREZ tense). When the duration of the state eventuality is limited the predication occurs in the PC and the de-durational changes into (timp de +) a unit of time measure phrase.
. English and Romanian events with for/de, pentru, timp de-phrases . On defining the PERF Following Kamp and Reyle (1993), we assume that the PERF VP describes a state that results from the occurrence of a certain eventuality. It is the have component of the PERF that acts on the recategorization of eventualities as result states (as we have seen above when we analysed Mary has written the letter → the letter is
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.15 (89)
Semantics of eventualities with measure phrases
written now). An event (accomplishment) description in the PERF cannot occur with the for-measure phrase. In a sentence such as *Mary has written the letter for 2 hours, have has already made its contribution in recategorizing the predication into a result state s. The contribution of the for-measure phrase would be that of measuring the duration of the underlying eventuality s ; only, there is no such state s unless the eventuality occurs in the progressive PERF for the durational to measure the amount of time of the state of writing (as in Mary has been writing the letter for 2 hours).
. English and Romanian event verbs that can occur with for/de-phrases As has already been stated, the third aim of our analysis is to prove that certain event descriptions in the PERF do occur with for-phrases, contra Kamp and Reyle who claim that in English for-phrases “never combine with accomplishment/achievement VPs” (1993: 631). Everything depends on the lexical semantics of the verb, viz. on the relative prominence (i.e. “event headedness”, Pustejovsky 1995) of the HEAD marker of the event (annotated e*). The “head” specifies what subevent of the whole event is being focused by the verb in question. In English, there is a small class of deictic verbs that are not based on participants’ actual location, but on their “normative location or home-base” (Levinson 1983) which do occur with for-phrases in the PERF. They are achievement verbs of the type come, run home, move in / leave, move out, go out which have the property of being right-headed or left-headed, respectively. They occur with for-phrases but in this case the semantics of the for-phrase changes and its reading becomes “prospective”. In Romanian, deictic event verbs of these types (i.e. a veni, a alerga acas˘a, a se muta aici / a pleca, a se muta de aici, a ie¸si) combine with the de-phrase in the PC and part of the semantics of the de-phrase is preserved.
. Romanian right-headed and left-headed event verbs with de-phrases ..
On the subeventual structure of events
I follow Pustejovsky in assuming a subeventual structure of events (which he calls TRANSITIONS (1995: 68)). Standardly, one of the subevents of a complex event will be more prominent and constitutes the HEAD of that event. “Event headedness provides a way of indicating a type of foregrounding and backgrounding of event arguments” (Pustejovsky 1995: 72). The head, i.e. the most prominent subevent in the event structure of a predicate, constitutes the “focus” in the predicate’s interpretation. The HEAD marker of an event is the relation that also temporally orders the subevents with respect to each other as in: * (e1 , e2 ), i.e. e1 is the head of e2 , where
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.16 (90)
Ilinca Cr˘ainiceanu
e1 ≤ e2 (cf. Pustejovsky 1995). Accomplishment verbs (such as: build the house, paint the picture, read a play, walk a mile) which involve a development process and a resulting state are taken as being left-headed transitions, while achievement verbs which are currently treated as all but culmination have, in this framework, a subinterval structure. They may be right-headed transitions as their head focuses on the “persistence of the final state” (Pustejovsky 1995: 73).4 Heads seem indeed to license certain types of modification as in: John ran home for an hour / Mary left town for two weeks (Pustejovsky 1995: 74). It is pretty clear that the argument for-phrase modifies the right-hand heads of the telic events rather than the entire event structures. They modify “the duration of the final states of the events” showing that “John spent an hour at home” and that “Mary was out of town for a period of two weeks” (cf. Pustejovsky 1995: 74).
.. Right-headed achievement verbs with the de-phrase Consider the following right-headed event verbs in the PC with the de-phrase (where the symbol ∼ = stands for an approximate English equivalent): (22) A venit de 2 ore He came (PC) for 2 hours ∼ = (He has come for 2 hours) (22 ) PAST [el a-veni* [De 2 ore]] (23) A ajuns acas˘a de o or˘a într-o fug˘a He ran (PC) home for an hour ∼ = (He has run home for an hour) (24) S-a mutat aici de 2 ani He moved (PC) in for 2 years ∼ = (He has moved in for 2 years) In Romanian, these events are interpreted as having entirely taken place before ST. The eventuality (rendered by the achievement verb) took place at the initial point in the past designated by the interval de x amount of time. I contend that the contribution of the de-durational is that of foregrounding the result state s of the event description in the PC. As the achievement verb is right-headed it lays itself open to having its subevent result state being measured by the durational. This time, the semantics of the de-phrase is that of measuring the result state s which begins at the end of the event and this state s holds true at ST. The interpretation of the eventuality can be paraphrased as: “he came 2 hours ago and he has been here since then”. With achievement verbs the de-phrase acts on the result state s (provided by the right-headed verb) and not on the whole underlying eventuality and thus, it
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.17 (91)
Semantics of eventualities with measure phrases
does not trigger coercion of the PC eventuality into a state eventuality (i.e. into a PREZ tense eventuality description). However, part of the semantics of the dedurational is preserved as long as it measures the result state up to and including ST. Compositionally, as in (22 ), the durational modifies only the right-hand head of the predicate and not the entire event that remains in the PC. Proof of the property of this class of right-headed event verbs with the de-durational to have current relevance at ST is their ability to conjoin with a PREZ tense sentence: (25) A venit de 2 ore s¸ i te a¸steapt˘a în sufragerie He came (PC) for 2 hours and is waiting for you in the living-room ∼ = (He has come for 2 hours and is waiting for you in the living-room) In contrast, sentence (26) below, where the current relevance at ST is cancelled, is ill-formed in Romanian: (26) *A venit de 2 ore dar neg˘asindu-te a plecat He came (PC) for 2 hours but since you were not here he left ∼ = (He has come for 2 hours but since you were not here he left) The grammatical structure of sentence (22) is given in (22 ) and its DRS interpretation in Figure 7. n
t
s
x
e
mt
t
Figure 7. A venit de 2 ore
.. Left-headed achievement verbs with the de-phrase In the same fashion, left-headed event verbs such as a pleca (leave), a se muta de aici (move out) in the PC with the de-durational evince sensitivity to where the head of the event is placed. With left-headed events, the de x amount of time phrase is in free variation with another adverb viz. acum x amount of time (= x amount of time ago, literally ‘now x amount of time’). Both adverbs have the same entailment: he left two hours ago and he has not been here since then.
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.18 (92)
Ilinca Cr˘ainiceanu
n
e
s
t
x
mt
t
Figure 8. A plecat de 2 ore/acum 2 ore
Consider sentence (27) below where the event is again interpreted as taking place at the initial point in the past designated by the interval covered by the two adverbs: (27) A plecat de 2 ore/acum 2 ore He left (PC) for 2 hours/2 hours ago (He left 2 hours ago) (27 ) PAST [el *a pleca [de 2 ore/acum 2 ore]] (28) S-a mutat de aici de 2 ani/acum 2 ani He moved (PC) out for 2 years/2 years ago (He moved out 2 years ago) In this case, we argue that the subevent foregrounded by the measure phrase argument of the achievement verb in the PC is situated to the left. The de-durational measures the result state between the end of the event of leaving and his not being here up to and including ST. Left-headed event verbs with the de-durational in the PC have current relevance at ST; however, in the “normative/base locations” interpretation, they can not be conjoined with a PREZ tense sentence in the affirmative due to the semantics of the verb. Compositionally, as in (27 ), the ET of the left-headed event verb will be designated by the denotation of α (i.e. 2 hours ago) and the output event is located in the past by the tense operator. The DRS interpretation of sentence (27) is given in Figure 8. Of course, a right-headed event verb (e.g. veni/come) can occur with acum 2 ore/2 hours ago and its interpretation is the same as that for a left-headed event verb (e.g. pleca/leave) in collocation with this adverb; the implication is that he came two hours ago but he is no longer here.
.. Romanian accomplishments with the de-phrase Romanian is a very permissive language in that it also allows the combination of all accomplishment verbs I am aware of, with the de-phrase on its latter reading (i.e.
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.19 (93)
Semantics of eventualities with measure phrases
when it partakes of the semantics of acum x amount of time phrase = x amount of time ago). This happens with all accomplishment verbs, either right-headed: a scrie o scrisoare / write a letter, a picta un tablou / paint a picture, a construi o cas˘a / build a house, a rezolva o problem˘a / solve a problem, a alerga o mil˘a / run a mile, a-¸si lua prînzul / eat lunch, a se muta din/în cas˘a / move out/in, a vinde o cas˘a / sell a house or two-headed accomplishment verbs: a da/give, a lua/take, a se c˘as˘atori/marry (cf. Pustejovsky 1995). Consider the following examples: (29) A construit casa de 2 ani He built (PC) the house for 2 years (= 2 years ago) (He built the house 2 years ago) (30) A luat prânzul de 2 ore He had (PC) lunch for 2 hours (= 2 hours ago) (He had lunch 2 hours ago) (31) I-am dat cartea de 2 luni I lent (PC) the book to him for 2 months (= 2 months ago) (I lent the book to him 2 months ago) (32) S-a mutat de aici de 2 ani He moved out (PC) for 2 years (= 2 years ago) (He moved out 2 years ago) The DRS interpretation of these accomplishment predications is the same as that offered for A venit/plecat acum 2 ore / He came/left 2 hours ago (see Fig. 8). The dephrase is used to characterise the result state described by the PC event predication. For instance, sentence (29) means that he finished building the house two years ago and the house has been in the state of being erected since then (up to and including the ST). The same goes for sentence (30): he ate two hours ago and has been in the state of having eaten for two hours. This reading is impossible with the English for-phrase in PERF accomplishments (see Kamp & Reyle 1993: 587, Note 48). Romanian exhibits a very regular semantics of both right/left-headed achievements and all accomplishments with this de-durational.
. English right-headed and left-headed event verbs with the for-phrase .. The deictic event verbs in English and their “prospective” reading English proves to be much less sensitive to where the head of deictic event verbs such as arrive, run home, move in vs. leave, go out, move out is placed. When these achievement verbs occur with the for-phrase, the semantics of the measure phrase changes (when compared with its interpretation with state eventualities). In this case, the for-phrase measures the duration of the result state s (and no longer characterises the underlying eventuality s ) of the event description in the PERF. Conse-
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.20 (94)
Ilinca Cr˘ainiceanu
quently, the contribution of the PERF to construe events as result states is not lost, while the contribution of the for-phrase argument is to foreground the limited duration of the result state s. The for-phrase acquires a “prospective” reading.
.. The lack of sensitivity to heads of English deictic event verbs Consider the following eventualities in the PERF with a for-phrase: (33) He has come only for 2 hours (34) He has left us for 2 years However marginal these sentences may sound, they are acceptable to many native speakers of English. In both cases, the eventualities conflate two properties: the eventualities took place at the initial point in the past (designated by the forphrase) and they strongly imply that the intention of staying/leaving is only for two hours/for two years (and he is presumably back or will be back after the two-year period of time is over). As the for-phrase may have both this interpretation and the one I discussed above with state eventualities, I conclude that it is a polysemantic preposition. Although the semantic interpretation of the for-phrase changes, the contribution of the PERF is not lost and in both (33) and (34) we have to do with the current relevance of the events at ST. The duration of the result states is measured by the for x amount of time which covers the span of time between the end of the event e and ST irrespective of where the head of the verb is placed. Thus, the grammatical structures of sentences (33) and (34) are given in (33 ) and (34 ) respectively, and their DRS interpretations in Figure 9 below: (33 ) PRES [PERF [he arrive* [For 2 hours]]] (34 ) PRES [PERF [he *leave [For 2 hours]]] n
e
s
t
x
mt
t=n sOt He (x) e ÉÌ s 2 hours (mt) dur(s) = mt e: x arrive* / *leave
Figure 9. He has arrived/has left for 2 hours
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.21 (95)
Semantics of eventualities with measure phrases
It might be the case that in larger contexts, the relative prominence of the head marker of the event comes into play in a more salient fashion. Consider the following contexts: (35) John has arrived only for 2 hours and he’s waiting for you in the livingroom. Go straight there and talk to him! (35 ) George has left for 2 years. He’s on a trip to the States for a two-year period and you can’t talk to him now. In (35), I contend that the result state purported by the PERF is conjoined with the right-headed verb’s semantics to ensure that John is here and waits for you in the living-room at ST. In contrast, in (35 ), it is the result state conveyed by the PERF plus the semantics of the left-headed event verb which render the fact that George is not here at ST for you to talk to him.
. Romanian achievement verbs with pentru-durational In Romanian, the limited duration interpretation of the English for x amount of time with achievement verbs is rendered by pentru (+ x unit of time) (= translated as for x amount of time). When pentru x amount of time occurs with both rightheaded and left-headed achievement verbs (e.g. a veni/come, a pleca/leave) in the PC, we have to do with the same conflation as in the case of the for-phrase discussed above. The result states tell us about his/her being or not being here at ST. The duration of these states is measured by the pentru x amount of time. Consider the following examples: (36) A venit pentru 2 ore (¸si vrea s˘a vorbeasc˘a cu tine) He came (PC) for 2 hours (and wants to talk with you) (He has come for 2 hours (and wants to talk with you)) (37) A plecat pentru 2 ore la ora 8 (¸si nu po¸ti vorbi cu el decît la 10) He left (PC) for 2 hours at 8 o’clock (and you can talk with him only at 10) (He has left for 2 hours at 8 o’clock (and you can talk with him only at 10)) The semantics of these achievement predications is exactly the same as that of their English equivalents above. The grammatical structures of sentences (36) and (37) are given in (36 ) and (37 ) and their DRS interpretation is offered in Figure 10. (36 ) PAST [el a-veni* [Pentru 2 ore]] (37 ) PAST [el *a-pleca [Pentru 2 ore]]
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.22 (96)
Ilinca Cr˘ainiceanu
n
e
s
t
x
mt
t
Figure 10. A venit/a plecat pentru 2 ore
. Problems raised by Romanian accomplishment verbs with pentru/timp de-phrases A quite puzzling problem is raised by the subclasses of Romanian accomplishment verbs of creation vs. accomplishment verbs that do not take created objects when they combine with timp de x amount of time and pentru x amount of time.
.. The timp de-measure phrase with accomplishment predicates The timp de-measure phrase with accomplishment predicates measures the duration of the process development phase of accomplishments while the event as a whole is situated entirely in the past by the PC. That is what happens with accomplishment verbs that involve verbs of creation such as: a scrie un roman / write a novel, a construi o cas˘a / build a house, a s˘apa o groap˘a / dig a hole, a picta un peisaj / paint a landscape. Consider the following examples: (38) A scris la roman timp de 3 ani He wrote (PC) AT/(la) novel for 3 years ∼ = (He worked on the novel for 3 years) (38 ) PAST [el a-scrie la roman [Timp de 3 ani]] (39) A construit la cas˘a timp de 3 ani He built (PC) AT/(la) house for 3 years ∼ = (He worked on building the house for 3 years) (40) A s˘apat la groap˘a timp de 2 ore He dug (PC) AT/(la) hole for 2 hours ∼ = (He worked on digging the hole for 2 hours) Note that the complement NPs of accomplishment verbs (i.e. roman, cas˘a, groap˘a) must be –definite and must be preceded by the preposition la (at, on); this preposition indicates that the processes of writing/building/digging unfolded in the past
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.23 (97)
Semantics of eventualities with measure phrases
n
t
x
e
mt
t
Figure 11. A scris la roman timp de 3 ani
during the period of time designated by the measure phrase. The creation object must be present in the structure as a kind of “partitive” for the activity of writing/building/digging respectively, to be exercised on these objects. Linguists (Dowty 1979 among many others) assume that accomplishments have a complex structure and the predicate CAUSE is a bisentential operator: its subject clause expresses the causing activity (with a causing agent) and the object clause expresses the resulting state: (41) John painted the picture (John paints) CAUSE (BECOME (a picture exists))) As the durational measures the process stage of the accomplishment verb, the implication of the predications above is that the novel/the house/the hole could or could not have been finished in the past. Consequently, they can or cannot exist as finished objects at ST (i.e. they are intensional in nature). It follows that these accomplishment VPs with timp de x amount of time seem to fit the scheme of process eventualities (rather than that of accomplishments) situated entirely in the past by the PC (i.e. if V is a process, then V(x) is only true of an interval longer than a moment). The syntactic structure of sentence (38) is offered in (38 ) and its DRS interpretation is given in Figure 11. On the other hand, accomplishment verbs such as: a lua/a da cartea (take/bring the book), a împrumuta cartea de la cineva/a împrumuta cartea cuiva (borrow the book from somebody/lend the book to somebody) fit the scheme of accomplishments when they occur with the measure phrase timp de x amount of time. In a sentence such as: (42) Ion i-a dat cartea Mariei John has given the book to Mary
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.24 (98)
Ilinca Cr˘ainiceanu
the book is not created, it does not come into existence on that occasion; it was just transferred from John to Mary. Both Romanian and English prove to be sensitive to the distinction between ±created objects as accomplishment predications can or cannot be compatible with for/timp de x amount of time durationals. Consider: (43) Ion i-a dat *romanul/*la roman Mariei timp de 2 ore John gave (PC) *the novel/*LA (at) novel to Mary for 2 hours (*John has given the novel to Mary for 2 hours) The above predications are ill-formed on the interpretation that John’s giving the novel to Mary lasted 2 hours.
.. The pentru/for-measure phrases with accomplishment predicates Quite expectedly, pentru x amount of time / for x amount of time that designates a limited duration of a state s cannot modify an accomplishment predication where the object is +definite. Since accomplishment verbs involve a development process and an outcome as a resulting state, that outcome is already concluded and it cannot be measured by a limiting durational in Romanian and in English. Consequently, all the following predications with accomplishment event verbs in the PC plus pentru x amount of time / for x amount of time are ill-formed: (44) *A construit casa pentru 2 ani (*He has built (PC) the house for 2 years) (45) *A pictat tabloul pentru 2 luni (* He has painted the picture for 2 months) I do not discuss the standard occurrence of accomplishments with in/în measure phrases that indicate the whole bound durational of the event. In this case, accomplishment predications with a +definite object are well-formed in both languages (e.g. A pictat tabloul în dou˘a luni / He has painted the picture in two months). On the other hand, the limiting durational pentru x amount of time / for x amount of time can occur with non-creation accomplishments, both in Romanian and in English. In this case, the result state s of the eventuality can be modified by the measure phrase as already seen: (46) I-am împrumutat (PC) cartea pentru 2 luni (46 ) PAST [eu a-împrumuta cartea [Pentru 2 luni]] (I have lent the book to him for 2 months) (47) Ti-am ¸ dat (PC)-o pentru 2 luni (I have given it to you for 2 months)
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.25 (99)
Semantics of eventualities with measure phrases
These accomplishment verbs behave like achievement verbs with pentru x amount of time/for x amount of time with the durational argument modifying the last subevent of the accomplishment verb irrespective of where their head is situated or how many heads a verb can have. I follow Pustejovsky (1995) in interpreting the verb give/a da as involving a relational predicate on each subevent; it is thus a two-headed verb. The syntactic structure of sentence (46) is given in (46 ) and its DRS interpretation could be similar to that offered for achievements with pentru x amount of time (Fig. 10).
.. Conclusions In principle, with events, what a measure phrase can do is: a.
either it pins down the moment when the entire eventuality took place (and measures the amount of time that lapses between that past moment and ST) b. or it measures the duration of the development process stage of events c. or it may foreground the final subevent of the event, in which case it measures the duration of the result state s. The measure phrases we have discussed are the English for-phrases and their Romanian equivalents: de-phrases (and acum x amount of time = x amount of time ago), timp de-phrases and pentru x amount of time when they appear with PERF and PC event descriptions, respectively. In English, possibilities (a) and (b) are not available: *He has built the house for 2 years. Possibility (c) is illustrated by eventualities containing achievement verbs: He has arrived for 2 hours / He has left for 2 hours and non-creation accomplishment verbs: I have taken it/lent it/given it for 2 weeks. This slim potential of English event descriptions with PERF and for-measure phrases must be due to the semantic contribution of the PERF that recategorizes all predications it occurs with into result states. In Romanian, all possibilities are available. Possibility (a) is manifested with left-headed achievement verbs: S-a mutat de 2 ani (= acum 2 ani) (He moved in 2 years ago) and with creation accomplishment predicates: A construit casa de 2 ani (∼ = He built the house 2 years ago and it is still erected). Possibility (b) is evinced by accomplishment predications of creation: A scris la roman timp de 3 ani (∼ = He worked on the novel for 3 years). Possibility (c) is illustrated by right-headed achievement verbs where the de-durational still preserves (i.e. as with state eventualities) its interpretation of a kind of “since” phrase. It links the result state s to ST: A venit de 2 ore (∼ = He came 2 hours ago and he has been here since then). This last possibility is also present with non-creation accomplishment predications: I-am împrumutat cartea pentru 2 luni (I have lent the book to him for 2 months).
TSL[v.20020404] Prn:17/10/2002; 14:58
F: PBNS05.tex / p.26 (100)
Ilinca Cr˘ainiceanu
Romanian displays more linguistic means of lexicalizing measure phrases and, at the same time, the semantics of the PC can be modified both by its argument NPs and by adverbials. All Romanian measure phrases prove to have a more regular semantics as shown by their consistent interpretation when they combine either with state or event descriptions.
Notes . In this paper we do not treat quantificational NPs (such as “every boy”). . In DRS interpretations and grammatical structures of eventualities, the Romanian infinitive will be represented notationally as a long infinitive (i.e. a-locui) for the short infinitive forms not to be mixed up with the PERFECT SIMPLU forms. . In this paper I do not discuss the distinction between the semantics of the simple present perfect and that of the present perfect progressive. . See also C. Smith (1991: 170) who analyses the temporal interval at which an achievement predicate occurs as: ti–1 ti ti+1 a b c (a) – at ti–1 the situation S does not obtain; (b) – at ti the change of state occurs; (c) – at ti+1 the result state of S obtains.
References Bach, E. (1986). The Algebra of Events. Linguistics and Philosophy, 9. Cr˘ainiceanu, I. (1999). Aspect and Coercion with Romanian Past Tenses. Bucharest Working Papers in Linguistics, 2. Dowty, D. (1979). Word Meaning and Montague Grammar. Dordrecht: D. Reidel. Kamp, H., & Reyle, U. (1993). From Discourse to Logic. Dordrecht: Kluwer. Krifka, M. (1992). Thematic Relations as Links between Nominal Reference and Temporal Constitution. In I. A. Sag & A. Szabolcsi (Eds.), Lexical Matters. Stanford: Centre for the Study of Language and Information. Landman, F. (1992). The Progressive. Natural Language Semantics, 1. Parsons, T. (1991). Events in the Semantics of English. Cambridge, MA: MIT Press. Pustejovsky, J. (1995). The Generative Lexicon. Cambridge, MA: MIT Press. de Swart, H. (1998). Aspect Shift and Coercion. Natural Language and Linguistic Theory, 16. Smith, C. (1991). The Parameter of Aspect. Studies in Linguistics and Philosophy. Dordrecht: Kluwer. Taylor, B. (1977). Tense and Continuity. Linguistics and Philosophy, 1. Vendler, Z. (1957). Verbs and Times. Linguistics and Philosophy. Ithaca. Verkuyl, H. (1993). A Theory of Aspectuality: The Interaction between Temporal and Atemporal Structure. Cambridge. Vet, C. (1994). Future Tense and Discourse Representation. In C. Vet & C. Vetters (Eds.), Tense and Aspect in Discourse. Berlin: Mouton de Gruyter.
TSL[v.20020404] Prn:29/08/2002; 15:45
F: PBNS06.tex / p.1 (101)
The present perfect in English and in Catalan* Hortènsia Curell Universitat Autònoma de Barcelona, Spain
.
Introduction
The aim of this paper is to study contrastively the use of the present perfect in English and in Catalan, using a corpus gathered from novels, plays and filmscripts.1 The model used as the basis for the comparison is taken, eclectically, from various proposals which will not be discussed in detail here for lack of space. The paper, therefore, does not include a review of the different theories about the present perfect (PrP), for which the reader can refer, among others, to Bauer (1970), Comrie (1976), Depraetere (1998, 2000), Huddleston, (1984), Leech (1987), Matthews (1987), McCawley (1971), McCoard (1978), Michaelis (1994), Palmer (1974), Schopf (1987), Schwenter (1994), Zydatiss (1978). These authors develop, in various degrees, the following issues, which will be considered here just in so far as they explain contrastive facts: the monosemic or polysemic view of the PrP; whether the perfect is a tense or an aspect; the relationship of the PrP with the other perfect forms (past and future); the interaction of the different types of perfect with the situation type or Aktionsart; time specification with the PrP and with the preterite. The cross-linguistic studies on the PrP carried out by authors such as Benveniste (1968), Comrie (1976, 1985), Givón (1979), Kurylowicz (1964, 1975) and Zieglschmid (1930) show that the general tendency in Indo-European languages is for the PrP to enlarge its contexts of use at the expense of the preterite. Surprisingly, though, since they are also Indo-European languages, this phenomenon does not take place either in English or in Catalan. Elsness (1997: 347) proposes three stages in this development: i.
reference to a present state or result of an action (the form then being a present tense rather than a present perfect);
TSL[v.20020404] Prn:29/08/2002; 15:45
F: PBNS06.tex / p.2 (102)
Hortènsia Curell
ii. the emphasis is on the past action that brought about the present result, but specification of a past time separated from the deictic zero-point is not allowed; iii. the present perfect is simply one more form to express a past situation, with no restrictions on temporal specification. According to Elsness, French would clearly be in stage three, German in an intermediate phase between the second and the third stages, and English would undoubtedly be in stage two. In this paper it is argued that Catalan is also in the second stage. At first sight, then, Catalan and English seem to have similar present perfects, since in both languages this form is used to express past actions and is incompatible with specification of a time detached from the deictic centre.
. The Catalan PrP as a hodiernal past2 The most significant characteristic of the Catalan PrP (which sets it apart from the French, Italian, German and English PrP, as well as apart from that of most Spanish dialects) is that it is what Dahl calls a ‘hodiernal past’. It is the default (perfective3 ) form to refer to situations that have obtained within the day, irrespectively of whether or not there is temporal specification. (1) A quina hora has entrat a treballar? ‘what time have you started work?’ (2) Abans m’ha preguntat si tenia set. ‘earlier you have asked me if I was thirsty’ There is no choice in Catalan, then, when one is narrating what has gone on earlier in the day, or more precisely what has happened since the moment one has gone to bed: the PrP is compulsory. This fact overrides all other possible conditions governing the use of the perfect in Catalan. The hodiernal use of the PrP is, by far, the most frequent one:4 76% of the 741 PrP tokens found in the five plays are instances of this use. However, this is not the only possible meaning of the PrP in Catalan, an issue that will be developed further in the following sections, when comparing it to the English PrP. The English PrP, as is well known, cannot be used as a hodiernal past, and this is, then, the first and probably most significant difference between the two languages under consideration. The hodiernal use of the Catalan PrP will not be studied in more depth in this paper, because it would not shed further light on the comparison of the PrP in English and in Catalan. Hodiernal perfects have been
TSL[v.20020404] Prn:29/08/2002; 15:45
F: PBNS06.tex / p.3 (103)
The present perfect in English and in Catalan
eliminated from the corpus, so that only non-hodiernal ones are considered in the statistics. However, reference to hodiernal perfects is made when necessary for the illustration of a given point.
. The present perfect across languages All the PrPs found in the corpus have been classified into four categories: existential, continuative, existential-continuative (those negative sentences that are ambiguous between the two meanings), and resultative, and the results are shown in Table 1 below. Table 1. Types of perfect across languages Type of perfect Existential Continuative Existential-continuative Resultative Total
English 30 31 6 114 181
Catalan 17% 17% 3% 63% 100%
14 20 7 139 180
8% 11% 4% 77% 100%
. Time-span perfects A sentence with the existential perfect expresses that the situation referred to has obtained at least once in a period of time that started at some point in the past and whose upper boundary is the moment of speech. (3) Larry and I have had our ups and downs. (4) Frank Sinatra. Have you ever met him? In Catalan this use is also possible for situations that have obtained at least once in a period of time leading up to the present, as can be seen in the following examples from the corpus (although no reference to it is made either by Badia i Margarit (1962, 1994) or Fabra (1956), no doubt the best known and most prestigious Catalan grammars). (5) Vostè no ha sentit mai ganes de matar una persona? ‘you’ve ever felt like killing someone?’ (6) És la bestiesa més gran que mai he sentit! ‘it’s the silliest thing I’ve ever heard’
TSL[v.20020404] Prn:29/08/2002; 15:45
F: PBNS06.tex / p.4 (104)
Hortènsia Curell
All existential PrPs must fulfil the ‘present-possibility’ requirement, defined by McCawley (1981: 82) as follows: ‘the speaker and addressee’s shared knowledge does not rule out the continued occurrence of events of the kind in question’ (quoted by Michaelis 1994: 138). For instance, in example (4) the PrP indicates that Frank Sinatra is still alive (when the question is uttered) and that it is still possible for the addressee to meet him. Otherwise the adequate choice would have been the preterite. As for aspectual character, in the two languages the existential PrP can occur both with telic and atelic situations, although the latter are much more frequent than the former (25–30% vs. 75–80%5 ). Around 75% of the existential perfect clauses in English and in Catalan contain a frequency or time span adverbial. The conditions for the use of the existential perfect (co-occurrence with time adverbials and the aspectual character of the situation), then, are very similar in English and in Catalan, although in general this use is much more frequent in the Germanic than in the Romance languages. With the continuative perfect, the utterance means that the situation expressed started at some moment in the past and still continues at speech time. (7) You’ve known that all along, haven’t you? (8) I’ve waited my whole life for a case like this. Comrie (1976) claims that the continuative perfect is typical of English and does not seem to be found in other languages. In Catalan, this type of perfect is by no means central, the meaning of a situation that started in the past and still continues at the moment of speech being generally expressed with a simple present (as in many other languages such as French, German, Italian or Spanish): (9) Tinc aquest cotxe des del 1989. ‘I have this car since 1989’ However, as stated in Fabra (1956, Appendix: 169), the continuative perfect does exist in Catalan, provided it co-occurs with adverbials such as sempre ‘always’, tota la vida ‘all (one’s) life’, d’ençà que vaig néixer ‘ever since I was born’, and the like, or else that a similar meaning is understood from the context. (10) Però és que jo sempre he estat una mica estrany. ‘but I’ ve always been a little strange’ (11) Jo he mamat l’odi per l’enemic d’ençà que vaig néixer. ‘I’ve grown up with hate for the enemy ever since I was born’ In both languages continuative PrPs occur with either atelic or negative telic situations (where the end-point is removed by the negation6 ) and in most cases (72% for English and 100% for Catalan) there is a time span adverbial in the same clause.
TSL[v.20020404] Prn:29/08/2002; 15:45
F: PBNS06.tex / p.5 (105)
The present perfect in English and in Catalan
Apart from a difference in relative frequency, the conditions for the use of the continuative are much more restrictive in Catalan than in English. In the Romance language, it can only occur with adverbials with a meaning similar to ‘always’ or ‘all one’s life’, either in the same sentence or understood from the context. In English, on the other hand, it is the form used whenever we want to express that a situation started somewhere in the past and continues until the moment of speech. In some negative sentences, existential and continuative PrPs are synonymous (Michaelis 1994: 137–138):7 (12) I haven’t had a drink in years. (13) No ho he intentat mai. ‘I have never tried’ The continuative reading of (12) would be ‘for all the times within some presentinclusive time span, the base sentence haven’t had a drink holds’, and the existential would be ‘there are no times within a present-time inclusive span such that the base sentence have a drink’ holds. This shows the basic similarity between these two types of perfect, namely that they both locate a situation with respect to a time span which includes the present: the existential PrP locates events within the time span, whereas with the continuative the event occupies the whole span.
. The resultative perfect The third category, the resultative PrP, is different from the other two in that it is not a time-span perfect, or, in other words, that the situation denoted in the sentence is located totally in the past. A sentence with a resultative perfect means that a past situation has a present resulting state.8 (14) She’s locked herself in the bathroom. (15) I aquell parell s’han escapat. ‘and those two have run away’ In this group we would also find examples of what other authors have considered ‘hot news perfect’ (used to ‘emphasize the significance and recency of the past event’ (Schwenter 1994: 995)), (16), and perfect of recent past (referring to a recent event), (17). (16) Rock musician Frank Zappa has died. A family spokesperson reported that the entertainer passed away at his home Saturday after a long bout with colon cancer. (Schwenter 1994: 1003) (17) ‘Is Red here?’ The waiter shook his head. ‘Ain’t been in tonight.’
TSL[v.20020404] Prn:29/08/2002; 15:45
F: PBNS06.tex / p.6 (106)
Hortènsia Curell
For the comparison, we will follow Michaelis (1994), and see whether or not her proposals apply both to our English corpus and to Catalan. According to Michaelis (1994: 114), the resultative perfect cannot be used to ‘(a) provide further information about a pragmatically presupposed event, (b) anchor the event in time by means of a temporal adverb, (c) assert the occurrence of an event complex or plural event [sequence of events], (d) specify the manner in which an action was performed’. All these restrictions seem to have a functional explanation: ‘the resultative PrP is used to focus upon the presently accessible consequences of a past event, rather than upon the past event itself ’ (Michaelis 1994: 114). In Catalan, all these constraints are overruled by the fact that the PrP is compulsory for situations that have obtained earlier in the day, that is, by the hodiernal character of the PrP.9 In other words, in Catalan the PrP can and actually must be used to give extra information about a pragmatically presupposed event, to specify the time of occurrence of an earlier event, to refer to an event complex, and to specify how an action has been performed, provided the events/actions have taken place within the day. In the rest of this section, then, we will address the issue of whether these constraints are seen to apply to the PrPs of our English corpus and to those Catalan perfects that are non-hodiernal.
.. Pragmatic presupposition According to Michaelis (1994: 143), the PrP ‘cannot be used to describe (‘elaborate upon’) a pragmatically presupposed event’. In her opinion, this constraint would explain why the following sentences are not possible in English: (18) *What has that been? (after a terrible noise has been heard) (19) *Where have you bought that nice sweater? In (18) the presupposed event would be the noise that has been heard, that is, the speaker assumes that the listener has also heard the racket, and in (19) the buying of the sweater, since both speaker and hearer can see the piece of clothing. This constraint would also explain why (20) is possible but (21) is not.10 (20) I can’t come tonight. I’ve broken my ankle in a skiing accident. (21) A: My God! Look at that cast! B: *I’ve broken my ankle in a skiing accident. In (20) the speaker is both establishing the new fact that she has broken her ankle and explaining the circumstances. In (21), on the other hand, the broken ankle is a known fact both to speaker and hearer (a presupposed event), so the PrP cannot be used to provide more information about it.
TSL[v.20020404] Prn:29/08/2002; 15:45
F: PBNS06.tex / p.7 (107)
The present perfect in English and in Catalan
There are some apparent counterexamples, included by Michaelis in her paper, such as (22) and (23), which she explains by saying that these are questions that make reference to the present result rather than to the past event. (22) Why have you signed your name in red ink? (23) Where have you hidden my watch? A possible explanation for this constraint would be that the PrP ‘cannot express anaphoric temporal reference’ (Michaelis 1994: 144), but, as she herself admits, there is a certain degree of idiosyncrasy in it, because (22) above is possible, but (24) is not:11 (24) ?What have you signed your name in red ink for? In our corpus, we have found one example in which the PrP is, unexpectedly, used to refer to a presupposed event: (25) Has it been a success or hasn’t it? The question makes reference to a first attack by the Australian troops in Gallipoli that has just taken place, which is, clearly, a presupposed event. This sentence is very similar to (18), and, in fact, it sounds a little odd in the context in which it appears. The outcome of this attack is decisive in the following course of action to be taken by the Australian soldiers, or, in other words, the context being military strategy, all that needs to be known is whether or not the attack has been a success. (25), then, might be explained along the lines of (22) and (23), that is, that the question is more about the present state than about the past event. More specifically, as Li et al. (1981) remark for the Mandarin particle le, the use of the PrP in (25) may be indicating that the course of action to be followed will be determined by the answer to the question. In that sense, then, it would be similar to (26),12 although in (26) there is no elaboration of a presupposed event: (26) My wallet has been stolen. (26) could be uttered when there is something to be done about the theft of the wallet, for instance, have the credit cards cancelled. If the wallet was stolen a month ago, then the preterite would be the correct option because all the actions deriving from the theft would already have been undertaken. On the whole, however, I would agree with Michaelis that this constraint explains many cases difficult to account for in theories such as ‘current relevance’, ‘extended now’ or ‘present result’. In Catalan, this constraint does not apply at all. To start with, the PrP is compulsorily used to refer to presupposed events having taken place earlier in the day. Moreover, the perfect can also be used to talk about such events having taken place
TSL[v.20020404] Prn:29/08/2002; 15:45
F: PBNS06.tex / p.8 (108)
Hortènsia Curell
earlier than today, the only condition being that the result of the past action still holds:13 (27) Qui t’ha regalat aquest rellotge? ‘who has given you this watch?’ (28) A: No sabia que eres aquí. Quan has arribat? ‘I didn’t know you were here. When have you arrived?’ B: Vaig arribar ahir. ‘I arrived yesterday’ There is a big difference, then, between English and Catalan in this respect. Whereas in English the PrP cannot be used to elaborate upon presupposed events, in Catalan it is perfectly possible, and indeed quite frequent, provided the event is either relatively recent and/or considered by the speaker to be new information for the hearer.
.. Temporal specification It is a well known fact of the English PrP that it is incompatible with definite time adverbials (regardless of whether or not they make reference to situations earlier in the day), unless they are of a cyclical nature. Hence, (29) would be ungrammatical, whereas (30) would be possible with an existential reading: (29) *The jury has reached a verdict at midnight. (30) This jury has reached a verdict at midnight (on previous occasions). In Catalan, the situation is the same, but only for times detached from the time of speech. That is, specific temporal specification is indeed possible if the adverbials make reference to the day in course. (31) *Hem firmat l’informe ahir. ‘we have signed the report yesterday’ (32) Hem firmat l’informe en un dilluns (en alguna ocasió). ‘we have signed the report on a Monday (on some occasion)’ On the other hand, both in English and in Catalan the resultative PrP can cooccur with adverbials expressing a period of time inclusive of the present, and with adverbs such as just, already, recently, yet, lately and the like: (33) That hasn’t been typed out yet. (34) But that’s a job I’ve already turned down once today. (35) Encara no m’he alliberat de tots els meus somnis. ‘I still haven’t freed myself from all my dreams’
TSL[v.20020404] Prn:29/08/2002; 15:45
F: PBNS06.tex / p.9 (109)
The present perfect in English and in Catalan
(36) . . . totes les que m’ha fetes al llarg de la nostra vida de conveïns. ‘everything she’s done to me during our life as neighbours’ In this respect, then, the English and the Catalan resultative PrP (provided it is not hodiernal) are very similar, not to say identical.
.. Sequence of events Michaelis (1994: 149–150) claims that the resultative PrP cannot be used to narrate a sequence of events, and that the preterite is the form used with this function. According to her, a sentence such as (37) can be expressing consecutive events, but only under an existential reading. If the PrPs are interpreted as resultative, the sentence simply means that at the moment of speech there is a clean house and a fed dog. She also states that there is no feature in the semantics of the resultative PrP that accounts for this restriction, that it is simply an idiosyncrasy. (37) I have cleaned the house and fed the dog. It could be argued, though, that two resultative PrPs can be interpreted as occurring in the order in which they are presented if they form a natural sequence, such as the following example:14 (38) I have bought a new broom and cleaned the house. Schwenter (1994: 1005) disagrees with Michaelis and claims that what he calls the ‘hot news perfect’ (a perfect that does not focus on the present result of a past event but on the event itself, due to its significance and recency) can indeed be used to express events in sequence, and he provides the following example : (39) J: Does anyone know if Clinton’s here yet? B: Yes, they’ve just pulled up out front; they’ve brought him up to the plaza; now I think they’ve put him up on stage. However, as suggested above, these events form a natural sequence, and this would explain the fact that they can be interpreted as consecutive. Schwenter is perfectly aware, though, that examples such as (39) are indeed rare, and that hot news perfects tend to be followed by preterites that elaborate on the past event, as illustrated in example (16), from a piece of news, and in (40). (40) Oh, I haven’t done anything? I let you kiss and grope me. This low frequency, according to him, is due to the fact that hot news events tend not to occur one after another. There are no examples in the English corpus of resultative PrPs used to express events in sequence.
TSL[v.20020404] Prn:29/08/2002; 15:45
F: PBNS06.tex / p.10 (110)
Hortènsia Curell
As for Catalan, as has already been mentioned, resultative PrPs not only can but actually must be used to narrate sequences of events that have taken place earlier in the day. Moreover, they can also be used to narrate non-hodiernal events with a present result: (41) He sortit de casa i he donat un cop de porta. ‘I have left the house and I have slammed the door’ (42) Han agafat els que passaven pel carrer i els han dit (. . . ) ‘they’ve taken those passing by the street and they’ve told them (. . .)’ It is also true, though, that in Catalan resultative PrPs are also frequently followed by sentences in the preterite, especially but not exclusively in news, both written and oral. In such cases, the PrP is used to set the temporal frame, and the preterite identifies particular events within that frame. (43) A: D’on les ha tretes? ‘where have you taken them from?’ B: Algú me les va donar. ‘someone gave them to me’ (44) La situació a l’illa filipina de Jolo s’ha complicat de manera inesperada en les últimes hores. Quan tothom esperava que la fi del malson (...) fos qüestió de dies, un altre comando terrorista va empitjorar les coses (. . .) ‘the situation in the Philippine island of Jolo has become more complex unexpectedly in the last few hours. When everybody expected the end of the nightmare to be a matter of days, another terrorist commando made things worse (. . . )’ AVUI, 30 August 2000 In this point, then, there is a considerable difference between English and Catalan. Although in both languages the PrP is often followed by sentences in the preterite elaborating on the past event, in English it is, if not impossible, at least extremely rare to find PrPs expressing events in sequence. In Catalan, on the contrary, this use is compulsory for hodiernal events, and perfectly possible for situations obtaining before today.
.. Manner adverbs According to Michaelis (1994), ‘pure’ manner adverbs15 are incompatible with the perfect, and she gives as a possible explanation the fact that the resultative PrP ‘focuses upon the consequences of an action, rather than the manner in which an action was performed’. However, she herself also points out that there is nothing in the semantics of the PrP that precludes the possibility of talking both about the manner in which an action is carried out and about its present results.
TSL[v.20020404] Prn:29/08/2002; 15:45
F: PBNS06.tex / p.11 (111)
The present perfect in English and in Catalan
(45) The president has (??angrily) called a halt to the press conference. (46) Judge Wapner has (??loudly) overruled the defense’s objection.16 The following counterexamples have been found in the corpus: (47) After a thorough investigation by the Firearms Discharge Review Board, this department has satisfactorily concluded that the use of deadly force in the shooting of James Henry Bowden was justifiable. (48) Its funny how our attitudes have suddenly changed. Michaelis (1994: 150–152) also presents sentences with a resultative PrP and a manner adverb in her paper, which she explains by making reference to the different types of manner adverbs existing in English, and claiming that this constraint only works with certain manner adverbs, such as loudly or angrily. In Catalan, this constraint does not apply at all: manner adverbials are perfectly compatible with the resultative PrP, be it hodiernal or not. (49) De quina manera he ficat la pota fins al capdamunt! ‘in what way I have put in my foot up to the top!’ What a way to put my foot in it right up to the waist! (50) El doctor Spinne és tan amable i m’ha atès tan bé (. . . ) ‘Dr Spinne is so kind and he has looked after me so well’ In order to verify the applicability of this constraint, one needs to make a very finegrained distinction between the different types of manner adverbials which, in our opinion, makes it far more difficult to establish as a general constraint than the other three.
.. Summary Of the four constraints proposed by Michaelis (1994), that of manner adverbials does not seem to be very conclusive. As for the other three, Catalan and English have been seen to be similar in temporal specification: in neither language is it possible for the resultative PrP to co-occur with definite time adverbials denoting a time separated from the moment of speech. In relation to pragmatic presupposition and sequence of events, they are completely different: in Catalan the PrP can be used to elaborate upon presupposed events and to narrate events, whereas in English it is totally impossible. This fact probably explains the higher frequency of resultative PrPs in Catalan than in English in the corpus, which is statistically significant. Both in English and in Catalan, we find telic and atelic situations with resultative PrPs, the former being more frequent. About one third of the sentences in the
TSL[v.20020404] Prn:29/08/2002; 15:45
F: PBNS06.tex / p.12 (112)
Hortènsia Curell
two languages contain an adverbial expressing recentness such as lately, recently, in the past two weeks, etc.
. Conclusion The first and clearest difference between the English and the Catalan PrP is that the latter has a hodiernal use, and is thus compulsorily used for events having taken place within the day, and the former does not. Establishing the various uses of the perfect for the two languages has proven to be useful, because in this way the similarities and differences are much easier to determine. As for the existential PrP, the two languages are very similar (in both atelic situations are more frequent than telic ones and there is a high frequency of time span adverbials), but this usage is more common in English than in Catalan. The continuative perfect is very typical of English, and is used in Catalan only in very restricted environments. In both languages, this type of perfect can occur with atelic and negative telic situations, and tends to occur with a time-span adverbial. In relation to aspectual character and temporal specification, the English and the Catalan resultative PrPs are very similar. They are also similar in their impossibility to co-occur with definite time adverbials denoting a time separated from the moment of utterance. They differ, though, in that in Catalan this type of perfect can elaborate upon presupposed events and narrate sequences of events, whereas in English it cannot. Both English and Catalan can be said to belong to Elsness’s (1997) second group of Indo-European languages,17 since in neither language is the PrP compatible with a time specification detached from the moment of speech. There is a crucial difference between the two languages, however, in what counts as a past time separated from the deictic zero-point. In English, any time specification of a past time not inclusive of the present is incompatible with the PrP, even time adverbials so close to the present moment as ten seconds ago. In Catalan, on the other hand, and as has already been mentioned, the PrP can occur with any time specification of the past provided it is included within the day.
Notes * The research reported here has been partially supported by the grant PB96-1158 from the Ministerio de Educación y Ciencia.
TSL[v.20020404] Prn:29/08/2002; 15:45
F: PBNS06.tex / p.13 (113)
The present perfect in English and in Catalan . See the appendix for a list of the works used. . Throughout the paper, unless otherwise specified, the illustrating PrP examples are all taken from the corpus gathered. . It is the tense used whenever a perfective form is required. Otherwise, the past imperfective is selected. . The percentages have all been rounded to the closest full figure. . Throughout the text, both the Catalan and the English statistics are from the corpus. . As stated by McCoard (1978:142). . Michaelis (1994) obviously only states this for the English PrP, but it is also true in Catalan. . For the different types of result and their interaction with the Aktionsart of the situation, see Depraetere (1998). . See Section 2. . These examples are from Michaelis (1994:143). . This is Michaelis’s opinion. However, other native speakers find this sentence perfectly grammatical. . Invented example. . Examples (27) and (28) are invented to make the point clearer. The examples found in the corpus would require too much contextualization. . Example provided by Alan Reeves. . ‘Pure’ manner adverbs are predicate adverbs specifying the way in which the action is performed. . These two examples are from Michaelis (1994:152). . See Section 1.
Appendix: List of works used Catalan theatre plays Altés i Campà, J. (1998). Montserrat, Marta i Maria. Editorial Millà, Barcelona. Batlle, C. (1999). Les veus de Iambu. Edicions 62, Barcelona. Baulenas, L. A. (1992). No hi ha illes meravelloses. Institut del Teatre, Barcelona. Bergoñó, J. (1974). Por. Edicions 62, Barcelona. Coquard, L. (1995). Ens ha caigut la sogra. Editorial Millà, Barcelona.
TSL[v.20020404] Prn:29/08/2002; 15:45
F: PBNS06.tex / p.14 (114)
Hortènsia Curell
English novels and films Novels Fearing, K. (1969 [1946]). The Big Clock. London: Hamish Hamilton Ltd. Hammet, D. L. (1981). The Big Knockover and Other Stories. Harmondsworth: Penguin Books. McCoy, H. (1984). Kiss Tomorrow Goodbye. Westlake, D. E. (1980 [1961]). Killing Time. New York: Charter Books. Films Deadly Force. (1986). USA. Prod.: Telecom. Dir.: Michael Miller. Gallipoli. (1981). Australia. Prod.: Robert Associated R & R Films. Dir.: Peter Weir. Riot. (1969). USA. Plaza Suite. (1971). USA. Prod.: Paramount. Dir.: Arthur Hiller.
References Badia i Margarit, A. M. (1962). Gramática catalana. Madrid: Editorial Gredos. Badia i Margarit, A. M. (1994). Gramàtica de la llengua catalana. Barcelona: Enciclopèdia Catalana. Bauer, G. (1970). The English ‘Perfect’ reconsidered. Journal of Linguistics, 6, 189–198. Benveniste, É. (1968). Mutations in linguistic categories. In W. P. Lehmann & Y. Malkiel (Eds.), Directions for Historical Linguistics (pp. 83–94). Austin: University of Texas Press. Comrie, B. (1976). Aspect. Cambridge: Cambridge University Press. Comrie, B. (1985). Tense. Cambridge: Cambridge University Press. Depraetere, I. (1998). On the resultative character of present perfect sentences. Journal of Pragmatics, 29, 597–613. Depraetere, I. (2000). Review article of Elsness (1997) and Michaelis (1998). Lingua, 110, 449–465. Elsness, J. (1997). The Perfect and the Preterite in Contemporary and Earlier English. Berlin/New York: Mouton de Gruyter. Fabra, P. (1956). Gramàtica catalana. Barcelona: Editorial Teide. Givón, T. (1979). Discourse and Syntax. New York: Academic Press. Huddleston, R. (1984). Introduction to the Grammar of English. Cambridge: Cambridge University Press. Kurylowicz, J. (1964). The Inflectional Categories of Indo-European. Heidelberg: Carl Winter. Kurylowicz, J. (1975). Esquisses Linguistiques. Munich: W. Finck. Leech, G. (1987). Meaning and the English Verb (2nd edition). London: Longman. Li, C. J., Thompson, S. A., & Thompson, R. M. (1982). The discourse motivation for the perfect aspect: the Mandarin particle Le. In P. J. Hopper (Ed.), Tense-Aspect. Between Semantics and Pragmatics (pp. 19–44). Amsterdam/Philadelphia: John Benjamins.
TSL[v.20020404] Prn:29/08/2002; 15:45
F: PBNS06.tex / p.15 (115)
The present perfect in English and in Catalan
Matthews, R. (1987). Present perfect tenses: Toward an integrated functional approach. In Essays on Tensing in English. Vol. I: Reference Time, Tense and Adverbs (pp. 111–176). Tübingen: Max Niemeyer Verlag. McCawley, J. D. (1971). Tense and time reference in English. In C. J. Fillmore & D. T. Langendoen (Eds.), Studies in Linguistic Semantics (pp. 97–113). New York: Holt Rinehart. McCawley, J. D. (1981). Notes on the English perfect. Australian Journal of Linguistics, 1, 81–90. McCoard, R. W. (1978). The English Perfect: Tense-Choice and Pragmatic Inferences. Amsterdam: North-Holland. Michaelis, L. A. (1994). The ambiguity of the English present perfect. Journal of Linguistics, 30, 111–157. Palmer, F. R. (1974). The English Verb. London: Longman. Schopf, A. (1987). The past tense in English. In Essays on Tensing in English. Vol. I: Reference Time, Tense and Adverbs (pp. 177–220). Tübingen: Max Niemeyer Verlag. Schwenter, S. A. (1994). ‘Hot news’ and the grammaticalization of perfects. Linguistics, 32, 995–1028. Zieglschmid, A. J. F. (1930). Concerning the disappearance of the simple past in various Indo-European languages. Philological Quarterly, 9, 153–157. Zydatiss, W. (1978). ‘Continuative’ and ‘resultative’ perfect in English? Lingua, 44, 339–362.
TSL[v.20020404] Prn:29/08/2002; 15:46
F: PBNS07.tex / p.1 (117)
A contrastive reading of temporal-aspectual morphemes in Swahili The case of ‘-li’ and ‘-me’ Frederick Kang’ethe Iraki University of Nairobi, Kenya
The distinction between past, present and future is only an illusion, even if a stubborn one. Albert Einstein
.
Introduction
The current paper summarises my observations on the pragmatics of two temporal morphemes in Swahili namely -li and -me. Swahili is a language largely spoken in Eastern, Central and Southern Africa. The study compares these morphemes with the Passé Simple (PS) and Passé Composé (PC) of the French language under the glare of the Relevance Theory (Sperber & Wilson 1989). Classical descriptions of these morphemes endow them with an invariable semantic content. In other words, they are assigned a stable meaning that is insensitive to pragmatic factors. For instance, li and the PS are thought to describe events that occurred in a distant past whereas me and PC describe recent events. I argue that this distinction is untenable in so far as cases abound that point to the contrary; me like PC can as well describe an event in the distant past as one in the immediate past. In addition, me seems to be the dominant morpheme describing states in Swahili, irrespective of the distance from the moment of speech. The morpheme li is excluded from describing states. It follows that the choice of me or li hinges on more than their semantics but rather on the effect that the speaker intends to produce in the hearer, that is how he intends to make the hearer view the eventuality (action, event, state) in question. The morphemes therefore appear to impart instructions or algorithms in a computational fashion as to how the eventu-
TSL[v.20020404] Prn:29/08/2002; 15:46
F: PBNS07.tex / p.2 (118)
Frederick Kang’ethe Iraki
ality should be conceived or imagined. And this can be argued to be the case for all temporal morphemes in language. In this sense, they are far from enjoying a stable semantic load. These morphemes, it is further argued, unlike nouns, adverbs and adjectives that carry conceptual information, purvey information of a procedural nature (Moeschler 1998). The paper offers evidence to the fact that the temporal referencing of an eventuality is dependent on lexical, semantic and pragmatic factors. In isolation, any of these parameters is inadequate, or simply misleading.
. Theoretical framework Swahili grammars, like French grammars, accord an invariable meaning to verb tenses. In this regard, me is associated with states, and events that happened in the immediate past, whereas li cannot represent states but points to an event that happened a long time ago.1 Contini (1989), quoting Comrie (1985), considers me and li respectively as relative and absolute representations of events. This thinking is reminiscent of the “24 hour rule” (Estienne 1954) used to distinguish the PC from the PS; the PC is limited to events happening within 24 hours, after which only the PS should be employed. Our theoretical standpoint differs considerably from this thinking in a number of ways. In the Relevance Theory developed by Sperber and Wilson (1989), and in the works of Moeschler (1998, 2000), there are three types of information: Conceptual, Procedural, and Contextual. Conceptual information is encoded by verbs, nouns, adjectives and adverbs, that is, they serve to designate objects and eventualities2 of the world. Procedural information points to the manner in which we should envisage the eventuality, and is encoded by prepositions, temporal morphemes, conjunctions, articles,3 etc. These categories give instructions, in the computational sense, as to how we should interpret things. Finally, the contextual information is what is accessible from memory (short and long term) and the sensory organs. This information helps situate eventualities in the arrow of time. The argument in the theory is that the interpretation of temporal morphemes is context-dependent and therefore their semantic charge changes depending on contextual information.
. The discussion . The distance hypothesis The argument of distance is quite interesting but there are counterexamples. In (1) li can be justified by the presence of asubuhi (morning), a temporal adverb that indicates some distance from the moment of speech. For convenience, we shall refer to me, li and later ka as Tense-Aspect Markers (TAM).
TSL[v.20020404] Prn:29/08/2002; 15:46
F: PBNS07.tex / p.3 (119)
A contrastive reading of temporal-aspectual morphemes in Swahili
(1) Rais/a/li/fika/asubuhi ya leo/ President/he/TAM/arrive/morning of today. ‘The President arrived this morning’ However, a simple commutation with me, like in (2), is equally acceptable and common: (2) Rais/a/me/fika/asubuhi ya leo Moreover, classical narratives4 in Swahili also offer many examples, like in (3), (4), and (5) where the morpheme me combines with temporal adverbials such as jana (yesterday), siku hiyo (that day) and alikuwako (there was once . . . ). (3) Bwana u me ni faa sana jana Ni me towelea harufu ya mbuzi wako hata nikashiba. Sir, you TAM me benefit a lot yesterday, me TAM use soup smell of goat yours until me TAM satisfy. ‘Sir, you really helped me yesterday, I used the smell of your roast goat to eat my food until I was full.’ (4) Na siku hiyo, bahari I me chafuka vikubwa, na upepo u na vuma mtindo mmoja. And day that, sea it TAM get dirty bigly, and wind it is blowing style one ‘That day, the sea was turbulent, and the wind was blowing hard.’ (5) Alikuwako mtu mmoja a me pata bahati ya kusikia lugha zote za wanyama There was man one he TAM get luck of to hear language all of animals ‘There was a man who had a gift of understanding animal language.’ In (3), the presence of the complement jana (yesterday) ought, according to traditional descriptions, to eliminate the usage of me in the predicate. But this is not the case here. Likewise in French, the Passé Composé (PC) can be readily employed in lieu of the Passé Simple: (3 ) Monsieur, Hier, tu m’as été fort utile, je me suis servi des arômes de ta chèvre comme soupe pour manger mon repas jusqu’à la satiété. No doubt, in this instance, me functions like the PC in French or the Simple Past Tense in English (see the translations). In (4), however, me operates more like the Imparfait (Imperfect) tense in French or Simple Past Tense in English: (4 ) ‘Ce jour-là, la mer était perturbé . . . ’ ‘That day, the sea was turbulent . . . ’
TSL[v.20020404] Prn:29/08/2002; 15:46
F: PBNS07.tex / p.4 (120)
Frederick Kang’ethe Iraki
The same description applies to (5) where me corresponds to the Imperfect tense in French or the Simple Past Tense in English. These examples attest to the fact that no matter how far in time the eventuality may have occurred, the morpheme me can be safely employed. This strikes as counterintuitive to some Swahili purists, especially with regard to adverbials like jana (yesterday), siku hiyo (That day), etc. Most speakers would rather elect for li to accompany such adverbial complements. Another example, borrowed from Luscher (1998), in French, shows that the PC has little to do with distance in time. In (6) the Passé Simple (PS) can be replaced by a PC, like in (7), thereby diminishing the purchase of the distance in time criterion: (6) Hugo écrivit Les Misérables (7) Hugo a écrit Les Misérables. The French examples incite us to re-evaluate the distinction between me and li. Translating these two utterances, me and li are acceptable: (6 ) Hugo aliandika Les Misérables. ‘Hugo wrote . . . ’ (7 ) Hugo a/me/andika /Les Misérables. Hugo/he/TAM/write/Les Misérables ‘Hugo has written “Les Misérables” ’ The argument is not that this criterion is false but that it is inadequate and further analysis is required. In (8), for instance, li cannot be combined with temporal adverbs denoting immediateness. The appropriate morpheme is me like in (9). (8) Sasa hivi a li* sema a na ku penda. Now like this he TAM say he TAM you like ‘Just now, he said he loves you.’ (9) Sasa/hivi/ a/me/sema . . . . However, with certain verbs like kuwa (to be), it appears to be possible to use either li or me, like in (10): (10) Sasa hivi, a li/me kuwa hapa. ‘Just now, s/he TAM was here.’ Utterance (10) underscores the need to examine what role the different verb predicates may play in the selection of me or li. For instance, in some other work (Kang’ethe 1999) we have shown that psychological verbs such as think, feel, know, understand favour li in describing states in the past.
TSL[v.20020404] Prn:29/08/2002; 15:46
F: PBNS07.tex / p.5 (121)
A contrastive reading of temporal-aspectual morphemes in Swahili
Conclusion: The criterion of distance in time, without being inaccurate in toto, cannot account for all manifestations of me in utterances. In addition, the morpheme me appears to be astride the two categories of eventualities (Events and States) as described by Vendler (1967). In effect, it means that me can as well describe an event or a state. Does this mean that me is a case of polysemy? It is this question that we shall attempt to answer in the next paragraph. . Me and li in Relevance theory In Relevance theory, me and li are temporal-aspectual morphemes that do not have any fixed semantic content; they are dependent upon contextual factors. In addition, they are of a procedural nature giving instructions on how to interpret the eventuality represented by the predicate. This precludes polysemy for the simple reason that in any given utterance5 the temporal morpheme receives one and only one interpretation. The reasoning is fundamentally non monotonic, therefore misinterpretations are possible; in which case the inferencing will be updated thanks to new contextual hypotheses. What then are the instructions of me and li?
. Interpreting me and li In the case of me, we argue that the speaker intends to make it manifest to the hearer that the eventuality (event or state) in question has some relevance to the moment of speech. In other words, he invites the hearer to react to the eventuality, the reason being that the eventuality in question is viewed as being of immediate consequence to both participants. In other words, me has a prodding effect on the addressee of the utterance. As for li, the eventuality is presented as having no bearing or consequence to the moment of speech, and therefore the hearer is not expected to react to it, but to record it. No prodding is necessary since the detachment vis-à-vis the eventuality is considerable for both participants. It is, therefore, the intention of the speaker that selects for the morpheme to create an effect on the hearer. Interpreting the utterance is tantamount to retrieving the informative intention of the speaker, that is, the way the hearer is meant to view the eventuality. The quest for relevance is what activates the inference procedure on the part of the hearer. In other words, upon hearing me, the listener’s mind is drawn to the relevance of the eventuality to the moment of speech. This implies that the so-called distance in time is, in effect, psychological distance, with me bringing eventualities closer to the hearer in order to engage his attention and prompt a reaction. This is what we propose to call the Prodding Effect of me.
TSL[v.20020404] Prn:29/08/2002; 15:46
F: PBNS07.tex / p.6 (122)
Frederick Kang’ethe Iraki
. Other effects of me Wilson (1997) and Contini (1989) argue that me represents states but not events. Besides, Wilson notes that me is to be employed like the Past Perfect Tense in English. These arguments are interesting but inadequate as we have shown above. To illustrate this further, we shall draw a parallel between me and the French Past tense (PC) Luscher (1998). Consider (11) where the interpretation of PC or me requires contextual information to enable us retrieve the intentions of the speaker. That is, the context guides the inference process. If I utter (11), my hearer will deduce whether I mean the dog has been out, the dog has gone out, or the dog is out depending on the information from the context (co-text, physical situation, environment, world knowledge, etc.). (11) Le chien est sorti. Mbwa/a/me/toka Dog/it/TAM/go out ‘The dog went out/has gone out/is out/has been out.’ This bolsters our argument that the temporal morpheme encodes instructions on how to view the eventuality in question. Contextual or pragmatic factors intervene to assist us in making the right deductions. This inferencing procedure is nondemonstrative, in the sense that wrong premises lead inexorably to wrong deductions, but true premises deliver true conclusions. In this sense, the reasoning is not truth-preserving.
. Special effects of me In (12) we have a rather surprising use of me that many purists of Swahili hesitate to accept: (12) Kesho, hii kazi ni me i maliza/ Tomorrow this work I TAM it finish ‘Tomorrow, this work is finished.’ Banfield (1995), in her remarkable exposé on indirect speech and its effects, introduces the notion of a subject of conscience who seems to be observing an eventuality from a certain perspective. We would like to suggest that the Swahili me, in the above utterance, introduces a subject of conscience who sees the work as already finished. Analogous observations have been made on the French Imparfait and Passé Simple (Sthioul 1998), Passé Composé (Luscher 1998), and Swahili (Kang’ethe 2000).
TSL[v.20020404] Prn:29/08/2002; 15:46
F: PBNS07.tex / p.7 (123)
A contrastive reading of temporal-aspectual morphemes in Swahili
. Me and discourse relations We intend to finish this section by proposing two hypotheses regarding Swahili discourse: one on me, and the second one on ka. Both hypotheses are bathed by works on time directionality proposed by Moeschler (1998, 2000) and on discourse relations proposed by Lascarides and Asher (1993). Hypothesis 1: Given two utterances α and β, if me is present in both, then a causal relation is to be deduced between them, β has caused α. (13) α-Maksi a me anguka/. β-Yohana a me m sukuma/. Max he TAM fall. John he TAM him push ‘Max fell. John pushed him.’ The second event explains why the first one occurred, and therefore it must have preceded it. This is what Asher (1993) in SDRT (Segmented Discourse Representation Theory) refers to as an Explanation. An alternative way is to view the discourse as a Narration whereby the events happened as they are mentioned, that is, Max falls and John pushes him. This latter interpretation is not as natural as the first, although there are contexts that can validate it. In Swahili discourse, however, the ambiguity is non-existent since in the case of a narration, the second verb must contain the morpheme ka, and not me. Although there are many uses for the morpheme ka in Swahili discourse, we have retained one, and probably the most important, of them. And this will form our second hypothesis: Hypothesis 2: Given two events α and β, if ka is present in the second event, then a narration is to be deduced, β happened after α. In (14), the presence of ka in the second event compels the discourse to move forward in time: (14) α-Yohana a li/me m sukuma Maksi. β-Maski a ka anguka. John he TAM him push Max. Max he TAM fall ‘John pushed Max. Max fell.’ Jean poussa Max. Max tomba. In (13) and (14), connectives like because (kwa sababu) and then (halafu) are simply redundant. This does not mean that they give conflicting instructions as to the directionality of time. They simply restate the instructions of the second morpheme (me or ka). Other manifestations of me in compound tenses, like in (15) seem to suggest that this morpheme actually invites the hearer to see the eventuality, no matter the distance in time:
TSL[v.20020404] Prn:29/08/2002; 15:46
F: PBNS07.tex / p.8 (124)
Frederick Kang’ethe Iraki
(15) Ni li mw ona Tina a me beba mtoto wangu. I TAM her saw Tina she TAM carry child mine. ‘I saw Tina carrying my child.’ This aspectual characteristic of me reinforces our argument that this tense creates effects. It is not just a way of expressing states, but a mode of creating a viewing effect, not dissimilar to the cinematographic effect created by the Present Tense (Kang’ethe 2000). We argue that the Present Tense, in particular na in Swahili, far from describing events happening at the moment of speech, merely invites the hearer to watch the events unfold before his eyes as if it were a film, hence the label cinematographic. It is immaterial whether the eventuality is temporally before, contemporaneous or after the moment of speech.
. Conclusions Temporal morphemes me and li should be viewed as procedural in nature, rather than conceptual. Their meaning is context-dependent and cannot be fixed once and for all. Moreover, the meaning is also a function of the intention of the speaker. In discourse, it is clear that the value of the morpheme is determined largely by the co-text or linguistic environment. That is, a succession of me or li in two utterances creates a causal link between them on one hand, and when me or li are followed by ka then a narrative is to be inferred between them. In addition, me creates psychological effects in the mind of the hearer rendering the events more lively and relevant to the speech moment. Put in another way, me prods the hearer to react to the events described. On the other hand, li represents the event as having little or no bearing on the speech moment and the hearer is invited just to shelve it in the archives. No doubt, the parallelism between me and li and Passé Composé and Passé Simple is instructive despite some notable differences. The same could be said mutatis mutandis of the English Present Perfect and the Simple Past Tense which have only been touched on incidentally in this discussion.
Notes . Contini (1989), Wilson (1997). . We employ this term here to mean states and events. . Swahili has no articles, definite or indefinite. . Here we borrow from Hekaya za Abunuwas (Tales of Abunuwas). . We regard an utterance to be what is uttered by a subject at a particular time and place.
TSL[v.20020404] Prn:29/08/2002; 15:46
F: PBNS07.tex / p.9 (125)
A contrastive reading of temporal-aspectual morphemes in Swahili
References Asher, N. (1993). Reference to abstract objects in discourse. Dordrecht: Kluwer. Ashton, E. O. (1989). Swahili Grammar. Nairobi: Metro Forms and Systems. Banfield, A. (1995). Phrases sans paroles: Théorie du récit et du style indirect libre. Paris: Seuil. Comrie B. (1985). Tense. Cambridge: Cambridge University Press. Contini, E. (1989). Discourse Pragmatics and Semantic Categorization – The case of Negation and Tense-Aspect with special reference to Swahili. Berlin: Mouton de Gruyter. Estienne, H. (1954). Grammar and style. Paris. Kang’ethe, F. (1999). La pragmatique des temps verbaux du Swahili: une étude comparée avec le français. M.Phil. dissertation, University of Geneva. Kang’ethe, F. (2000). Une lecture pragmatique des morphèmes temporels du Swahili: le cas de na. Cahiers de Linguistique Française, 22, 295–306. Luscher, J.-M. (1998). Procedure d’interprétation du Passé Composé. In J. Moeschler et al. (pp. 181–196). Macmillan (1999). Hekaya za Abunuwas na Hadithi Nyingine. Nairobi: Sunlitho Ltd. Moeschler, J., et al. (Eds.). (1998). Le Temps des événements. Paris: Kimé. Reader, J. (1998). Africa: A biography of the continent. New York: Alfred A. Knopf. Sperber, D., & Wilson, D. (1989). La pertinence, communication et cognition. Paris: Minuit. Sthioul, B. (1998). Temps verbaux et point de vue. In J. Moeschler et al. (pp. 197–220). Paris: Kimé. Vendler, Z. (1967). Linguistics in Philosophy. Ithaca: Cornell University Press. Wilson, P. M. (1997). Simplified Swahili. Longman.
TSL[v.20020404] Prn:23/08/2002; 14:35
P III
Modality
F: PBNSP3.tex / p.1 (127)
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.1 (129)
Semantic and pragmatic constraints on mood selection* Rui Marques Universidade de Lisboa, Portugal
.
Introduction
In this paper, I will address the issue of mood selection in two groups of languages. As has been noticed in the literature, there exist some differences in the distribution of indicative and subjunctive in languages as Rumanian, Hungarian and (Modern) Greek, on one hand, and Western Romance languages – French, Portuguese, Spanish, Catalan and Italian – on the other. I will first consider mood distribution in complement clauses. Subsequently, I will address the main semantic and pragmatic approaches to mood selection, and then present my own proposal. In the final part of the paper, the analysis will be extended to adverbial clauses.
. Mood selection in complement clauses As regards complement clauses, in all languages under consideration the predicates listed in (1) are subjunctive governors, while those listed in (2) select indicative: (1) a. b. c. d. e. f. g. h.
desideratives – e.g. want, hope . . . directives – e.g. order, ask, suggest . . . predicates of fear – e.g. be scared epistemic predicates expressing negative commitment – e.g. doubt negative declaratives – e.g. deny authorise, allow, forbid . . . be enough, cause, avoid, imply, try . . . be convenient, be urgent . . .
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.2 (130)
Rui Marques
(2) a. declaratives – e.g. the equivalents of say, claim, confess . . . b. fiction verbs (a term I borrow from Farkas 1992) – e.g. imagine, dream ... c. epistemic verbs expressing positive commitment – e.g. think, infer . . . d. commissives – e.g. promise . . . e. some factive verbs – e.g. know, foresee, realise, find out, ignore, notice ... In the definition of factive verb, I follow Karttunen 1971, assuming that factive verbs are those whose complement clause is taken to be true, regardless of the truth value of the main clause. In other words, factive verbs allow the inference of the truth of their complement proposition when they occur in either affirmative or negative sentences, as shown in example (3): (3) John knows/doesn’t know that Mary is in London. ⇒ Mary is in London. Now, while some factive verbs select indicative in all languages under consideration, others, such as the equivalents of those listed in (4), select indicative in Modern Greek, Rumanian and Hungarian, and subjunctive in Portuguese, Spanish, Catalan, French and Italian. (4) astonish, find strange, like, regret, deplore, forgive . . . This is the main difference between the two groups of languages. It constitutes a problem for practically all the semantic or pragmatic analyses of mood, which will next be considered.
. The state of the art on mood selection . The assertion/non-assertion hypothesis The first approach to be considered is advocated by Klein 1975, 1990; Hooper 1975; Bybee and Terrell 1990, among others, according to whom mood selection is related to speech acts, indicative being the mood of assertion, and subjunctive its complement. This hypothesis explains the selection of indicative in affirmative sentences, as well as its selection by declarative verbs, and it explains as well the selection of subjunctive by the equivalents of verbs like command, deny, doubt, regret and others, since they are not associated with an assertive speech act. Despite the general feeling that this hypothesis sheds some light on the issue, some questions have been raised, namely about the semantic basis for the assertion/non-assertion distinction. For instance, it is pertinent to ask whether verbs like know, realise, dream and others, which are indicative governors, can be
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.3 (131)
Semantic and pragmatic constraints on mood selection
considered assertive predicates. Another problem, as Palmer 1986 points out, is the failure of the hypothesis to explain the selection of indicative in some non-assertive contexts, such as interrogatives.
. The degrees of belief hypothesis Another hypothesis, sustained by Solano-Araya 1982, Palmer 1986, Bell 1990, and others, relates the selection of mood to the degree of belief being expressed. According to such a proposal, indicative is selected for those cases where the subject shows some positive degree of commitment to the truth of the proposition, while subjunctive is selected for those contexts where a negative commitment (as ‘I do not believe that p’, ‘I doubt that p’) or a lack of positive commitment (as ‘maybe p’, or ‘it is possible that p’) is expressed. This second theory explains the selection of subjunctive by predicates like order, want, or be scared, since they do not express a positive commitment with the truth of their complement proposition, and it explains as well the selection of indicative by the equivalents of verbs like know and by verbs like believe in affirmative sentences. Moreover, it explains why indicative is selected by verbs like believe in affirmative sentences, as (5) shows, while subjunctive may be selected if the verb occurs in negative sentences, as in (6), since in such case there is no commitment to the truth of the complement proposition: (5) Eu acredito que a Ana está doente. I believe that the Ana is-ind ill ‘I believe Ana is ill.’ (6) Eu não acredito que a Ana esteja doente. I not believe that the Ana is-subj ill ‘I don’t believe Ana is ill.’ Although this hypothesis accounts for the selection of mood in a considerable number of cases, it raises some new problems. First, as Farkas (1992) observes, some indicative governors do not commit anyone to the truth of the complement proposition. Such is the case of verbs, which she names ‘fiction verbs’, like the equivalents of imagine or dream. For instance, the truth of (7) does not allow the inference that John believes himself to be the king of Spain: (7) John dreamed that he was the king of Spain. Another problem faced by this analysis concerns the selection of subjunctive in sentences taken to be true, as are the complement clauses of a sub-group of factive verbs – as noticed above, a situation that is common to all Western Romance languages – complement clauses of some nouns, like the equivalent of fact in French,
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.4 (132)
Rui Marques
as shown in (8), and concessive clauses, as in (9). This example is given in French, but subjunctive is the only mood that may be found in concessive clauses in other of the languages under consideration. (8) Le fait qu’il soit/*est malade l’ennuie. The fact that he is-subj/*is-ind sick him bothers ‘The fact that he is sick bothers him.’ (9) Bien qu’il soit/*est malade, il finira le rapport. Although he is-subj/*is-ind sick, he finish the report ‘Although he is sick, he will finish the report.’ Since the complement clauses of the factive noun, as well as concessive clauses, are taken to be true, the degrees of belief hypothesis predicts, contra the facts in some of the languages under consideration, that the indicative is selected. The same problem is faced by other proposals that will now be briefly addressed.
. The veridicality hypothesis According to the Portuguese grammatical tradition, subjunctive is the mood associated with irreality, indicative being the default mood. The same can be said about the French grammatical tradition, which, notably in Guillaume 1929, relates the indicative/subjunctive distinction with the actuality/virtuality opposition. Such an idea is refined in the analysis of Farkas (op. cit.) and Giannakidou 1994, who propose that subjunctive is selected if the relevant clause is not taken to be true by some entity in some possible world, not necessarily the real world. Thus, one of the problems faced by previous analysis is resolved in the system of Farkas/Giannakidou, namely the selection of indicative by such verbs as imagine or dream. The complement clause of such verbs is not taken to be true by some entity in the evaluation world of the main clause. However, as Farkas proposes, they introduce a new possible world, where their complement clause is taken to be true, hence the selection of indicative. Such a view of the facts accounts for the distribution of subjunctive in Rumanian and Hungarian, as shown by Farkas, as well as in Modern Greek, as shown by Giannakidou. However, if one takes into consideration data from Portuguese and other Romance languages, the proposal faces the aforementioned problem of explaining the selection of subjunctive in the complement clauses of some factive verbs and in concessive clauses, where true propositions are involved. Such a fact constitutes a problem for any theory that makes mood selection depend on the assumption of the truth of a sentence. However, apparently a unified analysis of mood in the languages under consideration can be reached provided that the formerly known claim that mood is an expression of modality is taken seriously.
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.5 (133)
Semantic and pragmatic constraints on mood selection
. Toward an analysis In brief, the proposal being advocated here contends that, in most Romance languages, mood selection is not directly related to the (acceptance of the) truth of the clause. Instead, mood selection is assumed to be related to the attitude being expressed with respect to the proposition. More specifically, it is claimed that indicative is selected if an epistemic attitude is expressed, while subjunctive is not associated with any particular modal value, in the way of a default mood. In what follows, this hypothesis will be tested against data concerning mood selection in complement clauses and in adverbial clauses.
. Mood selection in complement clauses This section starts with mood selection by factive verbs, which pose the main problems to the majority of the analyses, as noted in the previous section. Next, mood selection by non-factive verbs will be examined, together with the influence of sentential negation in mood selection in complement clauses.
.. Mood selection by factive verbs As was shown previously, some factive verbs – such as those repeated in (10) – select indicative in all languages under consideration, while others, such as those repeated in (11), also select indicative in Rumanian, Hungarian and Modern Greek, but select subjunctive in Portuguese, Spanish, Catalan, French and Italian. (10) (the equivalents of) know, foresee, realise, find out, forget, ignore, notice, verify, ascertain, acknowledge . . . (11) (the equivalents of) astonish, find strange, like, regret, deplore, forgive, disturb, censure . . . The majority of the verbs listed in (10), that is, factive verbs that select indicative in all languages under consideration, indicate that their complement clause is taken to be true by the main subject, the only exceptions being the verbs forget and ignore. However, these two verbs express the knowledge of the speaker that their complement clause is true. Thus, all such verbs express an attitude of knowledge, an epistemic attitude. As for the factive verbs that select subjunctive in the majority of Romance languages, such as those listed in (11), they do not express an epistemic attitude. Rather, they are associated with an attitude that relates the main subject to an evaluation of the fact described in the complement clause. Thus, although such verbs also allow the inference that their complement proposition is taken to be true by the main subject and the speaker, they are associated with what Rescher 1968 calls an evaluative modality.
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.6 (134)
Rui Marques
Thus, as regards factive verbs, it seems that in most Romance languages, indicative is selected by those factive verbs associated with an epistemic modal value, while subjunctive is selected by those associated with an evaluative modal value.
.. Mood selection by non-factive verbs Taking now into consideration mood selection by non-factive verbs, the conclusion can be drawn that those that select subjunctive are associated with a wide variety of modal values, such as deontic modality (cf. order, ask, suggest . . . ), boulomaic modality (cf. want, hope . . . ), epistemic modality (cf. doubt), causative modality (cf. be enough) and others. Thus, it seems that subjunctive does not signal any particular modal value. On the other hand, taking into consideration the indicative governors, it can be seen that all of them are associated with values of epistemic modality. This is clearly the case of predicates like think or infer, and of factive verbs like know or ignore, as was seen above, but apparently of other indicative governors likewise. Fiction verbs, like imagine and dream, as proposed by Farkas (op. cit.), express the acceptance of the truth of their complement proposition in a possible world different from the real one, as said above. As for declaratives, assuming that Grice’s Cooperative Principle is being observed, it can be accepted that if one says p, one believes that p is true. On the same line of reasoning, it may be claimed that predicates like promise allow the inference that who makes the promise compromises himself to make sure that the complement proposition will become true, otherwise the speech act becomes infelicitous, in Austin’s sense. Accordingly, all such verbs are associated with an attitude that relates the main subject to the acceptance of the truth of their complement proposition. Things being so, it may be claimed that in what concerns Western Romance languages, subjunctive is not associated with any particular modal value, contrary to indicative, which requires an epistemic modal base. Hence, it follows that indicative is the marked mood, the one associated with a particular value, subjunctive being a kind of default mood, used for those contexts that do not require the selection of indicative. However, it should be noticed that subjunctive may also occur in epistemic environments, namely (i) (the equivalents of) doubt and deny select subjunctive in all languages under consideration, and (ii) at least in Portuguese, some verbs may select indicative or subjunctive in their complement clause, as those listed in (12), which are also epistemic predicates: (12) believe, admit, assume, suppose, suspect . . . Consequently, the hypothesis that epistemic environments lead to the selection of indicative must be refined, since subjunctive also occurs in epistemic contexts. One obvious difference between the verbs doubt and deny, on one side, and the majority of the verbs that select indicative, on the other side, is that the former
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.7 (135)
Semantic and pragmatic constraints on mood selection
do not commit the main subject to the acceptance of the truth of their complement clause, contrary to most indicative governors. In other words, they express a negative epistemic attitude. Other epistemic verbs that do not indicate that the main subject takes the complement clause to be true are ignore and forget, which, however, are indicative governors: (13) O Paulo ignorava/esqueceu-se que a Maria estava The Paulo ignored/forgot that the Maria was-ind em Londres. in London ‘Paulo ignored/forgot that Maria was in London.’ Nevertheless, such verbs allow the inference that according to the speaker their complement clause is true, while doubt and deny do not commit anyone to the truth of their complement clause. Therefore, ignore and forget express a positive epistemic attitude, although it relates to the speaker, not to the main subject, while doubt and deny express a negative epistemic attitude. As was pointed out, another epistemic environment where subjunctive occurs is the complement clause of verbs, like believe, which admit indicative or subjunctive in their complement clause. With such verbs, the selection of indicative expresses a stronger belief (of the main subject) on the truth of the complement clause than the one that is expressed if subjunctive is selected. If a strong belief is expressed, the selection of subjunctive appears as odd, the same being the case with the selection of indicative if a low degree of belief is being conveyed: (14) A Ana acredita sinceramente que ele está em Paris/?esteja The Ana believes sincerely that he is-ind in Paris/is-subj em Paris. in Paris ‘Ana really believes that he is in Paris.’ (15) A Ana acredita que ele está/?esteja em Paris, aliás The Ana believes that he is-ind/?is-subj in Paris, actually, [she] tem quase a certeza. has almost the certainty ‘Ana believes that he is in Paris, actually she is almost sure.’ (16) Custa acreditar que ele saia/?sai daqui. [It] costs believe that he leave-subj/?leave-ind from here ‘It is hard to believe that he leaves.’
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.8 (136)
Rui Marques
.. Sentential negation and mood selection in complement clauses Given that the verbs of the believe type select subjunctive when associated with a low degree of belief, the prediction should be that, in negative sentences, such verbs select subjunctive, since in such case there is no commitment of the main subject to the complement clause. Such prediction is confirmed by examples like (17): (17) A Ana não acredita que o Paulo esteja doente. The Ana not believes that the Paulo is-subj sick ‘Ana doesn’t believe that Paulo is sick.’ However, in sentences like (17) indicative might also occur: (18) A Ana não acredita que o Paulo está doente. The Ana not believes that the Paulo is-ind sick ‘Ana doesn’t believe that Paulo is sick.’ That is, verbs like believe may select indicative and subjunctive both in affirmative and in negative clauses. In affirmative clauses, it was claimed that such verbs select one mood or the other depending on the degree of belief being expressed. Obviously, the same explanation can not be given to their behaviour in negative clauses: the degree of belief of the main subject in the truth of the complement clause is null regardless of the mood being selected. Nevertheless, there is one contrast between the interpretations of (17) and (18). In (18), the speaker is claiming that he knows that the complement clause is true, while in (17) he remains neutral with respect to the acceptance of the truth of such clause. In other words, although both (17) and (18) state that the main subject does not believe in the complement clause, in (18), but not in (17), the speaker expresses his own belief in the truth of such a clause. That is why in (19) subjunctive can occur, but not the indicative: (19) Eu não acredito que o Paulo esteja/*está doente. I not believe that the Paulo is-subj/*is-ind sick ‘I don’t believe that Paulo is sick.’ If indicative is selected, a contradiction arises. On one hand, it is claimed that the main subject does not believe in the complement clause, but, on the other, by selecting indicative, it is claimed that the speaker believes that such a clause is true. Now, since the referent of the main subject is the speaker, it is claimed that such a person believes and at the same time does not believe that the complement clause is true. With the selection of subjunctive, no such contradiction ensues. The sentence just expresses the absence of belief of the main subject as regards the truth of the complement clause. Notice, however, that in example (19), the indicative could occur if the main clause was temporally located in the past, as in the following example:
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.9 (137)
Semantic and pragmatic constraints on mood selection
(20) Naquela altura, eu não acreditava que o Paulo estivesse/estava In that time, I not believe that the Paulo was-subj/was-ind doente. sick ‘In those days, I didn’t believe that Paulo was sick.’ In this example, if indicative is selected, then it might be inferred that at the utterance time the speaker believes in what he did not believe at some previous time, namely that Paulo was sick. So, in this case, there is no contradiction, since two stages of belief of the same person are taken into consideration. Thus far, the influence of sentential negation on mood selection was only considered with respect to verbs like believe. However, sentential negation also interferes with mood selection by other classes of verbs. Such is the case of predicates like doubt and deny, which, as seen before, are subjunctive governors in affirmative sentences, as is the case of predicates like say, promise and others, which select indicative in affirmative sentences. As shown by the following examples, in negative sentences such verbs may select a different mood from the one they select in affirmative sentences: (21) Eu não duvido (de) que a Ana está doente. I not doubt (of) that the Ana is-ind sick ‘I do not doubt that Ana is sick.’ (22) Eu não nego que a Ana está doente. I not deny that the Ana is-ind sick ‘I do not deny that Ana is sick.’ (23) Eu não digo que ele esteja enganado. I not say that he is-subj wrong ‘I am not saying he is wrong.’ (24) Eu não prometo que chegue a tempo. I not promise that arrive-subj at time ‘I do not promise to arrive on time.’ Moreover, if negation is present, verbs like doubt or deny – subjunctive governors in affirmative sentences – and verbs like say or promise – indicative governors in affirmative sentences – may select the same mood they select in affirmative clauses only if the negation is a metalinguistic negation, as in the following examples: (25) Eu não duvido (de) que a Ana esteja doente. I not doubt (of) that the Ana is-subj sick’ ‘I do not doubt that Ana is sick.’
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.10 (138)
Rui Marques
(26) Eu não nego que a Ana esteja doente. I not deny that the Ana is-subj sick ‘I do not deny that Ana is sick.’ (27) Eu não disse que ele estava enganado. I not said that he was-ind wrong ‘I did not say he was wrong.’ (28) Eu não prometi que chegava a tempo. I not promise that arrive-ind at time ‘I did not promise to arrive on time.’ In such cases, the complement clause is given information in discourse. That is, sentence (27), for instance, is interpreted as ‘You think I said he was wrong, but that was not what I said’. Thus, the role of the negative sentence is to correct a previous affirmative statement – the one whose informative content is being corrected, the negation operator not interfering with mood selection. Leaving aside these cases of metalinguistic negation, the point that was made was that sentential negation interferes with mood selection by those verbs associated with a value of positive or negative belief, making them select the mood they do not admit in affirmative sentences. Such a fact is not surprising, since negation reverses the degree of belief being expressed. Thus, not doubt expresses a positive belief of the main subject on the truth of the complement clause, while not say does not express a belief in the truth of the complement clause. Hence, the selection of mood is in accordance with the epistemic attitude being expressed: a negative epistemic attitude leads to the selection of subjunctive, while a positive epistemic attitude leads to the selection of indicative. In sum, what the data shows is that, in epistemic environments, indicative is selected if the relevant clause is taken to be true either by the subject or the speaker, while subjunctive is selected if the attitude being expressed is absence of belief or, at least, a low degree of belief. As regards those verbs that are not associated with the expression of belief, negation does not interfere with mood selection. In fact, predicates like want, be scared, avoid, regret and many others, which are associated with non-epistemic values, select subjunctive both in affirmative and in negative sentences: (29) O Paulo (não) queria que a Ana o visse. The Paulo (not) wanted that the Ana him see-subj ‘Paulo did not want/wanted Ana to see him.’ (30) O Paulo (não) lamenta que a Ana o tenha visto. The Paulo (not) regrets that the Ana him had seen-subj ‘Paulo does not regret/regrets that Ana saw him.’
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.11 (139)
Semantic and pragmatic constraints on mood selection
Regardless of the fact that the main sentence is negative or affirmative, the attitude being expressed is non-epistemic. Therefore, according to my hypothesis, since there are no grounds for selecting indicative, it is the subjunctive, the complement mood, that is selected. Finally, factive epistemic verbs, like know, find out or ignore, select indicative both in affirmative and in negative sentences: (31) O Paulo (não) sabe que a Ana o viu. The Paulo (not) knows that the Ana him saw-ind ‘Paulo does not know/knows that Ana saw him.’ (32) O Paulo não ignora que a Ana o viu. The Paulo (not) ignores that the Ana him saw-ind ‘Paulo does not ignore/ignores that Ana saw him.’ The reason why mood selection by such verbs is not affected by negation, I think, is that in both affirmative and negative sentences an epistemic attitude is expressed. The complement clause may be taken to be true just by the speaker, if the main predicate is ignore or not know, or it may be taken to be true by the speaker and the main subject, if the main predicate is know or not ignore, for instance, but in either case the expressed attitude is epistemic in nature.
.. General picture The observations made so far lead to the hypothesis that, in Western Romance languages, indicative is selected if a positive epistemic attitude is being expressed, otherwise subjunctive being selected. As was seen, both the properties positive and epistemic are necessary to explain the selection of indicative, since subjunctive may also occur in epistemic environments if the expressed attitude is lack of belief. Consequently, the concept of veridicality seems to be at work here: if the relevant clause is not taken to be true by some entity, indicative may not occur. However, it should be noticed that the concept of veridicality can not, just by itself, explain the distribution of subjunctive and indicative in all languages under consideration: in Western Romance languages, subjunctive occurs in veridical contexts, as complement clauses of some factive verbs. In such languages, veridicality seems to be a necessary condition for the selection of indicative, but not a sufficient one. The concepts that seem to rule mood selection in complement clauses of most Romance languages are [veridicality] and [epistemic modal value]. That is, indicative is selected for veridical epistemic contexts, while subjunctive is selected for other contexts. As for the other languages that were considered – Rumanian, Modern Greek and Hungarian – indicative does not seem to be sensitive to the epistemic value. For such languages, the distribution of the two moods can be explained by the concept of veridicality, as proposed by Giannakidou 1994; that is, the indicative
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.12 (140)
Rui Marques
Table 1. Mood selection in complement clauses Rumanian, Hungarian, Modern Greek Epistemic verbs (ex.: verify, know ignore, think, say, dream, promise) Veridical verbs
Portuguese, Spanish, Catalan, French, Italian INDICATIVE
INDICATIVE Non-epistemic verbs (ex.: regret, be odd)
Non-veridical verbs (ex.: be enough, be useful, doubt, avoid, forbid, allow, want)
SUBJUNCTIVE SUBJUNCTIVE
is selected if the complement clause is taken to be true, otherwise the subjunctive being selected. Table 1 summarises the observations that were made with respect to mood selection in complement clauses in the languages under consideration.
.. Mood selection in nominal or adjectival argument clauses Thus far, only verbal complement clauses were considered. However, a quick look at the nouns and adjectives that select indicative or subjunctive reveals that in Western Romance languages, those that select indicative – as exemplified in (33) – express a positive epistemic attitude, while none of those that select subjunctive – exemplified in (34) – express such an attitude. These facts corroborate the hypothesis sustained in the present paper. (33) Nouns: affirmation, certainty, conclusion, confession, conviction, knowledge, belief, discovery, ignorance, promise, verification . . . Adjectives: aware, conscious, convinced, clear, informed . . . (34) Nouns: acceptance, authorisation, desire, doubt, interest, fear, need, prohibition, rejection . . . Adjectives: tired, eager, hopeful, full, interested, indigent, fearful . . .
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.13 (141)
Semantic and pragmatic constraints on mood selection
. Mood selection in adverbial clauses The adequacy of the hypothesis formulated in the previous section will now be evaluated against mood distribution in adverbial clauses. In all languages under consideration subjunctive is selected by non-veridical operators, that is, operators that do not allow the inference that the clause they introduce is true according to some relevant entity. Such is the case of the equivalents of for, even if, without, before, until, unless, provided, as shown by sentences (35)–(41). Once again, these examples are from Portuguese, but subjunctive is also selected for those contexts in the other considered languages. (35) Ele saiu mais cedo, para que ninguém o visse. He left more soon, so that nobody him see-subj ‘He left sooner, (in order) for no one to see him.’ (36) Ele vem sempre a pé, mesmo que esteja a chover. He comes always at foot, even if [it] be-subj at rain ‘He always comes on foot, even if it is raining.’ (37) Ele saiu sem que alguém tenha reparado. He left without that anybody has-subj noticed ‘He left without anybody noticing.’ (38) Eu saí antes que ele entrasse. I left before that he come-subj in ‘I left before he would come in.’ (39) Ele fica a trabalhar até que alguém o chame. He stays working until that someone him calls-subj ‘He will stay working until someone calls him.’ (40) Ele vem a pé, a menos que esteja a chover muito. He comes at foot, unless that [it] be-subj raining a lot ‘He will come on foot, unless it rains heavily.’ (41) O Paulo aceita o cargo desde que possa escolher The Paulo accepts the duty provided that [he] can-subj choose os seus assessores. the his assessors ‘Paulo accepts the duty provided he can choose his assessors.’ Indicative, again in all languages under consideration, is the mood selected in causal clauses, cf. (42), and in temporal clauses if the sentence is taken to be true, cf. (43):
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.14 (142)
Rui Marques
(42) Ele saiu mais cedo porque tinha uma reunião. He left more soon because [he] had-ind a meeting ‘He left sooner because he had a meeting.’ (43) Ele saiu quando o chamaram. He left when him called-ind ‘He left when he was called.’ Therefore, it seems that in adverbial clauses indicative is selected if the sentence is taken to be true, otherwise subjunctive is selected. This seems to be the case in Rumanian, Hungarian and (Modern) Greek, as well as in Spanish and Catalan. However, taking into consideration data from Portuguese, French and Italian, it can be seen that there are two kinds of adverbial clauses where the selection of mood is not related just to the truth of the proposition, namely concessive clauses and some conditionals. In what concerns concessive clauses, as (44), the question is that subjunctive is selected, despite the fact that the sentence is taken to be true. (44) Embora seja domingo, o Paulo está a trabalhar. Although [it] is-subj Sunday, the Paulo is working ‘Although it is-subj Sunday, Paulo is working.’ As for conditional clauses, in most of the languages under consideration, indicative is selected if the conditional is factual, as in (45). If the conditional is hypothetical, as in (46), or counterfactual, as in (47), subjunctive is selected, (although in Portuguese, indicative may also be selected, but mainly in colloquial speech). (45) Se, como dizes, ele está doente, então está em casa. If, as [you] say, he is-ind sick, then [he] is at home ‘If, as you say, he is sick, then he is at home.’ (46) Se ele estiver doente, então está em casa. If he is-subj sick, then [he] is at home ‘If he is sick, then he is at home.’ (47) Se ele estivesse doente, então estaria em casa, mas If he was sick, then [he] be-subj at home, but não está. [he] not is ‘If he was sick, then he would be at home, but he isn’t.’ However, in French, indicative is used in all kinds of conditionals, that is, even if the antecedent is taken to be false, indicative is selected:
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.15 (143)
Semantic and pragmatic constraints on mood selection
(48) Si (comme tu dis) il est malade, alors il est chez lui. If (as you say) he is-ind sick, then he is at home ‘If (as you say) he is sick, then he is at home.’ (49) S’il était malade, il serait chez lui, mais il n’est pas If he was-ind sick, then he be at home, but he not is chez lui. at home ‘If he was sick, then he would be at home, but he is not at home.’ Accordingly, the veridicality hypothesis, which claims that indicative is selected if the sentence is taken to be true, otherwise subjunctive being selected, does not explain the selection of subjunctive in concessive clauses in Portuguese, French and Italian, as it does not explain the selection of indicative in counterfactual and hypothetical conditionals in French. The idea I postulated, when considering complement clauses, was that indicative is the strong mood, selected for veridical epistemic contexts. That is, indicative is selected if the sentence is taken to be true and the truth of the proposition is at stake. If the sentence is taken to be true but the attitude being expressed does not relate to the acceptance of its truth, subjunctive is selected. It is my claim that the same principle is at work in adverbial clauses. Concessive clauses, as is well known, express a contrast between the two connected propositions, the main proposition being taken as unexpected in face of the concessive subordinate. The proposition that is marked as unexpected is therefore the main clause, not the concessive one. By selecting the strong mood – indicative – for the main clause, and the weak mood – subjunctive – for the concessive clause, the speaker is clearly focusing on the main proposition, thus explicitly marking the contrast that such constructions are associated with. Hence, it seems reasonable to think that subjunctive is chosen for concessive clauses in Portuguese, French and Italian simply because the speaker is not concentrating on the acceptance of the truth of such clause. That is, the information that is being focused on is not the one provided by such clause, for which reason the selection of the strong mood is not justified. As for Rumanian, Modern Greek and Hungarian, it was claimed that indicative was chosen for veridical contexts, otherwise subjunctive being selected. Thus, indicative is the mood chosen by concessive operators because they are veridical. Notice that in concessive clauses, Spanish and Catalan behave like Rumanian, (Modern) Greek and Hungarian – since indicative is also chosen – while in complement clauses they behave like Portuguese, French and Italian, where the complement clause of verbs like regret – a veridical context – exhibits subjunctive. One can hypothesise that this is due to the fact that in concessive clauses a non-epistemic
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.16 (144)
Rui Marques
modal value is not being conveyed, contrary to what happens in complement clauses of such verbs, where the attitude being expressed is meta-epistemic. As regards the antecedent of conditionals, the problem is that in French indicative is selected even if the sentence is taken to be false, which suggests that mood selection does not depend on the truth value of the clause. However, subjunctive does not lead to ungrammaticality, mainly occurring in formal and literary speech, as in examples (50)–(52), quoted by Grevisse, which suggests that in French indicative is taking the place of subjunctive in the antecedent of conditionals. (50) S’il fût venu, je l’aurais su. If he had-subj come, I it would know ‘If he had come, I would have known.’ (51) Elle resta stupéfaite et en larmes comme si elle eût déjà vu sa destinée tout entière. (Michelet, Jeanne d’Arc) She stayed stupefact and in tears as if she had-subj already seen her destiny all entire ‘She stayed stupefact and in tears, as if she had already seen all her destiny.’ (52) Ils discutaient maintenant comme si la vieille femme n’eût pas été présente. (Mauriac, Noeud de vip) They discussed now as if the old woman not had-subj neg-particle been present ‘They discussed now as if the old woman had never been there.’ The central hypothesis I sketched was that in most Romance languages indicative is selected if the sentence is taken to be true and it occurs in an epistemic modal base. Conditionals clearly are an epistemic environment, for which reason it is only natural that in French indicative occurs very easily in this context. It seems that the epistemic feature is being made more prominent than the veridical one.
. Final remarks The general hypothesis I have been sustaining can be synthesised as follows: (i) in languages like Rumanian, Modern Greek and Hungarian, indicative is selected in veridical contexts and subjunctive in non-veridical ones; (ii) in languages like Portuguese, Spanish, Catalan, French and Italian, the selection of indicative is sensitive not just to a veridical value, but also to an epistemic one. That is, indicative is selected for those veridical contexts where the truth of the proposition is at stake. However, some variation is verified among the languages of the latter group, namely: in the first place, in Spanish and Catalan indicative occurs in concessive clauses, while the other Western Romance languages select subjunctive; secondly,
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.17 (145)
Semantic and pragmatic constraints on mood selection
in French indicative may occur in the antecedent of any conditional, while in the other languages of the group subjunctive is most easily found. Such differences suggest that, while in all these languages the selection of indicative is triggered by the values [+ veridical] and [+ epistemic modality], some languages may give more salience to one feature or the other. French would be more sensitive to the epistemic feature, thus allowing indicative to occur in non-veridical epistemic contexts – as the antecedent of counterfactual conditionals – while Spanish and Catalan would be more sensitive to the veridical feature, subjunctive being chosen only if the sentence is not taken to be true or if the acceptance of its truth is sheltered by a non-epistemic value. Portuguese and Italian would give no prominence to one feature over the other, subjunctive being chosen if the sentence is not taken to be true or if the acceptance of its truth is not being focused, either because the sentence occurs in a non-epistemic modal base, as in complement clauses of verbs like regret, or because it is not the main information that is being conveyed, as is the case with concessive clauses.
Notes * The participation of the author in the Second International Conference on Contrastive Semantics and Pragmatics was financed by the Fundação para a Ciência e a Tecnologia.
References Bell, A. (1990). El modo en español: Consideración de algunas propuestas recientes. In I. Bosque (Ed.), Indicativo y subjuntivo (pp. 81–105). Madrid: Taurus Universitaria. Bybee, J., & Terrell, T. D. (1990). Análisis semántico del modo en español. In I. Bosque (Ed.), Indicativo y subjuntivo (pp. 145–163). Madrid: Taurus Universitaria. Farkas, D. (1992). On the semantics of subjunctive complements. In P. Hirschbühler & K. Koerner (Eds.), Romance Languages and Modern Linguistic Theory (pp. 71–104). Amsterdam/Philadelphia: John Benjamins. Giannakidou, A. (1994). The semantic licensing of NPIs and the Modern Greek subjunctive. In Language and Cognition, 4, yearbook of the Research Group for Theoretical and Experimental Linguistics. University of Groningen. Grevisse, M. (1993). Le Bon Usage: Grammaire Française (13th edition by André Goosse). Paris: Duculot. Guillaume, G. (1929). Temps et verbe: Théorie des aspects, des modes et des temps. La Société de Linguistique de Paris, XXVII. Paris. Hooper, J. B. (1975). On assertive predicates. In P. Kimball (Ed.), Syntax and Semantics, 4 (pp. 91–124). New York: Academic Press.
TSL[v.20020404] Prn:29/08/2002; 15:48
F: PBNS08.tex / p.18 (146)
Rui Marques
Karttunen, L. (1971). The Logic of English Predicate Complement Constructions. Indiana University Linguistics Club. Klein, F. (1975). Pragmatic constraints in distribution: The Spanish subjunctive. Papers from the IIth regional meeting, Chicago Linguistic Society (pp. 353–365). Klein, F. (1990). Restricciones pragmáticas sobre la distribución del subjuntivo en español. In I. Bosque (Ed.), Indicativo y subjuntivo (pp. 303–313). Madrid: Taurus Universitaria. Palmer, F. R. (1986). Mood and Modality. Cambridge: Cambridge University Press. Rescher, N. (1968). Topics in Philosophical Logic. Dordrecht: D. Reidel Publishing Company. Solano-Araya, J. (1982). Modality in Spanish: An Account of Mood. Ph.D. dissertation, University of Kansas.
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.1 (147)
Dilemmas and excogitations Further considerations on modality, clitics and discourse A. Capone Università degli Studi di Messina
.
Introduction
The clitic lo in Italian, in combination with a verb of propositional attitude (e.g. sapere (know), sentire (hear), capire (understand), etc.), seems to connect anaphorically with a previous thought already vocalized in discourse and presupposed by the speaker and the hearer at the moment of utterance. In this article, I do my best to explain the nature of this discourse relation and raise various questions about the semantics and pragmatics of the clitic in question. Various crucial dilemmas arise. What is the linguistic nature of presupposition? How does this notion connect with the issue of clitics? In what ways can clitics affect our understanding of the notion of presupposition? Is the presuppositional function of clitics a semantic or a pragmatic phenomenon? The structure of this paper is the following. I discuss Stalnaker’s notion of presupposition which seems to be tailored to my considerations on clitics. Then I present the local satisfaction approach to presupposition. At this point I discuss the presuppositions of clitics, clitic-left-dislocation, and the semantics/pragmatics debate in connection with clitics. I demonstrate that clitics, surprisingly, complicate the projection rules for presuppositions. At last I consider various problems raised by my approach.
. Presupposition In this section I shall present some of the background on presuppositions which is going to be of use in the understanding of the conversational dynamics of clitics in languages such as Italian. This background is not going to be exhaustive con-
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.2 (148)
A. Capone
sidering the proliferation of studies on presuppositions, but will consist of considerations which specifically bear on the issue under consideration. I will consider here presupposition as preeminently a linguistic issue, in other words the study of the relation between the use of sentences and the contexts in which they can felicitously appear (Stalnaker 1974; Chierchia & McConnell-Ginet 1990). The pairing of speech acts and the contexts in which they can felicitously appear is an intriguing question and one which bears on the logic of use of linguistic expressions. Suppose we want to say that John regrets that he came late to the cinema. I believe that it is a property of the verb regret that the fact regretted must be presupposed for the sentence to be uttered felicitously. Thus, I believe we are faced with a bundle of linguistic properties. The presuppositional expression treats the embedded proposition as information which is either given or uncontroversial, in the sense that we would not expect the hearer to object to it. In fact consider (1) (1) John regrets that he came late to the cinema in a context where the hearer is not aware of the fact that John came late to the cinema. The hearer will presumably accommodate the presupposition that p is true in the sense that he will take the speaker to believe that this proposition is uncontroversial. Concerning the circumstances that render a proposition uncontroversial, I will rehearse here some judicious considerations by Soames (1982: 430): A speaker S takes a proposition P to be uncontroversial at t (or, equivalently, takes P for granted at t) iff, at t, S accepts P and thinks (a) that P is already part of the conversational context at t; (b) that the other conversationalists are prepared to add P to the context without objection.
Now, I think some reflection is required on what this is supposed to mean. What is indubitable is that the context might present an asymmetry between the assumptions that the speaker and the hearer take for granted. Once the sentence (1) is uttered, the hearers are entitled to infer that the speaker takes that proposition for granted. However, if they do not take that proposition for granted, we shall not say that the pairing of the context and of the utterance has been infelicitous. I want to point out that, whereas certain presuppositional expressions do not commit the speaker to having any beliefs about what the hearer might believe, we might discover a class of presuppositional expressions which require, among other things, that the speaker and the hearer share the same presuppositions at a time prior to the time in which the presuppositional expression is uttered. I will propose in a later section that clitics are candidates for membership in such a class. Let me introduce, for the time being, some terminology. If A is a presuppositional expression, then we might call Ap the presuppositions of A. We shall say that the context will satisfy the presuppositions of A, in other words, Ap, if the speaker
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.3 (149)
Dilemmas and excogitations
and the hearers have the required beliefs. One of the required beliefs is that the speaker takes for granted Ap and that he invites the hearer to assume that he does so. If the presuppositional expression, however, required both the speaker and the hearers to share a certain presupposition, then the statement will be satisfied by the context only if the context includes the assumption that the speaker and the hearer share the same presuppositions. Stalnaker (1974) has provided an interesting definition of speaker’s presupposition. According to him, a speaker presupposes p if he acts as if he is disposed to believe that p and to take the hearer to recognize that he is so doing. This is a deep definition, which is intended to encompass a range of situations. Clearly, there might be discourses where pretense is involved. For example, by uttering a linguistic expression, I might pretend not to know that certain facts have occurred. If the hearer takes my speech at face value, then it might be said that I am acting as if I had certain beliefs rather than others. Suppose John has a hideous defect and he wants others to pretend that they do not notice it. Then he will act as if he did not have that defect. The hearers, who know about that defect, might act as if they did not know that defect. John might know that his hearers know about that defect, but he is not ready to acknowledge that and goes on with his pretence, while, at the same time, the hearers go on with their pretence. In this case, we might be justified in saying that John is acting as if he had certain beliefs, certain presuppositions. It has sometimes been argued in recent linguistic theorizing that, by accommodation, a presupposition can be introduced as new information. Basically, Stalnaker (1973) and Soames (1982) have provided examples like the following: (2) A: Mary is quite pretty/ B: Her husband thinks that too. It has been claimed that here speaker B is providing the information that Mary has a husband as new. What is certain is that speaker B acts as if he takes p for granted, in the sense that he acts as if he takes p for granted and believes that the hearer might add p to his beliefs without controversy. He knows nonetheless that A has made a social gaffe and ignores the fact that Mary has a husband. The speaker’s intention in the statement is to apprise A of the fact that Mary has a husband without acting as if he were doing this. What he is officially doing is to provide the new information that Mary’s husband has a certain thought while he is covertly informing the hearer that Mary has a husband. I believe that this subtle difference between the official and the unofficial record of the conversation has to be made. The information that Mary has a husband will be considered as given or uncontroversial and the conversational effect of warning the hearer does specifically depend on that information being acted upon as given or uncontroversial. Actually, Stalnaker’s definition of speaker’s presupposition is somewhat more complicated, since he proposes we should say that A presupposes p in case he is
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.4 (150)
A. Capone
disposed to act as if he believes that p. I believe that this specification is intended to emphasise the fact that presupposition is taken to be a mental phenomenon and, in particular, as Stalnaker says, a propositional attitude. My interlocutor and I might presuppose p even if nothing I say or he says may lead others who overhear the conversation to form the belief that that assumption is presupposed. My interlocutor and I might be disposed to act as if we each believe that p is part of the common ground but we may not act so. We assume that those propositions which are shared by the speaker and the hearer constitute the common ground. Statements which have been uttered up to a point in a conversation are stored in memory and are labelled as being part of the common ground. Salient visual images that come from the immediate surrounding environment are also part of the common ground. We may ask the question why presuppositions must be satisfied by the relevant local contexts. A possible answer has been given by Stalnaker (1973). The point of uttering S is to restrict possible worlds that belong to the context by eliminating those worlds in which S is false. It follows from this that if the possible worlds which make the propositions accepted in the context C true do not contain the presupposition p, then the assertion of S is useless because it cannot restrict further the possible worlds of the context since these do not contain the possible world that makes the proposition true.
. Context and presupposition I would like to add some further considerations on the interesting interplay between context and presupposition. Examples of this intriguing interplay proliferate easily but I think it is best to consider some cases brought to our attention by Stalnaker (1974). Consider in fact (3) and (4): (3) If I regret later that I have not told the truth, I will confess it to everyone. (4) If I discover later that I have not told the truth, I will confess it to everyone. Stalnaker (1974: 477) has pointed out that unlike (3), (4) does not presuppose that the speaker has not told the truth. He imputes this presupposition failure to the following reasoning: if a speaker explicitly supposes something, he indicates that he is not presupposing it. So when the speaker says If I discover later that p, he indicates that he is not presupposing that he will discover later that p. But if it is an open question for a speaker whether or not he will at some future time have come to discover that p, he can’t be assuming that he already knows that p. And if
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.5 (151)
Dilemmas and excogitations
he is not assuming that he himself knows that p, he can’t be assuming that p. Hence p cannot be presupposed.
I believe that Chierchia and McConnell-Ginet (1990) must be right when, in discussing this example, they state the question differently. For them the question is not why the presupposition is cancelled but why the implicature fails to arise. The reason why the implicature fails to arise must be that the utterance of a conditional sentence such as If I discover that p, implicates that the speaker has not yet discovered that p and thus p cannot be presupposed. The following case is even more interesting: (5) Did you discover that you had not told the truth? (6) Did you regret that you had not told the truth? The question in (5) does not presuppose you had not told the truth while that in (6) does. Stalnaker claims that the inference does not arise because by asking a question one presumes that one does not have an answer to it. However, when he has to explain the different properties of discover in this context as compared with those of regret, he must provide the embarrassing answer that the differences reduce to the different properties of the verbs in question. A better explanation must be one along the following lines. S has asked the question Q: Did you discover that p? One purpose of this question could be that the speaker wants to know whether p is the case and one way to find out is to inquire into H’s knowledge state by asking the question. Another purpose is that he wants to know whether H is aware of the proposition p which he (the speaker) has come to know in a different way (from a different source). Thus, there is at least one interpretative reading of Did you discover that p? which cannot presuppose p (we need not posit a semantic ambiguity, but only an interpretative ambiguity, along the lines of Jaszczolt 1999). Consider now Did you regret that p? This question is one that is usually asked not to inquire into H’s state of knowledge but to inquire into H’s feelings. The purpose in asking this question cannot be that S wants to know whether p. Furthermore, if the speaker inquires into H’s feelings about p, then he is not doing that on the basis of a mere supposition that p is true but he must know that p is true. Hence the inference that p is true.
. The projection problem Now I want to move on to the projection problem as dealt with by linguists such as Karttunen (1974), Soames (1982, 1989), Chierchia and McConnell-Ginet (1990) and Heim (1992). For the sake of space, I will omit discussion of another strand of presupposition theory based on the treatment of presupposition as anaphora.
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.6 (152)
A. Capone
Especially considerations by Heim (1992) interact with the issue of clitics. Clitics, in fact, seem to require semantically those presuppositions which, according to Heim, are accommodated pragmatically. At the heart of the modern solutions to the projection problem is the assumption that any new assertion expands the context of utterance and thus leads to an incrementation of information which constitutes the context in which a new assertion is evaluated. An assertion of p will be admissible in the context C if: a.
the assertion of p is informative with respect to the context C (in other words the context C does not entail p); b. the conjunction of the propositions entailed by C and p maintains consistency (van der Sandt 1992). The presuppositions of a complex expression are satisfied if the presuppositions of each of its component clauses are satisfied locally. Those presuppositions which are not satisfied locally must be satisfied by the input context and thus are inherited as presuppositions of the complex expression. Consider: (7) If John has children, then his children must be happy. The presupposition of the consequent clause is not inherited by the whole sentence. Various treatments have been given for this empirical fact. One way to deal with it is merely to say that the presupposition of the consequent is entailed by the ifclause and thus it is filtered out. Gazdar (1979) has tried to give an explanation for this fact and has claimed that the Q-implicature triggered by the maxim of Quantity takes precedence over the presupposition of the consequent clause and thus defeats it. Another way to deal with this example is to say that the presupposition of the consequent is satisfied trivially by the if-clause and, thus, is not inherited by the complex expression. The way I understand this example is the following. The speaker makes the supposition that John has children; the consequent clause presupposes that John has children, this presupposition is satisfied in the context of the possible worlds invoked by the supposition and thus need not be satisfied by the worlds that belong to the context of utterance. However, if the supposed clause did does not entail the presupposition of the consequent, then one would have to find some possible worlds belonging to the context C that satisfy the presupposition. Thus if I say (8) If John is happy, his children must be happy, the presupposition that John has children has not been satisfied in the immediately preceding local context and, thus, it must be satisfied by C (hence it is inherited
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.7 (153)
Dilemmas and excogitations
by the complex expression). Let us consider now the following example by Soames (1982): (9) If the Poles defeat the Russians, then the Hungarians will defeat the Russians too. Soames’s point, in discussing this example, is to claim that Gazdar’s strategy here fails, in so far as his theory should predict that the complex sentence has a presupposition while it is intuitively clear that the complex sentence inherits no presupposition from the main clause. Gazdar’s strategy is to allow the presupposition of the consequent to be cancelled by a conversational implicature triggered due to the maxim of Quantity by the antecedent. But here no quantity implicature cancels the presupposition of the consequent; hence his prediction is false. Soames, on the contrary, is in a position to predict that the complex sentence inherits no presupposition, since the presupposition of the main clause is satisfied by the local context provided by the antecedent. We still need to deal with sentences in which the if-clause follows instead of preceding the main clause, as in (10): (10) Andy will probably meet with the PLO again this week, if he met with them last week. About this example, Soames (1982) considers that the presupposition of the main clause due to again is satisfied by the if-clause, hence it cannot ascend to become a presupposition of the complex sentence. For similar cases, Soames (1982: 445) devises the following projection rule: If S = B, if A then X satisfies S iff (a) X + A satisfies B relative to U (b) X + B satisfies A relative to U. Let us now consider conjunctions between statements, such as John has got three children and his children are happy. It is intuitively clear that this sentence presupposes nothing since the presupposition of the second conjunct is satisfied (in so far as it is entailed) by the first conjunct. Given that this sentence has been used to assert the presupposition of the second conjunct, it is evident that the sentence as a whole presupposes nothing. As was the case with conditionals, those presuppositions of constituent clauses of the whole sentence which are satisfied by the local context (the immediately embedding clause) do not ascend to become presuppositions of the whole sentence. We might give a projection rule both for conditionals and conjunctions such as the following, first suggested by Karttunen (1974) and then adopted by Soames (1982) and Heim (1992) in subsequent work:
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.8 (154)
A. Capone
(C + φ) + ψ. This basically means that the presuppositions of φ must be satisfied by C (in other words the context C must entail the presuppositions of φ), while the presuppositions of ψ must be satisfied by (C + φ) ((C +φ) must entail the presuppositions of ψ). Let us now see how things work for disjunctions. Consider the statement (11) Either there is no king of France or the king of France is in hiding. According to Gazdar, the presupposition of the second disjunct evaporates in so far as by uttering the disjunction one may be prepared to accept that one conjunct is false. This might be due to a conversational implicature through the maxim of Quantity. Scholars such as Karttunen (1974) and Soames (1982) have proposed the following projection rule for disjunction: (C + (–φ)) satisfies ψ and (C + (–ψ)) satisfies φ. According to this rule the presupposition that there is a king of France (of the second conjunct) must be satisfied by the negation of the other conjunct. The negation of There is no king of France is There is a king of France. This obviously satisfies the presupposition of the first conjunct, thus the presupposition is filtered out, since we have said previously that presuppositions which are locally satisfied do not ascend to become presuppositions of the whole sentence (the reader will find this projection rule more intelligible by considering the logical equivalence of p or q and If not p, q). Now I want to consider an interesting example from Chierchia and McConnellGinet (1990): (12) Joan managed to land a job on Wall Street and she doesn’t regret that she will be rich. This sentence is uttered in a context C that entails It would be difficult for Joan to land a job on Wall Street but if she does, she’ll be rich. S1 is Joan managed to land a job on Wall Street and its presupposition is S1p, that is it was difficult for Joan to land a job on Wall Street. S2 is Joan doesn’t regret that she will be rich and its presupposition is that Joan will be rich. It is felicitous to utter the complex conjunction S in C because the presupposition of S1 is satisfied by the context C. The presupposition of S2 is not satisfied by the context C, however by adding the assertion of S1 to context C one obtains a context (C + S1) that does entail the presupposition of S2, which is thus satisfied. I believe that this is an interesting case in which the context is incremented by what is asserted and this context incrementation is used as the common ground on which further assertions can be felicitously uttered.
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.9 (155)
Dilemmas and excogitations
Karttunen (1974) has provided various examples in which the context prior to the utterance contains some propositions which must be added to a clause of a conditional so that a presupposition of its main clause can be satisfied. Here is an interesting one: (13) If Miss Woods destroyed the missing tapes, Nixon is guilty too. The consequent clearly exhibits a presupposition which must be satisfied by the local context. In this case, however, the presupposition cannot be satisfied unless the immediate relevant local context (the preceding clause) is uttered in a background that contains the proposition destroying the missing tapes was a crime. It is by supplementing the local context with further propositions already contained in the common ground that the presupposition of the main clause can be locally satisfied. Another example that requires the interplay of the common ground with the local context is the following by Karttunen: John called Mary a Republican and she insulted him back. The presupposition of the second conjunct must be satisfied by the prior context and it could be satisfied locally by the preceding clause provided that the information supplied by that clause is incremented by information already contained in the common ground, such as, for instance, the information that, in a certain walk of life, calling someone a Republican amounts to insulting him. Both examples illustrate the interplay of the local context with the greater context furnished by the common ground.
. Clitics, constraints on dialogue and clitic-left-dislocation Consider the following exchange: (14) A: Dov’ è andato Mario? B: Sandro sa che Mario è andato al cinema. (A: Where did Mario go? B: Sandro knows that Mario went to the cinema) (15) ? A: Dov’ è andato Mario? B: Sandro lo sa che Mario è andato al cinema. (A: Where did Mario go? B: Sandro it knows that Mario went to the cinema (lit.)) In both dialogues the knowledge ascription will be uttered for the purpose of answering a question requiring new information. However, we see that, whereas the cliticless construction can be used to convey the information contained by the embedded clause as new, the clitic construction cannot be used for that purpose (see
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.10 (156)
A. Capone
also Uriagereka 1995). Strictly speaking, it might be more correct to say that the utterance of the clitic construction, in a context where the information contained in the that-clause has to be construed as providing new information, is infelicitous; the explanation that can be provided for this inappropriateness must rest on the assumption that the clitic essentially indicates that the information contained in the that-clause is given in discourse. If this is so, then the clitic construction in combination with sa is predicted to be used felicitously in dialogic contexts where the knowledge ascription statement is a reply to an utterance requesting information about a person’s knowledge state. Hence the following dialogue, in contrast to the previous one, will be felicitous: (16) A: Lo sa Maria che Giovanni è a casa? B: Certo che Maria lo sa che Giovanni è a casa. (A: Does Maria it know that Giovanni is at home? B: Of course Maria it knows that Giovanni is at home (lit.)) So far we have noticed a correlation between discourse uses of the predicate know that presuppose the information contained in the embedded clause and the use of a clitic. The correlation is of such a systematic nature that it might be unreasonable to deny it, although one, faced with further data, might be inclined to make further distinctions and to look at the same type of phenomena with a more critical eye. In addition to the constructions we have investigated so far, we have those constructions which have been called clitic-left-dislocation constructions in the relevant syntactic literature (e.g. Che Maria è a Parigi lo sappiamo bene). Clitic-left-dislocation is a syntactico/semantic phenomenon of a number of pro-drop languages (the phenomenon is found in non-Romance languages as well as in Romance languages). A clause, in virtue of this phenomenon, is left-dislocated (adjoined to S’) and in its place a matching clitic antecedent is left. This object clitic is, clearly, a pronominal element.1 The phenomenon is not only syntactic but also pragmatic, since the leftdislocated clause is moved (according to Cinque 1990) to Top position, a position reserved for the topic of the sentence. It goes without saying that the information contained in this left-dislocated phrase must have a content which has a special status. It is clear that whatever the content of this clause should be, it must be described (in general terms) as being topical – as presenting information which has been previously vocalised in discourse, information that is presupposed. For the time being we leave aside whether the phrase must have been vocalized in adjacent discourse; in fact, it is possible that it has been vocalized in previous non-adjacent discourse or that it has been inferred from another more general statement (or a conjunction of statements) or that it is “in the air”, presupposed in the sense of belonging to the set of assumptions shared by the dyad
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.11 (157)
Dilemmas and excogitations
speaker/hearer. We may safely assume that a disjunction of these conditions must obtain for left-dislocation to occur. We could check this by an example such as (17): (17) A: Dov’ è andata Maria? B: (i) Giovanni sa che Maria è andata al cinema. (ii) Che Maria è andata al cinema Giovanni lo sa (translated in (18)). We must note that (ii) is not a suitable reply to A’s question, while (i) might be a suitable reply (in so far as the speaker uses Giovanni’s knowledge to infer the proposition Maria è andata al cinema). This shows that clitic-left-dislocation in Italian is also a discourse-related phenomenon and that, as a consequence of left-dislocation, the presupposition associated with the left-dislocated clause gets through. That clitics in these constructions are used for discourse-related purposes is an important observation. We might consider two hypotheses: (a) the presupposition is associated with the use of clitics (directly expressed by the use of clitics); (b) the presupposition is an outcome of left-dislocation and it just happens that left-dislocation in Italian involves clitics as obligatory elements. The first hypothesis, appealing though it might appear in the light of its coherence with what was said above about clitics in positive statements which do not exhibit left-dislocated structures, seems weak and some facts cast serious doubt on it. We must notice that topicalized constructions in English carry presuppositions. This could be checked by the same kind of test as above: (18) A: Where did Mary go? B: (i) John knows/has learned that Mary went to the cinema (ii) That Mary went to the cinema, John knows/has learned So it appears that topicalization (which involves a gap left by the moved phrase (as in That John is a nice fellow, Mary knows . . . (perfectly well)) is also associated with a strong presupposition (the speaker assumes the hearer knows that the embedded clause is true). This comparison apparently casts doubt on the hypothesis that clitics in Italian in the corresponding left-dislocation constructions are directly associated with the presupposition, although the option is left open that clitic-left-dislocation constructions are associated with presuppositions and (incidentally) involve clitics. In this kind of construction, it happens that the presence of the clitic is obligatory and this is certainly a syntactic fact. The facts about clitics in these particular constructions are determined by syntactic facts, rather than by discourse factors. Do these facts militate against the broad generalization that clitics in positive assertions containing verbs of propositional knowledge are associated with certain presuppositions or are they to be considered an array of independent facts? If the clitic lo is considered what it is – that is a pronominal element, which must be
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.12 (158)
A. Capone
free in its governing category and can either be connected with a long-distance antecedent within the sentence or with an element given in discourse, we can perhaps understand that the clitic can be assigned distinct uses. In clitic-left-dislocation, it is connected with an antecedent within the sentence. Whereas topicalization in English involves a gap – a trace – clitic-left-dislocation involves a clitic for reasons explained in Cinque (1990). The clitic matches the left-dislocated phrase and is coindexed with that phrase. As far as non-clitic-left-dislocated constructions are concerned, the clitic is not left in place of a moved element; therefore it must have a role which has to do: a) with the semantics of the sentence; b) with the discourse function of the sentence. It is in this kind of constructions that the clitic has a genuine discourse function. I believe that this distinction must be made. Of course, what all clitics have in common is anaphoricity. The clitic can refer back to an antecedent in the sentence (as in CLLD) or in discourse (as in the normal cases). I believe that a separate question arises. Consider the use of the clitic in positive assertions. The question might be asked whether the presence of the clitic can be accounted for distributionally or whether it is a matter of the semantics of the clitic that the sentence has the discourse function it has. I believe that this question is reasonable. In other words it might be reasonable to ask oneself whether it is just the case that, given a certain discourse slot, a clitic must cooccur with a verb of propositional attitude and it is a peripheral question whether the clitic means anything at all or whether the clitic has a meaning and, in virtue of this meaning, the discourse function is determined. The clitic seems to be required by a certain discourse distribution of sentences containing verbs of propositional attitude. The discourse slot determines the presence vs. absence of the clitic. Now it is immaterial whether the clitic directly indicates that the sentence has a particular discourse function; whether it is part of its job to indicate this or whether it is simply involved in a discourse process. Since the clitic is involved in a certain discourse distribution, it is reasonable to say that expressions exhibiting the clitic have a certain function, leaving open the question whether it is the clitic that indicates (or marks) this function or whether this function is a consequence of the distribution of clitic-expressions in discourse. I believe that if, ultimately, the hypothesis that it is a use of the clitic expression that carries certain presuppositions proves correct, the clitic simply correlating with certain discourse-distributions, we might still be able to say that a clitic-expression – unlike a cliticless expression – has certain properties in positive assertions. This modest statement is not totally insignificant. I believe it is now time to extend the data on which this article is based. I believe that what, so far, has been uncontroversially proven is that utterances containing clitics in combination with verbs of propositional attitude have a special discourse distribution, which is different from that displayed by the corresponding utterances without clitics. I feel that the more interesting claim that clitics are associated with
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.13 (159)
Dilemmas and excogitations
a presupposition might give rise to some objections, on the grounds that I have so far examined the presuppositions of factives and factives, after all, are associated with presuppositions regardless of the presence of a clitic. Now there is one way in which we could answer this hypothetical objection. We could consider verbs of propositional attitude which are not factive and we could show that, thanks to a clitic, the speaker commits herself to a presupposition. In the following, I would like to consider the verbs pensare (think), sospettare (suspect), immaginare (imagine), sperare (hope), prevedere (predict), aspettarsi (expect), and ricordare (remember). As usual, we can construct the pairs from which to extract the differences between the use and the absence of the clitic lo. Consider the following pairs: (19) a.
Sandro temette che Maria avrebbe detto la verità (Sandro feared Mary would tell the truth). b. Sandro lo temette che Maria avrebbe detto la verità (Sandro it feared that Mary would tell the truth (lit.)).
(20) a.
Sandro pensava che Maria avrebbe detto la verità (Sandro thought that Mary would tell the truth). b. Sandro lo pensava che Maria avrebbe detto la verità (Sandro it thought that Mary would tell the truth (lit.)).
(21) a.
Sandro sospettava che Maria avrebbe detto la verità (Sandro suspected that Mary would tell the truth). b. Sandro lo sospettava che Maria avrebbe detto la verità (Sandro it suspected that Mary would tell the truth (lit.)).
(22) a.
Sandro immaginava che Maria avrebbe detto la verità (Sandro imagined that Mary would tell the truth). b. Sandro lo immaginava che Maria avrebbe detto la verità (Sandro it imagined that Mary would tell the truth (lit.)).
(23) a.
Sandro sperava che Maria avrebbe detto la verità (Sandro hoped that Mary would tell the truth). b. Sandro lo sperava che Maria avrebbe detto la verità (Sandro it hoped that Mary would tell the truth (lit.)).
(24) a.
Sandro si aspettava che Maria avrebbe detto la verità (Sandro expected Mary to tell the truth). b. Sandro se lo aspettava che Maria avrebbe detto la verità (Sandro it expected Mary to tell the truth (lit.)).
(25) a.
Sandro ricordava che Maria aveva detto la verità (Sandro remembered that Mary had told the truth). b. Sandro lo ricordava che Maria aveva detto la verità (Sandro it remembered that Mary had told the truth (lit.)).
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.14 (160)
A. Capone
All the second members of the pairs above share the fact that they contrast with the prior members of the pairs in that they (the b members) seem to express presuppositions and require some anaphoric links with antecedents in discourse. In other words, the truth of the proposition embedded in the verb of propositional attitude is taken for granted. While the first members of the pairs, in fact, suggest (in Italian) that the speaker’s commitment is neutral as to whether the embedded proposition is true, the second members presuppose the embedded proposition. One further question may now be of some importance. As I have shown in Capone (2000), the use of the clitic lo seems to be able to exclude psychologised uses of verbs of propositional attitude in languages such as Italian; thus, we wonder if the modal function of the clitic depends on the fact that the clitic has a discourse function or if the modal and the discourse function of the clitic are two coexisting functions in the same word. It might be suggested, in fact, that to answer the question one might use the following criterion: one must think of contexts where the assumption expressed by the that-clause has been previously vocalised in discourse but where it is assumed that the that-clause is false. If, in this case, it is possible to use the clitic, then it can be shown that the discourse function is independent of the modal function. The criterion above seems quite good to me and may enable us to check in various languages whether the same considerations applicable to the Italian language are applicable there. This criterion seems to have consequences on the issue of whether the inference regarding the factual content of the embedded clause is a context-dependent linguistic phenomenon or is, rather, due to the semantics of the clitic. If the possibility that the clitic may be essentially used for discourse purposes and that the modal strengthening effect may only be a consequence of the discourse context in which the expression occurs (assuming that the expression is licensed in contexts where a presupposition is in the air) is accepted, then one consequence might be that the modal strengthening effect may not (merely) be of a semantic nature. It is extremely difficult to come to a decision about this issue as contexts are notoriously difficult things to deal with and it is always possible to find some contexts in which clitic utterances show different behaviours from the ones we have so far observed. One further difficulty presumably derives from the fact that if we consider factive verbs, these already carry presuppositions on their own. Thus, one might very well be led to investigate those verbs of propositional attitude which do not belong to the class of factives. Consider for example sperare (hope). Presumably one might hear conversations such as: (26) A: Speravo di essere un bravo linguista/ B: Anche io lo speravo che sarei diventato un bravo linguista (A: I hoped to become a good linguist/ B: I also hoped to become a good linguist).
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.15 (161)
Dilemmas and excogitations
(27) A: Tullio de Mauro sperava di diventare un bravo linguista/ B: Anche io lo speravo (Tullio de Mauro hoped to become a good linguist/ I also hoped that). (28) A: Sandro disse che Maria era andata a Parigi/ B: Anche Giovanni l’ha detto che Maria era andata a Parigi (A: Sandro said that Mary had gone to Paris/ B: John too said that Mary had gone to Paris). (29) A: Maria è andata al cinema/ B: Te lo avevo detto che era andata al cinema (A: Mary went to the cinema/ B: John too said that Mary had gone to Paris). (30) A: Tutti SANNO che Mario è un bravo ragazzo/ B: Anche Jim lo dice sempre (A: They all KNOW that Mario is a good fellow/ B: Jim always says that). (31) A: Maria ha detto la verità/ B: Sandro lo temeva che avrebbe detto la verità (A: Mary told the truth/ B: Sandro feared she would tell the truth). When we investigate such pairs, it appears to be uncontroversial that the proposition embedded in the reply must have been mentioned before in the conversation, given the anaphoric nature of the clitic. The factivity status of the embedded clause of the reply seems to be modally subordinated to the modal status of the proposition that is taken up through the clitic. Thus, modal subordination in the sense of Roberts (1989) makes it the case that in (26) speaker B does not express a full-blooded commitment to the embedded proposition. In (27) the utterance by A does not, of itself, express a strong commitment to the embedded clause, although one undestands through pragmatic presupposition that the speaker must believe that the embedded proposition is true. However, since this is just a pragmatic inference, it is not necessarily transferred through modal subordination to the embedded clause in B’s utterance. In example (28) A’s utterance uses a verb of saying and thus the proposition that Mary had gone to Paris need not be true, as far as the speaker knows. Thus, through modal subordination the proposition embedded in B’s reply need not be true. In example (29), through modal subordination, the proposition embedded in B’s reply is true. The same is true of examples (30) and (31). We are now prepared to answer the previous question: is the modal function independent of the discourse function (or to state the question differently, are we faced with two separate functions?). If it could be proven that the clitic directly contributes to factivity, then the two functions, obviously, could not be separated. As we have considered a number of contexts in which clitics can appear and we have now resorted to the modest assumption that clitics are indeed responsible for the modal status of the clause embedded in the verbs of propositional attitude (when the clitic precedes the verb in question) through modal subordination, we now
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.16 (162)
A. Capone
think that the modal and the discourse function of the clitic are the two sides of the same coin (although, presumably, they might represent two different functions). The clitic indubitably connects the current utterance to a prior discourse. If this does not actually take place in conversation, the discourse sequence is not wellformed. As a consequence of this, modal subordination in the sense of Roberts (1989) is established between the interpretation of an utterance and that of a prior utterance, in the same way in which modal subordination occurs with pronouns in English in examples such as Jim wants to build a pyramid. It will be the highest pyramid in the world. Thus, I would think that the modal and the discourse effect are inextricably connected. The modal function is a result of the anaphoric properties of the clitic lo.
. Clitics, speaker/hearer presuppositions and some surprising results for the projection problem So far we have seen that clitics are tied to presuppositions, at least with a number of verbs of propositional attitude and that they represent perhaps the most convincing proof that presuppositions can be expressed in terms of constraints on the contexts in which expressions containing them can appear. Following Stalnaker (1973, 1974), we accept that a presuppositional expression places some constraints on the context in which it can be uttered (e.g. John stopped beating his wife can be uttered only in a context which entails at least the speaker’s belief that John had been beating his wife). A presupposition, following this conception, can be considered an inference from a linguistic expression to a context or the formal pairing of a linguistic expression and a context. Even in the case of factives, where presumably the presupposition is tied to the factive verb, the clitic has a role to play, in that it allows the presupposition to survive in conditional contexts, as we shall see later. Furthermore, the clitic qualifies the presupposition of factives in the way we shall see below. It might be useful to distinguish between speaker presuppositions and speakerhearer presuppositions. I propose we should follow Grice (1989) in assuming that certain presuppositional expressions (e.g. factives) simply indicate the speaker’s beliefs prior to and during the utterance of a certain statement, while they carry the expectation that the hearer should add those beliefs to hers in case she did not previously entertain them. Other scholars, such as Chierchia and McConnell-Ginet (1990) have taken presuppositional expressions, such as regret, to constrain the common-ground (speakers’ and hearers’ beliefs). According to them, given that a presupposition is taken as an instruction for the hearers to add a certain belief to their own beliefs, the presuppositional expression makes sure that, after the utterance has been proffered, speaker and hearer presuppositions coincide.
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.17 (163)
Dilemmas and excogitations
Granting that this view is tenable, one also ought to consider a possible alternative class of presuppositional expressions and, namely, that class in which a reference is made to the speaker’s and the hearer’s presuppositions before (as well as during and after) the utterance of a linguistic expression. This class ought to be kept logically distinct from the class of presuppositional expressions which simply indicate the speaker’s presuppositions before the time of utterance (and which license accommodation). Now, it is time to claim that clitics such as lo exactly belong to that class of expressions that carry speaker/hearer presuppositions. Let us consider more closely the distinction between uttering John regrets that p and Giovanni lo sa che p (John it knows that p (lit.)). The former expression, as I have said previously, following essentially Grice’s (1989) considerations on presuppositions, carries the implication that the speaker already believes that p by the time the assertion has been uttered. There is no expectation that the hearer should already believe p before the time the assertion has been uttered, although she might be expected to add p to her beliefs from that moment on, if she does not (explicitly) refute the assertion. Furthermore, in a context in which this speaker’s presupposition is not satisfied (a context, that is, that does not entail the presupposition), the utterance will be infelicitous. Consider now the clitic expression Giovanni lo sa che p. This linguistic expression must be understood both with reference to the speaker’s presuppositions and to the hearer’s presuppositions. We have said previously, correctly I believe, that such clitic expressions presuppose vocalization of a certain proposition in the actual previous discourse. Vocalization of a proposition, however, on closer inspection, is just one of the conditions for the acceptability of such discourses. What happens is not just that the clitic refers to a previously vocalised thought but that modal subordination occurs and the clitic confers a certain modal status to the embedded proposition depending on the modal status it has in previous discourse. Thus, as the hearer’s attitude to the proposition is preserved through modal subordination, speaker/hearer’s presuppositions must coincide when the interpretative process comes to an end. It might be objected that, after all, factives may express speaker/hearer presuppositions through contextual enrichments, just in the way in which clitics do. Thus, one could have a dialogue such as the following one: A: Mary went to Paris / B: I know Mary went to Paris. Due to contextual enrichments, the speaker will presuppose p and that the hearer knows p. However, given that (presumably) no modal subordination is involved in the use of factives in English, there is no linguistic constraint that the speaker’s and the hearer’s presuppositions should coincide. It is a fact that they coincide but not a linguistic fact. Of course, it is an open question whether languages like English where a phenomenon analogous to modal subordination occurs by focusing on the factive verb actually exhibit a null (syntactically non-active) object pronoun, in which case modal subordination could occur through such a null pronoun.
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.18 (164)
A. Capone
A few remarks ought to be added with regard to the interplay of clitics and the semantics of complex expressions. If it is part of the semantic potentiality of clitics that they add presuppositions, then we would expect that the presence of clitics in complex expressions complicates the projection problem considerably. Let us start with Heim (1992)’s considerations on want. Heim considered sentences such as John wants to sell his cello. In fact, she thought that this complex sentence presupposes that John believes he has a cello and that the alleged presupposition that the speaker believes John owns a cello is only a conversational effect (possibly an implicature, that is something one suggests without wanting to commit oneself to (see Geurts 1998, for the opposite view)). If we translate this sentence into Italian, we have two possibilities: (32) a. Giovanni vuole vendere il suo violoncello. b. Giovanni lo vuole vendere il suo violoncello. Possibility (a) clearly conforms to all of Heim’s considerations in so far as it carries the presupposition that Giovanni believes he owns a cello and gives rise to a conversational implicature to the effect that the speaker believes he owns a cello. Possibility (b), instead, due to the additional presence of the clitic lo, does not conversationally implicate that the speaker commits himself to the existence of the cello but establishes a modal subordination to the assumptions of the previous context and, if in that context the existence of the cello was taken for granted, that must be the case for the speaker of (32b) as well. Here lo has the usual anaphoric properties we attributed to clitics, in that the expression cello must have appeared before (existentially) in the conversation. Of course, I must mention that there might be different views concerning the semantics of the verb want. One need not commit oneself to a presumed semantic ambiguity concerning the verb want, as one might endorse unitary semantics along the lines of Jaszczolt (1999). Jaszczolt might probably say about this example that the referential intention determines the preferred intepretation of the utterance, contrary to Heim (1992). Consider now the sentence It is possible that John knows that Mary is in Paris. The discussion of this type of sentences induces us to believe that the most embedded complement, in English, is implicated conversationally; so much so that van der Sandt (1992) has used these examples as evidence in favour of the defeasibility view of presuppositions. I am not persuaded that this is a real case of presupposition and the fact that the inference is defeasible points to conversational implicature. When we consider the equivalents of this sentence in Italian, we have, again, two possibilities: (33) a. E’ possibile che Giovanni sappia che Maria è a Parigi. b. E’ possibile che Giovanni lo sappia che Maria è a Parigi.
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.19 (165)
Dilemmas and excogitations
We make the usual prediction that possibility (a) commits the speaker to the fact that Mary is in Paris by conversational implicature (a default pragmatic inference), whereas in (b) the clitic establishes a modal subordination between the utterance and the previous context. Thus, if the context contains the presupposition that Mary is in Paris, the speaker of (33b) has to commit herself to the fact that Mary is in Paris presumably through anaphoric connection. Again we notice the usual anaphoric properties of the clitic expression that requires an antecedent sentence in order to be used felicitously. Let us now consider slightly more complicated statements, such as: (34) Either John knows that p or John does not know that p. Given the standard entailments of know, we assume that John knows that p and John does not know that p are incompatible, contradictory assertions. Karttunen (1974) and Soames (1979) have proposed the following projection rule for or: (C + (–φ)) satisfies ψ and (C + (–ψ)) satisfies φ. Thus, according to the standard theory, the disjunctive statement Either John knows that p or he does not know that p has no presuppositions to the effect that p2 (in other words, since the speaker is not sure whether John knows or does not know p, he cannot presuppose p). Consider now the equivalent in a language such as Italian: we could have possibility (35a), which has exactly the same projection properties as in English, or (35b): (35) a. O Giovanni sa che p o Giovanni non sa che p. b. O Giovanni lo sa che p o Giovanni non lo sa che p. It follows both from our intuitions about this sentence and from our general discussion of the clitic lo that possibility (b) is a correct statement. (b), in fact, presupposes that p is true if it follows a context in which p is accepted as true. The usual projection properties of disjunction are overridden by the anaphoric properties of the clitic. The projection rule, in fact, says that p is not a presupposition of the complex sentence. The clitic complicates the projection rule in that, when it is present, all we could say is that the sentence (35b) need not (but may) presuppose p. Consider now another possible combination *O Giovanni lo sa che p o Giovanni lo sa che non p (Either John knows it that p or he knows it that not p (lit.)). Given our considerations about the clitic lo we are in a position to predict correctly that this statement is ungrammatical, since the two clitics might, through modal subordination, semantically indicate respectively the presuppositions p and non p. When we conjoin the statement that clitics indicate linguistically presuppositions with the statement by van der Sandt (1992) that the conjunction of the proposi-
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.20 (166)
A. Capone
tions entailed by C and an assertion uttered in C maintains consistency, we obtain the reason for the ill-formedness of the previous statement. I think we should now consider verbs of saying, which according to the standard authorities on presuppositions are plugs to presuppositions (they impede a presupposition’s survival). Interestingly clitics allow the presuppositions of embedded clauses to survive, as in the example: (36) Giovanni ha detto che lo sa che Maria è a Parigi (John said that he knows it that Mary is in Paris (lit.)). If the clitic is omitted, then the sentence, like the English counterpart, does not presuppose the truth of the last embedded clause. But when the clitic is in place, the speaker can commit herself, through modal subordination, to the belief that Mary is in Paris and, furthermore, believe that the hearer shares this belief. Verbs of saying are plugs because we must accept that belief through indirect evidence is weaker than belief through direct evidence. Once a verb of saying is used, one conversationally implicates that one is not sure about the information imparted. Verbs of saying (apart from announce) share the feature that the presupposition of the most embedded clause is not inherited by the whole sentence. The clitic is an interesting device, because, among the things that can be said, there are some true and some false propositions as well as those propositions about which we suspend our judgement. When the clitic is absent, there is no implication that the proposition said is either true or false, but there is a strong implication that the speaker is suspending her judgement and is inviting the hearer to suspend his judgement too (presumably through a conversational effect). When the clitic is present, the speaker can, through modal subordination, commit herself to the truth of the furthest embedded proposition, presumably in so far as the clitic connects the utterance to a prior context in which the proposition said was considered as given. Interestingly, even under negation the presupposition can survive provided that a clitic is added as in (37): (37) Giovanni non ha detto che lo sa che p (John did not say that he it knows that p (lit.)). We shall now consider speech acts other than assertions. A directive such as: (38) Tell Mary that Jim is in Paris, normally presupposes that Jim is in Paris, although, since we are faced with a verb of saying, we are open to the possibility that the speaker is inducing the hearer to tell a lie. In Italian, granting that some coherence relation must render the presence of the clitic permissible, we have two possibilities
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.21 (167)
Dilemmas and excogitations
(39) a. Dì a Maria che Jim è a Parigi. b. Dillo a Maria che Jim è a Parigi. It is possibility (b) that presupposes without equivocation that Jim is in Paris has been vocalised in the context of utterance. Promises are those speech acts which, for their felicitous utterance, require that the promised proposition be not true yet. Thus we can predict that in the context of promises, the presence of the clitic cannot mark the embedded proposition as true, as in Te lo prometto che verrò al cinema. Here the semantic job of the clitic is to render the promise more emphatic. However, If I am reporting a promise the clitic can express the speaker’s commitment to the embedded proposition as in Lo promisi che sarei andato a Londra (I promised it that I would have gone to London) through anaphoric connection with a prior context. Let us now move on to conditionals. (40) If John knows that p, q. (40) is a complex sentence which does not presuppose p, although a part of it does, according to some views. The reason why the presupposition is cancelled or filtered out is, according to the standard views, that the sentence John knows that p is uttered as part of the protasis of a conditional, in other words it is part of a supposition. Even in English, I believe, the analysis of this sentence is somewhat problematic, since it can be uttered in two types of contexts: (a) a context in which the speaker and the hearer have no certainty regarding p; (b) a context in which both the speaker and the hearer presuppose p and recognize the other as so doing. In (b) the statement presupposes that p. In any case, we have one interpretative reading of the sentence, in which a statement can be made and no presupposition is recognized by the hearer as being part of the statement. This sentence can be translated into Italian in two ways, without or with the presence of a clitic. Again we have two possibilities: (41) a. Se Giovanni sa che p, siamo fortunati. b. Se Giovanni lo sa che p, siamo fortunati. In possibility (a) no presupposition that p is heard by the hearer or meant by the speaker in the default context, while in (b) a modal subordination is established between the utterance and a prior context and if this context contains the assumption p, the speaker of (41b) commits herself to the truth of p in that she treats p as having been part of previous discourse and given for the hearer too. Consider now Stalnaker’s (1974, 1999) example reported below: (42) If I discover later that I have not told the truth, I will confess it to everyone.
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.22 (168)
A. Capone
By a discussion of this example, it emerged that for pragmatic reasons the speaker is not heard as committing herself to the truth of the proposition I have not told the truth. In fact, if the speaker believed that she had not told the truth, then she would be unjustified and even misleading in saying If I discover later that I have not told the truth. The pragmatic unacceptability of committing oneself to such a presupposition in a similar context is reflected by the fact that no clitic can be used in the protasis of this example in the Italian equivalent of this sentence: (43) *Se lo scopro più tardi che non ho detto la verità, lo confesserò a tutti. If, by modal subordination, the clitic admits a factive interpretation of the embedded clause, the sentence is heard as conveying a pragmatically unacceptable utterance. Consider now another example by Stalnaker: Did you discover that p? Presumably this sentence might admit a use in which the speaker’s purpose is to find out whether p by inquiring into H’s knowledge state. This is the case in which no presupposition arises. There is another use of this sentence in which the speaker presupposes p and wants to find out whether the hearer has knowledge of p. This double use of the sentence is more evident in Italian, where possibility (a) requires no clitic, while possibility (b) requires a clitic, as in Lo hai scoperto che p? Consider now Levinson (1983)’s cases of presuppositions as in: (44) When I know that p, I will tell you so. (45) When I knew that p, I told him so. We notice that a possible strategy is to claim that when admits a presupposition, which is cancelled in certain contexts. However, in a language like Italian, we have to posit an interpretative ambiguity for quando (when), which does explain why in certain contexts the inference that the clause of time is true evaporates. The interpretative ambiguity is demonstrated by facts concerning the clitic, since we now predict that no clitic can occur in the opaque context (in that, by modal subordination, the speaker’s commitment to p will not be neutral, as it is in (46a)), while a clitic can occur in a transparent context, as is evidenced by the following sentences: (46) a. Quando saprò che p, te lo dirò. b. Quando lo seppi che p, glielo dissi. As usual, possibility (b) can imply by modal subordination that p is true and that the hearer knows that p.
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.23 (169)
Dilemmas and excogitations
. On some problems and some possible solutions My approach to clitics may cause a number of perplexities, which deserve to be examined so that various causes of anxiety can be dispelled, if possible. First, it might be argued that if factives are tied to presuppositions, it might be redundant to claim that, in fact, the same presuppositions are tied to clitics. Although this objection might prima facie be a serious one, it is not difficult to see that it can be surmounted. In fact, we have seen that clitics are not combined only with factives but, in general, with verbs of propositional attitude which do not generally admit presuppositions (sperare, pensare, aspettarsi, etc.). Even when these clitics are associated with factives (e.g. know), it is clear that their surprising effect lies in the fact that the presuppositions of clitics survive when the presuppositions of factives evaporate (e.g. in modal contexts such as the protasis of a conditional). Most importantly, the real difference lies in the fact that factives, without clitics, can admit (through a principle of charity) psychologised uses (which are non-presuppositional), whereas these uses are not allowed by the presence of clitics through modal subordination to a context in which the proposition embedded in the factive verb is taken for granted. Another crucial difference is that factives, without clitics, admit the speaker’s presuppositions, while clitics require a notion of speaker/hearer presupposition, which seems to me to be of some importance. Another possible objection is that I should have analysed clitics as analogous to the pronominal that in English. It might be argued that the effect of clitics is essentially the same as that of the pronominal that, as used in I knew that, I heard that, etc. That this argument does not seriously militate against my view is not difficult to demonstrate as I have so far granted the anaphoric properties of the clitic and I have conceded that the modal effects of the clitic may be due to a modal subordination to a prior context. Surely, what the linguist propounding this view would like to stress is that both that and the clitic lo require a notion of anaphoricity. They both refer back to something, presumably a proposition uttered earlier in previous discourse by one of the co-present speakers/participants to the conversation. There may be some doubt as to whether clitic doubled object constructions require the truth of a proposition embedded in a verb of propositional attitude due to semantics or anaphoric uptake. I ultimately believe that this issue cannot be easily decided and that there are data that militate in favour of one view or the other. Nevertheless, I must admit that the least committal view that only anaphoricity is involved is indeed plausible. There would be, in this case, a modal subordination in the sense of Roberts (1989). Now, if this view is taken, the clitic is no longer an expression having an inherent modal semantic content but it would have a modal semantic content by inheritance. In other words, given a context C* (F) (S1, S2, S3) and given a clitic expression S(CL), a modal dependence would be established between the prior context which is specified for factivity (F) and the subsequent
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.24 (170)
A. Capone
clitic assertion. As a result of this, an embedded proposition in S(CL) will depend for its modal indicator (the speaker’s commitment to p) on the feature F which was exhibited by the prior context. There might be some residual problems for the presuppositional view of clitics due to the fact that, apparently, in some constructions the clitic is not directly tied to a presupposition. For example, in sentences such as Io non lo so se Maria è a Parigi and Te lo prometto che andrò a Parigi, the clitic does not appear to be tied to a speaker/hearer presupposition. Now, as a linguist, I tend to take into account counterexamples but I do not exaggerate their weight. The example with se, for example, does not seem to me to contradict decisively what we have said, because it is the semantic job of se (if) to cast doubt on the embedded proposition and thus, by a principled combinatorial interaction between the clitic and the dubitative connective, one sees that the dubitative se prevails. Of course this is what we expect, because otherwise it would have been more sensible to use the complementizer that. Since this is not used and, instead, se (if) is being used, the reason for that must lie in the intention to consider the embedded proposition as one on which one casts some doubt. We must not lose track of the combinatorial semantic principles in our linguistic analyses and, thus, the apparent counterexample, when these combinatorial facts are analysed, is actually disarmed. Of course, it might be argued (but this argument might not be necessary) that even in this combinatorial dynamics the clitic plays a role and this role is to mark the embedded proposition as being (merely) vocalised in previous discourse. In other words, a speaker/hearer presupposition is still signalled thereby. The example Te lo prometto che andrò a Parigi might also militate against the presuppositional story. In fact, here the clitic signals only emphasis. We would be surprised, if it did not. In fact, in a different context, where factivity is not overridden by the semantics of the speech act promise, the presupposition persists, as in Lo promisi che sarei venuto (I it promised I would arrive (lit.)). When we analyse these two different sentences, where the clitic is, respectively, non-presuppositional and presuppositional, we cannot fail to notice the combinatorial effects of inserting the clitic in the speech act promise and, subsequently, of inserting the clitic in a report of a speech act. Promises, if considerations on speech act theory are followed, require some conditions. One of these is that the action promised should not have been carried out yet (it is useless to promise that p if I know that p has already been carried out). If these considerations make sense, we understand why the semantics of promises (in the direct speech act) overrides the semantics of the clitics. It could not be otherwise, if the speech act promise must have its effects. Of course, things change when the clitic is inserted in a speech act report, since here the speech act is only reported and does not come off due to the utterance of the sentence. The utterance, in other words, is not a promise and the semantics of promises does not override the semantics of the clitic. These combinatorial stories must be taken into
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.25 (171)
Dilemmas and excogitations
account and cannot be easily dismissed in favour of the gratuitous assertion that the theory is jeopardised by some fatal flaws. Before ending this section, I would like to dwell again on the projection problem, to avoid generating some unnecessary confusion. When I said that clitics complicate the projection problem, I did not mean that we should dispense with the current projection rules, which in my opinion are indeed tenable. I never intended, by adding the complications due to the clitics, to maintain that the projection rules are wrong or weak or inadequate. Neither is it necessary to amend the projection rules in favour of some new rules that can accommodate the semantics of clitics. First of all, it should be said that clitics such as lo exist in some languages and not in all languages. Thus, if some ad hoc projection rules were created, these would be language-dependent, whereas it would be highly desirable to have some languageindependent projection rules. As far as we know, the current projection rules are not tied to a particular language and are presumed to work for all languages, reflecting some pragmatic universals on the way information is processed and stored. To say that these rules are not universally applicable on the basis of Italian clitics is not what I would like to propose; thus, it might be preferable to say that the standard projection rules are merely applicable to all languages and that, given that semantic constraints always override pragmatic principles, in the specific case of Italian the semantics of clitics prevails over the pragmatics of the projection rules. Specifically, it might be assumed that all semantic facts are processed first and that they provide input to pragmatics and, specifically, to the projection rules. On closer reflection, clitics do not provide exceptions to the projection rules, but a semantic layer from which the projection rules have to take input. Nevertheless, the results I have hightlighted in this chapter remain surprising and also interesting, as a principled interaction between semantics and pragmatics has now emerged. The last problem I want to inquire into is whether the notion of speaker/hearer presupposition is still tenable, albeit we have conceded that the clitics introduce modal subordination to a prior context. As this is probably the most original part of this paper and, in any case, this is of uppermost importance to me, I want to defend the notion against a possible objection, which might take the following form: after all, modal subordination might involve in some cases no speaker’s commitment to a presupposition. I would like to contend that in the most unfavourable case, a minimal presupposition is involved, namely that a proposition has been vocalised in discourse. Furthermore, if modal subordination means that the speaker’s commitment to a proposition is equal to that previously expressed by the hearer (the previous speaker), then one consequence of this is that if the previous speaker exhibited a feeble commitment to p, the subsequent speaker will share that commitment. If the previous speaker exhibited a full commitment to p, the subsequent speaker will share that commitment. If the previous speaker exhibited a zero degree of commitment to p (say a neutral attitude), the subsequent speaker will exhibit
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.26 (172)
A. Capone
the same degree of commitment to p. But this amounts to saying that presuppositions are preserved in the ongoing discourse and that speaker’s and hearer’s presuppositions coincide to some extent. Thus a potential cause of anxiety has been disinfected, I hope. Summing up, I think that clitics offer us a special insight into the issue of presupposition and the results deriving from this insight are difficult to accommodate within the existing theories. Nevertheless, in the long term, I am persuaded that clitics promise to change substantially our view of presupposition. It is not possible to spell out within the scope of this essay all the possible consequences. I will simply point out that the issue of the clitic lo in Italian and other pro-drop languages exhibiting clitic-left-dislocation seems to lead in the direction of a binding theory of presupposition. However, this will not be the binding theory by van der Sandt (1992), about which I have not said anything, but, presumably, the binding theory by Asher and Lascarides (1998), which looks like a very fruitful framework and which I will deal with in a sequel to this article.
Acknowledgments This paper has a long story and it is right to acknowledge my indebtedness to a number of scholars who contributed to put me on the right track. I would like to express my gratitude to Robyn Carston, Paola Radici Colace, James Higginbotham, Yan Huang, Kasia Jaszczolt, Maria Vittoria Macrì, Jeanne Martinet, Sally McConnell-Ginet, Grazia Ortoleva, Sorin Stati, the late Professor Lino Falzon Santucci, Peter Strawson, Ken Turner and Edda Weigand. Needless to say, I take on full responsibility for any remaining errors. I would also like to give thanks to the head of the department of Philology and Linguistics of the University of Messina, Prof. Zumbo, for financial support.
Notes . This element, in so far as it does not obey subjacency and does not license parasitic gaps, cannot be considered as an overt spelling out of Wh-traces. . In that the presupposition of the first conjunct is satisfied.
References Asher, N., & Lascarides, A. (1998). The semantics and pragmatics of presupposition. Journal of Semantics, 15 (3), 215–238. Capone, A. (2000). Dilemmas and excogitations: considerations on modality, clitics and discourse. Lingua e Stile, XXXV (3), 447–470.
TSL[v.20020404] Prn:17/10/2002; 13:17
F: PBNS09.tex / p.27 (173)
Dilemmas and excogitations
Chierchia, G., & McConnell-Ginet, S. (1990). Meaning and grammar. Cambridge, MA: The MIT Press. Cinque, G. (1990). Types of A-dependencies. Cambridge, MA: The MIT Press. Davies, S. (Ed.). (1991). Pragmatics. Oxford: OUP. Gabbay, D., & Guenthner, F. (Eds.). (1989). Handbook of Philosophical Logic, Vol. IV. Dordrecht: Reidel. Gazdar, G. (1979). Pragmatics: Implicature, presupposition and logical form. New York: Academic Press. Geurts, B. (1998). Presuppositions and anaphors in attitude contexts. Linguistics and Philosophy, 21 (6), 545–601. Grice, H. P. (1989). Studies in the way of words. Cambridge, MA: Harvard University Press. Heim, I. (1992). Presupposition projection and the semantics of attitude verbs. Journal of Semantics, 9, 183–221. Jaszczolt, K. M. (1999). Discourse, beliefs and intentions. Oxford: Elsevier. Karttunen, L. (1974). Presupposition and linguistic context. In S. Davis (Ed.), Pragmatics (pp. 406–415). Oxford: OUP. Levinson, S. C. (1983). Pragmatics. Cambridge: CUP. Roberts, C. (1989). Modal subordination and pronominal anaphora in discourse. Linguistics and Philosophy, 12, 683–721. Soames, S. (1979). A projection problem for speaker presuppositions. Linguistic Inquiry, 10, 623–666. Soames, S. (1982). How presuppositions are inherited: a solution to the projection problem. In S. Davies (Ed.), Pragmatics (pp. 428–470). Oxford: OUP. Soames, S. (1989). Presupposition. In D. Gabbay & F. Guenthner (Eds.), Handbook of Philosophical Logic, Vol. IV (pp. 533–616). Dordrecht: Reidel. Stalnaker, R. C. (1973). Presuppositions. Journal of Philosophical Logic, 2, 447–457. Stalnaker, R. C. (1974). Pragmatic presuppositions. In S. Davis (Ed.) (1991), Pragmatics (pp. 471–482). Oxford: OUP. Stalnaker, R. C. (1999). Context and Content. Oxford: OUP. Uriagereka, J. (1995). Aspects of the syntax of clitic placement in western romance. Linguistic Inquiry, 26 (1), 79–123. van der Sandt, R. A. (1992). Presupposition projection as anaphora resolution. Journal of Semantics, 9, 333–377.
TSL[v.20020404] Prn:26/08/2002; 8:44
P IV
Evidentiality
F: PBNSP4.tex / p.1 (175)
TSL[v.20020404] Prn:17/10/2002; 14:35
F: PBNS10.tex / p.1 (177)
Inferred evidence* Language-specific properties and universal constraints Sergei Tatevosov Moscow State University, Russia
.
Introduction
This paper concerns evidentiality, a category marking the source of information available to the speaker (see Givón 1982; Chafe & Nichols 1986; Willett 1988; Guentchéva 1996, to appear; Johanson & Utas 2000; Kozintseva, to appear). Crosslinguistically, three main evidential meanings are distinguished: ‘direct evidence’ (‘the situation in question is directly perceived by the speaker’), ‘inferred evidence’ (‘knowledge about a situation obtains via inference’), and ‘reported evidence’ (‘knowledge is acquired via a verbal report from some external source’). In what follows, I am going to examine in detail one of these meanings, namely, a meaning of inferred evidence illustrated by the famous ‘Kemal’s coat’ example from Turkish (Slobin & Aksu 1982: 187): (1) C(ontext): The speaker comes home in the evening. In the cloakroom he finds his friend Kemal’s coat. The speaker says: Kemal gel-mi¸s. Kemal come-infer {I see,} Kemal has come.1 In (1), the speaker does not see Kemal, nor the event ‘Kemal came’; nevertheless, Kemal’s coat in the cloakroom tells the speaker that this event has occurred. Observing the coat, the speaker infers that the proposition ‘Kemal came’ is true, and refers to it using a verbal form in -mi¸s, for which one of the possible readings is the inferential one. Grammatical categories marking inferred evidence (possibly, together with some other meaning(s)) fall into two major types: experiential and non-experiential
TSL[v.20020404] Prn:17/10/2002; 14:35
F: PBNS10.tex / p.2 (178)
Sergei Tatevosov
(the term ‘experiential inference’ is coined by Anderson 1986: 284). The latter are used if knowledge about a situation is inferred only by means of logic, reasoning, intuition, and other mental constructs. The former are only appropriate for asserting a proposition p if there is some q, observed by the speaker at the moment of speech, which enables the speaker to conclude that p took place. Thus, mi¸s-forms in Turkish exemplified in (1) are an example of experiential inferential evidentials. Given this general background, in what follows I will discuss a few problems related to experiential inferentials. First, what kind of evidence should be available at the moment of speech to enable the speaker to use experiential inferentials for asserting an antecedent situation? Second, what are semantic and pragmatic restrictions on the use of experiential inferentials? Third, I will address a more general issue: how do answers to the first two questions influence our understanding of universal restrictions on cross-linguistic variation associated with epistemic categories such as evidentiality and epistemic modality? My data come from three genetically unrelated languages: Bagwalal (NorthCaucasian, Nakh-Daghestanian, Andic), Tatar (Altaic, Turkic), and Mari (Uralic, Finno-Ugric ). In all the three languages there are general markers of indirect evidence expressing both ‘reported’ and ‘inferred’ evidence. Reportive uses of these markers stay in the main beyond the scope of this paper, so below only inferential uses are examined in detail. For further reference on Bagwalal, Tatar, and Mari see Kibrik 2001; Zakijev 1993; and Ivanov and Tuzharov 1970 respectively.
. Inferential evidentials: Previous approaches Inferentiality has been dealt with in a variety of studies, both language-specific and cross-linguistic. Here follow a few quotations that reflect generally accepted views of what experiential inferentials are: inferences must be based on any kind of sensory evidence of resultant state, with the provision that no aspect of the antecedent process has been present in the speaker’s consciousness. (Aksu-Koç & Slobin 1986: 160) When a speaker sees result of some action, s/he may use it as evidence to infer what the action was that produced the observed state of affairs. (Willett 1988: 61) Le locuteur emploie
. . . pour reconstruire, au moyen d’un raisonnement s’appuyant sur des indices sensoriels observables au moment de l’énonciation, les événements antérieurs auxquels il n’pas assisté . . . (Meydan 1996: 134) C’est en observant des résultats qu’on infère les événements qui les ont causés (Lazard 1996: 28)
TSL[v.20020404] Prn:17/10/2002; 14:35
F: PBNS10.tex / p.3 (179)
Inferred evidence
The Inferential is used when the speaker is involved him- or herself with the evidence to a certain degree. The speaker makes deductions on the basis of evidence. This evidence has been collected by the speaker. (de Haan 1998: 7) . . . the evidential meaning of an inference . . . indicates that a past action is known or inferred on the basis of a current state. (Bybee et al. 1994: 96) The speaker has not perceived P directly, . . . the speaker has (directly) observed another situation which he interprets as pointing towards P. (Plungian 2001: 352)
Although the general idea underlying the descriptions of inferentiality cited above is quite straightforward – ‘there is something in the actual world that tells the speaker that the asserted proposition is true’ – it is extremely difficult if at all possible to deduce predictions about the distribution of inferentials from the notions like ‘results’, ‘indices sensoriels’, ‘a situation pointing towards p’, etc. For example, to predict if an inferential evidential is allowed in a certain inferential context, the theory relying on the concept of ‘results’ requires a procedure of distinguishing between what does count as a result of a given situation and what does not. But even if such a procedure is supplied, it is still unclear if any result of a situation can serve as a basis for inference. Going back to ‘Kemal’s coat example’, if Kemal’s coat in the cloakroom provides enough evidence that Kemal is here, is the same true for a pack of Kemal’s favorite cigarettes found in the cloakroom or for Kemal’s car in front of the speaker’s house? If not every result of a situation favors the use of inferentials, what are those results on which inference can not be based? The same questions arise when we are dealing with the other notions mentioned above. Therefore, we have to find more precise generalizations accounting for the distribution of inferentials and a more accurate theory producing predictions that can be either falsified or verified.
. Inferentiality in Bagwalal, Mari, and Tatar . Basic uses of evidentials Partial evidential paradigms of Bagwalal, Tatar, and Mari are given in Tables 1– 3. In each language, there is a series of indirect evidence forms contrasting with unmarked forms. In both Bagwalal and Tatar, indirect evidence forms originate from perfects. Development of this type is regularly attested across languages (see Dahl 1985: 149–153; Bybee et al. 1994: 95–97; Guentchéva 1996, Part 1; and Tatevosov 2001); frequently cited examples are Bulgarian, Turkish, Georgian, and Ar-
TSL[v.20020404] Prn:17/10/2002; 14:35
F: PBNS10.tex / p.4 (180)
Sergei Tatevosov
Table 1. Bagwalal: Partial paradigm of the verb hecÁi ‘stand up, rise’ Title
Preterite series (unmarked)
Perfect series (indirect evidence)
Past | Perfect
Past hecÁi hecÁi-b-o hecÁ-ir¯a-X hecÁ-ir-¯o-b hecÁ-¯a-ł-o-b
Perfect hecÁi-b-o hecÁi-b-o hecÁ-ir¯a-X hecÁ-ir-¯o-b hecÁ-¯a-ł-o-b
Pluperfect Imperfect Past Habitual Past Future
b-ukÁa b-ukÁa b-ukÁa b-ukÁa
ekÁ˚ a b-ukÁa-b-o b-ukÁa-b-o b-ukÁa-b-o b-ukÁa-b-o
ekÁ˚ a ekÁ˚ a ekÁ˚ a ekÁ˚ a
Table 2. Tatar: Partial paradigm of the verb bara ‘go’; 3rd person singular Preterite series (unmarked)
Perfect series (indirect evidence)
Preterite Pluperfect
bar-d@ bar-gan
i-de
Imperfect
bar-a
i-de
Past Future
bar-ajaˇcak
i-de
Perfect Past Mirative Present Mirative Prospective Mirative
bar-gan bar-gan
i-kän
bar-a
i-kän
bar-ajaˇcak
i-kän
Table 3. Mari: Partial paradigm of the verb užaš ‘see’; 3rd person singular Synthetic (unmarked) Old Preterite New Preterite Present
už-o už-6n už-eš
Periphrastic (indirect evidence) – Past Unwitnessed Present Unwitnessed
už-6n už-eš
ul-maš ul-maš
menian. The Perfect in Tatar is expressed by a bound morpheme, and in Bagwalal it is formed from the perfective converb with the auxiliary. In both languages, then, there are periphrastic forms consisting of a non-finite category – a participle or a converb – plus the perfect of the auxiliary. In Mari, indirect evidence forms contain a special form of the auxiliary ulmaš. The diachronic source of the morpheme -maš is unclear. The range of interpretations associated with indirect evidence categories in these three languages is virtually the same. (2)–(4) show that they may mark both inferred and reported evidence: linking Context1 to Translation1 produces an inferential reading, linking Context2 to Translation2 produces a reportive reading. Consider, for example, (2) from Bagwalal: (2) C1 : The speaker watches Mohammed walk quickly down the street. Suddenly Musa comes round the corner. The speaker knows that Musa has
TSL[v.20020404] Prn:17/10/2002; 14:35
F: PBNS10.tex / p.5 (181)
Inferred evidence
promised to beat up Mohammed. As soon as Musa appears, Mohammed runs away. The speaker: C2 : Musa robbed the post office but was arrested later. The hearer wonders how the police found out who the thief was. The speaker: BAGWALAL maHammad-i-ba musa hÖ$-w-o ekÁ˚ a. Mohammed-obl-aff Musa see-m-conv aux:prs 1. {I see} Mohammed saw Musa. 2. {I have heard} Mohammed saw Musa {breaking the window of the post office}. Under the context C1 , the source of information about the situation ‘Mohammed saw Musa’ is a consequent situation ‘As soon as Musa appeared, Mohammed ran away’. Under the context C2 , the speaker relies on second-hand information obtained from the other person. Similar interpretations are allowed for (3) and (4): (3) C1 : The speaker notices that Daut has changed: he doesn’t eat, doesn’t drink, listens to nobody, and follows Alsu with his eyes all the time. C2 : X tells the speaker that Daut fell in love with Alsu. The speaker tells Y: daut alsu-n@ jarat-kan. TATAR Daut Alsu-acc love-pfct 1. {I see} Daut fell in love with Alsu. 2. {I was told} Daut fell in love with Alsu. (4) C1 : The speaker sees ashes in place of the hearer’s house. C2 : X was told that the hearer’s house burnt. t6j-6n pört-et jül-en ul-maš. 2sg-gen house-2sg burn-pst:3sg aux-evid 1. {I see} your house has burnt. {I am so sorry!} 2. {I was told} your house has burnt. {I am so sorry!}
MARI
Let us now consider inferential interpretation in more detail. First of all, all the three categories are experiential inferential. Consider (5): (5) C1 : The speaker comes home in the evening. In the cloakroom he finds his friend’s coat (‘Kemal’s coat example’). *C2 : X: Do you think X is at home already? The speaker: It’s seven o’clock already, the working hours are over. a. maHammad w-¯a-w-o ekÁ˚ a. BAGWALAL Mohammed m-come-m-conv aux:prs b. zehrä kil-gän. TATAR Zuhra come-pfct
TSL[v.20020404] Prn:17/10/2002; 14:35
F: PBNS10.tex / p.6 (182)
Sergei Tatevosov
c.
maša tol-6n ul-maš. Masha come-pst:3sg aux-evid 1. {I see} Mohammed | Zuhra | Masha has come. 2. *Mohammed | Zuhra | Masha must have come.
MARI
C1 is again a ‘Kemal’s coat’ example. The speaker saying that a person in question has come relies on the observable situation ‘his/her coat is here’, and the sentence is perfectly appropriate. But under C2 the statement is only based on a general assumption, such as ‘After one’s working hours are over, one normally comes home; Mohammed’s/Zuhra’s/Masha’s working hours are over, so he/she must have come’. In this case the sentence is totally unacceptable.
. Inference and resultant state of a situation Many linguists have claimed that when experiential inferential interpretation obtains, inference is based on results of a situation (cf. some of the quotations above, in particular, one from Willett 1988: 61; see also Woodbury 1986: 193). In fact, there are dozens of examples in which this type of inference is attested. Consider (6): (6) a.
C: The speaker sees a hat on Mohammed’s head. maHammad-i-r butuna e¯sˇa-m-o ekÁ˚ a. BAGWALAL Mohammed-obl-erg hat put.on-n-conv aux:prs {I see} Mohammed put on a hat. b. C: The speaker looks for an ax for some time and finally sees Daut who holds the ax. TATAR daut balta-n@ al-gan. Daut ax-acc take-pfct {I see} Daut took the ax! c. C: The speaker comes to Masha and finds out that the floor in her house is clean. MARI maša küvar-6m üšt-6n ul-maš. Masha floor-acc sweep-pst:3sg aux-evid {I see} Masha cleaned up the floor.
In each of these sentences, the speaker observes a result of a situation in a strict sense, namely, a resultant state of a situation. In (6a), presence of a hat on one’s head is a resultant state of a situation ‘put on a hat’; ‘hold something’ in (6b) is a resultant state of ‘take something’, and ‘the floor is clean’ in (6c) is a resultant state of ‘clean up the floor’. Having observed a resultant state, the speaker, who did not witness a preceding situation, acquires enough evidence to assert that this situation had occurred in the actual world.
TSL[v.20020404] Prn:17/10/2002; 14:35
F: PBNS10.tex / p.7 (183)
Inferred evidence
But it turns out that no implicational relation holds between observing a resultant state and using inferentials. Direct experience of the resultant state is not obligatory for enabling the speaker to infer the truth of the proposition. Consider (7): (7) C1 : The speaker meets his friend, a plowman, who is smeared with soil and diesel oil, and looks tired. *C2 : The speaker sees the field just ploughed and tries to guess who could have done it. a. maHammad-i-r h˜u¯sˇa b-eLÁi-b-o ekÁ˚ a. BAGWALAL Mohammed-obl-erg field n-plough-n-conv aux:prs b. daut k@r-n@ sukala-gan. TATAR Daut field-acc plough-pfct c. j6van pakˇcaž-6m kural-6n ul-maš. MARI Ivan field-acc plough-pst:3sg aux-evid {I see} Mohammed | Daut | Ivan ploughed the field. The situation described in C1 is not a resultant state of a situation ‘A plowman has ploughed the field’. Indeed, the resultant state is ‘the field is ploughed’, but not ‘the plowman is dirty’. (7) does not even imply that the resultant state is attained, that is, that the whole field is ploughed. It merely indicates that the plowman has been involved in ploughing activity for a certain time. Nevertheless, (7) in which inferential forms occur is perfectly acceptable under C1 . The same is true of the sentences (2), (4), and (5) above: here observed situations are ‘Mohammed saw Musa and ran away’, ‘Daut’s behavior changed in a certain way’, and ‘The speaker found his friend’s coat in the cloakroom’. None of these situations is a resultant state of a corresponding asserted situation, but inferential forms are not ruled out. Lack of the precise correspondence between observing a resultant state and possibility to infer an antecedent situation has been recently discussed in Nikolaeva (1999: 138–139) on Khanty material. Citing Nedyalkov 1988, Nikolaeva subsumes differences similar to those between C1 and C2 in (7) under the notion of trivial vs. non-trivial results. Trivial results are “components of lexicographic description of a verb (mostly, telic verbs like open, come, put, stand up, etc.) . . . Thus, a trivial result is unique for each event and predictable from it”. The notion of ‘trivial result’ is therefore precisely equivalent to the notion of a resultant state used throughout this paper. Non-trivial results are context-dependent and not included in the meaning of the verb. Non-trivial results can be associated with any verbs, either telic or atelic, and they do not necessarily involve the same set of participants as corresponding situations. Nikolaeva claims, then, that Khanty inferential evidentials are not restricted to the inference from trivial results, but also involve non-trivial results. Nikolaeva’s data together with those discussed above strongly suggests that
TSL[v.20020404] Prn:17/10/2002; 14:35
F: PBNS10.tex / p.8 (184)
Sergei Tatevosov
the concept of resultant state (= trivial result in Nedyalkov/Nikolaeva’s terms) is not suitable for drawing generalizations about the distribution of inferentials. On the other hand, perceiving a resultant state is not a sufficient condition for the use of inferential evidentials either. In (7) under C2 , the situation ‘The field is ploughed’ is a resultant state of the situation ‘Mohammed/Daut/Ivan ploughed the field’, yet the sentence cannot feed into this context. Therefore, a theory relying on the notion of resultant state only accounts for a restricted set of data. It fails to predict the distribution of inferentials in all cases, so we have to find a more comprehensive account for cases like (7).
. Semantico-pragmatic restrictions on the use of experiential inferential markers Let us consider (8) which is only appropriate under the context C2 . This appropriateness is not surprising, as ‘be dead’ is a resultant state of ‘be killed’: (8) ??C1 : The speaker watches the battle. The commander of those who attack falls to the ground. C2 : The speaker approaches the commander, takes his pulse and realizes that he is dead. a. kÁamandir kÁ˚ a¯-w-o ekÁ˚ a. BAGWALAL commander kill-m-conv aux:prs b. kamand@r-n@ üter-gän. TATAR commander-acc kill-pfct c. komandir-6m pušt-6n-6t ul-maš. MARI commander-acc kill-pst-3pl aux-evid {I see} The commander has been killed. The inappropriateness under C1 should be explained, however. One of the native speakers of Bagwalal commented on (8a) in the following way: ‘how can I use the form kÁ˚ a¯ wo ekÁ˚ a,’ he said, ‘I am not sure the commander has been killed, maybe he has been wounded. If I were sure he is dead, of course I could have used kÁ˚ a¯ wo ekÁ˚ a’. The speaker’s intuition seems to be quite revealing here. It indicates that not only should the observed situation point towards the unwitnessed preceding situation; also, it should not point towards any other situation. Translating this into a more formal language results in the following constraint. (9) Probability constraint (PC). If the speaker is entitled – for whatever reason – to assume that situation p is more likely to produce the observed state of affairs q than any other situation, he can use the experiential inferential forms to assert p. If – for whatever reason – the speaker believes
TSL[v.20020404] Prn:17/10/2002; 14:35
F: PBNS10.tex / p.9 (185)
Inferred evidence
that q may be caused by pi as well as by pj , he cannot use the experiential inferential forms to assert either pi or pj . This constraint accounts for the inappropriateness of (8) under C1 as well as for the inappropriateness of (7) under C2 . Indeed, in (8) the situation ‘the commander fell to the ground’ may be caused by a few different events: ‘the commander is killed’, ‘the commander is wounded’, ‘the commander broke his leg’, etc. The speaker has no a priori reasons to evaluate one of them as more likely than others. For the same reason, in (7) it is impossible to unambiguously identify an agent. Given the observed state of affairs, the speaker has a choice as to who ploughed the field. It may be X, Y, Z; it may also be Mohammed, but there is no evidence that Mohammed’s participation in the situation is more likely than X’s, Y’s or Z’s. The statement ‘Mohammed/Daut/Ivan ploughed the field’ is only a hypothesis, by no means more reliable than any other. Accordingly, it is ruled out by the PC. Note that the PC does not involve any reference to the notion of resultant state and thus correctly allows all the appropriate sentences discussed so far, including those in which inference is based on non-trivial results in Nedyalkov/Nikolaeva’s sense. In particular, the PC allows for (7) under C1 , as for a plowman to be smeared with soil and machine oil unambiguously implies that he has been working for some time, i.e. that he has been ploughing a field. The PC is not the whole story about experiential inferentials, however. Consider (10): (10) C1 : The speaker meets X who is cutting a bear/elk. (He knows that X went hunting alone.) *C2 : X, Y, and Z went hunting, and the speaker is aware of it. Later the speaker meets X who is cutting a bear|elk. (The speaker knows that all the hunters are expert shots.) ??C : X, Y, and Z went hunting, and the speaker is aware of it. Later the 3 speaker meets X who is cutting a bear|elk. (The speaker knows that X is the best hunter in the village while the other two went hunting for the first time.) maHammad-i-r sÖI kÁ˚ a¯-b-o ekÁ˚ a. BAGWALAL Mohammed-obl-erg bear kill-n-conv aux:prs daut poš@j-n@ üter-gän. TATAR Daut elk-acc kill-pfct j6van maska-m pušt-6n ul-maš. MARI Ivan bear-acc kill-pst:3sg aux-evid {I see} Mohammed | Daut | Ivan killed a bear | elk.
TSL[v.20020404] Prn:17/10/2002; 14:35
F: PBNS10.tex / p.10 (186)
Sergei Tatevosov
Under C1 , the sentence is appropriate: in fact, it does not violate the PC. Under C2 , as the PC predicts, (10) is inappropriate: here there is no obvious way to determine which of three possible candidates may have killed a bear/elk, and the statement ‘Mohammed/Daut/Ivan have killed a bear/elk’ is not more reliable than corresponding alternative statements about two other hunters. But under C3 , inappropriateness is not predicted by the PC, as only one of the three hunters is likely to kill the bear/elk. Commenting on C3 , native speakers have pointed out that there is not enough evidence to say it was Mohammed/Daut/Ivan who killed a bear/elk in spite of the fact that it is definitely the most probable option. The same restriction obtains in (7) (repeated here as (11)) under C4 : (11)=(7)C3 : The speaker knows that in the morning X started ploughing his field. In the evening, he comes to see if the job is done and finds the field ploughed. ??C : The speaker sees the field just ploughed and tries to guess who could 4 have done it. There is a team of plowmen consisting of X, Y and Z. The speaker: I know Z is gone; Y is sick, so {I see} X has ploughed the field. Here again, the context C4 is not compatible with the assertion performed with the experiential inferential form although X’s participation in the situation has the highest probability to cause the observed situation ‘the field is ploughed’. In contrast, C3 , which allows no choice as to who could have ploughed the field, is appropriate. I assume that inappropriateness of (10) under C2 and of (11) under C4 can be attributed to the Recoverability Constraint, that must be satisfied simultaneously with the PC: (12) Recoverability constraint (RC). Given the sensory evidence q, obtaining at the moment of speech, the speaker can assert the proposition p if the whole set of arguments of p is fully identifiable through q and conventionalized knowledge of the world. Given C3 in (11), the speaker observes a resultant state of the situation. Although observing the resultant state is enough to infer that the proposition ‘Somebody has ploughed the field’ is true, it is insufficient to identify the Agent. But the speaker also knows that in the past the situation ‘Daut is ploughing the field’ occurred. In addition, he can access the following conventionalized knowledge: (13) Participants identity principle: If at the moment ti a situation with a certain set of participants is at one of its intermediate phases, then at the
TSL[v.20020404] Prn:17/10/2002; 14:35
F: PBNS10.tex / p.11 (187)
Inferred evidence
moment tj , following ti , this situation is associated with the same set of participants unless otherwise is overtly indicated. This principle normally allows to identify participants of an unwitnessed situation. Thus, if there is evidence that at some moment in the past the proposition ‘X is ploughing the field’ is true, the speaker is entitled to believe that the same is true at any subsequent moment, unless he encounters an overt indication that some alternative participant entered into the process. This further entails that it is X (but not Y, Z, etc.) who has accomplished ploughing the field. Therefore, under the context C3 , (11) implies, according to (12) and (13), that the Agent of the ploughing event was Mohammed/Daut/Ivan, but not any other individual. Given the RC, the inappropriateness of (11) under C4 is not unexpected. As one of the native speakers commented on this example, ‘I can not be aware of the fact it was Daut who ploughed the field, it is only my suggestion’. Although the speaker observes the resultant state ‘the field is ploughed’, he has a choice as to which situation produced this result – ‘X ploughed the field’, ‘Y ploughed the field’, ‘Z ploughed the field’. The participant identity principle does not help to choose one of the alternatives, as the speaker has no premonitory awareness of who was involved in ploughing activity, and this is the main difference between C3 and C4 . It may appear that choice between three plowmen is predetermined by knowledge that Y is out and Z is ill so it must be X who has ploughed the field. However, here we are not dealing with conventionalized, but with episodic, occasional relation between two situations (‘The field is ploughed’ ⇐ ‘X has ploughed the field’). This knowledge is completely determined by the current state of the world, while from the point of view of conventionalized knowledge about possible relations between situations all the three alternatives are possible, and the Agent is thus not recoverable. Reasoning in a similar fashion, one can account for the inappropriateness of (10) under C3 . The participants identity principle restricts a set of those who could kill a bear/elk just to three persons who went hunting, but provides no clues for choosing one of them. Moreover, we know that even a person who went hunting for the first time may hit a target whereas the best hunter in the village may miss. Therefore, from the point of view of conventionalized permanent knowledge, none of the three participants can be excluded, and that is why the context C4 in (10) provides insufficient evidence for asserting the proposition. Note that the RC does not imply that the agent NP is definite; it only requires that an NP refers exactly to those individuals that can be identified on the basis of the observed state of affairs and conventionalized knowledge. In (11) under C4 , as well as in (7) under C2 , no definite NP can be used, but using a specific indefinite pronoun makes the sentence perfectly acceptable (only the Bagwalal example is given here; examples from Tatar and Mari are similar):
TSL[v.20020404] Prn:17/10/2002; 14:35
F: PBNS10.tex / p.12 (188)
Sergei Tatevosov
(14) C2 of (7): The speaker sees the field just ploughed and tries to guess who could have done it. ¯ło-r-w-ukÁa-w-ola h˜u¯sˇa b-eLÁi-b-o ekÁ˚ a. BAGWALAL who-erg-m-indef-m field n-plough-n-conv aux:prs {I see} Somebody ploughed the field. Here the agent NP refers exactly to the set of participants that is identifiable via the situation ‘the field is ploughed’ provided that no additional evidence is available to the speaker, and the RC is now satisfied. Typically, the PC and RC are either both satisfied or not. For example, in (6a) above the speaker sees Mohammed with the hat on his head, and there is no evidence that it is not Mohammed but somebody else who put the hat on Mohammed’s head. So Mohammed is unambiguously identified as the agent of a situation in question, and the RC is satisfied. Trivially, ‘Mohammed put on a hat’ is also the most probable cause for ‘The hat is on Mohammed’s head’, so the PC is satisfied as well. The same is true of (2)–(4), (5) under C1 , (6), (7) under C1 , (8) under C2 , and (10) under C1 . On the other hand, in (7) under C2 none of the constraints is satisfied. The agent of the situation ‘X ploughed the field’ is not recognizable through the resultant state only, and no additional evidence is accessible for the speaker. The most probable cause of the observed state of affairs cannot be inferred either. As a result, (7) is inappropriate under C2 . Nevertheless, being logically independent, the PC and RC are not extensionally equivalent. As has been discussed above, in (10) under C2 the PC is satisfied, but the RC is not. Right the other way round, in (8) under C1 the RC is only satisfied. In fact, there is conventionalized knowledge that if a person is killed, he is not able to retain a standing position, but this is also true, for example, of ‘be wounded’, and, as we have already mentioned, ‘be wounded’ is not less probable than ‘be killed’. Notions on which the PC and RC are based are essentially pragmatic: the PC relies on the epistemic evaluation of the probability of a situation, and the RC involves reference to the speaker’s conventionalized knowledge. Values of pragmatic variables associated with these notions, are, of course, subject to intra-linguistic variation. For example, replacing X’s coat in ‘Kemal’s coat’ example (see (5) under C1 ) with X’s car standing in front of the speaker’s house will yield different results depending on what is known about X and his car. If X always uses his car himself and never gives it to anybody, observing the car unambiguously signals that its owner has arrived. But if the car is sometimes used by X, and sometimes by his relatives and friends, presence of X’s car does not necessarily imply that he is here. But even if the outcome of applying the PC and RC to a particular sentence in a particular context is conditioned pragmatically, there is every reason to assume that
TSL[v.20020404] Prn:17/10/2002; 14:35
F: PBNS10.tex / p.13 (189)
Inferred evidence
coded meaning of inferential evidentials does not change from context to context. It can be summed up as the following generalization: (15) Inferential evidentials signal that the speaker, relying on permanent conventionalized knowledge of the world shared by members of a given community, evaluates unwitnessed situation p with a fully specified set of participants P as the most probable cause for the observed state of affairs q with a set of participants Q. This generalization has two clear advantages over definitions cited in Section 2. While notions like ‘consequences’, ‘traces’ and even ‘results’ are set up ad hoc, two concepts on which (15) is based – permanent knowledge and evaluation of the probability of events – are much better defined and motivated independently. First, distinction between permanent and episodic knowledge is widely accepted within cognitive science, artificial intelligence, and, in recent years, within the functional approach to language. Extensive reference to this distinction is made, for example in (Givón 1984/1990). Second, evaluation of the probability of a situation is one of the basic concepts adopted in studies of modality (see, for example, van der Auwera & Plungian 1998 for further reference). Another advantage is also quite straightforward. While the definitions in Section 2 say little or nothing about which ‘results’ or ‘traces’ can and which cannot be used for inferring antecedent situations, from generalization (15) testable predictions can be deduced.
. Wider theoretical implications: Inferred evidence and epistemic evaluation Both constraints discussed above reflect the same general idea. An unattested situation is accessible for inferential evidentials if it is fully recoverable through evidence available at the moment of speech, that is, if the speaker is able to identify its taxonomic type and the whole set of its participants. This identification involves different kinds of epistemic evaluation and adjusting this evaluation to the speaker’s conventionalized knowledge. But epistemic evaluation is also a key concept for epistemic modality, a category marking the speaker’s attitude toward the proposition. Therefore, a general question about the relation between evidentiality and epistemic modality is bound to arise. There is a great variation in answering this question: some linguists (e.g., Palmer 1986) consider evidentiality a subcategory of modality, others (Chafe & Nichols 1986) view evidentiality as a category coding a broad range of epistemic meanings, including modal ones, and not only the source of information. One
TSL[v.20020404] Prn:17/10/2002; 14:35
F: PBNS10.tex / p.14 (190)
Sergei Tatevosov
more option (van der Auwera & Plungian 1998) is that evidentiality and epistemic modality, being different by nature, have nevertheless an appreciable area of conceptual overlap. Discussion in Section 3 strongly suggests that at least in Bagwalal, Tatar, and Mari a fixed modal value is associated with all appropriate sentences containing inferential evidentials. All the verbal forms in the above examples not only signal that knowledge about the asserted proposition is obtained via inference, but also express the speaker’s commitment to the truth of this proposition; in other words, experiential inferentials express what Bybee et al. (1994: 320–321) call ‘inferred certainty’. In this respect experiential inferentials differ radically from non-experiential ones. Consider (16) from Lega (Botne 1997: 516) showing non-experiential inferential particle ¢7mbε: (16) a.
C: Moke has walked in unenergetically with shoulders slumped. ¢7mbε Mfk¢7 ákorwa poss M. 3s-prs-be tired-fv It’s possible (that) Moke is tired
b. C: The speaker knows Kiguma has been sleeping most of the day and that he likes to drink palm wine ¢7mbε Kiguma ánunε maku poss K. 3s-drink-pst 6-beer Perhaps Kiguma is drunk. (16a) shows experiential use of the particle (the statement is based on the observed situation ‘Moke has walked in unenergetically, etc.’), while (15b) illustrates nonexperiential inference in which the speaker only relies on various types of knowledge. As Botne (1997: 517) has pointed out, ¢7mbε rather expresses conjecture than viable assertion, and this is the case with both (15a) and (15b). Note that this has been completely impossible for experiential inferentials discussed in Section 3: none of them can involve any uncertainty about the truth of the proposition. There is, therefore, a strong correlation between experiential nature of inference and degree of speaker’s commitment: non-experiential inferentials, for which ¢7mbε in Lega is an example, are associated with a lower degree of certainty than experiential ones. On the other hand, the more experiential, the more certain. I suggest that this correlation is predictable from the PC and RC formulated above. The PC requires that there are no alternative explanations of the current state of affairs except the asserted one, while the RC leaves no possibility that participants of a situation remain unspecified (although they need not be definite). This further implies that no room for epistemic uncertainty is left and, therefore, that
TSL[v.20020404] Prn:17/10/2002; 14:35
F: PBNS10.tex / p.15 (191)
Inferred evidence
the speaker takes full responsibility for the assertion that the situation has indeed occurred in the actual world. In contrast, non-experiential inferentials have no restrictions like those discussed in Section 3, and the PC and RC are irrelevant for them. Consequently, they are not associated with the highest degree of certainty, and relevant propositions are rather hypothesized than asserted. Data from three languages discussed here are of course too restricted to serve as a basis for valid cross-linguistic generalizations. Nevertheless, striking similarities between unrelated languages can hardly be a matter of pure coincidence; these similarities are very likely to reflect universal properties of grammatical morphemes associated with the domain of inferentiality.
Notes * I am very much indebted to the inhabitants of the villages Kwanada (Daghestan Republic), Tatarskij Jaltan (Republic of Tatarstan), and Staryj Torjal (Marij Republic) who served as informants on Bagwalal, Tatar, and Mari respectively. The financial support from the Research Support Scheme (RSS No. 1474/1999) is gratefully acknowledged. . The following abbreviations are used throughout the paper: 2SG 2nd person singular; 3PL 3rd person plural; 3SG 3rd person singular; ACC accusative; AFF affective; AUX auxiliary; CONV converb; ERG ergative; EVID eviential; FV final vowel; GEN genitive; INDEF indefinite; INFER inferential; M masculine; N neuter; OBL oblique; PFCT perfect; POSS possibility; PRS present; PST past.
References Aksu-Koç, A. A., & Slobin, D. I. (1986). A psychological account of the developement and use of evidentials in Turkish. In W. Chafe & J. Nichols (Eds.), Evidentiality: The linguistic coding of epistemology (pp. 159–167). Norwood: Ablex. Anderson, L. B. (1986). Evidentials, paths of change, and mental maps: typologically regular asymmetries. In W. Chafe & J. Nichols (Eds.), Evidentiality: The linguistic coding of epistemology (pp. 273–312). Norwood: Ablex. Auwera, J. van der, & Plungian, V. A. (1998). Modality’s semantic map. Linguistic Typology, 2 (1), 79–124. Botne, R. (1997). Evidentiality and epistemic modality in Lega. Studies in Language, 21 (3), 509–532. Bybee, J., Perkins, R. D., & Pagliuca, W. (1994). The evolution of grammar: Tense, aspect and modality in the lanuages of the world. Chicago: University of Chicago Press. Chafe, W., & Nichols, J. (Eds.). (1986). Evidentiality: The linguistic coding of epistemology (pp. 273–312). Norwood: Ablex. Dahl, Ö. (1985). Tense and aspect systems. Oxford: Blackwell. Givón, T. (1982). Evidentiality and epistemic space. Studies in Language, 9 (1), 23–49.
TSL[v.20020404] Prn:17/10/2002; 14:35
F: PBNS10.tex / p.16 (192)
Sergei Tatevosov
Givón, T. (1984/1990). Syntax: A functional-typological introduction. Vol. I–II. Amsterdam: John Benjamins. Guentchéva, Z. (Ed.). (1996). L’énonciation médiatisée. Paris and Louvain: Peeters. Guentchéva, Z. (Ed.) (to appear). L’énonciation médiatisée II. Paris and Louvain: Peeters. Haan, F. de (1998). The Category of Evidentiality. Unpublished Ms. Ivanov, I. G., & Tuzharov, G. M. (1970). Marijskij jazyk. {Mari.}. Joshkar-Ola: Marijskoe knizhnoje izdatel’stvo. Johanson, L., & Utas, B. (Eds.). (2000). Evidentials in Turkish, Iranian, and neighbouring languages. Berlin: Mouton de Gruyter. Kibrik, A. E. (Ed.). (2001). Bagwalinskij jazyk. {Bagwalal}. Moscow: Nasledije. Kozintseva, N. A. (Ed.). (to appear). Tipologija evidentsial’nosti {Typology of evidentiality}. St. Petersburg: Nauka. Lazard, G. (1996). Le médiatif en persan. In Z. Guentchéva (Ed.), L’énonciation médiatisée (pp. 21–30). Paris and Louvain: Peeters. Meydan, M. (1996). Les emplois médiatifs de mi¸s en turc. In Z. Guentchéva (Ed.), L’énonciation médiatisée (pp. 125–144). Paris and Louvain: Peeters. Nedyalkov, V. P. (Ed.). (1988). Typology of resultative constructions. Amsterdam: John Benjamins. Nikolaeva, I. (1999). The semantics of Northern Khanty evidentials. Journal de la Société Finno-Ougrienne, 88, 131–159. Palmer, F. R. (1986). Mood and modality. Cambridge: Cambridge University Press. Plungian, V. A. (2001). The place of evidentiality within the universal grammatical space. Journal of pragmatics, 33, 349–357. Slobin, D. I., & Aksu, A. A. (1982). Tense, Aspect, and Modality in the use of the Turkish evidential. In P. J. Hopper (Ed.), Tense-aspect: Between Semantics and Pragmatics (pp. 185–200). Amsterdam: John Benjamins. Tatevosov, S. G. (2001). From resultatives to evidentials: multiple uses of the Perfect in NakhDaghestanian languages. Journal of Pragmatics, 33, 443–464. Willett, T. (1988). A cross-linguistic survey of grammaticization of evidentiality. Studies in Language, 12 (1), 57–91. Woodbury, A. C. (1986). Interactions of tense and evidentiality: a study of Sherpa and English. In W. Chafe & J. Nichols (Eds.), Evidentiality: The linguistic coding of epistemology (pp. 188–202). Norwood: Ablex. Zakijev, M. Z. (Ed.). (1993). Tatarskaja grammatika: Morfologija {Grammar of Tatar: Morphology}. Vol. 2. Kazan: Izadatel’stvo Kazanskogo Universiteta.
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.1 (193)
Extension of meaning* Verbs of perception in English and Lithuanian Aurelia Usonien˙e University of Vilnius, Lithuania
Polysemy, ambiguity, change of meaning in time and over time have been widely discussed in linguistics (Aijmer 1980; Viberg 1984, 1993; Bolinger 1991; Traugott 1989; Sweetser 1993; Bybee et al. 1994; Barron 1997). The present paper is basically concerned with the meaning of verbs of perception in terms of the semantic opposition of direct vs. indirect perception, which can be roughly accounted for by the structural differences in the complementation of these verbs. The focus of the analysis is on how syntax interacts with semantics and the purpose of the study is to compare the patterns and paths of extension of meaning of the given verbs in two languages that are genetically and structurally very different.
.
Background preliminaries
. Frequency data It is not surprising that verbs of perception (PV) are very common in our language. All our contacts and communication with the world surrounding us go via our senses. Sense data obtained are the raw material for further mental processing; therefore it is absolutely natural to expect that verbs of cognition and perception should be among the most frequently used words. What is of major interest for contrastive linguistics is what words appear the highest in the lists, how and why the given frequency is achieved. As Viberg’s (1993) research results of 20 most frequent verbs in 11 European languages show, the field of perception is represented by the key verb denoting visual perception: (1) Basic verbs in European languages Reflecting universal tendencies: The nuclear verbs
(Viberg 1993: 346)
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.2 (194)
Aurelia Usonien˙e
Motion
Possession Production Verbal Perception Cognition Desire communication
GO (10)
GIVE (11) MAKE (10) SAY (11) SEE (11)
KNOW (9) WANT (8)
COME (8) TAKE (11) Languages: English, German, Swedish; French, Spanish, Italian, Rumanian; Russian, Polish; Finnish, Hungarian. In the frequency lists from the words in the British National Corpus1 (BNC) produced by Kilgarriff (http://www.itri.bton.ac.uk/∼Adam.Kilgarriff/bnc-readme.htm) the verb SEE precedes KNOW: (2) Lemmatised Frequency List ordered by frequency (LEMMA. num.) 51 52 64 90 143 148 319 592
191661 185534 153881 111058 62445 62185 30595 17461
SEE KNOW THINK LOOK SEEM FEEL APPEAR HEAR
V V V V V V V V
As we can see from the list, other verbs of perception under investigation in this paper come very high in the list as well. The Frequency Dictionary of Modern Written Lithuanian published in 1997 offers a very similar picture. Perception and cognition verbs are the key verbs in their fields and among the top 20 verbs in the list: (3) Lithuanian (Grumadien˙e & Žilinskien˙e 1997): 50 (5) ŽINOTI V (‘know’) 2081 708 150 (21) MATYTI V (‘see’) 937 481 195 (30) ATRODYTI V (‘look/seem’) 754 409 At first sight it seems that the leading position of the key verb of seeing in the verb frequency lists of many languages can be at least partly attributed to its highly polysemous nature. As we can see in (4), the number of meanings given in dictionaries of the languages cited below ranges from 9 to 30: (4) Language English German Swedish
Number of meanings 28–30 11–17 4–5 (15)
Vision + + +
Mental process + + +
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.3 (195)
Extension of meaning
Icelandic 5 (17) French 22 Russian 9 Lithuanian 20
+ + + +
+ + + +
Moreover, a closer look at the meanings themselves will suggest further subdivision into two basic types of meaning, namely (a) those that are related to or based on vision and (b) those that are related to some mental processing. According to Sweetser (1993), one of the reasons accounting for the polysemy is a metaphorical change of meaning that basically proceeds ‘from concrete to abstract’ and what the scholar calls ‘the Mind-as-Body Metaphor’ (Sweetser 1993: 29). In the present study the given opposition will be regarded as somewhat parallel to direct vs. indirect2 perception and most of the attention will be devoted to the extension of meaning that is primarily ‘complement-type-dependent’. Thus, the focus will be on the interaction of syntax and semantics in the example of the analysis of verb complementation. As is well known in linguistics, directness vs. indirectness in complementation (Borkin 1973; Aijmer 1980; Duffley 1992) is very much dependent upon the choice of the complementizer introducing a clause (Riddle 1975; Frajzyngier & Jasperson 1991; Dixon 1995). That-clauses take a very special place among various kinds of clauses in that “all that-clauses have the property of being interpreted as being in the de dicto domain” to quote Frajzyngier and Jasperson (1991: 139), which means that they do not refer to something as a direct description of an event. The conjunction THAT comes very high in the frequency lists of the two languages, in fact it appears twice as frequent as THAT-DET in English and the Lithuanian equivalent KAD-CONJ (‘that’) comes the 6th in the list, as shown in (5): (5) English(http://www.itri.bton.ac.uk/∼Adam.Kilgarriff/bnc-readme.htm): 13 760399 THAT CONJ 27 384313 THAT DET Lithuanian (Grumadien˙e & Žilinskien˙e 1997): 6 KAD3 JNG (‘that’-CONJ) 9789 1141 Thus, it seemed that checking the occurrence of the complementizer THAT and KAD in the post-predicate position of these verbs might cast light on the ratio of perceptual ‘directness vs. indirectness’ in the situations described by PVs. However, the preliminary results of the BNC search were not very surprising. The occurrence of THAT-clauses with the verb SEE as compared to WH-clauses is really the highest, as can be seen in Figure 1. The percentage of the matches of SEE forms taking THAT-complementizer (see/sees/saw/seen+CJT) against the total number of occurrences of the basic forms SEE, SEES, SAW, SEEN has turned out to be quite low, i.e. from 3–5% only (e.g.:
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.4 (196)
Aurelia Usonien˙e
Figure 1. Finite complementation of the English verb SEE
SEE-113, 315 vs. SEE+CJT-2, 989), which suggested that other types of complements are more common with this verb. However, when the reduced sets of concordances for SEE/SEES/SAW/SEEN were analysed manually for the types of complements they take, the percentage of THAT-clauses appeared to be much higher, as shown in Table 1: Table 1. Finite complementation of SEE
SEE SEES SAW SEEN (in active voice forms)
S-that
CP-finite (total)
10% 2% 8% 6%
16% 5% 12% 10%
This can be explained by the fact that post-predicate THAT-clauses (see Biber et al. 1999 for more detail on retention and omission of the complementizer) may be placed far from the matrix verb, e.g.: (6) I saw it was an accident. . . . he was still able to see from their reflection in the mirror behind the bar that they were deep in conversation. Therefore in many cases manual analysis was preferable for accuracy reasons. In Lithuanian KAD (‘that’)-clauses seem to be in a more favourable position. Moreover, their frequency happens to be directly dependent upon the time reference of the matrix clause. As is shown in Figure 2 below, their occurrence with the present tense singular forms of MATYTI (‘see’) is considerably greater. Results of the comparison of the percentage of finite clausal complementation (including both THAT & WH-words as complementizer) with present and past tense 1st-
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.5 (197)
Extension of meaning
1st person 20
Present
10
Past 3rd person
2nd person
Figure 2. KAD (‘that’)-clauses with MATYTI (‘see’): Person & Time Reference
/2nd-/3rd-person singular forms of seeing verbs in the two languages are presented in the table below: Table 2. Frequency of finite CPs in PV complementation (Present/Past tense forms) Present
English
Lithuanian
1st: I see/Aš matau 2nd: You see/Tu matai (sg) 3rd: S/he sees/Ji/s/jie mato Past SAW & MAT*
8% 14% 5%
31.6% 33.4% 34.3%
9%
22%
As seen from Table 2, clausal complementation is much more common in Lithuanian. Moreover, variation in the percentage of personal forms of MATYTI (‘see’) used parenthetically seems to depend upon the same parameters. Compare Figure 3 below with Figure 2 above:
Figure 3. Parenthetical use of personal singular forms of MATYTI (‘see’)
An increase in the number of finite complement clauses as well as parenthetical use of the verb MATYTI (‘see’) in Lithuanian allows us to assume that it is dependent upon the time reference and the person. Mention should be also made of the fact that plural forms of the 1st- & 2nd-person show even higher percentage
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.6 (198)
Aurelia Usonien˙e
of parenthetical use. A shortened present tense 2nd-person plural form MATOT (‘you see’)(cf. full form MATOTE) is used basically as a parenthetical (76%). The percentage of the present tense 1st-person form MATOME (‘we see’) used parenthetically is 19%. The most frequent parenthetical use of the English verb SEE has been found to be with the 2nd-person and 1st-person subjects both singular and plural, as shown below: (7) You see – 69%;
I see – 49%;
As we saw – 32%;
Moreover, it must be noted that the number of occurrences of YOU SEE in the BNC is 99, 623, which makes up about 87% of the total matches of SEE (113, 315). Thus, preliminary calculations (without analysing other types of complex NP structures following PVs) presented in Table 3 show that both finite clause complementation of these verbs and their parenthetical use constitute about 50% of usage of this verb in the two languages under analysis: Table 3. Preliminary statistical findings on indirectness in perception CP-finite Parenthetical Reference marker Total “see that/Ø/wh-S” “(as) you see” “see Ch.4.” SEE MATAU (‘see’ 1Person-sg) MATAI (‘see’ 2Person-sg) MATO (‘see’ 3Person-sg)
16% 31% 38% 35%
16% 14% 12% 3%
29% – – –
61% 45% 50% 38%
In all of these types of use, the verb SEE has no meaning of perceptual directness in the situations described. This indirect perception reading is much more common for the present tense forms, while past tense forms seem to give preference to the transitive type of functioning, as can be seen below: Table 4. Transitive use of SEE (Lithuanian vs. English)
Transitive
PRESENT mat-au/ai/o (‘see’ 1st-/2nd-/3rd-sg)
PAST maˇciau/mat-ei/˙e (‘saw’ 1st-/2nd-/3rd-sg)
PRESENT see/sees
PAST saw
(49/41/57) ∼ 49%
(65/61/64) ∼ 63%
(64/95) ∼ 79%
∼ 87%
Note: analysis of the meaning of PVs taking various types of nominal and non-finite complements in terms of directness/indirectness of perception will be given in Section 2.
The move of English and Lithuanian 1st- & 2nd-person forms of SEE from a full verb to a parenthetical can be checked against Aijmer’s (1997) model of analysis accounting for the structural flexibility of I think on a cline of pragmaticalization. The scholar proposes three basic stages in this development:
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.7 (199)
Extension of meaning
(8) Stage 1 Stage 2 Stage 3 I think that Bill is I think Bill is at home Bill is at home I think at home (Aijmer 1997: 7) The first person subject is said to be an important factor for developing into a parenthetical. The basic features distinguishing between the three stages are: first, a weakened form of the verb and loss of its status as a main verb; second, deletion of the complementizer, i.e. zero complementation; and third, acquisition of a new status (a pragmatic discourse marker/modal particle). Thus, we can check how the behaviour of seeing verbs in the two languages fits the above criteria for pragmaticalization. The very first point to be mentioned is the second person subject with SEE that is absolutely dominant in the parenthetical use, which is worth special attention and emphasis. Moreover, in English the second stage does not seem to play a very important role since Ø-complemented clauses are not very common with SEE. They constitute only 1–2.5% of the total usage of the verb SEE. The Lithuanian verb seems to be closer in this sense to the development of the English I think as shown by Aijmer. The percentage of Ø-complementation with personal forms of MATYTI ranges from 2.5% (3rd-person past tense form of ‘see’) to 10% (2nd-person present tense form of ‘see’). The last point is related to finite clause complementation, which has been observed to be much more common for the Lithuanian verb MATYTI4 (‘see’) than for its English counterpart. However, the greatest difference between the two languages can be seen in the behaviour of the verbs of seeming in terms of the interaction between the tense, person, complementation and change of meaning. Consider the data in Table 5: Table 5. Seeming verbs in English and Lithuanian ˙ SEEM SEEMS SEEMED ATRODO ATRODE˙ RODOSI RODESI (‘seems’ 3) (‘seemed’ 3) (‘seems’ 3) (‘seemed’ 3) CP-finite 5% Parenth. 4%
22% 3%
5% 2%
22% 39%
24% 16%
26% 31%
32% 27%
In English, the verb SEEM has a very low percentage of finite THAT-clause complementation and it does not develop much of parenthetical use, though there are a few occurrences of cases like: Seems I was wrong. On the contrary, 3rdperson present tense forms of the Lithuanian seeming verbs ATRODYTI/RODYTIS (‘look/seem/appear’) demonstrate a dramatic move towards parentheticalization. They also have a high percentage of finite complementation. The results of the analysis have also shown that if a verb form does not take that-clauses for complementation, it is not used as a parenthetical, which is absolutely in line with logic: if there are no chances for the speaker to express his/her attitude towards a proposition expressed, there is no further development to-
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.8 (200)
Aurelia Usonien˙e
wards a free modal modifier/qualifier. This holds true for all 1st- & 2nd-person forms of ATRODYTI/RODYTIS (‘look/seem/appear’), which do not function as parentheticals. So far, the analysis has been mainly concerned with the formal and statistical aspects of the types of complementation of perception verbs in English and Lithuanian. A more elaborate semantic analysis of the types of complementation with these verbs will be given in Section 2.
. Verbs of perception: Exp(eriencer-perception)-verbs vs. St(imulus-perception)-verbs Following Rogers 1971, Viberg 1984, Quirk et al. 1985, Schlesinger 1992, Croft 1993, Levin 1993, verbs and verb phrases under analysis are subdivided into Expverbs and St-verbs and phrases, as in the examples in (9) and used in the types of sentences given in Table 6. (9) Exp-verbs & phrases E:
St-verbs & phrases
see
look, seem, appear be seen/visible LT: mat-yti (tr) (‘see’ Inf) at-rod-yti/rod-yti-s [Prf. ‘show’ Inf] (‘look/seem/appear’) matyti (intr) (‘see’ Inf = ‘be seen/visible’) The analysis will focus on the use of Exp-verbs (active and passive forms) that describe a relation between two entities (the perceiver and the perceived) and Stverbs that describe a situation as the perceived-oriented. Most of the attention will be devoted to the complement types occurring in post-predicate position of both subgroups of verbs and the way these complement types can affect the meaning of the verb. Thus for Exp-verbs it is the grammatical Object position, while for St-verbs it is the position of the grammatical Subject and its complementation. The purpose of the present study is to compare the paths of extension of meaning of perception verbs that are very much complement-type-dependent in two languages that are genetically and structurally very different.5 Basic issues to be dealt with: –
– –
structural and semantic differences in perception verb (PV) complementation that are considered to be crucial in distinguishing between the descriptions of direct vs. indirect cases of perception; how much and what kind of indirectness is present in the situations described by PVs; types of modal qualification that linguistic expressions denoting cases of indirect perception can develop.
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.9 (201)
Extension of meaning
Table 6. Use of English and Lithuanian PVs E: Exp-verbs
St-verbs & phrases
Do you see that big brown thing that looks a boat?
She looks the image of Mother. The badge looks to be from the ’50s. How slim and white they seem in the moonlight. The Officer was talking to two gunners who appeared to be Polish judging by their accents. Seated, she appeared smaller than usual, and more fragile. The narrating self not only sees but wants to be seen to be seeing: the registration of detail, and the reproduction of objects . . . Grammar, then, can be seen as a resource for the adaptation of lexis.
She was surprised to find the newly-weds still in the church forecourt, until she saw a photographer packing up his equipment. I love to see that sun go down. What they saw as a problem, he saw as a challenge.
LT: Jis mat˙e mergait˛e, galvojo apie ja.˛ (‘He saw a/the girl, thought about her’) . . . ja˛ matom stovinˇcia˛ ant savo namo laiptu˛ . . . (‘[we] see her standingPrPartSgFAcc on the staircase of her house’) . . . dažnai matom ji˛ einant ˛i miesta˛ . . . (‘often [we] see him goingPrsGer to the town’) . . . aiškiai matai, kad turi reikala˛ su robotais. (‘[you can] see clearly that [you] are dealing with robots’)
Jos nosis . . . atrodo juokingai ilga . . . (AP) (‘Her nose looks/appears ridiculously long’) Seni juvelyriniai dirbiniai dažniausiai atrodo nat¯uraliai. (AdvP) (‘Old pieces of jewellery often look natural’) O ta pelk˙e buvo matoma iš toli. (PrsPartPass) (‘That marsh was seen from afar’) Iš antro aukšto buvo matyti kiemas. (Inf) (‘From the first floor the yard was seen/ visible’) . . . matyti, kad žmogus didelis vargšas . . . (Inf) (‘it is seen that the person is very poor’) Ta gi, matyti, turi savo fantaziju˛, . . . (Inf) (‘That [woman] apparently has got her own fantasies’)
Basic claims to be made: –
–
there are common features in the path of extension of meaning of PVs and factors affecting change of meaning are finiteness/non-finiteness in clausal complementation, presence/absence of introducing elements in the PV object complementation, time reference; cases of indirect perception with Exp-verbs tend to develop a modal qualification of evidentiality while those with St-verbs and phrases tend to con-
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.10 (202)
Aurelia Usonien˙e
vey qualification that is basically speaker-“strength-of-knowledge”-oriented and should be regarded as belonging to the realm of epistemic possibility/uncertainty.
. Previous studies: Directness/indirectness of perception It is mainly post-predicative complementation and the stimulus position that play a crucial role in the distinction made between cases of direct vs. indirect perception. The most important features that are crucial for the distinction between the two types of perception are (a) time reference, (b) types of entities denoted by complements and (c) purely formal structural differences, which have been widely analysed by many linguists. a. Time Reference: tense sequence: simultaneity (Woodbury 1986); determined/ dependent time reference (Noonan 1987: 92) time dependent reference (Dik & Hengeveld 1991: 240), Simultaneity Condition (Felser 1998: 352). All scholars agree that simultaneous interpretation of the events described by the verbs in the matrix and embedded clause usually is a necessary condition for the feature of directness to be present in a perceptual situation. b. Types/characteristic of entities denoted by complements: Truth vs. Occurrence complements denoting ‘direct physical perception vs. abstract cognitive evaluations of the truth of a proposition’ (Ransom 1986: 63), perceptually vs. mentally accessible entities (Sweetser 1993), direct vs. indirect (Tasmowski 1989; Verspoor 1996; Felser 1998), immediate vs. mental perception (Dik & Hengeveld 1991: 239), ‘resultatively’ vs. ‘operationally’ evoked perception (Duffley 1992). There is no polemic that the more abstract the perceived entity is the more indirect is the perception. Quite a few linguists (Borkin 1973; Duffley 1992) observe the importance of purely formal elements. c. Formal features in linguistic representation: presence/absence of the particle TO with the infinitive form (full or bare infinitive in the Infinitival Perception Verb Complements (IPVCs). They claim that occurrence of the Infinitive with TO triggers a reading of indirect perception of the entity denoted by the complex structure. The most common example for the marked member of the opposition is as follows: (10) I saw him to be obnoxious. He saw the children to be eating their lunch.
(Duffley 1992: 30) (Duffley 1992: 31)
Among the works that have dealt with entity types of different order and their linguistic expression, one can mention the classifications developed by Lyons (1977,
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.11 (203)
Extension of meaning
1991) and Mackenzie (1996). When dealing with complementation of perception verbs, Dik and Hengeveld (1991) have worked out the typology of perception-verb complements which is based on the hierarchical clause structure. It is exceptionally the so-called Exp-verbs like the English see, hear that their research was devoted to, which can be illustrated by the following examples: (11) a.
Direct: I saw your brother yesterday. (Immediate Perception of Individuals – IPI) I saw him walk down the street. (Immediate Perception of State of Affairs – IPSoA6 ) b. Indirect: I saw that Mary had been crying. (Mental Perception – MP) (Dik & Hengeveld 1991: 237–239)
It is generally accepted that the binary semantic opposition of direct vs. indirect perception is based on the post-predicate complementation and is parallel to 1st/2nd-order entities vs. 3rd-order entities that roughly can be seen as finding their linguistic expression in non-finite vs. finite forms of complement clauses where the choice of the complementizer can be crucial for the interpretation of the clause in terms of denoting perceptually/mentally accessible phenomena.
. Previous studies: Modal qualification It is generally accepted that in cases of finite complementation the Exp-verb SEE is used to describe cases of indirect perception, which can be accompanied by some modal qualification. Chafe (1986) and Woodbury (1986) consider this modal qualification to be evidential, i.e. indicating the path/source of knowledge. Verbs of seeming in English are usually referred to as (semi-)copula verbs. Semantically, they are regarded as stimulus/flip perception conveying a kind of meaning that can be considered modal (Poutsma 1926; Austin 1962; Poldauf 1972; Aijmer 1980; Wierzbicka 1980) and the latest studies claim that they are used as evidentials (Chafe 1986; Mithun 1986; Aijmer 1996; Barron 1997; Gisborne 1998). There seems to be no polemic regarding the effect of raising which brings about “a meaning which combines temporal properties with epistemic modality” (Rooryck 1997: 44). In Usoniene (2000) an attempt was made to show that the meaning of English verbs of seeming can also be described in terms of directness vs. indirectness of experience. The given semantic opposition is claimed to be very much complement-type-dependent in that it is based on the presence/absence of TO BE, which is said to function as a kind of a proposition marker. Thus the structures LOOK/SEEM/APPEAR ØP are used to describe direct impressions about the stimulus of perception, while LOOK/SEEM/APPEAR to be P contain modal quali-
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.12 (204)
Aurelia Usonien˙e
fication in terms of speaker’s uncertainty regarding what is being asserted. A more detailed discussion will follow below.
. Data analysis . Direct vs. indirect perception: Exp-verbs As I have already mentioned, the study has been focused on the complement types occurring in the Stimulus position of both subgroups of verbs. The present analysis has shown that a crucial factor for the verb SEE to have an IPSoA reading is not the structure itself but the type of entity showing up in the Subject position of the embedded infinitival or participial clause. Compare the following sets of examples under (12), where structurally both sets (a) and (b) have the so-called infinitive/participial perception verb complement (IPVC/PPVC), however it seems that only those in set (a) can be regarded as visually perceivable phenomena: (12) a.
sensory: He was on the other side of the road, saw me watching him, stopped, turned and walked slowly for fifty yards in the direction he had come. ‘I saw his leg coming up so I pulled the rope off and I jumped,’ he added. I looked up and saw again the arrow cut into the tree. b. non-sensory (mental): He sees the cost of the trees planted near the entrance to the company’s building branching into profit. We see companies rethinking how they can manage product design, ...
Conversely, those in set (b) should be regarded as intellectual products, i.e. mentally perceivable entities that exist in the Experiencer’s mind. They describe cases of perception that should be considered physically indirect but mentally direct. Consequently, it does not look appropriate to regard all of them as cases of Immediate Perception of SoA. The crucial factor here seems to be the type of NP showing up in the Subject position of the non-finite embedded clause. When they contain abstract nouns like panic, demand, rise that do not denote any visual stimuli, the state of affairs described by the complement clause can be interpreted as cases of direct mental perception. Thus the given extension of meaning of the verb SEE is not dependent upon the structure of the complement but seems to be more lexical-meaning-dependent.7 The results of the data analysis suggest that there are at least two more types of complex complement structures following the English verb SEE that should be also considered as bringing about an indirect perception meaning of this verb. The
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.13 (205)
Extension of meaning
first one will comprise the structures where the active and passivized forms of the verb SEE are followed by AS-phrases (AS Ps), as in the following examples: (13) Such claims might be contentious, but it would be odd to see them as meaningless. These two chapters should, therefore, be seen as being theoretically and empirically linked. Lack of knowledge or skill was seen as a potential constraint by half the people. I mean, are we to see that as symbolic the setting fire to Rochester’s bed? But, although I heard that music, I never saw it as the kind of music I’d be involved in and that was because Kensal Green was a deprived place . . . As-phrases are used to give some qualitative characteristic about the stimulus of perception and are regarded here as conveying evaluatively charged information (cf. Poldauf 1968; Gisborne 1998). Traditionally, when used in the structures with transitive SEE, they are considered to be ‘introduced’ object complements (IOC) with the introducing element AS. I will not deal with the grammatical status8 of the introducing element (for more detail see Noël & Simon-Vandenbergen 1996). A statistical analysis of the active SEE/SAW NP as P and passive SEEN as P structures collected from the BNC has shown that the most frequently occurring phrases introduced by AS are NPs. ING-phrases come next, then APs follow and a few percentage points are taken by PPs/AdvPs/APs and finite clauses. Compare their percentage in Table 7:9 Table 7. Frequency of types of introduced Object/Subject complements with SEE
see AT0 N saw AT0 N see P.* Seen
as NP
as AP
as IngP
as edP
as S
As PP/AdvP
77% 55% 76% 70%
4% 4% 5% 14%
12% 21% 10% 11%
– 5% 2% 1%
6% 2% – –
– 1%/1% 2% 0.7%
A contrastive analysis of ‘opinion’ verb complementation in terms of the opposition of IOC and BOC (‘Bare Object Complements’) carried out by Defrancq (1996a) shows that the English verb SEE takes a somewhat middle position regarding the occurrence of AS as an introductory element of IOCs. The scholar claims the ratio to be 1 to 5, which means that compared to REGARD, which always takes AS, the verb SEE is used less often with AS, while BELIEVE never takes AS. Traditionally, these phrases are considered to function as attributes to NPs they modify. At first sight it might seem that as-phrases here are added to give some qualitative characteristic about the entity perceived (the grammatical Object or
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.14 (206)
Aurelia Usonien˙e
Subject) and they have no effect on the meaning of the verb SEE. One can expect (as in the previously mentioned cases with IPVCs) that it is mainly the type of entity showing up in the Stimulus position that determines whether it is mentally or perceptually observable. However, a preliminary survey of the examples from the BNC shows that even in cases when the stimulus is an individual (1st-order entity), the interpretation of the situation is not the ‘Immediate Perception of Individual’. It is something about the individual/s and that something is not a property that is directly perceivable/detectable by senses.10 Actually, it is an evaluative judgement about the individual (stimulus of perception), which presupposes some mental activity on the part of the Experiencer; consequently the verb SEE is used to describe mental perception, e.g.: (14) Does he see you as a partner, or is he already the boss? I never saw you as insincere. I do not want to be seen as a liability to you. They are seen as mainly young with insufficient life experience; they are considered intolerant and left wing Thus the given cases of perception cannot be immediate and direct for they are not dependent upon the functioning of the individual’s physical senses. The verb SEE comes to describe perception that is more inferential or intellectual than sensory and can be compared to the cases of direct mental perception described by the verbs TAKE or KNOW in sentences like: (15) I take it as a compliment when my students ask questions after class. Many know him as the British jazz singer. A very similar interpretation of this type of meaning for the verb SEE is given by Defrancq who distinguishes between “typical” and “exceptional” meanings of the “opinion” verbs which are based on the contrast of “unmarked or bare” vs. “marked or introduced” complements (1996b: 132). One more structure should be added to the type of ISC/IOCs the introducing element of which is TO BE, the function of which also seems to be that of an ‘indirectness/proposition marker’. The entities denoted by the phrases introduced with TO BE cannot be regarded as phenomena represented as happening in time or experienced directly by senses (see ‘indirect’ perception in Declerck 1991; “operative” meaning in Duffley 1992), e.g.: (16) The Bedouins, whom Cecil cared for and saw to be poor, were being paid in private out of Cecil’s own pocket. At this stage, Pip sees a gentleman to be someone who romantically escapes from the humble world, into a more fairytale-like world.
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.15 (207)
Extension of meaning
He is an extremely nice man and was seen to be a successful and popular president . . . All the phrases with the introductory element TO BE in (17) are used to characterize the stimulus of perception (grammatical subject/object). Syntactically they can be regarded as ISC/IOC. The most common phrases introduced by TO BE are APs, while the introductory element AS prefers NPs, which is shown in Table 8. Table 8. Frequency of types of phrases in the introduced complements with the passivized form SEEN: Introductory element TO BE vs. AS
Seen to be Seen as
NP
Adj
Ving
Ved
PP
S
18% 70%
33% 14%
17% 11%
21% 1%
9% 0.7%
0.6% 0%
It is the presence of the as-phrase or TO BE that seems to play a crucial role in initiating the change of meaning of the verb SEE in the given structures. The following sets of alternating structures have been chosen for comparison: –
– –
cases where the Subject/Object (the Stimulus of perception) of the verb SEE is followed by a BSC/BOCs where the attributive position is filled in with the Participle (-ing/-ed forms) and adjectives (APs), cases where the complement position is taken by ISC/IOCs with the introducing element AS, cases where the complement position is taken by ISC/IOCs with the introducing element TO BE.
As seen in cases of zero-complementation, the situations described contain information of the Experiencer’s direct perceptual experience (IPI/IPSoA) or direct (mental) impression (DI), while those with introducing elements seem to be a kind of report statements (indirect perception). The structures with AS-phrases contain an evaluative judgement and those with TO BE do not denote directly observable phenomena, they are mere reports on events/facts (cf. Palmer 1988). The basic claim to be made here is that AS has a few functions: to introduce an evaluation (an element of subjectivity) and to bring about a kind of distance/remoteness between a speaker/experiencer and the phenomenon described. The function of TO BE is assumed to be that of a proposition/indirectness marker in that it introduces a disjointed type of proposition (A school which is seen to be supporting its clients is much more likely to be supported in turn), which can be opposed to the structures with a bare complement phrase that describes direct perception, as in: Someone’s got to be seen eating here. My father’s figure could be seen frozen in a posture that suggested he was taking part in some ceremonial ritual.
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.16 (208)
Aurelia Usonien˙e
Table 9. Types of SEE Complements in Contrast SEE+NP+Ving
SEE+NP+as Ving
SEE+NP+to be Ving
Liz turned and saw Anna leaning sleepily against the frame. (IPSoA)
Later he saw Arnold as initiating the degradation of philosophy and religion.
Someone saw him to be walking away.
They saw costs escalating and sales slumping as the effect of oil crises and world recession hit the Company. (DI)
C. Wright Mills saw pro- fessionals as increasingly becoming the servants of the rich and powerful.
SEE+NP+Ved
SEE+NP+as Ved
SEE+NP+to be Ved
Midnight opened his eyes and saw Jess surrounded by a fuzz of sunlight. (IPI)
The Aristotelian paradigm saw the universe as divid- ed into two distinct realms.
Visual arts officer Mr Steve Chettle wants to see Cleveland taken seriously as a place to visit and admire its arts (DI)
They saw the present as marked by an absence of community, . . .
Until 1910 the chief sponsor of the union was the Liberal Party, which saw political capital to be made out of organizing farm workers in the Tory rural strongholds,
SEE+NP+AP
SEE+NP+as AP
SEE+NP+to be AP
He liked seeing Edith happy. (IPI)
He thought a little wistfully of Provence, tried and failed to see Edith as happy in the HUtel Paradis . . .
I saw him to be impolite.
SEEN+Ving
SEEN+as Ving
SEEN to be Ving
Councel’s second argument was that all the evidence presented to the court was that the defendant had been seen engaging in an actual violence, (IPSoA)
Even those texts which appeared concerned with uncontroversial matters were open to be seen as engaging in political and social questions in which the state could suddenly intervene . . .
Thus, unions are seen to be engaging in a conflict with firms managements as part of the class struggle . . . . . . pharmaceutical companies have been less willing to be seen to be working with cannabis-like substances.
In Lithuanian the passivized forms of MATYTI (‘see’) are usually used to describe direct stimulus perception/impression. KAIP (‘as’) phrases that modify (a) the grammatical subject11 or (b) the quality of the process are added to introduce comparison, which contains a kind of evaluation:
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.17 (209)
Extension of meaning
(17) a.
Saul˙e matoma kaip šviesus diskas tamsaus dangaus fone. (‘The Sun is visible as a bright disc against the dark sky’) ˇ panorama apie viena˛ kilometra˛ buvo matoma kaip ant delno. b. Cia (‘Here, the panorama for one kilometer was seen as if on one’s palm’)
The evaluative information conveyed in these examples is parallel to the evaluation expressed in the structures with seeming12 verbs both in English and Lithuanian. They are regarded as cases of direct impression unspecified in terms of sense modalities (Usonien˙e 1999), e.g.: (18) They looked as fresh as maids in May. Their colours and patterns seemed as fresh as the day they were made and she was amazed at their richness compared with the wartime materials she was used to. The door appears as ordinary dark wood door at first sight, with a pair of brass doorknobs. Mama Danut˙e – jauna, graži, šalia savo vaiku˛ atrodo kaip vyresnioji sesuo. (‘Mother Danute – young, nice, . . . looks like a senior sister’) All the instances of use in (17) & (18) are regarded as cases of direct perception/impression, i.e. the way the outer world looks to the Experiencer no matter how distorted the image or how subjective the impression might be. Another type of complementation that seems to affect the meaning of the verb SEE in the same way is its occurrence with an AP followed by a full infinitive in its complement position, as in the cases like: (19) As far as I have seen, the BBC has not even seen fit to broadcast a single programme in his memory. In this example, the post-predicate position of SEE is taken by an AP denoting a property, which is not perceptually accessible, hence SEE is not used to describe direct perception. It is indirect, as in the structures with a full IPVC following SEE: (20) In 1986, there was a nationwide appeal for anyone who may have seen Joan to come forward. Nobody did. . . . something’s looking white to someone is his having certain sensations which are excited in him by what we would ordinarily say was the object he saw to be white. To sum up, IPI and IPSoA structures taking only mentally accessible entities into their stimulus position are considered as cases of perceptually indirect but mentally direct perception (Direct Mental Impression). For simple structures this meaning might be purely lexical-meaning-dependent but in more complex structures this
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.18 (210)
Aurelia Usonien˙e
shift in meaning seems to depend upon the structure of the complement. Basic types of complements are given in Figure 4: NP2nd-order As P NInsP
NP SEE
to be P NP/AP to V-inf
NP
MATYTI
CP-finite
Figure 4. Indirectness of visual perception as dependent upon types of complements (post-predicate position)
. Direct vs. indirect perception: St-verbs Finite complementation and that-clauses following verbs of perception are usually associated with cases of indirect perception. There is actually no polemic regarding this issue in linguistics because in the sentences like I see that he was at home, the complement clause denotes a proposition, a 3rd-order entity that is only mentally accessible to the Experiencer, no simultaneity in the time reference between the ‘seeing’ and ‘being at home’, hence a non-experiential reading of the verb ‘see’ where the speaker’s judgment is based on inference as a source of knowledge. As Woodbury points out “it is a combination of complement type and tense SEQUENCE that determine evidential value” (Woodbury 1986: 199). Evidentials indicate the source of information, the path via which the evidence has reached the speaker’s awareness and are regarded as modal expressions. What I want to argue is that the English verb SEEM is also evidential. There is no doubt that verbs of seeming in English and Lithuanian can develop modal meaning, however it is the meaning of epistemic possibility and this kind of meaning is very much complementtype-dependent. The following evidence can be offered to support the claim. First, by saying I see that he is at home the speaker-Exp indicates (or at least s/he does not deny) that his/her knowledge comes from some sensory evidence. Consequently the sentence describes a case of indirect perception that is mental, which has also developed a modal meaning of evidentiality. The proposition asserted ‘he is at home’ is an entity that is non-perceivable directly by senses. What is perceptually accessible to our senses is ‘he, his things, his voice, etc.’ from the presence/absence of which one can infer that he is (not) at home. Thus, time reference does not seem to play a crucial role here, which is the case in comparison with such sentences as: “I see that Jack is drinking” where the verb see, as Woodbury observes,
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.19 (211)
Extension of meaning
might have an “experiential evidential value” (Woodbury 1986: 199). Moreover, all these facts are applicable to the Lithuanian data, e.g.: (21) Matau, (kad) jis (yra) namie. see 1SgPrs that he 3SgM be 3Prs homeLoc (‘I see that he is at home’) If we use the Lithuanian infinitive form MATYTI (‘see’), it will not report on the direct act of visual or mental perception of some entity (as in Namie matyti betvark˙e – ‘At home a mess is seen/visible’) but will inform of the speaker’s source of knowledge regarding ‘his being at home’. When using matyti, kad S, as in: Matyti, kad jis namie (‘It is seen that he (is) at home’), the speaker remains rather remote and not so much personally involved in making a judgement. The speaker relies on some more objective evidence than one’s self-inference. The use of St-perception verbs of seeming in English would mean a choice between two types of complements: either they are followed by ØP or with the introducing element TO BE. The claim to be made here is that the only alternative possible for a speaker to comment on somebody’s physical presence at home (location in space) by using verbs of seeming is the structure with the obligatory TO BE P, as in: He seems to be at home. The information conveyed would express the speaker’s uncertainty regarding ‘his being at home’, which is a hypothetical judgment. Thus, the verb SEEM here can be said to have a modal meaning of epistemic possibility (van der Auwera & Plungian 1998) like in ‘He might be at home’. The source of knowledge is not indicated directly but it can be some mental processing of the same sensory data like in the case with I SEE. Thus, the evidence is inferential, i.e. not direct but indirect, and the speaker expresses his/her attitude regarding the reliability of the propositional content (in terms of its truth value). This is the type of modal qualification that verbs of seeming develop when followed by TO BE P. Conversely, in cases of zero-complementation of the phrases like AT HOME to the verb SEEM, as in You don’t seem at home in the dark, the structure will convey information about somebody’s appearance/outward looks, which is a report on the speaker’s direct impressions. In the given example it is the property of not being comfortable and happy which is attributed to the Stimulus of perception. The perception might be either sensory or mental but it is considered as direct. The Lithuanian equivalent sentences to illustrate this opposition are: a.
in cases of indirect mental perception with modal reading, we find sentences with the 3rd-person forms of St-verbs ATRODYTI/RODYTIS (‘look/seem/ appear’) followed by a KAD (‘that’)-clause or used parenthetically as well as parenthetical MATYT(I) (‘see’ = ‘apparently/obviously’), e.g.:
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.20 (212)
Aurelia Usonien˙e
(22) Atrodo/rodos(i)/matyt, (kad) jis namie. seem 3Pres that he 3SgM homeLoc (‘It seems that he is at home/Apparently he’s at home’) b. in cases of direct perception-impression of Stimulus appearance (evaluative, subjective but with not modal qualification) – the Lithuanian ATRODYTI (‘look/seem/appear’) can be followed by a great variety of phrases, i.e. NPs with the head noun in the Nominative, Genitive, Instrumental case, APs, PPs, AdvP among them. As there is no equivalent phrase with the word ‘home’ to express a property of feeling happy and comfortable, other expressions can be chosen for illustration: (23) Jis nori atrodyti geras/gerai/geras žmogus/geru žmogumi. (ANom P/AdvP/NNom P/NIns P) (‘He wants to look/seem/appear good/well/a good man’) To sum up, personal forms like MATAU, KAD/I SEE [S] followed by finite clauses are regarded as cases of indirect mental perception with an evidential qualification (Givon 1982; Chafe 1986; Palmer 1990). The sentences with MATYTI (‘it is seen’) [S] and BE SEEN TO BE P (see examples in Table 10) should be also regarded as cases of evidentiality for the indication to the source of knowledge. The difference between the types of structures as I SEE [S] vs. SEEN TO BE P can be explained in terms of Frawley’s (1992) Deictic Categorisation and Scaling of Epistemic Modality. The inference of I SEE falls under the dimension of ‘To Self ’, which is either the speaker or the perceiver, hence “direct self-inference” and the speaker’s “personal responsibility” for his/her statement. The inference in the sentences with MATYTI, KAD S (‘it is seen that S’) seems to be closer to the dimension of “Other” and it can be rather intersubjective in this respect, however no modal meaning of uncertainty results from here. The speaker does not evaluate the probability/reliability of the propositional content but indicates only the path via which the evidence comes from, i.e. inference based on the sensory data available to the speaker. On the contrary, in the case of seeming verbs with full infinitves the dimension is within the scope of the “strength of knowledge” (which is possibility) and falls under “From Self ” or “Other”, which means that the knowledge can come from some external source, outside the speaker, hence “intersubjective inference” which can also imply a “shared” or common responsibility for the modal qualification, to paraphrase Nuyts (1993: 946). Thus, seeming structures with full infinitival forms can be regarded as conveying information that is epistemically modal because they indicate the speaker’s assessment of the situation in terms of greater/lesser degree of certainty regarding the truthfulness of the proposition asserted. By contrast, seeming structures with zero-complementation are considered non-modal unspecified perception reporting on the evaluatively charged qualitative characteristic of the stimulus.
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.21 (213)
Extension of meaning
. Concluding remarks An amazingly high frequency of the verb SEE in the two languages seems to have been achieved thanks to its functioning as a verb of non-visual perception. The most important factor in this process of change of meaning is complementation. In both languages the path of extension of meaning of Exp-verbs and St-verbs roughly goes from Direct to Indirect (Mental) perception, which can acquire different modal qualification that is either evidentiality or epistemic possibility: (24) Direct Perceptionspecified vs. unspecified (→Indirect perceptually = Direct mentally ) →Indirect Perceptionmental (CP-finite/to be P) →ModalEvidential vs.Epistemic possibility (Exp-verbs vs. St-verbs) In Lithuanian, finite complementation happens to dominate for describing indirectness of perception, while in English there is a preference to complex nominal structures in the PV complementation. However, the models they choose when developing alongside this path of change are different. Seeing verbs seem to adopt the model of pragmaticalization (Aijmer 1997). As has been shown, first and second person subject forms of SEE that have the highest percentage of parenthetical use cannot be analysed in terms of truth. In cases of finite complementation they develop a modal reading of evidentiality. They are used to indicate the source of knowledge without any assessment of the truth-value of the proposition. Conversely, seeming verbs look likely to undergo grammaticalization for the modal meaning they convey contains the speaker’s assessment of the propositional content in terms of truth. The speaker expresses his/her uncertainty regarding what is being said. A sum-up of what has been claimed is presented below: (25)
English Evidentiality SEE CP-finite BE SEEN to be Epistemic possibility LOOK (uncertainty) SEEM to be P APPEAR
Lithuanian MATYTI CP-finite ATRODYTI CP-finite RODYTIS
The basic observations and the data presented in the analysis of the English complement types with seeing and seeming verbs seem to be in line with Duffley’s claims that the to infinitive with perception verbs “involves a subtle shifting of the perceptual verb into the conceptual field with the consequent evocation of an inference and not just something directly perceived” (Duffley 1992: 37). An attempt has been made to give evidence for the claim that the opposition of ØP vs. TO BE P is meaningful in PV complementation. The function of TO BE is assumed to be that of a proposition marker.
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.22 (214)
Aurelia Usonien˙e
Notes * A greater part of the research for this paper was made possible by the Swedish Institute scholarship and the exchange programme between the British Academy and the Lithuanian Academy of Sciences (I owe a lot to Kasia Jaszczolt, Jan-Olof Svantesson for their kind attention and assistance). The attendance of the 2nd International Conference in Contrastive Semantics and Pragmatics was partly sponsored by the British Embassy and Vilnius University (special thanks are due to HMA Christopher Robbins and vice-rector Benediktas Juodka). I am also very grateful to Laima Erika Katkuvien˙e and Bonifacas Stundžia for the favour of a year’s study leave from teaching at the department as well as to my colleagues and friends for their help with the data and valuable advice (Jon˙e Grigali¯unien˙e, Raimonda Jonkut˙e, Ann Lindvall, Irena Marija Norkaitien˙e, Rasa Ruseckien˙e, Nijol˙e Regina Teiberien˙e, Johan van der Auwera, Caroline Willners). My special thanks are due to Evalda Jakaitien˙e for her comments on the Lithuanian data. . The data for the analysis have been collected from two corpora: 1.
The British National Corpus (BNC); use has been made of: –
– 2.
The British National Corpus Online service, managed by Oxford University Computing Services on behalf of the BNC Consortium. All rights in the texts cited are reserved. The BNC used at the Department of Linguistics and Phonetics, Lund University, Sweden.
The Lithuanian Language Corpus donelaitis: http://donelaitis.vdu.lt/tekstynas/ compiled at Vytautas Magnus University. The Lithuanian Corpus donelaitis (61 million words) consists of texts published since 1994 and reflects present day written Lithuanian.
If not indicated otherwise, the data cited are from the above corpora. It should be also noted that the Lithuanian corpus is not annotated. Therefore all the qualitative and partly quantitative analysis had to be done manually. A great number of sets of concordances for the linguistic phenomena (especially for the queries like SEE, SEEM, MATYTI, etc.) under investigation were usually reduced to a reasonable amount of data for a more accurate analysis (approximately 150–350 matches). In quite a few cases the same manual procedure has been also applied to the English data. This was considered necessary when dealing with the finite complementation of the verb SEE (see further for more detail). Moreover, this has also allowed detecting a slight mismatch in some of the query solutions. For instance, the percentage of THAT-DT0 cases found in the “I saw [pos = CJT]” query was about 9%. Those were the cases of THAT-anaphoric denoting eventualities (Asher 1993) and THAT-remote demonstrative. . There are different terms used to describe the given contrast: ‘experiential/non-experiential’ (Woodbury 1986), direct/indirect (Tasmowski 1989; Felser 1998; Declerck 1991), ‘immediate vs. mental’ (Dik & Hengeveld 1991). . In Lithuanian there is no complementizer derived from a demonstrative pronoun like the remote THAT in English. The Lithuanian KAD (‘that’) is a shortened form of KADÀ
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.23 (215)
Extension of meaning
<*KADA˛ which originally comes from the adverb ‘when’ (Zinkeviˇcius 1981, v 2: 194). JOG, another ‘that’-type complementizer in Lithuanian, is assumed to have been derived either from the relative pronoun or from the demonstrative pronoun plus the particle -gi, however its frequency is much lower. . The best example of parentheticalization seems to be a shortened form of the Lithuanian MATYTI (‘see’), i.e. MATYT (‘obviously’) with an epistemic meaning of weak possibility based on inference. The number of occurrences of the modal parenthetical MATYT in the ‘donelaitis’ corpus is greater than that of the infinitive form MATYTI, namely 6659 vs. 6155 respectively. Consider the following examples: Buvo matyti, kad tai veikli moteris. Matyti, (kad) meluoja. be 3Past seeInf that it active woman seeInf (that) lie 3Pres (‘It was seen that she was an active woman’) (‘It is seen that [s/he/they] lie’) Matyt, meluoja. obviously lie 3Pres (‘Obviously [s/he/they] lie’) Note: All the translations of the Lithuanian examples in this paper are nearly literal. . Lithuanian is a highly inflected (pro-drop) language with a well-developed agreement in the predicative and attributive relations. The verbs are inflected for person, number, gender, tense, voice, and mood, while nouns and adjectives - for case, gender and number. All these features are closely related to the verb forms under investigation because presence/absence of them is crucial when distinguishing between different forms of verbs (impersonal/neuter and personal) that formally are absolutely identical. Thanks to these agreement features, we can easily distinguish, for example, between the attributive use of the Past Passive Participle Masculinum Plural Nominative form MATYTI (‘seen’) and the Infinitive MATYTI (‘see’) (as in the examples given below). Moreover, agreeing forms can be both predicative and attributive, while non-agreeing forms are used only predicatively: Agr (Attr/Pred)
Non-agr (Pred)
a. Matyti darbai (PastPasPartMPlNom) matyti. (Inf) b. Matoma planeta (PrsPasPartFSgNom) vs. Kažkas matoma. (PrPasPart.N) c. Matyta moteris (PastPasPartFSgNom) matyta. (PstPasPart.N) (‘These are/is seen works/planet/woman’) (‘Something’s been seen/visible’) ¯ (‘be’), namely YRA In the cases of predicative use various tense forms of the copula BUTI (be.3Prs), BUVO (be.3Past) and BUS (be.3Fut) can be considered optional.With all the grammatical relations being morphologically marked, the sentence structure is flat with a relatively free word order within a clause: O – S – V; O – V – S; V – S – O; S – O – V, etc. In complex sentences the main clause as a rule comes first, and the subordinate clauses follow. However, topical fronting is not infrequent. Consider an example from the corpus below: Ar supras v˙eliau – matysime. (‘If [s/he/they] will understand later – [we]’ll see’) The Lithuanian complementizer KAD (‘that’-CJT) has not developed from the demonstrative pronoun, hence no ambiguity with ‘see that’ type clusters, where that can be used as a complementizer or a pronoun, for instance: see that S = matyti, kad S and see that thing
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.24 (216)
Aurelia Usonien˙e
= matai ta˛ daikta.˛ Anaphoric reference to abstract objects and eventualities in English and Lithuanian can be illustrated by the following pairs of examples: I’m trying to write. Can’t you see that? I can’t see that as being loving a woman. . . . mes matom tik tai, kas mums . . . naudinga . . . (‘We see only that which is useful for us’) Zero complementation can result in an absolute structural identity of predicative and parenthetical use of different forms of perception verbs. Disambiguation is possible only by context, punctuation, intonation: Predication (= Perception) Atrodo laimingas. (‘He looks/seems happy’) Gerai matyti. (‘One can see well’)
Parenthetical (= Modal qualification) vs. Atrodo, laimingas. (‘Seems/apparently he’s happy’) vs. Gerai, matyt(i). (‘Everything’s well, apparently’)
The semantics of the Lithuanian oblique mood is parallel to that of the category of evidentiality (Ambrazas 1977). Neuter forms of active and passive participles (Part) functioning as predicates are used to express modal qualification (hearsay, inference, surprise, uncertainty) that is usually attributable to the modal categories of evidentiality and mirativity: Užkastieji pinigai dega˛PrsActPartPlNom . (‘Buried money is said to burn’) Kaip tavo daug matytaNPstPasPart ! (‘[It’s amazing] how much you’ve seen’)
(Ambrazas 1997: 264) (Balkeviˇcius 1998: 96)
. Not all cases of IPI and IPSoA will be considered in the present paper as direct perception. There will be further subdivision of them, which will depend upon the meaning of the NP denoting perceptual stimulus. . Similarly, the meaning of the verb see in the ‘NP V NP’ type structures (cf. Dik & Hengeveld: IPI) is also very much dependent upon the type of NP that is chosen both in the Subject and Direct Object position. The same shift towards physically indirect but mentally direct perception occurs when the Subject or DO position is taken by abstract nouns, as in the following examples: When a woman looks in the mirror, she sees the totality of her being: because of the social brainwashing to which she has been subjected, . . . The seventh century sees the appearance of leadership by birth. Melsvu˛ žmonos akiu˛ gelm˙ej Arvydas mat˙e li¯udesi˛ ir kart˙eli˛ . . . (‘In his wife’s blue eyes Arvydas saw sadness and bitterness’) Mat ši gerašird˙e moteris . . . visoje šioje istorijoje mato daugyb˛e sutapimu˛. (‘This kind-hearted woman . . . in this story sees many coincidences’) . In the BNC as is tagged as an adverb, conjunction and a preposition, for instance:
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.25 (217)
Extension of meaning
Don’t see me as/CJS-PRP/ a spokesperson for anything . . . . . . some people see it as/AVO/ instant answer to every problem . . . . . . informal carers are seen as/AVO/ able to give assistance. . In some of the figures/tables I have used the notation of the Basic Tagsets used for word class annotation of the whole BNC, thus NPs cover all the expressions where the head word is a noun or pronoun, e.g.: AT0+N.*, AJ0+N, PNP, N.*, AV0+AJ0+N.*, etc. . There are but a few cases when the verb SEE in the given type of structures with asphrases is used to describe direct visual perception or can come to characterize the Experiencer, as in the examples below: The inaccuracy is possibly due to choosing too high an exposure, so that some black cells are seen as white. I think he saw life as a theatre-goer . . . . As a highly inflectional and synthetic language Lithuanian seems to prefer casemarking semantics and chooses Bare Object Complements instead of IOCs to describe cases of non-sensory perception (DI). The transitive MATYTI (‘see’) can be used to describe DI when followed by BOCs that contain complex NPs with a headword in the Instrumental case complemented to the Direct Object (MATYTI – NPACC – NINS P), as in the following examples: . . . prisipažino nori˛s Roka˛ matyti daniško ir kanadietiško tipo žemdirbiu. confess3Pst wantPtcNOM RokasACC seeInf Danish and Canadian type farmerINS (‘He confessed he wanted to see Rokas as a Danish and Canadian type of farmer’) . Very often the descriptions can focus on the causes effecting the given distortions in the appearance of the stimulus, for instance: The terrorist bombs and politically linked gang violence in Bombay in March have made Congress seem incapable of preserving law and order. At times his language turned crude and made him look foolish. Their training and the pressure to conform to state rules and regulations have over the years made them appear to their pupils, and many parents, as agents of the regime.
References Aijmer, K. (1980). Evidence and the Declarative Sentence. Stockholm: Almqvist and Wiksell International. Aijmer, K. (1996). Swedish Modal Particles in a Constractive Perspective. In K. Jaszczolt & K. Turner (Eds.), Contrastive Semantics and Pragmatics: Meanings and Representations, vol. 1 (pp. 393–427). Oxford, N.Y., Tokyo: Pergamon. Aijmer, K. (1997). I think – an English Modal Particle. In T. Swan & O. J. Westvik (Eds.), Modality in Germanic Languages (pp. 1–47). Berlin, N.Y.: Mouton de Gruyter. Ambrazas, V. (1977). Netiesiogin˙es nuosakos (Modus Relativus) paplitimas ir kilm˙es problema. Lietuviu˛ kalbotyros klausimai, 17, 7–54.
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.26 (218)
Aurelia Usonien˙e
Ambrazas, V. (Ed.). (1997). Lithuanian Grammar. Vilnius: Baltos lankos. Asher, N. (1993). Reference to Abstract Objects in Discourse. Dordrecht, Boston, London: Kluwer Academic Publishers. Austin, J. L. (1962). Sense and Sensibilia. Oxford: Clarendon. Balkeviˇcius, J. (1998). Lietuviu˛ kalbos predikatiniu˛ konstrukciju˛ sintaks˙e. Vilnius: Mokslo ir enciklopediju˛ leidybos institutas. Barron, J. (1997). LFG and the History of Raising Verbs. In M. Butt & T. H. King (Eds.), Proceedings of the LFG 97 Conference. CSLI Publications. Biber, D., Johansson, S., Leech, G., Conrad, S., & Finegan, E. (1999). Longman Grammar of Spoken and Written English. Harlow: Longman. Borkin, A. (1973). To Be and Not To Be. CLS, 9, 44–56. Bolinger, D. (1991). Parecer. The Syntax of Parecer. In J. H. Silverman (Ed.), Essays on Spanish. Words and Grammar (pp. 25-44). Newark, Delaware: Juan de la Cuesta. Bybee, J., Perkins, R., & Paglucia, W. (1994). The Evolution of Grammar. Chicago & London: The University of Chicago Press. Chafe, W. (1986). Evidentiality in English Conversation and Academic Writing. In W. Chafe & J. Nichols (Eds.), Evidentiality: The Linguistic Coding of Epistemology (pp. 261–272). Norwood: Ablex. Croft, W. (1993). Case Marking and the Semantics of Mental Verbs. In J. Pustejovsky (Ed.), Semantics and the Lexicon (pp. 55–72). Dordrecht, Boston, London: Kluwer Academic Publishers. Declerck, R. (1991). A Comprehensive Descriptive Grammar of English. Kaitakusha. Defrancq, B. (1996a). On the Form of French, English and Dutch Object Complements. Quarterly Newsletter of the Contrastive Grammar Research Group of the University of Gent No 5. Defrancq, B. (1996b). Object Complements in English, French and Dutch: Some Observations. In A.-M. Simon-Vandenbergen et al. (Eds.), Aspects of Contrastive Verb Valency (pp. 125–143). Gent: University of Gent. Dik, S. C., & Hengeveld, K. (1991). The Hierarchical Structure of the Clause and the Typology of Perception-verb Complements. Linguistics, 29, 231–259. Dixon, R. M. W. (1995). Complement Clauses and Complementation strategies. In F. R. Palmer (Ed.), Grammar and Meaning (pp. 175–220). Cambridge: Cambridge University Press. Duffley, P. (1992). The English Infinitive. London and New York: Longman. Felser, C. (1998). Perception and Control: A Minimalist Analysis of English Direct Perception Complements. JL, 34, 351–385. Frajzyngier, Z., & Jasperson, R. (1991). That-clauses and Other Complements. Lingua, 83, 133–153. Frawley, W. (1992). Linguistic Semantics. Hillsdale et al.: Lawrence Erlbaum Associates. Gisborne, N. (1998). The Atrributary Structure, Evidential meaning, and the Semantics of English SOUND-class verbs. UCL Working Papers in Linguistics, 10. Givon, T. (1982). Evidentiality and Epistemic Space. Studies in Language, 6, 23–49. Grumadien˙e, L., & Žilinskien˙e, V. (1997). Frequency Dictionary of Modern Written Lithuanian. Vilnius: Mokslo aidai.
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.27 (219)
Extension of meaning
Levin, B. (1993). English Verb Classes and Alternations: A Preliminary Investigation. Chicago: The University of Chicago Press. Lyons, J. (1977). Semantics. Cambridge: Cambridge University Press. Lyons, J. (1991). Semantic Ascent: A Neglected Aspect of Syntactic Typology. In D. Arnold et al. (Eds.), Essays on Grammatical Theory and Universal Grammars (pp. 153–186). Oxford: Clarendon Press. Mackenzie, L. J. (1996). English Nominalizations in the Layered Model of the Sentence. In B. Devriendt et al. (Eds.), Complex Structures. A Functionalist Perspective (pp. 325–353). Berlin, N.Y.: Mouton de Gruyter. Mair, Ch. (1990). Infinitival Complement Clauses in English. Cambridge: Cambridge University Press. Mithun, M. (1986). Evidential Diachrony in Northern Iroquoian. In W. Chafe & J. Nichols (Eds.), The Linguistic Coding of Epistemology (pp. 89–112). Ablex: Norwood. Noël, D., & Simon-Vandenbergen, A.-M. (1996). English AS, French COMME and Dutch ALS Introducing Objects Complements: A Metagrammatical Comparison. In A.M. Simon-Vandenbergen et al. (Eds.), Aspects of Contrastive Verb Valency (pp. 145– 164). Gent: University of Gent. Noonan, M. (1987). Complementation. In T. Shopen (Ed.), Language Typology and Syntactic Description, vol. 2 (pp. 42–140). Cambridge: Cambridge University Press. Nuyts, J. (1993). Epistemic Modal Adverbs and Adjectives and the Layered Representation of Conceptual and Linguistic Structure. Linguistics, 31, 933–969. Palmer, F. R. (1988). The English Verb. London and N.Y.: Longman. Palmer, F. R. (1990). Modality and the English Modals. London and N.Y.: Longman. Poldauf, I. (1968). Evaluative Predication. Philologica Pragensia, 11, 1–12. Poldauf, I. (1972). Factive, Implicative, Evaluative Predicates. Philologica Pragensia, 15, 65–92. Poutsma, H. (1926). A Grammar of Late Modern English. Part II. Groningen: P. Noordhoff. Quirk, R. et al. (1985). A Comprehensive Grammar of the English Language. London and N.Y.: Longman. Ransom, E. N. (1986). Complementation: Its meanings and Forms. Amsterdam/Philadelphia: John Benjamins Publishing Company. Riddle, E. (1975). Some Pragmatic Conditions on Complementizer Choice. CLS, 11, 467–474. Rogers, A. (1971). Three Kinds of Physical Perception Verbs. CLS, 7, 206–222. Rooryck, J. (1997). On the Interaction Between Raising and Focus in Sentential Complementation. Studia Linguistica, 51, 1–49. Schlesinger, I. M. (1992). The Experiencer as an Agent. Journal of Memory and Language, 31, 315–332. Sweetser, E. (1993). From Etymology to Pragmatics. Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Tasmowski, L. (1989). A Look at sembler + Infinitive from Different Angles. In D. Jaspers et al. (Eds.), Sentential Complementation and the Lexicon (pp. 403–438). Dordrecht/Providence: Foris Publications. Traugott, E. C. (1989). On the Rise of Epistemic meanings in English: An Example of Subjectification in Semantic Change. Language, 65, 31–55.
TSL[v.20020404] Prn:17/10/2002; 14:36
F: PBNS11.tex / p.28 (220)
Aurelia Usonien˙e
Usonien˙e, A. (1999). Perception Verbs Revisited. Working Papers, 47, 211–225. Lund: Lund University. Usonien˙e, A. (2000). On the Modality of the English Verbs of Seeming. Belgian Journal of Linguistics, 14, 183–205. van der Auwera, J., & Plungian, V. A. (1998). Modality’s Semantic Map. Linguistic Typology, 2, 79–124. Verspoor, M. (1996). The Story of -ing: A Subjective Perspective. In M. Putz & R. Dirven (Eds.), The Construal of Space in Language and Thought (pp. 417–454). Berlin, N.Y.: Mouton de Gruyter. Viberg, Å. (1984). The Verbs of Perception: A Typological Study. In B. Butterworth et al. (Eds.), Explanations for LanguageUniversals (pp. 123–162). Berlin, N.Y., Amsterdam: Mouton Publishers. Viberg, Å. (1993). Crosslinguistic Perspectives on Lexical Organization and Lexical Progression. In K. Hyltenstam & Å. Viberg (Eds.), Progression & Regression in Language (pp. 347–385). Cambridge: Cambridge University Press. Wierzbicka, A. (1980). Lingua Mentalis. Sydney: Academic Press. Woodbury, A. C. (1986). Interactions of Tense and Evidentiality: A Study of Sherpa and English. In W. Chafe & J. Nichols (Eds.), Evidentiality: The Linguistic Coding of Epistemology (pp. 188–202). Norwood: Ablex. Zinkeviˇcius, Z. (1981). Lietuviu˛ kalbos istorin˙e gramatika, t. 2. Vilnius: Mokslas.
TSL[v.20020404] Prn:26/08/2002; 12:04
P V
Perspectives on eventualities
F: PBNSP5.tex / p.1 (221)
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.1 (223)
Information structure, argument structure, and typological variation* Márta Maleczki University of Szeged, Hungary
.
Introduction
. The aim of the paper This paper presents arguments in support of the assumption that there are universal constructions in human languages, possibly used for different purposes, depending on the characteristics of the particular language. Thus typologically distinct languages can show parametrical variation along the same dimension, in the sense that they use similar structures, but with different functions. The two languages to be contrasted here are English and Hungarian, and it will be shown that in these typologically fairly distinct languages the very same structures can be discovered, and the differences in the configurationality of these languages can be brought in connection with the different aims these common structures are used for. In English, the basic structural varieties are used to express differences in the argument-structures of verbs; in Hungarian, the same structural alternatives express variations in the information structure of sentences. Particularly, some basic Hungarian sentence-types will be discussed, and the analyses offered here will support the long-standing observation that the actual form of the sentences may affect (or may be affected by) the discourse function of sentential constituents, or the type of judgement the whole utterance expresses. Put differently, the discourse function of the sentential constituents can serve as the basis of sentence structuring. I will argue that the very same structures can be used for making differences between the two basic judgement-types in Hungarian as the ones are used for explaining the differences occurring in the possible argument-structures in English. In this way, the results presented here will support the assumption that there may be the same structures present in typologically different languages.
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.2 (224)
Márta Maleczki
A novelty for Hungarian descriptive linguistics is that it will be shown that the structural distinction between the two basic judgement types (thetic and categorical) is more systematic and influential than is usually assumed, and this distinction may form the common background when accounting for several, seemingly unrelated descriptive issues. The most extensively studied syntactic structure/information structure parallel in Hungarian is built around the focused/neutral opposition. In this paper, however, the emphasis will be laid on the fact that neutral sentences belong to two basic structure-types as well, distinguished according to the thetic/categorical opposition. The basic assumption of the paper is that the analysis presented by Ritter and Rosen (1998) is on the right track when offering a syntactic solution to the problems raised by variations in the argument-structures of verbs. The authors argue that there are two basic syntactic structures: one is reserved for the so-called delimited events exclusively, and the other hosts non-delimited events. Accepting that the interaction of these structures with the lexical semantic properties of verbs yields the possible variations in the argument-structures of English verbs, it will be shown that essentially the same structures are used to distinguish the two basic neutral (non-focused) information structure types in Hungarian. Particularly, I will argue that the construction Ritter and Rosen (1998) calls delimited event structure is associated with the thetic statements in Hungarian (not exclusively, but regularly), while the other structure is connected to categorical judgements exclusively.
. Theoretical and methodological assumptions During the whole argumentation the interaction of lexical semantic, syntactic, and sentence-level semantic-pragmatic factors is strongly emphasized; moreover, it will be demonstrated that the crucial point is exactly this interaction when articulating the relevant empirical generalizations. In this way, the argumentation presented here is fundamentally constructional in its spirit (see e.g. Goldberg 1995; Lambrecht 1994), although the terminology used here does not always reflect this (the literature referred to here is not necessarily constructional). The general theoretical conclusion that can be drawn from the data and descriptions presented here is that strictly modular approaches to natural languages are certainly not capable of discovering such very interesting interdependencies as the ones presented here, so non-modular, declarative and constraint based grammars seem preferable to the ones adhering to some distinguished module as a most basic one, and using input-output rule types.
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.3 (225)
Information structure and argument structure
. The structure of the paper In the next section discourse configurationality will be discussed briefly, exemplified by some Hungarian data. Section 3 will present issues emerging in connection with information structure in general: in the first part the theme–rheme, topic–comment, given– new distinctions will be characterized, and in the second part the main judgement-types will be considered. The fourth section is divided into three subparts: the first one presents (on the basis of Ritter & Rosen 1998) the two basic structures the analyses of the following sections are based on; the second subpart shows the presence of these structures in Hungarian; and the third contrasts some Hungarian and English data. In Section 5 the two basic structures will be used for explaining the differences in the information structure of Hungarian neutral sentences. Section 6 summarizes the main results.
. Discourse configurationality It has become well-known by now that there are so-called discourse configurational languages, in which the different discourse-semantic functions (topic, focus) serve as the main sentence-constituting factors (see the Introduction of É. Kiss 1995 and the references cited there). The fundamental difference between discourse configurational (exemplified here by Hungarian) and subject-prominent (here English) languages is that while sentences of the latter ones can be characterized with the order – or, more exactly, structural position – of their syntactic constituents (that is, with the relative positions of the main verb and its arguments), discourse-configurational languages display a seemingly “free” word-order. However, this freedom is only present if we consider purely syntactic and morphological phenomena, like word-classes and cases, and it disappears if we take into account the information structure of sentences: (1) a.
Alice caught the cat/her. subj verb obj b. *The cat/her caught Alice. *obj verb subj
(2) a.
Alice megfogta a macskát. Alice pfx-caught the cat-acc. subj verb obj b. A macskát megfogta Alice. the cat-acc. pfx-caught Alice obj verb subj
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.4 (226)
Márta Maleczki
c.
Megfogta Alice a macskát. pfx-caught Alice the cat-acc. verb subj obj
Considering only the relation of the arguments to the verb, (2a), (2b), and (2c) are the same sentences, expressing the proposition that Alice stands in the caught relation to the cat (that is, all the statements expressed by (2a–c) are true if and only if Alice caught the cat). However, the logical structure or the information structure of these sentences are different: (2a) expresses something about Alice, answering the question What did Alice do?; (2b) gives information on the cat (to the inquiry What happened to the cat?); and (2c) is a statement describing a situation, answering the question What has happened? Consequently, although arranging the arguments in different orders in Hungarian leaves their grammatical roles intact, the configuration they appear in influences heavily the information structure of the sentence (or conversely; the different pieces of linguistic information are present simultaneously, and I assume no priority of any level or module, see Section 1.2). In sum, discourse configurationality means in general that topic, focus or both are associated with a particular structural position (É. Kiss 1995). In Hungarian, the fact that the topic and the focus have a fixed position in the sentence has been amply demonstrated in É. Kiss’ works (see e.g. É. Kiss 1987, 1994). However, it is striking that the research into the structure of Hungarian concentrated on the differences between sentences having or lacking a focused constituent, and the contrast between sentences with and without topic, although noticed by some authors (e.g. Kálmán 1985; É. Kiss 1995; Molnár 1998), has not been studied as extensively as it deserves. The main concern of the present paper will be this contrast, and issues related to sentences having a focused constituent will not be addressed; only the neutral (non-focused) Hungarian sentences will be studied here.
. Information structure . Theme, topic, and “given” According to Östman and Virtanen (1999), the information structure of sentences can be analysed from the point of view of three related, but different perspectives: theme–rheme, topic–comment, given–new. The authors claim that although the phenomena these terms describe may partly overlap, the three pairs of terms are to be considered as representing three different pillars of the same bridge. This bridge is the mediating device between grammar and use: the information structure. The three dichotomies express three different motivations, each of them affecting somehow the form of individual sentences (and larger texts as well); but
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.5 (227)
Information structure and argument structure
to what extent and in what ways sentence-forming of a particular language is influenced by these, are largely language-specific issues. The interpretation of the three dichotomies offered by Östman and Virtanen (1999) is given languageindependently, so it can serve as a good starting point for a contrastive analysis. The first pillar, the theme– rheme distinction expresses structural constraints on the sentence form, and represents the speaker’s point of view. The speaker has to set out from something when conveying information, and it is the theme that (s)he starts with: this is the “peg” the whole remaining part of the sentence will be hung on. Östman and Virtanen (1999) defines the theme as the part of the sentence that occupies the very first position, this being the crucial information the content of all remaining material in the sentence is to be attached to. The topic–comment terms, on the other hand, are used from the perspective of the communicative interaction: the topic expresses what the utterance is about from the listener’s point of view. It may or may not be assigned to a structural position of the sentence, depending on the sentence-constituting parameters of the language; however, its default place is at the beginning of the sentence crosslinguistically. The third pillar of the information-structure “bridge” is the given–new distinction, and it can be regarded as based on a cognitive notion: the accessibility in the memory. Accessibility depends on the extent to which some information is activated in the memory of the communicating parties, so the given–new distinction is gradual in nature. Typically, presuppositions are the kind of information that is (supposed to be) “given”, and new information is typically expressed by the focus. This cognitive perspective of information structuring usually puts constraints onto the larger text the sentence occurs in. It would be an interesting task to examine how the instantiations of these dichotomies coincide with or differ from each other in the typologically different languages, and to what extent and in which ways they are differentiated linguistically (if at all). The mere existence of discourse configurational languages (as opposed to languages where the arrangement of the parts of sentences reflects the grammatical structure (that is, some kind of argument-hierarchy)) shows that information structuring dichotomies are instantiated in different ways in different languages. Thus it is sensible to ask which are the universal characteristics of information structuring, which instantiations follow from the typological character of the language, and which are the presumably language specific properties. However, answering these questions would lead us too far; so let us restrict ourselves to the following generalizations. In discourse-configurational languages the topic–comment and/or the given– new distinctions can put constraints on the structure of the sentences. Narrowing down the discussion of these phenomena to the linguistic data to be examined here, it seems that the (sentence-structural notion) theme and the (utterance-
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.6 (228)
Márta Maleczki
constraining notion) topic systematically overlap in neutral (non-focused) Hungarian sentences. This is not so in English, although the default position of the topic is sentence-initial; however, what here really organizes the structure of sentences is the argument-structure of the (main) verb (and the argument-hierarchy), and discourse functions are not really important when arranging the constituents of a sentence in a certain order.1 The next issues to be examined are the ones that are sentence-constituting factors in Hungarian and English: the main types of judgements (Section 3.2), and the structures the predicates and their arguments may occur in (Section 4).
. The main types of judgements In their recent works several linguists tend to take advantage of the philosopher Brentano’s hypothesis according to which there are two fundamental judgement types: the categorical and the thetic (see e.g. Sasse 1991; Ladusaw 1994; Lambrecht 1994, and the literature referred by them).
.. Categorical judgements The logical ground of forming judgements is the predicative relation. The most simple manifestation of this relation is predicating a property of an individual: (3) Amanda is happy. In these statements – called categorical – the logical subjects must denote wellidentifiable individuals, because if we want to decide whether a judgement is true or false in a given situation, it is necessary to know the individuals the statement is about. The part of the sentence whose referent is the logical subject of the proposition does not necessarily coincide with the grammatical subject: (4) Hugót ma megvendégelték a barátai. Hugh-acc. today pfx-entertained the friends-his ‘Hugh was entertained by his friends today.’ The logical subject of (4) is Hugh, he is the person the statement is about, and the proper name referring to him is the grammatical object in the Hungarian sentence. As we have seen in Section 3.1, the linguistic expression whose referent functions as the entity the statement is about is called the topic of the sentence (independently of the grammatical role it plays). In Hungarian, where syntax is determined mainly by informational, not by grammatical units, the constituent interpreted as the topic of the utterance has to be placed at the very beginning of the sentence, before the other fixed positions (those of the quantifier phrase(s) and the focus). Taking the informational character of that syntactic position seriously, the constituent occu-
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.7 (229)
Information structure and argument structure
pying the topic position must always be interpreted as the topic of the discourse (cf. Lambrecht 1994: 130). Consequently, the expression occurring in that position must have semantic properties that make it suitable for serving as the ground for the main predication the utterance expresses. Identifiability of the topic of categorical statements is required because of the very nature of this type of judgements: they are double judgements in the sense that the referent of their logical subject (topic) has to be identifiable independently of the statement itself, as it is the part of the utterance that gives the ground for the predicate. We have seen that identifiability is definitely a cognitive notion, so these observations lead us to the conclusion that in Hungarian categorical judgements the theme of the sentence (as defined by Östman & Virtanen, 1999) is regularly the same as the topic of the utterance, and its state is activated or semi-activated (or at least it is supposed to be so); that is, its place is on the “given end” of the given–new scale. The consequences of this triple coincidence of the informational perspectives are pervasive and can give explanations to various, seemingly unrelated syntactic and semantic issues. For example, if an indefinite NP occupies the syntactic topic position, it must be specific because of the identifiability requirements of the categorical statements. However, the semantic properties of the relevant parts of the sentence may not allow the specific interpretation, and in that case, the indefinite NP cannot become specific simply by virtue of the “topic position” (as is usually assumed in works on Hungarian). If we cannot anchor the referent of an indefinite NP that occurs apparently in the syntactic topic position, and the sentence is acceptable, then the indefinite NP, being unable to function as discourse topic, is NOT in the topic position (even if it SEEMS to be the sentential topic). There being no topic in the sentence, it cannot express a categorical judgement: it will belong to the other main judgement type, called thetic.
.. Thetic judgements As we have seen, categorical judgements can be divided into a logical subject and a predicative part; but there are statements that cannot be partitioned in that way: thetic judgements are logically non-analysable descriptions. Non-analysability does not mean that it is impossible to make a difference between the grammatical parts of the sentence; but none of these parts can refer to an entity fit for the logical subject, that is, the topic function.2 As no other possible discourse topic is available, a thetic statement is a predication on the SITUATION it is uttered in. Hungarian, having an information structure based syntax, may omit the grammatical subject in these sentences altogether. Characterizing this judgement-type with the dichotomies presented in Section 3.1, it can be said that in these sentences the theme and the comment coincide in Hungarian, because it is the comment (the new information) that occurs in the first place. In English, syntax being organized according to the syntactic arguments of the main verb, this seems impossible. Here, be-
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.8 (230)
Márta Maleczki
side the intonation, special, expletive constructions show the thetic character of the statement, and the expletive expressions play the role of the theme in the sentence: (5) Havazik. snows ‘It is snowing.’ (6) Vannak kispandák az állatkertben. are red pandas the zoo-in ‘There are some red pandas in the zoo.’ (7) Bejött néhány kutya. in-came some dog ‘Some dogs came in.’ None of these and similar Hungarian sentences have topics: all of the arguments (if there are any at all) FOLLOW the verb. As the topic is always at the very beginning of the Hungarian sentences, the arguments occurring in (6) and (7) belong to the predicative part of the sentence. Therefore, (5)–(7) express thetic judgements (in accordance with the native speakers’ semantic intuition). Summarizing the observations made so far, the train of reasoning is the following. A categorical statement must have a topic by definition. The discourse referent playing the role of the topic has to be (to some extent) identifiable, that is, the topic NP is always specific. It follows then, that definite NPs, being inherently specific, are always good candidates for the topic position; but a discourse referent introduced by an indefinite NP becomes specific only if it is anchored to the context somehow. The most striking and unavoidable part of the context is the main predicate of the statement itself. Consequently, my claim is that the sentential predicates play the very same role in organizing the structure of sentences in Hungarian as in English; but the structures they are licensed to occur in are related to the information structure of sentences in Hungarian, and not to the argument-configurations (as in English).
. Event-structure The key notion when trying to discover the common structures used with language-specific functions in English and Hungarian will be the event-structure. This term, however, is ambiguous in several ways in the literature, so the very first task is to make clear how it will be used here. For the present purposes, the notion of event-structure used by Ritter and Rosen (1998) seems to be the most useful device. The reason is that the authors define event-structure as a syntactic notion; although they do not deny that the
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.9 (231)
Information structure and argument structure
lexical properties of particular verbs play some role when interpreting sentences, this role is not considered definitive, even when trying to explain the licit and illicit argument-configurations a verb can occur in. Ritter and Rosen present convincing arguments in their paper to support their opinion by which the structure and the interpretation of sentences are not lexically driven in general; the lexicon only contributes some information to the compositionally interpretable event-structure. Being so, syntactic information is crucial when calculating the whole interpretation of the sentence, or even when giving the exact meaning of verbs. Of course, this view is not without predecessors; but the form Ritter and Rosen offer, and the phenomena they describe with the aid of their fundamentally nonlexical approach, seems very promising with respect to the kind of problems raised here and the holistic view argued for. Thus we have to dwell on that paper in order to be able to exploit the possibilities it offers for discovering the similarities and differences between English and Hungarian.
. Result-oriented predicates and syntactic structure The aim of Ritter and Rosen (1998) is to discover the factors determining whether a certain argument-structure for a given verb is licit or illicit (in English). Their theoretical starting point is that the possible variations in argument-structures cannot be explained solely on the grounds of the lexically given properties of verbs. They argue that the various argument-structures and the differences in their meanings are fundamentally compositional in nature: they can be calculated from the meaning of the verb and its arguments, provided we know how they are put together. Notice, for instance, the following contrast (see Ritter & Rosen 1998: 135, (2)): (8) a. *John walked the letter. b. John walked the letter to the dean’s office. Examining various types of the relevant data, the authors conclude that the variability in the argument-structure may regularly depend on the presence or absence of other constituents in the sentence.3 On the basis of these observations Ritter and Rosen (1998) repeatedly emphasize the relevance of the syntactic construction(s) the different verbs may enter: they argue that the syntactic structure can be definitive in licensing certain arguments, and it often happens that even the exact meaning of the verb is computed post-syntactically. In order to solve the problems raised by Ritter and Rosen (1998), two basic syntactic structures are assumed: one is reserved for so-called delimited events exclusively, and the other hosts non-delimited events. I will refer to these structures as L-structure and A-structure, respectively (although this terminology departs from what is used by Ritter & Rosen 1998).4 Here event-structure is a purely syntactic
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.10 (232)
Márta Maleczki
notion, and it is equated with a specific construction (L-structure). The essential property of the L-structure is that it has two distinguished functional nodes supplied with special roles: one functional node is assigned the delimiter event-role (Largument in the following), the other is assigned the initiator event-role. The most important property of the L-structure is that it has an L-argument (delimiter) obligatorily, independently of whether the verb is intransitive or not. The lexical feature that invariably selects the L-structure when the verb is entering a construction is the result-oriented meaning. However, delimited event interpretations do not arise simply by virtue of that lexical property, since “the assignment of an initiator or a delimiter is not determined in the lexical representation of the verb, but rather is determined in the syntax” (Ritter & Rosen 1998: 138.) For instance, although activity verbs, being atelic, have no delimited event-structure lexically, they can occur in the L-structure if some other constituent of the sentence licenses it. This explains why (8b) is good as opposed to (8a): the goal phrase makes the activity result-oriented, so the verb may appear in the L-construction with the object argument having the delimiter event-role. Predicates that are lexically result-oriented are confined to the L-structure automatically (see (9), (10)), but it is important that this is not the only possibility for evoking this construction (see (8b), (11)). (9) Amanda opened the window. (10) The window opened. (11) Bill danced Sue across the room. From the obligatory presence of the L-argument in the L-construction it follows that the only argument of result oriented intransitive verbs, these being specified for the L-structure lexically, always will be an L-argument. This is exemplified in (10): the only argument is the subject, but it cannot have the initiator (or any other) role; it has the delimiter event-role, by virtue of its L-argument position in the Lconstruction, which construction is invariably selected by the meaning of the verb. In (9) and (11), however, the L-argument position (licensed by the adverbial phrase in (11)) is occupied by the grammatical object (the window, Sue), and the subject (Amanda, Bill) acquires the initiator event-role. In the other type of syntactic constructions (A-structure) there is no special event-structure, so there are no event-roles, either: the usual theta-roles and functional projections can work. This structure is evoked for instance when there is no L-construction licensing adverbial phrase in the sentence containing an (intransitive) activity verb (see (12), (13)). These sentences do not meet any L-structure licensing condition, so they can only appear in the A-construction, where the lexically determined theta-role (agent) is given to the subject:
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.11 (233)
Information structure and argument structure
(12) John walked. (13) Sue danced. In sum, when accounting for the possible argument-structures of a verb, its ability to appear in the A- or the L-construction is a determining factor. Ritter and Rosen (1998) argue that there are verbs that are not specified lexically with respect to their possible argument-structures (weak verbs), and their meanings are also variable according to the constructions they occur in (especially activities show this property, cf. (8), (11), (12), (13)). Weak verbs are underdetermined lexically as they allow more interpretations and argument-configurations, and the structure they enter will determine compositionally which argument-configurations are licit and which are illicit for them. Consequently, in the case of weak verbs it is impossible to define either their exact meaning(s) or the number and roles of their arguments relying purely on their lexically given properties. Note, however, that there are verbs that do have a well-delineated meaning and invariant argument-structure lexically (strong verbs: mainly stative verbs, specified for A-structure, and achievements, specified for the L-structure).
. L-structure in Hungarian I suppose that in Hungarian the difference between the L-structure and A-structure is present in the same way as in English: there are certain verbs that are lexically specified for the L-structure, so they appear in the L-construction obligatorily; there are other verbs that are confined to the A-structure; and there are verbs that can appear in the A-structure, their meaning being lexically not result-oriented, but L-structure can also be available for them if there is a suitable delimiting constituent in the sentence. Recall that the crucial feature of the L-structure is that it must contain an L-argument, in a particular position with a particular aspectual (delimiting) interpretation. The details of the interpretation and the exact position of the L-argument do not concern us now; it is enough for our present purposes that the L-argument position is under the V , in the Spec of a delimiting functional phrase (see Ritter & Rosen 1998). The simple starting assumption will be here that in Hungarian the position the L-argument occupies is the same as the position of the so-called verbal modifiers (whose exact placement is also an open question of Hungarian syntax, partly depending on the theory used for the description).5 Anyhow, identifying the position of verbal modifiers with the place of the L-argument opens the way to describing in a systematic way the interdependencies between the syntactic structure, the information structure, and the lexical-semantic properties of the verb (the main predicate). In this way, it becomes possible to gain a deeper insight into such longstanding puzzles of Hungarian linguistics as e.g. the definiteness restriction, the
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.12 (234)
Márta Maleczki
licit and illicit occurrence of bare common noun arguments, and the possible interpretations of indefinite subjects (see the details in Maleczki 2001). The shared properties of these issues are that they are not independent of the lexical properties of the verb (the main predicate), while at the same time syntactic explanations seem plausible as well; moreover, the whole thing seems to be connected with the interpretation of the arguments. Now, recall that the very same story can be told about the argument-structure alternations in English, and these problems can be arguably solved by differentiating between the L-structure and the A-structure. So I claim that a common starting point can be used for English and Hungarian: . . . “what distinguishes predicates syntactically and semantically is the difference between delimited events and all others” (Ritter & Rosen 1998: 137).
. Similarities and differences in the L-structures of English and Hungarian As described above, delimited events have a syntactically encoded event-structure (L-structure), where an L-argument (confined to a certain syntactic position and having a delimiting event-role) is obligatorily present. My claim is that these two kinds of basic syntactic structures are present in English and Hungarian alike; but the ways these structures are used are different, depending on the idiosyncratic properties of these languages. The most striking difference, the argument-structure prominence versus information-structure prominence is clearly reflected in the different “interpretation” of these common structures. In order to enlighten the similarities and dissimilarities, let us see some data contrasted: English Hungarian the delimiting argument-position (L-argument) is filled obligatorily with direct object (9) Amanda opened the window
verbal prefix (14) Amanda kinyitotta az ablakot. Amanda pfx-opened the window-acc. secondary predicate
(15) Amanda opened the window wide.
(16) Amanda szélesre nyitotta az ablakot. Amanda wide opened the window-acc. argument without any determiner (17) Amanda ablakot nyitott. Amanda window-acc. opened ‘Amanda opened/was opening the window.’
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.13 (235)
Information structure and argument structure
English
Hungarian
activities have no inherent event-structure, but allow the addition of a secondary predicate if the activity is located, L-structure; with a delimiting argument intransitive verbs: thetic judgements transitive alternation is licit because of a delimiting goal phrase (18)*Bill danced Sue.
(19) *Tündér táncol. fairy dances (11) Bill danced Sue across the (20) Tündér táncol a tó tükrén. room. fairy dances the lake mirror-its-on ‘A fairy is dancing on the surface of the lake.’ (21) *Csalogány dalolt. nightingale warbled (22) Csalogány dalolt az ablakomban. nightingale warbled the window-my-in ‘A nightingale was warbling in my window.’ In the first set of examples an inherently result-oriented verb (open) is the main predicate of the sentences. On the basis of the argumentation given by Ritter and Rosen (1998) these predicates occur in the L-structure. Assuming the same structures for Hungarian, the verb nyit ‘open’ being similarly result-oriented, it also has to occur in the L-structure. However, Hungarian has the additional languagespecific property that the position immediately before the verb is reserved for a special class of words, the verbal modifiers.6 These belong to various word classes: a verbal modifier can be a prefix (see example (14)); an adverb; a secondary predicate (see example (16)); or a bare noun (see (17)). Notice that in Hungarian the non-agentive argument ((az) ablak ‘the window’), contrary to the English examples, cannot occupy the L-argument position if it has a definite determiner (at least in neutral sentences with no special interpretation): (23) *Amanda minden/az ablakot nyitott(a). Amanda every/the window-acc. opened This phenomenon can be explained with the special interpretation attached to the verbal modifier (L-argument) position in Hungarian: verbal modifiers always have a special, property-like interpretation. Consequently, if an argument is to occupy that place, it has to loose its “argument-like” character. This is possible if it is not definite, that is, if it does not have a definite or specific determiner. Therefore, bare
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.14 (236)
Márta Maleczki
common nouns seem to be good candidates for the verbal modifier position, since they have inherently a property-like interpretation; that is why they can occupy the L-argument position while a definite or specific NP is excluded from there.7 In sum, the L-argument-position is to be filled by some material obligatorily (cross-linguistically), and, if there is a verbal modifier present in the sentence, it has to occupy that position (in Hungarian). Bare common nouns can be regarded as verbal modifiers because of their inherent semantic properties, so the only place available for them is the L-argument position. This means that they cannot occur in the A-structure: there being no L-argument position, bare common noun arguments are ruled out. Consequently, the whole argumentation concerning the bare nouns can be tested by examples where the main predicate is a verb lexically confined to the A-structure: the reasoning presented so far entails that in these sentences bare common nouns cannot occur. Ritter and Rosen (1998) argue on the basis of English data that stative verbs, being not result-oriented, always occur in the A-structure. We can reasonably assume that the semantics of the stative verbs is the same in Hungarian, so we expect on the grounds of our starting assumption that the same structural possibilities are available for them. This expectation is borne out: it is a fact of Hungarian that stative verbs do not tolerate bare common noun arguments. Examples (24) and (25) serve as illustrations: (24) contains stative verbs, and the sentences are definitely wrong, while example (25) with a result-oriented predicate is perfect. This distributional fact can be explained nicely with the aid of the two kinds of structures the two different types of verbs require: in (24), there is no position compatible with the semantic properties of the bare common noun, so it cannot occur there. (24) *Hugó szendvicset szeret/utál. Hugh sandwich-acc. loves/hates (25) Hugó szendvicset eszik. Hugh sandwich-acc. eats ‘Hugh is eating sandwiches.’ Turning now to the second set of Hungarian examples (see (19)–(22)), the most remarkable fact is that although bare common noun subjects are usually ruled out in Hungarian (see examples (19), (21)), examples (20) and (22) are wholly acceptable Hungarian sentences. It is striking that the only difference between the non-wellformed and well-formed examples is the absence versus the presence of a locative phrase, respectively. As compared with the English examples (8) and (11), the generalization emerges that the same type of licensing condition can be present in Hungarian and English: it seems to be a sound assumption that the licensed entity is an L-argument in English and Hungarian alike. However, in accordance with the different typological character of the two languages, it is an extra argument that
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.15 (237)
Information structure and argument structure
is licensed in English, while a special type of arguments (bare common noun) is licensed in Hungarian. That the licensed argument is really an L-argument can be shown by examples where the bare common noun argument is allowed only in the presence of some adverbial phrase (see (19)–(22)), while other types of arguments, supposedly entering the A-construction, result in acceptable sentences without any locative phrase: (26) A tündérek táncolnak. the fairies dance ‘The fairies are dancing.’ (27) Minden csalogány dalolt. every nightingale warbled ‘All the nightingales were warbling.’ However, it is remarkable that bare common noun subjects are often not allowed even if there is a locative phrase in the sentence: (28) *Seb fáj a kezemen. wound aches the hand-my-on (29) *Lányok megérkeztek. girls pfx-arrived It seems that in Hungarian the verbs licensing bare common noun subjects in the presence of a locative phrase belong to the same class as the ones allowing an extra argument in English: they are intransitive activity predicates. As we have seen, it is an important criterion in Hungarian as well that there be a locative (or temporal) adverb in the sentence, because this constituent licenses the L-structure if the predicate is not result-oriented lexically. It seems that this criterion is enough, but only if the main predicate is an activity verb, in English and Hungarian alike. Notice that the predicates in (28) and (29) are not activities; (28) contains a stative, while (29) an achievement verb. Achievement verbs, however, are usually lexically result-oriented, so the reason why (29) is not acceptable cannot be that it does not have the L-structure (as in the case of (28), where the A-structure is the only possible option for the stative verb). (29) is wrong because the L-argument position is filled with a verbal prefix, so this place is not available for the bare noun; and, as we have seen, bare noun arguments are restricted to that position (by their inherent semantic properties). This is supported by the examples where the same verb has no prefix, thus the bare noun can occupy the L-argument position: (30) Lányok érkeztek. girls arrived ‘Some girls arrived.’
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.16 (238)
Márta Maleczki
The perfect acceptability of this sentence (but only with a non-specific subject interpretation!) shows that it is really the L-argument position that must be available for bare common nouns. In this way the distributional peculiarities of Hungarian predicates and their arguments can be explained with the aid of the same two basic structures as in English. The difference is in what respect these structures are definitive in the two languages: in English, the argument-structure itself depends on them, while in Hungarian the allowed “type” of the arguments of the otherwise invariable argumentstructure is related to these structures. The “type” of the argument is to be understood here as its ability to acquire specific or non-specific interpretation. Specificity, as was mentioned in Section 3.1, is a notion connected to the information structure (more exactly, to the cognitive notion of “givenness”). Consequently, the choice between the L-construction and the A-construction in Hungarian is in a strong relationship with the thetic or categorical character of the judgement the sentence expresses.
. Information structure and syntax in Hungarian: Thetic and categorical judgements The next question is how the linguistic phenomena arguably originating from the L-structure – A-structure distinction can be brought into connection with the information structure as was characterized in Section 3. Let us consider examples (20), (22) and (30) again: (20) Tündér táncol a tó tükrén. fairy dances the lake mirror-its-on ‘A fairy is dancing on the surface of the lake.’ (22) Csalogány dalolt az ablakomban. nightingale warbled the window-my-in ‘A nightingale was warbling in my window.’ (30) Lányok érkeztek. girls arrived ‘Some girls arrived.’ In these examples, it is the subject that fills the L-argument position and there is no other constituent in the sentence preceding it. Being so, the sentence can express only a thetic judgement (in Hungarian): the L-argument belongs to the event-structure, with a special (delimiting) event-role, so its semantic properties are never compatible with the ones the topic of an utterance must have; it does not denote an identifiable (group of) individual(s), by the very nature of the arguments
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.17 (239)
Information structure and argument structure
that are licensed to occur in this position. In sum, thetic statements in Hungarian either have the verb as the theme (see examples (5) and (6) in Section 3.2.2), or the sentence-initial element is an L-argument. Categorical statements, on the other hand, always must have a topic in the very first place, so A-structures will be the main candidates for expressing that kind of judgements (an obligatory L-argument not being required here). This analysis opens the way for an interesting generalization concerning the specificity of arguments in Hungarian. Leaving now out of consideration sentences where the main verb is in the first place (as we have seen, these are undoubtedly thetic statements), it is remarkable that arguments that can receive non-specific interpretation always occur in the L-construction. Put in another way, arguments standing before the verb can acquire the non-specific interpretation only under the functional nodes of the L-construction. It follows then that the indefinite arguments appearing before the verb will have a non-specific interpretation only if the L-construction is licensed (either by the lexical semantic properties of the verb, or by the presence of a locative component, see Maleczki 1999, 2001). In the A-construction the specific interpretations are available exclusively. Non-specific and specific interpretations of Hungarian indefinite NPs are equally available if the indefinite NP is the initiator argument of the L-construction; when it is interpreted as the initiator, it gets a non-specific reading, but when it becomes the topic of the sentence (which is an open possibility for initiators, since only the L-argument position has to be filled obligatorily in the L-construction) it receives specific interpretation. The conclusion is that in Hungarian the NPs that can be interpreted exclusively as specific (definite NPs), appear necessarily in the A-construction. This is the source of the definiteness restriction in Hungarian (see Maleczki 2001). On the other hand, non-specific NPs before the verb are licensed only by the L-structure, as was demonstrated by the unacceptability of sentences (with the neutral intonation pattern) when the locative adverb was absent, see (19), (21) in Section 4.3. However, this does not mean that the non-specific NP must always be the L-argument; it may appear under the other functional node of the L-construction: non-specific interpretation can be assigned to the initiator argument of the L-structure as well, see (31). (31) Néhány béka legyeket fogdos a tóban. some frog flies-acc. is catching the lake-in ‘There are some frogs catching flies in the lake.’ Here the L-argument is the object, so the subject can only occupy the other functional node of the L-construction, which is that of the initiator. The initiator node is not necessarily filled (see Ritter & Rosen 1998), so the subject has the option ei-
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.18 (240)
Márta Maleczki
ther to remain there or to fill the topic position. In the former case the judgement is thetic with a non-specific subject, while in the latter case it will become a categorical statement, having a specific constituent as a topic. This explains that the subject in (31) can be either non-specific or specific. In the latter case, however, the intonation pattern changes; this also supports the claim that the two interpretational possibilities of the indefinite NPs (the specific and the non-specific) originate from two different syntactic (and informational) constructions. In sum, the main claim of this paper concerning the specificity of indefinite arguments in Hungarian is the following: (32) In Hungarian, non-specific arguments can appear either in postverbal position, or under (either of) the functional nodes of the L-construction. The interpretation of the indefinite NP is necessarily non-specific only if it appears as an L-argument, and necessarily specific interpretation arises in the A-construction. From (32) it follows that non-specific arguments cannot appear in the Aconstruction. This is borne out by the data containing stative verbs. As it has already been mentioned, these verbs are typically confined to the A-structure, and they do not tolerate weakly interpreted arguments, indeed: (33) Néhány béka utálja a legyeket. some frog hates the flies ‘Some frogs hate flies.’ (33) allows only the strong, partitive interpretation of the subject, shown by the fact that this statement involves that there are frogs that do not hate flies. In the case of the NPs with an indefinite article the necessarily strong interpretation is even more striking: (34) Egy béka utálja a legyeket. a frog hates the flies (34) is not acceptable with the usual neutral intonation pattern, but we have two options to make this sentence interpretable. One is that we put a heavy stress on the indefinite article, so our frog will be contrasted with other frogs, which is evidently a strong interpretation. The other option is that we try to interpret (34) as a normative-generic statement, in the way (34 ) indicates: (34 ) Egy (rendes) béka (az olyan állat, hogy) utálja a legyeket. a regular frog that such animal that hates the flies ‘A (regular) frog (is such that it) hates flies.’
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.19 (241)
Information structure and argument structure
Generic interpretations belong to the strong, specific interpretations, because they presuppose the identifiability of the kind they refer to (at least to some extent). The presuppositional nature of generics is especially well demonstrable in Hungarian, where the ordinary (non-normative) generic NPs are always expressed by the definite article. As can be expected, the very same specificity constraints arise with the activity verbs as with the stative verbs, if there is no locative adverb in the sentence. In the absence of locatives, activities appear in the A-structure, so either the partitive or the normative-generic interpretation is forced; the former kind of strong interpretation is illustrated by example (35), and the latter by (36): (35) Egy csalogány dalolt (de a többi hallgatott). a nightingale warbled but the others remained silent ‘One nightingale was warbling (but the others remained silent).’ (36) Egy (rendes) tündér (az olyan, hogy) táncol. a regular fairy that such that dances ‘A (regular) fairy (is such that she) dances.’ Summarizing the conditions licensing non-specific arguments, the conclusion is that if the indefinite argument is not in postverbal position (where it can receive either specific or non-specific interpretation, depending solely on the context and the lexical properties of the verb), only the L-construction allows the non-specific interpretation of indefinite arguments. In this way we can successfully explicate the relationship between the meaning of the verbs and the interpretational constraints their arguments have to meet.
. Summary This paper purported to show that common structural properties can be discovered behind as different-looking linguistic phenomena as the possible argumentstructure variations of verbs (English) and judgement types (Hungarian). The two basic structures that can be shown to be present in English and Hungarian alike are used for different purposes, according to the basic typological character of the two languages (whether it is argument-structure prominent (English) or information structure prominent (Hungarian)). It has proved possible to discover the similarities and differences accounted for here only because the joint effect of linguistic information belonging to different levels or modules (lexical semantics, syntax, and informational structure) was taken seriously into consideration during the whole argumentation.
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.20 (242)
Márta Maleczki
Notes * The research reported here was supported by a grant from the Hungarian National Research Fund (OTKA T029590). . A systematic exception to the overlapping of the theme and the topic are the yes-no questions in English (according to the definitions used by Östman & Virtanen 1999): in Did Amanda go to the movie? the theme per definitionem is the sentence-initial did auxiliary, and it is undoubtedly not the topic of the utterance. In Hungarian, if the topic is Amanda, its place is in the first position of the sentence, in the same way as in sentences expressing statements, and the question status is only indicated by the intonation pattern: Amanda elment a moziba? Amanda away-went the cinema-to? . It seems that thetic judgements cannot have focused parts either, because the presupposition induced by the focus plays the role of the discourse topic. Sentences containing a focused constituent necessarily assume some information given previously in the discourse, and since given (or so assumed) information belongs undoubtedly to the ground the utterance is about, it presumably cannot be stated on a focused sentence that it has no discourse topic at all. Therefore, focused sentences cannot be thetic; but it is an open question whether they can be regarded as categorical in general. It might be the case that it simply makes no sense to characterize focused sentences along the categorical – thetic distinction, since their main function is based on a different kind of perspective, see 3.1. . They also refer to the well-known argument presented by Verkuyl (1972) in order to enlighten the crucial role of syntax in determining the event-structure. This observation demonstrates that the semantic properties of certain arguments can change the aspectual property of predicates while others leave it intact: (i)
John ate a hamburger/the spaghetti in his bowl (in an hour/*for an hour).
(ii) John ate spaghetti/hamburgers (for an hour/*in an hour). . The reason for changing the original terminology is to avoid potential misunderstandings that might arise because of the overloaded use of the terms Ritter and Rosen (1998) introduce (they refer to what is dubbed here L-structure as D-event or simply event-structure). . Several different solutions have been offered concerning the exact place of the verbal modifiers in Hungarian: Kálmán et al. (1986) defines their position under the V as the sister of the V; É. Kiss (1994) argues that their best place is under the VP, sister to the V ; Kiefer (1990–1991) regards the verbal modifier and the V0 as a lexical unit, thus places them together under another V0 in the lexicon, while allowing that a V dominates them in the syntax; Szabolcsi (1997) assumes a separate PredOp functional node for the verbal modifiers; É. Kiss (1998) places them in the Spec of an AspP, etc. I do not want to take a stand on this issue; it is enough for making the generalization above that all the proposals place the verbal modifier in the possibly “nearest” place to the verb itself, and assume an especially close semantic relationship between them and the verb they “modify”.
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.21 (243)
Information structure and argument structure . Recall that only neutral sentences are considered here, so we need not care about the question whether this position is the same as that of the focus in the focused sentences. . The characteristic semantic property of Hungarian bare common nouns is their cumulative reference (see Maleczki 1992). This means that they are not specified with respect to the number of the individuals they denote. As common nouns refer to a (structured) set, their interpretation will be similar to a property, as opposed to the interpretation of the “real” arguments referring to individuals in more specific ways. Note, moreover, that examining bare plurals occurring as subjects in English generic and existential sentences, Glasbey (1995) comes to the conclusion that they are to be interpreted as properties. Thus the propertylike interpretation of bare common nouns seems to be their cross-linguistically valid feature (subject to parametrical variation as to whether the singular or the plural is their semantically unmarked form). Evidently, it also depends on the particular language what positions can be occupied by bare nouns and how their interpretation is influenced by their positions.
References É. Kiss, K. (1987). Configurationality in Hungarian. Budapest: Akadémiai Kiadó. É. Kiss, K. (1994). Sentence Structure and Word Order. In F. Kiefer & K. É. Kiss (Eds.), The syntactic structure of Hungarian [Syntax and Semantics, 27] (pp. 1–90). San Diego: Academic Press. É. Kiss, K. (1995). Introduction. In K. É. Kiss (Ed.), Discourse Configurational Languages. Oxford: Oxford University Press. É. Kiss, K. (1998). Verbal Prefixes or Postpositions? Postpositional aspectualizers in Hungarian. In C. de Groot & I. Kenesei (Eds.), Papers from the Amsterdam Conference, Approaches to Hungarian, Vol. 6 (pp. 123–148). Szeged: JATE. Glasbey, S. (1995). A Situation-Theoretic Interpretation of Bare Plurals. Ms., University of Edinburgh. Goldberg, A. E. (1995). A Construction Grammar Approach to Argument Structure. Chicago: The University of Chicago Press. Kálmán, L. (1985). Word Order in Neutral Sentences. In I. Kenesei (Ed.), Data and Descriptions, Approaches to Hungarian, Vol. 1 (pp. 13–23). Szeged: JATE. Kálmán, L., Prószéky, G., Nádasdy, Á., & Kálmán, C. Gy. (1986). Hocus, Focus, and Verb Types in Hungarian Infinitive Constructions. In W. Abraham & S. de Meij (Eds.), Topic, Focus and Configurationality (pp. 130–142). Amsterdam: John Benjamins. Kiefer, F. (1990–1991). Noun Incorporation in Hungarian. Acta Linguistica Hungarica, 40 (1–2), 149–177. Ladusaw, W. A. (1994). Thetic and Categorical, Stage and Individual, Weak and Strong. In M. Harvey & L. Santelmann (Eds.), Proceedings from Semantics and Linguistic Theory IV (pp. 220–229). Ithaca: Cornell University. Lambrecht, K. (1994). Information Structure and Sentence Form. Cambridge: Cambridge University Press.
TSL[v.20020404] Prn:17/10/2002; 14:38
F: PBNS12.tex / p.22 (244)
Márta Maleczki
Maleczki, M. (1992). Bare Common Nouns and Their Relation to the Temporal Constitution of Events in Hungarian. In P. Dekker & M. Stokhof (Eds.), Proceedings of the Eighth Amsterdam Colloquium (pp. 357–375). Amsterdam: Institute for Logic, Language and Computation, University of Amsterdam. Maleczki, M. (1999). Weak Subjects in Fixed Space. Acta Linguistica Hungarica, 46 (1–2), 95–117. Maleczki, M. (2001). Indefinite Arguments in Hungarian. In I. Kenesei (Ed.), Argument Structure in Hungarian (pp. 157–199). Budapest: Akadémiai Kiadó. Molnár, V. (1998). Topic in Focus. Acta Linguistica Hungarica, 45 (1–2), 89–166. Östman, J. A., & Virtanen, T. (1999). Theme, Comment, and Newness as Figures in Information Structuring. In K. van Hoek, A. A. Kibrik, & L. Noordman (Eds.), Discourse Studies in Cognitive Linguistics (pp. 91–110). Amsterdam/Philadelphia: John Benjamins Publishing Company. Ritter, E., & Rosen, S. T. (1998). Delimiting Events in Syntax. In M. Butt & W. Geuder (Eds.), The Projection of Arguments (pp. 135–164). Stanford: CSLI. Sasse, H.-J. (1991). Predication and Sentence Constitution in Universal Perspective. In D. Zaefferer (Ed.), Semantic Universals and Universal Semantics (pp. 75–95). Berlin: Foris. Szabolcsi, A. (1997). Strategies for Scope Taking. In A. Szabolcsi (Ed.), Ways of Scope Taking (pp. 109–155). Dordrecht: Kluwer. Verkuyl, H. J. (1972). On the Compositional Nature of the Aspects [Foundations of Language, Supplementary Series, Vol. 5]. Dordrecht: Reidel.
TSL[v.20020404] Prn:29/08/2002; 16:21
F: PBNS13.tex / p.1 (245)
The network of demotion Towards a unified account of passive constructions Andrea Sansò Department of Linguistics, University of Pavia, Italy
Introduction: The passive in current linguistic theory Although patient topicalization is said to be one of the functions of the passive, passive constructions do not always signal that the patient is topical, and neither do they topicalize the patient by recalling it from the previous discourse and making it the object of subsequent predication. Consider the following Italian1 examples: (1) “Ah, papà, se vedessi . . . ”, gli brucia la famosa gaffe sulle leggi razziali, lo si capisce per come risponde: “Mio nonno ha dovuto firmare le leggi razziali perché sono state firmate prima di lui dal Gran Consiglio e il Gran Consiglio rappresentava l’Italia in quel momento. Le leggi erano fatte da Mussolini . . . ” (La Repubblica, 14.02.2000). “Oh, dad, if you could see . . . ”, the memory of his gaffe about racial laws still rankles, as his answer demonstrates: “My grandfather was compelled to sign the racial laws, for they had already been signed by the Supreme Council, and the Supreme Council represented Italy at that moment. The laws were promulgated by Mussolini” (2) Qui la polizia ha trovato materiale edilizio e macchine agricole rubate per un valore di 80 milioni. Alcuni dei mezzi erano stati reimmatricolati con nuove targhe, e la polizia ora sta lavorando per capire come sono riusciti a farlo. Sono stati trovati e sequestrati anche 11 cavalli. L’indagine prosegue ora via Internet per controllare più aziende possibili nelle due regioni dove la banda aveva avviato il redditizio traffico. (La Repubblica, 12.12.1999) Here the police found 80 million Lira worth of stolen building material and agricultural machines. Some of the machines had been rematriculated with new number-plates, and the police is now trying to understand how
TSL[v.20020404] Prn:29/08/2002; 16:21
F: PBNS13.tex / p.2 (246)
Andrea Sansò
they managed to do this. Also, 11 horses were found and sequestrated. The investigation is now being conducted through the web, in order to inspect as many factories as possible in the two regions where the gang’s profitable traffic had begun. (3) Quel covo fu “perquisito” dagli uomini più fidati di Totò Riina, guidati da Giovanni Brusca che comandava una squadra di picciotti trasformati in muratori. Quella casa fu messa a soqquadro. “Facemmo scomparire ogni cosa. – ha raccontato Giovanni Brusca – Furono anche divelti i pavimenti ed abbattute pareti. Utilizzammo anche un aspirapolvere per evitare che i carabinieri potessero trovare anche qualche capello e risalire all’identità di chi frequentava quella casa.” (Il Corriere della Sera, 22.02.2000) That den was inspected by Toto Riina’s most trustworthy men, guided by Giovanni Brusca, who had command of a team of ‘picciotti’ transformed into bricklayers. That house was turned upside down. “We made everything disappear – said Giovanni Brusca – Also, the pavements were torn up and the walls were demolished. We also used a vacuum cleaner, so that not a single hair was found that would have allowed the policemen to reconstruct the identity of the people that frequented that house”. In (1), the subject of the passive clauses is the discourse topic (the newspaper article is about the racial laws). The use of a passive clause here allows the writer to create a topic chain: the underlying object (le leggi razziali ‘the racial laws’) is both syntactically and pragmatically promoted, i.e. it becomes subject and topic at the same time. On the other hand, in (2) and (3), the subjects of the boldface passive clauses are brand-new entities, and are not persistent (i.e. they are not referred to any further in the subsequent discourse span). They are not plausible candidates for topichood. Note that the same passive construction (essere ‘be’ + passive participle) is used in the three examples above. Siewierska (1984) describes the view of the passive as a patient topicalizing strategy as follows: The passive is seen as a topicalizing construction for it places a non-agentive NP in unmarked subject-topic position. The agentive passive simultaneously locates the agent in the focal position of the clause. The agentive passive thus fulfils two functions, a topicalizing and a focusing one (Siewierska 1984: 222).
Whatever notion of topic one resorts to (topic as given information, or as ‘what the clause is primarily about’), the statement above is grossly incomplete and onesided. It rules out passive clauses with non-topical subjects such as (2)–(3) above as functionally “abnormal”. I am not claiming, of course, that Siewierska’s statement is completely wrong. It is true that the passive subject usually displays a cluster of features such as animacy, prominence, givenness or accessibility. None of these
TSL[v.20020404] Prn:29/08/2002; 16:21
F: PBNS13.tex / p.3 (247)
The network of demotion
features alone, however, is able to explain why the passive occurs, nor is the cluster of them. Furthermore, there are passive clauses such as (2)–(3) in which the subject does not display any of these features. Givón (1981) assigns the passive two more functions. He outlines a typology of passivization as exemplification of the method of syntactic typology: syntactic typology is the study of the mapping between structures (characterized by word order, morphology or intonation) and specific functions, defined in terms of information sequencing or packaging. Givón argues that one is justified in talking about functional domains rather than simply of functions, because syntactic functions very commonly are not atomic or totally discrete. Functional domains may also cross: a structure belonging to one functional domain may also be a member of another functional domain. The passive, in its capacity of assigning the clausaltopic function to a non-agent argument, is a member of the domain of topic identification, ranging from easiest to most difficult. The passive also belongs to the domain of detransitivization, sharing this property with statives, middles and reciprocal/reflexives. Moreover, it can easily be shown to be a member of yet another functional domain, that of impersonalization or agent suppression. These functional domains are (at least theoretically) independent from syntax and universal. The function of any syntactic structure belonging to a given domain cannot be defined independently from its position on the scale with respect to the other structures on the same scale. It has been pointed out by Prince (1997) that such a coarse-grained view of discourse functions should be rejected. According to Prince, each syntactic structure has often more than one discourse function: the various discourse functions of a given syntactic structure are to be posited at a more specific level than Givón’s functional domains, and very often no one-to-one iconic correspondence between structures and functions can be singled out. Givón simply acknowledges a correlation between functions and structures, but notions such as agent demotion or detransitivization are too vague to be assumed as defining properties of morphosyntactic structures. According to Shibatani (1985), the basic function of the passive is the defocusing of the agent: All entities which correspond to the elements of a semantic frame or valence can be considered as focused to some extent [. . . ] These semantically coded entities are correlated with different degrees of importance; certain elements are more prominent than others, since they are most salient in the speaker’s mind [. . . ] Now, language provides various morphosyntactic means of distributing, among the semantically coded elements, the focus strength correlated with the amount of attention required [. . . ] An element which requires
TSL[v.20020404] Prn:29/08/2002; 16:21
F: PBNS13.tex / p.4 (248)
Andrea Sansò
the least amount of attention is subjected to a defocusing strategy, and the most obvious means of defocusing an element is not to encode it syntactically. (Shibatani 1985: 832)
In Shibatani’s view, the defocusing of an agent is not merely a consequence of object topicalization, but rather is the basic and primary function of the passive. This would explain the fact that many languages allow passives of intransitive clauses, as well as passives without object promotion. Shibatani explicitly considers agent defocusing as a broad pragmatic notion, comprising at least the following phenomena: absence of mention of an agent, mention of an agent in a non-prominent syntactic slot, and indirect reference to an agent by the use of an oblique case. The genericity of ‘agent defocusing’ as the core component of passives has been widely criticized. Moreover, as Haspelmath (1990: 35) correctly observes, Shibatani’s approach “has the problem that it is not clear why the passive should be in the center of the prototypically organized category, rather than, say, the reflexive, with the passive as a marginal category”. In my view, the main problem with Shibatani’s approach is that it seems to suggest that only the agent is missing. In passive clauses such as (2) and (3), the event too is characterized by low salience within the causal network in which it is embodied. In these clauses, the function of the passive cannot be reduced to the demotion of the agent. Take, for instance, example (3): the non-encoded agent of the boldface passive clause is syntactically encoded in the immediately preceding sentence (facemmo scomparire ogni cosa, ‘we made everything disappear’) and continues to appear after the passive clause (utilizzammo anche un aspirapolvere, ‘we also used a vacuum cleaner’). It is a very topical one, animate and human. Why then is it demoted by means of a passive clause immediately after its mention in a contiguous active sentence? What is distinctive about the passive clauses in (2) and (3) is that they add marginal information to the previous discourse span, which provides basic information (i.e., information belonging to the main narrative or rhetorical line). This marginal information is encoded as a naked statement, i.e. in a summary fashion, by means of a passive clause. In Rhetorical Structure Theory (Mann & Thompson 1988), passive clauses such as (1) and (2) are labeled ‘elaboration satellites’. Elaboration satellites depict events at a very low degree of salience. They can be easily replaced by other elaboration satellites without affecting the overall structure of the text. In this paper, I will examine the meaning and use of two pairs of passive constructions in Spanish and Italian – the periphrastic one, consisting of the verb ‘be’ (Spanish ser, Italian essere) and the passive participle, and the so-called impersonal passive, containing the ‘middle marker’ se/si (Spanish/Italian). Given the perfect formal overlapping of the two constructions, both deriving from the same (late) Latin sources, it is not difficult to find studies in which the Spanish se-passive is
TSL[v.20020404] Prn:29/08/2002; 16:21
F: PBNS13.tex / p.5 (249)
The network of demotion
implicitly assumed to have a meaning and a range of uses roughly analogous to that of the Italian si-passive, and vice versa; the same is true for the two periphrastic constructions. Most of the contemporary linguistic studies on Italian and Spanish passives focus exclusively on the syntax of the passive constructions of these two languages. In the vast majority of cases, such studies are concerned with artificial examples, and there is no consideration for real texts. The present study attempts to fill such a gap: my analysis of Italian and Spanish passive constructions is based on real texts, without which a full semantic account of the two pairs of constructions can hardly be imagined.
A new view of the passive In this section, I will propose a semantic representation of the passive meaning that explicitly deals with the problems outlined in the preceding section. My approach differs from other current approaches in some important respects. I aim to account for the complexity of the passive uses illustrated above by postulating a semantic core of the passive which is capable of explaining all its uses. The linguistic consistency of my proposal will be demonstrated in the next section through a comparison between Italian and Spanish.
Passive and perspective: Cornelis (1997) The representation of the passive semantics I am going to introduce in this section involves the notion of perspective as described in cognitive linguistics. This notion2 has been applied to the Dutch passive by Cornelis (1997). According to Cornelis, the Dutch passive signals that the causer’s perspective should not be taken. This negative signal leaves behind a gap: the passive does not specify whose perspective should be taken, it only specifies whose perspective not to take. In English, instead, there is a positive relationship between the conceptualizer and the patient (i.e. the patient perspective should be taken). According to Cornelis, this difference is triggered by a difference in the “building blocks” that the two languages use in their passive constructions. These building blocks are said to affect the overall meaning of the construction. The meaning of the passive auxiliary or the preposition usually conveying the agent thus make a given passive construction in a given language more likely to encode either the patient’s perspective or no perspective at all. Dutch worden, ‘become’, is a processual verb which yields a more “dynamic” reading of the passive, with a stronger involvement of the causer. On the other hand, a stative passive with be keeps the agent definitely apart from the on-stage region. This allows the speaker/conceptualizer
TSL[v.20020404] Prn:29/08/2002; 16:21
F: PBNS13.tex / p.6 (250)
Andrea Sansò
to adopt the patient’s perspective more readily. As far as the preposition introducing the agent is concerned, Cornelis argues that the core spatial meaning of Dutch door is “intermediary”, whereas English by means “located in the neighbourhood of ”. This also would account for the stronger involvement of the causer in Dutch. Cornelis’ generalizations are too narrow. A more complex view of the event structure of passive clauses is to be introduced. To say that the meaning of a given passive construction in a given language depends on the lexical items forming the skeleton of the construction, would lead us to reject all instances in which a passive clause in that language does not conform to that meaning. As we have seen above, the Italian periphrastic passive is used both to signal that the patient is topical (cf. (1) and (4) below), and when the event is conceptualized in a summary fashion, as scarcely salient within the discourse network (cf. (2)–(3) and (5) below): (4) Dentro si trova un giovane albanese clandestino soprannominato dagli stessi connazionali «Cinquelire». Da tempo è tenuto sott’occhio perché ruba auto una dietro l’altra. È stato sorpreso sempre il giorno dopo, mai in flagranza di reato. (Il Tirreno, 10.02.2000) Inside the car is a young Albanian immigrant named «Cinquelire» by his compatriots. He has been kept under observation for a while because he has stolen many cars. He has been always captured the day after and never in the act of crime. (5) Adesso sappiamo che Laura Antonelli raccontava la verità. Ma sono passati nove anni. Sono stati scritti troppi articoli e troppe cartelle cliniche. La vita di una persona non è un film. E questo, perciò, non è un lieto fine. (Il Corriere della Sera, 12.02.2000) Now we know that Laura Antonelli was telling the truth. But nine years have passed by. Many articles and many files have been written. A person’s life is not a film. And, of course, this is not a happy ending. In (4), the perspective of the patient is adopted; whereas in (5) no perspective is adopted and the event is depicted in a summary fashion, as a bare occurrence. The construction used in these two examples is the same (in Cornelis’ terms, the ‘building blocks’ are the same), yet the patient’s perspective is adopted only in (4). How can we account for this duplicity of the passive, which is rather independent from the building blocks of the construction? I maintain that the passive has an agent defocusing function: a passive construction is to be considered an instruction to the hearer that a different perspective from the normal one should be taken. The normal unmarked perspective is the perspective of the agent/subject of the two-participant transitive event (cf. Croft 1994), in which there is a transmission of force from the subject to the object, construed as respectively starting point and endpoint of this transmission. In my view, the passive is underspecified with
TSL[v.20020404] Prn:29/08/2002; 16:21
F: PBNS13.tex / p.7 (251)
The network of demotion
respect to which perspective must be taken. It nevertheless signals that the main actor’s perspective should not be taken. This is the abstract meaning component of the passive. Among the entities that are left on-stage (i.e. all the entities but the agent), the speaker can choose to focus on the patient, displaying empathy towards him, or can choose to embrace the maximal scope of what is on-stage, thus conceptualizing the event in a summary fashion, i.e. as a bare happening, with little or no internal structure. Passives such as the one in example (1) thus depict patientoriented processes whereas passives such as those in examples (2) and (3) depict bare happenings. These two kinds of passives occupy a central position in the semantic network I am going to outline.
The working hypothesis: A two-level analysis of passive constructions The problem with Cornelis’ analysis is that she singles out only one level of crosslinguistic variation of passive constructions. I believe that a two-level articulation of the passive meaning is needed in order to account for meaning differences such as the ones in examples (4)–(5). The first level is the level of the lexical meaning of the elements forming the template of the construction, and corresponds to Cornelis’ “building blocks”. This level is mainly responsible in determining the tendency for a given construction to encode one of the two meanings sketched above (i.e. a patient-oriented process vs a bare happening). The second level of meaning concerns the lexical elements that instantiate the construction (i.e. that fill in its slots).3 Given semantic distinctions such as those in (4)–(5), which are present in spite of the same “building blocks” used in the two examples, we must search for a higher level at which the meaning of a passive clause is determined. I believe that the ontological properties of arguments are likely to affect the overall meaning of a given passive construction in a given language.
A semantic network for the passive A more structured representation of the semantics of passive construction informally sketched above is now introduced. It will be conceived of as a semantic network that will be called ‘network of demotion’. This simple network is formed by two poles sharing the basic meaning ‘agent’s perspective not central’. In the first pole (patient-oriented process), the perspective of another participant is adopted (usually the patient, but also the theme in ditransitive constructions), whereas in the second pole (bare happening) no perspective is taken, and there is a decrease in the referentiality of all the participants (i.e. their ability to appear as arguments of a subsequent predication or to be cataphorically persistent), so that the “semantic
TSL[v.20020404] Prn:29/08/2002; 16:21
F: PBNS13.tex / p.8 (252)
Andrea Sansò
limit” of the bare happening can be reached. The semantic core shared by these two poles is the unimportance of the agent’s perspective. Whereas in the first pole of this network agent demotion takes place in function of the promotion of the salient argument (i.e. the patient), in the second pole, agent demotion takes place in order to increase focus on the bare occurrence of the event by eliminating the main actor, which is the one whose perspective is taken in the unmarked transitive event. This network is different from Cornelis’ representations of the passive meaning in English and Dutch in that it allows a single construction in a given language to range over its two poles, whereas Cornelis’ representations only suit a single construction, i.e. they are atomic and devised to account only for a single level of variation determined by the building blocks of the construction. The working hypothesis I want to validate in my corpus analysis is that different kinds of passives (both within a single language and cross-linguistically) displace on one pole or the other of the network of demotion, according to the two levels of articulation of the passive meaning introduced above. The inherent semantics of a passive construction (first level) determines its tendency to lean towards one of the two poles, while the lexical elements that fill the slots of the construction (second level) determine its actual meaning in a given utterance.
The passive in Italian and Spanish The passive in Italian texts Two Italian constructions have been examined: the periphrastic one, consisting of the auxiliary essere ‘be’ and the passive participle, and the so-called impersonal construction with si (cf. Manzini 1986; Cinque 1988). This construction is also used in reflexive, reciprocal and impersonal clauses. I will not address the question of the relatedness of these different meanings here (but cf. Sansò 2001). In the first construction, the underlying patient is the grammatical subject and governs agreement. The only argument of si-passive may or may not govern agreement (the question is a matter of debate, cf. Lepschy 1992). Table 1 shows the distribution of these two passive constructions in my corpus. The si-construction is much less frequent than the periphrastic construction and than its Spanish counterpart with se. This is not without reasons. Table 1. Distribution of periphrastic and si-passive in Italian
Si-passives Periphrastic passives
% of passive clauses
% of total clauses
14.38% 85.62%
2.24% 13.36%
TSL[v.20020404] Prn:29/08/2002; 16:21
F: PBNS13.tex / p.9 (253)
The network of demotion
Table 2. Distribution of different types of subject in periphrastic passive (both SV and VS) and si-passive in Italian; results widely exceed chance (χ2 = 135.42; d.f. = 4; p < 0.1) Type of passive subject
VS clauses
si-passives
SV clauses
Textually evoked – discourse topic Inferrable Unused Brand-new Total
14.01% 38.31% 7.47% 40.18% 100%
32.53% 19.27% 1.2% 46.98% 100%
66.14% 22.73% 1.55% 9.56% 100%
The nature of passive subjects and agents with respect to givenness has been taken into account. The results are displayed in Table 2.4 Italian seems to exploit word order within the periphrastic passive in order to introduce brand-new entities. I am not claiming that the VS word order is a marker of a bare happening passive. However, considering the role of word order in information packaging, and given the fact that VS order is used in Italian in intransitive sentences to shift attention from the subject to the event, we can say that this correlation is significant to a certain extent: a VS order periphrastic passive is more likely to express a bare happening, whereas SV order usually signals patient topicality. This is just a tendency, and passives with SV order that convey a bare happening are also found: (6) Ventisei, tra amministratori e dipendenti delle ferrovie dello stato sono stati raggiunti da avvisi di garanzia. I provvedimenti sono stati emessi dalla procura di Mondovì nell’ambito dell’inchiesta sui disservizi della linea ferroviaria Torino-Savona. L’inchiesta si è aperta all’inizio dell’anno dopo una serie di inconvenienti denunciati dai viaggiatori della linea. Negli scorsi mesi la magistratura ha acquisito la documentazione negli uffici delle ferrovie, riguardo soprattutto i diversi incidenti e la manutenzione del materiale rotabile. E ieri è stato posto sotto sequestro a Verona un locomotore, che si era incendiato nel tratto Torino-Mondovì a marzo. Un primo locomotore era già stato messo sotto sequestro a Fossano un paio di mesi fa. (La Repubblica, 26.06.2001) Twenty-six managers and employees of the National Railways received notification of investigation from the District Attorney’s office. These measures were taken by the District Attorney’s office of Mondovì during the investigation that is being carried out about the inefficiency of the railway section Turin-Savona. The inquiry was opened at the beginning of the year, after some inefficiencies denounced by the passengers. In the last few months the Bench acquired documentation concerning the various accidents and the rolling stock maintenance in the Railways offices.
TSL[v.20020404] Prn:29/08/2002; 16:21
F: PBNS13.tex / p.10 (254)
Andrea Sansò
And yesterday a locomotive that caught fire in the railway section TurinMondovì in March was seized in Verona. Another locomotive had been already seized in Fossano a couple of months ago. In order to test the statistical validity of my findings, I performed the same textual counts on a sample of active sentences. The results are displayed in Table 3. If we look at the percentage of brand-new active objects and then compare it with the percentage of brand-new passive subjects of both si-passive and VS periphrastic passive, we have to acknowledge that active objects are discourse-older than (at least) some passive subjects (i.e. the entities bearing the same semantic role of patient). This is a further demonstration that there is more to passive than a mere promotion of arguments. To put it in other words, when using a VS periphrastic passive or an impersonal si-construction, the speaker conceptualizes the event as a whole. The information conveyed by the clause is not stored under the entry for the patient in the hearer’s memory. Other significant differences can be observed if we look at the different behaviour of the three kinds of passives with respect to persistence.5 The subject of si-passives and VS periphrastic passives is more likely to decay than the subject of periphrastic SV passive, as Table 4 demonstrates. A very widespread discourse pattern for passive constructions is their tendency to require locative, temporal, or adverbial adjuncts in order to avoid a sense of anomaly. This is particularly true in the case of VS periphrastic passives and sipassives (cf. Table 5), given the fact that these two kinds of passives encode the mere Table 3. Subjects and objects in a sample of 300 active clauses; results widely exceed chance (χ2 = 85.44; d.f. = 3; p < 0.1)
Textually evoked/Discourse Topic Inferrable Unused/Unexpected Brand-new Total
Active subjects
Active objects
65% 16% 12% 7% 100%
32% 33% 9% 26% 100%
Table 4. Average persistence rate of passive subjects in the three constructions investigated (in terms of number of clauses to the right); results widely exceed chance (χ2 = 23.05; d.f. = 4; p < 0.1)
Average persistence rate
si-passives
SV passives
VS passives
0 81.93% 1–2 18.07% >2 0%
0 67.89% 1–2 21.81% >2 10.29%
0 77.77% 1–2 22.22% >2 0%
TSL[v.20020404] Prn:29/08/2002; 16:21
F: PBNS13.tex / p.11 (255)
The network of demotion
Table 5. Percentage of passive clauses with an adjunct; results exceed chance (χ2 = 13.7; d.f. = 2; p ∼ = 0.1)
An adjunct is present No adjunct is present
SV periphrastic passive
VS periphrastic passive
si-passive
36.18% 63.82%
12.14% 87.86%
30.12% 69.88%
Table 6. Distribution of periphrastic and se-passives in Spanish
se-passives periphrastic passives
% of passive clauses
% of total clauses
50.98% 49.02%
7.09% 6.82%
occurrence of an event: this means that, in order for the sentence to be felicitously uttered or relevant, an additional piece of information should be present, as in the following examples: (7) [. . . ] solo ad Agrigento si abbattono le case in nome della legalità Only in Agrigento houses are demolished in the name of the law (8) È stata issata una tenda verde sotto a un lampione, nel centro della piazza A green tent was pitched under a street-lamp, in the middle of the square This is not the case when the passive encodes a patient-oriented process. In this case, no adjunct is needed, because the information conveyed by the clause is stored under the entry for the current topic, thus being relevant in itself. In other words, since in this case perspective is assigned to some participant (whereas the same is not true in the bare happening passive), this is enough for the speaker to utter the sentence.
The passive in Spanish texts The Spanish data also confirm my hypothesis. If we look at the distribution of the two kinds of passives (periphrastic and se-construction), the difference between Spanish and Italian is striking (Table 6). Whereas in Italian the impersonal construction appears to be less widespread than the periphrastic passive, its Spanish correlate is even more frequent than the periphrastic construction. The reason for this different behaviour is that in Spanish word order evidently plays no role within the periphrastic construction (the VS periphrastic construction is very scarcely attested, around 2% of the periphrastic passives analyzed), and the se-construction represents the only way to convey a bare happening, as the following examples demonstrate:
TSL[v.20020404] Prn:29/08/2002; 16:21
F: PBNS13.tex / p.12 (256)
Andrea Sansò
(9) Durante estos 18 años ha residido en México, llegando a vivir de limosna en algunas épocas. Fue detenido el pasado 7 de julio por llevar documentación falsa y entregado a las autoridades españolas. During these 18 years, he lived in Mexico, and was sometimes reduced to beggary. He was arrested on July 7th because he presented false documents and he was handed over to the Spanish authority. (10) El obispo de Vercelli había apelado a Clemente V y éste había convocado una cruzada contra los herejes. Se decretó la indulgencia plenaria para todos aquellos que participaran en la misma, y se pidió ayuda a Ludovico de Saboya, a los inquisidores de Lombardia y al arzobispo de Milán. Fueron muchos los que cogieron la cruz para auxiliar a las gentes de Vercelli y de Novara, desplazándose incluso desde Saboya y Francia, y todos se pusieron bajo les órdenes del obispo de Vercelli. Los choques entre las vanguardias de ambos ejércitos se sucedían con frecuencia, pero la fortificacion de Dulcino era inexpugnable. The Bishop of Vercelli had appealed to Clement the Fifth, and a crusade had been called against the heretics. A plenary indulgence was granted to anyone taking part in it, and Louis of Savoy, the inquisitors of Lombardy, the Archbishop of Milan were prompt to act. Many took up the cross to aid the people of Vercelli and Novara, even from Savoy and France; and the Bishop of Vercelli was the supreme commander. There were constant clashes between the vanguards of the two armies, but Dolcino’s fortification was impregnable. The results displayed in Table 7 show that brand-new subjects are preferentially found in se-constructions. Therefore, this construction should not be considered a topicalizing strategy. This pattern has already been noticed by Hidalgo (1994). Hidalgo further claims that, since the agents of periphrastic passives in Spanish are very often present and very persistent, the periphrastic passive should be considered an instance of the inverse voice. Inverse voice is characterized by high Table 7. Types of passive subjects and agents in Spanish; results widely exceed chance (χ2 = 168.94; d.f.: 6; p < 0.1)
Textually evoked Inferrable Unused Brand-new Total
Passive subjects in periphrastic passives
Passive agents in periphrastic passives
Passive subjects in se-passives
61.78% 11.78% 4.02% 22.42% 100%
13.79% 8.05% 17.25% 60.91% 100%
26.52% 14.92% 1.94% 56.62% 100%
TSL[v.20020404] Prn:29/08/2002; 16:21
F: PBNS13.tex / p.13 (257)
The network of demotion
Table 8. Average persistence rate for patients in se-passives and periphrastic passives in Spanish; results widely exceed chance (χ2 = 45.1; d.f. = 2; p < 0.1)
Average persistence rate for patients
se-passives
periphrastic passives
0 89% 1–2 11% >2 0%
0 70.4% 1–2 24.48% >2 4.88%
topicality of agent and patient (the patient is more topical than the agent but the agent retains considerable topicality). There is no need to consider this passive an instance of inverse voice in order to account for the presence of agents: it follows directly from our description that passive agents are likely to be present especially in those passive constructions which preferentially express a patient-oriented process. The reason is straightforward: whereas in bare happening passives the speaker demotes or deletes the agent in order to convey the mere occurrence of the event, the demotion that takes place in patient prominent passives is a function of the higher salience of the patient with respect to the agent. i.e. this demotion is not total and agents retain some referential strength (i.e. they may be recalled in the subsequent discourse). As far as the persistence rate is concerned (Table 8), the data confirm the different patterns found for the two passive constructions examined. To sum up, in Spanish there is a polarization between two different constructions, whereas in Italian this polarization is located within a single construction (word order being sometimes exploited to convey a bare happening). The siconstruction in Italian appears to be less widespread than its Spanish counterpart because it is in competition with the periphrastic construction.
Conclusion and further research In this paper, I have limited my attention to two historically related languages. The differences between them are even more significant because of this relatedness. I believe that a thorough analysis of the passive in a limited number of languages, involving also a full consideration of the textual patterns of this construction, constitutes the basis for important generalizations about the semantics of the passive in general and contributes to an increased comprehension of passive systems in other languages as well. The network I introduced is typically associated with what traditional grammars label ‘passive constructions’. It must be noted, however, that there are, cross-linguistically, several constructions referred to by traditional grammars as passive constructions. They are very different from one another, when one looks at the details. The main thing that they have in common is that they are used by lan-
TSL[v.20020404] Prn:29/08/2002; 16:21
F: PBNS13.tex / p.14 (258)
Andrea Sansò
guage users when something like ‘agent demotion’ or ‘diminution of event individuation’ is going on. The network I proposed, I believe, is cross-linguistically valid: we are likely to find grammatical constructions that encode either of the two poles in many languages, although these constructions may dramatically differ with respect to structural properties (or ‘building blocks’). In many cases, however, the constructions encoding these prototypes are referred to by traditional grammars as ‘passive constructions’ and are somehow connected to passive morphology. Thus, it could turn out from further investigation that this network is not actually a network for the passive, but rather a network for a number of other constructions, or, generally, a network for the cognitive construct ‘agent’s perspective not central’: however, the correlation between this network and passive or passive-like constructions calls for explanation and should be seriously addressed. At the same time, the relatedness of the two prototypical poles forming this network is supported by linguistic evidence: when looking at Italian, this relatedness becomes immediately evident. I claim that this interelatedness is valid also when a language has two different structures encoding the two prototypes, as demonstrated by the fact, signalled by Spanish grammars, that the se-passive is spreading so as to cover the area of the periphrastic passive, which prototypically encodes only a patient-oriented process.
Notes . This research is based on a corpus consisting of 450 articles from both Italian and Spanish newspapers. The work was in part developed thanks to the funds obtained from the University of Pavia (Progetto Giovani Ricercatori 1999–2000). I would like to thank Anna Giacalone Ramat for her insightful comments on an earlier version of this paper. The usual disclaimers apply. . Perspective is used here in its Cognitive Grammar sense (cf. Langacker 1987: 128: “Every linguistic expression [. . .] structures a conceived situation (or scene) by means of a particular image. [. . . ] the speaker, by choosing appropriate focal “settings” and structuring a scene in a specific manner, establishes a construal relationship between himself and the scene so structured”). . It is possible that the aspect of a passive clause also concurs to determine whether the passive expresses a patient-oriented process or a bare happening. For a complete discussion of the correlation between passive and aspect, cf. Sansò 2001, where it is argued that this two-level articulation is what makes the passive a sentential phenomenon very similar to aspect, but not an aspect on its own. The two-level articulation I postulate affects the event structure of the sentence in the same way as aspectual operators do. According to de Swart and Verkuyl (1999), aspect is a sentential phenomenon characterized by two main dimensions: first, there is the level of the Aktionsart, which is a lexical property of verbs, and points to a general shape of the event. At a second level, every verb can be construed as whatever
TSL[v.20020404] Prn:29/08/2002; 16:21
F: PBNS13.tex / p.15 (259)
The network of demotion
kind of event, and this is determined by the ontological properties of the arguments (cf. well-known alternations such as John ate a sandwich [accomplishment]/John ate sandwiches [activity]). . I adopt the taxonomy of given/new information proposed by Prince (1981), with some minor modifications. Givenness is taken here to be a good index of topicality of a referent. . Following Givón (1981), persistence is measured in terms of number of clauses to the right of a sentence s in which the topic continues an uninterrupted presence as a semantic argument of the clause, an argument of whatever role and marked by whatever grammatical means. If a grammatical device strongly topicalizes an element, this element is also likely to rank high on the persistence scale (i.e. it is likely to be the object of some subsequent predication).
References Cinque, G. (1988). On “Si” constructions and the theory of “Arb”. Linguistic Inquiry, 19, 521–581. Cornelis, L. H. (1997). Passive and perspective. Amsterdam-Atlanta: Rodopi. Croft, W. (1994). Voice: beyond control and affectedness. In B. Fox & P. J. Hopper (Eds.), Voice. Form and function (pp. 89–117). Amsterdam-Philadelphia: Benjamins. Givón, T. (1981). Typology and functional domains. Studies in language, 5 (2), 163–193. Haspelmath, M. (1990). The grammaticization of passive morphology. Studies in Language, 14, 25–72. Hidalgo, R. (1994). The pragmatics of de-transitive voice in Spanish: From passive to inverse? In T. Givón (Ed.), Voice and inversion (pp. 168–186). Amsterdam-Philadelphia: Benjamins. Langacker, R. (1987). Foundations of cognitive grammar. Stanford: Stanford University Press. Lepschy, G. C. (1992). La linguistica del Novecento. Bologna: Il Mulino. Mann, W. C., & Thompson, S. A. (1988). Rhetorical structure theory: Toward a functional theory of text organization. Text, 8 (3), 243–281. Manzini, M. R. (1986). On Italian “Si”. In H. Borer (Ed.), Syntax and Semantics 19: The syntax of pronominal clitics (pp. 241–262). New York: Academic Press. Prince, E. F. (1981). Toward a taxonomy of given-new information. In P. Cole (Ed.), Radical pragmatics (pp. 223–255). New York: Academic Press. Prince, E. F. (1997). On the functions of Left-Dislocation in English discourse. In A. Kamio (Ed.), Directions in functional linguistics (pp. 117–144). Amsterdam-Philadelphia: Benjamins. Sansò, A. (2001). Passive and elaboration of events. Ph.D. dissertation, University of Pavia, Italy. Shibatani, M. (1985). Passives and related constructions: A prototype analysis. Language, 61 (4), 821–848. Siewierska, A. (1984). The passive. A comparative linguistic analysis. London: Croom Helm. de Swart, H., & Verkuyl, H. (1999). Tense and aspect in sentence and discourse. ESSLLI 1999 lecture notes (http://esslli.let.uu.nl/Courses/swart-verkuyl.html).
TSL[v.20020404] Prn:17/10/2002; 14:40
F: PBNS14.tex / p.1 (261)
Valence change and the function of intransitive verbs in English and Japanese* Mayumi Masuko Waseda University, Tokyo, Japan
.
Introduction
Verb valence is one of the most frequently discussed issues within linguistics. At the same time, however, it is one of the most obscure topics. There is no consensus as to how it is determined within one language (see, for instance, Suga & Hayatsu 1995 for Japanese). Similarly, why verbs with similar meanings are intransitive in some languages and transitive in others remains a mystery. In this paper, I shall examine English examples whose grammatical status is marginal. There are certain constructions in Japanese that are similar in their functions, and I shall discuss them in order to establish a generalisation which it is hoped will serve as a practical definition of the term ‘intransitive’. I shall argue that what is involved in both cases is valence reduction, or argument suppression, driven by pragmatic factors.
. The English data In English, there are verb forms that are used as both intransitive and transitive. These verbs, when used as intransitive, are classified as the middle: a wellknown example is melt. There are, however, other verbs whose intransitive usage some speakers find perfectly acceptable and others reject as ungrammatical. The following are such examples: (1) The car handles well. (2) She interviewed well. (3) Bean curd digests easily.
TSL[v.20020404] Prn:17/10/2002; 14:40
F: PBNS14.tex / p.2 (262)
Mayumi Masuko
(4) The dress zips up. (5) This passage just isn’t translating. (6) The river fords easily. (7) The pill swallows easily. (8) Mary scrubbed clean. (9) This wine drinks well. (10) The bridge crosses easily. (11) London approaches easily at this time of the day. (12) Dinner was preparing in the kitchen. (13) The bed hides under easily. (14) These bureaucrats bribe. (15) French acquires more rapidly than Russian. (16) As a pet, my goldfish owns more cheaply than my cat. (17) That constraint doesn’t violate. (18) Dictators don’t impeach. These are somewhat modified versions of examples used in Rosta (1995). It should be noted here that, with the possible exception of (2) and (8), these examples describe some state, rather than action or process, which corresponds to some characteristics that distinguish the referent of the sentence initial nominal from other, probably similar, entities. The ‘exceptional’ cases (2) and (8) may be taken as describing one-off events, but I should like to suggest that they, too, refer to states. That is, in (2) what is important is not the fact that she answered questions satisfactorily during the interview, but her performance during the interview had the quality of being good. Similarly, (8) describes the fact that Mary was clean after she scrubbed herself. What all the above examples have in common is that the verbs are usually used as transitive and yet they lack the agent. The subject expression in each example does not refer to the agent in the event, but rather to what would have been the object of the corresponding transitive verb, i.e. the patient. It should be stressed here that these sentences may not be judged perfectly grammatical by native speakers of English and that even those who found (some of) them acceptable, expressed differing degrees of acceptability. I should mention here, however, that some of them are listed in some dictionaries as intransitive verbs with the relevant meaning: for instance, drive as in (1) and digest as in (3) are included in Longman Dictionary of Contemporary English (Summers 1991). Why
TSL[v.20020404] Prn:17/10/2002; 14:40
F: PBNS14.tex / p.3 (263)
Valence change in English and Japanese
these examples are acceptable for at least some native speakers despite their ungrammaticality and why their acceptability differs would have to be accounted for.
. The middle in English It has been noted, for instance by Levin (1993), that there is a class of transitive verbs in English that can be used as intransitive verbs when accompanied by a manner adverb: (19) A freshly baked loaf doesn’t cut easily. (20) The meat cuts easily. (21) *The bread cut.
(LDOCE) (Levin 1993: 26) (Levin 1993: 29; her (16b))
Levin (1993) classifies examples like (19) and (20) as verbs of “middle alternation”. She explains that they are “characterized by a lack of specific time reference and by an understood but unexpressed agent . . . The middle alternation is described as being restricted to verbs with affected objects” (Levin 1993: 26). The following examples, however, show that this characterisation is not altogether appropriate: (22) This wine is drinking well. It should be noted that (22) is a progressive version of (9). Unlike (9), however, many speakers found this perfectly acceptable, which suggests that the “lack of specific time reference” may not be the characteristic of the middle, contrary to Levin’s claim. Furthermore, manner adverbs might not be obligatory as evidenced by (23): (23) . . . a wine that will drink for many years
(RHD)
This suggests that Levin’s description of the middle construction is not appropriate and that different and/or additional factors are involved. We shall come back to this in Section 6. We shall examine the examples (1)–(18) in more detail. Rosta (1995) argues, calling them “mediopassives”, that the subject in these sentences functions as the “archagonist” by which he means the person or object primarily responsible for the event. It, however, functions not as the agent but as the object, and is semantically a “theme” of the sententially denoted event. The following is the only place where Rosta comes close to giving a definition of the mediopassive: (24) Properties of the mediopassive: I Lect 1: The subject of a mediopassive is the object of the mediopassive.
TSL[v.20020404] Prn:17/10/2002; 14:40
F: PBNS14.tex / p.4 (264)
Mayumi Masuko
Lect 2: The subject of a mediopassive is whatever can be the subject of a passive – i.e. the object of the mediopassive or some other subordinate of the mediopassive, determined by principles as yet unknown. II Not every referent of the subject of a mediopassive is the er [= agent – MM] of the sense of the mediopassive. III The referent of the subject is the archagonist of the sense of the mediopassive. (Rosta 1995: 138) Rosta notes that the mediopassive is often used (a) with the adjunct (e.g. easily) which makes a non-agent more plausible as the archagonist, (b) when an emphatic auxiliary such as will is used, (c) when a negative auxiliary is used, or (d) when contrastive stress is placed on the verb. He explains that an adjunct, auxiliary or stress makes a mediopassive “newsworthy”. This makes the utterance acceptable even though the derived interpretations are unusual. He adds that even when such an expression is not used, the acceptability of the mediopassive will become high if the derived interpretations are “not contrary to prior expectations” (Rosta 1995: 133). Rosta goes further and contends that any verb could presumably be used as a mediopassive because whether a given verb can be used as the mediopassive depends on the availability of contexts that make the utterance acceptable; and as he admits, “of course, with enough ingenuity a plausible context can be found for any mediopassive” (Rosta 1995: 130). Unfortunately, Rosta does not explicate what he means by utterances being “newsworthy” or interpretations being “not contrary to prior expectations”, and since he does not, his argument lacks explanatory power as it can only claim that anything goes as long as it is pragmatically acceptable. This suggests that the mediopassive, at least as Rosta envisages it, cannot be clearly defined. We need to look at more data in order to determine whether Rosta’s account would be of any use. If it turns out that the notion of “newsworthiness” indeed explains the mediopassive, his insight can be utilised by defining the notion more rigorously and explicating the conditions. Before we attempt to do this, however, somewhat similar cases in Japanese will be examined in order to see whether there may be universally applicable conditions for using verbs as intransitive.
. Intransitive verbs in Japanese This section examines some Japanese constructions that are somewhat similar to the English cases discussed above. I shall start with a discussion of the transitiveintransitive distinction in Japanese.
TSL[v.20020404] Prn:17/10/2002; 14:40
F: PBNS14.tex / p.5 (265)
Valence change in English and Japanese
Transitive verbs in Japanese may or may not have their intransitive counterparts. When they do, the intransitive versions are morphologically distinct from the transitive. This incidentally means that Japanese differs from languages such as English in which the same forms of some verbs function both as the transitive and the intransitive (e.g. the middle). The following examples illustrate this: (25) Mado-ga ai-ta. window-nom open (intransitive)-past “[A] window opened.” (26) Taroo-ga mado-wo ake-ta. Taroo-nom window-acc open (transitive)-past “Taro opened [a] window.” In (25), the past tense form of aku ‘open’, aita, is used. As can be inferred from the case-marking particle ga, the sentence-initial noun, mado ‘window’, is usually interpreted as the subject; aku, thus, is intransitive. As the agent is not mentioned in (25), it is assumed that it is not salient in the discourse (cf. Masuko 1999) (26) contains the past-tense form of akeru, aketa, which is the transitive counterpart of aita. What complicates the matter is that there are cases where these intransitive verbs involve wo-marked postpositional phrases1 (PPs) as discussed by Suga (1981). This is problematic because these PPs are normally taken to be objects of transitive verbs. (27) Taroo-ga ookina-kuchi-wo ai-ta. Taroo-nom large-mouth-acc open (intransitive)-past (28) Taroo-ga ookina-kuchi-wo ake-ta. Taroo-nom large-mouth-acc open (transitive)-past “Taro made his mouth wide open.” (Suga 1981: 123; my gloss and translation) On the surface, (27) is just like (28) except for the verb form. The reason why the verb aita is treated here as intransitive even though it co-occurs with the womarked PP is because it has exactly the same form as in (25): since aita is usually classified as an intransitive verb, it would be consistent to treat it as such even when it co-occurs with a phrase that looks like an object. That is, the wo-marked PPs will be taken to be adjuncts and accompanying verbs therefore remain intransitive. Would this suffice as a reasonable explanation? If we wish to retain a putative generalisation that a wo-marked PP is by definition the object of a transitive verb, cases like (27) ought to be dismissed as a performance error. Such arguments have, indeed, been made by several grammarians, dating back to the nineteenth century: see, for instance, Ootsuki (1897). Suga (1981) refutes this by saying that there are
TSL[v.20020404] Prn:17/10/2002; 14:40
F: PBNS14.tex / p.6 (266)
Mayumi Masuko
examples that almost all native speakers would accept as grammatical, such as (29), analogously to their transitive equivalents like (30): (29) Yamada-kun-wa shigoto-wo owat-te ie-ni Yamada-Mr-top job-acc finish (intransitive)-and home-to kaet-ta. go back-past “Yamada finished for the day and went home.” (30) Yamada-kun-wa shigoto-wo oe-te ie-ni Yamada-Mr-top job-acc finish (transitive)-and home-to kaet-ta. go back-past “Yamada finished his job and went home.” (Suga 1981: 132; his 39a–b) There have been other theorists, like Matsushita (1930/1977), who argued that cases like (27) and (29) illustrate the possibility of intransitive verbs temporarily used as transitive. Suga (1981) contends, however, that there are some differences in meanings in pairs such as (27) and (28). He maintains that intransitive versions describe a situation where action of the sentence initial PP does not affect the wo-marked PP. The transitive versions, on the other hand, describe a situation where the sentence-initial PP causes some change in the wo-marked PP. This may be unclear in the pair (29) and (30). Suga gives the following examples: (31) Yamada-shi-ga shinsatsu-wo Yamada-Mr-nom medical examination-acc owat-ta. finish (intransitive)-past (32) Yamada-shi-ga shinsatsu-wo oe-ta. Yamada-Mr-nom medical examination-acc finish (transitive)-past “Mr. Yamada finished a medical examination.” (Suga 1981: 132; his 41a–b) While Mr. Yamada could be either the doctor or the patient in (31), he has to be the doctor in (32). Suga argues this is because, in the transitive version, Mr. Yamada has to cause some change in the object, the medical examination. It would be easier, and perhaps more reasonable, for him to be the agent than the patient (or experiencer) in order to cause change in the situation in which he is involved. Suga (1981) supports his argument further by giving pairs of examples such as (33)–(36):
TSL[v.20020404] Prn:17/10/2002; 14:40
F: PBNS14.tex / p.7 (267)
Valence change in English and Japanese
(33) *Keiko-wa megane-wo hazure-ta. Keiko-top spectacles-acc remove (intransitive)-past “Keiko was not wearing her spectacles.” [intended] (34) Keiko-wa megane-wo hazushi-ta. Keiko-top spectacles-acc remove (transitive)-past “Keiko took off her spectacles.” (Suga 1981: 131; his 31a–b) The act of taking off one’s spectacles affects the spectacles in a sense that its position changes. That is why (34) is unproblematic but (33) is unacceptable. Since it contains the putative intransitive verb hazusu, (33) implies that the object, megane ‘spectacles’, was not at all affected by the event and no change took place, which cannot be true. (35) Watashi-ga kokyoo-wo hanareta-no-wa I-nom hometown-acc left (intransitive)-nomi2-top juunen-mae-da. 10 years-ago-be “It was ten years ago that I left my hometown.” (36) *Watashi-ga kokyoo-wo hanashita-no-wa I-nom hometown-acc left (transitive)-nomi-top juunen-mae-da. 10 years-ago-be “It was ten years ago that I left my hometown.” [intended] (Suga 1981; his 32a–b) While (35) is acceptable, (36) is not. (35), with the intransitive hanareta ‘left’, does not imply that my leaving affects my hometown. In contrast, (36) is unacceptable because the use of the transitive verb hanashita implies that my leaving caused some change in the town, which is hardly likely. It should be noted here that intransitive verbs in general describe states rather than actions, and the sentence initial nominal does not function as an agent but as a participant in the event that has led to the state. I shall argue that it is placed at the beginning because it is salient. If such is the case, the function of ga, a particle traditionally taken to be the nominative case marker, is used solely to indicate saliency. It follows then that the ga-marked nominal does not have to be the subject. In the next section, slightly different cases will be examined. They nevertheless prove that the function of ga is not to mark the nominative case or indicate that the accompanying nominal is the subject but rather to indicate saliency. It will be also shown that the function of intransitive verbs is to explicate the fact that the sentence denotes a state rather than an event in its narrow sense.
TSL[v.20020404] Prn:17/10/2002; 14:40
F: PBNS14.tex / p.8 (268)
Mayumi Masuko
. Quasi-intransitive verbs in Japanese Similar to the English examples discussed in Sections 2–3, there are cases in Japanese where verbs that are normally considered transitive are used without “object”. This has been noted by Masuko (1998a, 1998b) in relation to examples like (37)–(40): (37) Ano-mise-de kakkoii T-shatsu-ga utteiru. that-shop-loc cool T-shirt-nom be on sale “In that shop, cool T-shirts are selling ø.” (lit.) ⇒ “In that shop, cool T-shirts are (currently) on sale.” [intended] (38) Hanzaisha-ga tsukamaeru-koto-ga deki-na-i. criminal-nom catch-event-nom able-neg-pres “Criminals cannot catch ø.” (lit.) ⇒ “Criminals cannot be caught.” [intended] (39) Shiai-ga kaishishi-mashi-ta. game-nom begin-polite-past “The game has begun ø.” (lit.) ⇒ “The game has begun.” [intended] (40) Keiba-wa Funabashi-ga kaisai-desu. horse racing-top Funabashi-nom opening-copula “As for horse racing, Funabashi is being held.” (lit.) ⇒ “Horse races are held at Funabashi.” [intended] Kaishisuru ‘begin’, tsukamaeru ‘catch’, kaishisuru ‘begin’ and kaisaisuru ‘hold’ are transitive verbs, and yet (37)–(40) lack the object, which would have to be the womarked nominal. Since it is possible to omit the subject and/or object in Japanese, provided that their referent is clear from the context or inferable from the preceding discourse, (39), for instance, could mean that the game started something. That, however, is rather odd: games normally do not start anything and in the unlikely event that they did, what exactly they begin would have to be mentioned. (39) could only mean that a game begun, and the ga-marked nominal is not used as the subject of kaishisuru. Mutatis mutandis for (38)–(40). As the function of a nominal can be determined by postpositions in Japanese, the above examples might simply be regarded as performance errors where the accusative wo should have been used instead. However, there might be other factors involved because postposition errors more frequently involve ga than any others. Furthermore, they are similar to the English cases presented in Section 2 in that they refer to some intrinsic characteristic of the sentence initial nominal. In these examples, the ga-marked PPs are not agents but patients in the event. They are placed at the very beginning of the sentence because they are salient in
TSL[v.20020404] Prn:17/10/2002; 14:40
F: PBNS14.tex / p.9 (269)
Valence change in English and Japanese
the current discourse and have come to the speaker’s mind first. Each sentence describes some state, rather than action, in which the referent of the ga-marked PP is involved; his or her participation is essential in this state. Neither of these sentences concerns a change of state. Furthermore, these examples do not express events in the usual sense but states. They describe some distinguishing characteristic(s) of the referent of the subject which appears in the sentential initial position because it is considered newsworthy in the discourse. The examples (37)–(40) are thus similar to those that were discussed in Section 4. These sentences with the sentence initial ga-marked nominal but without the wo-marked object can be interpreted only by invoking common-sense knowledge about sententially denoted events and information associated with lexical items, such as argument structure. For instance, when processing (38), the common-sense knowledge associated with hanzaisha ‘criminal’ may be invoked, which would include the fact that criminals are usually caught by the police and catching someone is not one of their distinguishing characteristics. Thus, hanzaisha that co-occurs with tsukamaeru ‘catch’ will have to be interpreted as the object of catching, even though it is not marked with wo. Similarly, the knowledge associated with horse racing in Japan would contain the fact that Funabashi is a well-known racecourse. Common-sense knowledge would tell us that racecourses, being inanimate entities, cannot hold anything, and this would rule out an agentive interpretation. In short, the cases discussed in this section are similar to the English examples in that their subjects do not refer to agent and that the referent of the subject expressions would normally be interpreted as the object of the verb which usually would be regarded as transitive. They are used when the agent in the transitive version does not play an important role and it is not worth mentioning. It can thus be argued that one basic function of intransitive verbs in English and Japanese is to describe states. To sum up the discussion so far, one of the main functions of intransitive verbs in English and Japanese is to describe states not involving change. Furthermore, in Japanese, intransitive verbs have two different usages. Firstly, there are verbs such as hanareru ‘leave’ in (35) that apparently take two PPs. It has been shown that the wo-marked PP in this construction is not the object in the usual sense and functions as an adjunct rather than an argument of the verb. These verbs are hence intransitive, and not transitive. The second involves the verbs that are normally considered as transitive and yet only have the ga-marked PP as their argument that denotes a salient participant in the event. The referent of the ga-marked PP used in both types neither causes nor is affected by change.
TSL[v.20020404] Prn:17/10/2002; 14:40
F: PBNS14.tex / p.10 (270)
Mayumi Masuko
. Newsworthiness In this section, we shall go back to the English middle cases and examine conditions that may make them more or less acceptable. Section 3 reported that Levin (1993) argued that there are certain requirements for verbs to be used as the middle and that one of them is that the middle verbs must have affected objects. It also questioned the validity of her argument. Her claim will be more closely examined here. Masuko (1998b) argued that the reason why an adjunct and/or a modal makes the mediopassive more acceptable was that the property of the subject in each of these sentences is not worth mentioning without the contribution made by the adjunct or the modal. That is, it is not at all surprising that, say, a loaf of bread or meat can be cut or wine can be drunk. It indeed is so ordinary that producing some utterance to describe such an event (or a state) seems a waste of one’s energy and effort. If the cutting is easy or difficult, then the fact that it was easy or difficult merits mention. The same argument was independently made by Fellbaum (1998) and Lemmens (1998). Fellbaum offered the following examples to provide further evidence: (41) This vegetable microwaves (easily). (42) This vegetable cooks *(quickly). (43) This suitcase zips shut (in a flash). (44) This suitcase closes *(easily).
(Fellbaum 1998: 97–98; her (41)–(44))
They show that adverbials are not obligatory when the intransitive verbs are specific enough. Pace Levin’s remark that the middle requires a manner adverb, when the verb itself specifies in what manner something is done, the manner adverb becomes redundant. These examples also prove that the middle verbs are typically used to describe characteristic properties of the sentence initial nominal. There is further evidence, taken from Imaizumi and Gunji (2000), that shows that what is important is whether the meaning conveyed by a sentence as a whole is specific enough: (45) a.
Kujoo-ga deru. complaint-nom appear “Complaints will be/are sent in.” b. *Kujoo-wo dasu. complaint-acc put out “(Someone) will send in complaints.” [intended]
TSL[v.20020404] Prn:17/10/2002; 14:40
F: PBNS14.tex / p.11 (271)
Valence change in English and Japanese
(46) a.
Hanketsu-ga deru. judgement-nom appear “(The) sentence is passed.” b. ?Hanketsu-wo dasu. judgement-nom pass “(Someone) will pass (the) sentence.” [intended] (Imaizumi & Gunji 2000: 50; their (26a) and (26f); their judgements; my gloss and translation)
My judgements do not completely coincide with Imaizumi and Gunji’s. I find (45b) marginally acceptable, though I admit it is not as good as (46b). I argue that the acceptability will be improved if it is passivised. And if it co-occurs with a verb of feeling (or so-called psych-verb), it seems much better. (47) Kujoo-wo das-are-ru. complaint-acc put out-pass-pres “Complaints are sent in.” ⇒ “Complaints are made against (something related to) me/us.” (48) Kujoo-wo das-are-temo komaru. complaint-acc put out-pass-though be at a loss “Even if complaints are sent in, nothing could be done about them” [intended] (49) Hanketsu-wo das-arete naita. judgement-acc pass-pass-and cried “The sentence was passed and (someone) cried.” [intended] This is because when dasu is passivised, the existence of someone who receives the complaints or sentence is highlighted even if that person or group of people is not linguistically expressed. The passive also implies that someone suffers from or is affected by the resultant state. In the case of (48), the existence of person or persons who will receive the complaints becomes even clearer as it must be the same as someone who is at a loss and does not know what to do with them. (49) can be similarly explained. A sentence is passed in a court of law by the judge and the defendant receives it. All these are associated with the meaning of hanketsu ‘sentence/judgement’. What all this suggests is that linguistic expressions can be omitted when their contribution can be inferred and/or recovered from linguistic and/or extralinguistic context. In some cases, the semantics of co-occurring noun or verb phrases will entail the existence of a missing entity or incorporate the meaning of a manner adverb which is otherwise required. In other cases, the referent of a noun phrase has already been mentioned and become salient, or is salient because it is the primary
TSL[v.20020404] Prn:17/10/2002; 14:40
F: PBNS14.tex / p.12 (272)
Mayumi Masuko
participant in the discourse, i.e. the speaker. In all cases, the ‘missing’ expression is considered superfluous: the utterance without it is enough, or sufficiently informative.3 This notion of sufficient informativeness appears to play a crucial role in determining the relative acceptability of utterances. (50) Sufficient Informativeness of an Utterance Utterances must be sufficiently informative to be acceptable. That is, an utterance must linguistically realise all the arguments required by the verb semantics and include manner adverbials unless it is otherwise capable of conveying enough information to ‘recover’ the missing argument(s) and encode a specific event or state. The notion of newsworthiness that was invoked in Masuko (2000) can be provisionally defined as follows: (51) Newsworthiness An utterance is considered newsworthy when it refers to a situation markedly different from others. The two notions (50)–(51) together state when a linguistic expression can be omitted: (52) Omission of Linguistic Expressions Linguistic expressions, even arguments of a predicate, can be omitted as long as the resultant utterance is sufficiently informative and newsworthy. All in all, people utter something that conveys a meaning that can be easily understood. At the same time, they will try to do this with minimum effort, for this drive for economy seems to underlie all human behaviour (cf. Zipf 1949). This can be captured by (50)–(52).
. Conclusion To sum up the discussion thus far, verbs in English and Japanese can be used as intransitive when they are sufficiently informative and refer to states, rather than events in their narrower sense. More data and analyses are required to decide whether this is universal or not. The data examined here suggest that linguistic expressions can be omitted when certain conditions are met. Many of the factors involved are pragmatic, but, pace Rosta (1995), it is not the case that any utterance is acceptable. Linguistic contributions of the co-occurring expressions such as verb semantics can easily be explicated in lexically-oriented theories such as the generative lexicon framework
TSL[v.20020404] Prn:17/10/2002; 14:40
F: PBNS14.tex / p.13 (273)
Valence change in English and Japanese
proposed in Pustejovsky (1995). A precise exposition will be left for further study, though an informal sketch was given in Masuko (1998b, 1999). During the discussion, previous theories that propose conditions for using verbs in English as the middle were examined and some conditions were rejected. One condition not examined is ‘affectedness’ that is also invoked in analyses of the passive (cf. Imaizumi & Gunji 2000). Lemmens (1998) argues that the affected entity involved in the situation described by the middle construction contributes to obtaining the described state. He claims that the middle construction focuses on ‘deliberately designed properties’ of the affected. More detailed analyses of the middle construction across languages will be necessary to judge whether this is the case. This study belongs to the Gricean pragmatics that tries to capture formally conditions and/or requirements involved in (largely) linguistic communication. The contention shared by work in this tradition is that most factors underlying linguistic behaviour can be explicated and formalised, though there are certain grey areas where formalisation seems extremely difficult, if not impossible. It is hoped that the work reported here has shown that some fuzzy notions can be formalised to a certain extent. The next logical step is to explicate and formalise precisely functions of intransitive verbs in general and conditions for using verbs as intransitive, which this paper did not accomplish. Such notions as saliency and affectedness, which so far have escaped theorists’ attempts to formalise them, would have to be formally defined in order to achieve the goal.
Notes * The work reported here was partially supported by Waseda University Grants for Special Research Projects (Individual Research: Nos. 99A-146 and 2000A-118), which support is gratefully acknowledged. Different versions of this paper were presented at Atsugi, Sendai, Paris and Cambridge. I should like to thank the audience, particularly Maarten Lemmens, for useful comments. I am grateful to my ‘informants’ for their judgements and also to Janusz Buda for valuable suggestions. Remaining errors are of course my own. . Noun phrases (NPs) in Japanese are usually followed by case-marking particles (kakujoshi). Quite often, an NP with one of these particle is also called a NP. However, I shall call them postpositional phrases (PPs) following Gunji (1987). . Nomi here is an abbreviation of nominalizer. That is, no here is not the genitive-marking particle, but a nominalizer that makes the preceding clause an event. See, for instance, Fuji (1999) who treats this as a determiner. . Lemmens (1998) uses similar examples and other evidence to refute the argument put forward by several theorists, including Rapoport (1993), that only ‘change of state’ verbs
TSL[v.20020404] Prn:17/10/2002; 14:40
F: PBNS14.tex / p.14 (274)
Mayumi Masuko
can be used as the middle. This issue needs to be further explored, though it is outside the scope of the current study. . I would have used the term specificity had it not been used widely in linguistic literature in a completely different sense.
References Fellbaum, C. (1998). Semantic network of English verbs. In C. Fellbaum (Ed.), WordNet: An Electronic Lexical Database (pp. 69–104). Cambridge, MA: The MIT Press. Fuji, M. (1999). Constructional meaning in compositional semantics – Internally headed relative constructions as concealed causatives. In K. Arimura et al. (Eds.), Papers from the Sixteenth National Conference of the English Linguistic Society of Japan (pp. 11–20). Tokyo: The English Linguistic Society of Japan. Gunji, T. (1987). Japanese Phrase Structure Grammar. Dordrecht: Reidel. Imaizumi, S., & Gunji, T. (2000). Goiteki hukugo ni okeru hukugo jisho – das, der ni mirareru sieki to judo no yakuwari [Complex events in lexical compounds – Causative and passive relations in the Japanese verbs das and der]. In T. Gunji (Ed.), Seiyaku ni Motozuku Bumpoo no Renzoku-ryo no Gainen wo Tori-ireta Kakucho no Kenkyu [Research on Extension of Constraint-Based Grammars Using the Concept of Incremental Enumeration] (pp. 39–65). Kobe Shoin Women’s University. Lemmens, M. (1998). Lexical constraints on constructional flexibility: English ‘middable’ verbs. Ms., Université Lille 3. Levin, B. (1993). English Verb Classes and Alternations: A Preliminary Investigation. Cambridge, MA: The MIT Press. Masuko, M. (1998a). How utterances are interpreted: Lexical information and meaning extension. The Cultural Review, 12, 71–80. Masuko, M. (1998b). Lexicon and how to recover missing information. The Cultural Review, 13, 133–145. Masuko, M. (1999). Conditions for argument suppression. In V. Kordoni (Ed.), Proceedings of the ESSLLI-99 Workshop on Lexical Semantics and Linking in Constraint-Based Theories (pp. 111–125). Utrecht: Universiteit Utrecht. Masuko, M. (2000). A brief note on intransitive verbs in Japanese. The Cultural Review, 16, 129–136. Matsushita, D. (1930/1977). Zouho Koutei Hyouhun Nihon Kougo Bunpoo [A Standard Grammar of Colloquial Japanese: Revised Edition]. Tokyo: Bensei-sha. Ootsuki, F. (1897). Kou-Nippon Bunten [A Comprehensive Dictionary of Japanese Sentences]. A private publication. Pustejovsky, J. (1995). The Generative Lexicon. Cambridge, MA: The MIT Press. Rapoport, T. R. (19933). Verbs in depictives and resultatives. In J. Pustejovsky (Ed.), Semantics and the Lexicon (pp. 163–184). Dordrecht: Kluwer. Rosta, A. (1995). ‘How does this sentence interpret?’ The semantics of English mediopassives. In B. Aarts & C. F. Meyer (Eds.), The Verb in Contemporary English: Theory and Description (pp. 123–144). Cambridge: Cambridge University Press.
TSL[v.20020404] Prn:17/10/2002; 14:40
F: PBNS14.tex / p.15 (275)
Valence change in English and Japanese
Suga, K. (1981). Jita chigai – Jidoshi to mokuiteki-go, soshite doushi no bunrui [Differences between intransitives and transitives: intransitive verbs, object and verb classification]. Reprinted in Suga and Hayatsu (1995: 122–136). Suga, K. & E. Hayatsu (Eds.). (1995). Doushi no Jita [Transitivity and Intransitivity of Verbs]. Tokyo: Hituzi Shobo. Summers, D. (Ed.). (1991). Longman Dictionary of Contemporary English (New edition). Harlow, Essex: Longman. Zipf, G. K. (1949). Human Behavior and the Principle of Least Effort. Addison-Wesley, MA.
TSL[v.20020404] Prn:26/08/2002; 16:00
F: PBNS15.tex / p.1 (277)
The transitive/intransitive construction of events in Japanese and English discourse* Patricia Mayes University of Wisconsin, Milwaukee
Introduction Over the past twenty years, much research in cognitive linguistics has been devoted to studying how speakers express event causation. Object manipulation, encountered very early in childhood, is often cited as the prototypical causative event upon which semantic categories such as agent and patient are based. Such events may be expressed in simple transitive clauses, which are claimed to be a canonical form in the languages of the world (Lakoff 1977, 1987; Slobin 1981, 1985; Langacker 1991). Thus it is not surprising that transitive clauses have been the object of many studies. Still, little is known about how frequently speakers actually use transitive/intransitive clauses. Indeed, several recent studies suggest that the majority of clauses occurring in natural discourse are intransitive (cf. Hopper 1991, 1992, 1997; Tao 1996; Kärkkäinen 1996; Thompson & Hopper 2001). An additional related question concerns the frequency with which transitive/intransitive clauses are used across languages. In this study, I add to the growing body of research surrounding these issues by comparing the frequency of transitive/intransitive clauses in data from Japanese and English. This question is of particular interest with respect to these two languages because many Japanese specialists have claimed that English tends to be higher in transitivity than Japanese (Ikegami 1978, 1981, 1985, 1991; Teramura 1978, 1980; Jacobsen 1981, 1989, 1992; Fujii 1993; Ohori 1997). The data I use are cooking class demonstrations. These data are particularly well suited for this investigation due to the fact that an important part of the teachers’ task in a cooking class is to explain how to perform the steps in recipes, many of which involve object manipulation. As will be made clear below, my findings support those of Thompson and Hopper (2001) and others who suggest that other genres such as conversation and informal narrative are also relatively low in transi-
TSL[v.20020404] Prn:26/08/2002; 16:00
F: PBNS15.tex / p.2 (278)
Patricia Mayes finite (independent) clause Clause non-finite (dependent) clause
Figure 1. Japanese clause types
tivity. Indeed, in both languages, my data show a relatively low frequency of transitive clauses. Thus I will argue that high- or low-transitivity is not an inherent semantic property of a language that is invariantly realized in the same way in all situations. Rather, it is a continuous, relative property that varies according to how a language is used to make meaning in a particular situation.
Data and method The data for this study are drawn from five cooking lessons in each language, which were audiorecorded in Tokyo, Japan and Santa Barbara, California. It is well-known that English and Japanese have very different clause structures. Therefore, before discussing the coding system for transitivity, I will explain the types of clauses that were included in the database for each language.
The Japanese database The Japanese database consists of 291 clauses, which I have divided into two types: finite and non-finite.1 Finiteness concerns the dependency relationship between clauses, and finite clauses may be considered independent, while non-finite clauses are dependent and connected to a subsequent finite clause. This relationship is illustrated in Figure 1. In example (1), lines 1 and 3 contain non-finite clauses, and line 4 contains a finite clause. (1) Sushi (S = Suzuki) 1. S: . . . hi o tsuke-mashi-tara, heat acc turn.on-pol-temp ‘after you’ve turned on the burner,’ 2. yahari, as.expected 3. . . . ano = futtoo suru made wa, hes boil do temp top
TSL[v.20020404] Prn:26/08/2002; 16:00
F: PBNS15.tex / p.3 (279)
The transitive/intransitive construction of events
4. sono mama oite itadaki-masu. that way place receive(hum)-pol ‘leave (it) there until (it) boils.’
The American database The American database contains 248 clauses. For this study, I examine independent clauses of two types (simple and complex) and adverbial clauses. Simple clauses are much more frequent than the other two types, accounting for 70% of the American data. I define simple clauses as main clauses that may or may not contain an infinitival complement. Example (2) contains an infinitival complement, and example (3) does not: (2) Generic Bread (P = Pam) P: uh you just want to soften it. (3) Generic Bread (P = Pam) P: and it wakes up, The motivation for treating infinitival complements as part of the main clause is that they often do not function like main verbs; rather they give information about modality or aspect (Quirk et al. 1985). For example, in cases like (2), Quirk et al. suggest that it is perhaps more accurate to treat want to as a pragmatic particle rather than a main verb. I refer to such pragmatic particles as modal particles. Thus in this example, want to is a modal particle signaling deontic modality, and soften is the main verb. I treat like to as a modal particle as well. The complex clauses in the American database contain an embedded relative clause. I consider such relative clauses to be modifiers, and thus I did not code the relative clause itself for transitivity. (4) contains an example. (4) Crab Cakes (G = Gary) → G: . . . the oil or the butter that you’re using .. breaks down, . . . after a few minutes, In this example, the relative clause that you’re using is a modifier of the NP the oil or the butter. I coded this example as a one-argument clause, which, as will be discussed below, is considered intransitive. The adverbial clauses in the data are of three types: temporal, reason, and conditional clauses. Figure 2 summarizes the clause types in the American database.
TSL[v.20020404] Prn:26/08/2002; 16:00
F: PBNS15.tex / p.4 (280)
Patricia Mayes simple (may have an infinitival complement) Independent complex (matrix + an embedded relative clause)
Adverbials
temporal (when, while, after) reason (because) conditional (if )
Figure 2. English clause types
Transitivity Research over the past twenty-five years suggests that transitivity is a complex set of semantic and morphosyntactic features that are associated with an entire clause (cf. Lakoff 1977, 1987; Hopper & Thompson 1980; Sugamoto 1982; Tsunoda 1985, 1991; DeLancey 1987; Langacker 1991). One of the earliest works to provide a comprehensive definition of transitivity was Hopper and Thompson (1980). Their definition suggests that, like other conceptual categories, transitivity can be understood in terms of a prototype, with central members and other more peripheral members. They define transitivity in terms of 10 interacting parameters. I use a simplified version of their model, which includes the two factors that seem to be most central in a prototypical transitive event: the number of participants and the type of event. The essence of the prototypical transitive event is an object-manipulating action in which there are two or more participants. This agrees with the definition of the transitivity prototype suggested by other researchers, including Slobin (1981, 1985), DeLancey (1987), and Langacker (1991), in particular, who suggests that the central characteristic of a transitive event is the transfer of energy from an agent to a patient. It is this type of event (i.e., an object-manipulating action) that is expressed in a prototypical transitive clause. Using this model of the prototypical transitive clause, I suggest that a minimal definition of a transitive clause is a clause with two or more arguments: an A and an O. Clauses with two or more arguments can then be ranked as high- or low-transitive along a continuum that varies according to the type of event expressed in the clause. This continuum, which I call “the Event Scale”, is shown in Figure 3. High Transitive object-manipulating action
Figure 3. The event scale
non-object-manipulating action
event
Low Transitive state
TSL[v.20020404] Prn:26/08/2002; 16:00
F: PBNS15.tex / p.5 (281)
The transitive/intransitive construction of events
I coded all clauses with two or more arguments according to the type of event they express.
Coding of transitive/intransitive clauses As was suggested above, I consider all one-argument clauses to be intransitive, and all clauses with two or more arguments to be more or less transitive. In order to determine where the clauses with two or more arguments lie on the transitivity continuum, I coded all of those clauses according to the event scale. Some examples of the coding system are given below.
Object-manipulating actions Clauses in which the main verb refers to a specific object-manipulating action are considered to express object-manipulating actions. (5) is an English example, and (6) is a Japanese example. (5) Vegetarian Cooking (J = Jean) J: . . . we’re gonna um, take these cucumbers, → . . . cut them up, (6) Sushi (S = Suzuki) → S: . . . hi o keshite itadaki-masu. flame acc turn.off receive(hum)-pol ‘turn off the burner.’
Non-object-manipulating actions Clauses with two or more arguments where the main verb does not refer to a specific object-manipulating action are considered to be non-object-manipulating actions. A list of verbs that were sometimes used in this way is given in Table 1, along with examples. The predicates in Table 1 do not form a coherent class, and if we examine them out of context, it might seem likely that they would be used to express objectmanipulation. However, in the cooking class context they often were not used in this way. Indeed, many of the example utterances in Table 1 seem to suggest that a series of actions will take place, but they do not specifically refer to a particular object-manipulating action. Though we can infer that the O argument will be affected, none of the predicates specifies how that might occur or in what way it would be affected. The predicate do and its Japanese equivalent suru merit separate discussion because they sometimes function to express object-manipulating actions, while in
TSL[v.20020404] Prn:26/08/2002; 16:00
F: PBNS15.tex / p.6 (282)
Patricia Mayes
Table 1. Non-object-manipulating verbs English verb
Example
make take use finish prove treat create check keep on get do
and then make a small sample now we wanna take . . . half the peppers we’ll use commercial stock finish it in the oven always prove your yeast and you have to treat it in that way you’re creating a surface tension let me just check our bread over here you keep it on low heat and you get em nice and brown and crispy on one side (discussed below)
Japanese verb tsukuru (‘make’)
Example osoosu o tsukutte mairimasu ‘(we’ll) make the sauce’ ano futtoo suru made wa, sono mama oite itadakimasu. ‘leave (the rice) as it is, until it starts to boil’ ookii katamari no ho o rebaa ni tsukaimasu ‘(we’ll) use the big piece (of ginger) in the liver’ chotto usu aji de, shiagete mitai to omoimasu ‘(we’ll) finish (the radish) with a mild flavour’ (discussed below)
oku (‘place’, ‘leave’) tsukau (‘use’) shiageru (‘finish’) suru, itasu, nasaru (‘do’)
other cases, they express non-object-manipulating actions.2 The reason for this in the case of the American database is that do primarily functions as a pro-form that is anaphoric for another verb. For example, in (7), it refers back to made. In this case, I coded both clauses as non-object-manipulating. (7) Vegetarian Cooking (J = Jean) J: and we made stock. that’s what you can do. → you can do that at home. On the other hand, in (8), do refers back to joining the pieces together (with ones’ knuckles), and both predicates were thus coded as object-manipulating. (8) French Bread (P = Pam) P: . . . and use your knuckles again to joi=n those pieces together. . . . then, turn it u=p, → . . . and do that again.
TSL[v.20020404] Prn:26/08/2002; 16:00
F: PBNS15.tex / p.7 (283)
The transitive/intransitive construction of events
In the Japanese data, suru (‘do’) is more frequent than English do (23% of the clauses with two or more arguments vs. 10%). One reason is that it functions in word formation processes to produce verbs from nouns (Shibatani 1990). Many of the instances of suru in my data are of this type. I consider these verbs to be lexicalized compound verbs, and they are primarily used to express object-manipulating actions. (9) is an example: (9) Sushi (S = Suzuki) S: . . . okome o arau to dooji ni, rice acc wash cond same.time at . . . kono .. awasezu wa, this vinegar.mixture top → . . . choogoo shite kudasai. mix do give(hon) ‘while (you)’re washing the rice, mix in the awasezu at the same time.’ In this example, I consider choogo shite to be a compound verb meaning mix; thus it refers to an object-manipulating action.3 On the other hand, there are a few cases in which suru functions in a way similar to the other predicates listed in Table 1. In other words, in these cases, suru does not refer to a specific object-manipulating action. (10) contains an example. (10) Sushi (S = Suzuki) 1. S: ano, hes 2. futtoo shita jiten de, boil do point at 3. . . . kobu o dashi-masu. seaweed acc take.out-pol ‘when (the rice) boils, take out the seaweed.’ → 4. . . . sono mama shite ori-masu to ne, that way do cont-pol cond it ‘if you leave it that way,’ In this example, in lines 1–3, the teacher tells the students to remove the seaweed when the rice begins boiling. In line 4, she begins to give a warning about what will happen if they do not do this (instead letting the rice continue to boil with the seaweed), using shite orimasu (‘doing’) to do so. In this instance, the verb suru does not refer to an object-manipulating action. There are only a few cases like this one where suru (‘do’) functions as a non-object-manipulating verb.
TSL[v.20020404] Prn:26/08/2002; 16:00
F: PBNS15.tex / p.8 (284)
Patricia Mayes
Events I consider events to be occurrences that are represented by a clause in which the A argument is non-agentive. Hopper and Thompson (1980) as well as others such as Lakoff (1977) and DeLancey (1987) suggest that volitionality is an important characteristic of a prototypical transitive event that is directly tied to the A argument. In other words, events that are caused by volitional agents are higher in transitivity. Following these scholars, I assume that the prototypical A argument must be human or at least animate in order to be capable of volition. Thus A arguments that refer to inanimate objects are not considered to be agents, and all such clauses were coded as events. (11) through (13) are examples. (11) Beef Carbonade (G = Gary) G: .. it’s just gonna cook all the juices out of it. (12) Crab Cakes (G = Gary) G: @ . . . the machine’s stopping me. ((THE FOOD PROCESSOR IS STUCK)) (13) Guratan (K = Kikuchi) K: . . . ano okona no ryuushi ne, hes flour lk particles it . . . abura no maku o, oil lk film acc . . . ano=, hes . . . (.7) koo karamete shimau tte ii-masu ka, this bind end.up qt say-pol q ‘the particles of flour end up binding the oil,’
States Clauses that express states refer to conditions that occur over a period of time and are usually not marked for a beginning or an endpoint. English clauses that have the main verb have or the modal particle like to were coded as states. (14) and (15) are English examples. (14) Crab Cakes (G = Gary) G: .. and I do have bay shrimp in here, (15) Vegetarian Cooking (J = Jean) J: . . . so I like to wipe mushrooms.
TSL[v.20020404] Prn:26/08/2002; 16:00
F: PBNS15.tex / p.9 (285)
The transitive/intransitive construction of events
In the Japanese database, there are a few clauses with verbs that express psychological states, including wakaru (‘understand’), mieru (‘see’), and wasureru (‘forget’). These were coded as states. (16) is an example. (16) Makizushi (S = Suzuki) S: . . . (6.6) konna ne, this it . . . wakari-masu kashira ne. understand-pol wonder it ‘(I) wonder if (you) understand this.’
Results In order to determine the frequency of transitive/intransitive clauses in the databases, I first counted the number of clauses with one argument and the number with two or more arguments. The clauses with two or more arguments were then coded along the event scale defined in Figure 3. The results are presented in Table 2. Recall the questions that were posed at the beginning of the paper: What is the frequency of transitive/intransitive clause types in the cooking demonstration genres? And what is the frequency of transitive/intransitive clauses found in these similar genres from the two languages? Table 2 shows that intransitive clauses are the most frequent in both languages. It also shows that high-transitive clauses are fairly infrequent in the American data, and this supports the findings of other researchers working on other genres (e.g., Hopper 1991, 1992, 1997; Kärkkäinen 1996; Thompson & Hopper 2001). Notice also that the Japanese data contain more high-transitive clauses than the American data, so perhaps we could say that Japanese is higher in transitivity than English. However, this does not seem to be correct either, considering the fact that there is also a larger number of intransitive clauses in the Japanese data than in the American data. I suggest that it is not accurate to say that either language is higher in transitivity than the other, but that Table 2. The frequency of transitive/intransitive clauses in the databases Japanese Intransitive (1-argument) clauses Transitive (2-argument) clauses High-transitive Object-manip Act Non-object-manip Act Event Low-transitive State
English
136
47%
105
42%
113 35 3 4
39% 12% 1% 1%
67 41 17 18
27% 17% 7% 7%
TSL[v.20020404] Prn:26/08/2002; 16:00
F: PBNS15.tex / p.10 (286)
Patricia Mayes
transitivity varies with the context – depending on what the participants are saying and doing. Table 2 also suggests that in the American database, the clauses with two or more arguments are distributed more evenly across the different event types. Thus in effect, there are more clauses with two or more arguments that are used to encode non-object-manipulating actions, events, and states than there are in the Japanese database. Figure 4 provides a closer look at this distribution. Figure 4 shows that in the Japanese teachers’ talk, there are fewer clauses at the low end of the scale. Thus clauses with two or more arguments seem to be used in a more restricted way in Japanese, primarily to refer to object-manipulating actions. On the other hand, the English-speaking teachers used clauses with two or more arguments to express not only object-manipulating actions but also nonobject-manipulating actions, events, and states. Perhaps this is what the researchers who suggest that English is higher in transitivity mean – that English speakers use clauses with two or more arguments to express a less restricted set of event types than Japanese speakers do. Indeed, the idea that in Japanese there is a restriction on agency such that A’s must be human is sometimes mentioned in this connection. This restriction does seem to hold in the cooking class data. (13), repeated below, is the only example in the Japanese database in which there is an inanimate A. In this example, an inanimate object okona no ryuushi (‘particles of flour’) is the A argument.
Event
Figure 4. Transitivity and actions, events, and states in clauses with two or more arguments
TSL[v.20020404] Prn:26/08/2002; 16:00
F: PBNS15.tex / p.11 (287)
The transitive/intransitive construction of events
(13) Guratan (K = Kikuchi) K: . . . ano okona no ryuushi ne, hes flour lk particles it . . . abura no maku o, oil lk film acc . . . ano=, hes . . . (.7) koo karamete shimau tte ii-masu ka, this bind end.up qt say-pol q ‘the particles of flour end up binding the oil,’ There is no restriction concerning the animacy of the A argument in English, and there are 11 clauses with inanimate A’s in the database (8% of the clauses with two or more arguments). Returning to my claim that neither of these languages can be considered “high in transitivity” because transitivity is not an inherent property of a language, I suggested that transitivity is not only continuous, it is a relative property, varying with the situation and genre. Of course, speakers of all languages are able to express prototypical transitive events, but whether they do so and how they do so depends on the situation. The relative nature of transitivity is well demonstrated by innovations such as the following: (17) Vegetarian Cooking (J = Jean) 1. J: a=nd, 2. just cook them for about an hour, 3. until, → 4. the leaves pull out. In this example, Jean is explaining how to cook artichokes. The clause in lines 3 and 4 contains the verb pull out, which would normally be assumed to be a transitive, 2-argument verb. Here, however, she uses it as a 1-argument verb. It seems that examples like this are in some ways both transitive and intransitive. Jean could have used the predicate come out, which is a canonical 1-argument verb, but it seems that pull out better expresses the important point: When cooking artichokes, the way to test if they are done is by trying to pull the leaves out. If they come out easily, the artichokes are done. If Jean had said cook them until the leaves come out, the students might have assumed that the leaves would fall out by themselves. In this case, the artichokes would be overdone. By using the predicate pull out, Jean seems to include the notion of an agent, while focusing on the patient and the result, which is the most important point in her instructions.
TSL[v.20020404] Prn:26/08/2002; 16:00
F: PBNS15.tex / p.12 (288)
Patricia Mayes
Conclusions In investigating transitivity, I defined it as a scalar property based on some of the parameters outlined by Hopper and Thompson (1980). In particular, I used the number of arguments and the type of event to define it. I assumed that clauses that express object-manipulating actions encode prototypical transitive events and are thus highest in transitivity; clauses expressing non-object-manipulating actions, events, and states are found along the scale, continuing from higher to lower transitivity; and one-argument clauses are intransitive. I found that when clauses were defined as transitive/intransitive using this scale, intransitives were actually the most frequent type in both the Japanese and American teachers’ talk. Though this may seem somewhat surprising given the cooking class context, it concurs with the findings of other researchers concerning transitivity variation in other genres of natural discourse. I also found that the claim that English is higher in transitivity than Japanese is not supported. In fact, the Japanese teachers’ talk contained a higher number of high-transitive clauses than the American teachers’ talk, but it also contained a higher number of intransitive clauses. Considering all of the variations in transitivity together, I have suggested that level of transitivity is not an inherent property of a language; rather, it is a cluster of semantic and morphosyntactic features (as others have shown) that are realized differently in different situations. Thus the frequency of transitive/intransitive clauses varies as participants use these constructions to make meaning in a particular situation.
List of abbreviations acc cond cont hes hon hum it lk pol q qt sof temp top
accusative conditional continuous aspect hesitation honorific humble interactional element (particle) linker polite interrogative marker quotative softening word temporal connector topic
TSL[v.20020404] Prn:26/08/2002; 16:00
F: PBNS15.tex / p.13 (289)
The transitive/intransitive construction of events
Transcription conventions : (colon) carriage return , (comma) . (period) . . . (n) . . . (3 dots) .. (2 dots) @ = (( ))
speaker attribution and turn beginning intonation unit boundary continuing intonation contour final intonation contour long pause, 1.0 seconds or more medium pause, .4–.9 seconds short pause, .3 seconds or less laughter lengthening researcher’s comment
Notes * I would like to thank Pat Clancy, Susanna Cumming, Shoichi Iwasaki, and Sandy Thompson, as well as two anonymous reviewers for comments on an earlier version of this article. I am also grateful to the participants of the Second International Conference in Contrastive Semantics/Pragmatics for their comments and suggestions. I am particularly grateful to the conference organizers Ken Turner and Kasia Jaszczolt for providing a forum for contrastive research. . Complex clauses, containing an embedded complement clause, account for only 2% of the data, and thus are not considered. . This also holds true for the honorific variants of suru: itasu (‘do(HUM)’ and nasaru (‘do(HON)’). . Shite is the -te form of suru (‘do’).
References DeLancey, S. (1987). Transitivity in grammar and cognition. In R. S. Tomlin (Ed.), Coherence and Grounding in Discourse (pp. 53–68). Amsterdam: John Benjamins. Fujii, S. Y. (1993). On the clause-linking to construction in Japanese. In P. M. Clancy (Ed.), Japanese/Korean Linguistics, 2 (pp. 3–19). Stanford: CSLI Publications. Hopper, P. J. (1991). Dispersed verbal predicates in vernacular written narrative. Berkeley Linguistics Society, 17, 402–413. Hopper, P. J. (1992). On the uses of the notion of EVENT in linguistics. Paper presented at the International Congress of Linguists, Quebec, Canada. Hopper, P. J. (1997). Dispersed verbal predicates in vernacular written narrative. In A. Kamio (Ed.), Directions in Functional Linguistics (pp. 1–18). Amsterdam: John Benjamins.
TSL[v.20020404] Prn:26/08/2002; 16:00
F: PBNS15.tex / p.14 (290)
Patricia Mayes
Hopper, P. J., & Thompson, S. A. (1980). Transitivity in grammar and discourse. Language, 56, 251–299. Ikegami, Y. (1978). How universal is a localist hypothesis? A linguistic contribution to the study of ‘semantic styles’ of language. In R. P. Fawcett & M. A. K. Halliday (Eds.), Semiotics of Culture and Language (pp. 49–79). London: Frances Printer Publishers. Ikegami, Y. (1981). “Suru” to “Naru” no Gengogaku [The Linguistics of “Do” and “Become”]. Tokyo: Taishukan Shoten. Ikegami, Y. (1985). Activity – accomplishment – achievement: A language that can’t say ‘I burned it, but it didn’t burn’ and one that can. In A. Makkai & A. K. Melby (Eds.), Linguistics and Philosophy: Essays in honor of Rulon S. Wells (pp. 265–304). Amsterdam: John Benjamins. Ikegami, Y. (1991). “Do-language” and “become-language”: Two contrasting types of linguistic representation. In Y. Ikegami (Ed.), The Empire of Signs: Semiotic Essays on Japanese Culture (pp. 286–326). Amsterdam: John Benjamins. Jacobsen, W. M. (1981). Transitivity in the Japanese Verbal System. Ph.D. Dissertation, University of Chicago. Jacobsen, W. M. (1989). Tadoosei to purototaipuron [Transitivity and prototype theory]. In S. Kuno & M. Shibatani (Eds.), Nihongogaku no Shintenkai [New Developments in the Study of the Japanese Language] (pp. 213–248). Tokyo: Kuroshio. Jacobsen, W. M. (1992). The Transitive Structure of Events in Japanese. Tokyo: Kuroshio. Kärkkäinen, E. (1996). Preferred argument structure and subject role in American English conversational discourse. Journal of Pragmatics, 25, 675–701. Lakoff, G. (1977). Linguistic gestaults. Chicago Linguistics Society, 13, 236–287. Lakoff, G. (1987). Women, Fire, and Dangerous Things. Chicago: University of Chicago Press. Langacker, R. W. (1991). Foundations of Cognitive Grammar, Vol. 2. Stanford: Stanford University Press. Ohori, T. (1997). Transitivity in grammar and rhetoric: A case from English and Japanese. In The Locus of Meaning: Papers in Honor of Yoshihiko Ikegami (pp. 389–406). Tokyo: Kuroshio. Quirk, R., Greenbaum, S., Leech, G., & Svartvik, J. (1985). A Comprehensive Grammar of the English Language. London: Longman. Shibatani, M. (1990). The Languages of Japan. Cambridge: Cambridge University Press. Slobin, D. I. (1981). The origins of grammatical encoding of events. In W. Deutsch (Ed.), The Child’s Construction of Language. New York: Academic Press. Slobin, D. I. (Ed.). (1985). Crosslinguistic evidence for the language-making capacity. The Crosslinguistic Study of Language Acquisition: Theoretical Issues, Vol. 2 (pp. 1157–1256). Hillsdale, NJ: Lawrence Erlbaum Associates. Sugamoto, N. (1982). Transitivity and objecthood in Japanese. In P. J. Hopper & S. A. Thompson (Eds.), Studies in Transitivity [Syntax and Semantics 15] (pp. 423–446). New York: Academic Press. Tao, H. (1996). Units in Mandarin Conversation: Prosody, Discourse, and Grammar. Amsterdam: John Benjamins.
TSL[v.20020404] Prn:26/08/2002; 16:00
F: PBNS15.tex / p.15 (291)
The transitive/intransitive construction of events
Teramura, H. (1978). “Naru” hyoogen to “suru” hyoogen: Nichiei “tai” hyoogen no hikaku. [“Become” expressions and “do” expressions: A comparison of voice expressions in Japanese and English]. In Nihongo to Nihongo Kyooiku: Moji-hyoogenhen [Japanese and Japanese Language Pedagogy: Writing and Expression] (pp. 49–68). Tokyo: Kokuritsu Kokugo Kenkyuusho. Teramura, H. (1980). Nihongo no bunpoo (ge) [Grammar of Japanese, Vol. 2]. Tokyo: Kokuritsu Kokugo Kenkyuusho. Thompson, S. A., & Hopper, P. J. (2001). Transitivity, clause structure, and argument structure: Evidence from conversation. In J. L. Bybee & P. J. Hopper (Eds.), Frequency and the Emergence of Linguistic Structure. Amsterdam: John Benjamins. Tsunoda, T. (1985). Remarks on transitivity. Journal of Linguistics, 21, 385–396. Tsunoda, T. (1991). Sekai no Gengo to Nihongo: Gengo Ruikeiron Kara Mita Nihongo [The Languages of the world and Japanese: A linguistic typological approach to Japanese]. Tokyo: Kuroshio.
TSL[v.20020404] Prn:26/08/2002; 16:02
F: PBNSP6.tex / p.1 (293)
P VI
Topics in grammar and conceptualization
TSL[v.20020404] Prn:17/10/2002; 14:41
F: PBNS16.tex / p.1 (295)
Towards a universal DRT model for the interpretation of directional PPs within a reference frame Didier Maillat Oxford University, UK / Université de Fribourg, CH
.
Setting the scene
A great deal of energy has been spent on the semantic analysis of temporal relations within a sentence and beyond that, between sentences.1 Crucially, it has been a well-established and intensively studied fact that eventualities (or events) are interpreted within a temporal framework, i.e. a temporal reference frame with respect to which the temporal relations which obtain between those events are interpreted (e.g. Reichenbach 1947; Kamp & Reyle 1993). For instance, a past tense verb tends, in the absence of any other external clue, to be temporally interpreted with respect to the time of utterance. That is to say that a typical semantic contribution of a simple past is assessed on a one-dimensional frame centred on the time of utterance, as is the case in sentence (1), where the past tense is understood as encoding a precedence relation between the event described by the sentence and the time of utterance n. (1) Alice took a walk by the river. Notice that this is not always the case. In a larger discursive context, a speaker can control some of these parameters, for instance by embedding a simple past within a larger narrative sequence. In this configuration, a simple past is then often interpreted as following the time denoted by the preceding simple past. So, if we insert (1) within a larger discursive environment, the past tense of took gets interpreted in a different way: (2) Alice slept in that morning. After a hefty breakfast, she needed some exercise. She [Alice] took a walk by the river.
TSL[v.20020404] Prn:17/10/2002; 14:41
F: PBNS16.tex / p.2 (296)
Didier Maillat
Although it is still the case that the action described by sentence (1) is located at a time t which precedes n, the whole context of (2) adds a further constraint to the interpretation of the past tense in took as it implies a narrative sequence by virtue of which, t has to follow – in the temporal dimension – t´, the time at which Alice had her breakfast. In contrast, similar facts in the spatial domain have been barely touched upon by the semanticists’ community. In particular, although the existence of spatial reference frames has been ascertained again and again, very little has been done in order to show how, e.g. spatial adverbs or spatial prepositional expressions interact with a spatial framework in order to yield a given meaning. Part of the failure to come up with a satisfactory account can be explained by the relative difficulty to tell the semantic from the pragmatic contributions. It is the goal of this paper to present a formalism that addresses these issues and tries to draw a line between what is encoded and what is contextually derived. In order to do so, I will focus on a set of framework-sensitive PPs, namely directional (or projective) PPs: to the left/right of, in front of/behind, above/below and cardinal directions. Researchers (see for instance Levinson 1996) have shown that human languages make use of only – and up to – three universal reference frames in order to express spatial relations. All the of them are instantiated in Indo-European languages. Thus, a single spatial configuration can be expressed in three different ways in these languages, depending on which reference frame is activated. All the examples below are to be interpreted as referring to a unique spatial setting. (3) Le chat est à l’est de la voiture (4) the cat is east of the car → ABSOLUTE FRAMEWORK (5) Le chat est derrière la voiture [with respect to the car] (6) the cat is behind the car → INTRINSIC FRAMEWORK (7) Le chat est à droite de la voiture [with respect to the speaker’s viewpoint] (8) the cat is on the right of the car → RELATIVE FRAMEWORK Crucially, while (5) and (7) – in the relevant readings – are cases of intrinsic and relative encoding, respectively, it should be clear that both sentences fit equally well in the other framework. That is to say that Indo-European languages are systematically ambiguous between the intrinsic and relative reference frames, since directional PPs can be successfully interpreted in both frameworks. Notice that this is also the case in other language families as similar ambiguities are observed in Japanese, Arabic and Sesotho. Most importantly, the spatial area denoted by each of the six directional PPs mentioned above (i.e. all but the cardinal directions) differs depending on the reference frame within which it is interpreted.
TSL[v.20020404] Prn:17/10/2002; 14:41
F: PBNS16.tex / p.3 (297)
Towards a universal DRT model
The strong claim that the present study will make is that – very much in the fashion of how DRT deals with tenses or temporal adverbials – spatial expressions, such as directional prepositions, can be ambiguous between several readings, because these expressions are underdetermined as to how they connect with the spatial framework. And this choice between alternative computations triggers the ambiguity in (5) and (7).
. Cross-linguistic variation Also, despite the striking recurrence of these facts throughout the Indo-European languages and other families, it should be pointed out that some languages display other properties. For instance, Tzeltal (Mexico) and Guugu Yimithirr (Australia) rely (almost) exclusively on the absolute framework. Contrary to what can be observed in Indo-European languages where the absolute frame is relatively rare and used in very restricted contexts. This type of cross-linguistic variation makes up a very strong case in favour of a unified model for all three frame types. If such a model can be constructed, the upshot would consist in being able to account for cross-linguistic variation in terms of a parametric set-up. Hausa (Africa) speakers, on the other hand, do not apply the Indo-European ‘mirror’ configuration when they construct a relative frame. In such a language, a tree’s front is not facing toward the speaker, but away from her/him (i.e. in a ‘tandem’ configuration). A suitable model will have to be flexible enough in order to capture this second type of parametric variation in a satisfactory way too.
. Untying the knot A first step towards a full understanding of the phenomena described above consists in a thorough analysis of the properties of the three frameworks. In fact, although superficially they appear to be fairly different, the three frameworks are very much alike. In order to see the similarity between them, let us start by assuming a set of conceptual primitives that will make up a spatial framework in the proposed model. Under the current view, all three frameworks combine three primitives: i. a set of three orthogonal axes (frontal, lateral and vertical) in a Euclidean space ii. a point of origin iii. an orientation vector
TSL[v.20020404] Prn:17/10/2002; 14:41
F: PBNS16.tex / p.4 (298)
Didier Maillat
where the orientation vector determines the orientation of the axial system by fixing the direction and orientation of the frontal axis (see below for a detailed explanation of this process). According to this model any spatial framework is uniquely and sufficiently defined by means of these three elements, only two of which are variables, viz. (ii) and (iii).2 In other words, any spatial reference frame is fully determined if one knows the value of its origin and that of its orientation vector. This means that this is all that is needed in order to evaluate (a) the truth-value of directional expressions in that reference frame and (b) the selection of an appropriate directional expression for that framework, given some spatial configuration. Notice that contrary to what is suggested in the literature on this topic (e.g. Zwarts & Winter 2000), the axial structure is not “polarised”, in the sense that each axis defines a dimension (frontal, lateral and vertical), but the sides are left unspecified. In other words, the distinction between, e.g. front and back is not built into the system and will have to be handled at a later stage. To pursue the analogy with the temporal domain, the spatial framework will have to be fixed, by means of assigning a value to the two parameters origin and orientation. Very much like the temporal dimension is organised around its origin, namely the time of utterance n in DRT. Notice, though, that the addition of two more dimensions in the spatial domain triggers the need for a second parameter, orientation. And as we will soon find out, a lot, if not all of the computation process needed to interpret a directional preposition goes precisely into the value assignment for the orientation parameter. However, what we have seen so far is the similarity between the various frameworks, but as we illustrated earlier on, they also differ in a crucial way, which accounts for the ambiguity observed in sentences (5) and (7). A close scrutiny of the three frame types reveals that they differ with respect to their orientation vector. Indeed the computation of the orientation vector appears to constitute the distinguishing factor between the three frames as is shown in Table 1. Table 1 illustrates that while the orientation parameter of an absolute frame is given as part of the lexical/encyclopaedic knowledge that the speaker has of a preposition like east of, the orientation parameter of an intrinsic frame is defined by the properties of the ground object (G in the table). For instance in our example (5), the functional and geometrical properties of the car will yield the correTable 1.
Absolute Intrinsic Relative
Origin
Orientation
G G G
given G VPT
TSL[v.20020404] Prn:17/10/2002; 14:41
F: PBNS16.tex / p.5 (299)
Towards a universal DRT model
sponding intrinsic orientation. For the relative system, the framework is oriented with respect to a salient viewpoint (VPT), which often coincides with the location of the speaker but may also differ at times. Table 1 also sums up the facts regarding the other parameter, namely the origin. We see that in all cases the origin of the framework is given by the ground object G (determined by the complement of P). In other words, we can conclude that in the present model, the type of reference frame is fully determined by the sole orientation parameter, which is formally represented by a vector whose origin corresponds to the point of origin of the axial system and whose endpoint is either lexically encoded, or depends on the intrinsic properties of G, or coincides with a salient VPT (viewpoint), in striking contrast with the suggestion made by Crow (1989) who makes use of no less than 4 different variables. This, however, leaves the question of the actual denotation of directional prepositions open. Building on and extending the kind of vector geometry used above and discussed extensively by O’Keefe (1996) and Zwarts and Winter (2000), the proposed analysis assumes that a directional preposition denotes an angular range. Crucially, this angular range needs to be fixed or anchored on a framework in order to be fully specified. The model poses that the angular templates are anchored on the reference frame’s orientation vector. An obvious advantage in favour of this approach lies in the fact that it readily accounts for the ambiguity observed in (5) and (7), in the sense that a single template can denote two different configurations depending on the value assigned to the orientation vector; hence, the ambiguity. Figure 1 below illustrates the kind of templates that are used in this model (in this case, a bird’s eye view of the derrière template). One should point out that the angular range is calculated clockwise from the orientation vector (represented by the arrow in the diagram). A quick word on the actual values of the angular range displayed here is required. Figure 1 instantiates the maximal denotation of the preposition derrière. 180°
90°
–90°
F
Figure 1.
TSL[v.20020404] Prn:17/10/2002; 14:41
F: PBNS16.tex / p.6 (300)
Didier Maillat
That is to say that this template covers the maximal extensional – and truthfunctionally correct – scope of the preposition, but it does not reflect the optimal angular range denoted by the preposition. Nevertheless, for the sake of brevity, I will have to gloss over the details of a more refined template (but see Maillat 2001). One should also point out the obvious relation between the proposed denotation for directional PPs and the denotation suggested by Kamp and Reyle for temporal adverbials headed by after and before. Thus, they write: What the phrase after α does is to divide the axis of time into two halves and to say of the described eventuality that it lies in the “upper-half ”. [1993: 626]
Clearly, the same remark applies to the type of directional adverbials under consideration in this paper, as Figure 1 above illustrates. The main difference between the two types of adverbials comes from the dimensionality of the respective domains: while the temporal domain is strictly one-dimensional, space, on the other hand, is three-dimensional. As a result, the purely linear model adopted by Kamp and Reyle to handle temporal relations is not sufficiently informative to tackle the spatial domain. However, the kind of vector geometry chosen here could of course be successfully applied to the temporal domain. Crucially, decreasing the dimensional arity makes the angular information about vectors irrelevant: in a one-dimensional domain all vectors are parallel to one another. In other words, before and after are the one-dimensional equivalents of in front of and behind.
. Towards a DRT model In the last part of this paper, I want to look at the kind of formalism that could handle the processes described in the previous sections. In that respect, Discourse Representation Theory seems to be quite appealing as it has been partly developed in order to come to terms with similar framework phenomena in connection with the temporal domain (Kamp & Reyle 1993 and Van Eijck & Kamp 1997). Perhaps I should first enter a caveat regarding the underlying ontology of this extended model. The extension of the DRT model in order to account for the interpretation of directional prepositional phrases requires that the ontology include some sort of spatial entity. Without further explanation, I introduce locations as a new type of entity. I can only but refer the reader to the relevant literature (Aurnague 1995 and Asher & Sablayrolles 1996). An alternative approach would have consisted in leaving out the spatial concepts and manipulating only existing discourse referents with the proviso that some conceptual module à la Jackendoff and/or spatial module would have to translate the DR structures. Here I have opted for a slightly more explicit representation.
TSL[v.20020404] Prn:17/10/2002; 14:41
F: PBNS16.tex / p.7 (301)
Towards a universal DRT model
The DRT element under scrutiny is the so-called construction rule for the preposition derrière. This choice emphasises the fact that the essential constituent which triggers the setting up of a spatial framework is the spatial adverbial. In this sense, this analysis is very much in line with the treatment of temporal adverbials offered in DRT. To put it slightly differently, it is the directional preposition itself which activates the frame assignment process. CR.Derrière Triggering configuration: PP
α P
β
derrière
Choose:
Orientation point, Opt, from the following items: (i) ST (β) (ii) ST (salient VPT) & polarity is switched to negative
Introduce into UK : Introduce into ConK :
where ST() is an operator which returns the spatiotemporal ‘slice’ denoted by its argument. l, l1 , l2 l ⊆ Opt l1 = ST(α) l2 = ST(β)
→ – l1 behind l2 & orientation vector = l2 l & relational –→ vector = l2 l1 ◦ & 90 < δ < –90◦ , where δ obtains between the relational vector and the orientation vector That being said, following the practice of Kamp and Reyle (1993), I assume that the crucial orientation parameter is encoded as an Orientation Point, Opt, reminiscent of the Temporal Perspective Point and Reference Point (see also the register r in the dynamic version of DRT elaborated by Van Eijck & Kamp 1997). This choice is pragmatic in nature and, hence, the proposed construction rule is unspecific with respect to the selection process. Consequently, the pragmatic orientation assignment process is appropriately encoded using the Choose operation. Below is a partial quotation, taken from Kamp and Reyle (1993: 621), presenting a typical DRS of a temporal adverbial, which relies on a similar device.
TSL[v.20020404] Prn:17/10/2002; 14:41
F: PBNS16.tex / p.8 (302)
Didier Maillat
CR.Sunday [. . . ] Choose:
Origin of Computation t´´ from the following items: (a) n (b) the current TPpt (c) The current Rpt [. . . ]
The proposed representation for derrière also makes use of the operator ST(), which extracts the spatio-temporal value of its argument. To be brief, ST() returns a set of ordered pairs of locations and times (Aurnague 1995 and Asher & Sablayrolles 1996). This step is crucial in order to interpret sentences like (9) They danced behind the car. involving an event dance (x, e), in which the adverbial describes the location of the whole activity denoted by the verb, i.e. the dancing. With respect to the original DRT environment, one could think of ST() as being modelled on the dur() operator, which returns the temporal value of its argument. Most of the semantic work done by the preposition is determined by the set of conditions added to the discourse representation structure. As we can see, the representation proposed here is relatively explicit, and therefore slightly at odds with standard DRT. But as I already argued, this does not constitute a serious obstacle and is meant to illustrate the whole process more clearly. In fact, this series of conditions simply formalises the information contained in the template illustrated in Figure 1, once it has been anchored on a reference frame. It says that the relational vector which obtains between the Ground object and the Figure object must be at an angle δ from the orientation vector, where δ is contained in the angular range defined above. If one were to prefer the more implicit kind of representation favoured by Kamp and Reyle, one should replace the last condition to be introduced in ConK with the following one: [. . . ]
→ – –→ l1 behind l2 & orientation vector = l2 l & relational vector = l2 l1
In this system, the absolute framework does not essentially differ from the other two frame types. The only divergence is that a preposition like east of is not ambiguous between frames. Therefore, its construction rule must reflect the fact that Opt is uniquely determined lexically (at least in Indo-European languages). Thus, CR.east_of does not include a Choose operation, instead a constraint is added to the set of conditions, such that:
TSL[v.20020404] Prn:17/10/2002; 14:41
F: PBNS16.tex / p.9 (303)
Towards a universal DRT model
x magnetic north (x) Opt := ST(x) In addition, because the model is applied across the board to all three frameworks, ‘absolute’ languages like Tzeltal and Guugu Yimithirr do not require any special device in order to be accounted for by the system. Essentially, these languages express directions like we do, they just have stronger selection restrictions on Opt. Also, Hausa can be readily handled by the theory by means of the polarity index defined in the Choose operation. This pragmatic index keeps track of the orientation assignment process and switches its default (positive) value in case Opt is matched with (ii). Hausa-behind – and for that matter, Hausa-in front of – then, simply makes use of this index to generate the opposite vector from the orientation vector. In Hausa, glossing over the syntactic differences, CR.Hausa_behind would thus be: [. . . ] Introduce into ConK : l ⊆ Opt l1 = ST(α) l2 = ST(β) → – l1 behind l2 & orientation vector = <polarity> l2 l & –→ relational vector = l2 l1 & 90◦ < δ < –90◦ , where δ obtains between the relational vector and the orientation vector The very same formalism elegantly captures the notorious lateral flip observed in Indo-European languages. In these languages, the prototypical denotation of to the left of is at a 90◦ angle, clockwise, from the front side as defined in a relative framework, but at a 90◦ angle, anticlockwise, from the front side as defined in an intrinsic framework. And vice versa for the prototypical denotation of to the right of. That is to say that the lateral axis is, as it were, flipped when you pass from one framework to the other. As a result, in a configuration where the intrinsic and relative denotations of in front of and behind coincide, the denotations of intrinsic to the left of will be that of relative to the right of. So far, this oddity has been a major obstacle in every formal approach which has tried to tackle the problem. For instance, Zwarts and Winter (2000) – in the latest attempt to date at formalising the semantics of spatial expressions – have skipped the lateral flip issue and decided to treat only the laterally neutral beside (2000: 182). In contrast, the proposed model offers a single treatment to and generalises over the Hausa tandem configuration and the Indo-European lateral flip. So, the very same polarity index will also capture the Indo-European lateral
TSL[v.20020404] Prn:17/10/2002; 14:41
F: PBNS16.tex / p.10 (304)
Didier Maillat
flip by generating the opposite vector in case option (ii) is selected in the Choose operation. Finally, as is the case in standard DRT with temporal reference points (see Kamp & Reyle 1993: 603), Opt is reset after the triggering clause has been fully processed. This last deletion step is needed so that sentences involving several directional phrases can be treated without imposing the requirement that Opt is held constant throughout, which would clearly be too restrictive. So, in sequence (10) below, we must be able to assign an intrinsic reading to the first directional PP, while, at the same, we are forced to select a relative reading for the second PP. (10) I can see three balls. The one in front of the car [Choose = (ii)] is identical to the one behind the tree [Choose = (i)]. In other words, the acceptability of (10) under the relevant reading indicates that the orientation assignment process is a “bookkeeping device” which must be reset after the triggering component has been processed (Kamp & Reyle 1993: 603).
. Concluding remarks To conclude, I hope to have shown that an elegant DRT model for directional prepositions can be constructed on a single parameter and that this model successfully captures some very puzzling cross-linguistic facts, while, at the same time, shedding some new light on the crucial interaction between spatial expressions and spatial reference frames.
Notes . This research project has been supported by a grant from the Berrow Trust. . Strictly speaking, a third parameter is needed in order to fully determine such a space, but, for brevity’s sake, I will gloss over the variability of this third parameter and assume that gravity fixes the vertical axis (see Levelt 1996 for further discussion of the issue).
References Asher, N. (1993). Reference to Abstract Objects in Discourse. Dordrecht: Kluwer. Asher, N., & Sablayrolles, P. (1996). A Typology and Discourse Semantics for Motion Verbs and Spatial PPs in French. In J. Pustejovsky & B. Boguraev (Eds.), Lexical Semantics (pp. 163–209). Oxford: Clarendon Press.
TSL[v.20020404] Prn:17/10/2002; 14:41
F: PBNS16.tex / p.11 (305)
Towards a universal DRT model
Aurnague, M., & Vieu, L. (1996). A three-level approach to the semantics of space. In C. Zelinsky-Wibbelt (Ed.), The semantics of prepositions from mental processing to natural language processing (pp. 393–439). Berlin: Mouton de Gruyter. Aurnague, M. (1995). Orientation in French Spatial Expressions: Formal Representations and Inferences. Journal of Semantics, 12, 239–267. Bloom, P., Peterson, M. A., Nadel, L., & Garrett, M. F. (Eds.). (1996). Language and Space. Cambridge, MA: MIT. Crow, J. (1989). Toward a Semantics for English Spatial Expressions. Ms. Ph.D. Thesis, University of Texas at Austin. Herskovits, A. (1986). Language and Spatial Cognition: An interdisciplinary study of the prepositions in English. Cambridge: CUP. Jackendoff, R. (1990). Semantic Structures. Cambridge, MA: MIT. Kamp, H., & Reyle, U. (1993). From Discourse to Logic. Dordrecht: Kluwer. Levelt, W. J. M. (1996). Perspective Taking and Ellipsis in Spatial Descriptions. In P. Bloom, M. A. Peterson, L. Nadel, & M. F. Garrett (Eds.), Language and Space (pp. 77–108). Levinson, S. C. (1996). Frames of reference and Molyneux’s question: Crosslinguistic evidence. In P. Bloom, M. A. Peterson, L. Nadel, & M. F. Garrett (Eds.), Language and Space (pp. 109–171). O’Keefe, J. (1996). The spatial prepositions in English, vector grammar, and the cognitive map theory. In P. Bloom, M. A. Peterson, L. Nadel, & M. F. Garrett (Eds.), Language and Space (pp. 277–316). Maillat, D. (1999). On the Interpretation of Directional Expressions: Empirical and Theoretical Considerations. Ms. M.Phil. Thesis, University of Oxford. Maillat, D. (2001). Which Template for behind? Empirical Considerations on the Meaning of Directional PPs. In Oxford Working Papers in Linguistics, Vol. 6. University of Oxford. Nam, S. (1995). The Semantics of locative prepositional phrases in English. Ms. Ph.D. Thesis, University of California, Los Angeles. Reichenbach, H. (1947). Elements of Symbolic Logic. London: Macmillan. Vandeloise, C. (1986). L’espace en français. Paris: Seuil. Van Eijck, J., & Kamp, H. (1997). Representing Discourse in Context. In J. Van Bethem & A. Ter Meulen (Eds.), Handbook of Logic and Language (pp. 179–237). Amsterdam: Elsevier. Zwarts, J., & Winter, Y. (2000). Vector Space Semantics: A Model-theoretic Analysis of Locative Prepositions. Journal of Logic, Language and Information, 9 (2), 171–213.
The interaction of syntax and pragmatics* The case of Japanese ‘gapless’ relatives Akiko Kurosawa King’s College London
.
Introduction
Dynamic Syntax (DS) (Kempson, Meyer-Viol, & Gabbay 2001) models the process of utterance interpretation in context as a left-right process of tree growth. This gives us a basis for natural-language syntax, and a formal characterisation of the syntax/pragmatics interface. This paper is a case study of relatives and anaphora construal. The interaction of relativisation and pragmatic processes, and the phenomenon of ‘gapless’ relatives in Japanese, are construed using (i) principles of how long-distance dependency is processed, (ii) a unified theory of discourse anaphora, and (iii) general pragmatic assumptions. Since the extensive work on Japanese noun modifying clauses by Teramura (1975–1978/1992), ‘gapless’ relatives, specifically ‘pseudo’ relatives, have been discussed in diverse perspectives such as cognitive grammar (Yamanashi 1995), Frame Semantics (Matsumoto 1997) and machine translation (Narita 1994). While there is an argument-predicate relation between the head nominal and the predicate in the regular relative clause such as Doa-o tataku hito, ‘the person who knocks on the door’, the pseudo relative lacks such a relation as in Dareka-ga doa-o tataku oto, (lit. ‘the sound that someone knocks on the door’). This paper is the first attempt to analyse the phenomenon using a formal model and a pragmatic theory, contrasting its account mainly with Narita’s two-place relation analysis. In the next two sections, I outline the framework of Dynamic Syntax, and how a verb-final language, Japanese, is modeled in DS. Then in Section 4, relatives in English are introduced, followed by Section 5 that discusses the structure building processes of regular relative clauses in Japanese. Finally in Section 6, ‘gapless’ relatives are analysed both in pragmatic and formal perspectives.
Akiko Kurosawa
. Dynamic Syntax (DS) – structural underspecification Dynamic Syntax models the parsing process as progressive structure building, using concepts of structural underspecification. In this section, I will informally introduce these concepts using two English examples.
. Structure building in English (1) I like John. (2) John, I like. These strings have an identical semantic formula. If the pronoun I is identified as Bill: (3) (Like (John)(Bill))
Bill likes John.
DS characterises the processes of how this interpretation is achieved. Some major symbols used in this section are as follows: ?: requirement ◊: pointer Ty(t): the semantic type of IP Ty(e): the semantic type of DP Ty(e→t): the semantic type of VP Ty(e→(e→t)): two-place predicate
. A sample derivation – example (1): I like John The overall goal is to derive a truth-evaluable proposition, and this is the starting point of the operation. To get to the goal of Ty(t), this root node must have a subject node and a predicate node as its daughters. This is an example of a computational action. The pronoun I provides the information that this is a subject, and slotted in the subject node. The node is filled with a metavariable, Upro, which is to be identified with an entity. Here Upro is immediately resolved as the speaker, e.g. Bill. Then a VP node of Ty(e→t) is built as a requirement in (a). In (b), this requirement of Ty(e→t) works as a trigger for the actions defined by the verb like. Make a left daughter node for the verb, and make a right daughter node for the object, as this verb is a two-place predicate of Ty(e→(e→t)). This is a lexical instruction. The next lexical input, John, satisfies the requirement of Ty(e) on the object node, updating the tree as in (c). Now all the terminal nodes are annotated with Fo (formula), requirements are removed and compilation takes place at the Ty(e→t) node and then at the root node. With no requirements outstanding, the goal is
Japanese ‘gapless’ relatives
achieved as ‘Bill likes John’, which represents an interpretation for the string, as in (d). (a)
Scanning I
(b) Scanning like
{? Ty(t)}
{? Ty(t)}
{Ty(e), Fo(Upro/Bill)} {? Ty(e t), }
{Ty(e), Fo(Bill)}
I = Upro = the speaker, e.g. Bill Upro is a place-holder, later filled with another formula
{Ty(e (e t)), Fo(Like), } {? Ty(e)}
(c) Scanning John
(d) Compilation at the root node {Ty(t), (Fo(Like (John)(Bill)), }
{? Ty(t)}
{Ty(e), Fo(Bill)}
{? Ty(e t)}
{? Ty(e t)}
{Ty(e (e t)), Fo(Like)}
{Ty(e), Fo(John), }
Bill
Like(Bill)
Like
John
In this way, the tree structure expands incrementally through requirements for something missing. The tree is unfolded from the top to the bottom. Each stage except the final one represents only partially specified structure, and this incompleteness till the process gets to the end is one primary form of structural underspecification.
. Left-dislocation – example (2): John, I like While the tree node location for each constituent in (1) is immediately identified as introduced, in (2), the object John is left-dislocated, and its location in the semantic tree is not known when initially scanned. Here is another type of structural underspecification. This underspecification of the tree node location is shared by constructions such as WH-questions, clefting and relative clauses. In DS, aiming to give a unified account for these long-distance dependency phenomena, a notion of unfixed node is introduced. An unfixed node is a node which is related to the root node weakly, not incorporated into the tree structure immediately, and left dangling until its location on the tree is found.
Akiko Kurosawa
(a) Scanning John Unlike in (1), in processing (2) an alternative strategy is adopted. It is constructing an unfixed node requiring Ty(e) described only as dominated. The word John then decorates the node with `Fo(John), Ty(e)’. Unlike pronouns, John has no information on the case, i.e. where in the tree to be located is not lexically instructed. It can be a subject, a fronted object, or the first item of a co-ordinate structure, John and Mary, and so on. There is a finite number of options. So it cannot be incorporated into the tree. It’s unfixed. {? Ty(t)}
{Ty(e), Fo(John), }
(b) Scanning I This input is processed in the same way as in the previous example. {? Ty(t)}
{Ty(e), Fo(Upro/Bill), }
{? Ty(e t)}
{Ty(e), Fo(John)}
(c) Scanning like From Ty(e→t) node, two daughters are projected. The left daughter is filled with the verb, but the right node is open with a requirement of Ty(e). There is one unfixed node of the same type. Merging John and the object node yields a tree structure from which all nodes can be decorated so that their requirements are met: {? Ty(t)}
{Ty(e), Fo(Bill)}
{? Ty(e t)}
{Ty(e (e t)), Fo(Like)} {? Ty(e), } {Ty(e), Fo(John)}
MERGE
(d) Compilation {Ty(t), (Fo(Like (John)(Bill)), ◊}
Japanese ‘gapless’ relatives
. Structure building in Japanese . The structural ambiguity Unlike English, Japanese is a head-final language, that is, verbs are preceded by their arguments and adjuncts, complementisers follow the complement clause, and relative clauses precede their head nominal. Arguments and adjuncts can be put in any order, and it is also a pro-drop language. This means the structural ambiguity multiplies as a sequence of NPs and PPs are scanned; each phrase may or may not be in the main clause. For example, the string of (4) may be followed by a COMP as in (5), or may be followed by a noun as in (6). In (5), John-ga ringo-o tabeta turns out to be a complement clause, and in (6), ringo-o tabeta a relative clause. John and ringo may or may not be in the matrix clause: (4) John-ga ringo-o tabe-ta John-nom apple-acc eat -past ‘John ate the apple’ (5) [John-ga ringo-o tabe-ta] -to Bill-ga it -ta. John-nom apple-acc eat -past-comp Bill nom say-past ‘Bill said that John ate the apple’ (6) John-ga [ ei ringo-o tabe-ta] hitoi -o sitteiru. John-nom apple-acc eat -past person-acc know ‘John knows the person who ate the apple’ It seems that there is no way to determine in parsing what structure to assign until the end or at least the head of the embedded clause (e.g. the complementiser, the head nominal) is met. To model the parsing processes in Japanese, DS models each node as being introduced by a lexical action, projecting only a very weak structural relation to the root so that lexical inputs do not directly combine together to form a tree starting from the root node. The overall goal is the same as in head-initial languages; to resolve all underspecifications and obtain the truth-evaluable proposition. In this sense, interpreting Japanese is still a goal-driven process, no matter how weak it is.
. A sample derivation – example (4): John-ga ringo-o tabeta (a) Scanning John-ga ‘John-NOM’ This may or may not be in the main tree, so its relation to the root node is the weakest form expressible. -Ga NOM specifies some subject node in the tree (i.e. its
Akiko Kurosawa
mother in the final tree must be of Ty(t)), but this still does not determine which subject it is. {? Ty(t)}
{Fo(John), Ty(e), ? < > Ty(t), }
(b) Scanning ringo-o ‘the apple-ACC’ The case, -o, indicates it must have a node of Ty(e→t) (a predicate node) as its mother in the final tree. Again the node itself is introduced as unfixed in relation to the root. {? Ty(t)}
{Fo(Ringo), Ty(e), ? < > Ty(e t), }
(c) Scanning tabeta ‘ate’ Verbs are a driving force in articulating a tree structure. A two-place predicate such as tabe ‘eat’ creates two argument-nodes and projects a requirement for a VP node of Ty(e→t). A sequence of actions is defined by the verb, with the requirement of Ty(t) at the root node as their trigger: {? Ty(t)} {? Ty(t)} {Ty(e), Fo(Upro)}
{? Ty(e t)}
{Ty(e), Fo(Vpro)}
{Ty(e (e t)), Fo(Tabe), }
(d) Merge If nothing follows, then two place-holders, Upro on the subject node and Vpro on the object node in (c), are merged with John-ga and ringo-o respectively. This is the same process as in the regular Left-Dislocation Structure (see (2)).
Japanese ‘gapless’ relatives {? Ty(t)}
John
{? Ty(t)}
Ringo
{? Ty(e t)}
{Ty(e), Fo(Upro)} MERGE MERGE
{Ty(e), Fo(Vpro)}
{Ty(e (e t)), Fo(Tabe), }
(e) Compilation The string is compiled at the root node as a proposition:1 {Ty(t), Fo (Tabeta (Ringo) (John)), ◊} ate apple John
. Relatives in English In DS, relatives are analysed as a pair of linked structures: a tree is defined as LINKed to another if it has a copy of an element in the main tree. In English, a relative pronoun introduces the copy in one such LINKed tree. (7) Johni , whoi I like ei , . . .
(I = the speaker = Bill)
(a) a LINK transition – result of processing who {? Ty(t)}
the head
{Ty(e), Fo(John)} {? Ty(e t)} LINK
the new tree
{? Ty(t)} {Ty(e), Fo(John), }
Linked node Unfixed node
The relative pronoun projects an annotation. The new tree has a requirement ? Ty(t) and a copy of Fo(John) at node n somewhere in the new tree. As the node location for Fo(John) is not known yet, it is represented as an unfixed node as in (a).
Akiko Kurosawa
(b) Scanning like
(c) Result of processing the relative {? Ty(t)} {? Ty(t)}
{Ty(e), Fo(John)} {? Ty(e t)} {Ty(e), Fo(John)} {? Ty(e t)}
LINK
LINK
{? Ty(t)}
Like(John)(Bill), {? Ty(e t)}
Bill/Upro
Bill Like {Ty(e), Fo(John)}
Like(John)
Vpro MERGE
Like
John
The rest of the updating process till the unfixed node is identified as the object node of the LINKed tree in (b) is exactly the same as that of the left dislocation case as in 2.3. Then the proposition of relative clause is compiled as (Like(John)(Bill)) as in (c).
. Relative clauses (with a gap) in Japanese As Japanese is a head-final language, the head nominal is preceded by the relative clause. An initially identical string may or may not involve an embedded clause as illustrated in examples (4)–(6). Modern Japanese lacks a relative marker which ensures the LINK transition as in languages such as English (e.g. who) and Arabic (e.g. illi), so each predicate merely shows that there is a disjunction: this is the end of a sentence OR the string is followed by something.2 Then how do we interpret (8) below as containing a LINKed structure? Are all the constituents initially introduced as unfixed nodes, with verbs constructing tree structures, just like a simple sentence in (4)? At which point is the LINK transition made? Consider the following example: (8) John-wa [ei okane-o tukanda] dorobooi -o tukamaeta. John-top money-acc grabbed thief -acc caught ‘John caught the thief who grabbed the money’
Japanese ‘gapless’ relatives
(a) John-wa okane-o tukanda | doroboo-o . . . {? Ty(t)} LINK {Ty(e), Fo(John)}
{Ty(e), Fo(Doroboo) ?< > Ty(e t), } ‘thief ’
{Ty(e), Fo(Okane) {? Ty(t)} ?< > Ty(e t)} ‘money’ {Ty(e), {? Ty(e t)} Fo(Upro)}
MERGE
{Ty(e), Fo(Vpro)}
{Ty(e (e t)), Fo(Tukanda)} ‘grabbed’ MERGE
The above figure illustrates the processing of doroboo-o. The first predicate, tukanda, does not work as a relativiser, as it no longer has a distinction as the adnominal form. The string John-wa okane-o tukanda might be interpreted as a simple sentence, ‘John grabbed the money’. The next lexical input, doroboo, is a good candidate for the head on the basis of our general knowledge in which a thief is associated with an action of grabbing money. However, it is not guaranteed structurally, as doroboo can belong to another embedded clause.3 What indicates that doroboo is interpreted as the head is phonological information. Okane-o tukanda | doroboo has a downstep after the verb, showing this string constitutes a phrase, ‘the thief who grabbed the money’.4 Using this information, the string is interpreted on-line. Phonological information interacts with lexical information in structure building and plays an important role when it reflects complex syntactic structure and has a distinctive function in case the structure is not yet fully specified. The downstep expresses a great deal how two adjacent constituents are related to each other; whether they form one domain or not. In the above case, if doroboo is not the head of a relative clause but a nominal that belongs to another embedded clause, then the downstep is blocked and doroboo is realised as high pitched as tukanda, indicating that this is the beginning of a new domain.5,6 Thus the downstep serves as a disambiguation device.7 So Doroboo is introduced as the head of the relative clause. Okane-o is merged with Vpro on the object node instructed by its case information, and the head, doroboo, provides a value to Upro decorating the subject node, hence providing the shared element in the LINKed tree. In (b) below, after the last constituent, tukamaeta ‘caught’, is scanned, dorobooo is slotted into the sister node projected by the matrix verb. Then the first constituent, John, is finally incorporated into the structure, merging with a placeholder on the matrix subject node.
Akiko Kurosawa
(b) Scanning tukamaeta {? Ty(t)} {Ty(e), Fo(John)} {? Ty(e t)}
{Ty(e), Fo(Doroboo) ‘thief ’ LINK
{Ty(e (e t)), Fo(Tukamaeta), } ‘caught’
{Ty(t), Fo(Tukanda(Okane)(Doroboo/Upro))}
{Ty(e), Fo(Doroboo/Upro)}
{Ty(e t), Fo((Tukanda)(Okane))} {Ty(e), Fo(Okane)} ‘money’
{Ty(e (e t)), Fo(Tukanda)} ‘grabbed’
. Gapless relatives – problems for the LINK analysis? In defining LINK structure as having a common element in the structure built from the two clauses, the DS analysis reflects the construal of regular relatives, in which some item needed to complete the interpretation of the relative is taken to be provided by a copy of the head nominal. However there are so-called gapless relatives which do not seem to have an empty node to be fulfilled, and this seems to pose a problem. Gapless relative clauses in Japanese are divided into three categories: a. ‘Pseudo relatives’ b. Appositional Gapless Relatives (‘Content’ clauses) c. Gapless relatives whose head leads an adverbial clause8 This paper focuses on the first one, ‘pseudo relatives’, followed by a brief look at the appositional gapless relatives in 6.6, which head-initial language such as English also have.
. ‘Pseudo relatives’ as gapless relatives ‘Pseudo-relatives’,9 are found in languages like Japanese:10
Japanese ‘gapless’ relatives
(17) [Keeki-ga yakeru] nioi-wa subarasii. cake-nom bake smell-top is-lovely (lit.) ‘The smell that the cake is baking is lovely’ ‘The smell of the cake baking is lovely’ (18) [Seetaa -ni wain-o kobosita] simi-ga oti -nai. Jumper-loc wine-acc spilled stain-nom come_off-neg ‘the stain that (I made by) spilling wine on the jumper doesn’t come off ’ In these examples, there is no ‘gap’ in the relative clause that is co-indexed with the italicised head noun. Contrast (18) and (19): (19) [ ei Seetaa -ni wain-o kobosita] hitoi jumper-loc wine-acc spilled person ‘the person who spilled wine on the jumper’ Teramura (1975–1978/1992) turns his attention to the range of nouns that are typically used as a head of this type of relative clause: ‘perception’ nouns such as nioi, ‘smell’, oto, ‘sound’, kookei, ‘sight’, koe, ‘voice’ and so-called relational nouns such as usiro ‘back’, yokuzitu ‘the next day’, and gen’in ‘cause’. Examples of relational nouns are given below: (20) [Humiko-ga suwatta ] usiro-no mado -ni -wa momizi-ga Humiko-nom sat back-gen window-loc-top maple-nom aokatta. green_was ‘Behind where Humiko was sitting, there were green maple leaves outside the window’ (Yasunari Kawabata Senba-zuru ‘The Thousand Cranes’) (21) [kodomo-ga umareta ] yokuzitu -wa Kurisumasu-datta. Child -nom was_born next_day -top Christmas-was (lit.) ‘The next day that the child was born was Christmas’ (22) [Kazi-ga hirogatta] gen’in-wa hakkiri_si-nai. fire-nom spread cause- top is_clear -neg (lit.) ‘the cause that the fire spread is unclear’ Relative clauses in (20)–(22) express what is the converse of the concept of the head; in (20), while the head noun, usiro, means back, the position where Humiko sat is in the front of the window. In (21), the meaning of the head, yokuzitu, is the next day, and the time when the child was born is the day before. In (22), whereas the head, gen’in, means cause, the fact that the fire spread is the consequence. This relation expressed with gapless relatives is often found in locational and temporal expressions, including nouns such as mae, ‘front’, usiro, ‘back’, tonari,
Akiko Kurosawa
‘next (position)’, hidari, ‘left’, migi, ‘right’, ue, ‘upper position’, sita, ‘lower position’, mae, ‘earlier time’, ato, ‘later time’, and so on.11 As all these relatives have no gap, a question arises: how do we interpret the relation between the head nominal and the relative clause? In the following subsections from 6.2 to 6.5, I will examine several possible answers to this question.
. The two-place relation analysis The first possible analysis for pseudo relatives is the two-place relation analysis. In examples from (23) to (25), the head noun seems to be in the two-place relation, ‘Y is the . . . of X’, with the NP in the relative clause: X-gen Y (23) keeki-no nioi
Y of X ‘the smell of the cake’
(24) wain-no simi
‘the stain of the wine’
(25) tabako-no kemuri
‘the smoke of the cigarette’
This interpretation involves construing the head nominal as a relation between two arguments. This adds an argument to the representation of the head nominal (e.g. ‘the cake’ is added to ‘the smell’ to show the meaning relation of the relative and the head). Although this analysis applies to a number of cases, it is inadequate to explain some others such as in (26): (26) [Dareka-ga otiba -o taku ] kemuri-ga someone-nom fallen_leaves-acc burn smoke-nom tati_nobotte iru. rising -is ‘The smoke of someone burning fallen leaves is rising’ (27) *otiba-no kemuri ‘the smoke of fallen leaves’ ‘The smoke’ and ‘the fallen leaves’ do not seem to have the relation that holds between ‘the smoke’ and ‘the cigarette’. Among Japanese linguistic literature for machine translation, Narita (1994) proposes the following treatment for pseudo relatives: (28) Change the structure into the regular relative clause which contains the gap, and then translate it into English. (Narita 1994: 86) For example, (29) [gasu-ga morete-iru ] nioi gas-nom leaking-is smell (lit.) ‘the smell that the gas is leaking’
Japanese ‘gapless’ relatives
is transformed to (30) whose structure allows a more natural translation: (30) [ei morete-iru] gasui -no nioi leaking-is gas-gen smell ‘the smell of the gas that is leaking’ With this treatment, (26), repeated here as (31), is converted to (32): (31) [Dareka -ga otiba -o taku ] kemuri-ga tati_nobotte iru. someone-nom fallen_leaves-acc burn smoke-nom rising -is (lit.) ‘The smoke that someone is burning fallen leaves is rising’ (32) [Dareka-ga ei taku] otibaI -no kemuri-ga someone-nom burn fallen_leaves-gen smoke-nom tati_nobotte iru. rising -is ‘The smoke of the fallen leaves that someone is burning is rising’ This manipulation of transforming the structure, motivated by technical needs for machine translation, resolves certain problematic cases like (26)/(31). The main factor that makes the interpretation acceptable is the verb in the relative clause that supplies the missing link in the meaning relation between ‘the smoke’ and ‘the fallen leaves’. However Narita’s treatment fails to give the right interpretation for such examples as in (33) and (38): (33) [Kuruma-ga torakku-to butukatta] oto -wa sugokatta. car -nom lorry -with crashed sound-top was-dreadful ‘The sound of the car crashing with the lorry was dreadful’ (34) ?[ei torakku-to butukatta] kurumai -no oto -wa sugokatta. lorry -with crashed car -gen sound-top dreadful ‘The sound of the car that crashed with the lorry was dreadful’ (35) ?[Kuruma-ga ei butukatta] torakkui -no oto -wa sugokatta. car -nom crashed lorry -gen sound-top dreadful ‘The sound of the lorry that the car crashed with was dreadful’ In this case, supplying the (regular) relative clause is not sufficient, as changing the structure into the regular relative does not change the respect that the relation is set between ‘the sound’ and an individual picked on by the NP (either ‘the car’ or ‘the lorry’). (36)
?kuruma-no oto
‘the sound of the car’
(37)
?torakku-no oto
‘the sound of the lorry’
Akiko Kurosawa
The argument to be added to the head nominal cannot be clearly specified in this case. Intuitively we know it is wrong to attribute the source of the sound just to one of them. Another counter example is: (38) [Dareka -ga to -o tataku ] oto-ga sita. someone-nom door-acc knock sound-nom did (lit.) ‘I heard the sound that someone was knocking on the door’ ‘I heard the sound of someone knocking on the door’ Changing the structure into a regular relative gives: (39) *[ ei to-o tataku ] dareka i -no oto door-acc knock someone-gen sound ‘the sound of someone who knocks on the door’ (40) ?[ dareka -ga ei tataku ] to i -no oto someone-nom knock door-gen sound ‘the sound of the door that someone knocks on’ From these observations, it is now clear that there is not a single NP (even with a regular relative clause) that has a meaning relation with the head. It is the whole gapless relative clause rather than an NP that projects the implied relation and that has to project the required argument. (41) [Someone is burning the fallen leaves] The smoke . . . (42) [The car crashed with the lorry] The sound . . . (43) [Someone knocked on the door] The sound . . . This view also accommodates examples (17) and (18) that seem to fit the two-place relation analysis that we have looked at above: (44) [The cake is baking] The smell . . . (45) [I spilled wine on the jumper] The stain . . . The two-place relations, ‘X-no Y’ or ‘Y of X’, are in fact semantically indeterminate in nature like the possessive construction,12 whose meaning relation is interpreted contextually.13 In (17), ‘the smell of the cake’ does not necessarily mean ‘the smell of the cake baking’; it can be used to describe ‘the smell of the cake’ which is baked this morning and now on the plate. Unlike the two-place relations, gapless relatives in (17) and (18) describe how the smell and the stain are made. What is not linguistically encoded is how these relatives are related to the head as relatives. In the next sub-section, I will take up a parallelism found between gapless relatives and bridging reference.
Japanese ‘gapless’ relatives
. Bridging reference14 and gapless relatives Interestingly Japanese gapless relatives look like bridging reference in English. (46) I walked into the room. The chandeliers sparkled brightly. (47) Peter went to a Japanese restaurant. The waitress was from Osaka. In interpreting The chandeliers and The waitress whose referents are not explicitly mentioned, the hearers add the (bridging) assumptions in (48) and (49) below to the context: Bridging assumptions: (48) The room had chandeliers. (49) The waitress was working at the Japanese restaurant. Now if a Japanese gapless relative clause and the matrix clause are divided into two simple sentences, it yields a bridging reference construction as follows (shown in English): (50) The cake is baking. The smell is lovely. (51) I spilled wine on the jumper. The stain doesn’t come off. (52) Someone is burning the fallen leaves. The smoke is rising. (53) The car crashed with the lorry. The sound was dreadful. (54) Someone knocked on the door. I heard the sound. (55) The fire spread. The cause is unclear. The italicised NPs which are the head in the gapless relative construction are bridging NPs in bridging reference.15 The referent of the bridging NP is pragmatically inferred through a bridging assumption. Likewise the relation between the head nominal and the pseudo relative clause is retrieved by pragmatic inference via a contextual assumption which is based on general knowledge schemas. Contextual assumptions: (56) If the cake is baking, there is a smell of it. (57) If one spills wine, it makes a stain. (58) If someone burns fallen leaves, it makes smoke. (59) If the car crashes with the lorry, it causes a (dreadful) sound. (60) If someone knocks on the door, there is a sound of it. (61) If the fire spread, there must be a cause.
Akiko Kurosawa
Contextual assumptions, being combined with the proposition of the relative clause as another premise, yield contextual implications in the following manner: (62) If the cake is baking, there is P→Q contextual assumption a smell of it. The cake is baking. P proposition in the relative clause There is a smell of it.
Q
contextual implication (head)
(63) If the car crashes with the lorry, it causes a (dreadful) sound. The car crashes with the lorry. It causes a (dreadful) sound. In (62) and (63), the conclusion part includes the head nominal, ‘smell’ and ‘sound’ respectively, showing how the relation which is not encoded is provided through inference. The assumption that allows bridging reference and pseudo gapless relatives is generally the one that is readily accessible in our general knowledge schemas. Accessibility of assumption affects the acceptability of bridging and gapless relatives. It is argued in Matsui (1995, 1998) that the accessibility factors involve a pragmatic criterion of consistency with the principle of relevance proposed by Sperber and Wilson (1986/1995) which hearers can evaluate the resulting interpretation with. Criterion of consistency with the principle of relevance An utterance, on a given interpretation, is consistent with the principle of relevance if and only if the speaker might rationally have expected it to be optimally relevant to the hearer on that interpretation. Communicative principle of relevance Every act of ostensive communication communicates a presumption of optimal relevance. Sperber and Wilson’s claim is that an utterance is optimally relevant if and only if it has enough cognitive effects with no unjustifiable effort for the hearer to obtain them. Conversely less effort ensures higher acceptability.16 To demonstrate this point, compare (64) with (65) and (66): (64) a.
[Dareka -ga otiba -o taku ] kemuri someone-nom fallen_leaves-acc burn smoke ‘The smoke of someone burning fallen leaves’
Japanese ‘gapless’ relatives
b. [Dareka -ga otiba -o taku (toki deru) ] someone-nom fallen_leaves-acc burn (when is-produced) kemurii smoke ‘The smoke (that is produced when) someone burns fallen leaves’ (65) a. ?*[Hito -ga ie -o tateta ] gomi people-nom house-acc have_built rubbish ‘the rubbish of people having built a house’ b. [Hito -ga ie -o tateta (toki deru) ] gomi people-nom house-acc have_build (when is-produced) rubbish ‘the rubbish (that is produced when) people have built a house’ (66) a. ?*[Hito -ga sekitan-o hotta] gasu people-nom coal -acc have_dug gas ‘the gas of people having dug coal’ b. [Hito -ga sekitan-o hotta (toki deru) ] gasu people-nom coal -acc have_dug (when is-produced) gas ‘the gas (that is produced when) people have dug coal’ (examples slightly altered from those in Abe 1994) In (64), the interpretation which is the explicature of (64a) is given in (64b). However, in (65a) and (66a), the gapless relatives are NOT acceptable, while (65b) and (66b) show the same kind of relation. Building a house may produce rubbish, and digging coal may produce gas, but the association between them is not as strong as the one between burning the fallen leaves and the smoke, or the cake baking and the smell, and not accessible enough to allow them taking the form of a gapless relative clause and its head. It is presumably due to the low frequency of use (of those combinations) and complexity of the assumption. The contextual assumption needed for (65a) has two steps; when a house is built, an old building on the site has to be demolished, and if an old building is demolished, it produces rubbish. If the step is reduced to one, the acceptability increases a great deal as in (67) below:17 (67) [Hito -ga ie -o kowasita] gomi people-nom house-acc demolished rubbish ‘the rubbish of people having demolished a house’ (65a) and (66a) are unacceptable as the contextual assumption is considered not accessible enough given normal expectation of effect.
Akiko Kurosawa
. Saturation of hidden indexicals and ‘free’ enrichment processes We assume that there is a pragmatic instruction interacting with computational rules and lexical actions in interpreting utterances, and Dynamic Syntax assumes Relevance Theory developed by Sperber and Wilson (1986/1995) as its background. In Relevance Theory, examples of ‘and’ conjunction such as (69) and (70) are argued as a case of pragmatic enrichment (Carston 1998, 2000): (68) It’s summer in England and it’s winter in New Zealand. (P and Q = Q and P) (69) Sue got a PhD and (then) became a lecturer.
(P and Q = Q and P)
(70) She shot him in the head and (as a consequence) he died instantly. (P and Q = Q and P) In (68), changing the order of two conjuncts/propositions does not affect the truthconditional meaning, whereas in (69) and (70) it does. A temporal relation in (69) and a cause-consequence relation in (70) are NOT linguistically encoded and determined but pragmatically inferred. The pragmatically inferred relation enriches the logical form which then develops into an explicature, i.e. a communicated assumption. While the enrichment is the process of recovering unarticulated constituents, there is a process of supplying contextual values to indexicals which is termed ‘saturation’ (Recanati 1993: 233–268). A saturation process does not only supplies values to overtly marked indexicals such as pronouns and demonstratives but also supplies values contextually to ‘hidden indexicals’, which are invisible but linguistically present constituents. Examples include certain ‘relational’ terms and genitive constructions (Carston 2000): (71) The winners each get £1,000.
[winners of what?]
(72) I like Sally’s shoes.
[shoes in what relation to Sally?]
There is a covert slot to be filled, and these are semantically incomplete until the constituent is contextually supplied in answering the bracketed question, and there is a lexical item which requires completion (e.g. winner, genitive marker). The question is whether the pseudo gapless relative is a case of enrichment or saturation. In case of pseudo gapless relatives whose head is a relational noun such as usiro ‘back’, yokuzitu ‘next day’ and gen’in ‘cause’ as in (20)–(22), they have a covert slot to be filled in like winner in (71), which calls for saturation (e.g. the back of what?). Then the proposition of the relative clause is there to fill the slot.18
Japanese ‘gapless’ relatives
However, no lexical item dictates ‘completion’ in, for example, (17), repeated here as (73), whose explicature is shown in (74): (73) [Keeki-ga yakeru] nioi -wa subarasii. Cake -nom bake smell-top is_lovely (lit.) ‘The smell that the cake is baking is lovely’ (74) [Keeki-ga yakeru (-toki suru) ] nioi-wa subarasii. cake-nom bakes (when it_smells) smell-top is-lovely ‘The smell (that is made when) the cake is baking is lovely’ None of the lexical items in Keeki-ga yakeru ‘the cake bakes’, and nioi ‘the smell’ has such a requirement. The grammatical structure, the relative clause in this case, constrains the interpretation, forcing the hearer to look for a certain meaning relation between the relative clause and the head, as between them there must be some relation as long as they are adjacent in the form of the relative clause and the head. In this sense, a hidden slot is present though in a very abstract way. One might say that nioi ‘the smell’ evokes a question, ‘the smell of what?’, and that there is a covert slot as well. But if so, it seems that more and more nonrelational expressions are subject to having this type of slot: The sound of what and caused by what, the rubbish of what, and the gas of what, and so on. Even for the nouns such as book and door, one can ask for slots: the book of whom, on what, the door of which room, etc. Then we are not sure how many ‘slots’ we will have. In this sense, words are ‘relational’ to some extent, and it is undetermined how many slots we want.19 However, more importantly, the pragmatic inference that supplies the value for the meaning relation to the pseudo gapless relatives is based on the contextual assumption accessible from encyclopaedic knowledge schemas, which is the mechanism of interpretation also for enrichment. As Carston (1998) puts it: . . . these relations being supplied perhaps by highly accessible general knowledge schemas concerning relevant ways in which events connect up. (Carston 1998: 225)
It is presumably general knowledge schemas on the basis of which the interpretation of (17)/(73) is determined, and this non-linguistic aspect of interpretation processes is what is shared by ‘and’ conjunction, bridging reference, and pseudo relatives. In the next subsection, I turn to the Dynamic Syntax proposal in which I shall argue that an event variable is employed in modeling the gapless relatives. The event variable stands for the proposition in the relative clause. However, unlike a bound variable in quantification, it is not sufficient to fully determine the interpretation itself; its role is to express that the whole relative clause has a meaning relation with the head. How it is related is only pragmatically inferred.
Akiko Kurosawa
. The event variable analysis As we have seen, the relation between the head and the gapless relative clause may vary. My proposal is to represent it employing an event variable so that these various and complex relations can be expressed in a unified way. The notion of event variable is not new. The idea that, the referential relation between a proposition and the event it denotes is one of reference, was introduced by Davidson (1967). Davidson directed his attention to the logical (entailment) relation between (a) and the others in (75) below: (75) a. b. c. d. e.
Jones buttered the toast in the bathroom with a knife at midnight. Jones buttered the toast in the bathroom. Jones buttered the toast with a knife. Jones buttered the toast at midnight. Jones buttered the toast.
Davidson proposes that (75a) has a logical form as in (76): (76) ∃x (butter (x, Jones, toast) & in (x, bathroom) & with (x, knife) & at (x, midnight)) The verb butter stands for an event, and the event variable, x, is bound by an existential quantifier, asserting the existence of the event. (76) is paraphrased as an event of Jones’ buttering the toast took place, its place was the bathroom, the instrument used for it was a knife, and the time of it was at midnight. Prepositions are treated as a predicate, with the event variable and a nominal as their arguments. Separating each prepositional phrase from the verb as a conjunct using the event variable, now it is possible to show the entailment relation between (a) and the others in (75), as a conjunction entails each of its conjuncts: (P & Q) → P. Davidson’s proposal is extended by Parsons (1990) to account for various phenomena such as the relation between ‘Agatha saw Brutus stab Caesar’ and ‘Agatha saw the stabbing of Caesar by Brutus’. Anaphoric properties of tense are expressed through the use of event variables by Kamp and Reyle (1993: 483–690), and Steedman (1997). In Dynamic Syntax, a metavariable is used as a label si for a propositional formula P as shown below: (77) si : P This variable can be used as an argument for any relation/predicate of appropriate type. Perrett (2000) gives an account of temporal and causal relations in Hadiyya, which is a verb-final language of Cushitic group of the Afro-Asiatic family:
Japanese ‘gapless’ relatives
(78) S1 precedes the Utterance time
Precede (SUTT ) (S1 )
(79) a. S1 precedes S2 b. S1 causes S2
Precede (S2 ) (S1 ) Cause (S2 ) (S1 )
Past tense
Tense has anaphoric properties in that it expresses a temporal relation between the event variable associated with the current sentence and some other event time, and, which event time that is, is determined in context. This anaphoric relation of tense is inferred like the case of pronouns, whose interpretation is fixed by other representations in context, e.g. from expressions in the preceding clauses and sentences. This observation underlies the DS representation of tense. The temporal metavariable in (78), like the Formula metavariables, [Fo(U)], has no denotational content and a value will have to be supplied through the interpretation process. Si is an argument of a tense relation indicated by the verbal morphology in (78) above, e.g. the suffix -ed in English, and -u (or -o), a simple perfect morpheme in Hadiyya. The time value for S1 , which interval at some time in the past of the utterance time, is contextually inferred by non-linguistic knowledge. In (79), the particular interpretation of the temporal relation in (a) and the causal relation in (b) between conjoined clauses may be pragmatically achieved. Similarly, the event variable that stands for the proposition is used to represent the whole event in the gapless relative clause.20 The anaphoric relation (i.e. there is a common element) between two clauses is expressed by using a metavariable. The variable provides the argument for the relation between the gapless relative clause and the head. The structure-building process of (17)/(73) is demonstrated below: (80) [Keeki-ga yakeru] nioi-wa subarasii. cake-nom bakes smell-top is-lovely ‘The smell of the cake baking is lovely’ As seen in Section 5, phonological information plays a role in determining the structure on-line. A downstep between yakeru and nioi provides information that they belong to the same phonological domain reflecting the structure, which indicates that nioi is the head of a LINKed structure constructed from the immediately previous clause rather than a constituent of some separate structure. At this juncture, keeki-ga yakeru nioi can be unambiguously identified as a relative clause and its head. However, unlike the regular relatives, the tree constructed from keeki-ga yakeru has no outstanding requirement and no node for nioi to merge.
Akiko Kurosawa { Ty(t)}
{Fo(S1, Nioi (S1)), {Fo(Subarasii) Ty(e)} ‘smell’ Ty(e t)} ‘lovely’ LINK {Ty(t), Fo(S1: (Yakeru(Keeki))}
{Fo(Keeki), ‘cake’ Ty(e) }
{Fo(Yakeru), ‘bake’ Ty(e t)}
S1 = bake (cake)
The event of the relative clause is expressed with S1 , which is the label for the proposition yakeru (keeki), ‘bake (cake)’, and it appears as an event variable on the node for the head, nioi, ‘smell’, showing that this is the common element between the two clauses. S1 is used as an argument of the head, nioi, expressing that the event of cake baking, S1 , has a property of smell. In this way the relation which is pragmatically inferred is represented. The event variable does not supply the interpretation of the relation. It is from a contextual assumption, ‘if the cake bakes, there is a smell of it’ and the relative clause, ‘the cake bakes’, as another premise, that the conclusion, ‘there is a smell of it’, is drawn. This inference provides the basis for understanding the relation between the proposition of the relative clause and the individual denoted by the head.
. Appositional gapless relatives (‘content’ clauses) Another form of gapless relative clause, the appositional relative is known as a ‘content clause’ in the Japanese linguistic literature (Masuoka 1994): (81) [ John-ga Mary-o aisiteiru ] (to iu ) zizitu John-nom Mary-acc love comp fact ‘the fact that John loves Mary . . . ’ All the argument nodes in the relative clause being filled with formulae, the head, zizitu ‘fact’, cannot have an argument-predicate relation with the verb, aisiteiru ‘love’, hence no ‘gap’. The head, as given by zizitu ‘the fact’, is identified with the structure projected by the whole relative clause; i.e. the relative clause gives the ‘content’ of the ‘fact’, and is in apposition to the head in this sense. The DS analysis is that the shared element is the label s1 for the formula at the root node projected from the relative clause, in the same manner as for pseudo relatives.
Japanese ‘gapless’ relatives ? Ty(t)
Ty(e), Fo(S1, Zizitu(S1)) ‘fact’ LINK S1 : Aisiteiru (Mary) (John)
John
Aisiteiru (Mary)
Mary zizitu (S1) ‘fact’
Aisiteiru ‘love’ S1 is the fact
The difference between the pseudo relatives and the appositional relatives is that pragmatic inference processes which are requisite for the former are not necessary for the latter. Despite differences between the two structures, we can nevertheless maintain the definition of LINK for both pseudo and appositional ‘gapless’ relative clauses. There is a common element present in both LINKed and main trees – the event variable. That the whole event expressed by the relative clause has a meaning relation with the individual denoted by the head is shown through it. Looking back and comparing head-initial (e.g. English) and head-final (e.g. Japanese) relative clauses, we see that there is asymmetry between them. In headinitial relative structures, only appositional (i.e. fact-that) constructions provide an event variable as the common element. In head-final relative constructions, the head dictates that an event-variable interpretation is available in both types of gapless relatives.
. Conclusion In this paper, Japanese relatives are analysed in the Dynamic Syntax perspective as linked structures sharing a variable. The process of projection is the reverse of English, an argument variable in the LINKed structure copied over as the head. On the assumption that the structure projected is semantically transparent, event variables labelling the propositional formula are used as a common element. This manipulation of event variables allows a natural characterisation of the structure required for relative clause construal while respecting the pragmatic nature of the inference process. The projection of structure for natural languages needs to allow integration between structural and pragmatic factors.
Akiko Kurosawa
Notes * This paper is an extension of a talk given jointly by myself and Ruth Kempson at the Second International Conference on Contrastive Semantics and Pragmatics in Cambridge. I am indebted to Ruth Kempson for insightful comments and constructive suggestions on earlier drafts of the paper; all remaining errors are mine. I have also benefited greatly from discussions with Tomoko Matsui. . In Japanese, nominative and accusative case markers are optionally omitted (e.g. John ringo tabeta). Each argument node is always introduced as initially ‘unfixed’ in the structure (see 2.3). Furthermore, arguments themselves need not be explicitly expressed. (e.g. e e tabeta). Even in such a case, the tree structure is articulated by the verb as in (c) in 3.2. Then metavariables on two argument nodes are substituted with John and ringo. This anaphora resolution may involve a pragmatic process. . In Classical Japanese, the adnominal form (rentai-kei, e.g. aru ‘exist’) of a predicate which is distinctive from the conclusive form (shushi-kei, e.g. ari ‘exist’) in most cases played a role of the relativiser, indicating that some kind of nominal should follow. However, in Modern Japanese, the distinction is lost from the majority of predicates, and its role is much weaker. Note that Modern Korean has the adnominal form which is distinctive from the conclusive form and works as a relativiser. . Strictly it is not the noun which is the head but the variable X which is an argument of the predicate projected by the noun, Doroboo, (X, Doroboo(X)). I sidestep this distinction in what follows, referring loosely to the noun as providing the head (see Kempson et al. 2001: 35, 115, 137 for the internal structure of the noun). . See Kubozono (1987: Chap. 5; 1995: 44, 98–116) for the detailed discussion on phonological phrasing reflecting structure. . For instance, (ii) below shows how the downstep is blocked after aisita while it occurs in (i): (i)
Watasi-ga aisita | supai I -nom loved spy ‘the spy I loved’
(ii) Watasi-ga aisita supai-ga katuyaku-suru monogatari I -nom loved spy -nom play_an_active_role story ‘the story I loved, in which a spy plays an active role’ In (ii), supai ‘spy’ is a constituent of the second relative clause, ‘(in which) a spy plays an active role’. . In Japanese, a downstep occurs before the head nominal if at least one of the two constituents of the relative clause is lexically accented. In Okane-o tukanda | doroboo, tukanda has a lexical accent on ka. If both constituents in the relative clause are accentless as in (i) below, the three constituents merge, forming a single intonational phrase. The second and the third constituents lose their original low pitch on the first nuclei in (ii), leaving the domain-initial nucleus, o, as low pitched as in (iii):
Japanese ‘gapless’ relatives
(i)
Okane-o suteta doroboo money-acc threw_away thief ‘the thief who threw away the money’
(ii) [o ka ne o] [su te ta] [do ro bo o] (iii) [o ka ne o su te ta do ro bo o] This intonational phrasing is blocked if doroboo is not the head nominal but a constituent of another syntactic domain (e.g. another relative clause), indicating that there is a border between okane-o suteta and doroboo. . This paper discusses phonological information confining it to the one which is crucial in structure building of LINK on given examples. . There are a considerable number of nouns which are the head of a relative clause and at the same time function as a (part of) subordinator leading an adverbial clause in Japanese. Among them are: toki ‘time/when’; mae ‘front/before (temporal adverbial), front/in front of (locational adverbial)’; ato ‘back/after’ (temporal); usiro ‘back/behind’ (locational); tokoro ‘place, point, scene/as’; tame ‘reason/because (causal), purpose/in order to (purpose)’; baai ‘case/if ’; ue ‘top/in addition to’. Each word has a different degree of grammaticalisation, but all belong to both lexical and functional categories to some extent. . Teramura (1975–1978/1992) used the term Soto-no Kankei ‘the outer relation’ for the relations between the head noun and the modifying clause for both pseudo and appositional (content) gapless relatives while that of the regular relatives is Uchi-no Kankei ‘the inner relation’. Two sub-types with nouns of perception and relation as the head are termed Pseudo relatives in Abe (1994), and the other three sub-types are called Content clauses in Masuoka (1994). . Korean also has this type of gapless relatives. . The same noun may be used in regular relatives and in appositional gapless relatives as well as in pseudo relatives as in the examples below. So it is NOT the noun itself that determines the type of relatives: (i)
[ei koohyoo sareta] kekkai was_published result ‘the result i that ei was published’
(regular relative clause)
(ii) Syusyoo -ga hatugen_si-te, [sizi -ritu -ga sagaru] (to iu) prime minister-nom speak -and approval rate-nom drop (comp) kekka-to natta. result-ptl became (appositional gapless relative clause) ‘The prime minister made a remark and (it) brought about the result that the approval rate (for the government) dropped’
Akiko Kurosawa
(iii) [Syusyoo -ga hatugen_sita] kekka, sizi -ritu -ga sagatta. prime minister-nom spoke result approval rate-nom dropped (pseudo relative clause) ‘As a result of the prime minister’s having made a remark, the approval rate (for the government) dropped’ . See Kempson (1977: 123–138) for a discussion. . See Blakemore (1992: 83) for examples in a context. . Cf. Matsui 1992, 1995, 1998, 2000; Wilson & Matsui 1998. . The pseudo relatives can invariably be paraphrased as bridging reference sentences. However the reverse is not always true; e.g. (46) and (47) are not acceptable as gapless relatives in Japanese. A general association between two NPs such as a room and chandeliers, or a restaurant and a waitress may allow a bridging reference interpretation as well as the relation between an event in the first clause and an NP in the second clause as in (50)–(55). However, in pseudo relatives, the relation is restricted to the one between the event described in the relative clause and the head NP. . This paper does not discuss an interesting case of gapless relatives seen in Japanese haiku poetry, each consisting only of 17 syllables: The contextual assumption which the reader is required to make in order to connect the head and the relative clause is sometimes far less accessible than in the examples we have seen, but nevertheless it is acceptable (examples below). Relevance Theory explains this in terms of the balance of the processing effort and the cognitive effect; extra processing effort demanded should be offset by extra effect, like any element of indirectness seen in the case of metaphor, repetition, etc. puts the hearer to extra effort, and promises extra effects. In case of haiku poetry, unusual, less frequently used general knowledge schemas (hence low accessibility) encourages a number of weak implicatures as well, (readers are responsible for their recovery), hence poetic effects. (i)
[Choo -no sita zenmai -ni niru ] atusa-kana butterfly-gen tongue flowering_fern-to resemble heat -ptl (lit.) ‘the heat that the butterfly’s tongue resembles a flowering fern’ (Ryunosuke Akutagawa)
(ii) [Uki tosi -no ima aratamaru] kiteki -kana sorrowful year-gen/nom now is-renewed steam_whistle -ptl The sorrowful year is now being renewed, which (is known by the sound of) the steamwhistle (that all the ships in port blow at midnight) (Tae Koizumi) The expected cognitive effects are more specific in these haiku examples, and the interpretation process is effect-driven, while the interpretation process with more accessible contextual assumptions in examples like (17) and (18) is effort-driven. Approaches such as cognitive grammar (Yamanashi 1995) and Frame Semantics (Matsumoto 1997) might not be able to provide very adequate accounts in this respect, as the head noun of the pseudo relatives is explained either as ‘a default value drawn from the event schema’ (Yamanashi 1995: 180–182) or a member which bears a role in a frame, where
Japanese ‘gapless’ relatives
the frame is associated with prototypical scenes (Matsumoto 1997). These descriptions do not seem to accommodate unconventional cases of gapless relatives. . This example is due to Jun’ichi and Chihiro Tsuchiya (personal communication). . To be exact, there needs to be the bracketed constituent with which the proposition in the relative clause fills the slot: (i)
[ Humiko-ga suwatta ] (basho -no) usiro Humiko-nom sat place-gen back ‘Behind (the place) Humiko was sitting’
(ii) [Kodomo-ga umareta ] (hi -no ) yokuzitu child -nom was_born day -gen next_day ‘the next day (of the day ) the child was born’ . It is of some interest that Marten (1999, 2002) advocates that ALL verbs underspecify their type. The specific choice of type is selected according to the number of optional arguments. This argument is well suited to explain this phenomenon. Carston (2000) argues that not all the ‘hidden indexicals’ receive a contextual value on a particular occasion of use, and only those constituents which are relevant are recovered (e.g. I’ve eaten might contain four hidden indexicals as in I’ve eaten [x] [in manner y] [at location l] [within time span t], but in I must wash my hands: I’ve eaten, only the manner constituent, using my hands, is necessary to be recovered and the others are redundant). . The motivation for the use of the event variable in DS is different from that of Davidson’s and Parsons’, and there is no need to separate prepositional phrases as conjuncts.
References Abe, Y. (1994). Rentai Shushoku no Sho-Mondai. [Issues on Noun Modification.] In Y. Takubo (Ed., 1994) (pp. 153–171). Blakemore, D. (1992). Understanding Utterances. Blackwell. Carston, R. (1998). Pragmatics and the Explicit – Implicit distinction. Ph.D. thesis, University College London. Carston, R. (2000). Explicature and Semantics. UCL Working Papers in Linguistics, 12, 1– 44. To appear in S. Davis & B. Gillon (Eds.), Semantics: A Reader. Oxford: Oxford University Press. Davidson, D. (1967). The logical form of action sentences. In N. Rescher (Ed.), The Logic of Decision and Action (pp. 81–95). Pittsburgh: University of Pittsburgh Press. Kamp, H., & Reyle, U. (1993). From discourse to Logic. Dordrecht: Kluwer Academic Publications. Kempson, R. M. (1977). Semantic Theory. Cambridge University Press. Kempson, R. M., Meyer-Viol, W., & Gabbay, D. (2001). Dynamic Syntax: The Flow of Natural Language Understanding. Oxford: Blackwell. Kubozono, H. (1987). The Organization of Japanese Prosody. Ph.D. thesis, University of Edinburgh.
Akiko Kurosawa
Kubozono, H. (1995). Go-keisei to On’in Kozo. [Word Formation and Phonological Structure]. Tokyo: Kuroshio Shuppan. Kurosawa, A. (2000). ‘Soto-no kankei’-no rentai shushoku ni tsuite. [On Gapless Relatives.] Paper presented at the 13th conference of Japanese Language Teaching, Vienna. Kurosawa, A. (in prep.). Japanese Relative Clauses from a Dynamic Syntax Perspective. Ph.D. thesis, King’s College London. Marten, L. (1999). Syntactic and Semantic Underspecification in the Verb Phrase. Ph.D. thesis, SOAS, University of London. Marten, L. (2002). At the Syntax-Pragmatics Interface: Verbal Underspecification and Concept Formation in Dynamic Syntax. Oxford: Oxford University Press. Masuoka, T. (1994). Meishi Shushoku Hyogen no Setsuzoku Keishiki -Naiyo-setsu o Chushi ni. [Optional Complementisers for Content Clauses in Noun Modification.] In Y. Takubo (Ed., 1994) (pp. 5–27). Matsui, T. (1992). Bridging Reference and the Notions of ‘Topic’ and ‘Focus’. In UCL Working Papers in Linguistics, 4 (pp. 239–258). University College London. Matsui, T. (1995). Bridging and Relevance. Ph.D. thesis, University College London. Matsui, T. (1998). Pragmatic criteria for reference assignment: A relevance-theoretic account of the acceptability of bridging. Pragmatics & Cognition, 6, 47–98. Matsui, T. (2000). Bridging and Relevance. Amsterdam: John Benjamins. Matsumoto, Y. (1997). Noun-modifying Constructions in Japanese – A Flame-semantic approach. Amsterdam: John Benjamins. Narita, H. (1994). Rentai Shushoku-setsu no Kozo Tokusei to Gengo Shori – Nihongo rashii Hyogen no Kikai Hon’yaku to Oyo Gijutsu. [Syntactic Idiosyncracy and Processing of Noun Modification – The Machine Translation of Idiosyncratic Japanese Expressions and Applied Technology.] In Y. Takubo (Ed., 1994) (pp. 67–126). Parsons, T. (1990). Events in the Semantics of English – A Study in Subatomic Semantics. Cambridge, MA: The MIT Press. Perrett, D. L. (2000). The Dynamics of Tense Construal in Hadiyya. Ph.D. thesis, SOAS, University of London. Recanati, F. (1993). Direct Reference – from Language to Thought. Oxford: Blackwell. Sperber, D., & Wilson, D. (1986/1995). Relevance [Second edition]. Oxford: Blackwell. Steedman, M. (1997). Temporality. In J. van Benthem & A. ter Meulen (Eds.), Handbook of Logic and Language, Chap. 16 (pp. 895–938). Amsterdam: Elsevier Science B. V. Takubo, Y. (Ed.). (1994). Nihongo no Meishi Shushoku Hyogen. [Noun Modification in Japanese.] Tokyo: Kuroshio Shuppan. Teramura, H. (1975–1978/1992). Rentai Shushoku no Shintakusu to Imi. [Syntax and Semantics of Noun Modification in Japanese.] In H. Teramura (Ed., 1992), Teramura Hideo Ronbun-shu I (pp. 157–336). Tokyo: Kuroshio Shuppan. Wilson, D., & Matsui, T. (1998). Recent approaches to bridging: Truth, coherence, relevance. In UCL Working Papers in Linguistics, 10, 173–200. University College London. Yamanashi, M. (1995). Ninchi Bunpo-ron [Cognitive Grammer]. Tokyo: Hitsuji Shobo.
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.1 (335)
Constraint interaction at the semantics/pragmatics interface The case of clitic doubling Javier Gutiérrez-Rexach The Ohio State University, USA
Introduction Clitic pronouns are prosodically weak proforms which are morpho-syntactically dependent and normally attach to a verbal host. The study of these proforms faces two main challenges. First, it is necessary to provide a formal semantics of clitics and clitic doubling constructions in a manner that goes beyond the impressionistic characterizations which are often found in the literature. Second, cross-linguistic variation has to be addressed in a way in which semantic and pragmatic factors are factored in as the main triggers of variation. In Gutiérrez-Rexach (1999a, 2001a), a strong correspondence hypothesis at the syntactic and semantic levels is proposed. In the syntax, clitics are specified as definite determiners, as proposed by Torrego (1995) and Uriagereka (1995). This assimilation of clitics to determiner heads is in the spirit of Postal’s (1966: 203) proposal: “My basic claim is that the so-called pronouns I, our, they, etc. are really articles, in fact types of definite articles.” As a matter of fact, English pronouns exhibit overt determiner-like behavior in some cases: (1) a. Did you see us guys? b. Who insulted you men? The phenomenon of clitic doubling, also called pronominal reduplication, can be seen as a clear instantiation of Postal’s idea. The difference between Spanish-type languages, which allow clitic doubling, and French-type languages, which do not, can be cast in the following terms: Spanish-type clitics are specified as selecting for
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.2 (336)
Javier Gutiérrez-Rexach
an argumental DP (the doubled element), whereas French or Italian clitics lack this specification. (2) a. Juan le ha dado libros a Isabel b. *Jean lui a donné des livres a Isabelle ‘Jean has given books to Isabelle’ In Gutiérrez-Rexach (1999a, 2001a), in accordance with the proposed strong correspondence hypothesis, it is also proposed that clitic pronouns denote determiner functions inherently restricted to context sets. Westerståhl (1985) proposed that pronouns should be treated as functions from context sets to generalized quantifiers, where a generalized quantifier is a function from sets to truth values. For example, the denotation of the pronoun they is, for a universe of individuals E and all sets A, C ⊆ E, the function THEY ∈ [P(E) → [P(E) → 2]] such that THEY(C)(A) = 1 iff C ⊆ A & |C| ≥ 2. The set C is the context set of the pronominal function. In (3), C is the set of two old men who are walking in the park. (3) Two old men are walking in the park. They are bald. Clitics are, thus, pronominal determiners. As is well known, clitic pronouns are normally morphologically specified as accusative or dative forms. Following Keenan’s (1989) semantic case theory, we can take the case feature to indicate the type of the function with which clitics combine. Let α be a clitic expression, E a universe, and [A → B] the set of functions from A to B. Then, (i) if α is [+nom.], [[α]] is a function F ∈ [P(E) → [P(E) → 2]]; (ii) if α is [+acc.], [[α]] is a function F ∈ [P(E) → [P(E2 ) → P(E)]]; and finally (iii) if α is [+dat.], [[α]] is a function F ∈ [P(E) → [P(E3 ) → P(E2 )]]. The derivation of the truth conditions of (4) is as in (5), where R ⊆ E2 , C ⊆ E, a, b ∈ E, Ip is the individual generated by Pedro (Keenan 1996) and LO(C)(R) = {a|C ⊆ {b| < a, b >∈ R} & |C| = 1}. (4) Pedro lo compró Pedro it bought ‘Pedro bought it’ (5) Ip (IT)(C)(BOUGHT) = 1 iff Ip ( {a|C ⊆ {b| < a, b >∈ BOUGHT} & |C| = 1}) = 1 iff p ∈ {a|C ⊆ {b| < a, b >∈ BOUGHT} & |C| = 1} Similarly, the derivation of the truth conditions of (6), where a dative pronominal clitic occurs, is as in (7), where R ⊆ E3 , C ⊆ E, a, b, c ∈ E, and LES(C)(R) = {< a, b > | C ⊆ {c| < a, b, c >∈ R} & |C| ≥ 2}.
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.3 (337)
Constraint interaction at the semantics/pragmatics interface
(6) Juan les dio algunos libros Juan them gave some-pl books ‘Juan gave them some books’ (7) JUAN ((LES(C)(GAVE))(SOMEpl BOOK)acc) = 1 iff Ij ( {a | | BOOK ∩ {b | C ⊆ {c | < a, b, c >∈ GAVE} & |C| ≥ 2}} | ≥ 2})= 1 iff j ∈ {a | | BOOK ∩ {b | C ⊆ {c | < a, b, c >∈ GAVE} & |C| ≥ 2}} | ≥ 2}. The determination of the relevant context set in cases such as the ones illustrated above is constrained by standard restrictions on pronominal anaphora resolution, such as the following ones: salience, parallelism, linear precedence in a text, focal stress assignment, and common sense reasoning. For instance, the discourse in (8) makes the woman who was walking salient and this singleton set is the context set for the interpretation of the pronoun. On the other hand, in (9a) there are two possible “antecedents” which are equally salient. A common sense reasoning constraint is the one that determines the context set of la in (9a). In (9b), where there would be two context sets available, the dominant constraint to determine the content of the context set is focal stress assignment. (8) Una mujer estaba andando. Pedro la vio a woman was walking. Pedro her saw ‘A woman was walking. Pedro saw her.’ (9) a.
María dijo que Luisa estaba enferma, pero Pedro la vio esta mañana ‘María said that Luisa was sick, but Pedro saw her this morning’ b. Pedro dijo que María y LUISA estaban enfermas, pero Pedro la vio esta mañana ‘Pedro said that María and LUISA were sick, but Pedro saw her(Luisa) this morning’
Finally, a parallelism constraint on ellipsis resolution is responsible for the determination of the context set in (10), where the singleton set whose only element is Juan cannot be the context set of the accusative pronoun. Here I am neutral about the concrete mechanism that derives this constraint, but see Fox (1995), Dalrymple, Shieber and Pereira (1991), and Gutiérrez-Rexach (1999b) for a thorough discussion of the issues involved. (10) Juan odia a Pedro y Luis lo odia también ‘Juan hates Pedro and Luis hates him too’ The main claim of this paper is that pronominal doubling is associated with a series of constraints that operate at the semantics/pragmatics interface and that are different from the ones described above. What this means is that when a pronominal
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.4 (338)
Javier Gutiérrez-Rexach
clitic is doubled, it has to satisfy stricter semantic/pragmatic requirements than its non-doubled counterpart. In recent years, Optimality Theory (Prince & Smolensky 1993, 1997) has developed as a formal theory of constraint interaction and satisfaction in several domains of linguistic inquiry. Applications to syntax and, particularly to phonology and morphology, are well-known (see Kager 1999; Archangeli & Langendoen 1997; etc. for introductory overviews). Recently, an Optimality Theoretic approach to semantic/pragmatic constraints has been defended in work by Hendricks and de Hoop (2001), Zeevat (1999), Blütner (1999), etc. What I will argue is that an optimality-theoretic approach to constraint interaction in the semantics/pragmatics interface can successfully account for dialectal and cross-linguistic variation in clitic-doubling constructions.
Semantic and pragmatic constraints on clitic doubling Clitic doubling constructions differ from regular pronominal constructions in the way in which the content of the context set is determined. Without exception, in clitic doubling structures the context set is retrieved from the denotation of the doubled expression. Consider the correlate of (9b) in (11), where the only difference is that in the latter the pronoun doubles the accusative argument. No matter what the focus structure of (11) is, i.e. no matter whether María or Luisa receive focal stress, the context set for la has to be the singleton set with the individual denoted by María. (11) Pedro dijo que María y Luisa estaban enfermas, pero Pedro Pedro said that María and Luisa were sick, but Pedro la vio a María esta mañana her saw A María this morning ‘Pedro said that María and Luisa were sick, but Pedro saw her(María) this morning’ The relevant conclusion is that there has to be “semantic matching” between the context set required by the pronoun and one of the elements of the associated generalized quantifier. Formally: (12) Let αd/acc be an accusative doubling clitic expression and αacc its nondoubling counterpart. Then, for all C ⊆ E, R ⊆ E2 , and quantifiers Q (with accusative extension Qacc ): [[αd/acc ]](C)(R)(Qacc) = [[αacc ]](C)(R) & C ∈ Q
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.5 (339)
Constraint interaction at the semantics/pragmatics interface
This characterization does not impose any restriction on the nature of the associated quantifier, so in principle it predicts that an accusative clitic may double any expression with the matching case feature. Nevertheless, this is not correct. I claim that accusative clitic doubling in Spanish obeys four semantic/pragmatic constraints. The first one is the Principal Filter/Referentiality Constraint (REF). In the terms of generalized quantifier theory, we can say that the generalized quantifier associated with an accusative clitic has to be a principal filter. A generalized quantifier Q over E is a principal filter iff there is a non-empty set A ⊆ E, such that for all B ⊆ E, Q(B) = 1 iff A ⊆ B. The set A is called the generator of Q. When a DP denotes a principal filter it is referential. It refers to a particular group. The context set of a clitic pronoun is the group referred to by the associated DP. The contrast between the sentences (13) and (14) follows from the REF constraint. (13) Los engañaron a los/esos/mis/todos los estudiantes them fooled A the/those/my/all the students ‘They fooled those/my/all the students’ (14) *Los engañaron a una mayoría de them fooled A a majority of los/varios/pocos/aproximadamente tres estudiantes the/several/few/approximately three students ‘They fooled most/several/few/approxiamtely three students’ The REF constraint applies not only when the complement of a verb is a DP but also applies to propositional attitude verbs. In this case, the relevant criterion is the distinction between those verbs that select propositions as their complements, such as know, and those that select sets of propositions, such as question embedding verbs of the wonder type (Karttunen 1977). Typically, the complements of the verbs in the first class satisfy REF vacuously and can be accusatively doubled (15a), whereas those in the second class do not satisfy REF and are not doubled (15b). (15) a.
Lo sé que te sientes mal it know-I that you feel bad ‘I know that you feel bad’ b. *Se lo preguntó si te sentías mal refl it wonder if you felt bad ‘He wondered whether you felt bad’
The second constraint that has to be satisfied in Spanish accusative clitic doubling constructions is the Presuppositionality Constraint (PRES). The generator of the generalized quantifier associated with an accusative clitic is a presupposed set (See Capone 2000 for an analysis of presuppositionality and non-doubling clitics). The relevant presupposition is an existence presuposition, so the existence of the indi-
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.6 (340)
Javier Gutiérrez-Rexach
viduals in the set is presupposed. The constraints REF and PRES are not in conflict. Actually, as formulated, PRES represents a specialization of REF, since PRES requires that the group referred to by the associated DP has to be presupposed. The effects of this constraint are clearly shown in the following examples. Typically, focused elements cannot be clitic doubled. All of the “clitic . . . DP” associations in (16) would satisfy REF because Juan denotes a principal filter. Critically, none of them satisfies PRES because Juan is focused and does not belong to the presuppositional content of the sentence. Thus, prosodically focused doubled constituents (16a), or doubled constituents occurring as answers to a question (16b), in cleft and pseudocleft constructions (16c, d), or in the scope of the focus operator sólo ‘only’ (16e) violate PRES. (16) a. *Yo lo ví a [Foc Juan] I him saw A [Foc Juan] ‘I saw [Foc Juan]’ b. ¿A quién viste? (*Lo) vi a Juan A who saw-you him saw A Juan c. A quien (*lo) vi es a Juan A who him saw-I is A Juan ‘Who(ever) I saw is Juan’ d. Es Juan a quien (*lo) vi is Juan A who him saw-I ‘It is Juan who I saw’ e. (*Lo) vi sólo a Juan him saw-I only A Juan ‘I saw only Juan’ Other potential semantic/pragmatic constraints are first, a Salience/Emphasis Constraint (EMPH). The context set of the clitic (the group referred to by the doubled quantifier expression) is salient in a scale ordering available context sets. Again, this constraint does not conflict with the REF constraint. On the other hand, a conflict may arise with PRES, when the context sets available are focused and the associated order reflects the salience order among novel discourse elements. In this case, EMPH may override PRES, and the above sentences are perceived as felicitous. They are also usually associated with a specific intonational contour (“emphatic” intonation). Nevertheless, the default scenario is that PRES and EMPH do not conflict and the element emphasized by accusative clitic doubling is a discourse topic. In many Spanish dialects there is an additional constraint that we call the Animacy Constraint (ANIM). ANIM requires that the doubled quantifier live on the set ANIMATE/HUMAN. The “live on” property is understood in the terms of Bar-
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.7 (341)
Constraint interaction at the semantics/pragmatics interface
wise and Cooper (1981), namely the intersection of the denotation of the head noun with the set of animate/human entities is non-empty. What this amounts to is that only expressions that denote animate/human entities may be doubled. In most Spanish dialects, ANIM conflicts with REF. Even if an expression denotes a principal filter, it cannot be doubled by an accusative clitic if the generator of that principal filter is not a set of animate individuals. The ungrammaticality of (17) shows, then, that ANIM is ranked higher than REF (ANIM REF), and that a violation of ANIM becomes fatal. (17) *Las limpié las mesas them cleaned-I the tables ‘I cleaned the tables’
Existential quantifiers and interactions with modality As has been observed in the literature (Suñer 1988, 1991; Sportiche 1996; Uriagereka 1995; etc.), specific objects tend to favor accusative doubling whereas non-specific ones block it. Thus, the examples in (18) are only grammatical/felicitous if the object is understood as “specific.” (18) a.
Las he visto a tres monjas/tres de las monjas them-fem have-I seen A three nuns/three of the nuns ‘I have seen three (specific) nuns’ b. Los enojó a algunos profesores them-pl annoyed A some-pl teachers ‘He annoyed some teachers’
What I want to defend here is that the “specificity effect” on doubled objects reduces to the combination of the REF and PRES constraints. REF requires that a doubled object denote a group and PRES requires that this group be presupposed. When applied to existential quantifiers, the combined satisfaction of both constraints is only compatible with “specific” readings. Furthermore, the specificity effect is in reality a referentiality effect. Something stronger than mere specificity, which equals presuppositionality in Diesing’s (1992) terms, is required. Scope marking is a property which is sensitive to the combination of the strong specificity attributes of a DP. In Spanish, there are two scope marking possibilities. In the neutral default case, word order determines the scopal order of constituents. Thus, (19a) is only compatible with the SUBJECT > OBJECT scopal order. Inverse scope readings, in which the indefinite object scopes over the subject, require either overt focusing of the object, as in (19b), or clitic doubling, as in (19c).
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.8 (342)
Javier Gutiérrez-Rexach
(19) a.
Tres three b. Tres three c. Tres three
estudiantes students estudiantes students estudiantes students
leyeron dos libros read two books leyeron [Foc dos libros] read two books los leyeron dos libros them read two books
The object wide scope reading of (19c) is a byproduct of the satisfaction of REF and PRES. A weaker notion of specificity based only on the satisfaction of PRES would not suffice. For example, it has been claimed that overt partitives are specific (De Hoop 1992; Enç 1991). The object DP in sentence (20a) is weakly specific. The speaker is asking the addressee to bring a group of several books from a presupposed set. In this respect, more than one group could satisfy the speaker’s request, as long as they belong to the presupposed set of books. Sentence (20b) requires a stronger notion of specificity, namely the combination of referentiality and presuppositionality and, as a consequence, the speaker is requesting a unique group of books. (20) a.
Quiero que me traigas varios de los libros want-I that me bring-you several of the books ‘I want you to bring me several of the books’ b. Quiero que me los traigas varios de los libros want-I that me them bring-you several of the books ‘I want you to bring me (a group of) several of the books’
Accusative doubling has also a “scope freezing” effect, so sentence (21) can only be construed in the object wide scope interpretation. (21) Tres profesores los engañaron a muchos de los estudiantes three professors them fooled A many of the students ‘Three professors fooled a group consisting of many of the students’ As observed by Montague (1969), the objects of intensional or opaque verbs are ambiguous. They can have a non-transparent or a transparent reading, the latter arising when the object scopes over the verb. (22) a. I am looking for a unicorn b. I want two secretaries In Spanish, the transparent reading requires the presence of the accusative doubling clitic. For example, (23) lacks the opaque reading. Similarly accusative doubling of a DP modified by a subjunctive relative clause is not possible. This follows from the fact that this type of subjunctive modification is intrinsically non-transparent (Rivero 1977).
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.9 (343)
Constraint interaction at the semantics/pragmatics interface
(23) Lo busco a un unicornio him look-for-I A a unicorn ‘I am looking for a unicorn’ (24) a.
Las busco a dos secretarias que saben them-fem look-for-I A two secretaries that know-ind-they Francés French ‘I am looking for two secretaries who speak (indicative) French’ b. *Las busco a dos secretarias que sepan Francés them look-for-I A two secretaries that know-subj-they French ‘I am looking for two secretaries who speak (subjunctive) French’
Variation and constraint satisfaction Dative doubling In the previous sections we have studied several constraints on Spanish accusative doubling. Dative clitic doubling is not sensitive to the constraints REF, and PRES, so there is not a contrast between the examples (25a, b). None of them violate the relevant semantic constraints. (25) a.
Les dieron un libro a los/esos/mis/todos los to-them gave a book to the/those/my/all the ‘They gave a book to the/those/my/all the students’ b. Les dieron regalos a muchos/varios/pocos to-them gave-they presents to many/several/few ‘They gave presents to many/several/few students’
estudiantes students estudiantes students
The animacy and the emphasis constraint do not have to be satisfied by dative doubling constructions either. Thus, dative clitic doubling can be considered a mark of syntactic agreement (case matching), wheras accusative clitic doubling is a mark of true semantic agreement. This property is related to two facts. First, since accusative clitics semantically select certain features, they are a marked and more stringent choice. Dative clitic doubling has become obligatory as the unmarked option in many dialects: sentences in which dative indirect objects are clitic doubled can be considered as the neutral variants. The only exception to this generalization is the use of le for certain types of semantic marking, such as “intensive” le in Mexican Spanish (Torres 2000), illustrated in (26). In this case, dative doubling satisfies REF, PRES, EMPH.
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.10 (344)
Javier Gutiérrez-Rexach
(26) Ándele, pues walk-cl-dat thus
Accusative doubling in Romanian and Greek The constraints that have been shown to be active in Spanish also have to be considered in the study of clitic doubling in Romanian and Greek. In Romanian, accusative doubling is obligatory when the object is preceded by the preposition pe. It always triggers also semantic agreement, in other words, it is not semantically inert. The constraints REF and PRES are satisfied. (27) L/I-am v˘azut pe Popescu/to¸ti/to¸ti copii him/them-have(I) seen PE Popescu/all/all children-the ‘I have seen Popescu/all/all the children’ Generalized existential quantifiers satisfying the intersectivity condition (Keenan 1996) may be doubled only when they are referential, that is to say, when they denote principal filters and satisfy REF. In the following examples from DobrovieSorin (1994), the objects are understood as referential. (28) a.
I-am dus pe dou˘a fran¸tuozaice la gar˘a them-have-I taken PE two Frenchwomen to-the station ‘I have taken two Frenchwomen to the station’ b. I-am p˘ac˘alit pe mul¸ti copii them-have-I fooled PE many children ‘I have fooled many children’
A final parallelism with the Spanish case is the doubling of indefinites in their strongly specific or principal filter denoting reading, as in (29), and their incompatibility with the subjunctive mood in a modifying relative clause (30). (29) O caut pe fat˘a de la noi din sat her am-I looking-for a girl from our village ‘I am looking for a girl from our village’ (30) a.
Caut un elev care s˘a-¸stie engleze¸ste look-for-I a student which speak-subj English ‘I am looking for a student who would be able to speak English’ b. *Il caut pe un elev care s˘a-¸stie engleze¸ste him look-for-I PE a student which speak-subj English
An important contrast between Spanish and Romanian surfaces in dative clitic doubling. Recall that we have shown above that Spanish dative clitic doubling dif-
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.11 (345)
Constraint interaction at the semantics/pragmatics interface
fers from accusative doubling in not being sensitive to any concrete semantic constraint. In Romanian, dative doubling is obligatory if the associated expression is referential and presuppositional. Thus, it also has to satisfy REF and PRES. The dative object in (31) lacks the weakly specific (partitive) and non-specific readings. (31) le caut˘a leac unor bloi incurabile cl-dat.pl search-he remedy some illness incurable ‘He searches remedies for incurable diseases’ In Greek, only expressions that are always referential may be doubled by an accusative clitic. Consequently, a contrast arises between Greek and Spanish existential quantifiers. In Spanish, these expressions may be doubled but in Greek they may not. According to Gutiérrez-Rexach (1999a), this shows that the referentiality restriction on accusative doubling is actually stronger in Greek and amounts to a co-intersectivity condition. Only co-intersective generalized quantifiers (universals, definites, proper names, etc. See Keenan 1996) are doubled in Greek. Thus, we may split the referentiality/principal filter constraint in two subconstraints: REFCOINT and REFINT/EXIST . In Greek, *REFEXIST is operative, and this explains the ungrammaticality of the doubled constructions in (32). (32) a. *Tin psaxno mia/kapja grammatea her look-for-I one/a secretary ‘I look for one/a secretary’ b. *Tous eksetase merikous apo tous asthenis them examined-he several of the patients ‘He examined several of the patients’ Anagnostopoulou and Giannakidou (1995) propose that clitic doubled objects have to be prominent, in Heim’s (1982) sense. In (33a), the non-doubled ton sigrafea is ambiguous. It may either refer to Arthur Miller (k = i) or to the (accommodated) author of the book that John read (k = j). The second option is not available under clitic doubling. Ton sigrafea in (33b) can only be understood as referring to Arthur Miller (k = i). (33) a.
O Jannis diavase [j ena vivlio jia ton [i Arthur Miller]], enthusiastike, ke thelise na gnorisi [k ton sigrafea] apo konta John read [j a book about [i Arthur Miller]], he got very enthusiastic, and he wanted to get to know [k the author] b. O Jannis diavase [j ena vivlio jia ton [i Arthur Miller]], enthusiastike, ke thelise na ton gnorisi [i ton sigrafea] apo konta John read [j a book about [i Arthur Miller]], he got very enthusiastic, and he wanted to get to cl know [k the author]
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.12 (346)
Javier Gutiérrez-Rexach
In sum, clitic doubling in Greek requires that the entity referred to by the noun phrase have been explicitly introduced in the previous discourse. We can call this constraint PROM. Since what is asserted cannot be presupposed, this condition is in conflict with the Presuppositionality condition that we formulated before. In Greek, PROM is ranked higher than PRES (PROM PRES) but in Spanish, PRES is ranked higher (PRES PROM), as was discussed above. Thus, in (34) the anaphoric linking in which j = k is the preferred one. The author that Juan wants to get to know is the author of the book, not Arthur Miller. (34) Juan leyó [j un libro sobre [i Arthur Miller]], se entusiasmó, y quiso conocerlo [k al autor] John read [j a book about [i Arthur Miller]], he got very enthusiastic, and he wanted to get to cl know [k the author] One consequence of the above contrast is that in Greek doubling of accommodative definites (35a), weak definites (Poesio 1994) (35b), novel proper names/definites (those mentioned for the first time) (35c) is not possible. All these expressions relate to presupposed elements which are not prominent. efaga tis sokolates (35) a. *I Maria mou estile glika; tis the Mary to-me sent sweets; them ate-I the chocolates amesos immediately ‘Mary sent me sweets; I immediately ate the chocolates’ b. *Xtes ton gnorisa ton fititi enos diasimou glossologou yesterday him met-I the student of-a famous linguist ‘ I met the student of a famous linguist yesterday’ c. *Xtes onirevtika oti ton gnorisa ton Clinton s’ena party yesterday dreamt-I that him met the Clinton in-a party ‘Last night I dreamt that I met Clinton at a party’ In Spanish, where (PRES PROM), doubling of all these types of definites is possible. They all require that the doubled element refer to a presupposed entity. (36) a.
María me envió dulces; me los comí los chocolates Mary to-me sent sweets; cl-dat them ate-I the chocolates inmediatamente immediately ‘Mary sent me sweets; I immediately ate the chocolates’ b. Ayer lo conocí al estudiante de un famoso linguista yesterday him met-I A-the student of a famous linguist ‘I met the student of a famous linguist yesterday’
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.13 (347)
Constraint interaction at the semantics/pragmatics interface
c.
Fíjate, ayer soñé que lo conocí a Clinton en look, yesterday dreamt-I that him met A Clinton in una fiesta a party ‘Look, yesterday I dreamt that I met Clinton at a party’
Economy of context There is an additional constraint on dative doubling which accounts for some puzzling patterns in Spanish doubling. We can formulate this Economy of Context (ECON) constraint as follows: Let Qi , Qj be two quantifiers satisfying REF such that they would yield an identical interpretation in a clitic doubling construction. Thus, Qi , Qj have to be principal filters and denote the same function. Then, if one is an inherently contextually restricted function and the other is not, only the contextually restricted quantifier can be doubled. A contextually restricted function is one that is specified as being restricted to a context set. This property is normally highlighted by morphological marking. In Spanish, the inherently restricted determiners are typically complex determiners formed by a basic determiner function and the determiner the. The existence of the constraint ECON explains why several Spanish quantifiers, apparently denoting the same function, nevertheless exhibit differential patterns with respect to doubling. The contrasting complex and simple determiners are todos los ‘all the’ and todo ‘every’; cada uno de los ‘each one of the’ vs. cada ‘each’; and algunos ‘some pl’ vs. unos ‘a-pl.’ Let us consider for example the universal determiners todos los ‘all the’ and todo ‘every’. We have claimed that the first one is inherently contextually restricted whereas the second one is not. What this means is that no matter what the nature of the surrounding discourse is, todos los is understood to be a contextually or discourse linked expression whereas todo can only be contextually dependent when it is coerced (Pustejovsky 1995) under very strict conditions. For instance, both sentences in (37) express a universal statement about men. Sentence (37a) can also be used when the speaker is referring to a particular group of men whereas this is not possible in the case of (37b). The latter sentence is understood as a generic statement about males. (37) a.
Todos los hombres aman a una mujer all-masc.pl the men love A a woman ‘All the men love a woman’ b. Todo hombre ama a una mujer every man loves A a woman ‘Every man loves a woman’
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.14 (348)
Javier Gutiérrez-Rexach
The discourses in (38) show the contrast between these determiners even more clearly. Recall that todos los denotes a function inherently restricted to a context set. In (38a), the set of twelve students who came becomes the context set of the determiner and the discourse is felicitous. On the other hand, todo in (38b) lacks a context set argument. The discourse becomes infelicitous because the context provided by the first sentence is not compatible with a universal/generic statement about Europeans. (38) a.
Vinieron doce estudiantes. Todos los europeos came-they twelve students. All the Europeans sonreían were-smiling ‘Twelve students came in. All the European ones were smiling’ b. Vinieron doce muchachos. *Todo europeo sonreía came-they twelve students. Every European were-smiling ‘Twelve students came in. Every European was smiling’
The other contrasting couples of simple and complex determiners considered above exhibit the same behavior with respect to these tests. According to the constraint ECON, only the quantifiers headed by the complex determiners may be doubled. This prediction is borne out, as the following constrasts show: (39) a.
Los vi a todos los hombres them saw-I A all-masc.pl the men ‘I saw all the men’ b. *Lo vi a todo hombre him saw-I A every man
(40) a.
Lo vi a cada uno de los estudiantes him saw-I A each one of the students ‘I saw each one of the students’ b. *Lo vi a cada estudiante him saw-I A each student
Similarly, algunos ‘some pl’ is a contextually restricted indefinite determiner, but unos ‘a-pl’ is not (Gutiérrez-Rexach 2001b). As expected, only DPs headed by algunos can undergo accusative doubling: (41) a.
Los vi a algunos hombres them saw-I A some-masc.pl men ‘I saw some men’ b. *Los vi a unos hombres them saw-I A a-pl men
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.15 (349)
Constraint interaction at the semantics/pragmatics interface
The constraint ECON is ranked lower than REF. In all the above examples, the doubled quantifiers are principal filters. If ECON were ranked higher than REF, it would be possible to double contextually restricted expressions even if they do not denote principal filters. In Porteño Spanish (Suñer 1988), *REFEXIST is active and doubling of quantifiers headed by intersective or generalized existential determiners is not possible (42a). Nevertheless, ECON is ranked higher than *REFEXIST (ECON *REFEXIST ), so a violation of *REFEXIST is possible in (42b) in order to prevent a violation of the higher ranked ECON. (42) a. *Los vi a tres them saw-I A three ‘I saw three men’ b. Los vi a tres them saw-I A three
hombres men de los hombres of the men
The activity of the ECON constraint can also be detected in the domain of negative quantifiers. The determiners ninguno de los ‘none of the’ and ningún ‘no’ and the quantifiers ninguno ‘nobody’ vs. nadie ‘nobody’ contrast again in that the former denotes a contextually restricted function. By applying the same reasoning as before, we explain the following contrasts as the result of the satisfaction or violation of ECON. (43) *No lo he visto a ningún político not him have-I seen A no politician ‘I have not seen any politician’ (44) No lo he visto a ninguno de los políticos not him have-I seen A none of the politician ‘I have not seen any of the politicians’ (45) a.
No lo he visto a not him have-I seen A ‘I have not seen anybody’ b. ??No lo he visto a not him have-I seen A ‘I have not seen anybody’
ninguno nobody nadie noone
Clitic doubling of interrogative quantifiers In this section, I will consider accusative doubling in Spanish and Romanian questions. I will treat wh-words in general as interrogative generalized quantifiers or determiners (Gutiérrez-Rexach 1997). Consequently, clitic doubling in questions
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.16 (350)
Javier Gutiérrez-Rexach
has to be treated as a case of pronoun-quantifier interaction in parallel with doubling in declarative sentences. Evidence for this parallelism comes from the fact that Spanish dative clitic doubling does not impose any semantic restriction on the associated wh-word: (46) a.
¿A quién le diste un libro? to who to-him gave-you a book ‘To whom did you give a book?’ b. ¿A cuál le molestan estos ruidos? A whicho to-him bothers these noises ‘Which one is bothered by these noises?’
Accusative doubling shows sensitivity to the nature of the associated interrogative quantifier. Spanish wh-words quién ‘who’ and qué ‘what’ contrast with cuál ‘which’ in that the latter denotes a contextually restricted determiner or generalized quantifier function (Gutiérrez-Rexach 1997). In (47), the variant with cuál expresses a query about a presupposed set of individuals (the context set); the variant with quién is not subject to this restriction. (47) ¿Quién/Cuál vino a la fiesta? who/which-one came to the party? Only the interrogative expression inherently restricted to a context set can be doubled by an accusative clitic, as (48) shows: (48) a. *¿A quién lo viste? A who him saw-you ‘Who did you see?’ b. *¿Qué lo compraste? what it bought-you ‘What did you buy?’ c. ¿A cuál de ellos lo viste? A which of them him saw-you ‘Which one of them did you see?’ d. ¿A cuál lo han matado? A which him have-they killed ‘Which one did they kill?’ This contrast constitutes strong evidence of the activity of ECON in the interrogative domain. The functions denoted by qué and quién only differ from cuál in that they are not restricted to a context set. Thus, if they are associated with an accusative clitic, a violation of ECON arises. The constraint PRES is also satisfied vacuously when ECON is satisfied because the context set represents pre-
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.17 (351)
Constraint interaction at the semantics/pragmatics interface
supposed information. The nature of the information exchange that is typical in question-answer dialogues renders the constraint REF inactive. Interrogative generalized quantifiers cannot be principal filters because in information questions the speaker is not referring to a particular individual or group of individuals. Rather, he is asking to identify one. There is dialectal variation with respect to how strong ECON is. For instance, in Porteño Spanish, ECON requires overt partitivity in all doubled quantifiers: (49) a.
¿A cuál de ellos viste? A which of them saw-you ‘Which one of them did you see?’ b. ??¿A cuál han matado? A which have-they killed ‘Which one did they kill?’
Free choice wh-elements also satisfy ECON. The free choice item cualquiera ‘whichever’ may be doubled by an accusative clitic whereas quienquiera ‘whoever’ may not. (50) a.
En este departamento, lo admiten a in this department, him admit A ‘In this department, they admit anyone’ b. *En este departamento, lo admiten a in this department, him admit A ‘In this department, they admit anyone’
cualquiera whichever quienquiera who-pl-ever
Accusative doubling in Romanian interrogatives also satisfies ECON. The interrogative quantifier expressions cine ‘who’ and ce ‘what’ do not denote contextually restricted functions, whereas care ‘which’ does. As was the case in Spanish, only the latter quantifier may be doubled, as shown in (52). (51) a. *Pe cine l-ai v˘azut? PE who him-have(you) seen ‘Who did you see?’ b. *Ce (roman) l-ai citit? what (novel) it-have(you) read ‘What (novel) did you read?’ (52) a.
Pe care l-ai v˘azut? PE which him-have(you) seen ‘Which did you see?’ b. Pe care b˘aiat l-ai v˘azut? PE which boy him-have(you) seen ‘Which boy did you see?’
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.18 (352)
Javier Gutiérrez-Rexach
Conclusions In this paper, I have argued that the problem of the association between a clitic and a generalized quantifier can be succesfully accounted for in an Optimality Theoretic framework. The essence of my proposal is that the nature of the constraints at play is semantic, and that the predicted restrictions take place in the semantics/pragmatics interface. Taking the arguments and data considered in GutiérrezRexach (1999a) as a departure point, I have formulated five main constraints on accusative doubling (REF, PRES, EMPH, ANIM, and ECON) and have shown that this type of marking can be considered truly semantic. On the other hand, dative clitic doubling is not sensitive to these constraints and cannot be considered a proper instance of semantic marking. The account extends successfully to clitic doubling in interrogative environments, where ECON is claimed to play a more significant role. Finally, I have discussed clitic doubling in Greek and Romanian and argued that cross-linguistic variation follows from different patterns of activity in the set of constraints discussed or by differences in constraint ordering. This confirms some of the main tenets of Optimality Theory and shows that the explanation of linguistic variation based on constraint interaction and satisfaction postulated by this approach is relevant for the analysis of semantic/pragmatic phenomena.
References Anagnostopoulou, E., & Giannakidou, A. (1995). Clitics and prominence, or why specificity is not enough. Chicago Linguistic Society 31: Parasession on Clitics. Archangeli, D., & Langendoen, T. (1997). Optimality Theory. An Overview. Oxford: Basil Blackwell. Barwise, J., & Cooper, R. (1981). Generalized quantifiers and natural language. Linguistics and Philosophy, 4, 159–219. Blütner, R. (1999). Some aspects of optimality in natural language interpretation. Ms. Berlin University. Capone, A. (2000). Dilemmas and Excogitations. An essay on modality, clitics and discourse. Messina: Armando Siciliano. Dalrymple, M., Shieber, S., & Pereira, F. (1991). Ellipsis and higher-order unification. Linguistics and Philosophy, 14, 399–452. De Hoop, H. (1992). Case Configuration and Noun Interpretation. Ph.D. dissertation, University of Groningen. Diesing, M. (1992). Indefinites. Cambridge, MA: MIT Press. DobrovieSorin, C. (1994). The Syntax of Romanian. Berlin/New York: Walter De Gruyter. Enç, M. (1991). The semantics of specificity. Linguistic Inquiry, 22, 1–26. Fox, D. (1995). Economy and scope. Natural Language Semantics, 3, 283–341.
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.19 (353)
Constraint interaction at the semantics/pragmatics interface
Gutiérrez-Rexach, J. (1997). Questions and generalized quantifiers. In A. Szabolcsi (Ed.), Ways of Scope Taking (pp. 409–452). Dordrecht: Kluwer. Gutiérrez-Rexach, J. (1999a). The formal semantics of clitic doubling. Journal of Semantics, 16, 315–380. Gutiérrez-Rexach, J. (1999b). Scope parallelism and the interpretation of ellipsis at the syntax/semantics interface. Syntaxis, 2, 51–79. Gutiérrez-Rexach, J. (2001a). Interface conditions and the semantics of argument clitics. In J. Gutiérrez-Rexach & L. Silva-Villar (Eds.), Current Issues in Spanish Syntax and Semantics (pp. 107–142). Berlin/New York: Mouton-De Gruyter. Gutiérrez-Rexach, J. (2001b). The semantics of Spanish plural existential determiners and the dynamics of judgment types. Probus, 13, 113–154. Heim, I. (1982). The Semantics of Definite and Indefinite Noun Phrases. Ph.D. dissertation, University of Massachusetts, Amherst. Hendricks, P., & de Hoop, H. (2001). Optimality theoretic semantics. Linguistics and Philosophy, 24, 1–32. Kager, R. (1999). Optimality Theory. Cambridge: Cambridge University Press. Karttunen, L. (1977). The syntax and semantics of questions. Linguistics and Philosophy, 1, 3–44. Keenan, E. (1989). Semantic case theory. In R. Bartch et al. (Eds.), Semantics and Contextual Expression (pp. 33–57). Dordrecht: Foris. Keenan, E. (1996). The semantics of determiners. In S. Lappin (Ed.), Handbook of contemporary semantic theory (pp. 41–64). Oxford: Basil Blackwell. Montague, R. (1969). On the nature of certain philosophical entities. The Monist, 53, 159–194. Poesio, M. (1994). Weak definites. In Proceedings of Semantics and Linguistics Theory, IV (pp. 282–299). Ithaca, NY: CLC Publications. Postal, P. (1966). On so-called “pronouns” in English. In D. Reibel & S. Schane (Eds.), Modern studies in English (pp. 201–224). Englewood Cliffs, NJ: Prentice Hall. Prince, A., & Smolensky, P. (1993). Optimality Theory. (MIT Press, forthcoming). Prince, A., & Smolensky, P. (1997). Optimality: From neural networks to universal grammar. Science, 275, 1604–1610. Pustejovsky, J. (1995). The Generative Lexicon. Cambridge, MA: MIT Press. Rivero, M. (1977). Estudios de Gramática Generativa del Español. Madrid: Cátedra. Sportiche, D. (1996). Clitic constructions. In J. Rooryck & L. Zaring (Eds.), Phrase Structure and the Lexicon (pp. 213–276). Dordrecht: Kluwer. Suñer, M. (1988). The role of agreement in clitic-doubled constructions. Natural Language and Linguistic Theory, 6, 391–434. Suñer, M. (1991). Two properties of clitics in clitic-doubled constructions. In J. Huang & R. May (Eds.), Logical Structure and Linguistic Structure: Crosslinguistic Perspectives (pp. 233–251). Dordrecht: Kluwer. Torrego, E. (1995). On the nature of clitic doubling. In H. Campos & P. Kempchimsky (Eds.), Evolution and Revolution in Linguistic Theory. Washington, DC: Georgetown University Press. Torres, R. (2000). Intensive “le” usage in Mexican Spanish. Talk at LSRL 30.
TSL[v.20020404] Prn:29/08/2002; 16:38
F: PBNS18.tex / p.20 (354)
Javier Gutiérrez-Rexach
Uriagereka, J. (1995). Aspects of the syntax of clitic placement in Western Romance. Linguistic Inquiry, 26, 79–124. Westerståhl, D. (1985). Determiners and context sets. In J. van Benthem & A. ter Meulen (Eds.), Generalized Quantifiers in Natural Language (pp. 45–71). Dordrecht: Foris. Zeevat, H. (1999). Explaining presupposition triggers. Proceedings of the Twelfth Amsterdam Colloquium, 19–24.
TSL[v.20020404] Prn:4/09/2002; 13:56
F: PBNS19.tex / p.1 (355)
Cross-language commutation tests and their application to an error-prone contrastive problem Ger. einige, Fr. quelques, Sp. algunos Eva Lavric Institut für Romanische Sprachen, Wirtschaftsuniversität Wien
As a linguist, and particularly as a contrastive linguist, I would like to start this contribution with three cheers to our students: Thank God we have students; thank God the students ask questions and thank God the students make errors! Because the errors and the questions about what would have been right and why, not only provide the contrastive linguistics branch of research with a first-class right to exist but also with a continuous stream of ideas of what to examine more minutely from what point of view. The investigation at hand thus starts with errors, errors of German-speaking students producing texts in business French, errors in particular areas of nominal determination and especially in the field of indefinite plural determiners. Those errors serve as the starting point of an investigation into the three languages German, French and Spanish, with a short look at the English language.1 The question how to carry out such a contrastive semantic investigation, i.e. what methods to apply to accomplish a satisfactory description of a particular area allowing for the deduction of rules suitable for teaching – that is a question which exceeds mere error analysis and which calls for the whole range of linguisticsemantic methodology. Personally, I see myself as part of a structuralist tradition as e.g. represented by Coseriu,2 Vater,3 Schifko.4 In addition to standard linguistic methods, there are of course special contrastive methods, in particular the translation comparison, which yields valuable results but which, however, did not seem sufficient for my question concerning determiner semantics. I would like to exemplify why this is the case. And my first example just happens to be an error example, and a three-fold one at that:5
TSL[v.20020404] Prn:4/09/2002; 13:56
F: PBNS19.tex / p.2 (356)
Eva Lavric
(1) Le premier impact [d’un accident] est celui du point de vue de *ces institutions qui doivent financer *ces frais qui sont directement liés aux accidents. *Ces frais qui se trouvent à gauche sont les frais directement liés à l’accident This is of course about determiners in noun phrases with restrictive relative clauses. Both the explanation and the respective rule are simple; what I would like to point out in this example (which is located halfway between syntax and semantics) is the way how to get there. This way is via another example, and it is that example by means of which I want to exemplify my semantic method: (2) J’étais entré dans le bonheur, qui est de faire le métier qu’on aime Ich genoß nun das Glück, das darin besteht, daß man jene Arbeit macht, die man gerne tut Comencé a disfrutar de la felicidad que supone dedicarse a la profesión que a uno le gusta (HPP 180–183) The example is taken from the corpus of my postdoctoral thesis,6 and one can see that this corpus is trilingual: It consists of texts with translations (half of them literary and the other half everyday texts). In this case, the original text is in French – and therefore comes first – the translations are into German and Spanish. (The other two directions of translation are of course also represented in the corpus.) The applied method is first and foremost a translation comparison.7 Example (2) is relevant for our first error example in so far as it includes a noun phrase with a restrictive relative clause and in its present form also provides the explanation of the error: In German, a restrictive relative clause can be signaled by a demonstrative determiner, namely by jener; the student has by analogy tried to do the same in French. One can already provide the learner with a rule of thumb here: In French, restrictive relative clauses cannot be signaled by a demonstrative determiner, one has to use the definite article. Linguistically, however, this rule is still unsatisfactory because it does not completely and explicitly represent the situation in both languages; moreover, Spanish is still missing to render the comparison a triple one. How is a restrictive relative clause signaled in Spanish? Example (2) shows: by the definite article, as in French. There is a parallel between the two Romance languages, which are in opposition to German, at least in example (2). Unfortunately however, this result is totally incorrect: mere translation comparison has led us astray in this case. With a mere translation comparison, one is far too much subject to the coincidences of translation, and that is why I system-
TSL[v.20020404] Prn:4/09/2002; 13:56
F: PBNS19.tex / p.3 (357)
Cross-language commutation tests
atically supplement the translation comparisons by commutation tests in all three languages. This means that in the same slot, I test all other determiners of the relevant language and note those which lead to the same meaning as the original text. I adapted this method from Vater (1963/1979), adding the contrastive dimension myself.8 The commutation tests restore the whole range of paradigmatic possibilities in all languages involved; under this method, one not only compares what is there in the text, but also what could have been there in this slot of the text while maintaining its meaning. The result for our example (2): (2 ) J’étais entré dans le bonheur, qui est de faire le métier qu’on aime Ich genoß nun das Glück, das darin besteht, daß man jene Arbeit macht, die man gerne tut daß man diejenige Arbeit macht, die man gerne tut daß man die Arbeit macht, die man gerne tut Comencé a disfrutar de la felicidad que supone dedicarse a la profesión que a uno le gusta a aquella profesión que a uno le gusta There are three forms yielding the same meaning in the same slot in German, two in Spanish and only one in French. And thus we can with the help of this characteristic example lay down all possibilities to signal a restrictive relative clause in the systems of the three languages: 1. In all three languages, there is the option of the definite article. 2. In addition, there are possibilities with demonstrative determiners in German and Spanish: Sp. aquel, Ger. jener and derjenige.9 3. In French, there is no possibility to signal a restrictive relative clause by a demonstrative determiner.10 This constraint is probably related to the fact that French does not really have a distal demonstrative,11 while both Ger. jener and Sp. aquel are distal demonstratives.12 So this would be the precise result of an empirically-based interlingual systems comparison, which the combination of translation comparison and commutation tests allows for. Although the rule of thumb for the student remains the same, we can now predict that e.g. German-speaking learners of Spanish will not make any interference errors with such examples, but that their transfer attempts will be successful.13 This brings us to the main part of this paper, the plural indefinites used for small numbers, i.e. Fr. quelques, Sp. algunos and Ger. einige. Again, we start with a few typical error examples:
TSL[v.20020404] Prn:4/09/2002; 13:56
F: PBNS19.tex / p.4 (358)
Eva Lavric
(3) Dans *quelques classe(s) d’école 70% des élèves ont besoin de leçons particulières (4) Son électorat est très hétérogène, mais *quelques groupes sont particulièrement attirés (5) *Quelques villes ont un certain pouvoir d’attraction, c’est la raison pour laquelle beaucoup d’entreprises s’y installent volontairement What these examples have in common is that the learners use the Fr. quelques in contexts in which it is not ‘appropriate’. By way of correcting, one could insert Fr. certains in all the examples; however, what the students actually ‘meant’ was not its German equivalent manche, but the German form einige, which was rashly equated with quelques. It is furthermore noteworthy that similar errors do not occur with Sp. algunos. We will therefore have to look at the French and Spanish equivalents of Ger. einige. Our ‘pre-investigation’ by means of error analysis suggests that Sp. algunos and Ger. einige could be synonyms, whereas Fr. quelques shows a narrower range of possible applications. The relation to certains furthermore suggests that the opposition [+/–specificity] plays a certain role regarding these differences. (What I mean by [+/–specificity] is the opposition between e.g. Engl. certain and any, i.e. the opposition constituted by random selection versus a selection of the referent which is governed by (knowledge of) its identity).14 In the linguistic literature, Fr. quelques is often compared with plusieurs (e.g. Gondret 1976; K˛esik 1978; Gaatone 1991). This, however, is of no help with our error examples, since it is not the interlingual problem concerning German. Here, errors occur because of the learners’ hypothesis that Fr. quelques is the equivalent of Ger. einige. This hypothesis is not totally unfounded, however. It is definitely valid in a number of cases (and the Spanish versions show that algunos is indeed the Spanish equivalent): (6) Nach hinten wird das Gelände von einigen Eichen begrenzt Vers le fond, quelques chênes bornent l’espace visible Hacia detrás, el terreno está delimitado por algunos robles (Wal 22/25/20) (7) Quelques femmes célèbres sont censées symboliser au Parlement et au gouvernement l’égalité des droits entre les sexes Einige berühmte Frauen gelten in Parlament und Regierung als symbolhafte Verkörperung der Gleichberechtigung von Mann und Frau Algunas mujeres célebres simbolizan en el Parlamento y en el gobierno la igualdad de derechos entre los sexos (Hen 9–16)
TSL[v.20020404] Prn:4/09/2002; 13:56
F: PBNS19.tex / p.5 (359)
Cross-language commutation tests
The forms quelques, algunos and einige have in common the semantic feature [small number],15 which means that the number of referents is lower than expected, lower than an objective or at least intersubjective standard. Since we are moreover interested in the aspect of [+/–specificity], I have chosen examples 6 and 7 to illustrate both poles of the opposition: (6) is non-specific and (7) is specific. Surprisingly, however, Fr. quelques can be found in both examples, which means it is not marked for the opposition [+/–specific]. The semantic restrictions of Fr. quelques compared with Ger. einige can thus not be located in this dimension. Therefore, we still have to look for authentic examples in which Ger. einige can occur, while Fr. quelques cannot. In my corpus, few but very characteristic examples can be found, the most meaningful of which is the following: (8) Ce truculent article ravira bien des présidents d’organisations agricoles! *quelques présidents Dieser gesalzene Artikel wird einige Präsidenten landwirtschaftlicher Organisationen erfreuen . . . ! ¡Este artículo encantará a algunos presidentes de organizaciones agrícolas . . . ! (Pre) A particular semantic feature can be tested against this example, namely the feature [considerable number]. While the opposition [small number] versus [large number] represents the objective aspect of the ‘number’ dimension, the opposition [considerable number] versus [inconsiderable number] denotes a subjective evaluation of the number on part of the speaker.16 The two aspects can be well differentiated in the semantics of the forms Fr. plus d’un / Sp. más de un, since both aspects are in a state of conflict here: Plus d’un /más de un denote an objectively small but subjectively considerable number. Fr. bien des signifies a large and at the same time considerable number.17 Example (8) shows that Fr. quelques cannot be used to refer to a considerable number and will therefore be substituted with bien des; quelques thus shows the feature [inconsiderable number], and that is exactly the difference to Ger. einige and Sp. algunos. Because einige and algunos are neutral concerning the opposition [+/–considerable], they can, as in the majority of examples (6)–(7), denote an inconsiderable number, but are also used to refer to a considerable number, as in example (8).
TSL[v.20020404] Prn:4/09/2002; 13:56
F: PBNS19.tex / p.6 (360)
Eva Lavric
[SMALL NUMBER] [CONSIDERABLE NUMBER]
einige algunos
[INCONSIDERABLE NUMBER]
quelques
From a didactic viewpoint, it may make sense to ask whether there is a determiner in German which, just like Fr. quelques, represents a combination of the features [small number] and [inconsiderable number]. There is indeed such a determiner: It is the form ein paar. This enables us to include Spanish again, because Sp. un par de is a synonym of Ger. ein paar, i.e. they can be applied in exactly the same contexts; e.g. in the following, which also confirms the equivalence with Fr. quelques: (9) Combien de prétendus savants sont aujourd’hui capables de faire une règle de trois ? Quelques mathématiciens ! Wie viele, die sich Wissenschaftler nennen, sind heutzutage noch fähig, einen Schluß zu rechnen? Ein paar Mathematiker! ¿Cuántos pseudo-científicos son capaces hoy en día de hacer una regla de tres? ¡Un par de matemáticos! (Wil 92) One can see that ein paar / un par de are German and Spanish determiners which represent the same features as quelques. Is it therefore possible to tell the learners to use Fr. quelques only when they could also use Ger. ein paar? It is possible, albeit not one hundred percent true. Because there is a certain group of examples by which quelques exceeds ein paar / un par de. The commutations for examples (6) and (7) are particularly meaningful here: (6 ) Nach hinten wird das Gelände von einigen Eichen begrenzt ein paar Eichen Vers le fond, quelques chênes bornent l’espace visible Hacia detrás, el terreno está delimitado por algunos robles un par de robles (7 ) Quelques femmes célèbres sont censées symboliser au Parlement et au gouvernement l’égalité des droits entre les sexes
TSL[v.20020404] Prn:4/09/2002; 13:56
F: PBNS19.tex / p.7 (361)
Cross-language commutation tests
Einige berühmte Frauen gelten in Parlament und Regierung als symbolhafte Verkörperung der Gleichberechtigung von Mann und Frau ??Ein paar berühmte Frauen Algunas mujeres célebres simbolizan en el Parlamento y en el gobierno la igualdad de derechos entre los sexos ??Un par de mujeres célebres
Example (6) is [–specific], and ein paar / un par de fit well, example (7) is [+specific], and it excludes ein paar / un par de. Ein paar / un par de are thus marked as [–specific]. We therefore have to enlarge our table by the ‘specificity’ dimension: [SMALL NUMBER]
‘number’ dim.
‘specificity’ dim. [+ SPEC.] [– SPEC.]
[CONSIDERABLE NUMBER]
einige algunos
[INCONSIDERABLE NUMBER]
quelques ein paar un par de
If we now want to briefly include the English language, we can try to translate our example sentences into this language and test certain commutations. The obvious form for a comparison with einige/quelques/algunos and ein paar / un par de resp., is of course Engl. a few, which signifies [small number] just like all the other examined forms. But in which variant? Is a few perhaps an equivalent of ein paar / un par de? Example (9) at least would suggest just that: (9 ) How many of those who call themselves scientists are nowadays able to do a rule-of-three? A few mathematicians! We see that a few can occur in [–considerable] examples. To be equivalent to ein paar / un par de, it would also have to signify [–specific]. So let’s check examples (6) and (7): (6 ) The back of the area is bordered by a few oaks (7 ) A few famous women in Parliament and in Government are deemed as token embodiements of equal rights of men and women Engl. a few is possible in both examples and is thus not marked for the opposition [+/–specific]. It could therefore be an equivalent of Fr. quelques or Ger. einige.
TSL[v.20020404] Prn:4/09/2002; 13:56
F: PBNS19.tex / p.8 (362)
Eva Lavric
Let’s check its behaviour with regard to example (8), which tests compatibility with [+considerable]: (8 ) This malicious article will be to the liking of *a few presidents of agricultural organisations A few completely changes the meaning and can thus not be inserted here. From this it can be deduced that it behaves more or less like Fr. quelques, which represents the features [small number] and [–considerable]. However, a few can also express [+considerable] if a relevant modifier is added:18 (8 ) This malicious article will be to the liking of quite a few presidents of agricultural organisations Views may differ as to how to interpret this possibility: A few can either be regarded as [–considerable] and quite a few as [+considerable], so that the two terms are in privative opposition; or a few is (like Ger. einige and Sp. algunos) not marked for the opposition [+/–considerable], so that quite a few only topicalises a possible facet of meaning – which would mean that this is a case of an inclusive opposition. In this case, however, one would have to explain why a few alone ‘fits’ well in example (9) and not at all in example (8), i.e. why without a modifier it tends towards [–considerable], whereas Ger. einige and Sp. algunos actualize [+considerable] in the respective examples. It can be seen that even with such seemingly simple (pseudo-)equivalences each language triggers completely new questions. After this short excursus into English we now return to German, French and Spanish and will, in the last part of this paper, enlarge our semantic table by several additional categories, which enable us to include further forms of determiners: If we e.g. want to locate the specific determiners Fr. certains / Sp. ciertos / Ger. manche we have to exceed the realms of [small number], because certains/ ciertos/manche are not specified according to number.
‘number’ dim. [LARGE NUMBER]
[SMALL NUMBER]
‘specificity’ dim. [+ SPEC.] certains / ciertos manche [– SPEC.]
[CONSIDERABLE NUMBER]
einige algunos
[INCONSIDERABLE NUMBER]
quelques ein paar un par de
TSL[v.20020404] Prn:4/09/2002; 13:56
F: PBNS19.tex / p.9 (363)
Cross-language commutation tests
In this context, I provide my students with the following rule of thumb: Einige in the sense of manche = certains; einige in the sense of ein paar = quelques. This rule of thumb is valid for certain areas and would e.g. have been sufficient to avoid the errors in examples (3) to (5). To be precise, however, one would have to say that Fr. quelques as opposed to Ger. einige conveys, as an additional feature, a subjectively negative evaluation of the number, i.e. the feature [inconsiderable number]. For the sake of completeness, we should now closer examine Sp. algunos. It showed solidarity with Ger. einige in all previous examples and could thus be considered a synonym of einige. This is not exactly true, however, because the range of algunos is actually broader than that of einige. Algunos can also occur in examples in which the referents exist merely hypothetically, i.e. in a possible world, whereas einige is restricted to the existence of referents in the real world, i.e. to actual existence:19 (10) Bringt er irgendwelche Ideen vor, ist man automatisch dagegen *einige Ideen S’il présente des idées, on est automatiquement contre *quelques idées Si expresa algunas ideas, se le lleva automáticamente la contraria (Pre) One can see that in the relevant examples, the German equivalent of algunos is the form irgendwelche, which refers to [–specific] referents, no matter whether they exist in possible worlds or simply in the real world.20 For a final table, the ‘specificity’ dimension therefore has to be broadened by a worlds aspect ([RW] = existence in the real world; [PW] = existence only in a possible world):
‘number’ dim. [LARGE NUMBER]
[SMALL NUMBER]
‘specificity’/ ‘worlds’ dim. [+ SPEC.] certains / ciertos manche
[CONSIDERABLE NUMBER]
einige
[RW]
quelques ein paar un par de
[– SPEC.] irgendwelche [PW]
[INCONSIDERABLE NUMBER]
algunos
TSL[v.20020404] Prn:4/09/2002; 13:56
F: PBNS19.tex / p.10 (364)
Eva Lavric
In all its complexity, this table shows how real equivalences – e.g. certains/ciertos/ manche or ein paar / un par de – are actually an interlingual exception. The plural indefinites of the three examined languages overlap in specific functional areas, but their borders are divergent in the different languages. In particular, there is actually a relation of inclusion between the pseudo-equivalents algunos, einige and quelques: quelques ⊂ einige ⊂ algunos. The investigation clearly shows how the three languages structure reality in a comparable, but not in the same way. The table developed here shows the limits of certain contrastive rules of thumb, because in the subtlety of analysis, it far exceeds the necessities of error explanation. It shows the problematic forms to be parts of a complex system, parts which can be compared in minute detail by means of a table divided according to semantic features and oppositions. By applying the combined contrastive method, it is possible to generate a comprehensive map of determiner meanings in three different languages. French ‘number’ dim. [LARGE NUMBER]
[SMALL NUMBER]
‘specificity’/ ‘worlds’ dim.
[CONSIDERABLE NUMBER]
[+ SPEC.] certains
[INCONSIDERABLE NUMBER]
quelques
[RW] [– SPEC.] [PW]
Spanish ‘number’ dim. [LARGE NUMBER]
[SMALL NUMBER]
‘specificity’/ ‘worlds’ dim.
[CONSIDERABLE NUMBER]
[INCONSIDERABLE NUMBER]
[+ SPEC.] ciertos [RW]
un par de
[– SPEC.] [PW]
algunos
TSL[v.20020404] Prn:4/09/2002; 13:56
F: PBNS19.tex / p.11 (365)
Cross-language commutation tests
German ‘number’ dim. [LARGE NUMBER]
[SMALL NUMBER]
‘specificity’/ ‘worlds’ dim. [+ SPEC.] manche
[CONSIDERABLE NUMBER]
[INCONSIDERABLE NUMBER]
einige
[RW]
ein paar
[– SPEC.] irgendwelche [PW]
French, Spanish and German ‘number’ dim. [LARGE NUMBER]
[SMALL NUMBER]
‘specificity’/ ‘worlds’ dim. [+ SPEC.] certains / ciertos manche
[CONSIDERABLE NUMBER]
einige
[RW]
quelques ein paar un par de
[– SPEC.] irgendwelche [PW]
[INCONSIDERABLE NUMBER]
algunos
Sources of the examples:21 Hen
Käthe Henschelmann: Technik des Übersetzens Französisch-Deutsch, Heidelberg: Quelle & Meyer 1980 (+ Spanish translation by Ángel Borda)
HPP Herbert Huber, Henri Perrin, Alain Pacthod: Deutsch-französische Übersetzungen mit Stilübungen. Thèmes et versions, Vienna: Österreichischer Bundesverlag 1973 (+ Spanish translation by Ángel Borda) Pre
“Pas facile d’être président” / “Von der Kunst, Präsident zu sein” (translated by Susanne Auer) / [“No es fácil ser presidente”], journal article of unknown origin, which has been taken from the internal bulletin of the SFG, “Lausanne Bourgeoise”, November 1983 (+ Spanish translation by Ángel Borda)
TSL[v.20020404] Prn:4/09/2002; 13:56
F: PBNS19.tex / p.12 (366)
Eva Lavric
Wal
Martin Walser: Eiche und Angora. Eine deutsche Chronik, Frankfurt a.M.: Suhrkamp 1963 Fr.: Chêne et lapins angora. Chronique allemande (traduit de l’allemand par Gilbert Badia), Paris: Gallimard 1968 Sp.: Roble y conejos de Angora (traducción Heleno Saña Alcón), Madrid: Edicusa 1970 (Cuadernos para el diálogo. Libros de teatro 19)
Wil
Marc Wilmet: La détermination nominale. Quantification et caractérisation (Linguistique nouvelle), Paris: Presses universitaires de France 1986 (German and Spanish translation by me, E. L.)
Notes . For error analysis and the notion of interference, see e.g. Rattunde 1977; FlamentBoistrancourt 1985; James 1990; Vogel 1990; Henrici & Zöfgen 1993. . See e.g. Coseriu 1973. . See the famous dissertation, Vater 1963/1979. . See e.g. Schifko 1975 and 1992. . The fact that the same mistake occurs three times in this student’s text shows that it is a competence mistake, i.e. that there is a wrong rule stored in the interlanguage. . Lavric 2001a. . The founder of this method is of course Mario Wandruszka (1969 and 1971). See also the recent symposium held in the course of the Romanistentag Osnabrück 1999 (Albrecht & Gauger 2001). . For a justification and explanation of the combined method (translation comparison + commutation tests), see among others Lavric 2001b. . Derjenige is actually confined to this function and cannot be used otherwise. (Although sometimes it also marks restrictive attributes in a form other than that of the relative clause.) . There are exceptions to this rule which correlate with clearly describable stylistic effects; see Lavric 1996, 1997 and 1998. . Ce. . . -là can only be used with explicit or implicit contrast. . Strictly speaking, the proximate demonstratives Ger. dieser and Sp. este are marked for non-restrictivity (= appositionality) of the following relative clause. The distal demonstratives are actually unmarked and thus theoretically comprise both possibilities (restrictivity or appositionality); in fact, however, they are used to mark restrictivity in the overwhelming majority of cases. (Strictly speaking, the definite article also embodies both possibilities). See Lavric (2001a: 900–914). . In addition, one could predict that Spanish-speaking learners of French will face problems similar to those of German-speaking ones when confronted with this matter, while native speakers of French learning German or Spanish will probably tend to exclusively
TSL[v.20020404] Prn:4/09/2002; 13:56
F: PBNS19.tex / p.13 (367)
Cross-language commutation tests
use the definite article with restrictive relative clauses and thus convey the so-called “overindulgence” phenomenon (Levenston 1972) – out of several varieties the one similar to the mother tongue is preferred – which, however, is obviously not too troublesome. . See e.g. Vater 1963/1979; Sroka 1983; Zhou 1985. This oposition should not be confused with the one between actually versus hypothetically existing referents, which is often referred to by the same term. For a more extensive discussion of the ambiguities of indefinite reference see Galmiche 1983; Martin 1983; and Lavric 1990. . Moreover, they of course show the features [indefinite], [more than one referent] and [–exact number], meaning that they are plural indefinites. (For the semantics of plural indefinites in French and Spanish, see Lavric 2000 and forthcoming). . A number of authors (e.g. Oomen 1977; Flückiger-Studer 1983; Zhou 1985) mix the two oppositions, which in fact can be distinguished in case of a few determiners only. Moxey & Sanford (1993), however, do distinguish between the two approaches to quantity, the objective and the subjective one. . Ger. so mancher and Fr. maint(s) refer to a number as being considerable without specifying whether it is large or small. . Of particular interest is the existence of a singular form many a (many a president . . . ), which is the exact equivalent of Fr. maint(s) and Ger. so manche(r). All three forms signify [+considerable] without specifying a large or small number. They are furthermore stylistically marked as slightly pompous. . For the concept of possible worlds see e.g. Martin 1983 and 1987, and subsequently Lavric 1990 and 1995. . Engl. a few is here subject to the same restrictions as Ger. einige; it has to be substituted with any (here equivalent of Ger. irgendwelche): (10 ) When he proposes any new idea(s), people are automatically against them *a few new ideas . All emphases in the examples are mine, E. L.
References Albrecht, J., & Gauger, H.-M. (Eds.). (forthcoming). Sprachvergleich und Übersetzungsvergleich. Leistung und Grenzen, Unterschiede und Gemeinsamkeiten. Frankfurt a.M. c.a.: Peter Lang. Coseriu, E. (1973). Probleme der strukturellen Semantik. Lecture given in the winter semester 1965/1966 at the University of Tübingen. Authorised and edited notes by Dieter Kastovsky (Tübinger Beiträge zur Linguistik 40). Tübingen: Narr. Flament-Boistrancourt, D. (1985). L’interférence : Un masque, une ambiguïté ? Acta universitatis wratislaviensis, 818, Romanica wratislaviensia, 23 (= Le français langue étrangère. Actes du colloque franco-polonais, Lille, avril 1983), 21–35.
TSL[v.20020404] Prn:4/09/2002; 13:56
F: PBNS19.tex / p.14 (368)
Eva Lavric
Flückiger-Studer, T. (1983). Quantifikation in natürlichen Sprachen. Zur Semantik und Syntax französischer und deutscher Beschreibungen (Linguistische Arbeiten 132). Tübingen: Niemeyer. Gaatone, D. (1991). Les déterminants de la quantité peu élevée en français. Remarques sur les emplois de quelques et plusieurs. Revue romane, 26 (1), 3–13. Galmiche, M. (1983). Les ambiguïtés référentielles ou les pièges de la référence. In G. Kleiber & M. Riegel (Eds.), Grammaire et référence (= Langue française, 57) (pp. 60–86). Gondret, P. (1976). ‘Quelques’, ‘plusieurs’, ‘certains’, ‘divers’: Étude sémantique. Le français moderne, 44, 143–152. Henrici, G., & Zöfgen, E. (Eds.). (1993). Fehleranalyse und Fehlerkorrektur (= Fremdsprachen Lehren und Lernen, 22). James, C. (1990). Learner language. Language teaching, 23 (4), 205–213. K˛esik, M. (1978). Quelques vs. plusieurs. In Actes de la IIe Conférence des linguistes romanisants polonais (pp. 59–64). Warsaw: Editions de l’Université de Varsovie. Lavric, E. (1990). Mißverstehen verstehen: Opake Kontexte und Ambiguitäten bei indefiniten und definiten Nominalphrasen (Grazer Linguistische Monographien 7). Graz: Institut für Sprachwissenschaft der Universität Graz. Lavric, E. (1995). Referential ambiguities, possible worlds, and universes – The problem of substitution in contexts of propositional attitude. In R. A. Geiger (Ed.), Reference in multidisciplinary perspective. Philosophical object, cognitive subject, intersubjective process (pp. 45–62). Hildesheim/Zurich/New York: Georg Olms. Lavric, E. (1996). ‘Cet air doublement immobile qu’ont les choses mobiles quand elles ne bougent pas...’: Haloeffekt und Relativsatz-Ankündigung. Moderne Sprachen, 40 (2), 157–170. Lavric, E. (1997). ‘Ese reino movible’ – Spanische, französische und deutsche Demonstrativa. In G. Wotjak (Ed.), Studien zum romanisch-deutschen und innerromanischen Sprachvergleich. Akten der III. Internationalen Arbeitstagung zum romanischdeutschen Sprachvergleich (Leipzig, 9.10.–11.10.1995) (pp. 515–543). Frankfurt a.M./Berlin/Bern/New York: Peter Lang. Lavric, E. (1998). Este, ese y aquel en función determinativa. In G. Ruffino (Ed.), Atti del XXI Congresso Internazionale di Linguistica e Filologia Romanza (Palermo, 18–24 settembre 1995), Vol. III: Lessicologia e semantica delle lingue romanze (pp. 405–418). Tübingen: Niemeyer. Lavric, E. (2000). Indéfinis pluriels français et espagnols. In A. Englebert, M. Pierrard, L. Rosier, & D. Van Raemdonck (Eds.), Actes du XXIIe Congrès International de Linguistique et Philologie Romanes, Bruxelles, 23–29 juillet 1998, Vol. VII: Sens et fonction (pp. 377–386). Tübingen: Niemeyer. Lavric, E. (2001a). Fülle und Klarheit. Eine Determinantensemantik Deutsch–Französisch– Spanisch (Stauffenburg Linguistik 9), Vol. I: Referenzmodell; Vol. II: Kontrastivsemantische Analysen. Tübingen: Stauffenburg. Lavric, E. (2001b). Übersetzungsvergleich mit Kommutationsprobe – eine paradigmatische Tiefendimension. In J. Albrecht & H.-M. Gauger (Eds.), Sprachvergleich und Übersetzungsvergleich. Leistung und Grenzen, Unterschiede und Gemeinsamkeiten (pp. 97– 129). Frankfurt a.M. c.a.: Peter Lang.
TSL[v.20020404] Prn:4/09/2002; 13:56
F: PBNS19.tex / p.15 (369)
Cross-language commutation tests
Lavric, E. (forthcoming). Indefinidos plurales españoles. In F. Sevilla & C. Alvar (Eds.), Actas del XIII Congreso de la Asociación Internacional de Hispanistas, Madrid 1998, Vol. III: Hispanoamericana. Lingüística. Teoría literaria. Madrid: Castalia. Levenston, E. A. (1972). Über- und Unterrepräsentation – Aspekte der muttersprachlichen Interferenz. In G. Nickel (Ed.), Reader zur kontrastiven Linguistik (pp. 167–174). Frankfurt a.M.: Fischer. Martin, R. (1983). Pour une logique du sens (Linguistique nouvelle). Paris: Presses universitaires de France. Martin, R. (1987). Langage et croyance. Les ‘univers de croyance’ dans la théorie sémantique (Philosophie et langage). Brussels: Pierre Mardaga. Moxey, L. M., & Sanford, A. J. (1993). Prior expectation and the interpretation of natural language quantifiers. European journal of cognitive psychology, 5 (1), 73–91. Oomen, I. (1977). Determination bei generischen, definiten und indefiniten Beschreibungen im Deutschen (Linguistische Arbeiten 53). Tübingen: Niemeyer. Rattunde, E. (1977). Transfer – Interferenz? Probleme der Begriffsdefinition bei der Fehleranalyse. In E. Rattunde (Ed.), Fehleranalyse/Fehlerbewertung (= Die Neueren Sprachen, 7 (1)) (pp. 4–14). Schifko, P. (1975). Bedeutungstheorie. Eine Einführung in die linguistische Semantik (Problemata 45). Stuttgart-Bad Cannstatt: Fromman-Holzbog. Schifko, P. (1992). Spanisch: Lexikologie und Semantik. In G. Holtus, M. Metzeltin, & C. Schmitt (Eds.), Lexikon der Romanistischen Linguistik (LRL) Vol. VI, 1: Aragonesisch/Navarresisch, Spanisch, Asturianisch/Leonesisch (pp. 132–148). Tübingen: Niemeyer. Sroka, K. A. (1983). Definiteness and truth relation. In R. Jongen, S. De Knop, P. H. Nelde, & M.-P. Quix (Eds.), Sprache, Diskurs und Text. Akten des 17. Linguistischen Kolloquiums, Brüssel 1982, Vol. I (Linguistische Arbeiten 133) (pp. 110–120). Tübingen: Niemeyer. Vater, H. (1963/1979). Das System der Artikelformen im gegenwärtigen Deutsch. Tübingen: Niemeyer, 1963. 2nd, revised edition (Linguistische Arbeiten 78). Tübingen: Niemeyer, 1979. Vogel, K. (1990). Lernersprache. Linguistische und psycholinguistische Grundfragen ihrer Erforschung (Tübinger Beiträge zur Linguistik 341). Tübingen: Narr. Wandruszka, M. (1969). Sprachen, vergleichbar und unvergleichlich. Munich: Piper. Wandruszka, M. (1971). Interlinguistik. Umrisse einer neuen Sprachwissenschaft. Munich: Piper. Zhou, H. (1985). Determination und Determinantien. Eine Untersuchung am Beispiel neuhochdeutscher Nominalsyntagmen (Bochumer Beiträge zur Semiotik 2). Bochum: Brockmeyer.
Language index
A Arabic 296, 314 B Bagwalal 178–185, 187, 188, 190, 191 C Catalan 101–113, 129, 130, 133, 140, 142–145 E English 3, 4, 7, 23, 26, 27, 29, 33, 35, 38, 41, 45–48, 51–54, 56, 61, 71, 75–78, 81–90, 93–95, 98, 99, 101–106, 108–114, 119, 120, 122, 124, 157, 158, 162–167, 169, 193–201, 203–205, 209–211, 213, 214, 216, 223–225, 228–231, 233–238, 241–243, 249, 250, 252, 261–265, 268–270, 272, 273, 277, 278, 280–288, 307, 308, 311, 313, 314, 316, 318, 321, 327, 329, 335, 344, 355, 361, 362 F Finnish 194 French 3, 4, 9, 10, 21, 26, 30, 45, 76, 84, 102, 104, 117–120, 122, 129–133, 140, 142–145, 194, 195, 262, 282, 336, 343, 355–358, 362, 364–367 G German 102, 104, 194, 355–358, 360, 362, 363, 365, 366
Greek 129, 130, 132, 133, 139, 140, 142–144, 344–346, 352 Guugu Yimithirr 297, 303 H Hadiyya 326, 327 Hausa 297, 303 Hungarian 30, 129, 130, 132, 133, 139, 140, 142–144, 194, 223–226, 228–231, 233–243 I Icelandic 195 Italian 3, 4, 7–9, 12–14, 16, 26, 31, 34, 38, 102, 104, 129, 130, 133, 140, 142–145, 147, 157, 160, 164–168, 171, 172, 194, 248–250, 252, 253, 255, 257, 258, 336 J Japanese 67, 261, 264, 265, 268, 269, 272, 273, 277, 278, 281–283, 285, 286, 288, 296, 307, 311, 314, 316, 318, 321, 328–332 K Khanty 183 L Lega 20, 190 Lithuanian 193–201, 208–217 M Mari 178–185, 187, 190, 191
Language index
P Polish 194, 201 Portuguese 31, 33, 34, 36–38, 40, 45–48, 51–53, 56, 58, 61, 67, 72, 129, 130, 132–134, 140–145 Brazilian Portuguese 58 R Romanian 75–81, 83, 84, 86–93, 95, 96, 98–100, 344, 345, 349, 351, 352 Russian 194, 195, 262
S Sesotho 296 Spanish 31, 34, 38, 45, 102, 129, 130, 133, 140, 142–145, 194, 248, 249, 252, 255–258, 339–347, 349–351, 355–358, 360, 362, 364–367 Swahili 117–120, 122–124 Swedish 194, 214 T Tatar 178–185, 187, 190, 191 Turkish 177–179 Tzeltal 297, 303
Name index
A Abe, Y.
323, 331
B Blakemore, D.
332
C Carston, R. 9, 172, 325, 333 Chierchia, G. 3–5, 7–13, 17, 26, 148, 151, 154, 162 Cinque, G. 156, 158, 252 Cornelis, L. 249–252 Croft, W. 200, 250 D Dahl, Ö 102, 179 Davidson, D. 326, 333 de Swart, H. 75–77, 81–83 F Fellbaum, C. 270 G Gabbay, D. 307 Gazdar, G. 152–154 Givon, T. 212 Grice, P. 134, 162, 163
K Kamp, H. 61, 68, 71, 73, 75, 77–79, 81, 82, 84, 85, 88, 89, 93, 295, 300–302, 304, 326 Kempson, R. 330, 332 Kubozono, H. 330 L Langacker, R. 258, 277, 280 Lepschy, G. 252 Levin, B. 200, 263, 270 Löbner, S. 6 M Mann, W. 358, 361 Manzini, R. 252 Marten, L. 333 Masuoka, T. 331 Matsui, T. 322, 330, 332 Matsumoto, Y. 307, 332, 333 Meyer-Viol, W. 307 Michaelis, L. 101, 104, 106, 107, 109–111, 113 N Narita, H.
307, 318, 319
H Haspelmath, M. 248 Hidalgo, R. 256
P Parsons, T. 76, 326, 333 Perrett, D. 326 Prince, E. 247, 259
J Jackendoff, R.
R Recanati, F.
300
324
Name index
Reyle, U. 61, 68, 71, 73, 75, 77–79, 81, 82, 84, 85, 88, 89, 93, 295, 300–302, 304, 326 Ritter, E. 224, 225, 230, 231, 233, 235, 236, 239, 242 Rosen, S. 224, 225, 230–236, 239, 242 Rosta, A. 262–264, 272 S Sansò, A. 245, 252, 258 Schwenter, S. 101, 109 Shibatani, M. 247, 248 Siewierska, A. 246 Soames, S. 148, 149, 151, 153, 154, 165 Sperber, D. 118, 322, 324
Stalnaker, R. 147, 149–151, 162, 167, 168 Steedman, M. 67, 326 Suga, K. 265, 266 T Teramura, H. 277, 307, 317, 331 Thompson, S. 248, 277, 280, 284, 285, 288, 289 W Wilson, D. 117, 118, 122, 124, 322, 324, 332 Y Yamanashi, M. 307, 332
Subject index
A accessibility 227, 246, 322 actions 102, 107, 267, 281–283, 286, 288, 308, 312, 324 activity verbs 232, 241 adverbial clause 331 ‘affectedness’ 273 agent 15, 185–188, 232, 246–252, 257, 258, 262–267, 269, 277, 280, 287 defocusing 248, 250 ago 45, 47–56, 58, 59, 90–93, 99, 107, 112, 118, 254, 267 aktionsart 65–68, 73, 77, 81, 88, 101, 113, 258 algunos 337, 341, 347, 348, 355, 357–360, 362–364 anaphora 6, 64–67, 70, 151, 307, 330, 337 see discourse anaphora anaphoric temporal locators 61, 62 anchor points 45, 47, 48, 57 animacy 246, 287, 340, 343 argument structure 223, 269 aspectual modifier 75, 76, 82 até 52, 53, 58, 62, 63, 141 atelic descriptions 54 aucune 9 B bare happening 251–253, 255, 257, 258 bare object complements 205, 217 bien des 359 bound duration adverbs 84
bridging reference 320–322, 325, 332 British National Corpus (BNC) 214 C certains 358, 362, 363 ciertos 362 clitic pronouns 335, 336 clitic doubling 335, 338–341, 343–347, 349, 350, 352 coercion 75, 76, 79, 80, 84, 91 cognitive grammar 258, 307, 332 collective predicates 6 combined contrastive method 364, 366 commutation tests 355, 357 complement clause 130, 132–136, 138–140, 143, 204, 210, 289, 311 see main clause complementation 193, 195–203, 205, 209, 210, 213, 214, 216 complex clauses 279, 289 compound verbs 283 computational action 308 conceptual information 118 see procedural information configurationality 223, 225, 226 constraint interaction 335, 338, 352 context 12, 16, 20, 22, 37, 40, 41, 50, 52, 62, 104, 105, 107, 122, 143, 144, 148–156, 160, 162–171, 179, 181, 184, 186–189, 216, 230, 241, 264, 268, 271, 281, 286, 288, 295, 296, 307, 321, 327, 332, 336–340, 347, 348, 350, 363
Subject index
context sets 336, 337, 340 contextual information 118, 122 contingent statement 23 continuative perfect 104, 112 coordination 15, 19 counting of temporally ordered entities 45 D demonstrative determiner 356, 357 dependency 278, 307, 309 determiner semantics 355 see map of determiner meanings universal determiners 347 unrestricted determiners 4 detransitivization 247 direct impression 209 direct mental perception 204, 206 direct vs indirect perception 193, 195, 200, 204, 210 discourse anaphora 6, 307 discourse configurationality 225, 226 discourse representation structure (DRS) 302 Discourse Representation Theory (DRT) 75 discourse topic 229, 242, 246, 253, 340 durante esse tempo 64, 68, 70, 71 during that time 62, 64, 66–68, 70, 71 Dynamic Syntax 307, 308, 324–326, 329 E ein paar 360, 361, 363, 364 einige 355, 357–359, 361–364, 367 entretanto 64, 65, 68, 70, 71 enquanto isso 62–67, 70–72 epistemic attitude 133, 135, 138–140 possibility/uncertainty 202 error analysis 355, 358, 366 evaluation 132, 133, 188, 189, 207–209, 359, 363
events 54, 57, 58, 75, 77, 78, 80, 83, 88–91, 94, 99, 104, 105, 107–112, 117, 118, 122–124, 185, 189, 202, 224, 231, 234, 248, 262, 269, 272, 277, 284, 286–288, 295, 325 see intervals causation 277 role 232, 234, 238 scale 280, 281, 285 structure 72, 73, 89, 224, 250, 258 variable 325–329, 333 eventuality 24, 48, 61–67, 71–73, 75–77, 79–91, 93, 98, 99, 117, 118, 120–124, 300 descriptions 61, 62, 65, 66, 73, 75, 76, 79–81 every 7, 11, 13, 53, 71, 72, 179, 188, 217, 237, 258, 264, 303, 322, 347, 348 evidentiality 177, 178, 189, 190, 201, 210, 212, 213, 216 existential perfect 103, 104 experiential inferentials 178, 185, 190 see non-experiential inferentials exp-verbs 200, 201, 203, 204, 213 extension of meaning 193, 195, 200, 201, 204, 213 F a few 357, 361, 362, 367 factive verb 130, 162, 163, 169 finite complement clause 197 for 45–54, 56, 57 frame adverbs 84 frame anaphoric adverbs 84 Frame Semantics 307, 332 free choice 351 G gapless relatives 316, 317, 320–325, 328, 329, 331–333
Subject index
appositional gapless relatives 316, 328, 331 generalized quantifier 4, 24, 336, 338, 339, 350, 352 Generalized Quantifier Theory 4, 339 generic statement 5, 347 genre 287 given-new 225–227, 229 goal-driven process 311
H há 38, 45, 47–53, 55–59 ‘hodiernal past’ 102 ‘hot news perfect’ 105, 109
I impersonalization 247 indefinite plural determiners 355 independent clause 278, 279 indicative 129–145, 343 see subjunctive inferred evidence 177, 189 infinitival complement 279 Infinitival Perception Verb Complements (IPVCs) 202 information structure 223–226, 229, 230, 233, 238, 241 intensive quantity 11, 27 interrogative quantifier 350, 351 intervals 46–50, 52, 54–57, 61, 62, 64, 73 see events intransitive clause 248, 277, 281, 285, 288 irgendwelche 363
J judgement types 224, 228, 241 categorical 224, 228–230, 238–240, 242 thetic 224, 228–230, 235, 238–240, 242
L le moindre 10, 13 left downward entailment 35 lexical instruction 308 LINK 313, 314, 316, 319, 329, 331 Lithuanian Language Corpus 214 long-distance dependency 307, 309 Longman Dictionary of Contemporary English 262 M main clause 57, 130, 132, 136, 143, 153, 155, 215, 279, 311 see complement clause maint(s) 367 más de un 359 manche 358, 362–364 (in the) meantime 62, 65, 70 map of determiner meanings 364 measure phrase 87, 88, 92, 93, 97–99 mental perception 202, 204, 206, 211, 212 merge 312, 327, 330 metalinguistic negation 137, 138 metavariable 308, 326, 327 modal particles 279 modal qualification 200, 201, 203, 211–213, 216 mood selection 129, 130, 132–134, 136–141, 144 N nadie 29, 30, 38, 349 natural discourse 277, 288 negative concord (NC) 16, 29–31, 33, 41 negative determiners 3, 5, 7, 10, 13, 26 negative polarity 10 negative quantifiers 12, 19, 349 N-words 16, 30, 35, 37, 38 nessuno 4, 7–15, 17–27, 29, 30, 38 network of demotion 245, 251, 252 newsworthiness 264, 270, 272
Subject index
“newsworthy” 264, 269, 272 niente 4, 13–27, 38, 39 ninguém 29, 30, 35, 38 ningún 38–40, 349 no 4–6, 13, 26, 33–35, 38, 41 nobody 27, 29, 30, 38, 41, 42, 141, 181, 209, 349 nomic character 25 nominal determination 355 non-experiential inferentials 190, 191 see experiential inferentials noun phrase 271, 346, 356 abstract mass nouns 5, 9 concrete mass nouns 9, 25 countable nouns 3, 7, 9, 11, 12, 14, 19, 20, 27 definites 345, 346 indefinites 344, 357, 364, 367 plural indefinites 357, 364, 367 relational noun 324 uncountable nouns 3, 4, 7–12, 18, 24, 25, 27 O object manipulation 277 non-object manipulating 281–283, 286, 288 object-manipulating actions 281, 283, 286, 288 Optimality Theory (OT) 338, 339, 352 P parallelism 71, 124, 320, 337, 344, 350 parametrical variation 223, 243 parenthetical use 197–199, 213, 216 partitive interpretation 18, 240 passive constructions 245, 248, 249, 251, 252, 254, 257, 258 mediopassives 263 patient 245, 246, 249–254, 257, 262, 266, 277, 280, 287
-oriented process 251, 255, 257, 258 topicalization 245 pentru 88, 95, 96, 98, 99 PERF 75–79, 81–83, 85–89, 93–95, 99 perspective 51, 78, 88, 122, 227, 242, 249–252, 255, 258, 301, 329 plus d’un 359 pragmaticalization 198, 199, 213 present perfect (PrP) 101 presupposition 4, 6, 20–22, 25, 26, 106, 111, 147–155, 157, 159–172, 242, 339 presuppositionality 339, 341, 342, 346 PREZ 75–81, 84–88, 91, 92 principal filter 339–341, 344 procedural information 118 see conceptual information propositional content 211–213 prototype 280 pseudo relative clause 321, 332 pseudo relatives 316, 318, 325, 328, 329, 331, 332 punctual adverbs 84
Q qualche 7, 9, 12, 246 quantized eventualities 76 quelques 355, 357–364
R referentiality 251, 341, 342, 345 relative clause 279, 307, 311, 314, 315, 317–333, 342, 344, 356, 357, 366 restrictive relative clause 356, 357 Relevance, Principle of 322 Relevance Theory (RT) 117, 118, 121, 324, 332 resultative 103, 105, 106, 108–112
Subject index
Rhetorical Structure Theory (RST) 248 S saliency 267, 273 saturation 324 Segmented Discourse Representation Theory (SDRT) 123 semantic case theory 336 semantic features and oppositions 364 semantic property 35, 243, 278 sentential negation 133, 136–138 simple clause 279, since 45, 47, 51–56, 58, 59, 91, 99 singular determiners 3, 7, 27 so mancher 367 spatial adverbs 296 spatial prepositional expressions 296 speaker’s assessment 212, 213 specific determiner 235 [+/–SPECIFICITY] 358, 359 specificity effect 341 speech time (ST) 76, 77 st-verbs 200, 201, 210, 211, 213 states 5, 20, 66, 75–77, 80, 82, 86, 88, 90, 94, 95, 99, 109, 113, 117, 118, 120, 122, 124, 262, 267, 269, 272, 284–286, 288 see events strongly discrete units 12, 13 structural underspecification 308, 309 subjunctive 129–145, 342–344 see indicative sufficient informativeness 272 T taxonomic readings 8 telic events 54, 90 tense operator 76, 77, 79, 80, 83, 84, 87, 88, 92 that-clause 156, 160, 199 theme-rheme 225–227
time measurement 47, 49 topic-comment 225–227 topic identification 247 see discourse topic transitive clauses 277, 278 transitivity 277–281, 284–288 translation comparison 355–357 type semantic 308, 309, 317, 325 types of entities 202 U un certain 11 un par de 360, 361, 364 unfixed node 309, 310, 313, 314 until 51, 53, 58, 62, 141, 287 V valence 247, 261 verb ellipsis 15 verb phrase bare infinitive 202 finite 196–199, 203, 205, 210, 212–214, 278, 310 full infinitive 209 non-finite 180, 198, 203, 204, 278 verbal modifiers 233, 235, 236, 242 verbless sentences/clauses 4, 14–17, 20, 26, 27 verbs of perception 193, 194, 200, 210 veridicality 37, 132, 139, 143 visual perception 193, 210, 217 volitionality 284 W weakly discrete units 12 wh-words 196, 349, 350 Z zero complementation 199, 216
Contents of Volume 2
Grammaticalization Distal aspects in Bantu languages Steve Nicolle From temporal to conditional: Italian qualora vs English whenever Jacqueline Visconti
3 23
Then – adverbial pro-form or inference particle? A comparative study of English, Ewe, Hungarian, and Norwegian Thorstein Fretheim, Stella Boateng, and Ildikó Vaskó
51
The polysemy of the Swedish verb komma ‘come’: A view from translation corpora Åke Viberg
75
Metaphor in contrast Studying metaphors using a multilingual corpus Kay Wikberg
109
Cross-language metaphors: Conceptual or pragmatic variation? Andreas Musolff
125
A contrastive cognitive perspective on Malay and English figurative language Jonathan Charteris-Black
141
Metaphorical expressions in English and Spanish stock market journalistic texts Anna Espunya and Patrick Zabalbeascoa
159
Contents of Volume 2
Cross-cultural pragmatics and speech acts Directions of regulation in speech act theory Susumu Kubo
183
On Japanese ne and Chinese ba Mutsuko Endo Hudson and Wen-ying Lu
197
‘I am asking for a pen’: Framing of requests in black South African English Luanga A. Kasanga
213
Cultural scripts for French and Romanian thanking behaviour Tine Van Hecke
237
Sociocultural variation in native and interlanguage complaints Ronald Geluykens and Bettina Kraft
251
A cross-cultural study of requests: The case of British and Japanese undergraduates Saeko Fukushima
263
Questions as indirect requests in Russian and Czech Michael Betsch
277
The language of love in Melanesia: A study of positive emotions Les Bruce
291
Everyday rituals in Polish and English Ewa Jakubowska
331
A question of time? Question types and speech act shifts from a historical-contrastive perspective: Some examples from Old Spanish and Middle English Verena Jung and Angela Schrott The contrasts between contrasters: What discussion groups can tell us about discourse pragmatics Piibi-Kai Kivik and Krista Vogelberg
345
373
The semantics/pragmatics boundary: Theory and applications Cross-linguistic implementations of specificity Klaus von Heusinger
405
The semantics– pragmatics interface: The case of grounding Esam N. Khalil
423
Contents of Volume 2
On translating ‘what is said’: Tertium comparationis in contrastive semantics and pragmatics K. M. Jaszczolt
441
Translation equivalents as empirical data for semantic/pragmatic theory Bergljot Behrens and Cathrine Fabricius-Hansen
463
Language index
477
Name index
481
Subject index
483
Contents of Volume 1
489
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: Responding in Conversation. A study of response particles in Finnish. 2001. 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. 2002. 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. 2001. 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. 2001. 87. MUSHIN, Ilana: Evidentiality and Epistemological Stance. Narrative retelling. 2001. 88. BAYRAKTAROG LU, ArFn and Maria SIFIANOU (eds.): Linguistic Politeness Across Boundaries. The case of Greek and Turkish. 2001. 89. ITAKURA, Hiroko: Conversational Dominance and Gender. A study of Japanese speakers in first and second language contexts. 2001. 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. 2001. 92. GARDNER, Rod: When Listeners Talk. Response tokens and listener stance. 2001. 93. BARON, Bettina and Helga KOTTHOFF (eds.): Gender in Interaction. Perspectives on femininity and masculinity in ethnography and discourse. 2002 94. McILVENNY, Paul (ed.): Talking Gender and Sexuality. 2002. 95. FITZMAURICE, Susan M.: The Familiar Letter in Early Modern English. A pragmatic approach. 2002. 96. HAVERKATE, Henk: The Syntax, Semantics and Pragmatics of Spanish Mood. 2002.
97. MAYNARD, Senko K.: Linguistic Emotivity. Centrality of place, the topic-comment dynamic, and an ideology of pathos in Japanese discourse. 2002. 98. DUSZAK, Anna (ed.): Us and Others. Social identities across languages, discourses and cultures. 2002. 99. JASZCZOLT, K.M. and Ken TURNER (eds.): Meaning Through Language Contrast. Volume 1. 2003. 100. JASZCZOLT, K.M. and Ken TURNER (eds.): Meaning Through Language Contrast. Volume 2. 2003. 101. LUKE, Kang Kwong and Theodossia-Soula PAVLIDOU (eds.): Telephone Calls. Unity and diversity in conversational structure across languages and cultures. 2002. 102. LEAFGREN, John: Degrees of Explicitness. Information structure and the packaging of Bulgarian subjects and objects. 2002. 103. FETZER, Anita and Christiane MEIERKORD (eds.): Rethinking Sequentiality. Linguistics meets conversational interaction. 2002. 104. BEECHING, Kate: Gender, Politeness and Pragmatic Particles in French. 2002. 105. BLACKWELL, Sarah E.: Implicatures in Discourse. The case of Spanish NP anaphora. 2003. 106. BUSSE, Ulrich: Linguistic Variation in the Shakespeare Corpus. Morpho-syntactic variability of second person pronouns. 2002. 107. TAAVITSAINEN, Irma and Andreas H. JUCKER (eds.): Diachronic Perspectives on Address Term Systems. n.y.p. 108. BARRON, Anne: Acquisition in Interlanguage Pragmatics. How to do things with words in a study abroad context. n.y.p. 109. MAYES, Patricia: Language, Social Structure, and Culture. A genre analysis of cooking classes in Japan and America. n.y.p. 110. ANDROUTSOPOULOS, Jannis K. and Alexandra GEORGAKOPOULOU (eds.): Discourse Constructions of Youth Identities. n.y.p. 111. ENSINK, Titus and Christoph SAUER (eds.): Framing and Perspectivising in Discourse. n.y.p. 112. LENZ, Friedrich (ed.): Deictic Conceptualisation of Space, Time and Person. n.y.p. 113. PANTHER, Klaus-Uwe and Linda L. THORNBURG (eds.): Metonymy and Pragmatic Inferencing. n.y.p.